A Promoter Methylation Pattern in the N-Methyl-d-Aspartate Receptor 2B Gene Predicts Poor Prognosis in Esophageal Squamous Cell Carcinoma

Esophageal squamous cell carcinoma (ESCC) is one of the most aggressive malignant tumors. Although surgical techniques and perioperative management have progressed, the prognosis for patients with ESCC remains poor (1, 2). Various factors are linked to the development of esophageal cancer, such as tobacco smoking and alcohol consumption, which are the major risk factors of the disease (3, 4). In addition to some of the genetic alterations associated with development or progression of ESCC reported (5), finding new molecular therapeutic targets for ESCC treatment is still a promising avenue of research that may help improve the survival of patients with this type of cancer.

Aberrant promoter hypermethylation found in neoplastic cells can be used as a molecular target for cancer detection of neoplasia. In ESCC, genes such as p16^INK4a (6), RASSF1 (7), trypsinogen-4 (8), HLA class (9), and MGMT (10) have been found to harbor hypermethylation of normally unmethylated CpG islands within the promoter regions.

N-Methyl-d-aspartate receptors (NMDAR) are the first class of glutamate receptors and the predominant excitatory neurotransmitter receptors in the mammalian brain (11). Some NMDAR subunits are expressed in skeletal muscle, heart, and pancreas (12, 13), as well as suprabasal keratinocytes (14). Interestingly, the activation of NMDARs inhibits keratinocyte outgrowth necessary for some epithelialization processes (14). The glutamate receptor ionotropic kinate 2 (GluR-6) was also identified as a candidate tumor suppressor gene in acute lymphocytic leukemia (15). In addition, aberrant CpG island methylation was reported in G protein–coupled metabotropic glutamate receptor 7 of chronic lymphocytic leukemia (16).

Previously, we identified NMDAR2B as a putative tumor suppressor and cancer-specific methylated gene in ESCC (17). In this study, we found that NMDAR2B promoter methylation in primary ESCC correlated strongly with poor prognosis. By multivariate analysis, NMDAR2B methylation in the promoter was shown to be an independent prognostic marker of ESCC survival.

ESCC, normal tissues, and cell lines. Twenty paired ESCC and normal esophagus (patient nos. 1-20) tissue gDNAs were obtained from the Gastroenterology Division, Department of Medicine, University of Maryland. Forty-seven cases of primary ESCC tumors (patient nos. 31-78) were obtained from patients who underwent surgery at the Medical Institute of Bioregulation Hospital, Kyushu University and the Saitama Cancer Center. All the patients had undergone a potentially curative resection of the primary carcinoma. Forty-three of 67 cases had clinicopathologic data, so they were applied for statistical analysis. These patients included 39 males and 4 females, and informed consent was obtained. The tumor was located in the upper esophagus (n = 2), the middle esophagus (n = 26), or the lower esophagus (n = 15). Eight tumors were well-differentiated squamous cell carcinomas, 24 were moderately differentiated, and 11 were poorly differentiated. Two were submucosal carcinomas and 41 were more advanced lesions with invasion to the muscularis propria (n = 7) or adventitia (n = 34). The presence (n = 37) or absence (n = 6) of lymph node metastasis was noted. Among the 43 cases, 9 patients survived for more than 5 years after surgery, whereas 21 cases died within 5 years after surgery. In the remaining 13 cases, patients are alive at last follow-up or died from causes unrelated to ESCC. Specimens were obtained from tumors, avoiding necrotic centers, immediately after resection. Corresponding normal mucosa specimens, which were at least 5 cm away from the tumor edge, were also obtained by sharply dissecting the mucosa off the muscularis propria. All specimens were quick-frozen in liquid nitrogen and stored at −80°C until processing.

Twelve ESCC cell lines, TE1, TE2, TE3, TE4, TE5, KYSE30, KYSE70, KYSE140, KYSE150, KYSE200, KYSE410, and KYSE520, were obtained from the Cell Response Center for Biomedical Research Institute, Department of Aging and Cancer, Tohoku University (TE series), and kindly provided by Dr. Shimada in the Department of Surgery, Kyoto University, Kyoto, Japan (KYSE series). Three CRC cell lines, HCT116, HT29, and DLD1, were purchased form American Type Culture Collection and were grown in McCoy, 5× supplemented with 10% fetal bovine serum.

Bisulfite treatment. For DNA denaturing, 1 μg of genomic DNA was incubated with 5 μg of salmon sperm DNA in 0.3 mol/L NaOH for 20 min at 50°C. The DNA sample was then treated with 500 μL of a 2.5 mol/L sodium metabisulfite/125 mmol/L hydroquinone/0.4 mol/L sodium hydroxide solution and placed at 70°C for 3 h. The sample was then applied to a column (Wizard DNA Clean Up System, Promega), incubated with 0.3 mol/L NaOH for 10 min, and treated with 3 mol/L ammonium acetate for 5 min. Then, 2.5× volume of 100% ethanol was added and DNA was allowed to precipitate for 1 h at room temperature. DNA was resuspended in 50-μL water and stored at −80°C.

Sequencing and combined bisulfite restriction analysis. All PCR reactions were done as described previously (17), and the primer sequence of bisulfite-DNA amplification for NMDAR2B is shown in Supplementary Table S1. For amplification of the F3 region of NMDAR2B, nested PCR was done with F4-R2 primers, and 3 μL of PCR solution were used as template after first PCR was done using F3-R2 primers. All the PCR products were gel extracted (Qiagen) and sequenced with internal primers (F2 or F4) using the ABI BigDye cycle sequencing kit (Applied Biosystems). For combined bisulfite restriction analysis (COBRA), eluted DNA after gel extraction was digested with BstU1 that recognizes the CGCG sequence (New England Biolabs, Inc.) for 3 h at 60°C. Samples were loaded on a 10% acrylamide gel, stained with 1× SYBR Green Gold (Molecular Probes), and visualized under UV light.

Real-time quantitative PCR (TaqMan methylation-specific PCR). For quantitative methylation analysis, PCR primers were designed to hybridize three separate sites (P1, P2, and P3) of MDAR2B promoter that were previously determined to be methylated in ESCC cell lines by sequencing and fluorescent probes to the amplified region of the DNA. All oligonucleotide primer pairs were purchased from Invitrogen, and the TaqMan probe from VWR. The NMDAR2B primer and probe sequences are shown in Supplementary Table S1. TaqMan methylation-specific PCR (MSP) was done as reported (17). To ensure the specificity of the TaqMan MSP analysis, each 384-well PCR plate had wells that contained bisulfite-converted DNA isolated from patient tissue samples and wells that contained the following controls: in vitro methylated leukocyte DNA (positive control), DNA from normal esophageal mucosa in which NMDAR2B is not methylated (negative control), and multiple water blanks (control for PCR specificity). Leukocyte DNA from a healthy individual was methylated in vitro with excess SssI methyltransferase (New England Biolabs) to generate completely methylated DNA, and serial dilutions (90-0.009 ng) of this DNA were used to construct a calibration curve for each plate. All samples were within the assay range of sensitivity and reproducibility based on amplification of internal reference standard [threshold cycle (C_t) value for β-actin, ≤40]. The methylation ratio was defined as the quantity of fluorescence intensity derived from the NMDAR2B promoter amplification divided by fluorescence intensity from β-actin amplification, and multiplied by 100 [TaqMan methylation value (TaqMeth V)]. This ratio was used as a measure for the relative level of methylated NMDAR2B DNA in samples. The samples were categorized as unmethylated or methylated based on the sensitivity of the assay.

Statistical analysis. We used the methylation levels (TaqMeth V) for P1, P2, and P3 sites of NMDAR2B to construct receiver-operator characteristic (ROC) curves for the detection of ESCC. Using this approach, the area under ROC identified optimal sensitivity and specificity levels (i.e., cutoff values) at which to distinguish normal from malignant esophageal tissues, and corresponding TaqMeth V thresholds were calculated for the three sites. The cutoff values determined from ROC curve were applied to determine the frequency of NMDAR2B methylation. Samples with TaqMeth V ≥0.85 were designated as methylated and given a value of 1 for statistical analysis, whereas samples containing TaqMeth V <0.85 were designated as unmethylated and given a value of 0. Factors associated with the methylation status of three sites, P1, P2, and P3, were selected based on cross tabulations using χ² tests. Time to event (overall or disease-specific death) was analyzed with the method of Kaplan and Meier (18), the log-rank statistics (19), and the proportional hazards regression model (20). The simultaneous effect of two or more factors was studied using multivariate Cox proportional hazards models. Factors tested for prognostic variables included the methylation status of three sites (P1, P2 and P3), depth of invasion, differentiation, location, lymph node metastasis, lymphatic retention, and vascular permeation. All statistical tests were two sided. The significance level used was 0.05 and all statistical analyses were conducted using STATA version 9 (STATA, Inc.). First, we computed bivariate associations between each putative confounder variable and survival and between each putative confounder and NMDAR2B methylation status. The only variable that was independently associated with survival, other than NMDAR2B methylation status, was tumor location. Consequently, we constructed a Cox proportional hazards model with NMDAR2B methylation status and tumor location as the covariates. In our search for independent clinicopathologic factors, we did not include tumor-node-metastasis (TNM) staging in our multivariate analysis because TNM staging itself was already determined by the depth of tumor invasion, lymph node metastasis, and the presence of distant metastasis.

Methylation at three different sites in the NMDAR2B promoter. In the promoter, NMDAR2B harbors more than one CpG island within 1 kb upstream of the transcription start site (Fig. 1A). The first CpG island that resides directly upstream of transcription start site is less dense with CpG sites (F1 region) than those at more upstream regions (F3 region). We reported in the previous study that the F1 region of NMDAR2B was methylated in primary ESCC with a high frequency (>90%) and in normal-appearing esophageal tissues with a very low frequency (<5%; ref. 17). By bisulfite sequencing analysis of the F1 region, we found that three CpGs in the P1 site outside the boundaries of the first CpG island were less frequently methylated in ESCC compared with the other CpGs in the F1 region (30%, 6 of 20; Fig. 2B). In 12 ESCC cell lines, the P1 site was completely methylated in eight cell lines, whereas in the other four cell lines (TE3, KYSE70, KYSE410, and KYSE520), methylated and unmethylated alleles coexisted (partial methylation; Fig. 2A). All normal esophageal mucosa samples completely lacked cytosine methylation at the P1 site (0%, 0 of 20). For more detailed examination of the NMDAR2B promoter, we analyzed the methylation status of the F3 region by bisulfite sequencing and COBRA. The methylation frequency of the F3 region in ESCC was very high compared with that in normal corresponding tissues (5%, 1 of 20), which was similar to that of the F1 region (Fig. 2B; ref. 17). In all 12 ESCC cell lines, the F3 region was also completely methylated (Fig. 2A). Representative sequencing results are shown in Fig. 1B.

TaqMan methylation-specific PCR (MSP) was then done with probes targeted to the P1 and P3 sites of NMDAR2B in primary tissues (as designated in Fig. 1A). Methylation of all three sites of NMDAR2B showed highly discriminative ROC curve profiles, clearly distinguishing ESCC from corresponding normal mucosa (P < 0.005; Fig. 3C). The cutoff for each site (0.85) was chosen from the ROC curve to maximize sensitivity and specificity. At this cutoff, the specificity was highest at the P1 site where >90% of normal-appearing tissues displayed methylation under the cutoff (92.9%, 13 of 14; Fig. 3A). In the P3 site, 78.6% (11 of 14) of normal-appearing tissues had values under the cutoff (Fig. 3A). On the other hand, 67.3% (35 of 52) and 77.6% (52 of 67) of ESCC tissues in the P1 and P3 site, respectively, displayed higher level of methylation than the cutoff, showing an excellent sensitivity. All the ESCC cell lines (12 of 12) showed methylation of the two sites by quantitative analysis. In addition, the TaqMeth V of primary ESCC in all three sites were substantially higher than those of normal tissues, and all differences were significant (P < 0.05). The mean and median values of TaqMeth V of the three sites are shown in Table 1. The cutoff values can be adjusted to put more weight toward higher specificity. Even at the highest specificity (100%), the sensitivity of the three promoter sites was still relatively high (53.8 for P1, 31.1% for P2, and 46.3% for P3). When we considered at least one of the three sites positive at the cutoff value of 0.85, the overall diagnostic sensitivity for ESCC was 66.7% (44 of 67 cases).

In 47 of 67 ESCC, all three sites were successfully analyzed by the quantitiative methylation analysis (TaqMan MSP). Interestingly, in 19 of 47 (40.4%) cases of ESCC, the NMDAR2B promoter exhibited mosaic methylation (Fig. 3B). All ESCC tissues except three cases harbored methylation in at least two sites. One case harbored methylation only in P1 (patient no. 48) and the other 2 (4.25%) cases showed no methylation in all three sites (patient nos. 33 and 49). Twenty-six (55.3%) cases showed methylation at all three promoter sites. The mosaic methylation pattern was also seen in some of the normal tissues.

Clinicopathologic correlation with promoter methylation of NMDAR2B. There was a strong correlation between clinicopathologic variables and methylation of NMDAR2B in ESCC patients. Cases with higher NMDAR2B TaqMeth V than the cutoff value of 0.85 in the P1 site showed poorer prognosis than those with lower NMDAR2B TaqMeth V. As shown in Table 2, factors associated with an increased probability of methylation of the P1 site included vascular permeation and tumor location in the lower portion of the esophagus. The higher NMDAR2B methylation group in the P1 region included 3 of 11 (27%) patients with no vascular permeation and 22 of the 32 (68%) with vascular permeation. This difference for vascular permeation was statistically significant (P = 0.03). The difference in the methylation status of NMDAR2B in the P1 site for the tumor location of the ESCC was marginally significant (P = 0.052). No other clinicopathologic factors showed a statistically significant correlation with NMDAR2B methylation status in the P1 site. In addition, no significant correlation was found between the methylation of P2 and P3 sites and any clinicopathologic variables.

The median disease-specific survival time for the 43 patients in this study was 25 months. Three patients died of other causes in this study and were censored at the time of their death. The median follow-up period of the 22 surviving patients was 48.5 months. Factors significantly decreasing survival time included methylation of any one of these three sites, P1, P2, or P3. Kaplan-Meier overall and disease-specific survival curves for all the patients with ESCC were constructed to analyze survival discrepancies between patients with NMDAR2B methylation levels above or below the cutoff value of 0.85. As shown in Fig. 4, high methylation of NMDAR2B in the P1 site yielded a poorer prognosis compared with the group with low methylation in both overall (P = 0.05) and disease-specific (P = 0.006) survival analyses. Results of multivariate Cox proportional hazards modeling indicated that the group with higher methylation at the P1 site had 11.84 times greater the hazard than the group without higher methylation (P = 0.018; Table 3). When the death was adjusted for disease-specific, the hazard ratio was 3.13 for higher methylation in the P1 site (P= 0.049). Patients with methylation at the P1 site showed 2.5 times greater the disease-specific death hazard than patients without methylation at the site when accounting for the effect of location (univariate Cox proportional hazards modeling). Higher methylation at the P2 site showed a trend toward poorer prognosis in disease-specific survival with marginal significance (P = 0.06). In contrast, the methylation status in the P3 site was not significantly correlated with overall and disease-specific survival (P = 0.38 and P = 0.25, respectively).

At the time of diagnosis, ESCC generally exhibit features of a more advanced stage and metastasizes to the lymph nodes (21). As a consequence, the prognosis of ESCC patients is usually very poor. Recently, a combination of several methylation markers was recently reported to be significantly correlated with patient prognosis in esophageal adenocarcinoma (22). We have intensively investigated methylation of gene promoters in primary tumors and in body fluids as tumor markers in cancer development (23–27) and also reported that PGP9.5 was an independent prognostic marker for ESCC (28).

We previously identified NMDAR2B promoter to be hypermethylated in primary ESCC (17). Here, quantitative TaqMan MSP revealed methylation of the P2 site in 90% of ESCC. A high theoretical sensitivity (67%) was achieved when maximizing specificity (100%) for ESCC when we combined results for all three sites at the best cutoff values. P1 was the best among all the three sites from a diagnostic point of view because of the highest mean value of TaqMeth V and overall sensitivity. In addition, multivariate regression analyses showed that P1 was the strongest predictor of survival. P2 methylation was also marginally associated with the disease-specific survival of patients. No correlation was found between the P3 site methylation and clinicopathologic variables, suggesting that methylation of NMDAR2B at a proximal region from transcription start site might have more potential to predict malignancy in ESCC.

Methylation of the F3 region might be more relevant than F1 region for gene silencing. As shown in Supplementary Fig. S2A, CRC cell lines harbored complete methylation in the F1 region, but expression of NMDAR2B mRNA was clearly detected in HCT116 (17), HT29, and DLD1 (data not shown), suggesting that the methylation in the F1 might not be enough to silence NMDAR2B expression. Interestingly, the F3 region was not methylated in all CRC cell lines tested in COBRA (Supplementary Fig. S2B and C). Promoter activity of the luciferase constructs containing unmethylated (−SssI) F1 or F3 portions of the NMDAR2B promoter sequence showed that the activity of F3 construct was about twice that of F1 construct in HCT116 cells (Supplementary Fig. S2D). All these results support that methylation of regions other than F1 in the NMDAR2B promoter might be more involved in NMDAR2B gene silencing. Further studies will be needed to verify mechanisms underlying relationships between methylation of specific promoter regions and gene silencing.

In conclusion, this study represents the first clinical profile of NMDAR2B promoter methylation in primary ESCC. We conclude that a clone harboring NMDAR2B promoter methylation may have malignant potential in human cancers. Methylation status of NMDAR2B in primary tumors may serve as a potential clinical biomarker for ESCC diagnosis and prognosis.