The relationship between airway immunoglobulin activity and eosinophils in COPD

Abstract In chronic obstructive pulmonary disease (COPD), the effects of inhaled corticosteroids are predicted by blood eosinophil counts. We previously briefly reported increased immunoglobulin (Ig)A and IgM levels in bronchoalveolar lavage (BAL) of COPD patients with higher (eosinophilhigh) compared to lower (eosinophillow) blood eosinophils (>250/μL versus < 150/μL), suggesting differences in adaptive immune function. An inverse relationship exists between eosinophil counts and airway pathogenic bacteria levels. The mechanistic reasons for these associations between eosinophils, corticosteroids and pathogenic bacteria are unclear. IgA, IgM and IgG levels were assessed in BAL, bronchial biopsies and epithelium collected from eosinophilhigh (n = 20) and eosinophillow (n = 21) patients. Bronchial B‐cell numbers were measured by immunohistochemistry. B‐cell activity was assessed in bronchial samples and following exposure to BAL from eosinophilhigh and eosinophillow patients. BAL levels of non‐typeable Haemophilus influenza (NTHi)‐specific immunoglobulins were quantified. Results showed airway expression of IgA, IgG1 and IgM were lower in eosinophillow compared to eosinophilhigh patients, with lower levels of NTHi‐specific IgA and IgM. Bronchial B‐cell numbers were similar in both groups, but B‐cell activity was lower in eosinophillow patients. In conclusion, COPD eosinophillow patients show differences in adaptive immune function compared to COPD eosinophilhigh patients. These differences may cause different microbiomes in these COPD phenotypes.

help guide the use of ICS containing combination treatments in COPD patients. 2 The mechanistic reasons for the increased ICS response in COPD patients with higher blood eosinophil counts are unclear. We previously performed a bronchoscopy study in 21 blood eosinophil low (<150 eosinophils/µl; eosinophil low ) and 20 blood eosinophil high (>250 eosinophils/µl; eosinophil high ) COPD patients to investigate biological differences associated with eosinophil counts. 3 We reported higher lung eosinophil numbers, thicker reticular basement membrane and differences in the levels of various inflammatory mediators including IL-5, IL-13, CCL24 and CCL26 in eosinophil high compared to eosinophil low patients. 3,4 Such findings are also observed in patients with asthma, 5 providing insights into potential reasons for differential ICS responses associated with blood eosinophil counts in COPD.
Our previous report noted higher IgA and IgM levels in the bronchoalveolar lavage (BAL) of COPD eosinophil high compared to eosinophil low patients, 3 suggesting differences in adaptive immunity between these groups. An inverse association has been reported between sputum eosinophil counts and the levels of pathogenic bacteria in the airways. [6][7][8][9] Additionally, COPD patients with blood eosinophil counts <100 cells/µl are more likely to have chronic bacterial airway infection. 10 Overall, these findings support the concept that COPD patients with lower eosinophil levels are more susceptible to bacterial airway infection. Furthermore, COPD or asthma patients with raised blood and sputum eosinophils demonstrate greater presence of bacterial phylum Bacteroidetes, 11 whereas low sputum eosinophil levels were associated with lower bacterial diversity and increased Proteobacteria, specifically the Haemophilus genus. 9,12 These altered microbiome profiles associated with different eosinophil counts may cause distinct airway inflammation profiles in the airway which respond differently to anti-inflammatory drugs. 13 Here, we report a further analysis using samples from our bronchoscopy study that compared COPD eosinophil high with eosinophil low patients; 3 we focus on differences in adaptive immune function that may cause altered immunity against bacteria.
We investigated IgA, IgM and IgG levels using different samples and techniques, and measured B-cell activation and bacterial opsonization. 21 eosinophil low (blood count < 150/µl) and 20 eosinophil high (>250/ µl) COPD patients were recruited for bronchoscopy. Cut-offs were chosen as they identified upper and lower tertiles for blood eosinophils counts in COPD patients previously studied at the Medicines Evaluation Unit, Manchester, UK. As fluctuations in blood eosinophil numbers in COPD patients are typically minor, particularly at lower blood eosinophil counts, 14-16 the exclusion of the middle tertile helped ensure the eosinophil high and eosinophil low patients were distinct populations. Initial results and clinical characteristics were reported previously 3 (Table 1)

| Gene expression analysis
Bronchial epithelial brushings were collected from the lower lobes (n = 20 eosinophil high ; n = 17 eosinophil low ), and blood was collected  Table 2, including a functional description. Detailed methods are in the online data supplement.

| BAL Immunoglobulin assessment
BAL was successfully collected from n = 15 eosinophil high and n = 15 eosinophil low . Levels of IgG1 and IgG2 in BAL fluid were determined by ELISA (ThermoFisher). Secretory IgA levels were assessed by ELISA (Demeditec). IgA and IgM levels have been previously assessed in the same BAL samples by multiplex assay (Myriad RBM). 3 Levels of immunoglobulins were normalized to patient-specific BAL urea concentrations, 17 measured by colorimetric assay (Biovision Inc). The use of urea for normalization is not universally accepted, 18 so for completeness, we also present non-normalized results in the online supplement.

| Bronchial biopsy Immunohistochemistry
Formalin-fixed, paraffin-embedded bronchial biopsies were assessed for levels of B cells and plasma cells, along with expression of IgA2, IgG1, IgG2 and IgM by immunohistochemistry. Detailed methods are in the online data supplement.

| In vitro B-cell treatment
B cells from the blood of a healthy non-smoking volunteer were exposed to BAL fluid from eosinophil high and eosinophil low patients.
Healthy donor B cells were preferred over COPD B cells, which may be dysfunctional giving data that is difficult to interpret. The effect of BAL on X-box-binding protein 1 (XBP1) gene expression was assessed by quantitative PCR. Detailed methods are in the online data supplement.

| NTHi-specific immunoglobulin binding assay
Methodology for immunoglobulin-specific opsonization of nontypeable haemophilus influenza (NTHi) was adapted from Staples et al 19 Detailed methods are in the online data supplement.

| Statistical analysis
All results were assessed using Graphpad Prism version 7.04.
Comparisons between eosinophil high and eosinophil low were by t test or Mann-Whitney test, with distribution of data assessed by D'Agostino & Pearson normality test.

| Subjects
Clinical characteristics for the 21 eosinophil low and 20 eosinophil high COPD patients are summarized in Table 1. Lung function, proportion of patients using ICS, symptoms and exacerbation rates were similar TA B L E 1 Demographics and clinical characteristics of blood eosinophil low and blood eosinophil high patients between the groups, with the majority of patients being GOLD stage 2. No history of immunodeficiency was reported for any patient, and blood immunoglobulin gene expression levels were similar in eosinophil high and eosinophil low patients, as were serum IgA and IgM levels (Table S1). Blood basophil numbers were higher in eosinophil high compared with eosinophil low patients.

| IgA and IgM BAL protein and epithelial gene expression
Previously, we briefly reported that COPD eosinophil high patients had higher absolute BAL IgA and IgM protein levels, compared with eosinophil low patients ( Figure S1A,B). 3 BAL recovery rates vary between patients, causing variable dilution effects. We now present BAL immunoglobulin results normalized to patient-specific BAL urea levels ( Figure 1A); BAL IgA and IgM protein levels remained higher in eosinophil high compared to eosinophil low patients. Immunoglobulin gene expression in epithelial brushings from the same patients ( Figure 1B) demonstrated higher expression of genes for the IgA and IgM heavy chains, and the J chain that enables IgA and IgM polymerization, in the eosinophil high COPD patients. Secretory IgA levels in BAL from eosinophil high patients were higher compared with eosinophil low patients ( Figure 1A; Figure S1C).

| IgG BAL protein and epithelial gene expression
Bronchial epithelial gene expression levels of IGHG1 and IGHG2 were greater in COPD eosinophil high patients ( Figure 1B). The mean levels of IGHG3 and IGHG4 gene expression were also higher in the eosinophil high group, but did not reach statistical significance ( Figure 1B).  Figure 1A with urea normalized results and Figure S1D,E with non-normalized results).

| Immunoglobulin immunohistochemistry
IgA was predominantly found on the apical surface of the bronchial epithelium ( Figure 2A). The percentage of the epithelium immunoreactive for IgA was significantly higher in the eosinophil high patients compared with the eosinophil low patients ( Figure 2E). showed similar trends ( Figure S2; P = .08 and P = .05, respectively).

| B-cell numbers and activity
Differences in immunoglobulin levels may be due to different B cells or plasma cell numbers. We assessed levels of CD19 + B cells and CD27 + plasma cells in bronchial biopsies by immunohistochemistry ( Figure 3A,B). B cells and plasma cell numbers were similar between the eosinophil high and eosinophil low groups ( Figure 3C,D). Similar gene expression of CD19 and CD27 in bronchial epithelial brush samples supported these findings ( Figure 3E,F).
We treated healthy B cells with BAL fluid from eosinophil high and eosinophil low patients to assess B-cell activation. Cells treated with eosinophil high BAL expressed higher levels of XBP1, a key regulator of plasma cell immunoglobulin production (Figure 4). We compared bronchial epithelial gene expression of mediators that influence plasma cell activity ( Table 2). Levels of NOS2 and the lymphotoxins LTA and LTB were significantly lower in eosinophil low patients.

| NTHi opsonization
Reduced immunoglobulin levels in eosinophil low COPD patients may confer reduced anti-bacterial activity. Immunoglobulin-specific NTHi opsonization experiments were carried out using BAL fluid.
The results showed that approximately twice as many bacteria were Pneumonia risk is greater in COPD patients with lower blood eosinophil counts. 10,21 Lower airway eosinophil numbers are associated with increased presence of colonizing bacteria in stable COPD patients, and ICS use was associated with an increase in bacterial load only in patients with lower eosinophils. [6][7][8][9] Overall, these previous findings suggest that COPD patients with lower eosinophil counts have a different susceptibility to bacterial infection. 22 Our data showing lower immunoglobulin levels and activity in COPD eosinophil low patients provide a potential mechanistic explanation for these observations. The risk of pneumonia in COPD patients is associated with certain clinical characteristics, such as previous pneumonia events, older age and lower body mass index. 2 The use of ICS in such individuals may increase pneumonia risk further. 2 Overall, there appear to be complex interactions between clinical characteristics, blood eosinophil counts and ICS use that influence pneumonia risk in COPD patients. 22 COPD and asthma patients with lower airway eosinophils have greater levels of proteobacteria, including Haemophilus. 9,11,12 NTHi is found in the airways of COPD patients, both at stable state and during exacerbation. 12 Although IGHG2 expression was lower in eosinophil low patients, higher protein levels were retained in the subepithelium. This may be due to IgG2 being actively retained within the tissue to combat subepithelial bacterial colonization associated with lower apical IgA expression.
Lymphoid follicles act as focal centres for antigen presentation and adaptive immune responses; increased numbers of B-cell rich lymphoid follicles are associated with COPD severity. 31 IgM + plasma cells are the predominant B-cell found in airway lymphoid follicles. BAL IgM levels were reported to be higher in COPD patients compared to controls. 19,32 We have shown lower levels of IgM BAL protein and epithelial gene expression in eosinophil low compared with eosinophil high COPD patients. It is probable that these epithelial gene expression results are due to bronchial brush samples contain a mixture of epithelial and other cell types including B cells.
We were unable to determine if the distribution of tissue IgM differed between the eosinophil high and eosinophil low groups.
Immunohistochemical analysis showed that IgM was predominantly found in mucosal glands. Previous work has shown that immunoglobulins produced in glands do not distribute across the airway mucosal surface, but are localized to mucous plugs. 27 Like IgA, we showed that eosinophil low COPD patients had reduced levels of NTHi-specific IgM in BAL compared with eosinophil high patients.
B-cell and plasma cell numbers are increased in both the large and small airways of COPD patients compared to controls. 33,34 A recent study reported that lymphoid follicles in severe COPD patients have increased IgA + plasma cells. 32 We found no differences F I G U R E 4 Activation of B-cell XBP1 gene expression following treatment with BAL from eosinophil high and eosinophil low patients. B cells were isolated from the blood of a healthy subject, and then treated with BAL fluid for 72 h. Gene expression of XBP1 was assessed by quantitative PCR, with data presented as fold change compared to untreated B cells. Comparison between eosinophil high and eosinophil low was by t test: *P < .05. Bars represent mean values ± standard deviation F I G U R E 5 Immunoglobulin-specific opsonization of NTHi with BAL from eosinophil high and eosinophil low patients.
NTHi bacteria was treated with BAL for 30 min and then labelled with fluorescent antibodies against IgA, IgM, IgG1 and IgG2 to allow levels of opsonization to be assessed by flow cytometry. Data are presented as the percentage of bacterial cells showing positive opsonization. Comparisons between eosinophil high and eosinophil low were by Mann-Whitney test: **P < .01. Bars represent median values with interquartile range in B-cell or plasma cell numbers in biopsies from eosinophil high and eosinophil low patients. However, BAL fluid from eosinophil low patients induced lower levels of XBP1 expression than BAL from eosinophil high patients. The transcription factor Xbp1 is required for plasma cell differentiation and is essential for high levels of antibody secretion, 35 and our findings suggest that the airways of eosinophil low patients may be less primed to stimulate immunoglobulin production. In smoke-exposed mice, efficient clearance of NTHi has been shown to be B-cell dependent and relates to levels of NTHispecific IgA. 36 The epithelial gene expression of NOS2 and lymphotoxins LTA and LTB was lower in the eosinophil low group; these findings are discussed further in the online supplement.
We previously reported higher BAL IL-5 levels in eosinophil high patients. 3 This cytokine is important for eosinophil maturation, ac- in Streptococcus pneumonia levels, with no effect on Haemophilus influenzae. 38 The authors proposed that Streptococcus pneumonia reduction may be due to macrophages being redirected from eosinophil efferocytosis towards bacterial phagocytosis after eosinophil depletion. Additionally, eosinophils have no direct anti-bacterial activity against Streptococcus pneumonia and Haemophilus influenza. 38 Overall, the eosinophil itself appears to have little direct anti-bacterial activity, but here we show that eosinophil low COPD patients display differences in B-cell activity and immunoglobulin levels that may cause differences in microbiome.
We have recently shown higher airway IL-13 levels in the same group of eosinophil high (versus eosinophil low ) patients in this report. 4 IL-13 promotes humoral responses. 42 It is possible that the T2 environment present in the airways of eosinophil high COPD patients facilitates better anti-bacterial defence through mechanisms such as IL-13-stimulated B-cell activation.
Hypereosinophilic syndrome commonly involves the lungs (63% of patients). 43 The airways from chronic eosinophilic pneumonia (CEP) patients contain many more eosinophils compared to eosinophil high COPD patients; for example, median BAL eosinophil percentage in CEP is 52% and 0.75% in eosinophil high COPD. 3,44 However, like eosinophil high COPD, CEP patients also have elevated airway type 2 inflammation 44 and increased BAL levels of IgA, IgM and IgG. 45 We did not include healthy controls, as it has already been reported that immunoglobulin expression levels are higher in mild COPD patients compared with controls, 19 but more severe patients have lower secretory IgA levels. 27 We cannot state if the immunoglobulin levels observed here are elevated or decreased compared to normal levels, but can state that immunoglobulin levels are higher in eosinophil high relative to the eosinophil low patients. There is growing evidence that low eosinophil numbers are associated with bacterial colonization. [6][7][8]10 A limitation is that we did not assess bacteriology in the current study, preventing comparisons between immunoglobulin levels and bacterial load. We had few severe COPD patients, as these individuals are practically more difficult to bronchoscope for research. The sample size of 41 subjects was modest, but sufficient to observe consistent biological differences between groups.
In conclusion, we have shown differences in adaptive immunity in COPD eosinophil low patients compared with COPD eosinophilhigh patients, associated with altered anti-proteobacterial immunity.

This research was supported by the NIHR Manchester Biomedical
Research Centre and the North West Lung Centre Charity, Manchester. This report is independent research and the views expressed in this publication are those of the authors and not necessarily those of the NHS, the NIHR or the Department of Health.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.