DNA Hypermethylation Contributes to Incomplete Synthesis of Carbohydrate Determinants in Gastrointestinal Cancer

Background & Aims: It has long been known that malignant transformation is associated with abnormal expression of carbohydrate determinants. The aim of this study was to clarify the cause of cancer-associated abnormal glycosylation in gastrointestinal (GI) cancers. Methods: We compared the expression levels of “glyco-genes,” including glycosyltransferases and glycosidases, in normal GI mucosa and in gastric and colorectal cancer cells. To examine the possibility that DNA hypermethylation contributed to the down-regulation of these genes, we treated GI cancer cells with 5-aza-2′-deoxycytidine (5-aza-dC), an inhibitor of DNA methyltransferase. Results: The silencing of some of these glyco-genes, but not up-regulation of certain molecules, was observed. The Sda carbohydrate was abundantly expressed in the normal GI mucosa, but its expression was significantly decreased in cancer tissues. When human colon and gastric cancer cells were treated with 5-aza-dC, cell surface expression of Sda and the transcription of B4GALNT2, which catalyzes the synthesis of the Sda, were induced. The promoter region of the human B4GALNT2 gene was heavily hypermethylated in many of the GI cancer cell lines examined as well as in gastric cancer tissues (39 out of 78 cases). In addition, aberrant methylation of the B4GALNT2 gene was strongly correlated with Epstein-Barr virus-associated gastric carcinomas and occurred coincidentally with hypermethylation of the ST3GAL6 gene. Conclusions: Epigenetic changes in a group of glycosyltransferases including B4GALNT2 and ST3GAL6 represent a malignant phenotype of gastric cancer caused by silencing of the activity of these enzymes, which action may eventually induce aberrant glycosylation and expression of cancer-associated carbohydrate antigens. Background & Aims: It has long been known that malignant transformation is associated with abnormal expression of carbohydrate determinants. The aim of this study was to clarify the cause of cancer-associated abnormal glycosylation in gastrointestinal (GI) cancers. Methods: We compared the expression levels of “glyco-genes,” including glycosyltransferases and glycosidases, in normal GI mucosa and in gastric and colorectal cancer cells. To examine the possibility that DNA hypermethylation contributed to the down-regulation of these genes, we treated GI cancer cells with 5-aza-2′-deoxycytidine (5-aza-dC), an inhibitor of DNA methyltransferase. Results: The silencing of some of these glyco-genes, but not up-regulation of certain molecules, was observed. The Sda carbohydrate was abundantly expressed in the normal GI mucosa, but its expression was significantly decreased in cancer tissues. When human colon and gastric cancer cells were treated with 5-aza-dC, cell surface expression of Sda and the transcription of B4GALNT2, which catalyzes the synthesis of the Sda, were induced. The promoter region of the human B4GALNT2 gene was heavily hypermethylated in many of the GI cancer cell lines examined as well as in gastric cancer tissues (39 out of 78 cases). In addition, aberrant methylation of the B4GALNT2 gene was strongly correlated with Epstein-Barr virus-associated gastric carcinomas and occurred coincidentally with hypermethylation of the ST3GAL6 gene. Conclusions: Epigenetic changes in a group of glycosyltransferases including B4GALNT2 and ST3GAL6 represent a malignant phenotype of gastric cancer caused by silencing of the activity of these enzymes, which action may eventually induce aberrant glycosylation and expression of cancer-associated carbohydrate antigens. See editorial on page 305. It has long been known that malignant transformation is associated with abnormal expression of carbohydrate determinants.1Hakomori S. Glycosylation defining cancer malignancy: New wine in an old bottle.Proc Natl Acad Sci U S A. 2002; 99: 10231-10233Crossref PubMed Scopus (796) Google Scholar Many glycosyl epitopes such as sialyl Tn, Tn, T, and sialyl Lewis x/a (sLex/a) have been reported to be cancer-associated antigens. Some of them show statistically significant correlations between the degree of their expression in cancer tissues and the postoperative prognosis of patients with many types of human cancers.2Nakamori S. Kameyama M. Imaoka S. et al.Increased expression of sialyl Lewisx antigen correlates with poor survival in patients with colorectal carcinoma: clinicopathological and immunohistochemical study.Cancer Res. 1993; 53: 3632-3637PubMed Google Scholar, 3Nakayama T. Watanabe M. Katsumata T. et al.Expression of sialyl Lewisa as a new prognostic factor for patients with advanced colorectal carcinoma.Cancer. 1995; 75: 2051-2056Crossref PubMed Scopus (146) Google Scholar, 4Futamura N. Nakamura S. Tatematsu M. et al.Clinicopathologic significance of sialyl Lex expression in advanced gastric carcinoma.Br J Cancer. 2000; 83: 1681-1687Crossref PubMed Scopus (36) Google Scholar In addition, sLex/a determinants are known to serve as ligands for E-selectin, which is inducibly expressed by endothelial cells, in hematogenous metastasis of cancers.5Walz G. Aruffo A. Kolanus W. et al.Recognition by ELAM-1 of the sialyl- Lex determinant on myeloid and tumor cells.Science. 1990; 250: 1132-1135Crossref PubMed Scopus (883) Google Scholar, 6Takada A. Ohmori K. Yoneda T. et al.Contribution of carbohydrate antigens sialyl Lewis A and sialyl Lewis X to adhesion of human cancer cells to vascular endothelium.Cancer Res. 1993; 53: 354-361PubMed Google Scholar A long-standing debate is which is more important in understanding cancer-associated carbohydrate antigens, “neo-synthesis” or “incomplete synthesis.” To verify the former hypothesis, the levels of many glycosyltransferases involved in “neo-synthesis” of tumor-related glycosyl epitopes and their mRNA expression have been studied; however, no conclusive results have been obtained to date.7Ito H. Hiraiwa N. Sawada-Kasugai M. et al.Altered mRNA expression of specific molecular species of fucosyl- and sialyl-transferases in human colorectal cancer tissues.Int J Cancer. 1997; 71: 556-564Crossref PubMed Scopus (118) Google Scholar, 8Salvini R. Bardoni A. Valli M. et al.β1,3-Galactosyltransferase β 3Gal-T5 acts on the GlcNAcβ 1→3Galβ 1→4GlcNAcβ 1→R sugar chains of carcinoembryonic antigen and other N-linked glycoproteins and is down-regulated in colon adenocarcinomas.J Biol Chem. 2001; 276: 3564-3573Crossref PubMed Scopus (50) Google Scholar, 9Kumamoto K. Goto Y. Sekikawa K. et al.Increased expression of UDP-galactose transporter messenger RNA in human colon cancer tissues and its implication in synthesis of Thomsen-Friedenreich antigen and sialyl Lewis A/X determinants.Cancer Res. 2001; 61: 4620-4627PubMed Google Scholar On the other hand, there is a group of carbohydrate determinants that is less expressed in cancer tissues when compared with their level in normal tissues. Because their structures are commonly more complicated, the concept of “incomplete synthesis,” that the synthesis of complex carbohydrate determinants in nonmalignant cells might be impaired upon malignant transformation, has been proposed as an important cause of cancer-associated abnormal glycosylation.10Hakomori S. Tumor-associated glycolipid antigens defined by monoclonal antibodies.Bull Cancer. 1983; 70: 118-126PubMed Google Scholar The blood group Sda carbohydrate antigen serves as a typical example among the latter group. This carbohydrate determinant is abundantly expressed on glycolipids and glycoproteins in the normal gastrointestinal (GI) tract mucosa in the majority of humans; however, its expression in cancer tissue is strikingly reduced or absent.11Morton J.A. Pickles M.M. Terry A.M. The Sda blood group antigen in tissues and body fluids.Immunol Invest. 1988; 17: 217-224Crossref PubMed Scopus (30) Google Scholar, 12Dohi T. Ohta S. Hanai N. et al.Sialylpentaosylceramide detected with anti-GM2 monoclonal antibody Structural characterization and complementary expression with GM2 in gastric cancer and normal gastric mucosa.J Biol Chem. 1990; 265: 7880-7885Abstract Full Text PDF PubMed Google Scholar The last step in the biosynthesis of Sda is catalyzed by β1,4-N-acetylgalactosaminyl-transferase (β1,4GalNAcT). The activity of the β1,4GalNAcT responsible for synthesizing the Sda determinant (Sda-β1,4GalNAcT) also dramatically decreases in gastric and colonic cancer tissue.13Dohi T. Nishikawa A. Ishizuka I. et al.Substrate specificity and distribution of UDP-GalNAc:sialylparagloboside N-acetylgalactosaminyltransferase in the human stomach.Biochem J. 1992; 288: 161-165Crossref PubMed Scopus (17) Google Scholar, 14Dohi T. Hanai N. Yamaguchi K. et al.Localization of UDP-GalNAc:NeuAc α 2,3Gal-R β 1,4(GalNAc to Gal)N-acetylgalactosaminyltransferase in human stomach Enzymatic synthesis of a fundic gland-specific ganglioside and GM2.J Biol Chem. 1991; 266: 24038-24043Abstract Full Text PDF PubMed Google Scholar Recently we reported that forced expression of Sda-β1,4GalNAcT in GI cancer cells reduced their expression of sLex/a carbohydrates and decreased their metastatic potential in nude mice, probably owing to competition with sLex/a synthases for acceptor carbohydrate.15Kawamura Y.I. Kawashima R. Fukunaga R. et al.Introduction of Sda carbohydrate antigen in gastrointestinal cancer cells eliminates selectin ligands and inhibits metastasis.Cancer Res. 2005; 65: 6220-6227Crossref PubMed Scopus (66) Google Scholar Thus, the lack of Sda antigens in cancer cells is functionally important; however, very little is known about the molecular mechanism underlying the regulation of Sda expression. In line with these hypotheses of “neo-synthesis” and “incomplete synthesis,” we compared the expression of “glyco-genes,” including glycosyltransferases and glycosidases, in normal GI mucosa with that in gastric and colorectal cancer (CRC) cells in this study. Recently, epigenetic changes, such as DNA hypermethylation, have been recognized as one of the important mechanisms for gene inactivation.16Jones P.A. Baylin S.B. The fundamental role of epigenetic events in cancer.Nat Rev Genet. 2002; 3: 415-428Crossref PubMed Google Scholar In this study, we investigated the possible role of aberrant methylation in the glycosyltransferase gene promoter region in human GI cancer cells. We also examined epigenetic changes in a group of glycosyltransferases in human gastric cancer tissues and analyzed their relation to clinicopathologic features of the cases. The gastric and colon carcinoma cell lines that were used in this study were obtained from the Japanese Collection of Research Bioresources (Tokyo, Japan) or the American Tissue Type Collection (Manassas, VA). Human CRC cell line HCT116 with genetic disruption of DNMT1 (DNMT1 KO) or both DNMT1 and DNMT3b (DKO) were established as described previously.17Rhee I. Bachman K.E. Park B.H. et al.DNMT1 and DNMT3b cooperate to silence genes in human cancer cells.Nature. 2002; 416: 552-556Crossref PubMed Scopus (1031) Google Scholar The 78 gastric tumor specimens and their paired normal tissue specimens were obtained from 78 randomly selected Japanese patients. Informed consent was obtained from all patients before the samples were collected. Quantitative polymerase chain reaction (PCR) of glyco-genes was performed by using ABI TaqMan probes (Applied Biosystems, Foster City, CA) as described previously.18Toyota M. Ho C. Ahuja N. et al.Identification of differentially methylated sequences in colorectal cancer by methylated CpG island amplification.Cancer Res. 1999; 59: 2307-2312PubMed Google Scholar, 19Suzuki H. Itoh F. Toyota M. et al.Inactivation of the 14-3-3 sigma gene is associated with 5′ CpG island hypermethylation in human cancers.Cancer Res. 2000; 60: 4353-4357PubMed Google Scholar Threshold cycle numbers (Ct) were determined with Sequence Detector software and transformed by using the ΔCt method as described by the manufacturer, with glyceraldehyde-3-phosphate dehydrogenase (GAPDH) used as the calibrator gene. Human glyco-genes examined in this study, 8 genes encoding fucosyltransferases (FUT1, FUT2, FUT3, FUT4, FUT5, FUT6, FUT7, and FUT8), 8 N-acetylgalactosaminyltransferase genes (A3GALNT, GBGT1, B3GALNT1, B3GALNT2, B4GALNT1, B4GALNT2, B4GALNT3, and B4GALNT4), 3 N-acetylglucosaminyltransferase genes (GCNT1, GCNT3, and GCNT4), 14 sialyltransferase genes (ST3GAL1, ST3GAL2, ST3GAL3, ST3GAL4, ST3GAL5, ST3GAL6, ST6GAL1, ST6GAL2, ST6GALNAC1, ST6GALNAC2, ST6GALNAC3, ST6GALNAC4, ST6GALNAC5, and ST6GALNAC6), 6 sulfotransferase genes (GALNAC4S-6ST, CHST1, CHST2, CHST3, CHST4 and GCNT5), and 4 sialidase genes (NEU1, NEU2, NEU3, and NEU4), and TaqMan probe kits used in this study are summarized in Supplementary Table 1 (see supplementary material online at www.gastrojournal.org). Human stomach and colon total RNA (BioChain, Hayward, CA) were used as normal controls; they were prepared from normal stomachs and colon mucosae pooled from healthy subjects. Flow cytometry was performed with a FACScan (BD Bioscience, Franklin Lakes, NJ). Monoclonal antibody (mAb) KM694 (directed against Sda) was provided by Tokyo Research Laboratories (Kyowa Hakko Kogyo Co, Ltd. Tokyo, Japan). We assessed gene methylation by using primers that were designed to amplify both the methylated and unmethylated alleles.20Xiong Z. Laird P.W. COBRA: a sensitive and quantitative DNA methylation assay.Nucleic Acids Res. 1997; 25: 2532-2534Crossref PubMed Scopus (1036) Google Scholar Bisulfite modification was carried out by using an EpiTect Bisulfite Kit (Qiagen, Tokyo, Japan). For combined bisulfite restriction analysis (COBRA), the PCR primers used for B4GALNT2 were 5′-ATTGGTTTTTYGTATAGGTGGTTG-3′ and 5′-CCRAACCRATTCCCACACTC-3′, yielding a PCR product of 174 bp. Primers for ST3GAL621Okajima T. Fukumoto S. Miyazaki H. et al.Molecular cloning of a novel α2,3-sialyltransferase (ST3Gal VI) that sialylates type II lactosamine structures on glycoproteins and glycolipids.J Biol Chem. 1999; 274: 11479-11486Crossref PubMed Scopus (104) Google Scholar were 5′-GTTTGTTATATYGGGTYGTAGAAG-3′ and 5′-AATTAAAACTAACRAAAACCTAAAACT-3′ (162 bp). The products were then digested with the restriction endonuclease HhaI (for B4GALNT2) or AfaI (for ST3GAL6), which cleave only methylated CpG sites. For bisulfite sequencing, the PCR primers used for B4GALNT2 were 5′-GAGAGGTGAAATTTYGGGAGTA-3′ and 5′-RACTATCCACAACCCRCAATC-3′ (430 bp). For sequencing of the bisulfite–PCR product, the DNA fragment was purified and cloned into pCR4-TOPO vector (Invitrogen, Carlsbad, CA). Clones for subsequent sequencing were randomly picked up. To detect the Epstein–Barr virus (EBV) genome in gastric tumors, we performed real-time PCR using 2 sets of primers as described previously.22Lo Y.M. Chan L.Y. Lo K.M. et al.Quantitative analysis of cell-free Epstein-Barr virus DNA in plasma of patients with nasopharyngeal carcinoma.Cancer Res. 1999; 59: 1188-1191PubMed Google Scholar Consistent results were obtained with both systems. Helicobacter pylori (HP) infection was identified by conducting histologic review of hematoxylin and eosin-stained tissue specimens and PCR assays as described by Clayton et al.23Clayton C.L. Kleanthous H. Coates P.J. et al.Sensitive detection of Helicobacter pylori by using polymerase chain reaction.J Clin Microbiol. 1992; 30: 192-200PubMed Google Scholar Genomic DNA was amplified by using exon-specific primers for p53 exons 2–11 and the mutations were examined as described previously.24Rhei E. Bogomolniy F. Federici M.G. et al.Molecular genetic characterization of BRCA1- and BRCA2-linked hereditary ovarian cancers.Cancer Res. 1998; 58: 3193-3196PubMed Google Scholar Frozen sections of 8-μm thickness were prepared from a surgical specimen. After blocking sections with 3% bovine serum albumin in phosphate-buffered saline and then incubating them with mAb KM694, bound mAbs were detected with fluorescein isothiocyanate–conjugated goat anti-mouse immunoglobulin M (Southern Biotechnology Associates, Inc, Birmingham, AL). Each tumor was classified based on tumor location; macroscopic type; lymphatic invasion; venous invasion (Japanese Gastric Cancer Association)25Japanese Gastric Cancer AssociationGastric Cancer. 1998; 1: 10-24Crossref PubMed Google Scholar; pathologic tumor, lymph node, metastasis (pTMN) classification26Sobin L.H. Wittekind C. TNM classification of malignant tumors. 5th ed. John Wiley & Sons, New York1997Google Scholar; and the Lauren classification.27Lauren P. The two histological main types of gastric carcinoma: diffuse and so-called intestinal-type carcinoma An attempt at a histo-clinical classification.Acta Pathol Microbiol Scand. 1965; 64: 31-49Crossref PubMed Scopus (4994) Google Scholar Methylation of B4GALNT2 was compared by using the Student t-test for age; the Mann–Whitney U test for tumor size, pT status, pN status, and disease stage; and the Fisher exact test for gender, tumor location, macroscopic type, histology, lymphatic invasion, venous invasion, pM status, EBV association, HP status, p53 mutation, and methylation of ST3GAL6. The Fisher exact test was carried out by using SAS (SAS Institute Inc, Cary, NC), and other statistical analyses were made with SPSS software (version 11.0; SPSS Inc, Chicago, IL). All tests were 2-tailed, and values of P < .05 were considered significant. It has been known that carbohydrate structures in GI cancers are quite different from those in normal GI epithelium. To clarify the cause of this abnormal glycosylation in GI cancer cells, we first examined the expression levels of 43 “glyco-genes,” including 8 genes encoding fucosyltransferases, 14 sialyltransferase genes, 8 N-acetylgalactosaminyltransferase genes, 3 N-acetylglucosaminyltransferase genes, 6 sulfotransferase genes, and 4 sialidase genes (Supplementary Table 1). There was no gene whose expression was universally up-regulated in the GI cancer cell lines examined when compared with normal tissues. On the other hand, we found approximately one third of glycosyltransferase genes that were expressed in normal GI mucosa but whose expression levels were decreased in many GI cancer cell lines (Figure 1A and B). This silencing of glycosyltransferases was the major cancer-associated change detected in glyco-gene expression. To examine the possibility that DNA methylation contributed to the low expression levels of these genes, we chose 12 genes containing CpG islands in their promoter region from among cancer-associated down-regulated glyco-genes. When GI cancer cells were treated with 5-aza-2′-deoxycytidine (5-aza-dC), a DNA methyltransferase inhibitor, the mRNA expression of glycosyltransferases was significantly induced in many of them (Figure 2A). On the other hand, the expression of ≥2 glycosyltransferases (FUT1 and FUT2) was not recovered by the 5-aza-dC-treatment, implying there might be certain glyco-genes whose expression was not controlled by DNA hypermethylation despite the presence of CpG islands. In the human CRC cell line HCT116 with genetic disruption of both DNMT1 and DNMT3b,17Rhee I. Bachman K.E. Park B.H. et al.DNMT1 and DNMT3b cooperate to silence genes in human cancer cells.Nature. 2002; 416: 552-556Crossref PubMed Scopus (1031) Google Scholar in which genomic DNA methylation was nearly eliminated, the expression of A3GALNT and B4GALNT2 was rescued (Figure 2B). Because it has been reported that promoter hypermethylation of the A3GALNT gene is associated with the loss of blood group A antigen expression in bladder cancer, oral squamous cell carcinoma, and gastric cancer cell lines,28Chihara Y. Sugano K. Kobayashi A. et al.Loss of blood group A antigen expression in bladder cancer caused by allelic loss and/or methylation of the ABO gene.Lab Invest. 2005; 85: 895-907Crossref PubMed Scopus (51) Google Scholar, 29Gao S. Worm J. Guldberg P. et al.Genetic and epigenetic alterations of the blood group ABO gene in oral squamous cell carcinoma.Int J Cancer. 2004; 109: 230-237Crossref PubMed Scopus (76) Google Scholar, 30Kominato Y. Hata Y. Takizawa H. et al.Expression of human histo-blood group ABO genes is dependent upon DNA methylation of the promoter region.J Biol Chem. 1999; 274: 37240-37250Crossref PubMed Scopus (82) Google Scholar our results suggest that aberrant methylation of the A3GALNT gene may lead to a cancer-associated reduction in the level of A antigen in colon cancers. Although remarkable induction of B4GALNT1 mRNA was observed after 5-aza-dC-treatment, we excluded the B4GALNT1 gene from subsequent analysis; because the expression of B4GALNT1 and GM2 gangliosides synthesized by B4GALNT1 is already known to be increased in GI cancers.31Yuyama Y. Dohi T. Morita H. et al.Enhanced expression of GM2/GD2 synthase mRNA in human gastrointestinal cancer.Cancer. 1995; 75: 1273-1280Crossref PubMed Scopus (31) Google Scholar In any case, these results strongly suggest that down-regulation of glycosyltransferases might be the leading cause of cancer-associated abnormal glycosylation and that the B4GALNT2 gene is a good representative of gene silencing by hypermethylation.Figure 2Expression of glycosylation-related genes in DNA methyltransferase-inhibited cells. (A) Six human gastric cancer (MKN45, MKN28, MKN7, MKN74, KATO III, and AZ521, as depicted in order from left to right by the gray bars) and 9 CRC (Caco2, Colo320, LoVo, SW480, HCT116, RKO, HT29, DLD1, and SW48, as depicted in order from left to right by the black bars) cell lines were treated with 2 μmol/L 5-aza-dC for 72 hours, and the expression level of each gene was then assessed by RT-PCR. The rate of induction is expressed as the ratio of treated to untreated cells. The genes that were analyzed are shown at the bottom of the bar graph. (B) RNA was harvested from HCT116 and DKO cells, and the expression level of each of the indicated genes was assessed by RT-PCR. The rate of induction is expressed as the ratio of induction in DKO cells to that in the parental HCT116 cells.View Large Image Figure ViewerDownload Hi-res image Download (PPT) The human B4GALNT2 gene encodes a β1,4GalNAcT that is responsible for the synthesis of Sda carbohydrate antigen (Sda-β1,4GalNAcT). A noteworthy characteristic of the Sda carbohydrate determinant is that its expression is restricted to normal GI mucosa and is strikingly reduced or absent in GI cancer tissue.11Morton J.A. Pickles M.M. Terry A.M. The Sda blood group antigen in tissues and body fluids.Immunol Invest. 1988; 17: 217-224Crossref PubMed Scopus (30) Google Scholar, 12Dohi T. Ohta S. Hanai N. et al.Sialylpentaosylceramide detected with anti-GM2 monoclonal antibody Structural characterization and complementary expression with GM2 in gastric cancer and normal gastric mucosa.J Biol Chem. 1990; 265: 7880-7885Abstract Full Text PDF PubMed Google Scholar So we asked if the membrane Sda structure could be detected in human CRC cell lines in which DNA methylation was suppressed. Treatment of T84 and HT29 human colonic cancer cell lines, which originally lacked the Sda carbohydrate, with 5-aza-dC resulted in an obvious increase in cell-surface expression of Sda along with the concomitant induction of B4GALNT2 expression (Figure 3A and B). When these cells were treated with butyrate, a histone deacetylase inhibitor, neither expression of Sda antigen nor B4GALNT2 mRNA was induced. We found that DNMT1 KO cells strongly expressed Sda determinants, whereas the parental HCT116 cells only weakly expressed it (Figure 3C). Furthermore, transcripts of B4GALNT2 were detected in DNMT1 KO cells, but not in the parental HCT116 cells. These results suggest collectively that DNA hypermethylation rather than histone deacetylation may contribute to the down-regulation of B4GALNT2 expression in cancer cells. Next, we examined the methylation status of the upstream of the B4GALNT2 gene in gastric cancer cell lines by COBRA. Hypermethylation in the B4GALNT2 gene was detected in 5 of 6 human gastric cancer cell lines tested (Figure 4A, left). Atypical methylation in the B4GALNT2 in a primary gastric carcinoma but not in the normal gastric mucosa adjacent to it was also found (Figure 4A, right). Because COBRA reflects the methylation status of only 2 adjoining CpG motifs, PCR products, extending from 169 bp upstream to 217 bp downstream from the translation start site and containing 39 CpGs, were subjected to bisulfite sequencing. Most of the CpGs examined were methylated in gastric cancer cells except in MKN45 cells, which were methylation negative by COBRA (Figure 4B). We also examined the methylation status of the B4GALNT2 gene in DNMT1/DNMT3b DKO cells. As expected, methylated CpGs were hardly seen in DKO cells, whereas most of the CpGs examined were methylated in the parental HCT116 cells (Figure 4C). Furthermore, it was clearly evident that the upstream of the B4GALNT2 gene was frequently hypermethylated in human gastric cancer tissues (Figure 4D). Sample KG5T, methylation negative by COBRA, and MKN45 cells looked less methylated but included apparently hypermethylated clones. These results imply that DNA hypermethylation in the promoter region of the B4GALNT2 gene may have contributed to the down-regulation of B4GALNT2 expression in gastric cancers. To understand the significance of hypermethylation in the B4GALNT2 gene, we analyzed the methylation status of the B4GALNT2 gene and clinicopathologic characteristics of patients with gastric carcinomas. We deemed that the B4GALNT2 gene was methylated when the percentage of methylated DNA was ≥10% by COBRA. Of the 78 primary gastric tumors studied, 39 were classified as methylation positive (Table 1). Univariate analysis revealed no difference between the methylation-positive and -negative groups with respect to age, gender, tumor location, macroscopic type, lymphatic invasion, venous invasion, or pT, pN, or pM status. However, there were significant differences between patients in the methylation-positive and -negative groups with respect to histology (P = .012) and EBV status (P = .001). EBV was detected in 10 of the 78 tumors, and all EBV-associated tumors were methylation-positive ones. No difference was noted in the frequency of p53 mutation or the infection of HP between the methylation-positive and -negative groups. To examine the correlation between the Sda expression and DNA hypermethylation of B4GALNT2, we determined the expression levels of Sda carbohydrates in freshly frozen gastric cancers, because Sda antigen is expressed as a glycolipid in the stomach; its reactivity to antibodies was lost in formalin-fixed paraffin-embedded samples that we used for our clinicopathologic analysis. Of the 15 freshly frozen gastric cancers studied, the expression of Sda determinants was totally lost in all cases as determined by immunohistologic staining; 7 cases were methylation positive by COBRA (data not shown).Table 1Clinicopathologic Features of Gastric Cancer With or Without Methylation of B4GALNT2Number of patients (%)CharacteristicsTotalMethylatedUnmethylatedP-valueNumber of patients7839 (50.0)39 (50.0)Mean age ± SD (y)63.6 ± 13.665.4 ± 10.3.531Gender Male52 (66.6)24 (61.5)28 (71.8).472 Female26 (33.3)15 (38.5)11 (28.2)Tumor location Upper one third22 (28.2)14 (35.9)8 (20.5).279 Middle one third23 (29.5)9 (23.1)14 (35.9) Lower one third33 (42.3)16 (41.0)17 (43.6)Macroscopic type 04 (5.1)2 (5.1)2 (5.1).98 16 (7.7)3 (7.7)3 (7.7) 230 (38.5)14 (35.9)16 (41.0) 330 (38.5)15 (38.5)15 (38.5) 48 (10.3)5 (12.8)3 (7.7)Histology (Lauren) Intestinal36 (46.2)12 (30.8)24 (61.5).012 Diffuse42 (53.8)27 (69.2)15 (38.5)Lymphatic invasion Negative20 (25.6)9 (23.1)11 (28.2).78 Positive58 (74.4)30 (76.9)28 (71.8)Venous invasion Negative37 (47.4)20 (51.3)17 (43.6).651 Positive41 (52.6)19 (48.7)22 (56.4)Pathologic tumor classification pT15 (6.4)3 (7.7)2 (5.1).407 pT243 (55.1)19 (48.7)24 (61.5) pT328 (35.9)15 (38.5)13 (33.3) pT42 (2.6)2 (5.1)0 (0.0)Pathologic lymph node status pN022 (28.2)9 (23.1)13 (33.3).373 pN128 (35.9)15 (38.5)13 (33.3) pN216 (20.5)8 (20.5)8 (20.5) pN312 (15.4)7 (17.9)5 (12.8)Pathologic metastasis status pM066 (84.6)36 (92.3)30 (76.9).114 pM112 (15.4)3 (7.7)9 (23.1)Stage (pTNM) I18 (23.1)8 (20.5)10 (25.6).804 II16 (20.5)8 (20.5)8 (20.5) III21 (26.9)12 (30.8)9 (23.1) IV23 (29.5)11 (28.2)12 (30.8)Helicobacter pylori Positive65 (83.3)33 (84.6)32 (82.1).999 Negative13 (16.7)6 (15.4)7 (17.9)Epstein–Barr virus Positive10 (12.8)10 (25.6)0 (0.0).001 Negative68 (87.2)29 (74.4)39 (100.0)p53 mutation Positive19 (24.4)6 (15.4)13 (33.3).112 Negative59 (75.6)33 (84.6)26 (66.7)SD, standard deviation; pTNM, pathologic tumor, lymph node, metastasis status according to the International Union Against Cancer classification system. Open table in a new tab SD, standard deviation; pTNM, pathologic tumor, lymph node, metastasis status according to the International Union Against Cancer classification system. Finally, we examined whether epigenetic changes occurred in the ST3GAL6 gene together with those in the B4GALNT2 gene in human gastric cancer cells. Of the cancer-associated down-regulated glyco-genes that we found in the present study, the ST3GAL6 was hypermethylated in concurrence with methylation of the B4GALNT2 in many of the gastric cancer cell lines as well as in gastric cancer tissues (Figure 4A and Figure 5). As shown in Table 2, aberrant methylation in the ST3GAL6 was detected in 24 of 32 primary gastric tumors with statistically significant correlation with the methylation of B4GALNT2 and EBV status (P < .01). No difference was noted in the frequency of p53 mutation between the ST3GAL6-methylated and -unmethylated groups. These results strongly suggest that epigenetic changes may occur in a group of glyco-genes including B4GALNT2 and ST3GA