European Respiratory Society

Disconnect between sputum neutrophils and other measures of airway inflammation in asthma

Joseph R. Arron, David F. Choy, Michel Laviolette, Steven G. Kelsen, Ammar Hatab, Richard Leigh, Neil C. Thomson, Eugene R. Bleecker, Ron Olivenstein, Pedro C. Avila, Nizar N. Jarjour, Mario Castro, Gail M. Gauvreau, James T. Good, Joel N. Kline, Adel Mansur, Irvin Mayers, Liam G. Heaney, Qutayba Hamid, Jeffrey M. Harris

To the Editor:

Asthma heterogeneity has been described by the nature and intensity of granulocytic infiltration into the airways [1, 2]. Sputum, endobronchial biopsies and bronchoalveolar lavage (BAL) sample different anatomical regions of the airways. Few studies have directly compared the inflammatory cell infiltrates in these regions within a large cohort of moderate–severe asthma patients using standard techniques.

In the BOBCAT (Bronchoscopic exploratory research study Of Biomarkers in Corticosteroid-refractory AsThma) study, we sampled multiple airway compartments concurrently, enabling us to evaluate relationships between granulocytic infiltrates within and between compartments in a large cohort of moderate–severe adult asthma patients. This prospective multicentre observational study was conducted in four visits over a 4–6 week period, as described previously [3].

The patients included had moderate–severe persistent asthma (forced expiratory volume in 1 s (FEV1) 40–80% predicted and Asthma Control Questionnaire (ACQ) [4] score >1.5) and, within the past 5 years, evidence of >12% post-bronchodilator reversibility or a provocative concentration of methacholine causing a 20% decline in FEV1 ≤8 mg·mL−1 despite high-dose inhaled corticosteroid (ICS) (≥1000 μg·day−1 fluticasone propionate equivalent) with or without long-acting β2-adrenergic agonist therapy. Key exclusion criteria included initiation or increase in systemic steroid use 30 days prior to screening, chronic or recent (within the past 30 days) use of immunosuppressive therapies, or other active lung disease. Patients had a prior established diagnosis of moderate–severe asthma for ≥6 months prior to screening while receiving a stable dose regimen (>6 weeks) of a high-dose ICS. Allowed concomitant medications included leukotriene receptor antagonists and oral corticosteroids.

78 patients with confirmed moderate–severe asthma were enrolled at 18 sites; 67 (86%) completed the study. All patients had persistently impaired lung function and uncontrolled symptoms despite high-dose ICS treatment of ≥1000 μg·day−1 fluticasone propionate equivalent. 11 (14%) patients did not complete the study, due to an adverse event (n=1), physician’s decision (n=5), subject’s decision (n=3) or the sponsor’s decision (n=2). One patient experienced increased bronchospasm following bronchoscopy, requiring additional observation and oral corticosteroid treatment prior to being discharged home from the final study visit.

We enumerated sputum, tissue and BAL eosinophils and neutrophils using standard techniques as described previously [5, 6]. Both granulocyte types spanned broad ranges and had a unimodal distribution in each compartment (fig. 1). The majority of nonsquamous cells in sputum from most subjects were neutrophils, while a much smaller percentage of sputum cells were eosinophils. The majority of BAL cells were macrophages (data not shown), with eosinophils and neutrophils comprising <10% of total BAL cells in most cases.

Figure 1–

Relationships between eosinophil and neutrophil levels in each airway compartment. a) Eosinophil and neutrophil counts in endobronchial biopsy tissue; median (interquartile range) neutrophils 31 (17–56) cells·mm−2 and eosinophils 23 (7–44) cells·mm−2. Percentages of nonsquamous cells in b) induced sputum; median (IQR) neutrophils 52 (37–71)% and eosinophils 5 (2–15)%; and c) bronchoalveolar lavage (BAL) fluid; median (IQR) neutrophils 2 (1–8)% and eosinophils 1 (0–3)%. rS: Spearman’s rank correlation.

Within each compartment, eosinophils and neutrophils were significantly intercorrelated. However, these correlations were positive in biopsy tissue and BAL but negative in sputum (fig. 1). The correlation within any single compartment was greatest in biopsy tissue (Spearman’s rank correlation (rS)=0.68, p<0.0001). Across compartments, sputum eosinophils were significantly, albeit weakly, positively correlated with both biopsy tissue eosinophils (rS=0.36, p<0.05) and BAL eosinophils (rS=0.33, p<0.05), while neutrophils were not intercorrelated across any compartments. Sputum neutrophils were negatively correlated with tissue eosinophils (rS= -0.37, p<0.01) and BAL eosinophils (rS= -0.34, p<0.05). These findings show generally positive relationships between eosinophils and neutrophils within and across airway compartments, with the exception of sputum neutrophil percentage.

Serum periostin and exhaled nitric oxide fraction (FeNO) are positively correlated with sputum and biopsy eosinophils in BOBCAT [3]. Both serum periostin and FeNO trend towards a positive correlation with biopsy neutrophils (rS=0.25, p=0.06 for periostin; rS=0.23, p=0.08 for FeNO), and exhibit significantly negative correlations with sputum neutrophils (rS= -0.31 for periostin, rS= -0.35 for FeNO; p<0.05 for each). None of the airway measures, nor blood biomarkers, exhibited significant correlations with lung function as assessed by FEV1 or asthma control as assessed by the ACQ (not shown).

Significantly elevated airway neutrophils are not typically observed in mild–moderate asthma patients not taking ICS or on low-dose ICS, but are often seen in severe asthma patients on high-dose steroids [7, 8], suggesting that chronic ICS treatment is related to elevated airway neutrophils. Our observations add to these findings by showing that, in patients with moderate–severe asthma, mucosal neutrophils are particularly elevated in subjects with tissue eosinophilia despite ICS treatment, and this relationship scales continuously.

Although eosinophil and neutrophil percentages in sputum are measured on a continuous scale, a desire to classify asthma into discrete subsets has yielded “cut-offs” for eosinophilia around 2–3%, whereas cut-offs for neutrophilia range from 40% to >60% [1, 2, 7, 9, 10]. This is because eosinophils are typically absent in nonasthmatic subjects, while a substantial proportion of sputum cells are neutrophils even in healthy subjects [9]. As a considerable fraction of sputum is composed of neutrophils, an increase in the proportion of another cell subset may come at the expense of neutrophil percentage, setting up a propensity for inherently negative correlations between proportions of the two cell types. While changes in the relative proportion of a minor component of the cellular content of a sample (e.g. sputum eosinophils) may be informative, caution must be exercised in interpreting the relevance of changes in the relative proportion of a component representing the majority or significant plurality of the cellular content of a sample (e.g. sputum neutrophils). Consistent with these considerations, sputum eosinophils exhibited substantial proportional variability, but remained a minor component of sputum cells and correlated with tissue and BAL eosinophils, serum periostin, blood eosinophils and FeNO [3]. Sputum neutrophils exhibited less proportional variability in this study and largely appeared to vary as a negative function of sputum eosinophil percentage (fig. 1).

The weakness of the positive correlations observed between eosinophils across sputum, biopsies and BAL may be due to several factors, including: 1) the three compartments sample different regions of the airways; 2) factors mediating transmigration of granulocytes through bronchial mucosal tissue into the airway lumen are poorly understood; 3) the techniques for enumerating and reporting granulocytes in each compartment vary; and 4) each assessment in this study is cross-sectional, with sputum sampling taking place close to, but not contemporaneous with bronchoscopy. Biomarkers that integrate total inflammatory burden systemically may present a means of circumventing the variability across airway compartments. As previously reported for the BOBCAT cohort, patients with low eosinophil levels in both sputum and bronchial tissue had the lowest serum periostin levels, while subjects with elevated eosinophils in either one or the other compartment had intermediate periostin levels, and subjects with elevated eosinophils in both sputum and bronchial tissue had the highest periostin levels [3]. Therefore, sampling multiple airway compartments and/or systemic biomarkers may be a more sensitive means to ascertain whether a given asthma patient has eosinophilic airway inflammation that may be missed by sampling only one compartment.

BOBCAT represents a large, well-characterised cohort of moderate–severe asthma patients resistant to high-dose ICS with intensive airway sampling. The key findings from this study are as follows. 1) Sputum, biopsy and BAL eosinophils are imperfectly but positively correlated, suggesting that eosinophilia in each of those compartments may be informative in moderate–severe asthma. 2) While neutrophils are positively correlated with eosinophils in biopsies and BAL, they are negatively correlated with eosinophils in sputum. Whether this is a technical artefact or a biologically relevant finding is unclear, but given recent attention devoted to “neutrophilic” phenotypes in severe asthma, this should be investigated further, particularly in the context of interventional studies of ICS and new investigational asthma therapies. 3) Eosinophil and neutrophil levels relate to each other on a continuous scale, which suggests that defining discrete cut-offs for airway phenotypes should be undertaken with caution. Future studies should also examine the longitudinal variability of granulocytic infiltration, as well as the ability of these phenotypic variables to predict outcomes in interventional studies.


We wish to thank the patients and clinical personnel who participated in this study.


  • Support statement: This study was funded by Genentech, Inc.

  • Conflict of interest: Disclosures can be found alongside the online version of this article at

  • Received July 9, 2013.
  • Accepted October 7, 2013.