Background

Chronic obstructive pulmonary disease (COPD) is a multi-component respiratory disease with recognized systemic impact [1]. Numerous studies performed in recent years provide overwhelming evidence of COPD as a condition characterized by an abnormal inflammatory response beyond the lungs with evidence of low-grade systemic inflammation [2–5]. Raised levels of acute phase proteins like C-reactive protein (CRP), fibrinogen and pro-inflammatory cytokines such as interleukin (IL)-6 were found in circulation of stable COPD patients [3, 6] and have been shown to be associated with impaired functional capacity [7], reduced daily physical activity [8] and decreased health status [5, 7, 9]. However, given the cross-sectional nature of most studies performed so far and possible confounding by a number of lifestyle factors associated with levels of inflammatory biomarkers [10], it is not clear whether these proteins are simply markers of the inflammatory process accompanying chronic diseases such as COPD or key players in the pathogenesis of disease.

Genome-wide scans, twin and family studies have shown that circulating levels of CRP, fibrinogen and IL-6 are heritable (estimated as 25%–40%) [11–15]. Furthermore, recently CRP and FGB polymorphisms/haplotypes have been described that may partly explain heritability of acute-phase protein and cytokine levels [16–19]. Genetic association testing of genotypes, which influence circulating levels of proteins and directly relate to the outcome of interest, was suggested as more accurate unconfounded estimate of whether systemic inflammation levels causally influence outcome [20].

In the present study we investigate whether common haplotypes in CRP, IL6 and FGB (encoding fibrinogen β chain) genes influence systemic inflammatory status in COPD, the risk for COPD and, eventually, different disease-related phenotypes. Some of the results of this study have been previously reported in the form of an abstract [21].

Methods

Study participants

The investigation was designed as a case-control association study, consisting of unrelated individuals recruited from the same geographical area (Limburg province, the Netherlands). A total of 556 Caucasian subjects were investigated. All subjects were current or former smokers. Three hundred and sixty-one patients with clinically stable moderate-to severe COPD entering pulmonary rehabilitation (Center for Integrated Rehabilitation of Organ failure (CIRO), Horn, The Netherlands) were enrolled for the study. Clinical history of COPD and the degree of the disease severity were assessed according to the published Global Initiative for Chronic Obstructive disease (GOLD) guidelines [22]. One hundred and ninety five healthy (ex-) smokers were recruited as controls. The healthy control subjects were volunteers recruited through advertisement in a local newspaper. Part of the healthy controls were also recruited through the COSMO study [23]. Inclusion criteria for both groups were: Caucasian origin, 40 years of age or older, smoking history of 10 pack-years or more, completed spirometry and blood sample donation. The ethical review board of the University Hospital Maastricht approved the study, and all subjects gave their written informed consent.

Clinical examination and inflammation measurement

Lung function was determined using spirometry, height and weight were measured in every participant and body mass index was calculated. Several COPD-related clinical characteristics were assessed only in COPD patients based on standard procedures (see Additional file 1). Plasma levels of CRP, IL-6 and fibrinogen were measured by high-sensitivity particle-enhanced immunoassay, ELISA and coagulation reaction respectively. Further details are provided in Additional file 1.

TagSNP selection and genotype determination

TagSNPs were selected for genotyping from the SeattleSNPs database http://gvs.gs.washington.edu/GVS/ using resequencing data from 23 unrelated European Americans. Polymorphisms with a minor allele frequency of less 5% were not included. Six polymorphisms in CRP, 8 SNPs in IL6 and 6 SNPs in FGB were selected. For two non-redundant SNPs of FGB the development of the genotyping assay failed (rs2227432, rs2227439), leaving 4 tagSNPs for the analysis. For further details on DNA processing and genotyping see Additional file 1.

Statistical analysis

Descriptive statistics and baseline comparisons

Variables are presented as proportions (percentage), mean ± standard deviation (SD) or median (range) depending on their measurement scale and distribution. Plasma levels of inflammatory mediators were skewed to the right. They were natural logarithmically transformed to achieve symmetric distribution. Baseline differences between groups were analyzed using Student's t test and Mann-Whitney test for continuous variables and χ2 square test for categorical variables. Prevalence of systemic inflammation has been accessed by number of inflammatory markers in the top quartile and by CRP levels. The correlations between inflammatory markers and the different clinical phenotype were estimated using Pearson's or Spearman's correlation coefficient. Multiple linear regression models were performed to search for the best predictors of CRP, IL-6 and fibrinogen levels.

Genetic association analysis

To determine the linkage disequilibrium (LD) we have calculated the D' statistics between SNPs. Hardy-Weinberg equilibrium for each tagSNP was tested by the exact χ2 statistic. SNP rs2069849 in IL6 was excluded from haplotype analysis because of low allele frequency (3%). To reduce the problem of multiple testing we selected haplotypes as major genetic variable. Differences in haplotype distributions were tested using the global statistic of Schaid under additive genetic model [24]. If the global haplotype test achieved a P-value < 0.1, haplotype-specific analysis was performed. Haplotype-specific effects were estimated using linear and logistic regression as implemented in the haplo.stats package in R. Analyses were adjusted for potentially confounding factors (see Additional file 1 for full description). P-values less than 0.05 were considered significant (two-sided).

Eight primary genetic tests were performed. First, each gene was tested for association with two major outcomes: either corresponding protein level or COPD. These comprised 6 tests. Additionally, given that IL-6 is a major regulator of hepatic production of CRP and fibrinogen, we also investigated the association of IL6 haplotypes with CRP and fibrinogen plasma levels (2 tests). To reduce the problem of spurious associations due to multiple testing, Bonferroni correction was applied for those tests (P corrected = 0.05/8 tests = 0.006).

Body-Mass Index, Airflow Obstruction, Dyspnea, and Exercise Capacity (BODE) index, 6 minute walking distance (6MWD), maximum workload at the cardiopulmonary test, Medical research council (MRC) score, body mass index (BMI) and forced expiratory volume in 1 sec (FEV1) were tested after initial 8 analyses have been performed and treated as secondary outcomes. Secondary COPD outcomes were not considered for a conservative Bonferonni correction, because of a high degree of correlation with each other.

Results

Systemic inflammation

Table 1 summarizes the demographic data and baseline characteristics of study groups. Compared with the healthy smokers, COPD patients were older, more likely to be men and had more pack-years smoking. On average, patients had moderate-to-severe airflow obstruction and clear functional exercise intolerance.

Table 1 Baseline characteristics of COPD patients and healthy controls*
Full size table

In general, COPD patients had higher baseline median levels of circulating inflammatory markers (P < 0.001) (Table 1). This difference was still significant after adjustment for age, sex, smoking status, BMI and pack-years smoked. However, the COPD population was very heterogeneous: 27.3% of COPD patients had CRP levels > 10 mg/L, which is a significantly higher proportion than in healthy smokers (5.2%), whereas about 40% of COPD patents had no evidence of systemic inflammation as assessed by CRP levels alone (< 3 mg/L) or by the number of inflammatory markers in their top quartiles (no markers in the top quartile).

Table 2 shows the comparison of clinical characteristics from patients with baseline CRP value < 3 mg/L versus > 3 mg/L. Patients from the "inflammatory" group (CRP > 3 mg/L) were older, more likely to be male, had a higher BMI, more co morbidities and worse results in the 6 minute walking test. Interestingly, besides age and sex, the 6MWD was negatively associated with all 3 inflammatory markers in the backward multiple linear regression analyses (see Additional file 2). Correlations between the inflammatory markers and the functional measures of disease are shown in Additional file 3.

Table 2 Comparison of clinical parameters between patients with initial CRP levels ≤3 mg/L versus > 3 mg/L
Full size table

Haplotype Tagging SNPs and COPD

From the data of SeattleSNPs, we selected 18 SNPs (Table 3) which together tag the 5 most common (frequency > 2.5%) haplotype groups of CRP, the 8 most common haplotype groups of IL6 and the 5 most common haplotype groups of FGB (Table 4). None of the tagSNPs showed a significant deviation from Hardy-Weinberg in a control group, and the allele frequencies were comparable to those reported in the Seattle SNP panel and other Caucasian populations and individuals of Dutch origin [18, 19, 25]. All polymorphisms within each gene showed a little evidence of historical recombination as measured by D' (see Additional file 4).

Table 3 Tagging SNPs characteristics and single SNP association analysis for COPD
Full size table
Table 4 Frequency distribution in patients and controls and global association tests for haplotypes of IL6, CRP and FGB
Full size table

Table 3 shows basic characteristics for SNPs and their allele frequency distribution in COPD patients and healthy smokers. Significant difference in allele frequency between patients and controls was found for the CRP synonymous SNP 2667G/C (rs1800947) and for the IL6 3'UTR SNP 7592G/A (rs1818879). Results of single-SNP analyses performed for all 18 SNPs for each of the major outcome are summarized in Table 3 and Additional file 5. Briefly, an increased risk of COPD was found for carriers of the minor alleles of the rs1800947 and rs1818879. For the tri-allelic CRP SNP rs3091244, association with increased CRP levels was observed. We also identified two CRP and one IL6 SNPs, which were significantly associated with decreased CRP levels (see Additional file 5).

Haplotype-based analysis, COPD and systemic inflammation phenotypes in COPD patients

Common haplotypes were predicted from genotypic data. First, global test for haplotype association were performed for all 3 genes to provide an overall test with 4 outcomes of interest (Table 4). A significant difference in haplotype distribution between patients and controls was observed for IL6 (P = 0.003). There was a trend for association with COPD for CRP haplotypes (P = 0.09). At the same time, CRP levels were the most significantly associated with CRP haplotypes (P = 0.0004) and less strongly with IL6 haplotypes (P = 0.08) (Table 4).

To investigate the cause of these associations further we calculated haplotype-specific scores and then estimated haplotype effects relative to the most common haplotype when all haplotypes were entered into regression model (Tables 5, 6). In particular, CRP haplotype H3 was associated with higher risk of being COPD and low CRP levels when tested against the most common haplotype H4 (Table 5). However, IL6 haplotype H2 was associated with even higher decrease in CRP levels (Table 6). Moreover, the IL6 haplotype H2 was differently distributed between cases and controls (6.2% versus 2.4%, P = 0.0006) and associated with almost 5 times higher risk of COPD compared to the most common haplotype H4 (OR = 4.82, 95%CI 1.64–4.18, P = 0.004) (Table 6). Thus, the H2 haplotype effect on COPD risk is stronger than the effect of the IL6 rs1818879 (OR = 1.49, 95%CI 1.08–2.04) (Table 3).

Table 5 Association of CRP haplotypes with COPD and ln(CRP)
Full size table
Table 6 Association of IL6 haplotypes with COPD and ln(CRP)
Full size table

Comparable associations were found when the total group of cases and controls was considered for analysis of systemic inflammation levels (N = 550) (data not shown).

Genetic variation and COPD-related phenotypes

Given that inflammatory markers were correlated with some disease characteristics (see Additional file 3), we further studied whether corresponding inflammatory genes are associated with important quantitative COPD phenotypes and performed the association analysis for 6 outcomes: BODE index, MRC score, BMI, FEV1, maximum workload and 6MWD. For all 3 genes, no significant association was observed when an overall test of association was applied, except for a marginal association of the CRP gene and maximum workload (Table 7). Further testing showed borderline association of the haplotype 121211 with maximum workload (P = 0.05).

Table 7 COPD-related phenotypes and global association tests (P-values) for haplotypes of IL6, CRP and FGB
Full size table

Discussion

Similar to previous studies [4, 9], we have shown that circulating levels of CRP, IL-6 and fibrinogen are increased in clinically stable COPD patients comparing to healthy (ex)smokers. Importantly, inflammatory markers distribution has shown a striking heterogeneity of COPD population in relation to the systemic inflammatory response. Furthermore, we observed a significant association between IL6 and CRP haplotypes and CRP levels in COPD patients as well as in the total study group, similar to the previous population-based and clinical studies [16, 26, 27]. Surprisingly, in our study those haplotypes were not related to any of COPD functional phenotype including BMI and 6MWD, which were the most strongly correlated with CRP plasma levels similar to the previous reports [4, 5, 9]. However, when compared to the controls, a new association of IL6 gene and COPD was identified. In particular, IL6 haplotype H2 (6.2% of cases) was associated with 5 times higher risk of COPD and significant decrease in CRP levels in COPD patients. To our knowledge, the present study is the first haplotype-based association analysis examining common genetic variation in CRP, IL6 and FGB genes on COPD risk and circulating levels of inflammatory markers in COPD patients and healthy smokers.

Although presence of systemic inflammation has been widely accepted in the last years as an essential characteristic of COPD [3, 28], our data clearly show that levels of CRP and 2 other widely studied in COPD markers, IL-6 and fibrinogen, vary significantly across the COPD population. This gives us strong evidence that not all COPD patients have a manifested low-grade systemic inflammatory response. In fact, comparable findings was reported in our previous study of CRP in relation to COPD phenotypes [7]. Further evidence of heterogeneity in relation to CRP comes from a recent intervention study [29], which has shown that COPD patients segregate for provastatin-responders (CRP > 1 mg/uL) and non-responders (CRP < 1 mg/L) based on their baseline CRP levels. Those non-responders may represent the distinct COPD subpopulation with low CRP levels which has impaired immune response to bacteria, viruses and other agents. Interestingly, our findings also suggest that low CRP levels may be associated with significantly higher COPD risk in carriers of a specific IL6 haplotype. Given that lung-expressed CRP could be up regulated by cytokines and has shown cyto-protective effect in innate immune response against bacteria and particulate matter [30–32], CRP and IL6 haplotypes should be further studied in relation to the local impairment in COPD.

In previous reports there seems to be inconsistency about the relationship involving biomarkers levels, corresponding genes and COPD phenotypes. On the one hand, significant associations between the plasma levels of CRP, IL-6 and fibrinogen and COPD phenotypes have been observed [2, 7, 8, 33]. Yet, on the other, no relationship between the polymorphisms influencing the protein levels and COPD phenotypes could be demonstrated in the current and some other smaller studies [34, 35]. Interestingly, three recent mRNA/protein profiling studies in quadriceps and vastus lateralis of COPD patients [36–38] failed to confirm a common hypothesis that increased expression of pro-inflammatory cytokines (IL-6, IL-8, IL-1 and TNF-α) induce muscle atrophy in COPD patients with muscle weakness. Several phenomena might explain differences between studies of inflammatory genes, mRNA and proteins in COPD. It is well-known that results of observational studies of inflammatory markers might be confounded because of residual association with numerous non-diseases related factors (i.e. lifestyle and socioeconomic) [10, 39]. Thus, association of CRP or IL6 variation with disease phenotype may suggest a causal relationship between CRP levels and disease [20]. However, genetic variation in CRP actually only explains a relatively small proportion of CRP levels variation (< 3%), compared with other environmental or genetic factors [16, 26, 40]. Moreover, given the fact that COPD might share systemic inflammatory phenotype with cardiovascular disease, cancer and diabetes [41, 42], which are known co morbidities of COPD, conflicting association data may be caused by different composition of the COPD populations used across different studies. Lastly, acute-phase and pro-inflammatory cytokine genes may be only modifiers of other, not widely studied, inflammatory genes, which are critically involved in COPD pathogenesis. Further hypothesis-free genome-wide high-throughput screening and gene expression profiling studies are needed to validate existing and find new candidate markers characterizing distinct clinical profiles of multi-faced COPD population.

Our study has several strengths. It is the first one that has been specifically designed to evaluate relationship between systemic inflammation and COPD on both genetic and protein level. For this purpose, both standard outcomes (COPD diagnosis and lung function) as well as functional disease-related phenotypes (exercise capacity and body composition) were taken into account. We used a tagging SNP based approach capturing all common variability across the genes of interest to investigate whether genetic variation in the genes of interest influences the risk of COPD. Next this study is based on a well-characterized ethnically homogeneous cohort of stable COPD patients and spirometry-proved healthy smokers collected from the same geographic region.

In our analysis of 3 genes in relation to 4 primary outcomes it is possible but unlikely that results were false positives obtained due to a type I error inflation. It is worth mentioning, two associations (IL6 haplotypes and COPD, global P = 0.003, CRP haplotypes and ln(CRP), global P = 0.0004) were still significant after applying the conservative Bonferroni correction. In addition, these associations remained also significant at P < 0.002 after permutation tests with at least 10000 simulations. Nevertheless, all reported associations between COPD and CRP and IL6 genes require replication in other independent studies. Similar to our results, three previous studies of the IL6 gene could not find any association between COPD and the SNPs rs1800795 and rs1800797 [43–45]. Interestingly, a new association was found with the IL6 SNP rs1800796 in COPD patients from Spain [44]. Because of a low allele frequency in Europeans (4%), this SNP was not selected for current study; however, further studies using large samples size of Dutch COPD patients are required to replicate this new association.

There are also some limitations of the present study. First, the patients with COPD included in the study were assessed to enter pulmonary rehabilitation and had predominantly moderate-to-severe stages of COPD. This could affect generalizability of our results. Next, a relatively small sample size could have underpowered some observed genetic associations. Namely, IL6 association with CRP levels and CRP association with COPD were marginally significant in the overall testing but have shown strong haplotype-specific effects in further analysis. Furthermore, as a result of the failure of 2 FGB tag SNPs for genotyping we were not able to discriminate one common (defined by rs2227439) and one rare (carrying the minor allele of rs2227432) haplotypes of the FGB gene [19]. They were non-discriminative from the FGB haplotypes 1111 and 1121, respectively.

It is also noticeable, that, although we found an increased risk of COPD for the IL6 H2 carriers, the functional SNP causing this risk still has to be identified. IL6 haplotype H2 is tagged by the minor allele of rs1818879. However, it is unlikely, that this SNP located in 3'UTR region of IL6, is functional because no association between haplotype H4 (tagged by the combination of rs2069840 and rs1818879) and COPD risk was found. Most likely, the functional SNP lies on the IL6 H2 haplotype and was not genotyped in our study. Future resequencing of IL6 haplotype H2 carriers may help to identify functional SNP(s). Given that a recent genome-wide association study [46] has found a SNP in IL-6 receptor gene as one of the top associated with six lung function phenotypes and the association we found between COPD and IL6 gene, it would be worthwhile to study IL-6 pathway further [47].

Conclusion

In conclusion, CRP and IL6 haplotypes were shown to be associated with systemic inflammation and COPD but not with exercise capacity, dysnoea, BMI and BODE index. Future studies are needed to replicate the observed associations in large well-defined COPD cohorts.