Chronic obstructive pulmonary disease (COPD) is highly prevalent and the only disease with increasing mortality rates worldwide.¹ The disease is characterised by irreversible air flow limitation and associated with an influx of neutrophils, macrophages and CD8 T lymphocytes in the airways.^2,3 Most patients with COPD show airway hyper-responsiveness (AHR),⁴ an exaggerated airway response to non-specific stimuli resulting in airway obstruction. The severity of AHR is positively associated with inflammation in lung tissue⁵ and numbers of CD8 cells in bronchial biopsies in COPD.⁶ One study reported an association of AHR with sputum inflammatory cells in COPD,⁷ whereas another study did not find this association.⁸

Cigarette smoking is by far the most important risk factor for COPD but there is a wide range in disease severity, irrespective of the number of pack-years of smoking. Furthermore, only a minority of smokers develop the disease, suggesting that besides smoking, an underlying genetic constitution has a role in the development and severity of COPD. A disintegrin and metalloprotease 33 (ADAM33) gene is a gene of putative interest for COPD. It was first identified as a susceptibility gene for asthma and AHR,^9,10 and subsequently as a susceptibility gene for COPD in the general population.¹¹ It is furthermore associated with accelerated decline in lung function in the general¹¹ and asthma population.¹²

ADAM33 belongs to the ADAM family of membrane-anchored metalloproteinases that has a role in a variety of processes such as cell fusion, adhesion and signalling.^13,14ADAM33 is expressed in both airways smooth muscle cells and lung fibroblasts. Although the overall substrates and function of ADAM33 are as yet unknown, it has been shown to be able to cleave α₂-macroglobulin,^15,16 which has a role in pulmonary defence. It is suggested that ADAM33 is involved in tissue remodelling,¹⁷ a physiological process intricately related to airway inflammation, hyper-responsiveness and airway obstruction.

The aim of this study was to test the hypothesis that SNPs in ADAM33 are associated with the severity of both AHR and airway inflammation in sputum and bronchial biopsies of patients with COPD.

METHODS AND MATERIALS

The extended version of the methods section is available in the online data repository. In all, 114 patients with COPD participating in the GLUCOLD study were included.¹⁸ Patient characteristics have been described in detail previously.¹⁸ In brief, all patients had irreversible air flow limitation and chronic respiratory symptoms, and were current or ex-smokers with at least 10 pack-years of smoking. Patients did not use a course of inhaled or oral corticosteroids within 3 months, or maintenance treatment with these drugs within 6 months before randomisation. None of the patients had a history of asthma. The medical ethics committees of the Leiden University Medical Centre (Leiden, The Netherlands) and the Groningen University Medical Centre (Groningen, The Netherlands) approved the study, and all patients gave their written informed consent.

Pulmonary function tests have been described previously.¹⁸ Spirometry and reversibility to salbutamol were measured. Methacholine challenge tests were performed with the 2-min tidal breathing method¹⁹ and expressed as the provocative concentration of methacholine causing a fall in forced expiratory volume in 1 s (FEV₁) of 20% (PC₂₀).

Sputum induction and whole-sample processing were performed according to a validated technique²⁰ as described in detail previously.¹⁸

Details on biopsy collection, processing and immunohistology have been published previously.²¹ In brief, we collected the two best biopsies out of four paraffin-wax-embedded biopsies per patient, and used specific antibodies against T lymphocytes (CD3, CD4 and CD8), macrophages (CD68), neutrophil elastase, mast cell tryptase (AA1) and eosinophils (EG2).

DNA was extracted from peripheral blood. Genotyping was performed as described previously¹¹ using primers and probes from Applied Biosystems TaqMan SNP Genotyping Assays (Nieuwekerk aan de IJssel, Dordrecht, The Netherlands).

Figure 1⇓ shows the eight SNPs in ADAM33 genotyped: F+1 (G/A), Q-1 (C/T), S_1 (Val–Iso), S_2 (G/C), ST+5 (A/G), T_1 (Met–Thr), T_2 (Pro–Ser), V_4 (C/G). We based the selection of the SNPs on previous associations with AHR, excess decline in FEV₁, and/or presence of COPD.^{9,11,12,22,23}

• Download figure
• Open in new tab
• Download powerpoint

Figure 1

 The exon–intron structure of a disintegrin and metalloprotease 33 (ADAM33; the eight genotyped single nucleotide polymorphisms are indicated above the gene) and the domain organisation of ADAM33. EGF, epidermal growth factor; Prodomain, proteinase domain.

We used Arlequin (V.2.000) to test whether SNPs were in Hardy–Weinberg equilibrium and linkage disequilibrium. We investigated whether ADAM33 SNPs are associated with PC₂₀ and the number and subset of inflammatory cells in sputum and bronchial biopsies. Our primary genetic model for all SNPs was a dominant model (homozygotes and heterozygotes for the minor allele being compared as a group with homozygotes for the major allele). In addition, SNPs with a minor allele frequency of ⩾0.30 were entered in (1) a codominant model (three genotype groups per SNP separately) and (2) a recessive model (homozygotes and heterozygotes for the major allele being compared as a group with homozygotes for the minor allele). PC₂₀ and inflammatory cells in sputum and bronchial biopsies were log transformed to obtain a normal distribution. We performed univariate analyses using the t test and ANOVA. Multiple linear regression analyses were performed to investigate the association of polymorphisms in ADAM33 with PC₂₀, sputum inflammatory cells and inflammatory cells in biopsies as dependent variables. Independent variables included in the model were sex, smoking status, lung function, genotype and the interaction of smoking status and genotype. To determine whether the regression estimates (with standard errors) and p values were not due to chance, we performed bootstrapping by resampling rows of the original data frame (using n = 5000 bootstrap replicates) on the main significant outcome variables.

RESULTS

Results of univariate analyses and multiple regression analyses assuming a recessive model are shown in the online data supplement. Multiple regression analyses assuming a dominant and codominant model are presented below. The low number of patients per haplotype did not allow haplotype analysis.

Multivariate association of ADAM33 SNPs with AHR

Dominant model: patients with a G-allele for SNP ST+5 had a significantly higher PC₂₀ compared with the AA genotype (geometric mean (GM) 0.61 v 0.27 mg/ml, p = 0.04). SNPs F+1, Q-1, S_1, S_2, T_1, T_2 and V_4 were not significantly associated with the level of AHR.

Codominant model: patients with the GG genotype for SNP ST+5 had a significantly higher PC₂₀ compared with the AA genotype (fig 2A⇓). SNPs F+1 and S_2 were not significantly associated with the level of AHR.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

 Association of a disintegrin and metalloprotease 33 single nucleotide polymorphism (SNP) ST+5 with airway hyper-responsiveness (provocative concentration of methacholine causing a decrease in forced expiratory volume in 1 s (FEV₁) of 20% (PC₂₀ methacholine)), sputum inflammatory cells and CD8 cells in bronchial biopsies in patients with chronic obstructive pulmonary disease. (A) PC₂₀ methacholine per SNP ST+5 genotype. (B) Total sputum cell count per SNP ST+5 genotype*. (C) The number of CD8 cells in bronchial biopsies per SNP ST+5 genotype. *Total sputum cell count refers to the total non-squamous cells in induced sputum. Solid squares represent the geometric mean; vertical bars represent the 95% CI. Different numbers for the SNP genotypes are due to missing genotype data.

Multivariate association of ADAM33 SNPs with inflammatory cells in bronchial biopsies

Dominant model: patients with a minor allele for SNPs T_1 and T_2 had a significantly lower number of CD8 cells than the wild type (T_1: p = 0.02; T_2: p = 0.01). We present only the results of SNP T_2 in fig 3A⇓ given the fact that SNPs T_1 and T_2 are in complete linkage disequilibrium (p = 10⁻¹⁶; see online data repository, fig E1) and the association of these SNPs with the number of CD8 cells in bronchial biopsies was similar.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

 Association of single nucleotide polymorphisms (SNPs) in a disintegrin and metalloprotease 33 with the number of CD8 cells in bronchial biopsies in chronic obstructive pulmonary disease. (A) Per genotype for SNP T_2. (B) Per genotype for SNP S_2. Solid squares represent the geometric mean; vertical bars represent the 95% CI. Different numbers for the SNP genotypes are due to missing genotype data.

Codominant model: patients with the GG genotype for SNP ST+5 had significantly lower numbers of CD8 cells in bronchial biopsies than the AA genotype (fig 2C⇑). Homozygotic individuals for the minor allele of SNP S_2 had significantly lower numbers of CD8 cells in bronchial biopsies compared with the wild type (fig 3B⇑) and lower numbers of plasma cells (S_2 GG: GM (95% CI) 13.5/0.1 mm² (8.9 to 20.7); GC: 11.8/0.1 mm² (7.7 to 18.2), p = 0.54; CC: 6.5/0.1 mm² (3.3 to 12.9); p = 0.04). Heterozygotic patients for SNP S_2 had significantly lower numbers of CD4 cells compared with the wild type (S_2 GG: GM (95% CI) 28.5/0.1 mm² (19.9 to 41.0); GC: 21.8/0.1 mm² (16.9 to 28.2), p = 0.04; CC: 25.6/0.1 mm² (17.0 to 38.3); p = 0.59). No significant associations were found with SNP F+1.

We found no significant associations between SNPs in ADAM33 and the number of mast cells or eosinophils in bronchial biopsies.

Results of bootstrapping confirmed the above-presented main outcome results, with comparable and significant p values.

DISCUSSION

The important message of this study is that SNPs in ADAM33 are associated with the pathophysiology of COPD. Patients with the AA genotype for SNP ST+5 had more severe AHR, higher numbers of sputum inflammatory cells and higher numbers of CD8 cells in bronchial biopsies than those with the GG genotype. Moreover, individuals with the minor allele of SNP T_1 and T_2, and homozygotic individuals for the minor allele of SNP S_2 had significantly lower numbers of CD8 cells in bronchial biopsies, cells relevant to the pathology of COPD.

Prevalences of the ADAM33 SNPs found in our COPD population are comparable to those recently reported by van Diemen et al.¹¹ We also confirm their findings that minor alleles for SNPs F+1, S_1 and S_2 are more prevalent in patients with COPD than in healthy participants. Regarding SNP Q-1 we showed a significantly higher prevalence of the minor allele in patients with COPD compared with healthy controls, whereas van Diemen et al¹¹ showed a trend in the same direction (see online data repository). In addition, we found a higher prevalence for the A-allele in SNP ST+5 in patients with COPD. As far as we know, this is the first study suggesting that a COPD susceptibility gene is also associated with the pathophysiological process in COPD.

ADAM33 is a member of the ADAM family, a group of membrane anchored metalloproteases that contain a disintegrin and a metalloprotease domain and is expressed in multiple tissues, including lung tissue.²⁴ As its function has not been unravelled yet, we can only speculate about its role in COPD. It is conceivable that ADAM33 has a role in both remodelling and inflammation by shedding growth factors, cytokines and their receptors from the cell surface. If a SNP alters ADAM33 function to increase protein production, this might result in enhanced airway inflammation. Alternatively, when ADAM33 has a diminishing effect on the release of proinflammatory cytokines, inflammation would increase by a decrease in ADAM33 production or function. The same could be true for the role of ADAM33 in airway remodelling, in case a change in ADAM33 results in an increase in growth factors resulting in proliferation of airway smooth muscle cells and lung fibroblasts. Whatever the functions of ADAM33 turn out to be, our data suggest that it is associated with inflammation and AHR.

We showed an association between ADAM33 and the severity of AHR in COPD. AHR is important to COPD given its association with accelerated FEV₁ decline²⁵ and increased risk of COPD mortality.²⁶ The exact pathophysiology underlying AHR is unclear, but it is thought to result from an inflammatory process in the airways in addition to geometric changes owing to airway remodelling. As discussed above, it is possible that ADAM33 has a role in both these processes thereby contributing to the severity of AHR.

ADAM33 SNP ST+5 is associated with the total sputum cell count in our COPD population. When we replaced this by numbers of sputum neutrophils or macrophages, we found significant associations with SNPs S_2, F+1 and Q-1 but not with the other SNPs. This suggests that not one specific cell type was predominantly accounting for the association of total cell count with SNP ST+5. Several studies have shown an increase in the number of neutrophils and macrophages and concentrations of proinflammatory cytokines like interleukin 8 and tumour necrosis factor α in induced sputum of patients with COPD.^27–,²⁹ Intuitively, one would think that the severity of AHR is associated with the degree of inflammation and, in fact, we previously did show an independent, positive association between severity of AHR and total sputum cell counts in these patients with COPD³⁰; perhaps ADAM33 is the missing link.

O’Shaughnessy et al³¹ showed an increased number of neutrophils and CD8 lymphocytes in bronchial biopsies of smokers with air flow limitation, and this increase was inversely associated with the level of lung function. We, therefore, hypothesised that individuals with a genetic predisposition for a higher number of CD8 cells were more susceptible to a further increase in CD8 cells, which might finally result in air flow limitation. Interestingly, we found an association of SNP ST+5 with both the presence of COPD (see online data repository) and the number of CD8 cells in our patients with COPD.

ADAM33 has previously been associated with asthma in some,^9,22,23 but not all studies.^32,33 We would like to emphasise that we are confident that our patients with COPD genuinely have COPD and no asthma given the fact they all had moderate to severe air flow limitation after bronchodilation, ⩾10 pack-years of smoking, and no history of doctor-diagnosed asthma. We selected eight SNPs in ADAM33 based on previous literature. One could argue that we did not genotype all known SNPs in ADAM33. However, we did find consistent and significant associations between the genotyped SNPs and the severity of AHR and airway inflammation in our COPD population, indicating linkage of ADAM33 with pathophysiological features of COPD.

A potential criticism of our study is that the sample size of 111 patients with COPD is relatively small. From a practical point of view, biopsy studies are very demanding and the issue of sample size would especially be of concern in case we did not find any associations. However, we did find significant associations between ADAM33 and features of COPD. We set out with the a priori hypothesis that the same SNPs in ADAM33 previously shown to be associated with COPD are also associated with its pathophysiology. To investigate our hypothesis, we performed several analyses to assess the association of ADAM33 with hyper-responsiveness and airway inflammation and found associations of moderate significance. This has affected the interpretation of the results. One could raise the issue of multiple testing being responsible for the current results and that we should have adjusted for this in our analyses. We do not agree with this for several reasons. Firstly, the independent variables in our analyses (eg, sputum total cell count and differential cell count) are mutually related, indicating that a rigid statistical procedure like a Bonferroni correction for multiple testing would not do justice to their biologically linked nature. Secondly, we did not randomly test for associations between ADAM33 and features of COPD but had a predefined hypothesis based on previous literature.

ADAM33 is a highly polymorphic gene containing at least 58 SNPs of which we investigated eight, based on previous literature.^{9,11,12,22,23} We found an association of SNP ST+5 with the severity of AHR and airway inflammation in induced sputum and bronchial biopsies in COPD. The ST+5 SNP is an intron SNP between the S-exon (transmembrane region) and the T-exon (which includes a SH₃ domain and a phosphorylation site). This may have functional relevance, as non-coding introns can exert their effect by influencing alternative splicing, splicing efficiency or messenger RNA turnover.

We furthermore found an association of SNP S_2 with the inflammatory cell profile in sputum and the number of inflammatory cells in bronchial biopsies. A higher number of airway wall CD8 cells is associated with more severe AHR,⁶ by itself a risk factor for accelerated lung function loss.²⁵ van Diemen et al¹¹ showed an association of the minor allele of SNP S_2 with accelerated lung function decline in a general population and of the minor allele of SNP T_2 with the presence of COPD. Yet, in this study the minor alleles of these SNPs were associated with a lower number of CD8 cells. How can we reconcile these seemingly discordant findings? SNP S_2, a silent mutation, is located in the S-exon and SNP T_2 is located in the T-exon. Both SNPs are in close proximity to SNP ST+5. The significant linkage disequilibrium between SNPs S_2, ST+5 and T_2 may suggest that the part of ADAM33 involved in genetic susceptibility and pathophysiology of COPD lies in the 3′ region. Alternatively, it may still be either at a different location within ADAM33 or at an adjacent gene. Our study was not designed to discuss this question and clearly, it needs further research.

In conclusion, our study confirms ADAM33 as a COPD susceptibility gene and is the first to extend this observation by revealing an association of ADAM33 with the severity of both AHR and airway inflammation in individuals affected with COPD. These findings constitute an important step forward in linking gene polymorphisms with COPD pathophysiology, thereby possibly contributing to better future treatments for this progressive and disabling disease.