Introduction

Basic information on whether lifestyle factors and candidate genes are independent of each other is valuable for genetic–environmental study. Such information is essential for future studies to reveal the interaction of lifestyle factors and genetics. Furthermore, an understanding of genes associated with lifestyle factors will contribute to more accurate risk identification and to establishing tailor-made prevention measures.

Two of the most common and important lifestyle factors, cigarette smoking and alcohol drinking, are related to many diseases, including lung cancer, cardiovascular disease, and other chronic diseases. Hence, genetic influence on the use of tobacco has been strongly implicated by cross-sectional studies in twins, association studies, and numerous other genetic epidemiology data (Carmelli et al. 1992; Hussain et al. 2001; Pianezza et al. 1998; Sabol et al. 1999; Yoshida et al. 2001). It is also recognized that the use of tobacco is often accompanied by alcohol consumption (Hopfer et al. 2001). Alcohol consumption is found to be as heritable as smoking (Heath et al. 1991; McGue et al. 2001), and the contribution of ALDH2 and ADH1C genes to habitual alcohol drinking has been well known. However, current knowledge on genetic polymorphism related to smoking or drinking behavior is far from sufficient. Information on associated genetic factors could provide a more rational basis for developing smoking or drinking cessation programs, including identification of persons at high risk and the introduction of suitable interventions to prevent smoking-related or drinking-related diseases.

We examined the relationship between the smoking and drinking habits of 339 Japanese men and their genotypes in 153 single nucleotide polymorphisms (SNPs) in 40 selected genes. The genes were selected based on possible involvement in genetic–environmental, life-style interactions and constitute the genes for xenobiotic metabolism enzymes, DNA repair enzymes, and other stress-related proteins. This population-based association study will provide a list of genes independent of habitual smoking and drinking, which may be included in a future case study. It will also establish the role of genetic influence on smoking and drinking behavior and thus contribute to the future development of genotype-based prevention of smoking-related and drinking-related diseases.

Subjects and methods

Subjects

The selection of subjects is described in detail by Tsugane et al. (1992a, b). Briefly, from 1989 to 1991, using a random sampling method employing resident registration rolls, we recruited men aged 40–49 years from five areas having a population of approximately 100,000 people. Between 170 and 195 subjects were selected from each area. The selected individuals were initially invited to participate in the study by e-mail, and a subsequent letter, telephone call, and home visit encouraged participation. A questionnaire-based interview elicited lifestyle details and health check-up information, as well as yielding blood and urine samples. The overall participation rate was 72%, or 634 out of 880.

Smoking and drinking habits

Subjects were asked whether they had ever used tobacco regularly (habitual smoking) and were then classified as “past smokers,” “current smokers,” or “never smokers.” With regard to alcohol drinking, those who did not habitually consume alcohol once or more per month were considered “nondrinkers;” otherwise subjects were considered “drinkers.” Whether they flush on drinking was recorded as well.

DNA samples and SNPs analyses

A total of 25 ml of blood was drawn by venipuncture, and genomic DNA was extracted from the buffy coat layer using a commercial kit (Wako, Osaka, Japan). Some samples were used in another study (Sugimura et al. 1998) and exhausted before the present study, leaving 339 cases with a representative sample of 0.5 μg or more of DNA. For each gene, 5–8 SNPs were chosen from public databases or published papers, with a total of 289 SNPs being typed for 44 genes using the mass spectroscopy-based technique, Mass ARRAY (Ross et al. 1998). The 44 were selected from genes encoding xenobiotic metabolic enzymes, DNA repair enzymes, and other stress-related proteins. Only SNPs in the Hardy–Weinberg equilibrium with a >0.05 chi-square and SNPs having a minor allele frequency of at least 1% were selected for analysis. Thus, 153 SNPs from 40 genes were finally included. Details of SNP selection, including allele frequency in the 153 SNPs, is described by Yoshimura et al. (2003). The genes included in the present study were cytochrome P450 genes (CYP1A1, CYP1B1, CYP2C9, CYP2C19, CYP2E1, CYP17A1, and CYP19A1), the aryl hydrocarbon receptor gene (AHR), estrogen receptor genes (ESR1, ESR2, and ERRRG), the progesterone receptor gene (PGR), epoxide hydrolase genes (EPHX1, and EPHX2), hydroxysteroid (17-beta) dehydrogenase genes (HSD17B2 and HSD17B3), glutathione S-transferase genes (GSTM2, GSTM3, GSTT2, and GSTP1), N-acetyltransferase genes (NAT1 and NAT2), the catechol-O-methyltransferase gene (COMT), alcohol dehydrogenase genes (ADH1A, ADH1B, and ADH1C), the aldehyde dehydrogenase gene (ALDH2), nitric oxide synthase genes (NOS2A and NOS3), interleukin genes (IL1A and IL1B), repair genes for oxidative DNA damage [OGG1 and NUDT1 (MTH1)], dopamine receptor genes (DRD2, DRD3, and DRD4), the serotonin transporter gene (SLC6A4), the glucocorticoid receptor gene [NR3C1 (GCCR)], the folate metabolizing enzyme gene (MTHFR), and the quinone oxidoreductase gene (NQO1).

Ethical issues

All DNA samples were anonymous and unlinked to specific individual information, i.e., any ID, name, or address. The protocol of the present study was approved by the ethics review committee of the National Cancer Center (protocol number G12-02).

Statistical analysis

First, each SNP was analyzed independently to screen the SNPs possibly related with smoking or drinking. In order to show the ratio of the odds of becoming a smoker or drinker in those exposed to targeted genotype relative to the unexposed individuals, we used an odds ratio as an indicator of the strength of association between genotypes and the smoking or drinking habit. A logistic regression analysis was used to obtain odds ratios, and 95% confidence intervals were calculated using the standard errors of the logistic regression coefficients. P values were calculated from the χ2 test. We also calculated the statistical power of the analyses by using the formula of Schlesselman (Schlesselman 1982). We then included all significant SNPs in the above analyses simultaneously in a stepwise logistic regression analysis to select substantially significant ones. We tested the statistical significance once and did not use a correction of multiple testing because we used a multivariate analysis. All computations were performed using the SAS software package Version 8.2 (SAS Institute, Inc., Cary, NC, USA).

Results

The number of never smokers, past smokers, and current smokers was 55, 71, and 213, respectively. We combined never smokers and past smokers as nonsmokers in order to gain greater investigative power. The number of nondrinkers and drinkers was 71 and 268, respectively. All 153 SNPs in the 40 genes examined are listed in Table 1. Among them, 11 SNPs of five genes (CYP17A1, ALDH2, NOS2A, OGG1, and SLC6A4) in the dominant model and three SNPs of three genes (ESR1, EPHX1, and GSTT2) in the recessive model were found to be associated with habitual smoking in a statistically significant manner (Table 2). Similarly, ten SNPs of five genes (HSD17B3, COMT, ADH1C, ALDH2, and NOS3) in the dominant model and five SNPs of four genes (CYP1B1, ESR1, GSTM3, and NUDT1) in the recessive model were found to be associated with habitual drinking in a statistically significant manner (Table 3). No relationship with smoking and drinking behavior was observed for 32 genes and 31 genes, respectively.

Table 1 List of 153 SNPs in 40 genes selected for association study
Full size table
Table 2 Odds ratios (OR) and 95% confidence intervals (CI) of SNPs associated with smoking behavior. Allele A represents major allele; allele a represents minor allele
Full size table
Table 3 Odds ratios (OR) and 95% confidence intervals (CI) of SNPs associated with drinking behavior. Allele A represents major allele; allele a represents minor allele
Full size table

Next, a multivariate analysis was performed using variables including drinking status and the genotype of 14 SNPs listed in Table 2. The odds ratios of five genes for smoking, selected by stepwise logistic regression analysis, are given in Table 4. While a negative association was found in the dominant model between habitual smoking and the minor allele of SNPs in the OGG1 and CYP17A1 genes, a negative association was found in the recessive model for the minor allele of the SNP in EPHX1. In ESR1 and SLC6A4, minor alleles of SNPs were related to an increased risk for habitual smoking in the recessive and dominant models, respectively.

Table 4 Adjusted odds ratios (OR) and 95% confidence interval (95% CI) of smoking status according to genotype of genes selected by stepwise logistic analysis. Allele A represents major allele; allele a represents minor allele
Full size table

Dividing nonsmokers into never and past smokers indicated that the genotype frequencies containing the minor allele of the SNP in OGG1 (Cys) were similar among current and past smokers but much higher among never smokers. While past smokers have a higher genotype frequency containing the minor allele of the SNP in CYP17A1 than do current smokers, a lower frequency of the genotype containing the minor allele of the SNP in ESR1 was observed among past smokers compared with current smokers.

When the covariates, including smoking status and genes presented in Table 3, were selected using stepwise logistic regression analysis, polymorphisms in the ALDH2 and NUDT1 genes were chosen as the only variants significantly associated with alcohol drinking in the dominant and recessive heritable models for the minor alleles of their SNPs. We further examined the influence of NUDT1 and ALDH2 on alcohol-induced response. While NUDT1 had no association with alcohol-induced flushing, the odds ratio of having the Lys allele of ALDH2 was as high as 33 for alcohol-induced flushing (data not shown).

Discussion

Among 153 SNPs in the 40 genes, we found that SNPs in five genes (EPHX2, ESR1, hOGG1, SLC6A4, and CYP17A1) were related to habitual smoking and SNPs in two genes (ALDH2 and NUDT1) with habitual drinking. No common allele responsible for both habits was found among the 40 genes investigated in the present study. We also found that 32 genes were independent of smoking and 31 genes were independent of habitual alcohol consumption. However, due to the limitation of sample size, the statistical power was generally not large enough to detect an association that might actually exist for some genes. Further study is required to confirm the results.

A previous Japanese study noted that polymorphism of the SLC6A4 gene in the 5’-flanking region influences smoking behavior. Individuals with the homozygous S (a 44-bp deletion) genotype were less inclined to smoke and/or could more easily stop smoking than L allele (a 44-bp insertion) carriers (Ishikawa et al. 1999), although opposite data also exist (Lerman et al. 1998). In the present study, the presence of the minor allele rs717742 SNP of the same gene appears to confer a more than three-fold increased risk for smoking. Moreover, a significantly increased frequency of the minor allele of the SNP in current smokers compared with that in past smokers suggests that individuals with the minor allele may experience difficulty quitting. The location of the SNP of the SLC6A4 gene in the present study is about 10 kb to the polymorphism in the 5’-flanking region of previous studies. The possibility may exist that the two loci are in linkage disequilibrium.

Previous studies indicated that polymorphisms of the genes in the dopaminergic system are candidate genetic markers for habitual smoking; a relationship between the Taq I A or Taq I B RFLP polymorphisms of DRD2 and a predisposition to smoking was suggested, although the results were controversial (Wu et al. 2000; Yoshida et al. 2001). In addition, the DRD4 VNTR polymorphism was noted as another candidate marker for habitual smoking (Shields et al. 1998). No relationship between the selected SNPs of DRD2, DRD3, or DRD4 and habitual smoking was found in the present studies; albeit, the negative findings may narrow the research field to target SNPs in the dopaminergic system for future studies searching for genetic determinants of habitual smoking.

We found that the OGG1 Ser/Cys or Cys/Cys genotypes were associated with a decreased risk for smoking compared with homozygous Ser carriers. However, the OGG1 Cys allele polymorphism is considered to be related to an increased risk for lung cancer and other smoking-related cancers (Goode et al. 2002; Le Marchand et al. 2002). Since smoking is the most potent carcinogenic factor in lung cancer, the negative association of Cys carriers with habitual smoking and the positive association of Cys carriers with lung cancer risk seem controversial. However, a similar paradoxical relationship has been observed between ALDH2 polymorphism, alcohol consumption, and liver cancer incidence. The ALDH2 Lys allele has been related to an elevated risk for liver cancer, and Lys carriers lacking ALDH2 isozyme activity ought to reduce alcohol consumption, a well-known factor in hepatocellular carcinoma (Eriksson 2001; Yokoyama and Omori 2001). However, when those individuals with the Lys allele do drink alcohol habitually, they are at much higher risk for different forms of cancer of the digestive tract, liver, and upper respiratory tract than those without the allele (Munaka et al. 2003; Yokoyama and Omori 2001). The Cys allele may preclude the use of tobacco but may exacerbate the impact of smoking on smoking-related cancer development.

Many previous studies pointed out that the single base mutation from glutamic acid (glutamate) to lysine at residue 504 in ALDH2 was the best-characterized genetic factor influencing alcohol consumption behavior (Goedde et al. 1992; Muramatsu et al. 1995). The findings of the present study are in agreement with the original studies showing Lys allele carriers suffer the alcohol-flush reaction.

None of the previous studies examined whether the EPHX2, ESR1, and CYP17A1 genes contribute to smoking behavior or the NUDT1 gene to alcohol use. We observed an increased frequency of the genotype homozygous for the minor allele of the rs1913474 SNP in ESR1, as well as a decreased frequency of the minor allele of rs743572 SNP in CYP17A1, among current smokers compared with past smokers. The findings may suggest involvement of the ESR1 and/or the CYP17A1 SNP in smoking, as well as in difficulty in quitting, among homozygous minor allele carriers of the ESR1 SNP and/or major allele carriers of the CYP17A1 SNP. An American study indicated that the ESR1 gene might play a role in anxiety (Comings et al. 1999). Anxiety, perhaps, drives individuals to use tobacco for comfort.

The present population-based study using a random Japanese sample screened genes related to habitual smoking and drinking. The findings provided basic but essential information on the association between candidate genes and smoking or drinking behavior for future case-only study, which could determine the interaction between genetics and behavior. The findings also confirmed that ALDH2 gene polymorphism is associated with habitual alcohol consumption, suggesting for the first time that polymorphism in the OGG1, SLC6A4, EPHX1, ESR1, CYP17A1, and NUDT1 genes may influence tobacco and/or alcohol use among Japanese men. Allelic association is a powerful method for detecting genetic influence on complex traits such as cigarette smoking and drinking; yet it cannot formally identify the functional polymorphisms responsible for the phenotype because the association analysis is based on the linkage disequilibrium between the marker SNPs and the causative genetic variation. Further studies are needed to confirm our findings, as well as to expose the underlying molecular mechanism.