Chronic obstructive pulmonary disease (COPD) is a leading cause of morbidity and mortality, and is predicted to become the 4^th leading cause of death by the year 2030 [1]. Whilst smoking is a significant environmental cause of COPD, not all smokers will develop disease. It is well recognised that COPD has a genetic component as well as environmental, and that this may account for these differences in susceptibility.

Many candidate gene studies have been carried out over the past few years, with varying degrees of reproducibility. Conflicting results may be due to population differences, spurious results caused by small sample sizes and the subsequent low power of the study to detect true associations or to variation in the phenotype. Meta-analysis can be used to pool results from genetic studies and give an overall conclusion, and this approach has been used to summarise the results for several COPD candidates [2]. This study revealed significant associations for three polymorphisms in transforming growth factor beta 1 (TGFB1), although it was acknowledged that this was a limited analysis based on a small number of studies.

Other strategies for identifying the genetic basis of complex human disease include integrative approaches which combine gene expression data with association studies, or unbiased genome-wide approaches. Both of these strategies have also been employed within the study of COPD.

A recent genome-wide association study highlighted a region on chromosome 15q25 which shows strong association to COPD [3]. This region contains several genes, including members of the nicotinic acetylcholine receptors (CHRNA5, CHRNA3 and CHRNB4) and the iron regulatory protein 2 (IREB2). Although this region has also been linked to lung cancer and nicotine addiction as well as COPD [4, 5], the strong levels of linkage disequilibrium make it difficult to refine the location of the true association signal and the functional variants. However, the expression of IREB2 has also been shown to be altered in lung tissue from COPD patients compared to controls [6], adding to the evidence for IREB2 acting as a COPD susceptibility gene.

The aim of the current study was to investigate polymorphisms in the TGFB1 and IREB2 genes in our existing large collection of COPD cases and controls.

The values for linkage disequilibrium (LD) between the SNPs are shown in Figure 1. This reveals very strong levels of linkage disequilibrium between groups of SNPs in both genes. Within TGFB1 there is strong LD between SNPs 1 and 2 (rs2241712 and rs1800469; r² = 0.97), which is reflected in the very similar odds ratios calculated as part of the published meta-analysis. Within IREB2, it is clear that the seven SNPs studied fall into three main groups. SNPs 2 and 5 are in linkage disequilibrium (rs2656069 and rs10851906; r² = 0.99), as are SNPs 3, 4, 6 and 7 (rs1964678, rs12593229, rs965604 and rs13180; all pairwise r² = 0.99). SNP 1 (rs2568494) is not in linkage disequilibrium with any of the other IREB2 SNPs (r² <0.34). For this reason, subsequent tables show information for TGFB1 SNPs 2, 3 and 4, and IREB2 SNPs 1, 2 and 4 only. There are 51 SNPs in the region of IREB2 in HapMap, and using the 3 tagSNPs as presented here captures 64% of the variation (33/51 variants). These calculations are harder to do for TGFB1 as two of the SNPs we have used are not in HapMap, so we cannot calculate the LD between these SNPs and other variants in HapMap. There are only 8 SNPs in the TGFB1 region in HapMap, and rs1800469 captures information about one additional variant, ie 2 out of 8 SNPs (25%). It is likely that we have captured more variation than this, but are unable to quantify it more precisely due to the lack of information about our other two tagSNPs in HapMap.

Though there were significant differences between centres for the allele distribution of some of the SNPs, our results are unlikely to be due to population stratification since cases and controls were matched within centres. We also examined the genotype odds ratios for all six SNPs in each centre (Additional file 1: Table S3). The within-centre odds ratios for the three TGFB1 SNPs were all non-significant, showed no particular pattern and fluctuated around the overall (all-centre) odds ratios. The within-centre analysis for the IREB2 SNPs showed that, though non-significant due to the smaller sample size, the direction of the relationship was the same in each centre as the all-centre results. Thus each centre is showing the same IREB2 SNP relationships with COPD.

This study has replicated the associations seen for polymorphisms in IREB2 which have been previously reported, but failed to reproduce the associations seen in a recent meta-analysis for TGFB1. Recent studies using cluster analysis suggest that SNP rs1800470 in TGFB1 is associated with the emphysema-predominant phenotype [13]. There is a plausible biological explanation for this based on the critical role that TGFB1 plays in reducing matrix metalloproteinase-12 activity, a key mediator of emphysematous change. This is further supported by recent replicated studies showing that MMP-12 variants are associated with a protective effect on lung function and COPD [14, 15]. The supposition is that low levels of MMP-12 reduce the likelihood of emphysematous change. Over-expression of TGFB1 would therefore predictably have the same protective effect and reduced expression would be a predisposing factor. This is also supported by work in a murine model of COPD, where activation of the TGF-Β signalling pathway was observed alongside a partial rescue of the emphysema phenotype [16]. We did not have phenotypic data to assess emphysema scores and so were unable to test this proposition. It is also possible that other genetic variants in the TGFB1 region which were not investigated as part of the current study may be involved in COPD susceptibility.

We have confirmed previous observations which suggest that IREB2 variants may play a role in COPD. This appears to be independent of an effect on lung function, as we failed to detect any association between FEV1 and any of the SNPs included in this study. This is in agreement with the genome-wide association study which looked for variants associated with lung function, which also fails to find evidence for a role of IREB2 [17]. The IREB2 gene is located within the region identified by the COPD GWAS, and the high levels of linkage disequilibrium in this area make it difficult to identify the specific functional variant or gene which is underlying this association. Only SNP 1 (rs2568494) is in relatively high levels of linkage disequilibrium with SNPs identified by the genome-wide study of COPD (HapMap CEU data: r² = 0.790 with rs8034191 and r² = 0.692 with rs1051730), although the other IREB2 SNPs are also in LD with variants in other genes in the region, including CHRNA5. Identification of the genes and variants contributing to the association with COPD will require targeted resequencing of the region and further functional work. In support of IREB2 involvement, a previous publication has shown increased levels of IREB2 mRNA in COPD patients versus controls [6]. This is an intriguing observation as this gene codes for an iron binding protein. Iron homeostasis and free iron concentration are likely to be important mediators of oxidative stress and iron could therefore contribute to local damage by this mechanism. IREB2 protein is expressed in the lungs and cigarette smoking has been associated with higher levels of iron in the lung [18]. IREB2 knock-out mouse models have a predisposition to developing neurodegenerative disease due to aberrant cellular iron homeostasis [19], though the lungs of these animals were not examined in any detail. A potential mechanism for distorted ion homeostasis relates to membrane serine proteases (matriptases) which regulate a number of biological effects. Genome-wide association studies have shown that mutations in matriptase result in refractory iron deficiency anaemia and the proteolytic activity of matriptase is critical in regulating iron in this form of anaemia as loss of function mutations result in the disease [20]. The catalytic domain of matriptase is inhibited by alpha1-antitrypsin [21] and it is therefore conceivable that two major pathways associated with progressive lung damage, namely proteolysis and oxidative stress, are linked through iron homeostasis. This could also explain the tendency for greater oxidative stress in alpha1-antitrypsin deficiency as more active matriptase may result in excessive free iron.

In summary, this study failed to replicate previous reports of associations between SNPs in TGFB1 and susceptibility to COPD, despite being adequately powered. This may be due to the possibility that the association is actually with specific emphysema phenotypes rather than overall susceptibility to disease, which could not be assessed in this patient group. The replication of the association between variants of IREB2 and COPD provides further evidence to support the role of this genomic region in COPD pathogenesis, and the role of iron regulation deserves further investigation.