Single nucleotide polymorphisms of ADH1B, ADH1C and ALDH2 genes and esophageal cancer: A population-based case–control study in China

Authors


Abstract

Alcohol drinking is a major risk factor for esophageal cancer (EC) and the metabolism of ethanol has been suggested to play an important role in esophageal carcinogenesis. Epidemiologic studies, including genomewide association studies (GWAS), have identified single nucleotide polymorphisms (SNPs) in alcohol dehydrogenases (ADHs) and aldehyde dehydrogenases (ALDHs) to be associated with EC. Using a population-based case–control study with 858 EC cases and 1,081 controls conducted in Jiangsu Province, China, we aimed to provide further information on the association of ADH1B (rs1229984), ADH1C (rs698) and ALDH2 (rs671) polymorphisms with EC in a Chinese population. Results showed that ADH1B (rs1229984) was associated with EC with odds ratios (ORs) of 1.34 [95% confidence interval (CI): 1.08–1.66] for G-allele carriers compared to A/A homozygotes. No heterogeneity was detected on this association across different strata of alcohol drinking and tobacco smoking. Statistical interaction between ALDH2 (rs671) and alcohol drinking on EC susceptibility in both additive and multiplicative scales was observed. Compared to G/G homozygotes, A-allele carriers were positively associated with EC among moderate/heavy drinkers (OR = 1.64, 95% CI: 1.12–2.40) and inversely associated with EC among never/light drinks (OR = 0.75, 95% CI: 0.54–1.03). In addition, statistical interaction between ALDH2 and ADH1B polymorphisms on EC susceptibility among never/light drinkers was indicated. We did not observe association of ADH1C polymorphism with EC. In conclusion, our findings indicated that ADH1B (rs1229984) was associated with EC independent of alcohol drinking and tobacco smoking status and alcohol drinking interacted with ALDH2 (rs671) on EC susceptibility in this high-risk Chinese population.

Alcohol consumption has been established as a major risk factor for esophageal cancer (EC), which remains one of the most common and fatal malignancies worldwide.1, 2 Around 26% of deaths from EC could be attributed to alcohol use with attributable fractions ranging from 24% in low- and middle-income countries to 41% in high-income countries.3 Although the biological mechanisms underlying alcohol-induced carcinogenesis have not been fully understood, the metabolism of ethanol has been suggested to play an important role.4, 5 In alcohol metabolism, alcohol dehydrogenases (ADHs) oxidize alcohol to acetaldehyde, which has been classified as a Group I human carcinogen by the International Agency for Research on Cancer (IARC).6 When further oxidized, acetaldehyde produces less toxic acetic acid by aldehyde dehydrogenases (ALDHs).7

Single nucleotide polymorphisms (SNPs) of ADH- and ALDH-related genes can lead to structural and functional changes of the enzymes which would influence acetaldehyde levels and may predispose people to cancers.8, 9 Among them, three functional SNPs, rs1229984 in ADH1B, rs698 in ADH1C and rs671 in ALDH2 have been frequently studied on their roles in alcoholism and carcinogenesis.8, 9 The ADH1B (rs1229984) A/A homodimer has been found to have a 40-fold higher enzyme activity than the G/G form.10 Enzymes encoded by ADH1C (rs698) A allele have been shown to have a 2.5-times higher capacity oxidizing ethanol compared to those encoded by the G allele.10 The ALDH2 rs671 A allele encoded an inactive subunit with restrained ability to metabolize acetaldehyde. Blood acetaldehyde concentrations after consuming alcoholic beverages in individuals carrying ALDH2 A/A and A/G genotype were 19- and 6-fold higher, respectively, than in those with the G/G genotype.11

Epidemiologic studies, including genomewide association studies (GWAS), have associated genetic variations in ADHs and ALDHs with EC susceptibility.12–16 However, most studies had relatively small sample sizes which can suffer from limited statistical precision to detect interactions. In addition, few studies have investigated ADH1C and EC association among Asian populations. The primary aim of this large case–control study was to replicate the associations between EC and genetic polymorphisms of ADH1B (rs1229984), ADH1C (rs698) and ALDH2 (rs671) in a Chinese population. Joint effects and gene–gene and gene–environment interactions with alcohol consumption and tobacco smoking on EC susceptibility were also evaluated.

Material and Methods

Study population

Study design has been previously described in detail.17, 18 In brief, this population-based case–control study was conducted from 2003 to 2007 in two counties, Dafeng and Ganyu, in Jiangsu province, one of the areas with the highest EC mortality in South East China.19 The annual average age-standardized incidence of EC was 36 and 24 per 100,000 in Dafeng and Ganyu during 2006–2008, respectively.

Eligible subjects were restricted to local residents who have lived in the study area for at least 5 years. Newly diagnosed primary EC patients were recruited as cases, using the information from local population-based cancer registries. From 2003 to 2007, 68% and 75% of eligible cases were recruited and interviewed in Dafeng and Ganyu, respectively. Because of the low proportion of histologically confirmed cases in rural areas (39%), patients who were diagnosed by endoscopic examination (40%) or radiology (11%) were also included. Controls were randomly selected from the same county as cases in the county demographic database. Cases and controls were frequency matched by gender and age (±5 years). The response rate of controls was 87% in Dafeng and 85% in Ganyu.

This study was approved by the Institutional Review Board of Jiangsu Provincial Health Department. With written informed consent, epidemiological data were obtained by face-to-face interviews using a standardized questionnaire. The questionnaire collected information on demographic characteristics, socioeconomic status, living environment, smoking history, alcohol consumption and dietary history. A 5-ml nonfasting blood sample was collected during interview for both cases and controls.

Laboratory analysis

DNA was isolated from blood clots using phenol–chloroform method. SNPs were genotyped using Applied Biosystems (ABI) Taqman platform (Foster city, CA) as previously described.20 Genotype detection was performed on an ABI 7900HT sequence detection system with SDS2.3 software. Around 10% of the samples were randomly repeated for quality control. Call rates were above 95% and reproducibility was observed at 99.3%.

Statistical analysis

Data were entered into an Epidata 3.0 (EpiData Association, Odense, Denmark) database and cleaned and analyzed using SAS v9.1 (SAS Institute, Cary, NC). Ever smokers were defined as those who have smoked for more than 100 cigarettes in their lifetime. Ever alcohol drinkers were defined as those who drank at least once per month. Average weekly consumption of ethanol (milliliters) was converted from weekly intake of six mostly consumed type-specific alcoholic beverages in Jiangsu area (high degree liquor, low degree liquor, beer, wheat liquor, rice liquor and wine) according to average frequency and amount of drinking. We used median levels in the control group by gender to impute for 53 (4.4%) alcohol drinkers with missing values on weekly ethanol intake and 185 smokers (14%) with missing values on pack-years of smoking.

We used Pearson χ2 test and student's t-test to compare difference of distributions of selected demographic factors among cases and controls. Unconditional logistic regression models were applied for estimating the associations with odds ratios (ORs) and 95% confidence intervals (CIs). Potential confounders were selected based on prior knowledge, including age, sex (male/female), education level (illiteracy, primary school, middle school and above), previous income (continuous), body mass index (BMI, continuous), smoking pack-years (continuous), family history of EC (any malignancy in first-degree relatives) and study site (Ganyu, Dafeng). To minimize age confounding and to account for age matching, we used fine categories of age (under 50, 50–51, 52–53, 54–55, 56–57, 58–59, 60–61, 62–63, 64–65, 66–67, 68–69, 70–71, 72–73, 74–75, 76–77, 78–79, 80 and over) in the adjusted models as suggested by previous study.21

Effect modifications were evaluated by stratified analyses. Gene–environment and gene–gene interactions were assessed at both additive and multiplicative scales. The stratum with the lowest risk in joint effect models was used as the reference category in interaction analyses as suggested by Knol et al.22 Multiplicative interaction was assessed by including both the main effect variables and their product terms in the logistic regression models. Three additive interaction measurements suggested by Knol et al.,23 relative excess risk due to interaction (RERI), attributable proportion due to interaction (AP) and synergy index (SI) were calculated. The 95% CI of RERI, AP and SI were estimated by the delta method.24, 25 In the absence of an additive interaction, RERI and AP amount to 0 and SI amounts to 1.

Abbreviations

ABI: Applied Biosystems; ADHs: alcohol dehydrogenases; ALDHs: aldehyde dehydrogenases; AP: attributable proportion due to interaction; BMI: body mass index; CIs: confidence intervals; EC: esophageal cancer; ESCC: esophageal squamous cell carcinoma; GWAS: genome-wide association studies; IARC: International Agency for Research on Cancer; LD: linkage disequilibrium; ORs: odds ratios; RERI: relative excess risk due to interaction; SI: synergy index; SNPs: single-nucleotide polymorphisms; UADT: upper aerodigestive tract

Results

From 2003 to 2007, 1,520 cases and 1,683 controls were recruited in this study. However, because the quality of DNA samples was greatly improved after 2004, genotyping was only performed among those recruited after 2004. We did not observe difference between those who recruited before and after 2004 on basic demographic characteristics. A total of 846 EC cases and 1,079 controls were included in this analysis. Compared to population controls, cases had lower levels of education, previous income and BMI (Table 1). More cases were males, smokers and had family history of EC than controls.

Table 1. Distributions of selected demographic characteristics among cases and controls1
inline image

Ever alcohol drinking was associated with increased risk of EC with OR of 1.43 (95% CI: 1.12–1.84), after adjusting for potential confounders (Table 2). Positive dose-response relationships were observed with increased frequency and amount of alcohol drinking (P for trend <0.001). Compared to never alcohol drinkers, the ORs for consuming ethanol for 250–500 ml/week and for at least 500 ml/week were 1.62 (95% CI: 1.16–2.26) and 1.72 (95% CI: 1.28–2.32), respectively. We found similar results using imputed weekly ethanol consumption.

Table 2. Association between alcohol drinking and the risk of EC
inline image

Genotype distributions of ADH1B (rs1229984), ADH1C (rs698) and ALDH2 (rs671) among controls were all in agreement with Hardy-Weinberg equilibrium (p > 0.05). After adjusting for potential confounders, the inactive ADH1B (rs1229984) G-allele was associated with EC with ORs of 1.19 (95% CI: 0.94–1.51) for A/G heterozygotes and 1.88 (95% CI: 1.34–2.64) for G/G homozygotes, as compared to A/A homozygotes (Table 3). The OR was 1.34 (95% CI: 1.08–1.66) in dominant model. We did not observe clear association of ADH1C (rs698) and ALDH2 (rs671) with EC susceptibility.

Table 3. Distribution of ADH1B, ADH1C and ALDH2 polymorphisms and their associations with EC
inline image

ADH1B (rs1229984) G-allele carriers had consistent 30% increased odds of having EC compared to A/A homozygotes across different strata of alcohol drinking and tobacco smoking (Tables 4 and 5). In join-effect analysis, the highest odds were observed among moderate/heavy drinkers (weekly ethanol intake of 250 ml or more) with the G/G genotype (OR = 3.58, 95% CI: 2.20–5.84) as compared to never/light drinkers (weekly ethanol intake of less than 250 ml) with the A/A genotype, and among ever smokers with the G/G genotype (OR = 3.62, 95% CI: 2.23–5.87) as compared to never smokers with the A/A genotype. ALDH2 (rs671) A-allele carriers were associated with increased odds of EC among moderate/heavy drinkers and reduced odds of EC among never/light drinkers, while compared to G/G homozygotes. Statistical interaction was detected between ALDH2 (rs671) and alcohol drinking on EC susceptibility in both additive and multiplicative scales. Moderate/heavy drinkers with the ALDH2 A/G genotype had the highest odds of EC (OR = 2.34, 95% CI: 1.52–3.61) in joint-effect analysis, as compared to never/light drinkers with the G/G genotype.

Table 4. Joint effects between ADH1B, ADH1C and ALDH2 polymorphisms and alcohol drinking on EC
inline image
Table 5. Joint effects between ADH1B, ADH1C and ALDH2 polymorphisms and tobacco smoking on EC
inline image

Although gene–gene interaction was not detected, among moderate/heavy drinkers, the joint effect of ALDH2 and ADHs polymorphisms showed that the highest odds of EC was observed among those carrying ALDH2 A and ADHs G alleles (OR = 2.35, 95% CI: 1.40–3.94 for ADH1B; OR = 1.96, 95% CI: 0.94–4.09 for ADH1C), as compared to those with ALDH2 G/G and ADHs A/A genotype (Table 6). Among never/light drinkers, statistical interaction between ALDH2 and ADH1B on EC susceptibility was detected in both additive and multiplicative scales.

Table 6. Joint effects between ALDH2 and ADHs polymorphisms on EC, stratified on alcohol drinking status
inline image

Discussion

In this population-based case–control study among Chinese population, we reported a consistent association between ADH1B (rs1229984) and EC across different strata of alcohol drinking and tobacco smoking status. We observed statistical interaction between alcohol drinking and ALDH2 (rs671) on EC susceptibility, with A allele associated with increased odds of EC among moderate/heavy drinkers and decreased odds of EC among never/light drinkers. In join-effect analysis of ALDH2 (rs671) and ADHs among moderate/heavy drinkers, the highest odds of EC were observed among those carrying ALDH2 A allele and ADHs G allele. Among never/light drinkers, we detected statistical interaction between ALDH2 and ADH1B polymorphisms on EC susceptibility.

Our results on ADH1B (rs1229984) were in accordance with previous studies.12, 14, 16 In a meta-analysis across Asian populations, the ORs for those with the A/G and G/G genotype compared to the A/A genotype were 1.60 (95% CI: 1.25–2.00) and 2.17 (95% CI: 1.08–4.34), respectively.16 Recent GWAS have also reported ADH1B (rs1229984) to be associated with EC with ORs of 1.79 (95% CI: 1.69–1.88) for the G allele in Japanese12 and 0.38 (95% CI: 0.24–0.59) for the A allele in European populations.14 Several reasons could contribute to the excess risk of the G allele. First, in contrast with the less active G allele, the fast-metabolizing A allele may prevent people from heavy drinking because of higher acetaldehyde concentration after drinking which results in ethanol intolerance at low doses. Several studies have reported that G allele was associated with increased intensity of alcohol drinking.26, 27 However, we consistently observed the association across different alcohol drinking strata. Second, G allele carriers may experience longer exposure time to both ethanol and acetaldehyde after drinking than A allele carriers. Yokoyama et al. have demonstrated that salivary and blood ethanol and acetaldehyde levels were higher in G allele carriers than those carry the A allele.28 Increased salivary acetaldehyde production could result from oral microorganism overgrowth due to prolonged ethanol exposure resulted from less ADH1B activity.

The inactive ALDH2 (rs671) A allele is rare in Western populations, but is highly prevalent and mostly studied among Eastern Asians on its association with cancer.8, 12, 13, 15, 16, 29 In agreement with most studies, we detected statistical interaction between ALDH2 and alcohol drinking on EC susceptibility. We observed that while compared to G/G homozygotes, A allele carriers were associated with increased odds of EC among moderate/heavy drinkers, and they were associated with decreased odds of EC among never/light drinkers. A Chinese GWA indicated multiplicative interaction between alcohol drinking and rs11066015 of ACAD10 [in high linkage disequilibrium (LD) with rs671, r2 = 0.79] on esophageal squamous cell carcinoma (ESCC) risk, with more pronounced risk seen in drinkers (interaction P = 4.54 × 10−34).15 In a meta-analysis including 18 studies,16 increased EC risk was found only among moderate/heavy drinkers, but not among never/rare drinkers. The increased risk of A-allele carriers among moderate/heavy drinkers was biologically relevant, indicating the harmful effect of accumulated acetaldehyde after alcohol drinking.9 The reduced risk among never/light drinkers, however, were in agreement with some studies30, 31 and in disagreement with some others32–34 and warrants further investigation. A Japanese GWA study suggested reduced risk of EC for A/A homozygotes compared to G/G homozygotes (OR = 0.47, 95% CI: 0.28–0.78).12 Prevention of alcohol drinking among A/A homozygotes because of severe alcohol flush responses has been proposed as one of the mechanisms for the risk reduction.12 In this study, we observed similar inverse association for A/A homozygotes. However, the small sample size of the A/A homozygotes make the effect estimates vulnerable to shift and further elucidation is needed.

Alcohol drinking could mediate the association of ADHs and ALDHs SNPs with EC and we found that subjects with the fast genotype of ADH1B (A/A) and the slow genotype of ALDH2 (A/A) drank less, even within strata of alcohol drinking intensity (Supporting Information Tables S1 and S2). To examine whether effects of SNPs on EC are mediated through alcohol drinking, we further adjusted on weekly ethanol intake for the main- and stratified-association for ADH1B (rs1229984) and ALDH2 (rs671) on EC and did not find much difference of the observed associations either with or without the adjustments (data not shown). Furthermore, we found that both polymorphisms were associated with EC among never alcohol drinkers (ORA/G+G/G vs. A/A = 1.41, 95% CI: 1.00–2.01 for ADH1B; ORA/G+A/A vs. G/G = 0.69, 95% CI: 0.48–1.00 for ALDH2; data not shown), which suggested that ADH1B and ALDH2 may be associated with EC through multiple pathways in addition to alcohol drinking.

In join-effect analysis of ALDH2 (rs671) and ADHs among moderate/heavy drinkers, the highest odds of EC were found on those carrying ALDH2 A allele and ADHs G allele. Similar associations have been reported by several studies.12, 16, 31, 32, 34–37 In a meta-analysis, the highest risk of EC was observed among heavy drinkers with ADH1B G/G and ALDH2 A/G genotype (OR = 12.45, 95% CI: 2.9–53.46), as compared to those with ADH1B any A and ALDH2 G/G genotype.16 In a Japanese GWA study, individuals with ADH1B G/G and ALDH2 A/G genotype had a remarkably higher risk (OR = 16.17, 95% CI: 11.55–22.65) than those with ADH1B any A and ALDH2 A/A or G/G genotype.12 Interestingly, statistical interaction of ALDH2 and ADH1B on EC was indicated among never/light drinkers in our study, with the highest odds observed among those with ALDH2 G/G and ADH1B any G genotype. This association may need further investigation.

We did not observe association of ADH1C (rs698) with EC in this study. Different from ADH1B and ALDH2, ADH1C polymorphism is the rate-limiting factor in alcohol metabolism among Western populations and studies from European origins have associated ADH1C polymorphism with EC.14, 38 ADH1B and ADH1C genes are closely located in the short arm of chromosome 4, and strong LD (D′ > 0.65) has been reported by previous studies including Asians.27, 38–41 The role of ADH1C in esophageal carcinogenesis independent of ADH1B has been observed to be controversial. A Japanese study reported the association between ADH1C and EC disappeared after the adjustment on ALDH2 and ADH1B genotypes in multiple logistic models.42 However, a study in Europe indicated that ADH1B and ADH1C had independent association with upper aerodigestive tract (UADT) cancers, despite of their strong LD.38 LD between ADH1B (rs1229984) and ADH1C (rs698) in our study was minor (r2 = 0.16, D′ = 0.41) and could possibly explain the lack of association between ADH1C and EC. Results on ADH1C polymorphism and EC remain sparse and inconsistent, and need to be further elucidated.

There are several limitations in this present analysis. First, we had missing information on weekly ethanol intake and pack-years of smoking. However, instead of using medians in controls for missing imputation, we also performed multiple imputations in SAS with the Proc MI and the Proc Mianalyze procedures and the results are essentially the same with both methods. Second, although the questionnaire had been tested in previous studies, the self-reported alcohol drinking behaviors may be vulnerable to subjective judgment and recall bias could cause exposure misclassification. However, the strength of the associations for EC with alcohol consumption, particularly the dose-response trend indicates good validity and sensitivity of our study. Third, cases were in mixed histology in this study because of the low proportion of pathological examinations in less developed rural areas. However, previous reports have indicated that more than 95% of EC in China are ESCC.43 And last, only subjects recruited after 2004 were involved in this study. However, we did not find major differences on basic characteristics between this study population and the complete population.

In conclusion, ADH1B (rs1229984) polymorphism was associated with EC in this high-risk Chinese population. Gene–environment interaction between alcohol drinking and ALDH2 (rs671) polymorphism on EC susceptibility was observed. Moderate/heavy drinkers carrying ALDH2 A allele and ADHs G allele had the highest risk of EC. Genetic predispositions, together with lifestyle factors may ultimately determine individual's risk of EC.

Acknowledgements

The authors thank the study participants for their voluntary participation, as well as the staff of local Health Bureaus and CDCs in Dafeng and Ganyu for their assistance in fieldwork.

Ancillary