Polymorphisms in metabolic genes, their combination and interaction with tobacco smoke and alcohol consumption and risk of gastric cancer: a case-control study in an Italian population

Background The distribution and the potential gene-gene and gene-environment interaction of selected metabolic genetic polymorphisms was investigated in relation to gastric cancer risk in an Italian population. Methods One hundred and seven cases and 254 hospital controls, matched by age and gender, were genotyped for CYP1A1, CYP2E1, mEH, GSTM1, GSTT1, NAT2 and SULT1A1 polymorphisms. Haplotype analysis was performed for EPHX1 exons 3 and 4, as well as CYP2E1 RsaI (*5 alleles) and CYP2E1 DraI (*5A or *6 alleles). The effect modification by alcohol and cigarette smoking was tested with the heterogeneity test, while the attributable proportion (AP) was used to measure the biological interaction from the gene-gene interaction analysis. Results Gastric cancer risk was found to be associated with the inheritance of GSTT1 null genotype (OR = 2.10, 95%CI: 1.27–3.44) and the SULT1A1 His/His genotype (OR = 2.46, 95%CI: 1.03–5.90). No differences were observed for the haplotype distributions among cases and controls. For the first time an increased risk was detected among individuals carrying the *6 variant allele of CYP2E1 if ever-drinkers (OR = 3.70; 95%CI: 1.45–9.37) with respect to never-drinkers (OR = 0.18; 95% CI: 0.22–1.46) (p value of heterogeneity among the two estimates = 0.001). Similarly, the effect of SULT1A1 variant genotype resulted restricted to ever-smokers, with an OR of 2.58 (95%CI: 1.27–5.25) for the carriers of His allele among smokers, and an OR of 0.86 (95%CI: 0.45–1.64) among never-smokers (p value of heterogeneity among the two estimates = 0.03). The gene-gene interaction analyses demonstrated that individuals with combined GSTT1 null and NAT2 slow acetylators had an additional increased risk of gastric cancer, with an OR of 3.00 (95%CI: 1.52–5.93) and an AP of 52%. Conclusion GSTT1, SULT1A1 and NAT2 polymorphisms appear to modulate individual's susceptibility to gastric cancer in this Italian population, particularly when more than one unfavourable genotype is present, or when combined with cigarette smoke. The increased risk for the carriers of CYP2E1*5A or *6 alleles among drinkers need to be confirmed by larger prospective studies.

acetylators had an additional increased risk of gastric cancer, with an OR of 3.00 (95%CI: 1.52-5.93) and an AP of 52%.
Conclusion: GSTT1, SULT1A1 and NAT2 polymorphisms appear to modulate individual's susceptibility to gastric cancer in this Italian population, particularly when more than one unfavourable genotype is present, or when combined with cigarette smoke. The increased risk for the carriers of CYP2E1*5A or *6 alleles among drinkers need to be confirmed by larger prospective studies.

Background
Gastric cancer is the second most common cause of mortality from cancer, with 647,000 deaths reported worldwide in 2002 [1]. In many populations, particularly in high-income countries, in the last decades its incidence has gradually decreased, however it still represents the fifth most common type of cancer in Europe and the fourth internationally [1]. The development of gastric cancer appears to be the result of a complex interaction between lifestyle and genetic factors. Among the lifestyle and related risk factors, Helicobacter pylori infection, tobacco smoking, a high intake of salt and lack of food refrigeration all seem to play a major role [2]. Additionally, gastric cancer shows a familial clustering [3]. With regards to genetic factors, several Single Nucleotide Polymorphisms (SNPs) might potentially alter the individual susceptibility to gastric cancer, among them genes coding for metabolic enzymes [4].
A major part of carcinogenic substances require metabolic activation by enzymes to be genotoxic, and inherited variations in carcinogens metabolizing genes may alter enzyme activity and subsequently carcinogens activation or deactivation. Phase I enzymes, including Cytochrome P450 (CYP) and microsomal Epoxide Hydrolase (mEH), activate several compounds to form genotoxic electrophilic intermediates. Activated metabolites are then, in part, detoxified by phase II enzymes, such as glutathione S-transferase (GST), N-acetyltransferase (NAT) and Sulfotransferase (SULT) [5]. We recently showed, for the first time, that SULT1A1 Arg213His polymorphism might affect the risk of gastric cancer [6], while contradictory results concerning several SNPs in metabolic genes have been reported [7][8][9][10][11][12][13][14][15].
Based on the knowledge that metabolic genes are presumed to modulate an individual's susceptibility to cancer by interacting with carcinogens, and since the inheritance of several unfavourable genotypes is supposed to additionally increase the risk of gastric cancer [8,9,11], this hospital-based case-control study aims to investigate the effect on gastric cancer of selected SNPs of CYP1A1, CYP2E1, mEH, GSTM1, GSTT1, NAT2, SULT1A1, and their differential effect according to tobacco smoking and alcohol habits. We also investigated to what extent the inher-itance of more than one unfavourable genotype affects the risk of gastric cancer.

Study population
The study subjects were selected according to a case-control study design as previously described [16]. Briefly, cases were consecutive primary gastric adenocarcinoma patients, with histological confirmation, who underwent a curative gastrectomy in the "A. Gemelli" teaching hospital, located within the Università Cattolica del Sacro Cuore in Rome. We defined gastric cancer cases as including International Classification of Disease Ninth revision codes 151.0-151.9. Controls were selected from cancerfree patients, with a broad range of diagnoses including around 15% of blood donors, admitted to the same hospital during the identical time period and were frequency matched to cases for age (± 5 years) and gender. All subjects were Caucasians born in Italy. According to the Lauren histotype classification [17], the majority (57.8%) of the gastric cancer cases were intestinal. The tumours were located in the antrum (39.3%), in the corpus (14.8%), in the antrum/corpus (28.0%), in the cardia (10.3%), stumps (5.6%) and in the fundum (2.0%). Based on the cytological and architectural atypisms, as well as the histopathological reports [18], patients' tumours were classified accordingly: 68.3% scarcely differentiated, 29.2% moderately differentiated, 2.5% well-differentiated, while 53.8% were staged I-II and 46.2% staged III-IV. With a response rate of 95% and 90% respectively for cases and controls, 102 gastric cancer and 254 controls were recruited.
A venous blood sample was drawn from each participant, collected into an EDTA-coated tubes from which DNA was isolated from peripheral blood lymphocytes. The study was approved by the local review board and written informed consent was obtained from each subject. The procedures followed were in accordance with the Helsinki Declaration.

Data collection
Cases and controls were interviewed by trained medical doctors using a standard questionnaire to elicit information on demographic variables, tobacco smoking (includ-ing cigarette, cigar and pipe) and drinking history, dietary habits and family history of cancer. Questions pertaining to lifestyle focused on the time period ending one year prior to diagnosis. Smoking status was categorized as never and ever-smokers (including both current and former smokers) and alcohol consumption as drinkers/ non-drinkers (the latter including individuals whose alcohol intake less than 7 g/day). Fruit and vegetables intake was classified as high if the participant consumed at least two portions of fruit and two portions of vegetables per day. Meals salt addition was referred to the use of adding salt to the entrées during the main meals. Family history (including non-melanoma skin cancer) of cancer referred to parents, siblings and offspring. Data concerning previous Helicobacter pylori infection was not available for either cases or controls. The response rate for completing the interview was 99.1% for cases (106/107) and 99.6% (253/254) for controls, with the exception of data relating to a family history of cancer [unknown in 7.4% (8/107) of cases and 3.5% (9/254) of controls].

Statistical analysis
The relationship between gastric cancer and putative risk factors were measured using the adjusted odds ratios (ORs) and their 95% CI derived from logistic regression analysis using STATA software (version 8.2). We considered possible risk factors for gastric cancer as potential confounders if the addition of that variable to the model changed the OR by 10% or greater. Confounding checks were performed in both of the univariate and final multivariate models. If a factor was identified as a confounder of any estimated main effect, it was kept in all models. Based on these criteria, we controlled for age, gender, alcohol consumption and family history of cancer, when appropriate. In the multivariable model, we adjusted for the continuous variables of age and alcohol (g/day).
The genotypes of GSTM1 and GSTT1 were dichotomized according to the presence versus absence of the null allele, and NAT2 was dichotomized according to the inferred phenotype (slow versus fast). We analyzed exon 3 and exon 4 mEH genotypes by "imputed phenotype" as suggested from Smith and Harrison [24]. Lastly, we conducted haplotype analysis for EPHX1 exons 3 and 4, as well as CYP2E1*5 and *5A or *6 using Cocaphase software. Hardy-Weinberg Equilibrium (HWE) was tested for separately all of the case and control SNPs.
In order to assess if the effect of the studied polymorphisms is modified by tobacco smoking and alcohol consumption, we performed a stratified logistic regression analysis. An heterogeneity test was then used to test differences among the strata.
Biological interaction between two genes was estimated using departure from additivity of effects as the criterion of interaction, as suggested by Rothman [25]. To quantify the amount of interaction, the attributable proportion (AP) due to interaction was calculated as described by Andersson et al [26]. The AP due to interaction is the proportion of individuals among those exposed to the two interacting factors that is attributable to the interaction per se and it is equal to 0 in the absence of a biological interaction [25]. Finally, in order to test for more than multiplicative effect among two genes, the likelihood ratio test was used, with the homozygous wild-type individuals for both genes as the reference group.

Results
General characteristics of the study population are presented in Table 1. Alcohol consumption and family history of cancer were associated with an increased risk of gastric cancer, with ORs of 2.10 (95% CI: 1.22-3.60) and an OR of 1.93 (95%CI: 1.14-3.26), respectively ( Table 1). The genotype frequencies of our control group were in line with those for Caucasians and were in HWE both for cases and controls (p > 0.05) [5,20]. As shown in Table 2, we found a significant difference in the distribution of GSTT1 null and SULT1A1 His/His genotype amongst cases and controls: 37.1% versus 22.4% (OR = 2.10, 95% CI: 1.27-3.44) and 10.3% versus 5.1% (OR = 2.46; 95% CI: 1.03-5.90), respectively. An increased risk was also detected for NAT2 slow acetylators (OR = 1.38, 95% CI: 0.88-2.19), however not statistically significant. Haplotype analyses indicated that there was no significant linkage disequilibrium between EPHX1 exons 3 and 4, as well  as CYP2E1*5 and *5A or *6, amongst the cases and the controls. Furthermore, the frequency of the estimated haplotypes was the same among the groups (data not shown).
From the stratified analysis according to smoking status (Table 3), the significant association for SULT1A1 observed in the overall analysis seems to be limited to ever smokers, with a p value for heterogeneity among the two strata of 0.03 (Table 3). On the other hand, the increased risk for GSTT1 null individuals was significant regardless of the smoking status (  Table  3).
To reduce the chance of multiple testing, we limited the gene-gene interaction analyses to the three SNPs that exhibited the most prominent association with gastric cancer. It was observed that in all of the combinations individuals carrying two risk genotypes had an additional risk compared to those with only one risk genotype, with an AP greater than 0, however there was no evidence of multiplicative interaction (p values > 0.05, Table 4). The observed effect was particularly high amongst individuals with both GSTT1 null and NAT2 slow (OR = 3.00, 95% CI: 1.52-5.93; AP = 52%) (Table 4). Additionally, by stratifying these data according to smoking status (data not shown), ever-smoker individuals with combined GSTT1 null and NAT2 slow had an OR of 4.23 (95% CI: 1.49-12.01) compared to ever-smokers with combined normal variants, while an OR of 2.60 (95% CI: 1.00-6.67) appeared using the same comparators amongst neversmokers (p value of heterogeneity among the two estimates = 0.49).

Discussion
This case-control study of 107 surgical cases of gastric adenocarcinoma and 254 controls born in Italy evaluated the effect on gastric cancer risk of several metabolic gene polymorphisms simultaneously. Results showed a significantly increased risk for GSTT1 null and for SULT1A1 homozygotes, and an additional risk for NAT2 slow acetylator individuals, although not statistically significant. Risks associated with those genes became substantive when two unfavourable genotypes were combined, with evidence of biological interaction between them. From the gene-environment interaction analysis, we showed effect modification of the association between SULT1A1 and gastric cancer by tobacco smoking, and CYP2E1 (*5A or *6 alleles) by alcohol drinking. In addition, our results confirm previous findings of gastric cancer risk to be increased by alcohol intake and family history for cancer [27,28].
Several limitations should be taken into account in the interpretation of our results. Firstly, based on the prevalence of the analyzed genotypic variants in our population (Table 2), our study was powered to detect an OR of 2.0 for common polymorphisms (with a significance level of 5%), however not for CYP2E1*5 allele carriers, CYP2E1*5A or *6 allele carriers and the homozygotes variants of SULT1A1, EPHX3 and EPHX4. The study's sample size limits the ability to explore the combined effects of the genotypes, or gene-environment interactions, which highlights the need to increase the sample size in order to confirm our results. However, when appropriately conducted, large and small studies should give, theoretically, the same results, with just a more precise effect measure estimate from the larger ones [29]. Secondly, as in all casecontrol studies information bias may exist, leading to biased ORs related to the gene-environment interaction results. Thirdly, data on Helicobacter pylori infection were not available in our population.
This is the first study conducted on an homogenous ethnic group who evaluated the effect on gastric cancer risk of several metabolic genes SNPs contemporarily, and the effect of their combination with tobacco and alcohol. One of the main source of confounding in the genetic association studies arises from population stratification, since the ethnicity itself may be related to a specific disease and to the allele frequencies as well [30,31]. Our study showed a significant association between GSTT1 null genotype and gastric cancer, which is in keeping with the results of a recent meta-analysis considering only high-quality papers [8]. Individuals who have the homozyogous deletion in GSTT1 have no enzyme activity, and thus are more susceptible to carcinogens such as benzo [α]pyrene-7,8-diol epoxide and smaller reactive hydrocarbons, such as ethylene oxide and diepoxybutane [8]. We also reported that individuals carriers of the SULT1A1 variant allele, who have limited detoxification capability of xeniobiotics through sulfonate conjugation, have an additional risk of gastric cancer if smokers.
To our knowledge, we reported for the first time a strong effect modification by alcohol of the association between CYP2E1*5A or *6 alleles and gastric cancer, with an increased risk among ever-drinkers. Two previous studies [32,33] reported no association between CYP2E1*5A or *6 allele and gastric cancer, however no one of them stratified data according to alcohol habits. Additionally, one study evaluating the identical association among black South-African males showed an increased risk of oesophageal cancer among drinkers carrying the CYP2E1*5A or *6 alleles [34]. CYP2E1 is a naturally ethanol-inducible enzyme that is mainly involved in the metabolic activation of N-nitrosamines present in tobacco smoke and some dietary compounds, for which a causative role in gastric carcinogenesis has been hypothesised [2], and in the metabolism of fatty acids and several halogenated and aromatic compounds [35]. Additionally, CYP2E1 plays a minor role in alcohol metabolism, through the oxidation of ethanol to acetaldehyde and 1-hydroxyacetyl radicals [35]. The *5A or *6 alleles of CYP2E1 is characterized by some studies in an increased gene expression [36], so that individuals carrying the unfavourable variant might be at higher risk of gastric cancer because of: i) hyper activation of N-nitrosamines in more reactive species, especially among drinkers since enzyme activity is induced by alcohol; ii) hyper production of reactive oxygen species and subsequent cell toxicity generated by ethanol metabolism among drinkers. We expected to gain similar results for CYP2E1 RsaI polymorphism, identically associated with increased enzyme activity, however the few subjects in the stratified analysis probably did not show it. Since these results, however, are based on very few subjects (only one case drinker bearing *5A or *6 alleles) they need to be confirmed by larger studies.
Among the main results of our study, we found that GSTT1 null genotype individuals contemporarily NAT2 slow acetylators have a strongly increased risk of gastric cancer, with a more than just the additive effect of the risks associated with each of the two inherited SNPs. N-acetylation is considered a major detoxification step for carcinogenic aromatic arylamines, while GSTT1 is involved in the detoxification of polycyclic aromatic hydrocarbons, so individuals with one or both depleted phase II enzyme activities might be particularly susceptible to gastric damage from carcinogens, which is supported by the finding of an additional risk for ever-smokers. We used the attributable proportion due to interaction as a measure to quantify the biological interaction between those combined SNPs and showed a strong interaction between them. Assuming that the relationships studied are causal and based on the definition of biological interaction among two component causes [25,37], our results suggest that 52% of gastric cancer cases among GSTT1 null individuals with combined NAT2 slow acetylator phenotype are caused through a mechanism in which both risk factors are biological dependent in the same disease process. In other words, since biological interaction among two causes occurs when the effect of one is dependent from the presence of the other, in the absence of either of the two components (GSTT1 null or NAT2 slow), than a substantial number of gastric cancer cases would not occur. Given that in our population 25% of cases had a combination of those unfavourable genotypes, this means that a non negligible proportion of gastric cancer cases would have never developed if those enzymatic activities were adequate.

Conclusion
This study suggests that in this Italian population, GSTT1, SULT1A1 and NAT2 polymorphisms may modulate an individual's susceptibility to gastric cancer, particularly when more than one unfavourable genotype is present and in combination with cigarette smoke. Additionally, we showed that individuals carrying the *5A or *6 alleles of CYP2E1 are at increased risk for gastric cancer in drinkers. Clearly, since our study is based on a limited number of cases, it is critical that larger prospective studies possibly based on a single ethnic group confirms our results.