Candidate genes may be suggested by pathogenesis, or vice versa. Variation, or polymorphism, within the gene can be classified in different ways22, such as the structural nature of the change in the DNA, or its effect on the protein it codes for. Two common structural changes are microsatellites – multiple repeats of a short segment of DNA, and single nucleotide polymorphisms (SNP's) – a change of a single base. The latter are the most common type of polymorphism in the human genome23. Such changes may occur in coding regions of DNA (those that contribute to the making of a protein) or non-coding regions. If a change occurs in a coding region it can be described as non-synonymous or synonymous, depending on whether it affects the amino acid sequence of the gene product or not. Generally speaking non-synonymous changes in coding regions are more likely to alter the function of a protein24, and hence to be related to disease. International projects, such as the SNP consortium25, which catalogues common SNP's in the human genome, and HapMap2627, which has genotyped SNP's in 4 major ethnic groups, have contributed to the many databases available on genetic variation. Researchers can use such resources to identify potential disease causing polymorphisms, and their likely population frequencies, allowing the design of case-control association studies, looking for the polymorphism in those with and without the disease. This is a widely used approach, though often producing inconsistent results 28; this may be because of variation in the definition of cases and controls, underpowered studies, racial differences and population heterogeneity. The issue of power is particularly important when examining a complex disease such as COPD, as each gene may contribute only a small amount to the clinical phenotype: if this results in a genotype relative risk of developing the trait of less than 2, then adequate power may not be achievable29.

Linkage studies look for haplotypes, or short segments of the genome, conserved between generations by virtue of their size 30 – anything larger has the potential to be changed by recombination during meiosis. If a haplotype can be found that is passed down through a family, alongside a disease, then it suggests that there is a gene within or close to it that may have a functional effect on the disease. Haplotype analyses can also be useful in association studies, though difficult to perform31. This is because they allow for the possibility that a combination of SNP's within a gene may be causing the trait in question, rather than one of the SNP's alone32.

Linkage is usually reported as a logarithm of the odds (LOD) score33, which is a form of likelihood ratio derived from the recombination fraction between the marker and the proposed locus of the disease-causing gene. The threshold level of LOD score needed for genome-wide significance at 5% (p = 0.05) varies dependent on the study design, from 3.3 for family studies, using a proposed mode of inheritance (parametric linkage analysis) to 3.6 in sibling pair (non-parametric, no model of inheritance) studies 34. Linkage analyses need to be followed by case-control association analyses for any genes in the area of interest that have a plausible link to disease – established because of potential in pathogenesis, or differential expression in the target tissue 35.

The major linkage study in COPD is the Boston early-onset COPD cohort 36373839, which performed genomewide linkage analysis in 585 members of families with early onset COPD, looking for linkage to pulmonary function test results. Areas of linkage were found for FEV1/FVC on chromosome 2, chromosome 1 and chromosome 17. The area on chromosome 2 was subsequently investigated by the same group, identifying SERPINE2 as a potential candidate gene. This area also contains the gene for the IL8 receptor, which might contribute to COPD since IL8 is involved in neutrophil chemotaxis to the lung 40(a critical process in delivery of destructive proteases). FEV1 linked to chromosomes 12 and 19 38 in areas containing the genes for microsomal GST1 and TGFβ respectively. Mid expiratory flow, which is reduced in small airway disease in COPD, linked to chromosomes 2 and 12, together with a broader area on chromosome 19 39.

There are three classes of protease that have been studied in COPD – the serine proteases, which includes neutrophil elastase (NE)and proteinase 3, the cysteine proteases, such as cathepsin-B, and the matrix metalloproteases (MMP's) 41. In general the serine proteases are capable of degrading elastin and some forms of collagen 41, whilst the MMP's have more of an effect on collagen, gelatin and laminin 16. Each enzyme is inhibited by one or more anti-proteases, may inactivate other anti-proteases, or activate pro-inflammatory cytokines, such as TNFα, by interacting with proteinase activated receptors (PARs) 42. The proteases function to clear debris and damaged tissue, but if their action is not effectively controlled they may produce excessive lung damage. The relationships between these enzymes, their inhibitors and some inflammatory mediators are shown in Table 1.

The actions of MMP's include degradation of collagen, inactivation of AAT and activation of TNFα. Their action is reduced by tissue inhibitors of metalloproteinases (TIMP's). Studies using knockout mouse models have supported a role for MMP's in COPD. Mice over-expressing MMP1 develop emphysema 51, whilst those deficient in MMP12 are relatively protected 52. Further support comes from clinical studies showing increased concentrations of MMP's in the bronchoalveolar lavage fluid of COPD patients 53.

The most widely studied MMP gene polymorphism is in the MMP9 gene, located on chromosome 20. A SNP in the promoter region (C→T, position -1562), which increases its activity has been described 54, and linked to COPD in both Chinese 55 and Japanese populations 56. The Chinese cases were defined by airflow obstruction, according to the GOLD criteria, whilst the Japanese cases were defined by the degree of emphysema on CT scan. This latter group found that airflow obstruction did not correlate with presence of the T allele, though gas transfer corrected for alveolar volume (KCO) did. A later study has narrowed the emphysema phenotype linked to this polymorphism to upper zone predominant disease 57.

SNP's in MMP1 and MMP12 have also been studied in COPD. An insertion in the promoter region of MMP1 (G→GG, position -1607) that increases its transcription 58 by creating an extra transcription factor binding site has been described. This SNP occurs in 30% of the general population 58 and was negatively associated with rapid decline of lung function (defined by FEV1) in one case-control study59. This does not, however, have an explanation from its function. It would be expected that the GG variant would be positively associated if the higher level of MMP1 transcription lead to more lung damage. A role for polymorphisms of MMP12 was investigated by the same group, but an association with declining lung function was not seen 59. A haplotype containing the MMP1 G→GG SNP, together with an MMP12 SNP that results in a change in protein composition (Asn357Ser), was found more commonly in the rapid declining group59. The authors suggested that this may be because the gene actually responsible for rapid decline lies in linkage disequilibrium with these two SNP's.

There are four TIMP's (TIMP1-4) that inhibit active forms of MMP. Although all TIMP's are capable of inhibiting any MMP their affinity for each MMP varies and TIMP2 has been shown to have a greater affinity for MMP2 and MMP960. The contribution of various MMP's, TIMP1 and TIMP2 to emphysema have been investigated, and a key role for the MMP2-TIMP2 system proposed 61. Two SNP's in TIMP2 are more common in Japanese subjects with COPD. One in the promoter region (that may cause reduced TIMP2 levels) and a second synonymous change in exon 3 62. This result has not yet been reproduced in other ethnic groups, and the functional consequences of each SNP remain theoretical, rather than proven. This should be an area for future research.

α-1 antichymotrypsin (SERPINA3) inhibits cathepsin G and mast cell chymase in a reversible fashion. Two SNPs in SERPINA3, associated with low SERPINA3 levels, have been associated with COPD in Swedish subjects 6364, though their cases were defined by a measure of airway resistance, rather than standard spirometric parameters. The positive results for these SNP's were not reproduced in Japan in patients with airflow obstruction and low FEV1 6566, though a non-synonymous mutation affecting the signal peptide region was found more commonly in the COPD group. All 3 of these mutations were examined in an Italian study of patients with airflow obstruction and were not found to be associated with disease, though their cases included subjects with bronchiectasis as well as COPD 66. The variation in results between the studies could be explained by the different diagnostic criteria used by each group; it may be that the mutations are linked to airway resistance, but not to airflow obstruction, perhaps emphasising the heterogeneity of COPD.

The glutathione-S- transferase (GST) genes code for a family of enzymes that detoxify some of the harmful contents of tobacco smoke 69. Polymorphisms in the genes are known to have functional consequences, and have been examined in COPD 707172. The two variants with the most evidence supporting a role in the disease are GSTP1 and GSTM1.

GSTP1 contains two known SNP's, though only one is known to have an effect on the catalytic activity of the enzyme. This is an A→G change at nucleotide +313, resulting in a single amino acid substitution (Ile105Val) 73 shown to increase the metabolism of carcinogenic aromatic epoxides 74. Studies of the relationship of this variant to lung disease have varied in their results. It would be expected that the 105Ile variant would be associated with higher levels of lung damage, since it is less active against oxidants – this was confirmed by an association with airflow obstruction in a Japanese population 71and replicated in a Caucasian population in the Lung Health Study (LHS), where this polymorphism together with a family history of COPD was linked to rapid decline of FEV1 (OR = 2.20, p = 0.01)70. Conversely the same group showed that the 105Val variant was associated with low baseline lung function (OR = 1.69, p = 0.016) and rapid decline in the higher baseline group (p = 0.017) 72, whilst Gilliland 75demonstrated reduced annual growth rates for FEV1 and FVC in children homozygous for the 105Val variant. The latter results are difficult to explain on the basis of this gene's action alone, but might be understandable if there are gene-smoking or gene-gene interactions affecting the expression of the gene product. No gene-smoking affects were seen in the LHS 72, but there may be an additive effect of polymorphisms in GSTP1 and other GST genes70, suggesting that a consequence might not be seen unless a change in several gene products were present.

GSTM1 has 3 known alleles, one of which is a null allele, such that homozygotes for the null allele have no detectable GSTM1 activity. This genotype has been associated with emphysema 76 and chronic bronchitis 77, with conflicting results concerning its role in lung cancer 7678. In common with most other genetic studies in COPD the positive results have been difficult to replicate 28, though this may be because studies looked at different subgroups of patients with COPD. The negative studies defined their cases by airflow obstruction 79 and rapid decline in FEV1 70, hence might not have picked up a change in gene prevalence in chronic bronchitics. This difference in case definition remains a common theme in COPD genetics studies.

There are three superoxide genes, coding for scavengers of reactive oxygen species (ROS) 68. Extracellular superoxide dismutase (SOD3) is present at high concentrations in areas of the lung containing large amounts of type 1 collagen, especially around large airways and also adjacent to alveoli 80. It is thought to have a role in protecting the lung, particularly during inflammation 8182. A SNP (C→G substitution at +760) of SOD3 that increases plasma enzyme levels has been examined in 2 studies relating to COPD 6883, and found to have a protective effect. A case-control study was carried out in New Zealand, where the mutation was found more frequently in resistant smokers than in those with COPD(OR = 4.3, p < 0.05) 68. The second study was part of the Copenhagen City Heart Study, which examined 9258 individuals in both cross-sectional and prospective study designs 83. This demonstrated a reduced risk of developing COPD in smokers (OR = 0.4) and a reduced risk of hospital admission or death due to COPD (hazard ratio = 0.3) in those carrying the mutation. Since this effect was not seen in non-smokers, whose odds ratio of developing COPD when they carried the mutation was 1.5, it suggests a gene-smoking interaction, though this could not be statistically proven.

Microsomal epoxide hydrolase (EPHX1) is expressed in bronchial epithelial cells and metabolises highly reactive epoxide intermediates in cigarette smoke 8485. There are 2 known SNP's in this gene that affect enzyme activity by a single amino acid substitution. The first SNP is in exon 3 (Tyr113His), the second in exon 4 results in a further change in protein constitution (His139Arg). In both cases the His variant is associated with lower levels of enzyme activity 8687. Both polymorphisms only account for a modest change in activity level 87, so it may be that there is also variation in the gene's regulatory regions 88.

Patients carrying both His variants were at the highest risk of developing COPD (OR = 4.1, p < 0.001) and emphysema (OR = 5, p < 0.001) in a Scottish population89. This result was replicated in those with more advanced COPD in Japan (OR = 2.9, p = 0.02)90 despite the differing frequency of genotypes between the two racial groups. The LHS demonstrated a relationship with rapid decline in lung function (FEV1) for the same haplotype, though this was only statistically significant for those with a family history of COPD 91. The His139 variant alone was associated with a spirometric diagnosis of COPD in the Boston early-onset COPD cohort 28.

The contribution of this gene to the heterogeneity of COPD has been examined in more detail in the National Emphysema Treatment Trial (NETT) Genetics Ancillary Study 92. The authors studied a number of polymorphisms and looked for correlation between genotype and functional capacity phenotypes in two separate patient groups, hypothesising that there is a genetic basis to the observed phenotypes. The exon 3 SNP (Tyr113His) was associated with poor exercise capacity, whilst the exon 4 SNP (His139Arg) was connected to relatively greater gas transfer (DLCO). This study was powered to detect a moderate effect of each genotype on overall phenotype, so taken with the previous positive studies it seems likely that these polymorphisms contribute to the COPD phenotype. Their link to specific subgroups of COPD patients will need further study.

Heme oxygenase-1 (HMOX1) is an enzyme important in heme metabolism, which catalyses the oxidative cleavage of heme, resulting in the release of carbon monoxide, bilverdin and iron 93. Bilverdin is then broken down into bilirubin, which scavenges local ROS; thus HMOX1 contributes to the generation of antioxidants. It is present at higher concentrations in the lungs of smokers than non-smokers, suggesting up-regulation in these circumstances 94, presumably because of a response to increased ROS.

A microsatellite (GT)n repeat in the 5' region of HMOX1 has been described that seems to alter the level of transcription when under thermal stress 95. When the microsatellite is longer, it is not induced as effectively by ROS 96. This suggests that in the presence of a long GT repeat (for instance n ≥ 30) smokers would not be able to protect their lungs from the damage induced by ROS in smoke, and thus would be more susceptible to emphysema. Two clinical studies have shown a link between this HMOX1 polymorphism and COPD. A Japanese case-control study showed that patients with 30 or more GT repeats in the microsatellite region were more likely to have emphysema, diagnosed by CT scan96. A larger study in France showed that 33 or more GT repeats was associated with airflow obstruction and more rapid decline of lung function, particularly in smokers 97. They were able to show a significant gene-smoking interaction (p = 0.0006 for FEV1/FVC decline). This effect on decline was not, however, reproduced in the LHS 70.

TNFα mediated inflammation is thought to play a key role in both the respiratory 100 and systemic features of COPD 98. A SNP in the promoter region of the TNFα gene (G→A at position -308) directly affects gene regulation, and is associated with high TNFα production 101. This polymorphism has been studied in several COPD related phenotypes, with differing results. An initial case-control study in Taiwan examined subjects with chronic bronchitis, hypothesising that this was linked to increased airway inflammation 102. They found an increased prevalence of the polymorphism in cases relative to controls (p < 0.01, OR = 11.1). It has also been linked to airflow obstruction without chronic bronchitis, and severity of emphysema in Japanese subjects 103104. Studies in Caucasians have not been able to reproduce these results 2891 which might be explained by variation in genotype frequencies between races (data available from HapMap27), or by linkage dysequilibirum with HLA alleles, seen previously in the Caucasian population 105.

TGFβ1 regulates extra-cellular matrix production, cell growth and differentiation, tissue repair and some immune responses 106. Mice who are unable to activate latent TGFβ develop emphysema via alterations of MMP12, suggesting that disordered activation relates to the pathogenesis of COPD 107. A linkage analysis in the Boston early-onset COPD study showed association between an area of chromosome 19 containing the TGFβ1 gene and FEV1108. Three SNP's in this gene had a significant association with severe COPD in the NETT cohort 108. This association was replicated for two of the SNP's by Hersh et al92, who linked them both to subjective measures of dyspnoea, though not objective measures of exercise capacity. This apparent discordance may be important when defining phenotypes within COPD.

The two SNP's identified by Hersh et al both have an effect on TGFβ1 levels. The first is a C→T change at position -509, in the promoter region, which enhances promoter function, thus increasing levels of TGFβ 109. The second is a C→T change at position 613, which leads to an amino acid substitution (Leu→Pro) and higher production of TGFβ1 110. If both of these polymorphisms are implicated in COPD, it suggests that TGFβ may have a protective role. A case-control study examining the latter SNP in COPD subjects, resistant smokers and healthy controls concurred, finding that the Pro allele was less common in COPD subjects relative to resistant smokers (OR = 0.59, p = 0.01) and controls (OR = 0.62, p = 0.005). Further research on the role of TGFβ in COPD may help to clarify if this association has credibility in relevant pathogenic processes.

Vitamin D binding protein, also known as Gc globulin, is a precursor of macrophage activating factor (MAF) 111 and enhances the neutrophil chemotactic properties of C5 derived peptides 112. The latter function is prevented by neutrophil elastase inhibitors 113, suggesting a relationship between the protease-antiprotease pathway and inflammation. This would fit well with a role for vitamin D binding protein in the pathogenesis of COPD.

A number of studies have looked for links between polymorphisms in this gene (GC) and COPD. Two non-synonymous SNP's have been identified, which represent the GC2 and GC1S alleles. The GC2 allele has been found to be protective in studies of Caucasian subjects 114115, consistent with the fact that only 10% of this form can be converted to MAF 116. No role has been proven for this allele in neutrophil chemotaxis 114. The GC1S allele has not been shown to have a significant association with COPD 28. In Japanese subjects the GC1F allele has been linked to an increased risk of developing airflow obstruction, emphysema and a rapid decline of FEV1 117118. Caucasian patients homozygous for this allele were at increased risk of developing COPD in one study 115 but not in another 114. Neither could the link to rapid decline be reproduced in this racial group 91. The difference in allele frequency between racial groups may explain why studies in Caucasians (who have a lower frequency of the 1F allele) have been unable to detect an association, as they would have required greater patient numbers to be adequately powered. An alternative explanation is that there is racial variation in gene associations with COPD.

Studies in transgenic mice have shown that if IL13 is over expressed, it results in cathepsin and matrix metalloproteinase dependent emphysema with mucus metaplasia 119. A polymorphism in the promoter region (C→T, position -1055) is associated with increased IL13 production 120, with the T genotype being more common in COPD patients 121. In mice IL13 induced emphysema is characterised by excessive pulmonary mucus production, so further studies looking for the prevalence of this polymorphism in the subgroup of COPD patients with chronic bronchitis might be worthwhile.

The surfactant proteins are hydrophobic proteins that contribute to regulation of surface tension in the alveoli. Components of surfactant also have a role in host defence and control of inflammation. Alterations of surfactant might therefore be a factor in COPD, as suggested by mathematical models of emphysema 122 although this has yet to be studied in vivo. A SNP in the gene coding for surfactant protein B (SFTPB), which causes a single amino acid substitution (Thr131Ile), has been associated with COPD in the Boston Early-onset COPD cohort 28, and in a case-control study in Mexico 123. In the NETT cohort this was also seen when gene-environment interaction was taken into account, where it was associated with dyspnoea score and exercise capacity92. In the Mexican study a number of SNP's and microsatellites were examined, with mutations in SFTPB (or microsatellite markers linked to it) being the most closely associated with COPD.

The SERPINE2 gene was identified as having a potential role in COPD by a novel method. Firstly linkage of airflow obstruction to an area on chromosome 2 in the Boston early-onset COPD cohort 3738, followed by integration of these results with knowledge of genes expressed during murine lung development, together with human lung microarray datasets from control subjects and those with severe COPD 35. Multiple SNP's in this gene were examined in patients from the NETT cohort, with several being significantly associated 35. A subsequent large case-control study did not, however, find any association with COPD in European patients 124 and questioned the validity of some of the results reported in the original study.

SERPINE2 has not been studied in COPD. It is known to be an inhibitor of trypsin-like serine proteases, but not neutrophil elastase 125, which might have indicated a role in the protease-antiprotease pathways. Its major function is in coagulation and fibrinolysis 126. Although enhanced prothrombotic markers have been linked to decline of FEV1 in one small study in COPD 127 this has not been widely investigated.