Alcohol Drinking Mediates the Association between Polymorphisms of ADH1B and ALDH2 and Hepatitis B–Related Hepatocellular Carcinoma

Discussion

Although previous studies have at times examined the statistical interaction between alcohol drinking, and SNPs on ADH1B and ALDH2, they did not examine the mechanistic coordination of genetic factors and alcohol consumption on hepatocellular carcinoma risk. We utilize mediation analyses to investigate the indirect and direct effects of these SNPs on hepatocellular carcinoma risk with respect to alcohol drinking. To our knowledge, this is the largest study to examine the relationship between the SNPs rs1229984 and rs671, alcohol drinking, and future risk for hepatocellular carcinoma among individuals chronically infected with HBV. Whereas other studies have focused on simple adjustment for confounding factors, this study instead analyzed mediation effects using a mechanistic approach, which enables the examination of both direct and indirect effects, of these SNPs, while also accounting for SNP-by-alcohol interaction.

This study has two main findings. First, this study confirmed the protective effect of rs1229984 and rs671 against alcohol drinking among a large cohort of individuals with chronic HBV in Taiwan. Interestingly, we found that only individuals who carried at least one deficient allele for both polymorphisms were significantly protected against alcohol drinking, with a 4-fold decrease in risk.

Second, this study also found that SNPs on ADH1B and ALDH2 had significant effects on hepatocellular carcinoma risk, mediated through habitual alcohol drinking. Traditional multivariate analyses in this study found significant association between alcohol drinking and hepatocellular carcinoma, with a greater than 2.5-fold OR, but did not show any direct association between hepatocellular carcinoma and the SNPs rs1229984 and rs671. Past studies examining the association between alcohol drinking and ADH1B/ALDH2 also did not find significantly increased risk for hepatocellular carcinomaamong drinkers with chronic liver disease or chronic HBV who carried at least one polymorphism (25–27).

However, this does not mean that these SNPs did not play a significant role in hepatocarcinogenesis, as mediation effects were not considered. The seemingly null overall effect of ALDH2 is the consequence that direct and indirect effects mask each other. Further mediation analyses showed that carrying deficient alleles for both of these polymorphisms did in fact have a strong association with hepatocellular carcinoma risk and that this effect was highly mediated through the environmental risk factor of habitual alcohol consumption. The SNPs rs1229984 and rs671 were protective against the development of hepatocellular carcinoma through their protective effect against habitual alcohol drinking (indirect effect). Put in another perspective, it can be said that carriers of normal genotypes for these two SNPs are more likely to drink alcohol, thus increasing their risk for hepatocellular carcinoma. When examined separately, although ADH1B (rs1229984) did not have any significant direct effect on hepatocellular carcinoma incidence, ALDH2 (rs671) did have a marginally significant direct effect on increasing hepatocellular carcinoma risk, independent of habitual alcohol drinking (Table 3). It is interesting to note that we also did not see any significant marginal effect of the two SNPs, as their effect was largely masked by the downstream heterogeneity of habitual alcohol drinking. The interesting findings revealed by mediation analyses demonstrate its advantage compared with the conventional main effect or statistical interaction approach. Another example of effect cancellation by a positive direct effect and a negative indirect effect can be found in the study by Huang and colleagues (2015; ref. 32).

In addition, we did not find any significant SNP-by-alcohol interaction, meaning that the effects seen did not vary according to alcohol drinking habit. Had there been significant interaction, it would suggest that the SNPs perhaps also altered the susceptibility of patients to the effects of alcohol. Such gene–environment interactions have been seen in other studies of mediation effects between genetic factors, smoking, and lung cancer (35). However, this was not seen in the current study; it can be inferred that rs1229984 and rs671 significantly affects alcohol consumption behavior itself, rather than susceptibility to the effects of alcohol, which in turn subsequently significantly reduces risk for developing hepatocellular carcinoma. However, future studies should examine whether this is mechanistically the case.

In addition, the marginal increase in hepatocellular carcinoma risk that was seen for carriers of the rs671-deficient allele is consistent with previous studies, which suggested that increased acetaldehyde accumulation among drinkers who carry these SNPs can lead to increased risk for certain cancers, including liver cancer (14,23). However, the true effect of ALDH2 on hepatocellular carcinoma risk remains to be clarified. It is known that other sources of acetaldehyde exposure exist, including cigarette smoke and indoor/outdoor air. Clarification of the direct effects of ALDH2 would require a cohort of nondrinkers who are exposed to these secondary sources. However, this was not possible due to a lack of sample size, and future studies should investigate other acetaldehyde sources as well.

The results of this study also emphasize the importance of alcohol consumption as a nongenetic risk factor for hepatocellular carcinoma. As seen in Table 2, habitual alcohol consumption still confers significantly higher risk for hepatocellular carcinoma, again confirming the significant role of alcohol consumption, even after adjustment for other well-known risk factors.

Unfortunately, despite strong evidence supporting alcohol as an environmental risk factor for hepatocellular carcinoma, the exact pathways by which alcohol causes hepatocellular carcinoma are still unknown. Many mechanisms of alcohol-induced hepatocarcinogenesis have been suggested. One commonly hypothesized pathway is through oxidative stress; the metabolism of ethanol by ADH leads to an accumulation of acetaldehyde, as well as the production of free radicals (36). Acetaldehyde has been shown to affect DNA replication and repair mechanisms. In addition, consuming large amounts of ethanol induces microsomal ethanol metabolism by CYP2E1, which leads to additional production of acetaldehyde, as well as an increase in free radicals that can lead to cell death, DNA damage, and even production of other carcinogenic substances (13,14,36–38).

Other hypothesized pathways have included the decreased DNA methylation of tumor promoter genes (39,40). It has also been suggested that alcohol inhibits natural killer (NK) immune surveillance; studies have shown decreased numbers of NK cells in humans with alcoholic cirrhosis, and alcohol administration has also caused increases in lung cancer metastasis in NK cell–controlled cancer, implying that alcohol use can affect the ability of the body to detect and eliminate early cancer (41).

Finally, strong synergistic interaction also exists between alcohol consumption and chronic hepatitis virus infection. In one study, compared with nondrinkers uninfected with chronic hepatitis, heavy drinkers with chronic hepatitis had an adjusted OR (95% CI) of 53.9 (7.0–415.7) for developing hepatocellular carcinoma, compared with 2.4 (1.3–4.4) and 19.1 (4.1–89.1) for alcohol drinking alone and chronic hepatitis alone, respectively (42). These results suggest a possible common pathway for hepatocarcinogenesis, through both cirrhosis and/or oxidative stress, although as discussed above, the exact pathways are still unclear and need to be elucidated.

This study also highlights the importance of taking mediation effects into consideration in future genetic studies. As seen in this study, traditional analyses showed no significant effect of rs1229984 and rs671 on hepatocellular carcinoma risk. However, this study has shown that results of these direct analyses can be misleading if mediation effects are not considered. Moreover, in diseases such as hepatocellular carcinoma that have well-established nongenetic risk factors, future studies should focus more carefully on biologically plausible and mechanistic models that take into account gene–environment interaction, rather than on purely GWAS-based studies, which would have missed the SNPs discussed in this study entirely. As these SNPs did not have a significant direct effect on hepatocellular carcinoma, traditional GWAS studies would be unable to detect them, and, in addition, would have been unable to determine that the joint effect of both genes was needed to see significant mediation effects. Future genetic studies must also consider the possibility that certain genes may act to determine environmental exposure instead of the disease itself and should also be sure to take variant-by-mediator interaction into effect, so as to clearly delineate the different effects of gene–environment interaction as true interaction versus mediation.

For causal inference, mediation analyses assume that after adjustment for covariates, there is no confounding of the association between (i) SNPs and hepatocellular carcinoma, (ii) alcohol drinking and hepatocellular carcinoma, (iii) SNPs and alcohol drinking, and (iv) no common cause of alcohol drinking and hepatocellular carcinoma that are affected by the SNPs (33). Because PCAs showed no population stratification within the current cohort (Supplementary Fig. S1), assumptions (i) and (ii) hold. Assumption (ii) has been shown to be true in previous studies (43), and assumption (iv) holds based on specific functions of these two SNPs as discussed above.

This study was conducted in a homogeneous cohort of patients of Han Chinese descent, which has been shown in the 1000 Genomes Project to have much higher population frequencies of the rs1229984 “T” allele and the rs671 “A” allele than other populations. For example, frequencies of the rs1229984 “T” allele were 0%, 6%, and 3% in African, American, and European populations, respectively, compared with 72.9% in the REVEAL-TLCN cohort. More strikingly, reported frequencies of the rs671 “A” allele were 0% for African, American, and European populations compared with 28.8% in the REVEAL-TLCN cohort (44). Therefore, the results of this study would only apply to East Asian populations, and mediation effects of these two SNPs would not be able to be measured in other populations. However, the mediation effects of other reported SNPs affecting alcohol metabolism should be explored in non–East Asian populations.

This study has several limitations. Detailed information on exact drinking volume was not available for all participants, and therefore, drinking was classified as habitual drinking or not, defined as drinking four or more times per week or for a duration of at least one year. We are confident that our definition of drinking adequately separates heavy drinkers from light-to-moderate drinkers. Moreover, alcohol use was self-reported via questionnaire and for the TLCN cohort was reported after diagnosis. Therefore, the definition of alcohol use in this study may be subject to biases such as recall bias, resulting in possible misclassification of alcohol drinking status. Although we also did not have information on other environmental risk factors for alcohol dependence, we did include age and gender. Participants were from 30 to 65 years old and mostly infected perinatally with HBV genotypes B and/or C. These results need to be clarified in individuals infected with other HBV genotypes, such as A and D. Data on some predictors, such as serum HBsAg levels and family history, were not available in the TLCN patients. Therefore, these predictors could not be considered in this analysis.

In conclusion, the association between ADH1B and ALDH2 polymorphisms and hepatocellular carcinoma risk is significantly mediated by habitual alcohol consumption. Therefore, management of patients with chronic HBV should, in addition to focusing on viral suppression, also particularly in patients with wild-type ADH1B and ALDH2 polymorphisms, reduce the harmful effects of important environmental risk factors such as alcohol consumption.