We have demonstrated and replicated genetic associations between SNPs at the CHRNA3/5 locus and COPD in four study populations. The estimated population attributable risk from this locus was 12.2% and represents the discovery of a common major locus contributing to COPD in the general population. However, a potential complication with the interpretation of these findings is the possibility that differences in smoking behavior, likely related to nicotine addiction, between COPD cases and controls may drive the observed association. This is similar to the recently reported association of CHRNA3/5 SNPs with lung cancer 
In the current study populations, only limited assessment of nicotine addiction is available: (i) whether subjects were still smoking at the time of study participation, and (ii) their lifetime smoking intensity. Thus, we have limited ability to disentangle a genetic determinant of smoking behavior from a genetic determinant of COPD through an alternative pathway. There are several pieces of evidences to suggest that there could be a direct effect of CHRNA3/5 locus on COPD, independent of smoking behavior. First, to the extent that smoking behavior is captured in pack-years, this effect should be factored out by the statistical design in which the discovery analyses used a logistic regression model incorporating pack-years, age and gender as covariates. The adjustments for pack-years smoked, age and gender were also performed in all the replication analyses. However, pack-years smoked only partially capture smoking behavior. Many other factors, such as depth of inhalation, number of puffs per cigarette, and age of starting smoking are also likely to affect the toxicant exposure and effect. Second, we also tested directly for an association between the amount of smoking, measured as pack-years, within cases and controls with the SNP rs8034191. There was no significant association between the SNPs and pack-years of smoking in the Bergen and NETT/NAS populations. This is consistent with the observed allele frequency among the Norwegian pediatric general population sample (minor allele frequency
551) which is between that observed for cases and controls and not significantly different from either.
We observed a genotype-by-environment interaction between the risk of the rs8034191 genotype and current smoking status on COPD in the Norwegian sample (P
0.002, ), showing a substantially higher risk of COPD in current smokers carrying the rs8034191 C allele (OR
2.0) than in former smokers (OR
1.1). There are several possible explanations for this gene-by-environment interaction. First, it could relate to nicotine addiction; smokers that have greater difficulty quitting smoking may be more likely to develop COPD. Alternatively, it could indicate that a subset of individuals is at greater risk for developing COPD if they continue to smoke. A similar interaction with current smoking was not identified in the ICGN families. Since all the COPD patients in the NETT population were former smokers, we could not address this question in the NETT/NAS study.
The association of smoking dependence was explored in the lung cancer report by Hung and colleagues 
who did not detect any association with individual Fagerstrom indices of nicotine addiction or when comparing controls with a heaviness of smoking index (HSI). Another lung cancer report by Amos and colleagues 
did show weak evidence for association with smoking behavior, while a further report by Thorgeirsson and colleagues
showed very strong association with smoking behavior. Association of this locus with nicotine dependence has been reported in two other studies 
. Thus, it is reasonable to conclude that the variants may affect smoking behavior, at the same time as having a significant effect on COPD and other smoking related diseases such as lung cancer and peripheral arterial disease 
The CHRNA3/5 SNPs were also associated with lung function (FEV1
) in the ICGN and BEOCOPD populations. These SNPs were shown to be associated with FEV1
in the British Birth Cohort (rs8034191 and rs1051730, p
0.029 and 0.023, respectively (http://www.b58cgene.sgul.ac.uk/
, accessed [3/7/2008]). Historically, nicotinic receptors are classified as neuronal or muscle-type, based on their initial site of identification and composite subunits 
. Cholinergic activity in the airways primarily induces tracheo-bronchial smooth muscle contraction and mucous secretion. However, there is an increasing body of literature showing the importance of extra-neuronal cholinergic signaling 
in the lung.
The association of the SNPs at the chromosome 4 HHIP (Hedgehog-Interacting Protein) locus is also interesting, though it did not reach the stringent genome-wide significance levels in the populations studied in this manuscript. These SNPs were also associated with FEV1
in the BEOCOPD study (rs1828591 and rs13118928, p
0.0025 and 0.0014). The same SNPs are also associated with FEV1
in the British Birth Cohort (rs1828591 and rs13118928, p
0.039 and 0.038, respectively) but were not associated with FEV1
in the ICGN population.
In another manuscript in this issue of the journal, genome-wide association analysis results for FEV1/FVC in the Framingham Heart Study (FHS) are reported (Wilk et al). Due to differences in genotyping platforms, the most significantly associated SNPs on chromosome 15 in our study were not genotyped in FHS. Analysis of the genotyped SNPs in the chromosome 15 region in the FHS indicated no significant association with COPD, but association with FEV1 percent predicted was observed with one SNP in LD with rs8034191 (rs11636431 p value 0.007). Evaluation of the imputed data for the most significantly associated SNPs in our populations did not show association with COPD in FHS. Several factors could contribute to the absence of association to the COPD phenotype in FHS: (1) The FHS cohort is a population-based collection, while our studies evaluated populations ascertained specifically for COPD; (2) The FHS cohort was recruited over three decades, while our cohorts represent more recent recruitments (in the last 5–10 years)-smoking habits have changed over time, and it is also possible that COPD clinical characteristics have changed over this period; (3) Our cohorts include a greater proportion of severe COPD subjects than in FHS; and (4) There could be differences in linkage disequilibrium patterns between study populations. Further studies will be required to define the specific genetic determinants influencing COPD on chromosome 15, the relationship of these genetic factors to smoking behavior, and the characteristics of COPD subjects influenced by these genetic determinants.
The association of the Chromosome 4 region in the FHS cohort was genome-wide significant for the FEV1/FVC ratio and was also associated with COPD. This association was subsequently replicated in the Family Heart Study population. Though the HHIP locus association in our study did not reach genome-wide significance, the additional evidence from the FHS and Family Heart Study underscore the importance of this locus on COPD susceptibility.
We used independent populations with varying COPD severity, independent genotyping platforms and stringent statistical significance criteria to define genome-wide significant associations. We used consensus criteria for replication using a multi-stage replication design with similar phenotypes, the same genetic model and direction of association 
. The levels of statistical significance of the association for our most significant results in the CHRNA3/5 region were consistent in all of the populations studied and are unlikely to be false positive results. The p values after adjusting for multiple testing using the most conservative Bonferroni correction were 7.3×10−5
for the SNPs rs8034191 and rs1051730 respectively. Though this can be considered as strength, the conservative approach for SNP confirmation that we have used may lead to larger false negative rates. However, with the inconsistent results of previous complex disease genetic association studies, we contend that a conservative approach is appropriate. We selected only the top 100 SNPs from the GWAS for subsequent replication study and a larger number of significant associations may have been uncovered if more of the most promising SNPs had been followed up. A negative association in the replication studies may not rule out a true association, since the power to detect association in the replication populations may be limited. The primary replication cohort (ICGN) is moderately powered to detect the replicated associations. Though the sample sizes of the NETT/NAS and BEOCOPD studies are relatively low, these studies include a large percentage of severely affected individuals, who may be enriched for COPD susceptibility genes. This likely account for the high rate of replication in these populations. COPD is a heterogeneous disease and we used a spirometry-based definition for COPD in all of the populations. Differences in smoking exposure, current smoking status, entry criteria and geographic origin of the cohorts might contribute to phenotypic heterogeneity and may lead to lack of replication. The fact that the replicated associations holds-up strongly and consistently in all the populations shows that phenotypic heterogeneity likely has little effect on the most significant results.
In summary, we have identified robust evidence of association of COPD with the α-nicotinic receptor (CHRNA 3/5) and HHIP loci. The hedgehog (Hh) gene family encodes signaling molecules that play an important role in regulating morphogenesis and the HHIP locus may play a role lung development. Although there is evidence of association of CHRNA 3/5 locus with nicotine addiction, both this study and recent reports of a lung cancer association 
with the same alleles suggest that this region may be involved in more than nicotine addiction, and may potentially have direct functional relevance in the development of COPD, lung cancer, peripheral arterial disease, and other smoking related conditions. The first-degree relatives of both lung-cancer patients and COPD patients have higher rates of impaired forced expiratory flow rates than relatives of patients with non-pulmonary disease 
, suggesting that susceptibility to lung cancer and COPD share common familial components. The association of CHRNA 3/5
locus with COPD, lung cancer, and peripheral arterial disease is powerful enough to make genetic screening of smokers an attractive interventional strategy.