The characteristics of the light and heavy smokers included in this study are shown in . While there are roughly equal numbers of men and women among the heavy smokers, only 31% of the light smokers are men. These gender differences in smoking prevalence are consistent with those documented for Americans in the birth cohorts of the study population and are likely due to both cultural and biological factors (19
). The heavy smokers were slightly less educated and likely to consume more alcohol than the light smokers. The percentages that were current smokers were low in both groups (1.5% for light smokers and 11.9% for heavy smokers), likely reflecting their advanced age (smoking prevalence of US adults aged ≥65 is less than 9%) (20
). The characteristics of the individuals excluded from the analysis because of ethnic background or genotyping problems were similar to those included in the study (data not shown).
Frequencies* of selected characteristics of the light and heavy smokers included in this study.
Nineteen SNPs on chromosome 15 in the region that encompasses the CHRNA5-CHRNA3-CHRNB4
gene cluster were strongly associated with heavy smoking ( and ). All of these associations were highly statistically significant, even after conservative Bonferroni correction for multiple comparisons (p-values ranged 0.019-2.8E-07 after correction). In addition, all the associations remained significant for both men and women when gender-specific analyses were done (data not shown). These findings replicated the association for nine of the SNPs that had been identified by a GWAS (7
) and parallel candidate gene study (8
) of nicotine dependence and identified ten new associated SNPs in the same region.
Characteristics, minor allele frequencies, and associations with nicotine dependence for the CHRNA5-CHRNA3-CHRNB4 SNPs
Association of SNPs in the CHRNA5-CHRNA3-CHRNB4 gene cluster with heavy smoking
The minor alleles of eight of the statistically significant SNPs were associated with an increased risk of heavy smoking (, highlighted with a black background). Seven of these SNPs (all but rs1996371) share a high degree of linkage disequilibrium (LD, D' between 0.81 and 1.0, r2 between 0.61 and 0.99). The rs1996371 SNP is also in LD with the other 7 SNPs, although less so (D' between 0.55 and 0.72, r2 between 0.22 and 0.39). The association between rs1996371 and heavy smoking was completely attenuated (per allele OR = 1.02, 95% CI: 0.89, 1.18) when any other of the risk SNPs in this group were included in the logistic regression model. Thus, these eight SNPs represent a single correlated cluster that is very significantly associated with increased risk of heavy smoking.
The minor alleles of the other eleven SNPs with significant p-values in the CHRNA5-CHRNA3-CHRNB4 gene cluster were associated with decreased risk of heavy smoking (, highlighted by gray background). As with the risk SNPs, these protective SNPs share varying levels of LD (D' between 0.64 and 1.0, r2 between 0.39 and 0.98) and represent a single group of correlated SNPs. Although the LD across this region appears high when measured by D', the correlation between the risk and protective SNPS is low (r2≤0.2), indicating that they represent two unrelated associations (see ).
The genotype associations for the risk SNPs in the CHRNA5-CHRNA3-CHRNB4 gene cluster show a dose response [, for rs16969968, GA OR=1.45 (95% CI: 1.24, 1.70), AA OR=1.77 (95% CI: 1.37, 2.29)]. The same is true for the protective SNPs [, for rs3743078, CG OR=0.68 (95% CI: 0.58, 0.80), GG OR=0.49 95% CI: 0.34, 0.68)]. The highly significant p-trends for these results support both sets of variants following an additive genetic model. All the findings in remain statistically significant after correction for multiple comparisons except for the results for rs1996371.
The risks of heavy smoking associated with each genotype of the significant SNPs in the A5A3B4 gene cluster are shown. Associations are adjusted for age and gender
Because the risk and protective SNPs represent two separate associations, the combined influence of the variants was assessed to determine how heavy smoking was affected by genotype combinations of SNPs from each group. The nonsynonymous SNP rs16969968 was used to represent the risk group and rs3743078, one of the most significantly associated SNPs, was used to represent the protective group. The combination of the wild-type genotype of rs16969968 (GG) and homozygous variant genotype of rs3743078 (GG) is associated with the lowest risk for heavy smoking (, shaded cell) whereas the AA/CC genotype combination is associated with the highest risk (, outlined cell). The odds of heavy smoking relative to light smoking among smokers with the highest risk genotype combination was 2.43 times higher than in those with the lowest risk genotype combination.
Table 4 Combined influence of a SNP with a minor allele associated with increased risk (rs16969968) and a SNP with a minor allele associated with decreased risk of heavy smoking (rs3743078). The numbers at the top of each cell indicate the number of cases and (more ...)
Two of the correlated SNPs we found associated with increased risk of heavy smoking (rs803191 and rs1051730) have also been found to be associated with increased risk of lung cancer by three large GWASs (10
). In our study, 10 light smokers and 75 heavy smokers had been diagnosed with incident lung cancer. To determine if any part of the association we observed between these SNPs and heavy smokers was due to a direct association of the SNPs with lung cancer, we conducted a sensitivity analysis excluding the 85 lung cancer cases. The associations of both the risk and protective SNPs with heavy smoking changed very little (≤0.01 for per-allele and ≤0.02 for genotype ORs). As reflected in the trend p-values, which are listed in the column labeled excluded LCa p in , all previously significant results remained significant. The p-values for the risk SNPs (dark rows) were slightly attenuated while those for the protective SNPs (gray rows) were somewhat more significant.
We also assessed the association of the risk and protective SNPs with lung cancer directly while controlling for smoking phenotype. No association was observed for the protective SNPs and lung cancer risk [for rs3743078, CG OR=0.89 (95% CI: 0.55, 1.44), GG OR=1.31 (95% CI: 0.51, 3.36)]. However, the homozygous variant genotype of the risk SNPs was significantly associated with an increased risk of lung cancer [for rs16969968, GA OR=0.67 (95% CI: 0.41, 1.10), AA OR=1.80 (95% CI: 1.02, 3.20)].
The associations with heavy smoking for the SNPs in the CHRNB3-CHRNA6 gene region or in the CHRNA4 and CHRNB2 genes are shown in . Two SNPs (rs7012713, in the CHRNB3 gene, and rs7828365, located 5' to the CHRNA6 gene) were significantly associated with heavy smoking [for rs7012713, per allele OR=1.42 (1.05, 1.94), p=0.023, for rs7828365, per allele OR=0.84 (0.72, 0.99), p=0.039]. However, these associations were no longer statistically significant after correction for multiple comparisons. No other statistically significant associations were found for these four genes.
Characteristics, minor allele frequencies, and associations with heavy smoking for various nicotinic receptor SNPs