Genetic variation in CYP2A6 and in the nicotinic receptor subunit gene cluster, CHRNA5-CHRNA3-CHRNB4 (CHRNA5-A3-B4), has been modestly associated with lung cancer susceptibility in independent studies (1–5). CYP2A6 and CHRNA5-A3-B4 are both involved in the pharmacology of nicotine and nitrosamines, which are lung cancer procarcinogens (6). CYP2A6 inactivates nicotine and also activates tobacco-specific nitrosamines (1,7), whereas nicotinic receptors mediate nicotine-induced reward (8) and nitrosamine-induced carcinogenic signaling (9). Thus, genetic variations in CYP2A6 and CHRNA5-A3-B4 have the potential to influence lung cancer risk directly and indirectly through alteration of smoking behavior (10–13).

We investigated the associations between CYP2A6 and CHRNA5-A3-B4 polymorphisms, alone and in combination, with smoking behaviors and lung cancer risk. Ever-smoking non–small cell lung cancer patients (n = 417) and control subjects (n = 443) of European ancestry matched by age, sex, and smoking variables were selected from our previously published genome-wide association discovery set (4) (Supplementary Table 1, available online). Participants were genotyped for CYP2A6*2, *4, *9, and *12 (14,15), reduced enzymatic function alleles common in those of European ancestry, and grouped by predicted metabolic activity. CYP2A6 carriers were defined as either CYP2A6 normal metabolizers (participants who lacked variant alleles) or CYP2A6 reduced metabolizers (participants who carried at least one variant allele) (14,16).

Participants were also grouped according to genetic variation in the CHRNA5-A3-B4 cluster, represented by the tag single-nucleotide polymorphism rs1051730 G>A (4). The AA genotype is the established risk genotype for smoking intensity and lung cancer risk (4,5,17), hence rs1051730 GG and GA individuals were grouped together. Individual allele and combined group genotype results are available in Supplementary Table 2 (available online). We used the Kruskal–Wallis test to compare smoking variables among genotype groups for all participants and separately in control subjects who were current smokers. Lung cancer odds ratios (ORs) were estimated using logistic regression analysis. This study was approved by Review Boards at the M.D. Anderson Cancer Center (Houston, TX), the Institute for Cancer Research Foundation (London, UK), and the University of Toronto (Toronto, ON, Canada).

Daily cigarette consumption was statistically significantly associated with the CYP2A6 and CHRNA5-A3-B4 genotypes, alone and in combination, in the overall study population (Supplementary Table 3, available online). Here, we present the association in control subjects who were current smokers (n = 209) (Figure 1, A), because cancer diagnosis and recall bias among former smokers are minimized as confounders in this analysis. In control subjects who were current smokers, CYP2A6 normal metabolizers smoked statistically significantly more cigarettes per day (CPD) (mean = 25.9 CPD, 95% CI = 24.1 to 27.8 CPD) compared with reduced metabolizers (mean = 20.2 CPD, 95% CI = 15.6 to 24.8 CPD) (P < .001). The CHRNA5-A3-B4 AA genotype was associated with a statistically significant increase in CPD compared with the CHRNA5-A3-B4 GG and GA genotypes in the overall study population (Supplementary Table 3, available online) but did not reach statistical significance among control subjects who were current smokers (P = .416).

To assess the association of CYP2A6 and CHRNA5-A3-B4 genotypes in combination with smoking behavior, we separated participants into three groups as follows: CYP2A6 reduced metabolizers with the CHRNA5-A3-B4 GG and GA genotypes were identified as the low-risk group, CYP2A6 normal metabolizers with the CHRNA5-A3-B4 AA genotype were identified as the high-risk group, and control subjects with the either the CYP2A6 normal metabolizer genotypes or the CHRNA5-A3-B4 AA genotype were identified as the intermediate-risk group. Among control subjects who were current smokers, cigarette consumption was statistically significantly different (P < .001) between the low-risk group (mean = 20.8 CPD, 95% CI = 15.7 to 25.9 CPD), intermediate-risk group (mean = 25.3 CPD, 95% CI = 23.3 to 27.2 CPD), and high-risk group (mean = 27.9 CPD, 95% CI = 22.6 to 33.2 CPD) and increased linearly across these three genotype groups (P_trend = .042). Similar results were observed for the overall study population (Supplementary Table 3, available online).

We also investigated the relationship between nicotine dependence and CYP2A6 and CHRNA3-A5-B4 genotypes, as assessed by the Fagerström Test for Nicotine Dependence (FTND). FTND scores were statistically significantly associated with the CYP2A6 and CHRNA5-A3-B4 genotypes, alone and in combination, in the overall study population (Supplementary Table 3, available online). In control subjects who were current smokers, statistically significantly higher FTND scores (P = .036) were observed for CYP2A6 normal metabolizers (mean = 5.1, 95% CI = 4.7 to 5.5) compared with reduced metabolizers (mean = 4.2, 95% CI = 3.6 to 4.9). The CHRNA5-A3-B4 AA genotype was associated with a statistically significant increase in FTND scores compared with the CHRNA5-A3-B4 GG and GA genotypes in the overall study population (Supplementary Table 3, available online) but did not reach statistical significance among control subjects who were current smokers (P = .137).

Nicotine dependence was also associated with the CYP2A6 and CHRNA5-A3-B4 genotypes in combination. Among control subjects who were current smokers, FTND scores were statistically significantly different (P = .036) between the low-risk group (mean = 4.3, 95% CI = 3.6 to 5.0), intermediate-risk group (mean = 4.9, 95% CI = 4.5 to 5.3), and high-risk group (mean = 5.9, 95% CI = 4.9 to 6.9) and increased linearly across these three genotype groups (P_trend = .013). Similar results were observed for the overall study population (Supplementary Table 3, available online).

We then investigated the impact of CYP2A6 and CHRNA5-A3-B4 on lung cancer susceptibility (Figure 2, A). Because both CYP2A6 and CHRNA5-A3-B4 were associated with cigarette consumption, we adjusted odds ratios for cigarette pack–years. CYP2A6 normal metabolizers had a non-statistically significant increase in lung cancer risk compared with reduced metabolizers (OR = 1.26, 95% CI = 0.90 to 1.76; P = .180), whereas the CHRNA5-A3-B4 AA genotype was statistically significantly associated with increased lung cancer risk (OR = 1.57, 95% CI = 1.06 to 2.31; P = .024). Of note, the combination of both risk genotypes, CYP2A6 normal metabolizer and CHRNA5-A3-B4 AA, was statistically significantly associated with increased lung cancer risk (OR = 2.03, 95% CI = 1.21 to 3.40; P = .007).

Adjusting the odds ratio for each genotype alone by variation in the other gene did not affect the association between genotype and lung cancer risk suggesting that variation in CYP2A6 and CHRNA5-A3-B4 independently affects lung cancer risk (for unadjusted, pack-year, and genotype-adjusted odds ratios, see Supplementary Table 4, available online). A statistically significant interaction between CYP2A6 and CHRNA5-A3-B4 was not found in a logistic regression model of lung cancer risk, suggesting that the combined genotype effects are additive.

In subgroup analyses, the CYP2A6 normal genotype was statistically significantly associated with lung cancer risk in the lighter-smoking stratum in which CPD was 20 or less, as defined by median CPD in patients and control subjects (OR = 1.60, 95% CI = 1.03 to 2.49; P = .036) (Figure 2, B). A statistically significant association with lung cancer risk was observed among those with both risk genotypes, CYP2A6 normal and CHRNA5-A3-B4 AA (OR = 3.03, 95% CI = 1.38 to 6.66; P = .006), whereas among the heavier-smoking stratum (CPD >20), no association with lung cancer risk was noted for either CYP2A6 or CHRNA5-A3-B4 alone or in combination (Supplementary Table 5, available online). This pattern of higher genetic risk in the lighter-smoking stratum supports the notion of a direct genotype contribution to lung carcinogenesis vs a sole contribution of genotype from altered smoking quantity and has been previously reported for the CHRNA5-A3-B4 genotype (5). Low-exposure gene effects (18) have also been observed for other polymorphic drug-metabolizing enzymes such as CYP1A1, NAT2, and MPO (19–22) and merit attention. Lighter smokers, a growing segment of the smoking population (23,24), tend to have reduced concerns about the negative health effects associated with smoking. Our findings, however, suggest that the genetic risk for lung cancer may remain high among lighter smokers. Furthermore, the levels of smoking may obscure genetic signals in association studies conducted in smoking populations. We also conducted a subgroup analysis by histology (Supplementary Table 6, available online) and found evidence of a stronger association of CYP2A6 and CHRNA5-A3-B4 with adenocarcinoma (OR = 2.09, 95% CI = 1.08 to 4.03, P = .029) vs squamous cell carcinoma (OR = 1.44, 95% CI = 0.65 to 3.20, P = .372).

The major limitation of the current study is the use of self-reported cigarette consumption without biochemical verification, such as plasma cotinine levels or total urinary nicotine equivalents. In addition, neither the smoking behavior analyses nor our lung cancer risk models were able to incorporate potential changes in smoking patterns of study participants over time.

This study demonstrates for the first time, to our knowledge, that genetic variation in CYP2A6 and CHRNA5-A3-B4 combines to increase cigarette consumption and nicotine dependence and independently and additively combines to increase lung cancer risk. Our results also suggest that variation in CYP2A6 appears to have a larger relative role in smoking behaviors, whereas variation in CHRNA5-A3-B4 may play a larger relative role in lung cancer risk. Given that metabolic activation of procarcinogens is an early step along the pathway to cancer (6), altered CYP2A6 nitrosamine activation may make a small contribution to lung cancer risk. The α5, α3, and β4 subunits are expressed in the respiratory tract (25) and have been implicated in the pathological effects of nitrosamines on epithelial cells (26). Thus, genetic variation in these subunits could influence nitrosamine carcinogenic signialing. These findings further our understanding of genetic risk factors for smoking and lung cancer and provide insight into mechanisms of lung cancer carcinogenesis.