Impact of PSCA Polymorphisms on the Risk of Duodenal Ulcer

Background While duodenal ulcer (DU) and gastric cancer (GC) are both H. pylori infection-related diseases, individuals with DU are known to have lower risk for GC. Many epidemiological studies have identified the PSCA rs2294008 T-allele as a risk factor of GC, while others have found an association between the rs2294008 C-allele and risk of DU and gastric ulcer (GU). Following these initial reports, however, few studies have since validated these associations. Here, we aimed to validate the association between variations in PSCA and the risk of DU/GU and evaluate its interaction with environmental factors in a Japanese population. Methods Six PSCA SNPs were genotyped in 584 DU cases, 925 GU cases, and 8,105 controls from the Japan Multi-Institutional Collaborative Cohort (J-MICC). Unconditional logistic regression models were applied to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for the association between the SNPs and risk of DU/GU. Results PSCA rs2294008 C-allele was associated with per allele OR of 1.34 (95% CI, 1.18–1.51; P = 2.28 × 10−6) for the risk of DU. This association was independent of age, sex, study site, smoking habit, drinking habit, and H. pylori status. On the other hand, we did not observe an association between the risk of GU and PSCA SNPs. Conclusions Our study confirms an association between the PSCA rs2294008 C-allele and the risk of DU in a Japanese population.


INTRODUCTION
Peptic ulcer-duodenal ulcer (DU) and gastric ulcer (GU)-is one of the most common gastrointestinal diseases, with an estimated lifetime prevalence of 5-10% in the general population. 1 Peptic ulcer is defined as a mucosal defect which penetrates through the muscularis mucosa with a diameter of at least 0.5 cm. 2 Among the various risk factors of peptic ulcer reported to date, including smoking and drinking, one of the main factors is infection with Helicobacter pylori (H. pylori), 3,4 which is an established risk factor of gastric cancer. 5 While both DU and gastric cancer (GC) are H. pylori infectionrelated diseases, individuals with DU are well known to have a lower risk for GC. 6,7 It is increasingly clear that these heterogeneities are influenced by not only bacterial but also host factors. A genome-wide association study (GWAS) found an association between PSCA rs2294008 T-allele and the risk of GC (per allele OR 1.67, P value 2.2 × 10 −15 ) in a Japanese population, 8 while a second GWAS found associations between variations in PSCA gene and ABO gene and the risk of DU in an Japanese population. 9 These latter authors reported that the PSCA rs2294008 C-allele increased the risk of DU (per allele OR 1.84, P value 3.92 × 10 −33 ) but decreased that of GC (per allele OR 0.79, P value 6.79 × 10 −12 ). 9 In addition, they reported that the PSCA rs2294008 C-allele increased the risk of GU (per allele OR 1.13, P value 5.85 × 10 −7 ); however, ABO polymorphisms are not significantly associated with the risk of GU. 10 They also reported the function relevance of PSCA for peptic ulcer and gastric cancer 8,9 ; on the other hand, the function relevance of ABO for peptic ulcer has not been clarified yet. The association between DU=GC and these risk factors is summarized in Figure 1. Following these initial reports, however, few studies have since validated this association.
Here, we conducted a cross-sectional study to replicate the association between the variations in the previously reported PSCA loci 9,10 and risk of DU=GU, and evaluate the interaction between these variations and smoking=drinking status and H. pylori status on the risk of DU=GU in a Japanese population. We also evaluated the ABO loci, as reported in two previous studies. 9,10 MATERIAL AND METHODS

Study subjects
The Japan Multi-Institutional Collaborative Cohort (J-MICC) study is a large cohort study launched in 2005 to confirm and detect gene-environment interactions in lifestyle-related disease. Details of the J-MICC study have been reported elsewhere. 11 Briefly, the study includes 92,647 participants aged 35-69 years from 13 areas throughout Japan (Aichi, Chiba, Fukuoka, Iga, Kagoshima, Kyushu-KOPS, Kyoto, Okazaki, Sakuragaoka, Saga, Shizuoka-Daiko, Takashima and Tokushima sub-cohorts) as at end of March 2014. All participants gave written informed consent to participate; answered a questionnaire that inquired about lifestyle-related factors, past medical history, medication status and anthropometric characteristics; and provided a blood sample. The study protocol was approved by the Ethics Committees of Nagoya University Graduate School of Medicine and the other institutions participating in the J-MICC study. The present study was conducted in accordance with the principles expressed in the World Medical Association Declaration of Helsinki. A total of 14,539 participants were randomly selected to be genotyped from 47,163 participants in 12 areas (except for the Iga sub-cohort, where the survey was conducted from 2013 to 2014) recruited between 2004 and 2013. Subjects were selected as shown in Figure 2. We excluded 26 subjects because of inconsistent baseline information between the questionnaire and genotyping on sex; 422 whose genotype data did not meet quality control (QC) filters; 32 because of a lack of questionnaire data; 2,743 with a history of cancer; and 1,239 with a lack of data on ulcer status. In addition, to clarify the substantial impact on the risk of DU or the risk of GU, we excluded 463 with DU=GU overlap cases. Finally, we selected 9,614 subjects for participation in this study.

Past medical history and lifestyle-related factors
The questionnaire for the J-MICC study included questions on past medical history, and cigarette smoking, alcohol drinking, and coffee drinking habits. Medical histories for DU and GU were enquired about in the three categories of never, past, and current. The combination of a past and current medical history of DU=GU was considered positive, and otherwise as negative. Smoking= drinking habits were enquired about in the three categories of never, former, and current. Former smokers=drinkers were defined as those who had quit smoking=drinking for more than 1 year. Never smokers were defined as those who smoked less than 100 cigarettes in their lifetime. We defined the combination of former and current smokers=drinkers as ever smokers= drinkers. Smoking habit was evaluated in pack-years, calculated by multiplying the number of packs consumed per day by the number of years of smoking. Alcohol consumption of each beverage type (Japanese sake, beer, shochu, whiskey, and wine) was estimated as the average number of drinks per day, which was converted into a Japanese sake equivalent. One "go" of Japanese sake contains 23 g of ethanol, which is equal to one large bottle (633 mL) of beer, 108 mL of shochu (distilled spirit), two shots (57 mL) of whiskey, or two and a half glasses of wine (200 mL). Total alcohol consumption was determined as the total sum of pure ethanol consumption (g=day) of each alcohol beverage. Coffee consumption was obtained in terms of the frequency and amount of cups according to the following categories: almost none, 1-2 cups=week, 3-4 cups=week, 5-6 cups=week, 1-2 cups=day, 3-4 cups=day, and ≥5 cups=day. We classified coffee consumption based on its distribution among the subjects as almost none, <1 cup of day, and ≥1 cup of day. This study is based on the data version J-MICC_CS_20180111.
Genotyping and quality control filtering DNA was prepared from buffy coat fractions using a BioRobot M48 Workstation (Qiagen Group, Tokyo, Japan) at the central study office. For the samples from two areas (Fukuoka and Kyushu-KOPS), DNA was extracted from samples of whole blood using an automatic nucleic acid isolation system (NA-3000; Kurabo, Osaka, Japan

Genotype imputation
Genotype imputation was performed using SHAPEIT 17 and Minimac3 18 software base on the 1,000 Genomes reference panel (phase 3). After genotype imputation, variants with an imputation quality r 2 < 0.3 were excluded, resulting in 12,617,547 variants. In our primary analysis, we used imputed genotype data, namely GT format output by Minimac3, estimation of most likely genotype. We also evaluated allele dosage data by imputation as well to evaluate consistency with genotype data analysis.

Candidate SNP selection
To reduce the number of SNPs tested in this analysis, we prespecified tagSNPs based on HapMap-JPT data using National  Figure 2. Study subject selection. A total of 14,539 participants were randomly selected for genotyping from 47,163 participants. We excluded 26 subjects because of inconsistent baseline information between the questionnaire and genotyping on sex; 422 whose genotype data did not meet quality control (QC) filters; 32 because of a lack of questionnaire data; 2,743 with a history of cancer; 1,239 with a lack of data on ulcer status; and 463 with DU/GU overlap cases. Finally, we selected 9,614 subjects for participation in this study. QC, quality control; DU, duodenal ulcer; GU, gastric ulcer Association Between PSCA Variants and Duodenal Ulcer Risk Institute of Health (NIH) LD TAG SNP Selection. 19 We selected PSCA six SNPs and ABO 18 SNPs from eFigure 1 and eFigure 2.
We applied an R 2 threshold of 0.8 for SNPs with a MAF of more than 0.05. We forced the inclusion of previously reported SNPs (PSCA rs2294008 and ABO rs505922) due their association with the risk of DU in a previous study. 9 In addition, we assessed accordance with Hardy-Weinberg equilibrium using the chisquared test.

Serum sample measurement
We also evaluated H. pylori status by measuring anti-H. pylori IgG serum antibody in 2,760 samples from four among twelve study sites of the J-MICC Study (Daiko, Kyoto, Aichi Cancer Center, and Okazaki). Serum samples were immediately stored at −80°C until measurement. Anti-H. pylori IgG serum antibody was measured using a direct ELISA kit, "E plate 'Eiken' H. pylori Antibody" (Eiken Kagaku, Tokyo, Japan), with values of 10.0 units=mL or higher regarded as seropositive according to the manufacturer's instructions.

Statistical analysis
First, to narrow down the number of SNPs for interaction analysis, we analyzed the association of the SNPs with the risk of DU=GU by unconditional logistic regression analysis adjusted for age (continuous), sex, and study site. We also analyzed the association of the SNPs with the risk of H. pylori infection by the same model. We applied Bonferroni corrected P values of 0.05=24 to avoid false positive associations. Second, we examined for interaction between selected SNP and smoking= drinking status for the risk of DU=GU. We included interaction term between minor allele numbers (0, 1, and 2) of corresponding SNP and status (ever vs never) in the models. We added packyears for ever smokers and total sum of pure alcohol consumption (g=day) for ever drinkers as covariates in the models. Finally, we examined interaction between selected SNP and H. pylori status for the risk of DU=GU among available data. We included interaction term between the minor allele numbers (0, 1, and 2) of corresponding SNP and H. pylori status. Throughout the analysis, we estimated per allele odds ratios (ORs) and their 95% confidence intervals (CIs) using the major allele homozygote as reference. All analyses were performed using STATA version 15.1 software (Stata Corp., College Station, TX, USA).

RESULTS
Our analysis included 584 DU cases, 925 GU cases, and 8,105 controls. Table 1 summarizes the demographic, lifestyle, and medical characteristics of the study subjects by DU=GU status. Mean age was 56.2 years in DU, 56.6 years in GU, and 53.1 years in the controls, respectively. The proportion of males was higher in the case groups than in the control group (63.4% in DU, 57.6% in GU, and 41.5% in the controls). The proportion of never drinkers was lower in the case groups than in the control group (30.3% in DU, 36.5% in GU, and 42.1% in the controls). Similarly, the proportion of never smokers was lower in the case groups than in the control group (40.4% in DU, 44.5% in GU, and 62.8% in the controls). Although H. pylori status data was only available for some participants, the proportion who were H. pylori status-positive was higher in the case groups than in the control group (12.0% in DU, 13.3% in GU, and 9.6% in the controls). eTable 1 shows allele frequencies of PSCA and ABO SNPs at the survey; the r 2 at the imputation; and MAF in the HapMap-JPT data set, Human Genetic Variation Database (HGVD) 20 and Integrative Japanese Genome Variation Database (IJGVD). 21 All SNPs were in accordance with the Hardy-Weinberg equilibrium. Imputation quality for all SNPs showed high accuracy (r 2 > 0.8).
All SNPs showed a difference in MAF of less than 0.1 between this survey and the HapMap JPT dataset, HGVD or IJGVD. 21 The association of PSCA and ABO SNPs with the risk of DU and GU is shown in Table 2. The PSCA polymorphisms were significantly associated with the risk of DU (rs2294008, rs2920296, and rs2976397), while PSCA polymorphisms were not associated with the risk of GU. The ABO polymorphisms were not associated with the risk of DU or GU. We also analyzed data of allele dosage imputed by Minimac3, 18 and consistent results were observed (as shown in eTable 2). For PSCA rs2294008, the P value was 2.28 × 10 −6 for DU (per allele OR 1.34; 95% CI, 1.18-1.51). Rs2920296 and rs2976397 had also significant P values, and rs2920296 had the lowest P value, but we did not choose rs2920296 and rs2976397 for further detailed analysis based on the fact rs2294008 is the truly functional locus. 9 The association of PSCA rs2294008 with the risk of DU=GU stratified by smoking=drinking status is shown in Table 3. PSCA rs2294008 was significantly associated with the risk of DU regardless of smoking status. Similarly, after stratification by drinking status, PSCA rs2294008 was also significantly associated with the risk of DU. Regarding GU, we did not observe any association with PSCA rs2294008 after stratification by smoking=drinking status. We did not observe obvious multiplicative interaction between PSCA rs2294008 and smok-ing=drinking status for the risk of DU=GU. We also examined interaction between the selected PSCA rs2294008 and coffee consumption for the risk of DU=GU, but again saw no obvious multiplicative interaction (data not shown). Addition of packyears for ever smokers and g=day for ever drinkers as covariates resulted in no significant change in point estimates (per allele OR 1.29; 95% CI, 1.10-1.51 in ever smokers; per allele OR 1.33; 95% CI, 1.15-1.55 in ever drinkers).
We did not observe an association of PSCA and ABO SNPs and the risk of H. pylori infection (as shown in eTable 3). The association of PSCA rs2294008 with the risk of DU=GU stratified by H. pylori status among available data is shown in Table 4. This analysis included 150 DU cases, 297 GU cases, and 2,313 controls. PSCA rs2294008 was significantly associated with the risk of DU regardless of H. pylori status. Regarding GU, we did not observe significant association with PSCA rs2294008 after stratification by H. pylori status. We did not observe obvious multiplicative interaction between PSCA rs2294008 and H. pylori status for the risk of DU=GU.

DISCUSSION
We found a significant association between variations in PSCA and risk of DU. This association was consistent regardless of age, sex, and study site. However, we did not find an association between variations in PSCA and the risk of GU and an association between variations in ABO and risk of DU=GU. After stratification of the environmental factors smoking=drinking and H. pylori status, PSCA rs2294008 was also significantly associated with the risk of DU. No obvious multiplicative interaction between PSCA rs2294008 and smoking=drinking and H. pylori status was observed.
Our study suggests that PSCA rs2294008 C-allele was associated with an increased risk of DU in the Japanese population. This is consistent with previous studies. 9,22 The PSCA gene is located on chromosome 8q24.2 and encodes a cell membrane glycoprotein which belongs to the Thy-1=Ly-6 family. Several reports suggest that this glycoprotein is involved in cell renewal and proliferation. [23][24][25] While overexpressed in some types of cancers, including prostate, bladder, and pancreatic cancer, 23,26,27 it is downregulated in esophageal and gastric cancer. 28 Functional analysis in a previous GWAS revealed a considerable function for PSCA rs2294008. Tanikawa et al reported that PSCA protein encoded by the rs2294008 T-allele with an additional fragment of nine amino acids at the N-terminus and its localization changes from the cytoplasm to the cell surface, whereas short PSCA protein encoded by the rs2294008 C-allele is localized to the cytoplasm. They also suggested that the shorter PSCA protein encoded by the rs2294008 C-allele might result in insufficient epithelial proliferation to counteract the damage due to a lack of functional cell surface PSCA, resulting in slow recovery from duodenal tissue damage. 9 Those reports may support a significant association between the PSCA rs2294008 C-allele and risk of DU.
In contrast with DU, we did not observe the association between PSCA polymorphisms and the risk of GU. Although it is difficult to completely rule out the possibility of lack of statistical power, but our result indicates that PSCA polymorphisms have higher impact on the risk of DU than on the risk of GU. This difference in the magnitude of association by site is consistent with the previous reports. 9,10 In general, DU and GU have a different etiological spectrum, and differ with regard to the severity and distribution of background gastritis. 29 DU is usually diagnosed in patients with high antral inflammatory scores and high acid secretion, whereas GU is diagnosed in patients with corporal gastritis or pan-gastritis and with normal or decreased acid secretion. These differences might be related to the different impact of these PSCA polymorphisms on the risk of DU and GU. Further clarification of the biological mechanism is required.
In addition, in contrast with previous reports, 9 we did not observe the association between ABO polymorphisms and the risk of DU. One of the reason for this discrepancy might be difference of study population between previous study and this study. Previous study employed subjects from individuals with 47 diseases at the hospital. 9 On the other hand, this study was population-based cohort study mostly from the general population. 11 Further evaluation in different population is needed.
This study had several strengths. First, to our knowledge, it is the largest replication studies of this association in an Asian population following the initial report. Second, it is the first study to evaluate the interaction between PSCA rs2294008 and smoking=drinking and H. pylori status on the risk of DU=GU. Finally, although it is said that candidate gene approach tends to have greater statistical power than GWAS, 30 we did not observe previously reported associations.
Several limitations of this study should also be mentioned. First, it was based on a cross-sectional study in which exposure and outcome measurements were performed concurrently. This design generally does not allow proof of causality because the causal sequence may remain unclear. However, the observed association between these PSCA polymorphisms and risk of DU= GU is likely relatively valid, given that genetic polymorphisms are determined by nature. Second, as the participants in this study were recruited from selected areas, they may have differed from the general population. However, the equivalence of genotype   Association Between PSCA Variants and Duodenal Ulcer Risk distributions between our subjects and those in another Japanese database indicates a lack of such bias. Moreover, our analysis considered study site in the models, which also reduces the likelihood of such bias. Third, we assessed H. pylori status for only some subjects, and the evaluation may be insufficient. However, many previous studies reported a lack of association between PSCA rs2294008 and H. pylori infection prevalence. 9,31-34 and Hishida et al also reported a lack of association in this population. 35 Our result is in accordance with this finding. Thus, H. pylori infection is less likely to bias the association between PSCA rs2294008 and DU=GU risk. Fourth, nonsteroidal anti-inflammatory drugs and psychological stress are also known risk factors of peptic ulcer, 1 but we did not consider these variables in our analysis due to their lack of inclusion in the questionnaire. Further analysis which considers them would be informative. Finally, as the information about past medical history was collected from the questionnaire, it might have been affected by information bias. However, participants answered the questionnaire without knowledge of their genotype, making information bias unlikely; moreover, if any misclassification were present, it would likely be nondifferential and, therefore, likely to underestimate the causal association.
In conclusion, this study confirms an association between the PSCA rs2294008 C-allele and risk of DU in the Japanese population. This association was independent of age, sex, study site, smoking habit, drinking habit, coffee consumption, and H. pylori status. Further studies examining the biological mechanism behind these associations is required.