Background
COPD is a heterogeneous condition, composed of different clinical and pathophysiological components that vary both in presence and severity between patients [
1]. This heterogeneity causes variability in the responses to pharmacological treatments. Biomarkers that predict treatment responses to anti-inflammatory drugs may be useful for optimising the benefit versus risk ratio.
Sputum eosinophil counts in stable COPD patients predict the clinical response to corticosteroids [
2,
3]. However, measuring sputum eosinophils is time consuming, and some patients do not provide adequate samples for analysis. Blood eosinophil measurements are more practical, and appear to be a surrogate biomarker for sputum eosinophils as these measurements show a degree of correlation within the same individual, both in stable COPD patients and during exacerbations [
4‐
7]. Retrospective analysis of COPD clinical trials have shown that higher blood eosinophil counts predict a greater reduction in exacerbation rates with inhaled corticosteroid/long acting beta agonist (ICS/LABA) combinations compared to LABA [
8‐
10]. Furthermore, oral corticosteroid treatment during exacerbations has a greater effect in COPD patients with higher blood eosinophil counts [
11].
Chronic bacterial infection in COPD patients causes greater neutrophilic inflammation in the lungs [
12‐
14]. Neutrophilic lung inflammation responds poorly to corticosteroids [
15‐
17], implicating bacterial infection as a potential cause of corticosteroid insensitivity in COPD patients. The relationship between bacterial infection and eosinophil counts is not established in COPD. There may be an inverse relationship between these parameters, as blood eosinophil counts are known to be reduced during severe bacterial infection [
18,
19]. The existence of such an inverse relationship would suggest that the interaction between bacterial infection and eosinophils determines the corticosteroid response in COPD patients.
The primary aim of the analysis reported here using data from the COPDMAP cohort was to test the hypothesis that the bacterial load and eosinophil counts are inversely related in COPD patients. We investigated the relationship between eosinophil counts (in the blood and sputum) and the bacterial load in the stable state and during exacerbations in COPD patients.
Discussion
We observed evidence of an inverse relationship between bacterial infection and eosinophil counts in COPD. Firstly, sputum eosinophil counts were lower in COPD patients with bacterial infection in the stable state. Secondly, COPD patients with bacterial infection during exacerbations had a significant decrease in the blood eosinophil absolute count compared to the stable state, while no blood eosinophil count changes were observed in patients without bacterial infection. These observations during the stable state and exacerbations showing an inverse relationship between bacterial infection and eosinophil counts may be relevant to corticosteroid responsiveness in COPD; the increased corticosteroid responsiveness observed with higher eosinophil counts could be due, at least partly, to lower levels of bacterial infection..
This inverse relationship between bacterial counts and blood eosinophils was present during exacerbations but not in the stable state. This is likely to be due to the stronger association between blood and sputum eosinophils at exacerbation compared to the stable state. Blood eosinophils have been proposed as a surrogate biomarker of sputum eosinophils, but the weak relationship between these measurements in the stable state, demonstrated here and in previous studies [
6,
7], highlights a limitation of blood eosinophil counts in this context.
The prevalence of bacterial colonisation at baseline in our study (52%) was comparable to previously published studies that used qPCR to detect PPM presence; Bafadhel et al., [
12] reported a 51% prevalence while, Barker et al., [
26] reported a 77% prevalence. In the stable state, higher bacterial counts were associated with higher sputum neutrophil % and lower eosinophil %. This relationship between higher sputum bacterial counts and increased sputum neutrophil % is known [
12,
13,
29]. The lower sputum eosinophil % in COPD patients with bacterial colonisation is partly due to the bacterial driven increase in absolute and percentage neutrophil counts causing a lower calculated eosinophil %. However, the absolute sputum eosinophil count is not dependent on any calculation involving the neutrophil count, and we observed that the absolute sputum eosinophil count was also lower in patients with sputum eosinophils <3%, indicating a genuine reduction in eosinophil numbers in the airways of patients with higher bacterial counts.
The inverse relationships between bacterial counts and sputum eosinophils provide a potential mechanism that determines corticosteroid responsiveness in COPD patients. Individuals with lower sputum eosinophil counts have higher levels of bacterial infection; perhaps the presence of more bacteria, with associated neutrophilic inflammation [
12‐
14], contributes to the lower corticosteroid response previously reported [
16,
17]. Neutrophilic airway inflammation appears to be poorly responsive to corticosteroid treatment [
15,
17,
30], and the absence of bacterial inflammation may therefore favour a greater corticosteroid response through a shift in the balance of airway inflammation away from neutrophilic inflammation towards more eosinophilic inflammation.
Previous studies have failed to show a significant difference in sputum eosinophil % between stable COPD patients with bacteria present versus those without, although a numerical decrease in the former group was observed [
12,
26]. Compatible with our results is the previous finding of a significantly lower level of sputum CCL13 (an eosinophil associated cytokine) in PPM positive COPD patients [
26].
It has been reported that the effects of oral corticosteroids are dependent on the blood eosinophil count during COPD exacerbations, with more likelihood of treatment failure or a longer hospital stay with lower blood eosinophil counts [
11,
31,
32]. The data reported here show lower eosinophil counts during acute bacterial infections, and it could be inferred that treatment failure with oral corticosteroids in such cases is related to the presence of airway bacteria. During exacerbations both sputum and blood eosinophil measurements showed similar relationships to bacterial counts. However, the sputum data were less robust due to smaller sample size.
Blood neutrophil counts are known to increase during infection, irrespective of bacterial presence [
33]; we observed the same during COPD exacerbations. The decrease in blood eosinophil absolute count during exacerbation cannot be due to the increase in blood neutrophil counts for two reasons. First, there was a similar increase in blood neutrophil counts in both PPM positive and negative patients; second, we measured blood eosinophil absolute numbers which are not influenced by calculations of percentage relative to neutrophils.
In COPD patients hospitalised for exacerbations, blood eosinopenia is an independent predictor of mortality and length of stay [
34,
35]. The exacerbations in this study were treated in the outpatient setting (only 3/109 exacerbations resulted in hospitalisation). These less severe events had few cases of eosinopenia (5/109 patients). Nevertheless, our results demonstrate an effect of airway bacteria on blood eosinophil counts even in patients without overt bacterial sepsis. Similarly, blood eosinophil counts decrease in asthma patients with bacterial infections [
36].
Rhinovirus infection had no effect on blood eosinophil counts. Experimental rhinovirus infection causes an increase in the airway bacterial load, demonstrating the complexity of the relationship between virus and bacterial infection during COPD exacerbations [
37]. Our results indicate that bacterial, rather than viral, infection modulates blood eosinophil counts. However, it is worth noting that due to insufficient sputum samples being available, rhinovirus detection was performed in a smaller proportion of patients (75/109 patients). Consequently, the lack of effect of rhinovirus on eosinophil counts could potentially be attributed to limited power in the analysis.
The mechanism responsible for the decrease in eosinophil counts during bacterial infection is unclear. Eosinophils can trigger innate immune responses to pathogens through the release of extracellular DNA traps and the expression of specific pattern recognition receptors including Toll like receptor 4 [
38,
39]. Furthermore, eosinophil granule proteins have bactericidal activity [
40]. A decrease in circulating eosinophils may be a result of adrenal glucocorticoid stimulation in response to the stress of bacterial infection or the rapid accumulation of eosinophils at the inflammatory site [
41,
42].
A limitation of this study is that we defined a bacteria positive sample based on the total load of bacteria measured rather than a species specific count. This was done to increase statistical power to address the relationship between eosinophils and the common pathogenic bacteria in COPD. However, it is important to note that the total bacterial load from each bacteria positive sample had at least one of the three quantified pathogens above the 1 × 10
4 threshold used. The presence of other important bacteria, such as
Pseudomonas aeruginosa, was not assessed using qPCR. Additionally, we cannot discount the possibility of selection bias as a proportion of our patients failed to provide an adequate, high quality sputum sample for differential cell count analysis. The majority of the sputum samples used in our study were spontaneously produced and although cell viability can be lower in spontaneous samples [
43], we observed no differences in % sputum counts between spontaneous and induced samples, including the sputum eosinophil % count as previously reported [
43]. Finally, the majority of our patients were using ICS, but we found no difference in sputum or blood cell counts or bacterial loads due to ICS use. This is in keeping with studies that have shown no effect of ICS on blood or sputum eosinophil counts [
2,
44].
Future studies of the effects of ICS during the stable state could measure sputum (and blood) eosinophil counts and bacterial loads to test the hypothesis that increased bacterial load associated with reduced sputum eosinophil counts predicts reduced therapeutic response. Similar studies using oral corticosteroids during acute exacerbations measuring blood eosinophil counts and bacterial loads could test the hypothesis that increased bacterial load associated with reduced blood eosinophil counts predicts reduced therapeutic response.
Acknowledgements
The authors thank staff and participants of the COPD MAP consortium study.
Competing interests
UK, VG, RS, BB,,LG, AW, ST, TMcH, AB have no competing interests; GD received personal fees from MiCom SRL; CEB has received grants and or consultancy paid via his Institution from GSK, AZ/MedImmune, Novartis, Chiesi, BI, Pfizer, Theravance, Vectura; JW reports personal fees and non-financial support from Novartis, Pfizer, Astra Zeneca, Boehringer Ingelheim, grants, personal fees and non-financial support from GlaxoSmithKline, Takeda, grants and personal fees from Johnson and Johnson and Vifor Pharma, personal fees from Bayer, Chiesi, Napp; DS has received sponsorship to attend international meetings, honoraria for lecturing or attending advisory boards and research grants from various pharmaceutical companies including Almirall, AstraZeneca, Boehringer Ingelheim, Chiesi, Genentech, GlaxoSmithKline, Glenmark, Merck, NAPP, Novartis, Pfizer, Respivert, Skypharma, Takeda, Teva, Therevance and Verona.