Glutamate Receptor, Ionotropic, Kainate 2 Silencing by DNA Hypermethylation Possesses Tumor Suppressor Function in Gastric Cancer

IJC International Journal of Cancer

Glutamate receptor, ionotropic, kainate 2 silencing by DNA hypermethylation possesses tumor suppressor function in gastric cancer

Chi-Sheng Wu1, Yen-Jung Lu2, Hsin-Pai Li1,3, Chuen Hsueh4, Chang-Yi Lu2, Yu-Wei Leu5, Hao-Ping Liu3, Kwang-Huei Lin1,3, Tim Hui-Ming Huang6 and Yu-Sun Chang1,3

1 Graduate Institute of Biomedical Sciences, Chang Gung University, Taoyuan, Taiwan 2 Institute of Microbiology and Immunology, National Yang-Ming University, Taipei, Taiwan 3 Molecular Medicine Research Center, Chang Gung University, Taoyuan, Taiwan 4 Department of Pathology, Chang Gung Memorial Hospital, Lin-Kou, Taoyuan, Taiwan 5 Department of Life Science, Institute of Molecular Biology, National Chung-Cheng University, Chia-Yi, Taiwan 6 Department of Molecular Virology, Immunology & Medical Genetics-Human Cancer Genetics, Ohio State University, OH

Aberrant DNA methylation is considered a major mechanism for silencing tumor suppressor genes in gastric cancer. We used CpG microarray and differential methylation hybridization strategies to identify potential tumor suppressor genes and recovered glutamate receptor, ionotropic, kainate 2 (GRIK2) as a novel epigenetic target in gastric cancer. Additional experiments showed that the promoter region of GRIK2 was hypermethylated in 3 of the 4 tested gastric cancer cell lines, and its expression was restored by treatment of cells with the DNA methylation inhibitor, 50-aza-dC. In clinical samples, the GRIK2 promoter was differentially hypermethylated in tumor tissues compared with adjacent normal tissues (p < 0.001), and this methylation was inversely correlated with the expression level of GRIK2 mRNA (r 520.44). Functional studies further showed that GRIK2-expressing gastric cancer cell lines showed decreased colony formation and cell migration. Taken together, these results suggest that GRIK2 may play a tumor-suppressor role in gastric cancer. Future studies are warranted to examine whether DNA hypermethylation of the GRIK2 promoter can be used as a potential tumor marker for gastric cancer. Cancer Cell Biology

Gastric cancer is one of the most common human cancers with chemotherapy is a common treatment for gastric cancer. worldwide and is a major upper gastrointestinal tract malig- Factors such as diet, tobacco use and Helicobacter pylori nant disease with poor prognosis for patients suffering from infection have been reported as the major risk factors for gas- more advanced stages of the disease.1 In the 2009 Taiwan tric carcinoma.2 Cancer Registry Report, this cancer was ranked as having the Numerous studies have sought to identify biomarkers ca- seventh highest incidence among cancers. Surgery combined pable of improving the prognosis for patients with gastric cancer.3 Similar to other cancers, the development of gastric cancer is a multistep process involving a variety of genetic Key words: gastric cancer, GRIK2, tumor suppressor gene, and epigenetic modifications. Some of the cellular molecules hypermethylation related to gastric cancer development and progression seem Additional Supporting Information may be found in the online to be regulated via DNA methylation.4 DNA methylation fre- version of this article quently occurs at CpG islands, which are short stretches of The first two authors contributed equally to this work. GC-rich sequences frequently located on promoters and in Grant sponsor: Ministry of Education, National Science Council; the first exon of genes.5 Such methylation can have a pro- Grant numbers: NSC 94-2314-B-182A-188, 94-3112-B-182-005, 95- found effect on the expression of tumor suppressor genes in 2320-B-182-001, 97-3112-B-182-008; Grant sponsor: Chang Gung various cancers. Therefore, aberrant DNA methylation of tu- Memorial Hospital, Taiwan; Grant numbers: CMRPD150961, mor suppressor genes may be useful as a good biomarker for CMRPG360221, CMRPG360262 cancers.6 In gastric cancer, genes such as p14ARF, p16INKaa,7 DOI: 10.1002/ijc.24958 MGMT, hMLH18 and APC9 reportedly play tumor-suppres- History: Received 23 Apr 2009; Accepted 29 Sep 2009; Online 12 sor roles and are regulated by DNA modification. During the Oct 2009 development of gastric cancer, it is believed that some tumor Correspondence to: Yu-Sun Chang, Chang Gung Molecular suppressor genes undergo DNA hypermethylation, leading to Medicine Research Center and Graduate Institute of Biomedical their decreased expression.10 Sciences, Chang Gung University, 259 Wen-Hwa 1st Road, Glutamate receptors are membrane proteins responsible Kwei-Shan, Taoyuan, 333 Taiwan, Tel: 886-3-211-8800 x5131; for mediating most excitatory neurotransmissions in the 886-3-211-8683, Fax: 886-3-211-8683, E-mail: [email protected] mammalian central nervous system (CNS).11 They are

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC Wu et al. 2543

Table 1. Clinical characteristics of gastric cancer samples Case no. Hypermethylated Hypomethylated p Gender Female 11 9 2 0.231 Male 16 9 7 Age (yr) <65 13 9 4 1 65 14 9 5 Depth of invasion (pT) T1, T2 8 5 3 1 T3, T4 19 13 6 Lymph node status (pN) N0 6 5 1 0.628 Non N0 21 13 8 Histological type Intestinal 11 6 5 0.393 Diffuse 14 10 4 Mix type 2 2 0 Helicobacter pylori infection Negative 21 15 6 0.367 Positive 6 3 3 Paired samples Adjacent normal 27 2 25 <0.0011 Tumor 27 18 9 Cancer Cell Biology P values were calculated using Pearson Chi-square test. 1statistically significant when <0.05. classified into 2 major groups: the metabotropic glutamate re- Although GRIK2 has been implicated in several neurologi- ceptor (mGlu) group, which is composed of mGlu1-8, and cal diseases, such as Huntington’s disease,19 autosomal reces- the ionotropic glutamate receptors, which include the N- sive mental retardation,20 autism21 and manic-depressive ill- 22 methyl-D-aspartate (NMDA), a-amino-3-hydroxy-5-methyl- ness, little is known about its role in tumor progression. A iso-xazole-4-propionate (AMPA) and kainate glutamate report by Sinclair et al. (2004) suggested that GRIK2 may be receptors. Glutamate receptor, ionotropic, kainate 2 (GluR6 a candidate tumor-suppressor gene for acute lymphocytic or GRIK2) is 1 of the 5 members of the kainate glutamate re- leukemia (ALL), but no subsequent study has examined the ceptor subgroup, which also includes GluR5-7, KA1 and relevant biological functions or regulation mechanisms. We KA2. Although kainate receptors are distributed throughout show for the first time that GRIK2 is regulated by DNA the CNS, their physiological significance is not yet known. methylation in gastric cancer and further report that colony GRIK2, located on chromosome 6q16.3-q21, is a 17-exon formation and cell migration are suppressed in gastric cancer gene12 that may be transcribed into at least 6 different splice cell lines expressing GRIK2, strongly indicating that GRIK2 is variants.13 Under physiological conditions, GRIK2 forms a a tumor-suppressor gene. homomeric receptor channel14 or a heteromeric receptor channel with GluR5.15 Recent studies have indicated that the Material and Methods ionotropic glutamate receptors, NMDA2A and NMDA2B, Cell culture, clinical samples and 50-aza-dC treatment may play tumor-suppressor functions in esophageal cancer,16 AGS cells cultured in Ham’s F12 medium containing 10% fe- gastric cancer17 and colorectal carcinoma.18 The expression tal bovine serum (FBS). TMC1 cells were maintained in level of these NMDA receptors is regulated at least in part RPMI medium containing 10% FBS. AZ521 and KATO III through changes in DNA methylation. However, no func- cells were kindly provided by Dr. Lin KH of Chang Gung tional analysis has suggested that kainate receptors may play University (Taiwan) and were cultured in RPMI containing a tumor-suppressor role or that their expression might be 10% FBS. Twenty-seven paired gastric cancer/normal adja- regulated by DNA methylation in cancer cells. cent tissue samples were collected from patients between

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC 2544 Glutamate receptor, ionotropic, kainate 2 silencing

2000 and 2003 and were obtained from the Chang Gung Me- Bisulfite treatment and bisulfite sequencing morial Hospital (CGMH) Tumor Bank (Taoyuan, Taiwan) Genomic DNA (1 lg) was modified by sodium bisulfite using (Table 1). Two normal adult stomach genomic DNA samples the EZ DNA methylation kit (Zymo Research, USA). The (N24 and N33) were purchased from BioChain Institute CpG islands of the GRIK2 promoter region were amplified (USA). This study was reviewed and approved by the institu- using primers 50-GTTTGGTAAAATTTTTGTTAGTAAAG-30 tional review board and ethics committee of CGMH. and 50-AATTCCTTAAAAATATCCAATCCAC-30. The PCR Informed consent was obtained from all patients and healthy products were cloned into a TA vector and sequenced. The controls. sequencing results were analyzed with the Vector NTI 9.0 One day before experiments, gastric cancer cell lines software (Invitrogen, USA). (AGS, AZ521 and TMC1) were seeded in 10-cm dishes at a confluence of 20–30% per dish. KATO III cells were seeded 0 Quantitative methylation-specific PCR in 25-T flasks at a density of 20–30% per flask. For 5 -aza-dC Bisulfite-modified DNA was subjected to real-time quantita- l 0 treatment, cells were treated with 5 M5-aza-dC (Sigma, tive methylation-specific PCR (Q-MSP) using a Bio-Rad iCy- USA) for 5 days. The culture medium was replaced every l 0 cler (Bio-Rad, USA). Each reaction contained 7.5 lof2 24 hr with fresh culture medium containing 5 -aza-dC. SYBR Green supermix (Bio-Rad), 0.2 lM of each primer and 10 ng of bisulfite-modified DNA in a total volume of 15 ll. Differential methylation hybridization and CpG microarray The reaction conditions consisted of 95C for 3 min, fol- Differential methylation hybridization (DMH) was performed lowed by 50 cycles of 95C for 15 sec, 60C for 20 sec, 72C 23 l as previously described. In brief, 2 g genomic DNA was for 20 sec and 80C for 10 sec. A CpG-free region of the digested with MseI, and the digested DNA fragments were 0 ACTB gene was used as an internal reference, as previously ligated with the synthetic adaptor, H24/H12 (5 -AGG described.26 The amount of methylated DNA was determined CAACTGTGCTATCCGAGGGAT-30,50-TAATCCCTCGGA- 0 for each sample by plotting the threshold cycle number 3 ). The resulting fragments were further digested with the against a standard curve generated from CpGenomeTM CG methylation-sensitive enzymes, BstUI and HpaII (New Universal Methylated DNA (Chemicon, USA). To determine England Biolabs). After amplification by PCR using H24 as the relative amount of methylated DNA in each sample, the the primer, the amplicons were purified using a NucleoSpin values of the target gene were compared with that of the Extract II kit (Macherey-Nagel, Germany), labeled with Cy3 internal reference gene to obtain a ratio, which was then dye and hybridized with a homemade CpG microchip con- Cancer Cell Biology multiplied by 100 to give a percentage value. taining 9216 CpG-rich DNA clones (adapted from that of Dr. Tim Hui-Ming Huang, Ohio State University, OH).24,25 Resequencing of these plasmid clones confirmed that this RNA extraction, cDNA synthesis and quantitative library contains 3,832 individual clones (http://163.25.91.176). real-time RT-PCR The microarray was scanned using a Gene Pix 4200 scanner Cells were cultured in culture medium, and total RNA was (Axon, CA, USA), and the results were analyzed using the isolated with TRIZOL reagent according to the manufac- GenePix 6.0 software. turer’s protocol (Invitrogen). Total RNA of human adult nor- Duplicate CpG microarray data from the 4 gastric cancer mal gastric tissue was purchased from BioChain Institute cell lines (experimental group) and 2 purchased normal (USA). cDNA was generated by reverse transcription from l stomach tissues (control) were analyzed using the Genedata 1 g of RNA, using Impron II reverse transcriptase (Prom- Expressionist Pro3.0 software (Genedata AG, Basel, Switzer- ega) and oligo-dT primers. The resulting cDNA was diluted land). The signals from these samples were normalized by 1/20 and subjected to real-time PCR, using a Bio-Rad iCycler with 7.5 ll SYBR green super-mix and 30 pmol GRIK2 gene global median normalization, with the median intensity value 0 0 0 > < primers 5 -GCTGCTCTAATGTATGATGCTGT G-3 and 5 - set to 250 (cancer cell line 500 and normal 250). Because 0 cancer cells presumably have a higher methylation level and TGTGATGTTCGCTGGCTTTCC-3 . The reaction conditions consisted of 95C for 3 min, followed by 50 cycles of 95C therefore a higher CpG microarray signal intensity than normal tissues, we selected genes whose normalized intensity val- for 15 sec, 60 C for 20 sec, 72 C for 20 sec and 80 C for ues were 2-fold or more greater in the gastric cancer cell lines 10 sec. GAPDH was used as an internal control. compared with the normal controls. In addition, genes whose p value was <0.001 (student t test) were selected. Further- Transfection and immunofluorescence more, redundant CpG clones and clones whose sequences The human GRIK2 (NM_021956.2) expression vector did not match those of a known promoter region (2 kb pCMV6-Entry-GRIK2 (Origene, USA) was transfected into upstream or downstream from the transcription start site) AGS and AZ521 cells using Lipofectamine 2000 (Invitrogen). were excluded from this study. In the end, we identified 107 The transfected cells were grown on cover slides for 24 hr, genes that were hypermethylated in gastric cancer cell lines fixed with 3.7% formaldehyde for 30 min at room tempera- compared with normal stomach tissues. Our detailed analysis ture and permeabilized and blocked with 0.1% saponin con- strategy is shown in Supporting Information Figure S1. taining 1% bovine serum albumin for 15 min. The cover

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC Wu et al. 2545 slides were incubated with an anti-GRIK2 antibody (1:100 Cell proliferation assay Chemicon, USA) for 2 hr at room temperature, rinsed twice AGS (5 104) and AZ521 (2.5 104) cells were plated on with 1 phosphate-buffered saline (PBS) and incubated with 6-well plates in culture medium, and total cell numbers were an anti-rabbit FITC-conjugated secondary antibody (1:200, counted for 4 days. The experiment was performed at least 3 Jackson ImmunoResearch Laboratories, USA) for 45 min at independent times, each time in duplicate. room temperature. For visualization of individual cells, the cover slides were incubated with DAPI solution for 5 min and washed several times with 1 PBS. Finally, the cover slides were mounted with Vectashield reagent (Vector Labo- Wound closure assay ratories, CA) and stored at 4C. Stably transfected GRIK2-expressing AGS and AZ521 clones were grown to confluence in 35-mm dishes containing culture medium. For wounding, the cells were scraped mechani- Establishment of stable transfectants cally with a sterile 10-ll plastic pipette tip. The wounded sur- AGS and AZ521 cells were transfected with control vector face was then rinsed with 1 PBS, the image was captured (pCMV6-Entry)orGRIK2 expression vector (pCMV6-Entry- (AXIOVERT 200 MAT, ZEISS) at time 0 hr, the culture me- GRIK2) using Lipofectamine 2000 (Invitrogen). Transfectants dium was refreshed and the cells were cultured for an addi- were selected by exposure to G418 (800 lg/ml) for 10 days. tional 9, 18 and 24 hr for AGS cells, 48 hr for AZ521 cells. Single clones were amplified, and GRIK2 expression was Cells were counted from 10 random fields within the residual examined by Western blotting. wound area under 200 magnification, and averages were calculated. Immunohistochemical staining analysis Immunohistochemical analysis was performed using an auto- matic immunohistochemistry staining device, according to Transwell migration assay the manufacturer’s suggested procedure (Vision BioSystems, Transwell migration assays were performed in a 24-well VIC, Australia). Tissue sections and normal tissue array Transwell chamber (Corning, USA) fitted with a polycarbon- (#BN1002, US Biomax) were retrieved using Bond Epitope ate membrane (8-lm pore size). AGS (5 104) and AZ521 Retrieval Solution 1 (pH 6.0) on a Bond-max automated (1 105) cells were washed twice with serum-free medium, immunostainer (Vision BioSystems, VIC, Australia), and resuspended in 100 ll of serum-free medium and added to Cancer Cell Biology stained with anti-GRIK2 antibody (1:100; Chemicon, USA). the upper chamber. The lower chamber contained 10% FBS A polymer detection system (Bond Polymer Refine, Vision medium. After 2.5 (for AGS cells), 7 and 24 hr (for AZ521 BioSystems) was used to reduce nonspecific staining. Tissue 0 cells), the migrated cells were fixed and stained for 15 min sections were treated with liquid 3,3 -diaminobenzidine rea- with 0.25% crystal violet, 10% formaldehyde and 80% metha- gent, using 30-diaminobenzidine tetrahydrochloride as the nol, and then washed 5 times with ddH2O for removal of chromogen and hematoxylin as the counterstaining reagent. nonadherent cells. Ten random fields under 100 magnification were captured for each membrane, and the migrated Western blot analysis cells were counted. For spectrophotometric detection of Cells were lysed in RIPA buffer, and 50 lg of total protein migration ability, membranes were stained with crystal violet was resolved on a 10% SDS polyacrylamide gel and electro- and soaked in 10% acetic acid, and the OD570 was transferred to a nitrocellulose membrane. The membrane was measured. blocked and then incubated with anti-GRIK2 (1:1000; Chemi- con, AB5683) or anti-tubulin (1:10000; MDBio, Taiwan) antibodies overnight at 4C. Horseradish peroxidase-conjugated Statistical analysis immunoglobulins were used as secondary antibodies, and Comparisons of GRIK2 methylation or gene expression levels proteins were detected using an ECL system (Amersham, between paired tumor and adjacent normal tissue samples UK) and X-ray films. were performed using the Wilcoxon signed rank test. For the anchorage-independent colony formation, cell proliferation Anchorage-independent colony formation assay and migration ability assays, statistical significance was exam- For anchorage-independent assays, 3 103 AGS cells and ined using t-tests. The cutoff value for Q-MSP was deter- 1 104 AZ521 cells were stably transfected with GRIK2- mined by ROC analysis.27 The chi-square test was used to expressing or empty vectors were seeded in 6-well plates with calculate differences in gender of the patient, tumor stage 0.35% top agar and 0.7% bottom agar. After 4 weeks, foci and methylation status between adjacent normal tissues and were stained with 0.005% crystal violet and counted under a tumor tissues. These analyses were performed using SPSS dissecting microscope. The experiment was performed at least 13.0 software (SPSS, Chicago, IL). P values <0.05 were con- 3 independent times, each time in duplicate. sidered statistically significant.

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC 2546 Glutamate receptor, ionotropic, kainate 2 silencing

Results adjacent normal tissues. Immunohistochemistry analysis of Hypermethylation of GRIK2 CpG islands in GRIK2 expression in normal tissue array also supported that gastric cancer cell lines GRIK2 is expressed at higher level in normal stomach samples DMH experiments were performed on DNA from 4 gastric (Supporting Information Fig. S2a and Supporting Information cancer cell lines (AGS, AZ521, KATO III and TMC1) and 2 Table I). Furthermore, bisulfite sequencing of DNA from the normal adult stomach samples (N24 and N33). The detailed paired tissue samples showed that the GRIK2 CpG sites were procedures are described in the ‘‘Material and Methods’’ sec- more heavily methylated in tumor samples vs. adjacent nor- tion and Supporting Information Figure S1a. In brief, using a mal tissues (70 vs. 21% in sample #89001, and 78 vs. 20% in homemade CpG island microarray (comprising 8000 CpG sample #92024; Fig. 2b). On the basis of these findings, we clones corresponding to 3,832 individual genes) adapted from further analyzed 27 pairs of clinical samples, using Q-MSP Dr. TH Huang,28 we performed DMH and identified 107 and quantitative real-time PCR to test whether there was a genes that were differentially hypermethylated in the 4 gastric correlation between GRIK2 promoter methylation and gene cancer cell lines (AGS, AZ521, KATOIII and TMC1) but not expression in these samples. As shown in Figure 2c, a statisti- in the 2 adult normal stomach tissue samples. Among these cally significant difference (p < 0.001) was observed between individual genes, there were 7 probes located within the pro- the normal and tumor tissues of the 27 paired samples, indi- moter and exon 1 of GRIK2 gene, and all 7 probes were cating that GRIK2 gene is differentially methylated in gastric shown very significant p values. The average GRIK2 probe in- cancer. Across all of the clinical samples, the tumor tissues tensity of AGS, AZ521, KATOIII and TMC1 was significantly tended to be hypermethylated, whereas their adjacent normal higher than that of N33 and N24 (Supporting Information Fig. tissues tended to be hypomethylated (Supporting Information S1b). In addition, GRIK2 may be a candidate tumor suppressor Fig. S3). Using 40.8% methylation (per Q-MSP) as the cutoff gene for ALL.29 Furthermore, several ionotropic glutamate value, hypermethylation of GRIK2 could be used to reliably receptors such as NMDA2A and NMDA2B whose expression differentiate tumor samples from adjacent normal samples level is regulated at least in part through changes in DNA (Table 1; chi-square analysis, p < 0.001). To test whether methylation and may play tumor-suppressor functions in sev- GRIK2 promoter hypermethylation was correlated with lower eral solid tumors, including gastric cancer.17 This prompted us gene expression in clinical samples, Q-RT PCR was per- to investigate whether this gene could be a tumor-suppressor formed. As shown in Figure 2d, GRIK2 expression was signifi- gene in gastric cancer. The promoter region of GRIK2 (http:// cantly lower in tumor tissues compared with adjacent normal www.urogene.org/) contains 2 CpG islands (CpG1, 413 to tissues across the 27 paired biopsy samples (p ¼ 0.001) and Cancer Cell Biology 212; CpG2, 110 to þ95) comprising a total of 30 CpG there was an inverse correlation between GRIK2 methylation dinucleotides (Fig. 1a). Bisulfite sequencing showed that the and its gene expression level (r ¼0.44 and p ¼ 0.022; Fig. CpG sites of the GRIK2 gene were densely methylated in AGS 2e). Collectively, these data strongly suggest that GRIK2 (93%), AZ521 (51%) and TMC1 (97%) cells, but only sparsely expression is very likely to be regulated through promoter methylated in KATO III cells (26.6%) and the 2 normal stomach methylation in patients with gastric cancer. samples (8.6% in N24 and 20% in N33) (Fig. 1b). These results were confirmed by the use of Q-MSP (Fig. 1c). To determine GRIK2 does not alter cell proliferation but does affect whether GRIK2 gene expression is regulated by DNA methyla- anchorage-independent colony formation tion, we used quantitative real-time PCR to compare GRIK2 Because GRIK2 expression seems to be regulated by methyla- gene expression in gastric cancer cell lines treated with and with- tion and GRIK2 expression is lower in tumor samples com- out 50-aza-dC. In AGS, AZ521 and TMC1 cells, but not KATO pared with adjacent normal tissues, we next sought to deter- III cells, GRIK2 mRNA levels were restored after treatment with mine whether GRIK2 can function as a tumor suppressor. The 50-aza-dC for 5 days (Fig. 1d). These results suggest that GRIK2 GRIK2 expression plasmid, pCMV6-Entry-GRIK2, was trans- expression in AGS, AZ521 and TMC1 cells is likely to be regu- fected into AGS cells, and GRIK2 expression was observed by lated through DNA methylation of the promoter region. immunofluorescence and Western blotting (Fig. 3a). GRIK2 expression was undetectable in parental AGS cells and cells transfected with the vector control, pCMV-Entry (Fig. 3a). Hypermethylation of the GRIK2 promoter region Similar results were also observed in AZ521 cells (Supporting in clinical samples Information Fig. S4a). To investigate the biological functions Because the GRIK2 promoter region was found to be hyper- of GRIK2, we examined the proliferation of an AGS cell line methylated in 3 of the 4 gastric cancer cell lines, we specu- stably expressing GRIK2 (G8), a GRIK2-expressing mixed lated that GRIK2 expression might be decreased in tumor tis- clone (GM) and 2 AGS cell lines stably transfected with empty sues. Twenty-seven paired clinical samples were examined for vector (C1 and C2). After 4 days, there was no difference in immunohistochemical staining, Q-MSP and quantitative real- growth rate among these cell lines (Fig. 3b). In contrast, an- time RT-PCR. As shown in Figure 2a and Supporting Infor- chorage-independent colony formation assays performed on a mation Figure S2b, immunohistochemistry showed that vector-transfected mixed clone (CM) and GM showed that the GRIK2 expression was lower in tumor tissues compared with colony formation ability of CM clones was significantly higher

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC Wu et al. 2547 Cancer Cell Biology

Figure 1. Methylation status for the CpG islands of GRIK2 in gastric cancer cell lines. (a) Genomic map of the GRIK2 promoter region. The 2 CpG islands, 413 to 212 and 110 to þ95, contain 30 CpG dinucleotides; þ1 is the GRIK2 transcription start site. Each vertical bar represents 1 CpG dinucleotide. (b) Bisulfite sequencing analysis of the GRIK2 CpG islands. Bisulfite sequencing was performed with samples from 4 gastric cancer cell lines (AGS, AZ521, TMC1 and KATO III) and 2 stomach samples from healthy individuals (N24 and N33). Five individual clones from each sample were analyzed. The open and filled squares represent unmethylated and methylated CpG sites, respectively. Horizontal bars indicate the positions of the utilized methylation-specific PCR primers. The % methylation of the 30 CpG sites is shown in parenthesis. (c) Quantitative methylation-specific PCR analysis of GRIK2 methylation in gastric cancer cell lines and stomach samples from healthy individuals. Results were obtained from 3 individual experiments. (d) Recovery of GRIK2 expression in gastric cancer cell lines after 50-aza-dC treatment. The relative fold change was calculated by dividing the quantitative RT-PCR data from cells treated with 50-aza-dC by those from untreated cells. [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC 2548 Glutamate receptor, ionotropic, kainate 2 silencing Cancer Cell Biology

Figure 2. Analysis of GRIK2 expression and the gene methylation status of GRIK2 CpG islands in clinical samples. (a) Paired sample sets from patients with gastric cancer were analyzed by IHC. GRIK2 positivity was mainly detected in the adjacent normal tissues (brown). Bar, 100 lm; 400 magnification. (b) Methylation status of GRIK2 CpG islands in paired gastric cancer samples. Hypermethylation of GRIK2 CpG islands was present in the tumor samples but not the adjacent normal tissues of the samples shown in (a). Bisulfite sequencing analysis of tumor and adjacent normal tissues from 2 patients with gastric cancer was performed to determine the % methylation of the GRIK2 CpG islands. Open and filled squares represent unmethylated and methylated CpG sites, respectively. Horizontal bars indicate the positions of the quantitative methylation-specific PCR primers. (c) GRIK2 gene methylation status, as analyzed by Q-MSP assay of 27 paired clinical samples. Box blots showed significantly higher % methylation in tumor tissues compared with paired adjacent normal tissues (p < 0.001). (d)AnalysisofGRIK2 mRNA expression in 27 paired clinical samples. Box blots showed significantly lower GRIK2 expression in tumor tissues compared with adjacent normal tissues (p ¼ 0.001). (e) Inverse correlation between GRIK2 methylation and gene expression. The differences in the methylation of GRIK2 CpG islands (Q-MSP of T minus N) and the relative fold change of mRNA expression (T over N) were significantly and inversely correlated (r ¼ 0.44, p ¼ 0.022). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

than that of GM clones after 4 weeks (p ¼ 0.02; Fig. 3c). Simi- GRIK2 suppress cell migration lar results were obtained in AZ521 cells (Supporting Informa- Next, we tested GRIK2 for another tumor-suppressor activity, tion Fig. S4b). These results suggest that GRIK2 inhibits col- namely the ability to inhibit cell migration. By using the ony formation but does not affect cell proliferation in gastric above-described GRIK2-expressing AGS cell clones, we per- cancer cell lines. formed wound healing assays. As shown in Figures 4a and

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC Wu et al. 2549 Cancer Cell Biology

Figure 3. Effects of GRIK2 expression in AGS cells. (a) Image analysis of GRIK2 in transiently expressing cells. One lgofpCMV6-Entry-GRIK2 expression vector or empty vector was transiently transfected into AGS cells. GRIK2-positive cells appeared as green fluorescent cells; protein expression was confirmed by Western blotting. (b) Cell proliferation assay. AGS clones stably expressing GRIK2 G8, and GRIK2- expressing mixed clone (GM), and 2 clones stably transfected with control vector alone (C1 and C2) were checked by Western blotting for GRIK2 expression (G8 and GM cells showed various levels of GRIK2 expression), and then subjected to cell proliferation assays. (c) Anchorage-independent colony formation by AGS cells. Vector-transfected mixed cells (CM) and GRIK2-expressing mixed cells were grown in 0.35% top agar and maintained in 400 ng/ml G418. Colonies were counted after 4 weeks, and the results from 3 independent experiments were averaged. GRIK2-expressing cells generated significantly fewer colonies than cells transfected with the control vector (p < 0.05). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

4b, the cell migration ability of CM cells was significantly 0.001; Fig. 4f). Similar results were obtained in AZ521 cells lower than that of CM or parental AGS cells, as measured af- (Supporting Information Figs. S5c and S5d). ter 18 and 24 hr of wound healing. Similarly, the number of cells migrating to the scraped space was significantly lower in Discussion G8 cells vs. C1 cells (p < 0.0001; Fig. 4d). Similar results Previous studies have shown that many tumor suppressor were obtained in AZ521 cells (Supporting Information Figs. genes are inactivated by epigenetic modification, specifically S5a and S5b). In a trans-well migration assay measuring the by DNA methylation.30 In this study, we used DMH com- number of AGS GRIK2-expressing cells that migrated bined with CpG microarray analysis28 to identify genes that through a polycarbonate membrane over the course of 2.5 hr were differentially methylated in gastric cancer, in the hope (Fig. 4e), significantly fewer G8 and GM cells migrated com- that 1 or more of them would prove to be novel candidate pared with the number of migratory C1 and C2 cells (p < tumor suppressors. We demonstrated that GRIK2 promoter

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC 2550 Glutamate receptor, ionotropic, kainate 2 silencing Cancer Cell Biology

Figure 4. Wound closure and migration by AGS cells stably expressing GRIK2.(a) Wound closure assays were performed using AGS cells, AGS cells expressing mixed clones of GRIK2 (GM) and a mixed vector control (CM). Images were obtained at 0, 18 and 24 hr. Dashed lines indicate the scraped edges at 0 hr. Cells that had migrated into the area between 2 dashed lines were counted. The relative fold differences in the numbers of migrated cells are shown in (b). AGS and CM cells showed better migration ability than GM at 18 and 24 hr. (c) Wound closure by AGS cells stably expressing GRIK2. Images were obtained at 0, 9 and 18 hr. Dashed lines indicate the scraped edges at 0 hr. Cells that had migrated into the area between the 2 dashed lines were counted. The relative fold difference in the number of migrated cells is given in (d). C1 cells showed significantly better migration ability than GRIK2-expressing G8 cells at 18 hr (p < 0.001). (e) Trans-well migration assay. AGS clones stably expressing GRIK2 or vector control were subjected to Trans-well migration assays as described in the ‘‘Material and Methods’’ section. Images were captured at 2.5 hr under 200 magnification. Cells were counted from 10 randomly picked fields and averages were calculated; results were obtained from 3 independent experiments. The relative fold change in the number of migrated cells is shown, with the results from C1 cells given as 1. The migration abilities of GRIK2-expressing G8 and GM cells were significantly lower than that of C1 cells (p < 0.05) (f). [Color figure can be viewed in the online issue, which is available at www.interscience.wiley.com.]

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC Wu et al. 2551 methylation was inversely correlated with its gene expression (data not shown). Furthermore, the expression level of Cyclin in clinical samples and that ectopic expression of GRIK2 D1, a well-defined target gene of the b-catenin signaling path- inhibited the colony formation and cell migration abilities of way, remained unchanged in GRIK2-expressing AGS cells com- gastric cancer cells. To our knowledge, this is the first report pared with parental AGS cells (data not shown). In contrast, a to demonstrate that GRIK2, a gene belonging to the kainate previous report noted that the interaction of GRIK2 and b-cat- receptor subgroup of the ionotropic glutamate receptor fam- enin in Cos-7 cells was partially blocked by overexpression of ily, is a novel tumor-suppressor gene in gastric cancer. the binding site-competing b-catenin-associated protein, PSD- 33 Studies regarding the association of GRIK2 with cancer are 95. Notably, we cannot rule out the possibility that unidenti- limited.29 We found that the GRIK2 gene promoter was highly fied nutrients or growth factors in the culture medium may methylated in the majority (85%) of gastric cancer tumor tis- have acted as agonists of GRIK2, thereby affecting cell migra- sues compared with normal adjacent tissues (as assessed by Q- tion activity in our assays. Previously, glutamate insert in cul- MSP using 40.8% methylation as a cutoff). For comparison, we ture medium was shown to activate GluR3 (an AMPA receptor also analyzed the methylation status of dlc1, apreviously subtype) expressed on T cells, thereby triggering CXCR4-medi- 35 known tumor suppressor gene31 that was among the genes ated T-cell chemotactic migration. Future studies will be war- identified in our DMH analysis. Among our clinical samples, ranted to assess the detailed mechanisms involved in GRIK2- 33% showed hypermethylation of the dlc1 promoter region mediated inhibition of cell migration. (data not shown); this was consistent with the results from an Recently, increasing numbers of reports have been dedicated earlier study.31 The GRIK2 gene is located on chromosome to the emerging role of neuronal receptors in cancer. Studies 6q16.3-q21, which is frequently deleted in patients with ALL.29 have shown that neuronal receptors, such as metabotropic glutamate receptors, not only mediate neurotransmission but also However, we were able to amplify the genomic sequence of participate in cellular transformation as oncogenes.36 In this this gene from all clinical samples (Fig. 2c), suggesting that this study, we found that the ionotropic GRIK2 may be a tumor gene was not commonly deleted in gastric cancer tissues. suppressor. The average mRNA expression level of GRIK2 was GRIK2 is the second ionotropic glutamate receptor family about 90% lower in gastric cancer tissues compared with adja- member to reportedly demonstrate a tumor-suppressor function cent normal tissues (Fig. 2d), whereas the latter level was less in gastric cancer; the first was the NMDA receptor subgroup.16,17 than one sixth of that found in brain (data not shown). In non- However, the tumor suppression mechanisms of GRIK2 and neoplastic gastric mucosa, the expression of GRIK2 was strong NMDA receptors seem to differ. NMDAR2A induces apoptosis in parietal cells and chief cells (Fig. 2a, left upper), whereas it and abolishes the colony formation ability of colorectal carci- was weak in mucous epithelium, with slightly increased expres- Cancer Cell Biology noma cell lines.18 However,NMDAR2Apromotedproliferation 32 sion in intestinal metaplasia (Fig. 2a,leftlower).GRIK2proteins of MKN45 gastric cancer cells by accelerating cell cycle. In are expressed in multiple organs; a higher level expression is contrast, we found similar cell proliferation levels among detected in the normal stomach than other organs (Supporting GRIK2-expressing AGS cells and vector controls (Fig. 3b). In Information Fig. 2a and Supporting Information Table I). In addition, flow cytometry showed that cell cycle progression was this study, we found that GRIK2 promoter was hypermethylated similar among the stably transfected clones (data not shown). in gastric cancer cell lines and tumor region of clinical samples, Thus, GRIK2-mediated tumor suppression seems to be mediated but not in 2 healthy stomach samples or adjacent normal of via mechanisms other than suppression of cell proliferation. clinical samples, suggesting GRIK2 promoter methylation may Notably, reintroduction of GRIK2 expression in gastric be an early tumorigenic event. Thus, it is possible that, in the cancer cells decreased tumor cell migration. Previous studies tumor microenvironment, GRIK2-mediated signaling is pre- have shown that GRIK2 can be recruited by the cadherin/cat- vented because of no or low expression of the receptor, even if enin complex through an interaction with b-catenin at Cos-7 the GRIK2 ligands may exist. Because the microenvironment of cell–cell junctions.33 b-Catenin signaling plays a dual role in a tumor may affect disease development, it would be interesting the process of cell migration: it links cadherins to the cytoskele- to investigate the function of GRIK2 in the gastrointestinal tract. ton, thereby allowing tight intercellular adhesion; and, upon This may provide an opportunity to study an unexpected mech- stimulation, it translocates into the nucleus to serve as a tran- anism for gastric cancer development. scription cofactor for target gene regulation.34 In gastric cancer cells, GRIK2 was found to colocalize with b-catenin in GRIK2- Acknowledgements expressing AGS cells, but its presence or absence did not seem We thank the Bioinformatics Core and Pathology Core of Chang Gung Mo- to affect the amount of b-catenin at AGS cell–cell junctions lecular Medicine Research Center for technical support.

References

1. Catalano V, Labianca R, Beretta GD, Gatta 2. Rocco A, Nardone G. Diet. H pylori 3. Yasui W, Oue N, Aung PP, Matsumura S, G, de Braud F, Van Cutsem E. Gastric infection and gastric cancer: evidence and Shutoh M, Nakayama H. Molecular- cancer. Crit Rev Oncol Hematol 2009;71: controversies. World J Gastroenterol 2007; pathological prognostic factors of gastric 127–64. 13:2901–12. cancer: a review. Gastric Cancer 2005;8:86–94.

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC 2552 Glutamate receptor, ionotropic, kainate 2 silencing

4. Yasui W, Oue N, Kuniyasu H, Ito R, 17. Liu JW, Kim MS, Nagpal J, Yamashita K, 26. Chan MW, Wei SH, Wen P, Wang Z, Tahara E, Yokozaki H. Molecular Poeta L, Chang X, Lee J, Park HL, Matei DE, Liu JC, Liyanarachchi S, Brown diagnosis of gastric cancer: present and Jeronimo C, Westra WH, Mori M, Moon R, Nephew KP, Yan PS, Huang TH. future. Gastric Cancer 2001;4:113–21. C, et al. Quantitative hypermethylation of Hypermethylation of 18S and 28S 5. Nephew KP, Huang TH. Epigenetic gene NMDAR2B in human gastric cancer. Int J ribosomal DNAs predicts progression-free silencing in cancer initiation and Cancer 2007;121:1994–2000. survival in patients with ovarian cancer. progression. Cancer Lett 2003;190:125– 18. Kim MS, Chang X, Nagpal JK, Yamashita Clin Cancer Res 2005;11:7376–83. 33. K, Baek JH, Dasgupta S, Wu G, Osada M, 27. Buffart TE, Overmeer RM, Steenbergen 6. Jones PA, Baylin SB. The fundamental role Woo JH, Westra WH, Trink B, Ratovitski RD, Tijssen M, van Grieken NC, Snijders of epigenetic events in cancer. Nat Rev EA, et al. The N-methyl-D-aspartate PJ, Grabsch HI, van de Velde CJ, Carvalho Genet 2002;3:415–28. receptor type 2A is frequently methylated B, Meijer GA. MAL promoter 7. Iida S, Akiyama Y, Nakajima T, Ichikawa in human colorectal carcinoma and hypermethylation as a novel prognostic W, Nihei Z, Sugihara K, Yuasa Y. suppresses cell growth. Oncogene 2008;27: marker in gastric cancer. Br J Cancer 2008; Alterations and hypermethylation of the 2045–54. 99:1802–7. p14(ARF) gene in gastric cancer. Int J 19. Rubinsztein DC, Leggo J, Chiano M, 28. Yan PS, Chen CM, Shi H, Rahmatpanah F, Cancer 2000;87:654–8. Dodge A, Norbury G, Rosser E, Craufurd Wei SH, Caldwell CW, Huang TH. 8. Oue N, Shigeishi H, Kuniyasu H, Yokozaki D. Genotypes at the GluR6 kainate Dissecting complex epigenetic alterations in H, Kuraoka K, Ito R, Yasui W. Promoter receptor locus are associated with variation breast cancer using CpG island hypermethylation of MGMT is associated in the age of onset of Huntington disease. microarrays. Cancer Res 2001;61:8375–80. with protein loss in gastric carcinoma. Int J Proc Nat Acad Sci USA 1997;94:3872–6. 29. Sinclair PB, Sorour A, Martineau M, Cancer 2001;93:805–9. 20. Motazacker MM, Rost BR, Hucho T, Harrison CJ, Mitchell WA, O’Neill E, 9. Tsuchiya T, Tamura G, Sato K, Endoh Y, Garshasbi M, Kahrizi K, Ullmann R, Foroni L. A ﬂuorescence in situ Sakata K, Jin Z, Motoyama T, Usuba O, Abedini SS, Nieh SE, Amini SH, Goswami hybridization map of 6q deletions in acute Kimura W, Nishizuka S, Wilson KT, James C, Tzschach A, Jensen LR, et al. A defect lymphocytic leukemia: identiﬁcation and SP, et al. Distinct methylation patterns of in the ionotropic glutamate receptor 6 gene analysis of a candidate tumor suppressor two APC gene promoters in normal and (GRIK2) is associated with autosomal gene. Cancer Res 2004;64:4089–98. cancerous gastric epithelia. Oncogene 2000; recessive mental retardation. Am J Hum 30. Choi IS, Wu TT. Epigenetic alterations in 19:3642–6. Genet 2007;81:792–8. gastric carcinogenesis. Cell Res 2005;15: 10. Johnson IT, Belshaw NJ. Environment, diet 21. Dutta S, Das S, Guhathakurta S, Sen B, 247–54. and CpG island methylation: epigenetic Sinha S, Chatterjee A, Ghosh S, Ahmed S, 31. Kim TY, Jong HS, Song SH, Dimtchev A, signals in gastrointestinal neoplasia. Food Ghosh S, Usha R. Glutamate receptor 6 Jeong SJ, Lee JW, Kim TY, Kim NK, Jung Chem Toxicol 2008;46:1346–59. gene (GluR6 or GRIK2) polymorphisms in M, Bang YJ. Transcriptional silencing of 11. Ozawa S, Kamiya H, Tsuzuki K. Glutamate the Indian population: a genetic association the DLC-1 tumor suppressor gene by

Cancer Cell Biology receptors in the mammalian central study on autism spectrum disorder. Cell epigenetic mechanism in gastric cancer nervous system. Prog Neurobiol 1998;54: Mol Neurobiol 2007;27:1035–47. cells. Oncogene 2003;22:3943–51. 581–618. 22. Shaltiel G, Maeng S, Malkesman O, 32. Watanabe K, Kanno T, Oshima T, Miwa 12. Paschen W, Blackstone CD, Huganir RL, Pearson B, Schloesser RJ, Tragon T, H, Tashiro C, Nishizaki T. The NMDA Ross CA. Human GluR6 kainate receptor Rogawski M, Gasior M, Luckenbaugh D, receptor NR2A subunit regulates (GRIK2): molecular cloning, expression, Chen G, Manji HK. Evidence for the proliferation of MKN45 human gastric polymorphism, and chromosomal involvement of the kainate receptor cancer cells. Biochem Biophys Res Commun assignment. Genomics 1994;20:435– subunit GluR6 (GRIK2) in mediating 2008;367:487–90. 40. behavioral displays related to behavioral 33. Coussen F, Normand E, Marchal C, Costet 13. Barbon A, Vallini I, Barlati S. Genomic symptoms of mania. Mol Psychiatry 2008; P, Choquet D, Lambert M, Mege RM, organization of the human GRIK2 gene 13:858–72. Mulle C. Recruitment of the kainate and evidence for multiple splicing variants. 23. Huang TH, Perry MR, Laux DE. receptor subunit glutamate receptor 6 by Gene 2001;274:187–97. Methylation profiling of CpG islands in cadherin/catenin complexes. J Neurosci 14. Li G, Oswald RE, Niu L. Channel-opening human breast cancer cells. Hum Mol Genet 2002;22:6426–36. kinetics of GluR6 kainate receptor. 1999;8:459–70. 34. Muller T, Bain G, Wang X, Papkoff J. Biochemistry 2003;42:12367–75. 24. Wei SH, Chen CM, Strathdee G, Regulation of epithelial cell migration and 15. Paternain AV, Herrera MT, Nieto MA, Harnsomburana J, Shyu CR, Rahmatpanah tumor formation by beta-catenin signaling. Lerma J. GluR5 and GluR6 kainate F, Shi H, Ng SW, Yan PS, Nephew KP, Exp Cell Res 2002;280:119–33. receptor subunits coexist in hippocampal Brown R, Huang TH. Methylation 35. Ganor Y, Besser M, Ben-Zakay N, Unger neurons and coassemble to form functional microarray analysis of late-stage ovarian T, Levite M. Human T cells express a receptors. J Neurosci 2000;20:196–205. carcinomas distinguishes progression-free functional ionotropic glutamate receptor 16. Kim MS, Yamashita K, Baek JH, Park HL, survival in patients and identifies candidate GluR3, and glutamate by itself triggers Carvalho AL, Osada M, Hoque MO, epigenetic markers. Clin Cancer Res 2002;8: integrin-mediated adhesion to laminin and Upadhyay S, Mori M, Moon C, Sidransky 2246–52. fibronectin and chemotactic migration. J D. N-methyl-D-aspartate receptor type 2b 25. Yan PS, Chen CM, Shi H, Rahmatpanah F, Immunol 2003;170:4362–72. is epigenetically inactivated and exhibits Wei SH, Huang TH. Applications of CpG 36. Shin SS, Martino JJ, Chen S. Metabotropic tumor-suppressive activity in human island microarrays for high-throughput glutamate receptors (mGlus) and cellular esophageal cancer. Cancer Res 2006;66: analysis of DNA methylation. J Nutr 2002; transformation. Neuropharmacology 2008; 3409–18. 132:2430S–2434S. 55:396–402.

Int. J. Cancer: 126, 2542–2552 (2010) VC 2009 UICC