Impact of PSCA variation on gastric ulcer susceptibility.

This research project was approved by the ethical committees at the University of Tokyo, RIKEN, and Aichi Cancer Center. All participants provided written informed consent as approved by the ethical committees of the University of Tokyo and Aichi Cancer Center. The demographic details of study participants are summarized in Table 1 . A total of 3,866 gastric ulcer patients, and 20,791 gastric ulcer negative controls were obtained from BioBank Japan that was initiated in 2003 with the funding from the Ministry of Education, Culture, Sports, Science and Technology, Japan [37] . In the BioBank Japan Project, DNA and serum of patients with 47 diseases were collected through collaborating network of 66 hospitals throughout Japan. The list of participating hospitals is shown in the following website ( http://biobankjp.org/plan/member_hospital.html ). A total of 425 gastric ulcer cases and 1,874 healthy controls were obtained from the Aichi Cancer Center. The diagnosis of gastric ulcer was based on clinical, endoscopic, and histological features. List of disease-mix control samples used in this study was shown in Table S1 . We excluded patients with duodenal ulcer or gastric cancer from both cases and controls. Deregulation of PSCA was reported in many types of malignancy such as prostate, pancreatic, lung, bladder, gastric, cholangiocarcinoma, and esophageal cancer [14] – [16] , [20] , [38] , [39] . In addition, ABO locus was previously shown to be associated with various diseases such as myocardial infarction and pancreatic cancer [40] , [41] . Therefore, we excluded subjects with these diseases from case mix controls. We also excluded the subjects with continuous NSAID intake. Genotyping platforms used in this study are shown in Table 1 . A total of 1,862 gastric ulcer cases and 20,791 gastric ulcer negative control samples were genotyped with Illumina Human Hap610-Quad or with Human Hap550v3. The other samples were genotyped by the Invader assay system (Third Wave Technologies, Madison, WI) or Taqman assay. The association of SNPs rs2294008 and rs505922 with gastric ulcer was tested by chi-square test. The Odds ratios were calculated by considering the protective allele as the reference allele. The association of SNPs genotyped by Illumina Human Hap610-Quad with gastric ulcer was tested by multivariate logistic regression analysis upon adjusting for age at recruitment and gender using PLINK [42] . Heterogeneity across two stages was examined by Cochran Q test [43] .

A total of 4,291 gastric ulcer cases and 22,665 controls without having the past history of duodenal ulcer or continuous NSAID intake were recruited from the BioBank Japan and the Aichi Cancer Center ( Table 1 ). We then genotyped SNP rs2294008 and rs505922 in two case-control sample sets and examined the association with gastric ulcer in three genetic models (additive, recessive, and dominant model) ( Table 2 ). To increase the statistical power of this study, we used subjects with either of 22 diseases as control samples. Therefore we evaluated the confounding effect of disease mix control samples used in this analysis. SNPs rs2294008 and rs505922 did not show significant association between case-mix controls (n = 19,884) and healthy volunteers (n = 2,781) ( Table S1 ). In addition, both SNPs did not show the significant deviation from HWE (Hardy-Weinberg equilibrium) in each disease group. Therefore disease mix controls seem not to largely affect the association result in our analysis. Subjects with a history of gastric cancer or duodenal ulcer were excluded from cases and controls. Control samples consist of patients with colon cancer, breast cancer, diabetes, arteriosclerosis obliterans, atrial fibrillation, brain infarction, drug response, amyotrophic lateral sclerosis, liver cancer, liver cirrhosis, osteoporosis, fibroid, cervical cancer, chronic hepatitis B, ovarian cancer, pulmonary tuberculosis, keloid, drug eruption, hematological cancer, uterus cancer, heat cramp, endometriosis, and 907 healthy volunteers. We analyzed 4,291 gastric ulcer cases and 22,665 controls. rs2294008 (C allele) and rs505922 (T allele). P values were obtained using chi-square test. To calculate odds ratios (OR), non risk alleles were considered as references. Heterogeneity across two stages was assessed by Cochran Q test. OR and P values were obtained using the Mantel-Haenszel fixed-effects model in the meta analysis. The results of association analyses revealed that gastric ulcer patients had a higher frequency of C allele at rs2294008 than the control group in both sets (39.7% vs 36.9% and 40.1% vs 37.0%, respectively). A meta-analysis of the two studies showed the significant association of rs2294008 in an additive model with no evidence of heterogeneity (P = 5.85×10 −7 with OR of 1.13), although the association was not statistically significant among Aichi Cancer Center cohort probably due to smaller sample size. Risk alleles (C allele at rs2294008) in the two sample sets were consistent between duodenal ulcer and gastric ulcer, indicating the role of PSCA variation as common genetic factors for peptic ulcer. However impact of this variation on gastric ulcer risk was not as strong as those on duodenal ulcer reported previously [13] . On the other hands, SNP rs505922 showed inconsistent results between two cohorts. A T allele of rs505922 increased gastric ulcer risk in all three genetic models in BioBank Japan cohort. However, gastric ulcer patients exhibited lower frequency (53.5%) of a T allele than the healthy controls (55.1%) in the Aichi Cancer Center cohort. Therefore, further association analysis is essential to determine the role of ABO variations on gastric ulcer susceptibility. Since we have genotyping results of 1,862 gastric ulcer cases and 17,482 controls analyzed by Illumina Human Hap610-Quad genechip, we conducted whole genome screening using these sample set. Although 62 SNPs exhibited suggestive associations with P values of less than 1×10 −4 , no SNPs cleared genome wide significant threshold ( Table S2 and Figure S1 ). Thus, our sample set did not have sufficient statistical power to detect gastric ulcer susceptibility loci by GWAS. We also investigated the association of previously reported genes with gastric ulcer ( Table 3 ). We selected 32 SNPs at five gene loci that had been genotyped by Illumina Human Hap610-Quad genechip. As a result, two loci at LTA and PTGS1 indicated suggestive association (P = 1.64×10 −3 and 0.0376), although these associations were not statistically significant after Bonferroni’s correction ( P <0.00156 = 0.05/32). Thus further analyses are necessary to elucidate the role of these variations on gastric ulcer. We analyzed 1,862 gastric ulcer cases and 17,482 controls in this analysis. Chr.,chromosome; Postion in the NCBI Build 36.3. P values were calculated by Cochran Armitage trend test. OR, odds ratio was calculated by considering the major allele as the reference.

The development of gastric ulcer is determined by the interplay between gastric acid secretion and mucosal resistance, however their underling pathogenesis has not been fully elucidated. Gastric mucus, a gelatinous material secreted by gastric mucous cells, serves as an unstirred layer through which the diffusion of acid and pepsin is reduced. We here found that variation in the PSCA gene was significantly associated with gastric ulcer. PSCA was initially identified as a tumor antigen that was highly expressed in prostate, bladder, and pancreatic cancer tissues [15] , [16] . Since tumor cells treated with anti-PSCA antibody exhibited a growth suppressive effect [17] , [18] , cell surface-PSCA is considered to play an important role in cell proliferation. In contrast, down-regulation of PSCA in gastric and esophageal cancer tissues was also reported [19] , [20] . Thus the role of PSCA in carcinogenesis is still controversial [21] . These diverse effects of PSCA among various cancer types might be partially explained by the effect of genetic variation. Individuals carrying the T allele at rs2294008 express PSCA proteins with an additional fragment of nine amino acids at the N-terminal portion [13] . On the other hand, individuals carrying the C allele at rs2294008 express a shorter PSCA protein which lacks the signal peptide and is predicted to be localized in the cytoplasm without glycosylation [22] . We also found that the cytosolic shorter PSCA protein was more susceptible to proteasomal degradation than the long PSCA protein at the cell-surface. Since PSCA-derived peptides were reported to be a target of T-cell-based immunotherapy for advanced prostate cancer [23] , the shorter PSCA protein would cause the activation of CD4-positive and/or CD8-positive T cells and subsequently promote epithelial mucosal injury [24] . In contrast, the long PSCA protein at the cell surface might facilitate mucosal repair by enhancing epithelial cell proliferation. In addition, T allele of SNP rs2294008 was shown to be associated with higher mRNA and protein expression [25] . Thus the impact of PSCA on gastric ulcer and carcinogenesis could be regulated by the PSCA variation. H. pylori plays an important role in the development of gastritis, peptic ulcers, and gastric cancer, and the eradication of H. pylori was shown to reduce the recurrence of gastric ulcer [26] and prevent the onset of gastric cancer [27] . Since vertical transmission during childhood is the major source of infection, family history of H. pylori infection or H. pylori -related diseases is a risk factor for H. pylori infection [28] – [30] . In addition, recent accumulated evidences revealed a number of risk factors of gastric cancer (T allele at rs2294008, blood type A, decreased gastric acid, intake of proton pump inhibitor/H 2 blocker, and CagA in H. pylori [31] ) or peptic ulcer (C allele at rs2294008, blood type O, NSAID intake, dupA in H. pylori ) [32] . In addition, CYP2C19 genotype was associated with the response to triple anti- H. pylori therapy including proton pump inhibitor [33] . However, our previous analysis revealed that SNP rs2294008 and rs505922 did not associated with H. pylori prevalence [13] . Taking the above information into account, the estimation of disease risk and drug efficacy would enable us to determine the appropriate treatment protocol for H. pylori carriers. Here we found that PSCA variant was significantly associated with gastric ulcer. In our previous analysis, PSCA variation did not associate with H. pylori prevalence [13] . Since H. pylori infection was associated with many diseases such as MALT lymphoma [34] , idiopathic thrombocytopenic purpura [35] , atrophic gastritis [36] , and NSAID-induced gastric ulcer, it is very interesting to evaluate the effect of PSCA variation on these diseases. We hope our findings would contribute to the elucidation of disease pathogenesis as well as to the establishment of personalized medical treatments in the future.

Peptic ulcer is the most common disease in the gastrointestinal tract with symptoms of nausea, vomiting, and abdominal pain, and sometimes causes bleeding and perforation with acute peritonitis. Lifetime prevalence of peptic ulcer is 10–15% in the Japanese and 4–10% in Caucasians [1] – [3] . Approximately 70% of gastric ulcer patients and 90% of duodenal ulcer patients are associated with H. pylori infection [4] . Since eradication of H. pylori by antibiotics in combination with proton pomp inhibitor can effectively cure peptic ulcer [5] , H. pylori is shown to be the major cause of peptic ulcer. Although nearly 50% of individuals on the earth are infected with H. pylori, most of them remain asymptomatic indicating that the clinical outcome after the H. pylori infection varies substantially between individuals. These inter-individual diversities are affected by various factors including bacteria subtypes, host response, and their interaction. Duodenal ulcer promoting gene A ( dupA ) in H. pylori was indicated to induce interleukin (IL)-8 that increases the risk of duodenal ulcer and decreases the risk of gastric cancer [6] , [7] . Nonsteroidal anti-inflammatory drugs (NSAIDs) and smoking are known risk factors for peptic ulcer [8] , [9] . In addition to these bacterial and environmental factors, host genetic factors had been implicated to have some roles in the risk of peptic ulcer. Proband-wise concordance rate of peptic ulcer in monozygotic twins was as high as 23.6% while that in dizygotic twins was 14.8%. Several candidate gene approaches revealed the possible association of genetic variations in IL-6 , IL-8 , IL-10 [10] , TNF , LTA [11] , and COX1 [12] with peptic ulcer risk. In our previous genome wide association study (GWAS) of duodenal ulcer using a total of 7,035 cases and 25,323 controls, we identified the significant association of genetic variations at PSCA ( prostate stem cell antigen ) and the ABO blood group with duodenal ulcer [13] . The C allele of rs2294008 at PSCA increased the risk of duodenal ulcer (odds ratio (OR) of 1.84 with P value of 3.92×10 −33 ) in a recessive model, while it decreased the risk of gastric cancer (OR of 0.79 with P value of 6.79×10 −12 ) as reported previously [14] . Our functional analyses revealed that the T allele of SNP rs2294008 creates an upstream translational initiation codon and add the signal peptide sequences at the N-terminal portion, resulting in alteration of the protein subcellular localization from cytoplasm to cell surface. SNP rs505922 on ABO was also associated with duodenal ulcer in a recessive model (OR of 1.32 with P value of 1.15×10 −10 ). Since H. pylori infection and non-steroidal anti-inflammatory drugs induce gastroduodenal mucosal injury which would cause duodenal and gastric ulcer, we examined the role of variants in the PSCA and ABO genes on gastric ulcer risk among Japanese population.

Manhattan plot showing the genome-wide P values of association. The P values were obtained by logistic regression analysis upon adjustment for age and gender. The y-axis represents the –log10 P values of 480,566 SNPs, and their chromosomal positions are shown on x-axis. (TIF) Click here for additional data file. Genotype frequency of two SNPs in disease mix controls. (DOCX) Click here for additional data file. The result of association analysis of Gastric ulcer in GWAS. (DOCX) Click here for additional data file.