Aberrant Methylation of RASGRF1 Is Associated with an Epigenetic Field Defect and Increased Risk of Gastric Cancer

Gastric cancer is a major cause of cancer-related mortality, worldwide. Helicobacter pylori (H. pylori) plays an important role in gastric carcinogenesis, although the majority of the individuals with H. pylori infection do not develop gastric cancer (1). Histologically, gastric cancers are divided into 2 subgroups, intestinal and diffuse, which are thought to develop through separate pathologic pathways (2). Etiologic analysis has shown that individuals with H. pylori–related gastritis, severe atrophy, and intestinal metaplasia are at high risk of developing intestinal type gastric cancers, which are often associated with metachronous gastric cancer development. On the other hand, individuals with H. pylori–related pangastritis and enlarged-fold gastritis, which are lesions without mucosal atrophy or intestinal metaplasia, are at increased risk of developing diffuse type gastric cancers (3). Surveillance of these high-risk patients using reliable and accurate predictive markers is important for reducing the incidence of gastric cancer and its mortality.

Aberrant DNA methylation is one of the most common molecular alterations found in neoplasias; CpG island hypermethylation is associated with the silencing of tumor suppressor genes and other tumor-related genes, whereas global hypomethylation is thought to induce oncogene activation or chromosomal instability (4). The list of genes aberrantly methylated in gastric cancer is growing and now includes genes involved in cell-cycle regulation, apoptosis, immune function, cell signaling, and tumor invasion and metastasis (3, 5). In addition, aberrant DNA methylation is frequently observed in noncancerous gastric mucosa in H. pylori–infected patients, suggesting aberrant DNA methylation is an early step during gastric carcinogenesis (6, 7). We previously showed hypomethylation of LINE-1 repetitive elements and hypermethylation of CDH1 in enlarged-fold gastritis, which is an indicator of a high risk for diffuse type gastric cancer (8). More recently, we and others reported frequent hypermethylation of tumor suppressive miRNA genes in the noncancerous gastric mucosa of patients with gastric cancer (9, 10). Taken together, these results suggest that aberrant methylation in the gastric mucosa could be a useful biomarker for evaluating gastric cancer risk.

Our aim in the present study was to identify predictive biomarkers by screening for DNA methylation in the noncancerous background gastric mucosa in cases of gastric cancer. For this purpose, we carried out high-throughput CpG island methylation profiling in a set of noncancerous gastric mucosa specimens from patients with gastric cancer and from cancer-free individuals. We identified a number of aberrantly methylated genes in the gastric mucosa of the patients with gastric cancer, including RASGRF1, which was frequently methylated in cases of both intestinal and diffuse type gastric cancer. We provide evidence that RASGRF1 is a novel target for epigenetic silencing in gastric cancer, and that its methylation in the gastric mucosa is strongly associated with an elevated risk for both types of gastric cancer.

A total of 130 primary gastric cancer specimens were obtained through surgical resection or endoscopic biopsy. Samples of noncancerous gastric mucosa were obtained through endoscopic biopsy from 91 patients with gastric cancer and 69 healthy individuals. From each patient, biopsy specimens of noncancerous gastric mucosa were taken from the gastric body and antrum. H. pylori infection was assessed using a rapid urease test, a serum antibody test, or a urea breath test. If any one of these assays was positive, the patients were considered to be H. pylori–positive. The updated Sydney system and serum pepsinogen test was used to estimate the degree of gastritis (11). The serum pepsinogen test was carried out by assessing the serum pepsinogen I (PGI) and pepsinogen II (PGII) levels; the criteria for positivity were PGI ≤ 70 ng/mL and a PGI/PGII ratio ≤3.0 (12). Informed consent was obtained from all patients before the collection of specimens. Approval of this study was obtained from the Institutional Review Board of Akita Red Cross Hospital (Akita, Japan) and Sapporo Medical University (Sapporo, Japan).

Gastric cancer cell lines (MKN7, SH101, SNU1, SNU638, JRST, KatoIII, AZ521, AGS, and NCI-N87) were obtained and cultured as described previously (9, 13). SH101 and HSC43 cells were kindly provided by Dr. Kazuyoshi Yanagihara, Yasuda Women's University (14, 15). In some instances, cells were treated with 2 μmol/L 5-aza-2′-deoxycytidine (5-aza-dC; SIGMA) for 72 hours, replacing the drug and medium every 24 hours. Genomic DNA was extracted using the standard phenol–chloroform procedure. Total RNA was extracted using TRIZOL reagent (Invitrogen), and then treated with a DNA-free Kit (Ambion).

Methylated CpG island amplification (MCA) was conducted as described previously (16, 17). Briefly, 500 ng of genomic DNA was digested with the methylation-sensitive restriction endonuclease SmaI (New England Biolabs), after which it was digested with the methylation-insensitive restriction endonuclease XmaI. The adaptors were prepared by addition of the oligonucleotides RMCA12 (5′-CCGGGCAGAAAG-3′) and RMCA24 (5′-CCACCGCCATCCGAGCCTTTCTGC-3′). After the ligation of the digested DNA to the adaptors, PCR amplification was carried out. Using a BioPrime Plus Array CGH Genomic Labeling System (Invitrogen), MCA amplicons from gastric cancers and samples of H. pylori–positive noncancerous gastric mucosa were labeled with Alexa Fluor 647, and those from pooled samples of a mixture of H. pylori–negative normal gastric mucosa were labeled with Alexa Fluor 555. Labeled MCA amplicons were then hybridized to a custom human CpG island microarray containing 15,134 probes covering 6,157 unique genes (G4497A; Agilent Technologies; ref. 18). After washing, the array was scanned using an Agilent DNA Microarray Scanner (Agilent Technologies), and the data were processed using Feature Extraction software ver. 10.7 (Agilent Technologies). The data were then analyzed using GeneSpring GX ver. 11 (Agilent Technologies) after which unsupervised hierarchical clustering analysis were carried out using JMP ver. 8 (SAS Institute). The microarray data in this study have been submitted to the Gene Expression Omnibus (GEO) and accession number is GSE39175.

Genomic DNA (1 μg) was modified with sodium bisulfite using an EpiTect Bisulfite Kit (QIAGEN) after which methylation-specific PCR (MSP), bisulfite sequencing, and pyrosequencing were carried out as described previously (9). Briefly, for bisulfite pyrosequencing, the biotinylated PCR product was purified, made single-stranded, and used as a template in a pyrosequencing reaction run according to the manufacturer's instructions. The pyrosequencing reaction was carried out using a PSQ96 system with a PyroGold Reagent Kit (QIAGEN), and the results were analyzed using Q-CpG software (QIAGEN). For bisulfite sequencing, amplified PCR products were cloned into pCR2.1-TOPO vector (Invitrogen), and 12 to 14 clones from each sample were sequenced using an ABI3130x automated sequencer (Applied Biosystems). Primer sequences and PCR product sizes are listed in Supplementary Table S1.

Single-stranded cDNA was prepared using SuperScript III reverse transcriptase (Invitrogen), after which the integrity of the cDNA was confirmed by amplifying glyceraldehyde-3-phosphate dehydrogenase (GAPDH). Primer sequences and PCR product sizes are shown in Supplementary Table S1. Quantitative reverse transcription-PCR (RT-PCR) was carried out using TaqMan Gene Expression Assays (RASGRF1, Hs00182314_m1; GAPDH, Hs99999905_m1; Applied Biosystems) and a 7500 Fast Real-Time PCR System (Applied Biosystems). SDS ver. 1.4 software (Applied Biosystems) was used for comparative ΔC_t analysis.

A full-length RASGRF1 cDNA was amplified by PCR using cDNA derived from RASGRF1-expressing AZ521 cells as a template and then cloned into pcDNA3.2/V5/GW/D-TOPO (Invitrogen). The sequence was then verified. Primer sequences and PCR product sizes are shown in Supplementary Table S1.

Western blot analysis was carried out as described previously (9). Mouse anti-V5 monoclonal antibody (mAb; Invitrogen), rabbit anti-RASGRF1 polyclonal Ab (sc-863, Santa Cruz Biotechnology), and mouse antiactin mAb (Chemicon) were used as instructed by the manufacturers. The immunoreactive bands were visualized using enhanced chemiluminescence (Amersham Biosciences).

Colony formation assays were carried out as described previously (13). Briefly, cells (1 × 10⁵ cells) were transfected with 4 μg of RASGRF1 expression vector or empty pcDNA3.1 using Lipofectamine 2000 (Invitrogen) according to the manufacturer's instructions. Cells were then plated on 60-mm culture dishes and selected for 10 days in 0.4 mg/mL G418. Colonies were stained with Giemsa and counted using the NIH IMAGE software.

Cell invasion was assessed using Matrigel invasion assays as described previously (19). Gastric cancer cells (2 × 10⁶ cells) were transfected with 5 μg of RASGRF1 expression vector or a negative control using a Cell Line Nucleofector kit V (Lonza) with a Nucleofector I electroporation device (Lonza) according to the manufacturer's instructions. After incubation for 24 hours, 1 × 10⁵ transfectants suspended in 500 μL of serum-free RPMI-1640 medium were added to the tops of BD BioCoat Matrigel Invasion Chambers (BD Biosciences) prehydrated with PBS, and 700 μL of RPMI-1640 media supplemented with 10% FBS were added to the lower wells of the chambers. After incubation for 22 hours, the invading cells were stained with 1% toluidine and then counted in 3 randomly selected microscopic fields per membrane.

To compare differences in continuous variables between groups, t tests or ANOVA with posthoc Tukey tests were conducted. Fisher exact test or χ² test was used for analysis of categorical data. Receiver operator characteristic (ROC) curves were constructed on the basis of the levels of methylation. Values of P < 0.05 (2-sided) were considered statistically significant. Statistical analyses were carried out using SPSS statistics 18 (IBM Corporation) and GraphPad Prism ver. 5.0.2 (GraphPad Software).

To screen for methylation changes early during gastric carcinogenesis, we compared the methylation status of noncancerous background gastric mucosa from patients with gastric cancer with that in samples of gastric mucosa from healthy individuals. To accomplish this, we carried out methylated-CpG island amplification coupled with CpG island microarray (MCAM) analysis using a set of H. pylori–positive gastric mucosa specimens from otherwise healthy individuals (gastric antrum, n = 11; gastric body, n = 15), noncancerous gastric mucosa from patients with gastric cancer (gastric antrum, n = 10; gastric body, n = 10), gastric cancer tissues (n = 10), and gastric cancer cell lines (AGS, HSC43, KatoIII, MKN74, and NUGC4). The gastric mucosa specimens from the healthy individuals were divided into 2 histologic groups: antrum-predominant gastritis (n = 14), which has a low risk for gastric cancer and pangastritis or corpus-predominant gastritis (n = 12), which has a high risk for gastric cancer. We thus aimed to identify genes that showed greater methylation in noncancerous antral gastric mucosa from patients with gastric cancer than in mucosa from healthy individuals with antrum-predominant gastritis. MCAM analysis revealed 889 unique genes that were hypermethylated in the background gastric mucosa in intestinal type gastric cancer, as compared with antrum-predominant gastritis (Fig. 1A, Supplementary Table S2). In addition, we identified 478 unique genes that were methylated in the background mucosa in diffuse type gastric cancer (Fig. 1A, Supplementary Table S3). Subsequent Venn diagram analysis identified 224 genes that were methylated in the background gastric mucosae of patients with either type of gastric cancer but not in healthy individuals (Fig. 1A, Supplementary Table S4).

Unsupervised hierarchical clustering analysis using MCAM data for the 224 selected genes revealed that the genes could be categorized into 3 subclasses (Fig. 1B, Supplementary Table S4): group 1 genes (81 unique genes) were methylated in the majority of antral gastric mucosae from patients with gastric cancer and in gastric cancer tissues; group 2 genes (35 unique genes) were prevalently methylated in patients with gastric cancer and in otherwise healthy individuals with pangastritis or corpus-predominant gastritis, suggesting the methylation was inflammation-related and less cancer-specific; and group 3 genes (108 unique genes), which gave an elevated signal in the antrum of patients with gastric cancer and patients with pangastritis or corpus-predominant gastritis, but overall, the levels of methylation were relatively low in all of the specimens tested. These results suggest that genes predictive of gastric cancer risk are likely enriched in group 1. Among the group 1 genes, we selected 11 (RASGRF1, SOX5, GALNT14, RGS20, RPIB9, SYT5, WNT3, BASP1, ITGA4, KCNV1, and PAX5) that gave the highest microarray signals in the gastric cancer tissues and background gastric mucosa. Using MSP, we tested their methylation status in a small set of clinical specimens and found that 3 genes (RASGRF1, GALNT14, and SOX5) strongly discriminated between healthy individuals and patients with gastric cancer (Fig. 1B, Supplementary Fig. S1). We therefore used quantitative bisulfite pyrosequencing to assess their methylation levels in a set of antral mucosae from H. pylori–positive healthy individuals (n = 22) and noncancerous gastric mucosae from the antrum of patients with intestinal (n = 16) or diffuse (n = 13) type gastric cancer (Fig. 1C, Supplementary Fig. S2). Consistent with the MCAM and MSP data, the levels of methylation of these 3 genes were elevated in the background gastric mucosae from patients with gastric cancer, although methylation of GALNT14 in diffuse type patients with gastric cancer was less pronounced. In contrast, methylation of RASGRF1 was significantly elevated in patients with either type of gastric cancer, suggesting it could be a useful biomarker for predicting gastric cancer risk.

On the basis of the results summarized earlier, we selected RASGRF1 for further analysis. The promoter region and exon 1 of RASGRF1 are embedded within a typical CpG island (Supplementary Fig. S3). Bisulfite pyrosequencing revealed that RASGRF1 was highly methylated in a majority of the gastric cancer cell lines tested (Fig. 2A), and that expression of RASGRF1 mRNA was absent in those cells. Treatment with a DNA methyltransferase inhibitor, 5-aza-dC, restored RASGRF1 mRNA expression in multiple cell lines, suggesting RASGRF1 is epigenetically silenced in gastric cancer cells (Fig. 2A). In contrast, methylation levels were low in MKN74 and AZ521 cells, in which RASGRF1 mRNA was abundantly expressed (Fig. 2A and data not shown). Elevated levels (>15%) of RASGRF1 methylation were also frequently detected in both types of primary gastric cancer (intestinal type, 31 of 62, 50.0%; diffuse type, 41 of 68, 60.3%; Fig. 2B, Supplementary Table S5). We also found that RASGRF1 methylation correlates inversely with higher pathologic T (pT) categories and distant metastasis (Supplementary Table S5). When the methylation status of RASGRF1 was further assessed in selected specimens, we observed that its CpG island is densely methylated in gastric cancer cell lines, primary tumors, and background noncancerous gastric mucosa, whereas normal gastric mucosa from a healthy individual showed only limited methylation (representative results in Fig. 2C). To test whether methylation of RASGRF1 is associated with its downregulation in primary gastric cancer, we carried out bisulfite pyrosequencing and quantitative RT-PCR with a set of tissue specimens, which revealed an inverse relationship between methylation and expression (Fig. 2D).

The elevated levels of RASGRF1 methylation in the background gastric mucosa of patients with gastric cancer suggest its involvement in an epigenetic field defect. We therefore next assessed its clinical usefulness as a predictive biomarker of gastric cancer risk by comparing the levels of RASGRF1 methylation in a set of gastric mucosa specimens from healthy individuals (H. pylori–negative, n = 12; H. pylori–positive, n = 50) and noncancerous gastric mucosa specimens from patients with gastric cancer (intestinal type, n = 55; diffuse type, n = 36). All specimens were collected from the gastric antrum, and the clinicopathologic features of the samples are shown in Table 1. In healthy individuals, the levels of RASGRF1 methylation did not significantly differ between H. pylori–negative and -positive gastric mucosae (5.4% and 9.0%, average 8.3%), suggesting that severe RASGRF1 methylation is not induced by H. pylori infection alone (Fig. 3A). In contrast, methylation levels were significantly elevated in noncancerous mucosae from intestinal and diffuse type patients with gastric cancer (22.5% and 19.4%, average 21.3%), indicating that RASGRF1 methylation in noncancerous gastric mucosae may be associated with gastric cancer risk (Fig. 3A).

We also generated a ROC curve to assess the clinical use of RASGRF1 methylation for prediction of gastric cancer. RASGRF1 methylation was highly discriminative between noncancerous gastric mucosa from patients with gastric cancer and gastric mucosa from healthy individuals (Fig. 3B and Table 2). Earlier studies showed that severe gastric mucosal atrophy induced by H. pylori infection is a hallmark of gastric cancer risk, and that its evaluation based on serum pepsinogen tests is a useful means of assessing cancer risk (12, 20). To evaluate the relationship between RASGRF1 methylation and mucosal atrophy, we divided the samples based on the pepsinogen test results and then generated respective ROC curves. Notably, RASGRF1 methylation was highly discriminative between healthy individuals and patients with gastric cancer irrespective of the pepsinogen test results (Fig. 3C and Table 2), which suggests that the strong association between RASGRF1 methylation and gastric cancer is independent of gastric mucosal atrophy. This could make RASGRF1 methylation a powerful biomarker with which to identify individuals at high risk for gastric cancer.

Finally, we tested whether RASGRF1 serves as a tumor suppressor in gastric cancer. Gastric cancer cell lines were transfected with a RASGRF1 expression vector or a negative control, after which colony formation assays were carried out with the transfectants. Western blot analysis confirmed that the transfectants expressed exogenous RASGRF1 (Supplementary Fig. S4). Moreover, introduction of RASGRF1 markedly suppressed colony formation by the cell lines tested (Fig. 4A and B). When we then carried out Matrigel invasion assays to test the effect of RASGRF1 expression on gastric cancer cell invasion; we observed marked inhibition of cell invasion by SNU638 cells expressing RASGRF1 (Fig. 4C). These results suggest that RASGRF1 may play a tumor suppressor role that is itself suppressed in gastric cancer.

Identification of individuals at high risk of developing gastric cancer is essential for the prevention and early detection of gastric cancer. Esophagogastroduodenoscopy (EGD) is the most useful method for detecting gastric cancers, although population-based screening for gastric cancers using only EGD is generally considered ineffective for reducing mortality (21, 22). Severe atrophic gastritis is strongly associated with an increased risk of intestinal gastric cancer, and screening for high-risk individuals based on serum pepsinogen levels followed by careful observation using EGD is an effective strategy for surveillance of this disease (12, 23). In contrast, diffuse type gastric cancers lack those characteristic features because they do not progress through atrophic gastritis. Consequently, identification of sensitive and reliable biomarkers for diffuse type gastric cancer would be highly desirable. A number of studies have shown aberrant DNA methylation in precancerous lesions, including chronic gastritis and intestinal metaplasia, and detection of such an epigenetic field defect would provide useful information for identifying individuals at high risk for developing gastric cancer (7, 24, 25). The majority of those studies focused on well-studied genes, such as CDH1 and p16 (26, 27), but recent advances in microarray technology have enabled us to conduct a genome-wide analysis of CpG island methylation status. For instance, a recent study reported by Nanjo and colleagues identified a series of 7 methylation markers that can predict gastric cancer risk in individuals with past H. pylori infection (28).

In this study, we carried out high-throughput methylation analysis using a set of gastric mucosa specimens from healthy individuals and patients with intestinal or diffuse type gastric cancer. Our MCAM analysis identified a number of methylated genes in noncancerous gastric mucosae from the patients with gastric cancer. A larger number of methylated genes were identified in gastric mucosae from patients with intestinal type gastric cancer than with diffuse type gastric cancer, which is consistent with the earlier observation that aberrant DNA methylation is not induced by the presence of H. pylori, itself, but by inflammatory processes triggered by the infection (29). Our list of genes methylated in the background mucosa in intestinal type gastric cancer includes a number of methylation-prone genes (e.g., SFRP2 and IRF4) that confirms the reliability of our screening method (13, 30). The methylation status in the background mucosa of diffuse type gastric cancer had remained largely unknown and, to our knowledge, this study is the first to examine the genome-wide CpG island methylation status in the gastric mucosa from diffuse type patients with gastric cancer. It is noteworthy that we found that approximately half of the genes methylated in diffuse type gastric cancer were also methylated in intestinal type gastric cancer. It is generally believed that intestinal and diffuse type gastric cancers develop through distinctly different molecular pathways; however, our data may be indicative of a pathogenic mechanism common to both types. Furthermore, our results suggest that methylation of these genes could be an ideal molecular marker for assessing the risk for both gastric cancer types.

Among the genes identified, we selected RASGRF1, GALNT14, and SOX5 for further analysis and found that the elevation of their methylation levels was specific to patients with gastric cancer. SOX5 is a member of the high-mobility group superfamily and is reportedly overexpressed in several malignancies, including nasopharyngeal carcinoma and prostate cancer, which suggests it has oncogenic properties (31, 32). On the other hand, one recent study showed that SOX5 suppresses platelet-derived growth factor B–induced gliomas (33). GALNT14 belongs to a large subfamily of glycosyltransferases, and its expression in cancer cells is associated with cellular sensitivity to the proapoptotic ligand Apo2L/TRAIL (34). Up to now, however, methylation of GALNT14 and SOX5 has not been reported in human cancer, and further study will be needed to clarify their functional significance.

RASGRF1 and RASGRF2 constitute a gene family encoding guanine nucleotide exchange factors (GEF), which activate Ras GTPase by promoting the release of bound GDP, enabling activating GTP to take its place (35). RASGRF proteins are predominantly expressed in adult neurons in the central nervous system, and are involved in a wide range of neuronal functions. In mice, Rasgrf1 is an imprinted gene. The imprinted Rasgrf1 locus is methylated on the paternal allele at a differentially methylated region (DMR) located 30 kb upstream of the promoter, and it is expressed only from the paternal allele (36). Interestingly, a recent study showed that Piwi-interacting RNAs (piRNAs), a subset of noncoding small RNAs, play a pivotal role in the establishment of methylation at the Rasgrf1 DMR (37). In contrast, we found the promoter CpG island of RASGRF1 to be hypermethylated in gastric cancer, and that this methylation is unlikely to be associated with gene imprinting. Levels of RASGRF1 methylation are also significantly elevated in the noncancerous background gastric mucosa in both the intestinal and diffuse types of gastric cancer and are highly discriminative between gastric mucosa from cancer-free individuals and patients with gastric cancer. This suggests RASGRF1 methylation may be a gastric cancer risk factor that is independent of gastric mucosal atrophy, and that RASGRF1 methylation could be a predictive marker of gastric cancer risk that would overcome the disadvantages of other screening methods, such as the serum pepsinogen test and EGD.

The function of RASGRF1 in normal stomach and during gastric tumorigenesis is largely unknown, but it may exert oncogenic effects through activation of Ras proteins (35). On the other hand, one recent study showed that RASGRF proteins bind directly to Cdc42, another Ras-related GTP-binding protein, and suppress Cdc42-mediated cellular processes, including tumor cell invasion and transformation (38). In this study, we found that ectopic expression of RASGRF1 suppressed proliferation and invasion by gastric cancer cells, which is suggestive of its tumor suppressor role, although we carried out only overexpression experiments. Alteration of RASGRF1 gene in human cancer has not yet been reported; in fact, this is the first report of its epigenetic silencing, and further study will be needed to clarify its function in carcinogenesis.

In summary, we have comprehensively analyzed the DNA methylation status of gastric mucosa specimens from patients with gastric cancer. We identified a number of methylated genes that might be involved in an epigenetic field defect in the stomach. Among them, RASGRF1 is a novel gastric cancer–associated gene prevalently methylated in the background mucosa in both intestinal and diffuse type gastric cancer. The combination of a DNA methylation test using RASGRF1 as a marker with a pepsinogen test or EGD would greatly improve the efficacy of risk assessment and surveillance of gastric cancers.

No potential conflicts of interest were disclosed.

Conception and design: E. Yamamoto, H. Suzuki, T. Sugai, K. Imai, M. Toyota, Y. Shinomura

Development of methodology: E. Yamamoto, H. Yamano

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): H. Takamaru, E. Yamamoto, H. Yamano, K. Yoshikawa, T. Harada, R. Suzuki, H. Yamamoto, M. Kai, T. Sugai

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): H. Takamaru, E. Yamamoto, M. Nojima, R. Maruyama, T. Sugai

Writing, review, and/or revision of the manuscript: H. Takamaru, E. Yamamoto, H. Suzuki, T. Sugai

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): E. Yamamoto, H. Yamano, T. Kimura, M. Ashida, R. Suzuki, H. Yamamoto, T. Sugai

Study supervision: H. Suzuki, T. Tokino, K. Imai, M. Toyota, Y. Shinomura

The authors thank Dr. Yutaka Kondo for technical assistance with the MCAM and Dr. William F. Goldman for editing the manuscript.

This study was supported in part by a Grants-in-Aid for Scientific Research (B) from the Japan Society for Promotion of Science (to Y. Shinomura), a Grant-in-Aid for the Third-term Comprehensive 10-year Strategy for Cancer Control (to M. Toyota and H. Suzuki), a Grant-in-Aid for Cancer Research from the Ministry of Health, Labor, and Welfare, Japan (to M. Toyota and H. Suzuki), A3 foresight program from the Japan Society for Promotion of Science (to H. Suzuki), and the Takeda Science Foundation (to E. Yamamoto).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.