Common genes underlying asthma and COPD? Genome-wide analysis on the Dutch hypothesis

Introduction

Asthma and chronic obstructive pulmonary disease (COPD) are two common respiratory diseases. Their estimated prevalence ranges from ∼1% to 18% in different countries [1–3]. Both diseases may lead to airway obstruction, which is reversible in asthma but not in COPD. However, the diagnosis cannot rely on reversibility as it can disappear with asthma progression, making both asthma and COPD harder to distinguish. The immune mechanisms underlying the two diseases are thought to be very different, but similarities in inflammatory processes have recently been reported in both disease entities [4]. Classically inflammation in asthma is represented by elevated numbers of CD4+ lymphocytes and eosinophils, while in COPD there are CD8+ lymphocytes, macrophages and neutrophils [5]. However, severe asthma can be accompanied by neutrophilia [6] and COPD exacerbation by eosinophilia [7].

Over 50 years ago, the so called “Dutch hypothesis” was formulated by Orie et al. [8] stating that asthma and COPD are two features of one disease entity, referred to as chronic nonspecific lung disease (CNSLD). CNSLD was defined to result from the interplay of endogenous factors like genetic predisposition, and exogenous factors like viral infections, air pollution, tobacco smoking and allergen exposures. The timing of this interplay would then determine which clinical syndrome developed during a lifetime, i.e. asthma or COPD or features of both asthma and COPD.

So far this hypothesis has neither been confirmed nor refuted completely [9], but several common environmental exposures have been unequivocally identified as shared risk factors for both asthma and COPD, e.g. maternal smoking during pregnancy, air pollution and active smoking [10]. Genetic factors have been associated with either asthma or COPD using linkage [11–15], candidate gene [16–19] and genome-wide association studies (GWAS) [20, 21]. These studies elucidated genetic factors unique either to asthma or COPD, but in addition potentially shared genetic risk factors including TGFB1, TNFA, GSTP1, IL13 [22] and SERPINE2 [23]. ADAM33 has been linked to the presence of asthma [24], COPD and accelerated lung-function decline in the general population and in asthma [25, 26], suggesting common underlying genetic factors for both onset and course of asthma and COPD. So far, hypothesis-free GWAS studies that aim to identify novel genes underlying both asthma and COPD in the same source population are lacking. The aim of our study was to identify shared genetic-risk factors for asthma and COPD using an unbiased GWAS approach. We first performed a GWAS on asthma and COPD separately using individuals from Dutch descent and subsequently combined these in a meta-analysis, followed by three replication studies.

Results

GWAS and meta-analyses

GWAS were performed on both asthma (921 cases and 3246 controls) and COPD (1030 cases and 1808 controls). The genomic inflation factors (λ) were 1.01 for both asthma and COPD, indicating no population stratification (fig. S1). Individual p-values and odds ratios (ORs) were combined in a directional meta-analysis using a fixed-effects model (meta-analysis 1, fig. 1]). All 2048 SNPs with p≤0.001 were selected for a first-replication phase analysis in asthma (534 cases and 2568 controls) and COPD (711 cases and 1854 controls) cohorts separately. Subsequently results were combined in a meta-analysis (meta-analysis 2, fig. 1).

20 SNPs replicated at p<0.001 (table 2) in the combined meta-analysis 1 and meta-analysis 2, one SNP reached genome-wide significance.

19 of the 20 SNPs map to three genomic locations: 2p24.3, 5q23.1, and 13q14.2 (table S2).

The chromosome 2p24.3 locus spans ∼380 kb and contains genes encoding functional units, like processed transcripts, pseudogenes and RNA genes (fig. 2). The nearest gene with a known function, DEAD-box polypeptide 1 (DDX1), is ∼139 kb away from the top associated 2p24.3 SNP rs1477253. The locus on chromosome 5 is ∼328 kb and contains a single gene: COMM domain containing 10 (COMMD10) (fig. 2). The locus on chromosome 13 spans ∼320 kb and only contains a pseudogene: guanine nucleotide binding protein (G protein), gamma 5 pseudogene 5 (GNG5P5) (fig. 2). SNP rs9534578 in GNG5P5 reached genome-wide significance (p=9.96×10⁻⁹).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2–

Regional association plots for loci a) DDX1, b) COMMD10 and c) GNG5P5. The plots were generated using R and regional association plot script from the BROAD institute (Cambridge, MA, USA).

Second replication phase of top 20 SNPs

The top 20 markers from the combined analysis were further evaluated in an independent sample of the LifeLines cohort (LifeLines 2) and the SAPALDIA, RS-I, RS-II, RS-III, MESA and ARIC cohorts. Full details of subject numbers are given in table 1. None of the SNPs replicated at a nominal p-value <0.05. The meta-analysis of all cohorts together did not result in GWSA (table 2 and fig. 3).

View this table:

• View inline
• View popup

Table 1– Characteristics of the identification and replication cohorts

View this table:

• View inline
• View popup

Table 2– Top 20 single nucleotide polymorphisms (SNPs) resulting from the identification from the first meta-analysis (meta 1) and the first replication phase (meta 2)

• Download figure
• Open in new tab
• Download powerpoint

Figure 3–

Forest plots of the three top single nucleotide polymorphisms (SNP) in the meta-analysis of the asthma and chronic obstructive pulmonary disease cohorts. a) SNP rs9534578 in GNG5P5. b) SNP rs1477253 in DDX1. c) SNP rs254149 in COMMD10. SAPALDIA: Swiss Cohort Study on Air pollution and Lung Diseases in Adults; RS: Rotterdam Study; MESA: Multi-Ethnic Study of Atherosclerosis: ARIC: Atherosclerosis Risk in Communities Study; Meta fixed: meta analysis with fixed effect. ^#: includes the Dutch–Belgian Randomised Lung Cancer Screening (NELSON) trial and the Dutch Asthma Genome-wide association studies (DAG).

SNPs in the DDX1 and COMMD10 loci were associated with both asthma and COPD (table S3). The meta-analysis results of the GNG5P5 locus were driven by the association with the COPD phenotype, since none of the GNG5P5 SNPs were significantly associated with the asthma phenotype.

eQTL analysis of top 20 SNPs

Three of the top 20 SNPs from the combined analysis showed a cis-eQTL effect, when correlating the genotypes with gene expression levels in 1469 peripheral blood mononuclear cell samples with both GWAS and genome-wide gene expression data available [29]. The three SNPs were located in COMMD10. Figure 4 shows that the risk guanine (G) allele and SNP rs10043228 thymine (T) is in perfect linkage disequilibrium (r²=1) with rs10036292, increased COMMD10 expression levels in blood mononuclear cells, with similar findings to those found in lung tissue.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4–

Expression quantitative trait loci (eQTL) identified for COMMD10 single nucleotide polymorphisms (SNPs) rs10036292 in a) blood (p=3.68×10⁻⁴) and b) lung tissue (meta-analysis p=5.24×10⁻⁹). Order on x-axis is from nonrisk homozygote to heterozygote and risk homozygote. Note in the lung tissue dataset, the risk homozygotes were not present.

Network analysis

The genes found were investigated with GeneMANIA, which does not support pseudogenes. Hence we queried only COMMD10 and DDX1. This gene enrichment approach resulted in a set of genes, two genes (RAD50 and MRE11A) being involved in regulation of mitotic recombination (Bayes factor 11, p<0.0001) and telomere maintenance (Bayes factor 6, p<0.0001), possibly implicating COPD as a disease of rapidly aging lungs [35]. Another gene involved in telomere maintenance (BICD1) was previously reported in emphysema [36].

Moreover, products of DDX1 and COMMD10 interact with nuclear factor (NF) κβ2. COMMD10 has a direct interaction, while DDX1 interacts with RELA and RELB, known to interact directly with NF-κβ2 and to function in the same pathway (fig. 5).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5–

Gene enrichment plot using DDX1 and COMMD10 genes as a query. CSTF2: cleavage stimulation factor, 3' pre-RNA, subunit 2, 64kDa; HNRNPK: heterogeneous nuclear ribonucleoprotein K; MRE11A: MRE11 meiotic recombination 11 homolog A (Saccharomyces cerevisiae); RAD50: RAD50 homolog (S. cerevisiae); NBN: nibrin; MTHFSD: methenyltetrahydrofolate synthetase domain containing; NFKB2: nuclear factor of kappa light polypeptide gene enhancer in B-cells 2 (p49/p100); CDK5: cyclin-dependent kinase 5; SETD3: SET domain containing 3; DAB2IP: DAB2 interacting protein; NAGLU: N-acetylglucosaminidase, alpha; CEP250: centrosomal protein 250kDa; SAP30L: SAP30-like; ROGDI: rogdi homolog (Drosophila); SAP30: Sin3A-associated protein, 30kDa; SBDS: Shwachman-Bodian-Diamond syndrome; COMMD1: copper metabolism (Murr1) domain containing 1; RELB: v-rel reticuloendotheliosis viral oncogene homolog B; RELA:v-rel reticuloendotheliosis viral oncogene homolog A (avian); ATP2A2: ATPase, Ca++ transporting, cardiac muscle, slow twitch 2; COMMD10: COMM domain containing 10; DDX1: DEAD (Asp-Glu-Ala-Asp) box polypeptide 1.

Discussion

This is the first investigation of shared genetics for asthma and COPD in a hypothesis-free manner using a genome-wide screening in asthma and COPD in large population-based cohorts. We report three novel loci as potentially shared genetic factors between asthma and COPD, none reaching genome-wide significance in the discovery sample or seven replication cohorts. None of these three loci were previously reported to be associated with either asthma or COPD. However, DDX1 locus was reported in a recently published meta-analysis of lung function [37], a p-value of 9×10⁻⁶. The T allele of rs2544527 in DDX was associated with a reduced lung function and in our study with a risk for both asthma and COPD.

The shared 5q23.1 risk locus contains the COMMD10 gene. COMMD10 is a member of COMM domain containing proteins [38] with a largely unknown function. COMMD10 has been shown to form a complex with COMMD1, another member of this family of proteins, which regulates copper metabolism and sodium uptake and inhibits NF-κβ activation [39]. Copper and sodium levels are inversely regulated, i.e. when copper levels increase, sodium import in cells is inhibited and vice versa. Both ion levels can be regulated by COMMD1, with sodium control mediated through epithelial sodium channels (ENaCs) that are abundantly present in lung epithelial cells [40]. Sodium is crucial for maintaining a fluidic layer in the alveolar part of the lungs and ENaCs play a crucial role in this process [41]. It is tempting to speculate that COMMD10 is involved in this maintenance either through interaction with COMMD1, or independently by displaying similar functions as COMMD1. Also, its function in inhibition of NF-κβ activation could play a role in regulating inflammatory processes in airways diseases. Our eQTL studies support a functional role of COMMD10, since we established that two SNPs in the COMMD10 region influence expression of this gene in both blood cells and lung tissue.

The 13q14.2 locus contains the guanine nucleotide binding protein (G protein) (GNG5P5). Poliseno et al. [42] recently showed that pseudogenes can have a pronounced role in regulation of their putative transcripts by competing in noncoding RNA binding. It needs to be tested whether GNG5P5 can affect GNG5 levels, but it is interesting to note that the pseudogene is processed and has a transcript (ENST00000420444). The biological consequence of a change in GNG5 levels in relation to asthma and COPD pathology is unclear but it is well established that G proteins play a crucial role in signal transduction from cell surface to its interior. It is also known that G-protein coupled receptors (GPCRs) are involved in asthma and more generally are a target of many of the currently used asthma drugs [43].

A third locus on 2p24.3 is bordered by the DDX1 gene, encoding DEAD-box protein 1, RNA helicase I, and the MYCN genes whereas the locus itself contains nonprotein-coding genes including lincRNAs, ncRNAs, pseudogenes, processed transcripts and one newly discovered, protein-coding gene. Theoretically, any of these could be involved in asthma and COPD, hindering interpretation of our findings. However, the regional association plot (fig. 2) shows that the signal is mostly confined to AC008278.3 and AC008271.1. Further refinement of the region and functional assessment of the associated variants could help to potentially pin-point the actual causal gene. DDX1 is a plausible candidate for both asthma and COPD since it interacts with RELA, one of NF-κβ subunits, upon which it acts as a co-activator of NF-κβ mediated transcription [44]. Since this is a central and common pathway of inflammation present in the airways of both asthma and COPD, this may signify a unifying underlying mechanism of both disease entities. Further studies are needed to confirm this hypothesis.

The strengths of our study are the data quality of the cohorts involved, the design of the study and the analysis strategy of the discovery and replication phases. There are some limitations to our study as well. We found no overall replication in six out of eight replication cohorts. One explanation for the lack of replication might be the differences in asthma and COPD patients in the replication cohorts compared with the identification cohort. For instance there was a somewhat lower prevalence of asthma in LifeLines 2 (7.5% versus 8.5% in LifeLines 1) due to the average increased age of the subjects included in LifeLines 2. This could reflect a cohort effect or some asthma remission for the elder ages [45]. Furthermore, most studies used an asthma definition of self-reported asthma diagnosis. Self-reported asthma has led to firm GWAS findings in the GABRIEL study (a multidisciplinary study to identify the genetic and environmental causes of asthma in the European community) [46]. However, it cannot be excluded that our asthmatic groups consisted in part of individuals diagnosed with asthma in childhood, who now are in complete remission. The GABRIEL cohort studies suggested that the genetic background of early-onset and adult-onset asthma is different. It would be of interest to assess whether COPD would have more overlap in genetic background with either childhood-onset than adult-onset asthma. A previous study from our group [47] showed overlap between candidate genes for COPD and early childhood wheeze and lower lung function, suggesting there is some overlap in genetic background in early childhood characteristics. This clearly needs further study, since we could not analyse this adequately in our cohort, where the prevalence of childhood asthma was 82% in our identification cohort and 41 in the verification cohort. Similarly, the diagnosis of COPD was based on lung function only, and this could have led to inclusion of different types of COPD in the various replication cohorts. For instance the prevalence of never-smokers was 41% in SAPALDIA, whereas this was 0% in the identification and LifeLines 1 and 2 cohorts and ranged from 10% to 24% in the other cohorts. Furthermore some cohorts were consisted of subjects that were of an increased age (e.g. mean age ∼65 years in RS-I and RS-II and this may have led to inclusion of elderly asthmatics in the COPD group, since significant persistent airway-obstruction may occur in asthma with increase in age [48]. This may reflect an important limitation common to most GWAS, i.e. the heterogeneity of the phenotypes assessed and heterogeneity between discovery and replication samples. Table S3 shows the heterogeneity per meta-analysis performed, i.e. for each asthma–COPD meta-analysis. It differs substantially and due to specificity of the study we could not account for the heterogeneity between meta-analyses. We did not find as prime hits a gene that was associated with asthma and with COPD previously. For instance ADAM33 was not significantly associated with either asthma or COPD or represented in their overlap. This may either be due to the fact that not all SNPs were captured in the GWAS analyses, or that ADAM33 was only found by positional cloning when hyperresponsiveness was present in asthmatics [49]. The latter was not a prerequisite in our asthma definition, just as in other GWAS studies, where ADAM33 was also not found as a significant gene associated with asthma.

Do our findings then refute the Dutch hypothesis? This hypothesis states that both genetic and environmental factors contribute to the phenotypic outcome and that there is a common genetic background. Indeed the current study did not find significant genetic similarities between asthma and COPD, apart from the identification cohort and LifeLines 1. As highlighted by the Dutch hypothesis the importance of both type and temporal sequences of environmental exposures contribute to the occurrence of either phenotype. This may have affected the phenotypic outcome considerably and, hence, a crude covariate adjustment may represent an underestimated challenge to identify common genetic determinants of asthma and COPD. Finally, our study has power to identify strongly prevalent SNPs, yet not rare variants that may have an impact on asthma and COPD. Our findings either suggest that there is no common genetic component in asthma and COPD or, alternatively, different environmental factors, like lifestyle and occupation in different countries and continents may have obscured the genetic common contribution.

Recent efforts to characterise the substantial number of patients diagnosed with both asthma and COPD [50] show the increasing scientific interest in the phenotypic overlap between asthma and COPD. Future studies on the underlying genetics in this group of overlap patients would be of interest, specifically comparing outcomes with our results.

Overall, our results may suggest a role of the NF-κβ pathway, a key transcription factor in the inflammatory response, in both asthma and COPD, suggesting that the Dutch hypothesis may have some validity. However, we could not replicate associations in both asthma and COPD in most replication cohorts, thus this could refute the genetic background that the Dutch hypothesis implied to be common in asthma and COPD. Further studies including lifelong lifestyle factors across all cohorts need to be performed to assess whether this approach elucidates a common genetic background of asthma and COPD. Since none of the SNPs reached genome-wide significance further investigation of the loci should be performed to assess their role in both asthma and COPD. Although inflammatory processes differ in asthma and COPD, they are unequivocally mediated by NF-κβ, and as suggested by our current results, they could be driven by the same underlying genes, COMMD10 and DDX1. Our eQTL studies support a functional role of COMMD10, since we established that two SNPs, therefore, the natural next step is to perform genome-wide epistatic analysis in large cohorts of asthma and COPD patients to reveal the complex nature of interactions between SNPs and loci and their impact on the ultimate phenotype.