Introduction

Chronic Obstructive Pulmonary Disease (COPD) is one of the foremost causes of chronic morbidity and mortality worldwide. Airway inflammation is central to the development and progression of COPD, leading to destruction of lung parenchyma, goblet cell hypertrophy, and tissue remodeling. COPD is also associated with systemic effects in some patients, including cachexia, skeletal muscle wasting, and increased comorbidity. These effects may be caused or sustained by an enhanced systemic inflammation seen in some patients with this condition.

Proinflammatory mediators are currently being investigated both to improve our understanding of this disease and to provide targets for therapies. Two cytokines that have been subject to investigation are TNFα and IL-1β. During exacerbations of COPD, they are increased in airway secretions [1, 2] and correlate significantly with other inflammatory mediators and cells when stable [3]. Transgenic and KO mice models have provided insights into the effects of either cytokine [46] and in vitro, IL-1β expression in COPD neutrophils correlates with disease severity [7]. There is increased interest in utilizing these biomarkers in interventional studies and their receptors as specific anti-inflammatory therapies in COPD.

Studies have not clarified the relationships between these cytokines and their respective antagonists in health compared with COPD nor the relationships between these mediators in plasma and airway secretions in disease. This would be important to provide insight into the regulation and role of these proteins in COPD. TNFα is associated with increased apoptosis of skeletal muscle and weight loss in some patients [8, 9], but it is unclear whether these systemic changes are driven by pulmonary inflammation or reflect an independent inflammatory phenotype [10].

The aims of the present study were, therefore, fourfold. Firstly, we compared systemic (plasma) concentrations of both cytokines and their relevant antagonists (for IL-1β; IL-1 receptor antagonist (IL1-RA) and IL-1 soluble receptor 2 (IL-1sRII): for TNFα; TNFα soluble receptors 1 and 2 (TNFsR1, TNFsR2)) in COPD and in age-matched healthy controls to assess any differences. Secondly, we assessed whether there is an imbalance between the proinflammatory moieties and their antagonists in COPD, both at the systemic and pulmonary level. Thirdly, we investigated whether circulating and sputum levels of endogenous IL-1β, TNFα, and their antagonists correlated with parameters of disease severity. Finally, we assessed the relationship between sputum and plasma cytokine concentrations in COPD to determine whether there is direct evidence of an “overspill” of either mediator from lung to plasma.

Methods

Study Subjects

The patient group consisted of subjects with moderate to severe COPD defined by GOLD criteria [11]. They were daily sputum producers and current or ex-smokers, aged between 50 and 80. All patients were clinically stable for at least 8 weeks prior to recruitment with no changes in medication during this time. Alternative and concomitant lung disease was excluded clinically and physiologically, and by high resolution computed tomography, they had no other significant medical conditions. Healthy controls were matched for age and gender, had never smoked, medication free, and had no significant medical history.

Study Design

Patients were asked to complete daily diary cards (validated and described in 12) to record any changes in symptomatology. The variability of cytokine concentrations in airway secretions is high [3]; thus, it is either necessary to study a large number of patients or a smaller number of well-defined subjects studied on several occasions. We adopted the latter approach and both COPD and control subjects were seen on five consecutive days over 1 week. At each visit, subjects were examined, daily diary symptom scores were noted, and samples of blood (all subjects) and spontaneous sputum (COPD patients only) were collected. The 5-day mean concentrations of cells, mediators, and antagonists were calculated for each individual. Spirometry was assessed on days 1 and 5.

Sample Collection and Processing

At each clinic visit, a 10-mL blood sample was collected from all subjects in both the COPD and healthy control groups and used to assess total and absolute cell counts of neutrophils, eosinophils, lymphocytes, and monocytes [13]. A plasma sample was obtained by centrifuging the blood at 3,000 rpm for 10 min at 4°C. This was stored in aliquots at −70°C and subsequently used to measure IL-1β, IL-1RA, IL1sRII, TNFα, TNFsR1, and TNFsR2.

In the COPD group, spontaneous sputum samples were collected on the same day as the blood sample. Sputum samples were collected over four hours (from rising) following mouthwashing procedures to minimize salivary contamination. Sputum collection and analysis occurred at the same time on each visit. The samples were divided into three aliquots: the first was ultracentrifuged (50,000×g for 90 min at 4°C) to prepare a sol phase sample to determine mediator levels of IL-1β, IL-1β antagonists, TNFα, and TNFα antagonists. Samples were stored in aliquots at −70°C until analyzed. The second aliquot was treated with dithiothrietol and used to assess total cell numbers [14], and cytospins were prepared for differential cell counts (squamous cells, neutrophils, eosinophils, macrophages, and lymphocytes). The third aliquot was used for microbiological analysis.

Mediators were measured using Enzyme Amplified Sensitivity Immunoassay (R&D Systems, Abingdon, UK) and are expressed in molar concentrations. All assays were validated as described previously [13] to determine their working range, the variability of mediator measurements, and the spike recovery [15]. The TNFα and IL-1β assays measured both free and bound cytokine.

Statistical Analysis

Data analysis was performed using SPSS 16.0 for Windows (SPSS, Chicago, IL, USA). Numbers of subjects included in the study were based on power calculations using the known variability of TNFα and IL-1β concentrations in COPD [3]. Normally distributed data are expressed as means, categorical data as percentages. Differences between groups were assessed using unpaired t test or Mann Whitney U test. Correlations between data sets were assessed by Pearson’s correlation coefficient (PCC). Bonferroni adjustment was applied to correct for multiple comparisons [16].

The study was approved by the local research ethics committee and all subjects gave informed consent.

Results

Baseline Characteristics

Patients and healthy subjects (15 in each group) were enrolled in the study, and their baseline characteristics are summarized in Table I. There were no exacerbations of COPD (as characterized by Anthonisen et al. [17]) before or during the study period and no significant changes in lung function, symptoms, or diary scores. All patients had mucoid sputum on enrolment and during the course of the study [12], and two patients were colonized with low bacterial numbers (<106 colony forming units per milliliter) throughout the study (one with Moxerella catarrhalis and the other with non-typeable Haemophilus influenzae), but their inflammatory mediator concentrations were similar to uncolonised COPD pateints.

Table I Demographics for Subjects

Assay Variability and Sputum Characterization

The lower limit of detection and quantification for each assay was assessed, as described previously [15]. In all cases, the intra- and inter-assay variability was less than 10% and all assays showed greater than 90% recovery of known concentrations of mediator spike. The median percentage of squamous cells in sputum samples was 0.8% (IQR = 0–9.1%) and the median percentage of viable cells (assessed by trypan blue exclusion) was 83% (IQR = 63–92%). All samples were assayed and the mean result for each subject was obtained for the five consecutive days.

Comparisons Between COPD Patients and Healthy Controls

The intra-subject CVs in plasma were low and similar to the inherent variability for the assays (<10%). Figures 1a and b show the mean plasma concentrations of both cytokines and their antagonists in both the COPD and control groups. There was no difference in the IL-1β concentrations between COPD patients and controls (mean concentrations were 0.04 ± 0.004 and 0.03 ± 0.006 pM, respectively). However, mean concentrations of IL-1RA and IL-1sRII were markedly reduced in COPD patients (IL-1RA; 0.008 ± 0.001 vs. 12.85 ± 2 pM, p = 0.001 IL-1RsII; 0.22 ± 0.006 vs. 325 ± 17 pM, p < 0.001; see Fig. 1a). In COPD, plasma IL-1β concentrations were five times lower than IL-1sRII but five times greater than IL-1RA. In the control group, plasma IL-1β concentrations were over 1,000 times lower than IL-1sRII and over 100 times lower than IL-1RA.

Fig. 1
figure 1

A comparison of mean mediator and soluble antagonist concentrations in COPD and healthy age-matched controls. All scatter plots show the mean concentration of cytokine or antagonist for each participant, calculated over five visits. COPD indicates results from COPD patients (black circles) and HC indicates results from healthy subjects (gray circles). The vertical axes are the log of the protein concentration. a Comparison of plasma IL-1β, IL-1RA, and IL-1sRII in COPD and age-matched healthy controls. There was no significant difference in IL-1β concentrations between the two groups (average difference 0.15 pM, p = 0.07). Both plasma concentrations of IL-1RA and IL-1sRII were significantly higher in healthy subjects compared with COPD subjects (p = 0.001 and p < 0.001, respectively). b Comparison of plasma TNFα, TNFsR1, and TNFsR2 in COPD and age-matched healthy controls. There were no differences in TNFα, ΤΝFsR1, or TNFsR2 concentrations between the groups

There were no differences in the mean plasma concentrations of TNFα, TNFsR1, or TNFsR2 between COPD patients and controls. Mean TNF plasma concentrations are shown in Fig. 1b and are as follows (COPD followed by control values): TNFα 0.086 ± 0.01 vs. 0.091 ± 0.01 pM; TNFsR1 20.6 ± 0.9 vs. 18.6 ± 1.0 pM; TNFsR2 25.5 ± 1.6 vs. 25.1 ± 1.7 pM. In both groups, plasma concentrations of TNFsR1 and TNFsR2 exceeded TNFα by a ratio of 200 to 1.

Corresponding Mean Systemic and Pulmonary Mediator Concentrations in COPD

Table II summarizes the mean concentrations, the standard error of the mean, and the intra-patient CV of cytokines, antagonists, and cells in corresponding sputum and plasma samples for the COPD patients. The intra-patient CVs for sputum TNFα, IL-1β, and the receptors were derived from five consecutive sputum samples and were low, being less than described for other biomarkers using alternative sampling methodologies [18, 19].

Table II The Concentrations of Mediators and Cell Counts in Sputum and Plasma in COPD

In the sputum samples from COPD patients, mean concentrations of TNFα antagonists exceeded those of TNF by a ratio of approximately 10 to 1. In the same sputum samples, the mean concentration of IL-1β was ten times higher than IL-1sRII; however, the mean concentration of IL-1RA exceeded IL-1β concentrations by a ratio of 50 to 1. Sputum concentrations of IL-1β correlated significantly with sputum neutrophil and macrophage counts (PCC 0.83, p < 0.001 and PCC 0.3 p < 0.02, respectively) but no other cell type. Plasma IL-1RA correlated negatively with absolute sputum neutrophil and macrophage counts (PCC −0.54, p < 0.001 and PCC −0.35, p = 0.001, respectively).

The Relationships Between Sputum and Systemic Biomarkers and COPD Demographics

Sputum neutrophils correlated negatively with body mass index (BMI) (r = −0.7, p = 0.02; Fig. 2a). There were no correlations between TNFα or its antagonists and any patient features. Sputum IL-1ß correlated negatively with BMI (r = −0.7, p = 0.01; Fig. 2b). Plasma IL-1β correlated negatively with forced expiratory volume in 1 s (FEV1) expressed as a percent predicted (r = −0.56, p = 0.02; Fig. 2c). Plasma IL-1RA correlated negatively with smoking history expressed as pack year history (r = −0.7, p = 0.01) and positively with BMI (r = 0.78, p = 0.001; Fig. 2d).

Fig. 2
figure 2

The relationship between sputum and systemic biomarkers and COPD patients’ demographic data; all data points represent the log mean concentration of cytokine, antagonist or cell count for each COPD patient for five visits. Pack year history and body mass index (BMI) are based upon a single assessment. FEV1 percent predicted is the average of two measurements for each patient. a The relationship between BMI and sputum neutrophils. The regression lines are drawn and the correlation coefficient (r) and significance (p) is shown. b The relationship between sputum IL-1β and BMI. The regression lines are drawn and the correlation coefficient (r) and significance (p) is shown. c The relationship between plasma IL-1β and FEV1 percent predicted. The regression lines are drawn and the correlation coefficient (r) and significance (p) is shown. d The relationship between plasma IL-1RA and BMI (black line) and smoking history (gray line). Individual mean patient values are shown related to BMI (filled circles) and smoking history (unfilled circles). The regression lines are drawn and the correlation coefficient (r) and significance (p) are shown

Evidence of Direct Overspill from Lung to Plasma

Concentrations of TNFα were 50 times lower in plasma compared with sputum and concentrations of IL-1β were 6,000 times lower in plasma compared with sputum. There were no correlations between sputum and plasma levels of TNFα, TNFsRI, or TNFsRII or between sputum and plasma IL-1β, IL-1RA, or Il-1LsRII.

Discussion

The current study provides evidence that IL-1β (and in particular, a paucity of IL-1RA and IL-1sRII) may be related to pathophysiology in COPD. Although there were no significant differences in IL-1β concentrations between groups, plasma concentrations of IL-1RA and IL-1sRII were greatly reduced in COPD patients. The concentrations of IL-1β, IL-1RA, and IL-1sRII in our healthy subjects were comparable with previous studies of healthy volunteers [20, 21]. Furthermore, the levels of both antagonists seen in our healthy controls were comparable to those reported in inflammatory bowel disease [22], suggesting that the low levels of IL-1RA and IL-1sRII seen in our COPD patients is not a feature of chronic inflammation per se. Indeed, IL-1RA production appears to be raised in a number of other inflammatory lung conditions, including fibrosing alveolitis [23], cystic fibrosis [24], and sarcoidosis [25], and so the decreased expression seen in the current study may be specific to COPD. Furthermore, consistent with our in vivo work, in vitro IL-1RA production is pathologically downregulated in both monocytes and neutrophils from patients with COPD compared with controls [26, 7].

In the current study, in COPD, systemic concentrations of IL-1β exceeded IL-1RA and both IL-1β (negatively) and IL-1RA (positively) correlated with a clinical parameter of disease severity (BMI), consistent with an effect of this imbalance. In animal models, low levels of IL-1RA are associated with a low BMI, potentially by unregulated IL-1 mediated appetite suppression [27]. IL-1β increases protease activity [28, 29] while IL-1RA decreases it [28] and again, in animal models, IL-1β overexpression is associated with a neutrophilic infiltrate, distal airspace enlargement, increased thickness of the conducting airways, and enhanced mucin production [6]. IL-1β secretion increases monocyte recruitment [30], enhances secretion of pro-migratory chemokines, and increases neutrophil infiltration into the lung [31] which is consistent with the correlations seen here with plasma and sputum cells. The imbalance we have observed and correlations with disease features are, therefore, consistent with the in vitro data cited above.

Less is known of the role of IL-1sRII in health or disease. In vitro work has shown that the simultaneous presence of IL-1RA and IL-1sRII is highly efficient at abrogating IL-1β [32]. IL-1RA binds to IL-1R1 with a much higher affinity than IL-1β [33] and IL-1sRII binds to the precursor form of IL-1β with a low dissociation constant [34]. When considering the concentrations measured and the dissociation constants for IL-1β, IL-1RA [35], and ILsRII [34, 36], IL-1β is more likely to be active in plasma than sputum in the stable state. However, the concentration of IL-1RA measured in airway secretions could be artificially high, due to the high concentrations of IL-1RA in saliva [37] which may contaminate sputum (although we minimized any contamination using mouth washing techniques); therefore, local IL-1β activity in the pulmonary secretions cannot be confidently excluded.

There were no significant differences in plasma concentrations of TNFα or its antagonists in the two subject groups, and all measurements were similar to those previously reported in COPD suggesting that our patient group did not have unusually low levels of circulating or pulmonary TNFα [38]. In both groups, the plasma concentrations of the soluble TNF receptors were significantly higher than those of TNFα, and in COPD patients, sputum concentrations of both TNFsR1and TNFsR2 also exceeded TNFα. When the dissociation constants of the receptors and binding half lives are taken into consideration [3941], there remains an apparent excess of antagonist, both in the lung and systemically in the COPD patients, and systemically in healthy controls.

Previous studies have suggested that the maximal shedding capacity of TNF soluble receptors is reached at approximately 100 pM TNFα [42, 43]. In a study of limb perfusion, TNFα-related systemic effects were only noted when TNFα concentrations exceeded those of the soluble receptors [44], and the same is likely to be true in COPD, where greater expression (>100 pM) is needed before systemic effects are seen.

In our patients, the overall concentration of TNFα was not sufficient to overcome the naturally occurring inhibitors and the lack of relationship between TNFα and the clinical features of COPD would support this concept. Interestingly, a trial of anti-TNFα antibody in a similar group of stable COPD patients found no improvement in symptom scores or clinical outcomes compared with placebo following 24 weeks of therapy [45]. In conditions where anti-TNFα therapy has been shown to be effective, median levels of systemic TNFα have been reported to be 70 times higher [46] than those described in both the current study and the trial outlined above (70 vs. 0.9 pM [45, 46]) which may explain the lack of efficacy in stable COPD.

Previously, TNFα has been shown to correlate with BMI and cigarette smoke exposure [47, 48] and other inflammatory mediators [3] in COPD. It remains possible that TNFα is quiescent when COPD is stable and only becomes biologically active (with increased concentrations) during exacerbations [49]. However, there is evidence of a subset of patients with polymorphisms of the TNFα gene (that influence gene expression) who have an increased severity of COPD [50] or a mucus producing phenotype [51]. It would, therefore, be of interest to repeat the current studies in these subgroups to ascertain whether the relative concentrations of TNFα and antagonists are reversed.

There were no correlations between TNFα or IL-1ß in sputum and plasma, which would suggest that there is no simple overspill of either cytokine from lung to blood driving the systemic inflammation seen in these patients.

The relationships described above were investigated using plasma and sputum samples from a homogenous group of patients with moderate to severe COPD with a chronic bronchitis phenotype and plasma from age-matched healthy controls. Patients with chronic bronchitis were chosen because airways inflammatory burden is higher in this group of patients compared with patients who do not expectorate, irrespective of smoking status [52]. Sputum production may be related to an accelerated decline in lung function in COPD [53, 54], and mucus hyper-secretion is associated with death from respiratory infections [55]. Given the inflammatory and prognostic significance of chronic bronchitis in COPD, it is important to understand the relationships between inflammatory mediators in this subgroup of patients. Furthermore, spontaneous sputum collection is noninvasive and allows repeated measurements without inflammatory sequelae, and there are published data regarding reproducibility and intra-patient variability for a number of mediators [3].

We included never smokers as our age- and gender-matched healthy controls as we wished to compare plasma data from truly healthy normal individuals with patients with COPD, 60% of whom had stopped smoking. Although it remains possible that smoking in itself may influence the inter-relationships seen here, we found no difference in cytokine, antagonist concentrations, or relationships with clinical parameters between current and ex smokers in COPD. This suggests that the plasma results presented here represent predominantly a feature of COPD rather than smoking alone. In keeping with this, it has been shown that there is no difference in cells mRNA positive for TNFα in smoking and ex-smoking patients with COPD [56], which is consistent with our findings.

The data presented are the average concentrations of mediators and cells in samples taken on five consecutive days as this is the optimal number to assess relationships between mediators, cells and markers of disease burden, by reducing intra-patient sample variability [3].

There are some minor limitations to the current study. Firstly, the assays used measure both free and bound cytokines, confirmed by spiking experiments performed during the validation of all ELISA assays (data not shown). The assays do not measure biological activity, and small amounts of free cytokine could still exert a direct inflammatory effect. However, the in vivo data presented in the current study are in keeping with in vitro data that have been published previously [7, 26], suggesting that the imbalances described are real and that the ability to control IL-1β (but not TNFα) induced inflammation may be compromised in COPD.

Secondly, the low numbers included in the study may prevent determination of key factors that may alter cytokine and antagonist expression, including smoking status or medication exposure, although repeated sampling is beneficial and enables variability of data to be markedly reduced.

Conclusions

IL-1RA and IL-1sRII concentrations were reduced systemically in COPD compared with controls, and there is evidence that IL-1β is active in COPD, with a relative physiological excess of the cytokine compared with its antagonists. Furthermore, IL-1β and IL-1RA were associated with clinical disease parameters including FEV1, BMI, and smoking history suggesting a key role of smoking and this cytokine/antagonist ratio in the pathophysiology of COPD. TNFα, TNFsR1, and TNFsR2 concentrations were not raised; there was an abundance of antagonist compared with cytokine and TNFα did not correlate with markers of disease suggesting that TNFα is unlikely to be highly active in stable COPD. The relationship between lung and systemic inflammation may explain the high level of comorbidity seen with COPD; however, there is no evidence to support direct overspill of either cytokine from lung to plasma.