Analyses of asthma severity phenotypes and inflammatory proteins in subjects stratified by sputum granulocytes

Analyses of asthma severity phenotypes and inflammatory proteins in subjects stratified by sputum granulocytes Annette T. Hastie, PhD, Wendy C. Moore, MD, Deborah A. Meyers, PhD, Penny L. Vestal, MS, Huashi Li, MS, Stephen P. Peters, MD, PhD, Eugene R. Bleecker, MD, and the National Heart, Lung, and Blood Institute Severe Asthma Research Program Winston-Salem, NC Background: Patients with severe asthma have increased granulocytes in their sputum compared with patients with mild to moderate asthma. Objective: We hypothesized that inflammatory granulocytes in sputum may identify specific asthma severity phenotypes and are associated with different patterns of inflammatory proteins in sputum supernatants. Methods: This hypothesis was tested in 242 patients with asthma enrolled in the Severe Asthma Research Program who provided sputum samples for cell count, differential cell determinations, cell lysates for Western blot, and supernatant analyses by inflammatory protein microarrays and ELISAs. ANOVA and multiple linear regression models tested mediator associations. _2% Results: Stratified by sputum granulocytes, <2% or > _40% neutrophils, subjects with both eosinophils and <40% or > increased eosinophils and neutrophils had the lowest lung function and increased symptoms and health care use. Subjects with elevated eosinophils with or without increased neutrophils had significantly increased fraction exhaled nitric oxide (FeNO) and serum eosinophils and greater frequency of daily b-agonist use. Microarray data stratified by granulocytes revealed 25 to 28 inflammatory proteins increased >2-fold in _40% neutrophils. Microarray analyses stratified sputa with > by severity of asthma identified 6 to 9 proteins increased >2-fold in sputa in subjects with severe asthma compared with nonsevere asthma. ELISA data stratified by sputum granulocytes showed significant increases in brain-derived neurotrophic factor, IL-1b, and macrophage inflammatory _40% neutrophils; these protein 3a/CCL20 for those with > mediators demonstrated positive associations with neutrophil counts. Conclusion: Combined increased sputum eosinophils and neutrophils identified patients with asthma with the lowest lung

From the Center for Human Genomics, Wake Forest University Health Sciences. Supported by the NHLBI Severe Asthma Research Program Awards (HL69167) and the Wake Forest University General Clinical Research Center (MO1 RR07122). Disclosure of potential conflict of interest: D. A. Meyers and S. P. Peters have received research support from the National Institutes of Health/National Heart, Lung, and Blood Institute. The rest of the authors have declared that they have no conflict of interest. Received for publication June 9, 2009; revised November 12, 2009; accepted for publication February 5, 2010. Available online April 16, 2010. Reprint requests: Annette T. Hastie, PhD, Center for Genomics and Personalized Medicine Research, NRC-G70, Wake Forest University Health Sciences, Winston-Salem, NC 27157. E-mail: [email protected]. 0091-6749/$36.00 Ó 2010 American Academy of Allergy, Asthma & Immunology doi:10.1016/j.jaci.2010.02.008

1028

function, worse asthma control, and increased symptoms and health care requirements. Inflammatory protein analyses of sputum supernatants found novel mediators increased in patients with asthma, predominantly associated with increased sputum neutrophils. (J Allergy Clin Immunol 2010;125:102836.) Key words: Asthma phenotypes, protein microarrays, BDNF, CXCL13, TNFSF14, CCL20, CCL18

Eosinophils and neutrophils have each been separately observed in sputum from severe, poorly controlled, or persistent asthma.1-7 Severe asthma, as well as milder asthma, is therefore heterogeneous and composed of subgroups with potentially different underlying inflammatory pathologies. Subjects may be stratified according to sputum granulocytes to determine whether inflammatory differences are associated with asthma severity phenotypes. In the past, patients with asthma with both elevated eosinophils and elevated neutrophils were assigned to either the group with high eosinophils8 or the group with high neutrophils.9 This approach potentially obscures distinguishing features for mixed granulocytic inflammation. One study evaluated these subjects as a separate group but found no significant differences in clinical characteristics between the groups except for age.10 The small number of subjects with mixed granulocytes may have reduced differentiation from other phenotypes. We hypothesized that subjects with asthma with both granulocytes increased in sputum represent a separate phenotypic group different from those with only increased eosinophils, only increased neutrophils, or neither granulocyte increased. In addition, only a few reports have investigated protein constituents of sputum, despite recognition that mucosal leukocytic infiltration is regulated by inflammatory mediator release. Earlier reports have measured limited numbers of proteins present in sputum, primarily those predicted for a single cell type, such as eosinophil cationic protein or IL-8, and compared these to normal controls.11-14 We hypothesized that the patterns of infiltrating leukocytes in sputum are determined by differences in the inflammatory proteins contributing to asthmatic phenotypes. To test these hypotheses, comprehensively characterized subjects with severe and nonsevere asthma enrolled in the Severe Asthma Research Program of the National Heart, Lung, and Blood Institute (SARP) provided sputa and were stratified on the basis of granulocyte percentages. A subset of sputa was screened with a commercial protein microarray assessing 120 inflammatory proteins to identify mediators differing between subjects characterized by both granulocytes increased, eosinophils

HASTIE ET AL 1029

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

Abbreviations used BDNF: Brain-derived neurotrophic factor BLC: B-lymphocyte chemoattractant/CXCL13 EGF: Epidermal growth factor FVC: Forced vital capacity ICS: Inhaled corticosteroid MCP: Monocyte chemoattractant protein MIP-3a: Macrophage inflammatory protein 3a/CCL20 PARC: Pulmonary and regulated chemokine/CCL18 SAM: Significance Analysis of Microarrays SARP: Severe Asthma Research Program of the National Heart, Lung, and Blood Institute TNFSF14: TNF superfamily factor 14/LIGHT, Lymphotoxin homologous, Inducible expression, competes with HSV Glycoprotein D for HVEM receptor on T lymphocytes

increased, neutrophils increased, or neither granulocyte increased. Selected proteins were confirmed by standard ELISA in sputum supernatants from the larger group of patients with asthma. Mediator levels were examined for association with specific cell counts and clinical characteristics. Sputum cell lysates were analyzed by Western blotting for the presence of specific mediators.

METHODS Characterization of severe and not severe asthma was performed according to the SARP protocol.15 Nonsmoking subjects (<5 pack-years) met American Thoracic Society criteria for diagnosis of asthma and provided informed consent approved by the institutional review board. Comprehensive evaluation included spirometry, bronchodilator reversibility and bronchial responsiveness, assessment of atopy, collection of blood, exhaled nitric oxide, sputum induction, and an administered questionnaire that characterized asthma symptoms, quality of life, medications, and health care use15 (see Methods in this article’s Online Repository at www.jacionline.org).

Subjects Subjects from SARP with severe (N 5 48 as defined by the ATS Workshop on Refractory Asthma16) and nonsevere asthma (N 5 194, subjects with asthma not meeting ‘‘severe’’ criteria), enrolled at Wake Forest University, had sputum induction, or, if safety criteria were not met, a spontaneous sputum collected. Wake Forest University collected a majority of the sputum samples, and not all SARP sites participated in the sputum protocol. Demographics and clinical characteristics for the subjects in the sputum subset matched those previously reported for SARP subjects from all sites15 (see this article’s Table E1 in the Online Repository at www.jacionline.org).

Sputum induction and processing The sputum induction method was adopted from the Asthma Clinical Research Network.17 Sputum was processed immediately; cell cytospins were stained for differential count of leukocytes and bronchial epithelial and squamous cells, and aliquots of sputum supernatant were stored at –808C. The sputum cell differential counts from 175 subjects were adequate for further analyses.17 Cell lysates were examined by Western blotting for the presence of specific proteins (see Methods in this article’s Online Repository at www. jacionline.org).

Inflammation microarrays and ELISAs Inflammation protein microarray analyses (RayBiotech, Norcross, Ga) were performed on sputum supernatants from subjects with asthma (N 5 12

nonsevere, no inhaled corticosteroid [ICS] treatment; N 5 12 nonsevere, ICS-treated; and N 5 12 severe, high-dose ICS treatment) diluted to 1 mg total protein (see Methods in this article’s Online Repository at www.jacionline. org). Densities of duplicate reactions were averaged and normalized to controls on each membrane. Proteins identified from analyses based on corticosteroid use, granulocyte percentage, or asthma severity were investigated by specific ELISA (R & D Systems, Minneapolis, Minn; except pulmonary and regulated chemokine/CCL18 [PARC] and eotaxin 2 from RayBiotech, Norcross, Ga).

Statistics Microarray density data were log2-transformed and analyzed by Significance Analysis of Microarrays (SAM).18 A score greater than 1, a false discovery rate less than 5%, and a greater than 2-fold change were the criteria for significance. The SAM program is designed to address a chance identification problem for larger datasets than the 120-protein microarray.18 Demographic and ELISA data are presented as means 6 SDs, SEs, or medians (25% to 75% quartiles). Measures not meeting Kolmogorov-Smirnov test for normal distribution were transformed to log or square root values. A 0 in cell differentials or ELISA assays was replaced with a value half of the lowest observed before log transformation.19 Continuous variables were tested by ANOVA, or by Student t test if parametric, or by Kruskal-Wallis and Mann-Whitney if nonparametric (SAS 9.2, SAS Institute, Inc, Cary, NC; or Sigmastat 2.03, Systat Software, San Jose, Calif). Initial analyses with a significant difference were further explored by post hoc pairwise analyses (Tukey). Categorical variables were analyzed by using x2 tests. Multivariate linear regression models examined mediators for association with sputum cells, age, and clinical characteristics. Bonferroni correction was applied to variables with a P value <.05 to determine significance.

RESULTS Subject stratification on the basis of sputum inflammation percent eosinophils and percent neutrophils There were no significant differences in sputum cell percentages for severe compared with nonsevere subjects (see this article’s Table E2 in the Online Repository at www.jacionline. org), and therefore, these were combined for stratification by granulocytes: <2% eosinophils 1 <40% neutrophils, <2% eosin_40% neutrophils, > _2% eosinophils 1 <40% neutroophils 1 > _2% eosinophils 1 > _40% neutrophils (Table I and phils, and > see Table E7 in this article’s Online Repository at www. jacionline.org). The group with both increased eosinophils and increased neutrophils had lowest lung function (FEV1% predicted, FEV1/forced vital capacity [FVC] ratio, change FEV1% predicted [maximum – baseline]), highest frequency of daily wheeze, and most frequent health care use (Table I). Subjects _2% eosinophils and <40% neutrophils had the highest with > _40% neutrophils, with or withFeNO (P < .001). Subjects with > _2% eosinophils, were significantly older, but there was no out > difference in age of asthma onset. No differences were observed for positive skin test frequency or number. A greater frequency of positive responses to ‘‘daily use of b-agonist’’ and daily _2% eosinophils with or without wheeze were associated with > > _40% neutrophils, whereas a greater frequency of positive responses to ‘‘>1 urgent health care visit in the past year’’ was _40% neutrophils with or without > _2% eosinoassociated with > phils. Although only >1 urgent health care visit in the past year reached significance, other measures of health care use were higher in the combined increased granulocyte group than the other groups.

1030 HASTIE ET AL

J ALLERGY CLIN IMMUNOL MAY 2010

TABLE I. Demographics of subjects stratified by sputum percentage eosinophils 1 percentage neutrophils Categories

N 5 175 Sex Female Age (y) Age at onset (y) Duration (y) Lung function Baseline FEV1 % Predicted Baseline FVC % Predicted FEV1/FVC Maximum FEV1% Predicted %Reversible to 2 puffs b-agonist Change FEV1% predicted (maximum – baseline) Log PC20* FeNO (ppb) Freq 1 skin test No. 1 skin tests Serum Eos IgE Asthma control and health care use  Daily use b-agonist Ever ED visits >1 urgent HC visit in past year Symptoms worse if reduce Corticosteroid >3 oral Corticosteroid bursts/y Ever spent night in hospital Ever admitted to ICU Symptoms  >Daily wheeze

<2% Eos <40% Neu

<2% Eos _40% Neu >

_2% Eos > <40% Neu

_2% Eos > _40% Neu >

63 (36%)

50 (29%)

42 (24%)

20 (11%)

81% 33 6 1.6 10 (4-24) 17 6 1.6

76% 42 6 1.9 13 (5-26) 25 6 2.1

74% 34 6 1.8 13.5 (2-23) 19 6 1.9

60% 40 6 2.1 7.5 (3-30) 23 6 2.5

.30 <.001 .92 .014

6 6 6 6 6

76.6 6 2.8 90.5 6 2.7 0.70 6 0.02 90.3 6 2.6 14.5 6 12.3

66.6 6 4 81.7 6 3.7 0.66 6 0.03 80.8 6 4.2 15.8 6 15.5

<.001 .01 <.001 .003 .005

13.7 6 10.3

14.7 6 10.1

.002

–0.23 6 0.3 N 5 14 30.9 6 1.2 89% 4 (2-7) 0.30 (0.2-2.1) 100 6 1.3

.006 <.001 .66 .59 .001 .06

86.5 93.8 0.77 95.4 9.89

6 6 6 6 6

2.0 1.8 0.01 1.8 10

80.4 87.0 0.75 88.8 7.98

2.3 2.0 0.01 1.9 5.8

P value

9.1 6 6.5

8.3 6 6.4

0.47 6 0.1 N 5 55 22.9 6 1.1 83% 4 (2-6.8) 0.20 (0.1-3.3) 126 6 1.2

0.41 6 0.1 N 5 46 20.0 6 1.1 78% 3 (1-5.3) 0.20 (0.1-2.5) 78 6 1.3

24% 57% 22% 49%

28% 70% 40% 39%

60% 69% 19% 38%

55% 80% 70% 70%

<.001 .21 <.001 .07

17% 25% 6%

12% 40% 8%

9% 38% 21%

30% 40% 25%

.17 .07 .009

13%

14%

26%

50%

.002

–0.013 6 0.1 N 5 34 49.0 6 1.1 86% 4 (2-7.3) 0.30 (0.2-2.5) 191 6 1.3

Results from ANOVA are presented as means 6 SEMs; results from Kruskal-Wallis analysis of variance are presented as median (25% to 75%), P values meeting Bonferroni correction are in boldface. ED, Emergency department; Eos, eosinophils; HC, health care; ICU, intensive care unit; Neu, neutrophils; Freq + skin test gives the % of the group with at least one 1 skin test; No. 1 skin tests gives the median (quartiles) number of 1 skin tests. *Not all subjects met safety criteria for methacholine challenge.  Percentage of category answering ‘‘yes.’’

Stratification of microarray data on the basis of sputum percent eosinophils and percent neutrophils Microarray data were analyzed on the basis of sputum eosinophils and neutrophils of groups divided by granulocyte percentage. Corticosteroid effects20,21 were considered by first comparing nonsevere without corticosteroid treatment to nonsevere with corticosteroid treatment. Only monocyte chemoattractant protein (MCP)–3 was significantly reduced >2-fold in the corticosteroid treatment group. Nevertheless, to control for corticosteroid effects, only ICS-treated subjects were examined in the comparison of subjects with primarily elevated eosinophils (N 5 6) with subjects with primarily elevated neutrophils (N 5 7). Twenty-five proteins were increased >2-fold in the group with elevated neutrophils (Table II, comparison 1). For subjects with elevated eosinophils (N 5 6) compared with subjects with both granulocytes elevated (N 5 8), 30 proteins were significantly increased >2-fold for those subjects with both granulocytes elevated

(Table II, comparison 2). Eight new proteins were observed in this comparison, and 22 of the increased proteins were previously observed in comparison 1. Subjects with elevated neutrophils (N 5 7) were also compared with those with both granulocytes elevated (N 5 8) but showed no protein differences despite differing percentages of eosinophils (not shown). Microarray data stratified by subjects with <2% eosinophils _2% eosinophils, without considering percompared with > centage of neutrophils, showed just 2 proteins, IL-11 (2.2X) and IL-2 receptor a (2.3X), significantly increased >2-fold _2% eosin(false discovery rate, 0%) in those subjects with > ophils (not shown). In contrast, microarray data stratified by _40% neutrosubjects with <40% neutrophils compared with > phils without considering percentage of eosinophils had 28 _40% proteins significantly increased in those subjects with > neutrophils (Table II, comparison 3). The 3 comparisons identified 19 proteins increased in common (underlined). Limited supernatant fluid available required reducing the

HASTIE ET AL 1031

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

TABLE II. Three comparisons of inflammatory protein microarray data in subjects (all treated with corticosteroids) stratified by sputum eosinophils and neutrophils _2% Eos1 <40% Neu (N 5 6) vs <2% Eos1 1. > _40% Neu (N 5 7) > False-positive rate, 0% to 1% Protein

BDNF BLC BMP-4 BMP-6 CK b8-1 EGF Eotaxin-2 Eotaxin-3 FGF-7 GCP-2 GDNF GM-CSF HGF I-309 IFN-g IGFBP-1 IGFBP-2 IGF-1 IL-1b IL-2 IL-4 IL-6 IL-10 IL-13 IL-15 LIGHT MCP-1 MCP-2 MDC MIP-1d MIP-3a NT-3 SCF SDF-1 TARC TIMP-2 TNF-a TRAIL R3 uPAR

Criteria failure*

_40% Neu Increase for > fold change

_2% Eos1 <40% Neu (N 5 6) vs > _2% Eos 2. > _40% Neu (N 5 8) 1> False-positive rate, 0% Criteria failure*

a a

3.3 3.2 3.8 6.2

3.3 5.5 a, b

Criteria failure*

a a, b a a

3.4 5.3 3.6 2.5 3.2 2.5 3.8

2.2 2.4 3.2 2.2 2.0 2.1 2.7 b 2.7 a, b b

NS

3.1 2.3

2.6 2.6 11.7 3.1 9.0 4.1 5.9 3.1 2.7 2.8 2.1 2.1

a 2.7 16.1 3.3 8.4 2.9 4.2 3.5 3.8 a

3.8 a, b 3.9 3.0 2.4 2.4 a a NS

b a, b 2.1 7.7 b 6.8 2.1 2.2 3.2 2.1 a

3.5 3.3 2.7 2.5 3.0 3.6 4.8

a

2.4 2.8 2.4 a b 2.6 3.0 2.3 2.1

a, b 2.6 3.3 a, b NS

4.8

5.5 NS NS

_40% Neu Increase for > fold change

a

b 2.9 3.0 3.6 2.2 4.0

NS NS

_40% Neu Increase for > fold change

_40% Neu (N 5 15) 3. <40% Neu (N 5 9) vs > False-positive rate, 0% to 1%

a, b 2.4 2.1 2.8 2.5 3.3

Comparison 1 examines differences in mediators between elevated Eos 1 low Neu and elevated Neu 1 low Eos; comparison 2 examines differences between elevated Eos 1 low Neu and elevated Eos 1 elevated Neu; comparison 3 examines differences between low Neu and elevated Neu without considering Eos. Underlined proteins were observed to meet criteria in all 3 comparisons. BMP, Bone morphogenic protein; CK b8-1, creatine kinase b8-1; Eos, eosinophils; FGF-7, fibroblast growth factor-7; GCP-2, granulocyte chemotactic protein-2; GDNF, glial cell-derived neurotrophic factor; HGF, hepatocyte growth factor; IGFBP, insulin-like growth factor binding protein; IGF, insulin-like growth factor; MDC, macrophage derived chemokine; Neu, neutrophils; NT, neutrophin; SCF, stem cell factor; SDF, stromal cell derived factor; TARC, thymus and activation regulated chemokine; TIMP, tissue inhibitor of metalloproteinases; uPAR, urokinase type plasminogen activator receptor. *NS, Not significant (q>5%); a, scored <1; b, fold increase <2.

large number of potential biomarkers for specific ELISA confirmation. Therefore, microarray data were alternatively stratified by asthma severity.

Stratification of microarray data on the basis of asthma severity Comparisons of microarray data for subjects with nonsevere asthma without and with ICS treatment (N 5 12 and N 5 12,

respectively) with subjects with severe asthma (N 5 12) included (1) all subjects with nonsevere asthma with severe asthma, (2) subjects with nonsevere asthma receiving corticosteroids with subjects with severe asthma, and (3) subjects with nonsevere asthma receiving corticosteroids and FEV1% predicted >80% with subjects with severe asthma. These 3 comparisons yielded 9, 6, and 33 proteins, respectively, with significant increases >2-fold in sputa of subjects with severe asthma. Three proteins matched in all 3 comparisons (highlighted in this article’s Table E3 in the Online

1032 HASTIE ET AL

J ALLERGY CLIN IMMUNOL MAY 2010

TABLE III. Sputum supernatant mediator levels stratified by sputum percentage eosionphils and percentage neutrophils

BDNF pg/mL BLC/CXCL13 pg/mL BMP-4 pg/mL EGF pg/mL Eotaxin 2/CCL24 pg/mL IFN-g pg/mL IL-1b pg/mL IL-8 ng/mL IL-13 pg/mL LIGHT/TNFSF14 pg/mL MIP-3a/CCL20 pg/mL PARC/CCL18 pg/mL TNF-a pg/mL

<2% Eos 1<40% Neu

_40% Neu <2% Eos 1 >

_2% Eos 1 <40% Neu >

_2% Eos 1 > _40% Neu >

P value for 4 groups*

9.5 (6-14.5) 110 6 29 4.9 6 1.5 159 (108-205) 0.88 (0.01-2.38) 77 (37-149) 104 6 14 1.5 6 0.1 91 6 18 36 (12-69) 390 6 41 6.7 (0.4-19) 0.3 (0.01-1.24)

18.4 (11.2-29) 223 6 48 2.8 6 0.6 190 (116-263) 1.13 (0.01-6.78) 47 (2-128) 224 6 56 2.1 6 0.2 74 6 9.9 92 (36-156) 781 6 85 12.7 (4.7-59) 1.1 (0.01-2.6)

15.4 (8-26.5) 94 6 20 1.8 6 0.3 171 (90-212) 2.39 (0.38-4.63) 108 (49-175) 76 6 12 1.6 6 0.2 64 6 12 33 (13-108) 341 6 52 8.3 (1.7-22) 0.39 (0.01-1.9)

20 (14-39) 142 6 49 3.7 6 1.1 199 (145-256) 8.90 (2.5-14.1) 91 (23-125) 228 6 65 1.9 6 0.2 81 6 18 58 (24-184) 668 6 121 11 (4.6-40) 0.7 (0.01-8.1)

<.001 .005 .33 .13 .022 .11 <.001 .017 .55 .021 <.001 .037 .44

P values meeting Bonferroni correction are in boldface. BMP, Bone morphogenic protein; Eos, eosinophils; Neu, neutrophils. *ANOVA (mean 6 SEM) or Kruskal-Wallis (median [25% to 75% quartiles]).

Repository at www.jacionline.org): brain-derived neurotrophic factor (BDNF), B-lymphocyte chemoattractant (BLC/CXCL13), and epidermal growth factor (EGF). The comparison of nonsevere with mild asthma to severe asthma found 25 proteins significantly increased >2-fold in the subjects with severe asthma that were not observed in the other 2 comparisons (see this article’s Table E4 in the Online Repository at www.jacionline.org). On the basis of the microarray results, proteins were selected for ELISA assay on the larger panel of sputa available (N 5 175). IL-13 and IFN-g were also examined although identified in only 1 microarray comparison (Table II, comparison 1).

ELISAs on selected inflammatory mediators stratified by sputum percent eosinophils and percent neutrophils ELISA data stratified into the 4 groups on the basis of sputum _2% eosinophils 1 <40% or > _40% granulocytes, <2% or > neutrophils, revealed significant increases in mediators for those _40% neutrophils (Table III). BDNF, IL-1b, and subjects with > macrophage inflammatory protein 3a (MIP-3a/CCL20) were sig_40% neutrophils, either nificantly increased in sputa containing > _2% eosinophils. Only Eotaxin 2 with <2% eosinophils or with > _2% eosinophils, which showed an increase associated with > _40% neutrophils, but did was enhanced in combination with > not reach significance after correction. Associations of mediators with leukocytes and clinical characteristics Associations of each inflammatory mediator’s concentration with the actual count of specific cell types occurring in the sputum sample were investigated (Table IV). Age is positively associated with percentage of neutrophils22 and therefore is included as an independent variable. Only neutrophils were observed to have positive associations with those mediators showing significance. Despite positive association of neutrophils with age and with several mediators, age independently showed associations with only LIGHT/TNF superfamily factor 14 (TNFSF14) and PARC/ CCL18, which were not significant. Inflammatory cells and mediators were also examined for association with clinical characteristics including severity,

spirometric measures, FeNO, number of positive skin tests, and IgE levels (see this article’s Results, Fig E1, Table E5, and Table E6 in the Online Repository at www.jacionline.org). IL-8 and PARC/CCL18 had negative associations with baseline FVC % predicted, and LIGHT/TNFSF14 had a negative association with baseline FEV1 % predicted. TNF-a had a positive association with FeNO. Neither IgE nor number of positive skin tests showed associations with any mediator.

Western blots of sputum cell lysates probed for mediator presence Sputum cell lysates were analyzed by Western blots for BDNF and IL-1b. Mean densities for both BDNF and IL-1b in lysates from subjects with combined elevated eosinophils 1 neutrophils were higher than in those with elevated eosinophils, elevated neutrophils, or neither granulocyte elevated but did not reach significance (see this article’s Fig E2 in the Online Repository at www.jacionline.org). DISCUSSION Sputum represents the best available noninvasive assessment of bronchial inflammation in asthma and reflects underlying pathology caused by infiltrating cells and soluble mediators. Instead of comparing sputum cellular and biochemical components between healthy subjects and subjects with asthma, our objective was to assess whether comprehensive analysis of induced sputum over a spectrum of asthma severity improves our understanding of the factors that characterize different asthma phenotypes. Currently, definitions of asthma severity are based primarily on the recommendations of asthma guidelines and on consensus statements. Unfortunately, these definitions may not rely on evidence-based science for classification of asthma severity. Currently, the SARP network is investigating new approaches to classify asthma severity, including new imaging techniques, unbiased cluster methods23 similar to these reported by Haldar et al,24 and the assessment based on sputum granulocytes described in this article. Our hypothesis that sputum inflammatory granulocytes identify phenotypic subgroups of differing pathology and clinical characteristics was examined in a unique, well characterized SARP population including patients with and without corticosteroid treatment.

HASTIE ET AL 1033

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

TABLE IV. Multiple linear regression analyses for mediators significantly associated with specific sputum cells and age Cytokine or growth factor

BDNF BLC/CXCL13 BMP-4 EGF Eotaxin 2 IL-1b IL-8 IL-13 IFN-g LIGHT/TNFSF14 MIP-3a/CCL20 PARC/CCL18 TNF-a

R and P value

Bronchial Epithelial cell count*

0.51 <.001 0.388 .008 0.393 .122 0.508 <.001 0.27 .29 0.486 <.001 0.585 <.001 0.171 .86 0.324 .136 0.651 <.001 0.621 <.001 0.553 <.001 0.567 <.001

0.414 .032

Macrophage/ Monocyte count*

Lymphocyte count*

Neu count*

0.769 <.001 0.265 .003

32.6 .014

Eos count*

Age*

0.238 .064 –0.131 .065 –0.197 .0167

26.1 .02 0.28 .02

–0.12 .06

0.174 .016 0.382 <.005

0.569 .033

19.21 .08 3.76 .014

0.613 .062

2.22 .007 5.59 <.001 0.338 .006

0.097 .018 –2.24 .014 0.184 .025

0.016 .026

0.458 .007

The overall R and P values for each mediator are reported with the coefficient and individual P values for specific cell types or age necessary to predict a linear model for the mediator. P values meeting Bonferroni correction are in boldface. BMP, Bone morphogenic protein; Eos, eosinophils; Neu, neutrophils. *Values under cell type are regression coefficient and P value for that specific cell. Blank indicates that the cell type or age was not found to be necessary for linear model of association with the cytokine.

Other investigators have observed that asthma, including severe asthma, contains subgroups categorized by sputum granulocytes, eosinophilic or neutrophilic, or ‘‘noneosinophilic,’’25,26 and further subdivided in a 4-way stratification based on both eosinophils and neutrophils.10 However, this latter report observed only age differences for a small number of subjects with mixed granulocytes. In contrast, our larger groups stratified by granulocyte percentages showed significant differences in lung function, asthma control, health care use, and symptoms. Although most of the soluble mediators examined did not differ between the elevated eosinophils 1 elevated neutrophils group compared with the elevated neutrophils group, Eotaxin 2 and clinical characteristics support a distinction between these 2 groups. Phenotypic characteristics associated with the 4 asthma groups revealed that combined increased sputum eosinophils and neutrophils defined the most severe patients with lowest lung function measures, worse asthma control, and greatest symptoms and use of health care resources. Inflammatory mediators in sputum have been reported for specific mediators, predicted by increases in specific cell types such as eosinophil cationic protein (ECP) for eosinophils or IL-8 for neutrophils.12,26,27 Additional mediators reported in sputum for asthma include vascular endothelial growth factor, basic fibroblast growth factor, elastase, GM-CSF, IFN-g, IL-4, IL-5, IL-13, endostatin, MIP-3a/CCL20, metalloproteinase 9, tissue-inhibitor metalloproteinase 1, PARC, RANTES, and a1-antitrypsin.12,28-33 However, most increased inflammatory proteins are reported for subjects with

asthma compared with healthy subjects, instead of within-asthma severity subgroups, in which proteins may reflect important mechanisms determining disease severity. We used a more comprehensive approach with a focused 120– inflammatory protein microarray to identify mediators in sputum supernatants from patients with severe and nonsevere asthma. A recent report28 used a similar protein microarray approach but examined fewer proteins and compared only 9 patients with asthma with atopic and normal controls. PARC, growth regulated protein alpha (GRO-a), and eotaxin-2 were elevated in sputum from subjects with asthma, but subgroups of asthma were not investigated.28 A 4-group stratification of our microarray data by sputum percent eosinophils and percent neutrophils identified significant _40% increases in 19 inflammatory mediators in sputa with > neutrophils. ELISA data confirmed the significant increases in _40% neutrophils, either with <2% eosinophils or with sputa with > > _2% eosinophils, for BDNF, IL-1b, and MIP-3aa/CCL20. Eotaxin 2 increases in ELISA data were observed with increased eosinophils but did not reach significance. Nevertheless, higher levels of Eotaxin 2 were observed for sputum having combined increased eosinophils and increased neutrophils. Regression analyses showed significant positive associations of mediators with neutrophil counts, and less robust positive or negative associations with eosinophils counts. The positive association of neutrophils with BDNF, IL-1b, and MIP-3a raises the question whether neutrophils produce or are

1034 HASTIE ET AL

elicited by the actions of these proteins. Immunohistology in mouse models and isolated human peripheral monocytes identified cellular sources for BDNF as bronchial epithelium, activated T cells, and macrophages.34,35 Likewise, IL-1b is synthesized by multiple cells, including macrophages, epithelial cells, and neutrophils.36-38 Both BDNF and IL-1b levels in our sputum cell lysates were higher for combined increased eosinophils 1 neutrophils but not significantly, suggesting contribution from other cells. Cell sources for MIP-3a include airway epithelium, monocytes, eosinophils, and neutrophils; release is either constitutive or in response to LPS, particulates, allergens, or TNF-a, IL-1b, IL-4, or IL-13.33,39,40 Thus, any one or a combination of sputum cell types may contribute to the levels of MIP-3a observed, including neutrophils. The results demonstrate that mediators previously less well recognized for involvement in asthmatic inflammation may be identified by this approach. BDNF, BLC/CXCL13, and EGF were found to be significantly increased by microarray analysis of sputum from subjects with severe compared with nonsevere asthma. ELISA results on the larger panel of sputum supernatants showed that higher levels of these mediators occurred in the severe group but did not reach significance (see Table E5 in this article’s Online Repository at www.jacionline.org). Brain-derived neurotrophic factor has been implicated in bronchial hyperresponsiveness and inflammation41 but not specifically connected to asthma severity. Patients with allergic asthma have higher levels of BDNF, promoting eosinophil survival, in bronchoalveolar lavage fluid after segmental allergen challenge.35,42 Both epithelial cells and monocytes release BDNF in response to TNF-a, IL-1b, and IL-6.35,43,44 B-lymphocyte chemoattractant/CXCL13 has not previously been associated with asthma severity. CXCL13 upregulation in signal transducer and activator of transcription 6–deficient mice after repeated allergen challenge is associated with neutrophilic inflammation.45 Similarly, we observed a positive association of CXCL13 with neutrophils. Monocytes and mature macrophages treated with LPS secrete CXCL13.46 IL-6 signaling results in accumulation of B-cell follicles in the lung expressing CXCL13.47 IL-6 was increased in microarrays for sputa with elevated neutrophils. Epidermal growth factor stimulates IL-8 release and enhances TNF-a–induced IL-8 release from epithelium, and is thereby linked to neutrophilic inflammation in severe asthma.48,49 There is evidence of increased EGF receptor in bronchial biopsies from severe compared with mild asthma,48 but EGF either was similar50 or increased51 in 2 studies examining small numbers of patients with asthma compared with controls without asthma. Our study confirms association of EGF with neutrophils and higher amounts in subjects with severe asthma, although not significantly. Paired microarray comparisons of dialyzed with autologous undialyzed samples and of protease inhibitor–treated supernatants with autologous untreated supernatants indicated minimal consistent alteration of sputum proteins as a result of either dithiothreitol use or proteolysis detected by microarrays. Moreover, known amounts of standards for each ELISA generally displayed full recovery when assessed in normal sputum supernatants or buffer solution containing dithiothreitol (see Methods in this article’s Online Repository at www.jacionline.org). The processing method using whole sputum potentially contributes variable amounts of saliva but is counteracted by a 2-step collection.52 In fact, percentage of squamous cells did not differ

J ALLERGY CLIN IMMUNOL MAY 2010

for any stratification. Observed significant association of mediators in the supernatant with actual cell counts further supports that variable amounts of saliva had little dilutional effect on mediators. Woodruff et al21 reported that age, sex, ethnicity, and use of ICSs were important confounders of cellular inflammation in asthma. This contrasts with the report of Thomas et al,22 which demonstrated that age significantly affected percentage neutrophils but sex did not in a normal population. We confirm a positive association between age and percentage of neutrophils in patients with asthma. Inclusion of age into our models examining mediator association with specific sputum cell types showed associations of age with LIGHT/TNFSF14 and PARC/CCL18 (neither significant) but otherwise had little effect on neutrophils’ association with mediators. Subject stratification by sputum granulocytes did not show a difference in age of asthma onset, although subjects with a greater percentage of neutrophils were older and therefore possibly had a longer duration of asthma. We did not find any significant effect for race or for corticosteroid use on sputum cell counts or on mediator levels. The white and black proportions in our population (53% to 77% and 23% to 45%, respectively) are slightly different than those reported by Woodruff et al21 (68% white and 12% black); however, lack of an effect for race/ethnicity in our data cannot be attributed to insufficient numbers of minority subjects. Certain mediators identified, IL-1b, IL-8, and TNF-a, are usually associated with innate immunity and a TH1 response, rather than the TH2 mediators, IL-4, IL-5, and IL-13, typical of allergic asthma. This finding is consistent with the heterogeneity of clinical asthma that may be influenced by different pathologic mechanisms. However, we did observe increased MIP-3a/ CCL20, recently reported as a critical link between tumor necrosis factor-related apoptosis-inducing ligand (TRAIL) and the activation of TH2 cells in allergic airway disease.33 TNF-a, which we found increased in microarrays, induces increased MIP-3a/ CCL20 secretion from primary bronchial epithelium,33 indicating overlap of TH1 and TH2 inflammatory pathways. It is interesting that Kikuchi et al53 reported both IL-8 and neutrophils were necessary to promote greatest trans-basement membrane migration of eosinophils. The group of subjects with both elevated sputum eosinophils and neutrophils had equally high if not the highest levels of mediators, supporting this observation. In conclusion, stratification of patients with asthma by sputum percentage of eosinophils and percentage of neutrophils demonstrated that combined increased granulocytes identify those patients with asthma with lowest lung function and increased frequency of symptoms and health care use. Protein microarray screening of sputa identified novel proteins that have been less well recognized for participation in asthma and suggested a TH1 component to inflammation in more severe asthma. The levels of these mediators were generally higher in subjects with severe asthma but showed a stronger association with neutrophils than with eosinophils. These approaches delineating cellular and biochemical proteomics provide better understanding of the pathogenesis of asthma subphenotypes and may lead to development of biomarkers differentiating heterogeneity in asthma. We acknowledge the essential contributions of Jeffrey Krings, RN, and Regina Smith, clinical coordinators; Xingnan Li, statistician; and Min Wu, laboratory technician, in the recruitment, characterization, and analyses of all subjects enrolled in this study.

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

Key messages d

Four-way stratification of subjects, based on high or low percentage of eosinophils and high or low percentage of neutrophils in sputum, identified an association of subjects characterized by both high percentage of eosinophils and high percentage of neutrophils with the lowest lung function, worse asthma control, and increased use of health care resources.

d

Protein microarray data revealed novel proteins increased in sputum with both increased neutrophils and eosinophils, suggesting the importance of combined granulocyte effects on asthma severity.

REFERENCES 1. Louis R, Lau LCK, Bron AO, Roldaan AC, Radermecker M, Djukanovic R. The relationship between airways inflammation and asthma severity. Am J Respir Crit Care Med 2000;161:9-16. 2. The ENFUMOSA Study Group. The ENFUMOSA cross-sectional European multicentre study of the clinical phenotype of chronic severe asthma. Eur Respir J 2003;22:470-7. 3. Bartoli ML, Bacci E, Carnevali S, Cianchetti S, Dente FL, DiFranco A, et al. Clinical assessment of asthma severity partially corresponds to sputum eosinophilic airway inflammation. Respir Med 2004;98:184-93. 4. Lemiere C, Ernst P, Olivenstein R, Yamauchi Y, Govindaraju K, Ludwig MS, et al. Airway inflammation assessed by invasive and noninvasive means in severe asthma: eosinophilic and noneosinophilic phenotypes. J Allergy Clin Immunol 2006;118:1033-9. 5. Green RH, Brightling CE, Woltmann G, Parker D, Wardlaw AJ, Pavord ID. Analysis of induced sputum in adults with asthma: identification of subgroup with isolated sputum neutrophilia and poor response to inhaled corticosteroids. Thorax 2002;57:875-9. 6. Jatakanon A, Uasuf C, Maziak W, Lim S, Chung KF, Barnes PJ. Neutrophilic inflammation in severe persistent asthma. Am J Respir Crit Care Med 1999;160:1532-9. 7. ten Brinke A, Zwinderman AH, Sterk PJ, Rabe KF, Bel EH. Refractory" eosinophilic airway inflammation in severe asthma: effect of parenteral corticosteroids. Am J Respir Crit Care Med 2004;170:601-5. 8. Berry M, Morgan A, Shaw DE, Parker D, Green R, Brightling C, et al. Pathological features and inhaled corticosteroid response of eosinophilic and non-eosinophilic asthma. Thorax 2007;62:1043-9. 9. Simpson JL, Grissell TV, Douwes J, Scott RJ, Boyle MJ, Gibson PG. Innate immune activation in neutrophilic asthma and bronchiectasis. Thorax 2007;62:211-8. 10. Simpson JL, Scott R, Boyle MJ, Gibson PG. Inflammatory subtypes in asthma: assessment and identification using induced sputum. Respirology 2006;11:54-61. 11. Daldegan MB, Teixeira MM, Talvani A. Concentration of CCL11, CXCL8, and TNF-a in sputum and plasma of patients undergoing asthma or chronic obstructive pulmonary disease exacerbation. Braz J Med Bio Res 2005;38:1359-65. 12. Dente FL, Carnevali S, Bartoli ML, Cianchetti S, Bacci E, Di Franco A, et al. Profiles of proinflammatory cytokines in sputum from different groups of severe asthmatic patients. Ann Allergy Asthma Immunol 2006;97:312-20. 13. Pizzichini E, Pizzichini MMM, Efthimiadis A, Evans S, Morris MM, Squillace D, et al. Indices of airway inflammation in induced sputum: reproducibility and validity of cell and fluid-phase measurements. Am J Respir Crit Care Med 1996;154:308-17. 14. Simpson JL, Scott RJ, Boyle MJ, Gibson PG. Differential proteolytic enzyme activity in eosinophilic and neutrophilic asthma. Am J Respir Crit Care Med 2005; 172:559-65. 15. Moore WC, Bleecker ER, Curran-Everett D, Erzurum SC, Ameredes BT, Bacharier L, et al. Characterization of the severe asthma phenotype by the NHLBI Severe Asthma Research Program. J Allergy Clin Immunol 2007;119:405-13. 16. Proceedings of the ATS Workshop on Refractory Asthma: current understanding, recommendations, and unanswered questions. Am J Respir Crit Care Med 2000; 162:2341-51. 17. Fahy JV, Boushey HA, Lazarus SC, Mauger EA, Cherniack RM, Chinchilli VM, et al. Safety and reproducibility of sputum induction in asthmatic subjects in a multicenter study. Am J Respir Crit Care Med 2001;163:1470-5. 18. Tusher V, Tibshirani R, Chu C. Significance analysis of microarrays applied to ionizing radiation response. Proc Nat Acad Sci U S A 2001;98:5116-21. 19. Berry MA, Shaw DE, Green RH, Brightling CE, Wardlaw AJ, Pavord ID. The use of exhaled nitric oxide concentration to identify eosinophilic airway inflammation: an observational study in adults with asthma. Clin Exp Allergy 2005;35:1175-9.

HASTIE ET AL 1035

20. Fahy JV, Boushey HA. Effect of low-dose beclomethasone dipropionate on asthma control and airway inflammation. Eur Respir J 1998;11:1240-7. 21. Woodruff PG, Khashayar R, Lazarus SC, Janson S, Avila P, Boushey HA, et al. Relationship between airway inflammation, hyperresponsiveness, and obstruction in asthma. J Allergy Clin Immunol 2001;108:753-8. 22. Thomas RA, Green RH, Brightling CE, Birring SS, Parker D, Wardlaw AJ, et al. The influence of age on induced sputum differential cell counts in normal subjects. Chest 2004;126:1811-4. 23. Moore WC, Meyers DA, Wenzel SE, Teague WG, Li H, Li X, et al. Identification of asthma phenotypes using cluster analysis in the severe asthma research program. Am J Respir Crit Care Med 2010;181:315-23. 24. Haldar P, Pavord ID, Shaw DE, Berry MA, Thomas M, Brightling CE, et al. Cluster analysis and clinical asthma phenotypes. Am J Respir Crit Care Med 2008;178: 218-24. 25. Haldar P, Pavord ID. Noneosinophilic asthma: a distinct clinical and pathologic phenotype. J Allergy Clin Immunol 2007;119:1043-52. 26. Gibson PG, Simpson JL, Saltos N. Heterogeneity of airway inflammation in persistent asthma: Evidence of neutrophilic inflammation and increased sputum interleukin-8. Chest 2001;119:1329-36. 27. Simpson JL, Timmins NL, Fakes K, Talbot PI, Gibson PG. Effect of saliva contamination on induced sputum cell counts, IL-8 and eosinophil cationic protein levels. Eur Respir J 2004;23:759-62. 28. Kim H-B, Kim C-K, Iijima K, Kobayashi T, Kita H. Protein microarray analysis in patients with asthma: elevation of the chemokine PARC/CCL18 in sputum. Chest 2009;135:295-302. 29. Kanazawa H, Yoshikawa J. Effect of beclomethasone dipropionate on basic fibroblast growth factor levels in induced sputum samples from asthmatic patients. Ann Allergy Asthma Immunol 2005;95:546-50. 30. Asai K, Kanazawa H, Otani K, Shiraishi S, Hirata K, Yoshikawa J. Imbalance between vascular endothelial growth factor and endostatin levels in induced sputum from asthmatic subjects. J Allergy Clin Immunol 2002;110:571-5. 31. Vignola AM, Riccobono L, Mirabella A, Profita M, Chanez P, Bellia V, et al. Sputum metalloproteinase-9/tissue inhibitor of metalloproteinase-1 ratio correlates with airflow obstruction in asthma and chronic bronchitis. Am J Respir Crit Care Med 1998;158:1945-50. 32. Park S-W, Jangm HK, An MH, Min JW, Jang A-S, Lee J-H, et al. Interleukin-13 and interleukin-5 in induced sputum of eosinophilic bronchitis: comparison with asthma. Chest 2005;128:1921-7. 33. Weckmann M, Collison A, Simpson JL, Kopp MV, Wark PAB, Smyth MJ, et al. Critical link between TRAIL and CCL20 for the activation of TH2 cells and the expression of allergic airway disease. Nat Med 2007;13:1308-15. 34. Braun A, Lommatzsch M, Mannsfeldt A, Neuhaus-Steinmetz U, Fischer A, Schnoy N, et al. Cellular sources of enhanced brain-derived neurotrophic factor production in a mouse model of allergic inflammation. Am J Respir Cell Mol Biol 1999;21: 537-46. 35. Hahn C, Islamian AP, Renz H, Nockher WA. Airway epithelial cells produce neurotrophins and promote the survival of eosinophils during allergic airway inflammation. J Allergy Clin Immunol 2006;117:787-94. 36. Arend WP, Palmer G, Gabay C. IL-1, IL-18, and IL-33 families of cytokines. Immunol Rev 2008;223:20-38. 37. Hastie AT, Everts KB, Cho S-K, Zangrilli J, Shaver JR, Pollice MB, et al. IL-1b release from cultured bronchial epithelial cells and bronchoalveolar lavage cells from allergic and normal humans following segmental challenge with ragweed. Cytokine 1996;8:730-8. 38. Sabnis AS, Reilly CA, Veranth JM, Yost GS. Increased transcription of cytokine genes in human lung epithelial cells through activation of a TRPM8 variant by cold temperatures. Am J Physiol Lung Cell Mol Physiol 2008;295:L194-200. 39. Reibman J, Hsu Y, Chen LC, Bleck B, Gordon T. Airway epithelial cells release MIP3a/CCL20 in response to cytokines and ambient particulate matter. Am J Respir Cell Mol Biol 2003;28:648-54. 40. Pichavant M, Charbonnier A-S, Taront S, Brichet A, Wallaert B, Pestel J, et al. Asthmatic bronchial epithelium activated by the proteolytic allergen Der p 1 increases selective dendritic cell recruitment. J Allergy Clin Immunol 2005;115:771-8. 41. Nockher WA, Renz H. Neurotrophins and asthma: novel insight into neuroimmune interaction. J Allergy Clin Immunol 2006;117:67-71. 42. Virchow JC, Julius P, Lommatzsch M, Luttmann W, Renz H, Braun A. Neurotrophins are increased in bronchoalveolar lavage fluid after segmental allergen provocation. Am J Respir Crit Care Med 1998;158:2002-5. 43. Fox AJ, Patel HJ, Barnes PJ, Belvisi MG. Release of nerve growth factor by human pulmonary epithelial cells: role in airway inflammatory diseases. Eur J Pharmacol 2001;424:159-62. 44. Schulte-Herbruggen O, Nassenstein C, Lommatzsch M, Quarcoo D, Renz H, Braun A. Tumor necrosis factor-a and interleukin-6 regulate secretion of brain-derived neurotrophic factor in human monocytes. J Neuroimmunol 2005;160:204-9.

1036 HASTIE ET AL

45. Fulkerson PC, Zimmermann N, Hassman LM, Finkelman FD, Rothenberg ME. Pulmonary chemokine expression is coordinately regulated by STAT1, STAT6, and IFN-g. J Immunol 2004;173:7565-74. 46. Carlsen HS, Baekkevold ES, Morton HC, Haraldsen G, Brandtzaeg P. Monocyte-like and mature macrophages produce CXCL13 (B cell-attracting chemokine 1) in inflammatory lesions with lymphoid neogenesis. Blood 2004;104: 3021-7. 47. Goya S, Matsuoka H, Mori M, Morishita H, Kida H, Kobashi Y, et al. Sustained interleukin-6 signalling leads to the development of lymphoid organ-like structures in the lung. J Pathol 2003;200:82-7. 48. Hamilton LM, Torres-Lozano C, Puddicombe SM, Richter A, Kimber I, Dearman RJ, et al. The role of the epidermal growth factor receptor in sustaining neutrophil inflammation in severe asthma. Clin Exp Allergy 2003;33:233-40. 49. Subauste MC, Proud D. Effects of tumor necrosis factor-alpha, epidermal growth factor and transforming growth factor-alpha on interleukin-8 production by, and

J ALLERGY CLIN IMMUNOL MAY 2010

50.

51.

52.

53.

human rhinovirus replication in, bronchial epithelial cells. Internat Immunopharcacol 2001;1:1229-34. Polosa R, Puddicombe SM, Krishna MT, Tuck AB, Howarth PH, Holgate ST, et al. Expression of c-erbB receptors and ligands in the bronchial epithelium of asthmatic subjects. J Allergy Clin Immunol 2002;109:75-81. Amishima M, Munakata M, Nasuhara Y, Sato A, Kakahashi T, Homma Y, et al. Expression of epidermal growth factor and epidermal growth factor receptor immunoreactivity in the asthmatic human airway. Am J Respir Crit Care Med 1998;157:1907-12. Gershman NH, Wong HH, Liu JT, Mahlmeister MJ, Fahy JV. Comparison of two methods of collecting induced sputum in asthmatic subjects. Eur Respir J 1996;9: 2448-53. Kikuchi I, Kikuchi S, Kobayashi T, Hagiwara K, Sakamoto Y, Kanazawa M, et al. Eosinophil trans-basement membrane migration induced by interleukin-8 and neutrophils. Am J Respir Cell Mol Biol 2006;34:760-5.

HASTIE ET AL 1036.e1

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

METHODS

Sputum processing

Standard operating procedures were established, reviewed by an independent data safety monitoring board, and approved by Institutional Review Boards at Wake Forest University before enrollment of subjects in SARP. After written informed consent, phenotypic characterization of subjects included an administered clinical questionnaire, medical history, standard lung function tests with measures of bronchodilator reversibility and airway hyperresponsiveness, assessment of atopy by skin prick tests to common allergens, and collection of blood, exhaled nitric oxide, exhaled breath condensate, and induced sputum. Detailed medication use was obtained; no medication was excluded, and none was withheld before sputum induction. More detailed information regarding the methods and results for phenotypic characterization of severe and nonsevere asthma has been reported.E1

The method of sputum processing was adopted from the Asthma Clinical Research Network.E6 Briefly, the sputum sample was weighed and an equal volume added of 0.1% dithiothreitol in sterile PBS solution. After incubation with intermittent mixing at 37 8C in a shaking water bath to disperse mucus, the total cell count per milliliter (equal to cell count per gram sputum) and the percent viable cells were determined. Cytospin slides were stained for differential count on at least 500 white blood cells (WBC: macrophages 1 monocytes [Mac], lymphocytes, neutrophils, and eosinophils), and bronchial and squamous epithelial cells. Cell pellets (1 3 106 cells/pellet) and supernatants were prepared by centrifugation and stored (–80 8C) for further analyses. Total cell counts were reported as number 3 106/g. The %WBC and percent bronchial and percent squamous epithelial cells were calculated for total cell counts, and percent macrophages, percent lymphocytes, percent neutrophils, and percent eosinophils for WBC count. Cell number per gram for WBCs and bronchial epithelial cells were calculated from total cell number per gram; number per gram for macrophages, lymphocytes, neutrophils, and eosinophils were derived from WBC number/g (1 g was considered equivalent to 1 mL). Any sputum cell counts with >80% squamous cells have previously been considered inadequate.E6-E8 Our data confirmed an increased frequency of WBC differential counts that yielded ‘‘0’’ eosinophil counts across quintiles of percent squamous counts (0% to 20%, 21% to 40%, 41% to 60%, 61% to 80%, and 81% to 100% squamous cells). x2 Analysis of percent of differential counts yielding ‘‘0’’ eosinophil counts found no significant difference between the 41% to 60% and 61% to 80% squamous groups, whereas inclusion of the 81% to 100% squamous group contributed to a significant difference (P 5 .016). Thus, differential counts for samples with >80% squamous cells affect the determination of ‘‘0’’ eosinophils compared with differential counts with <80% squamous cells, and therefore, these samples were excluded from further analyses (N 5 67). Although the contamination of sputum with saliva may be variable, the induction method uses a 2-specimen collection system for separation of saliva and sputum.E7 The mean contamination by squamous epithelial cells was equal to or less than 40% across all groups and did not differ between groups for any stratification either by granulocytes (not shown) or by asthma severity (Table E2).

Subjects Wake Forest University SARP subjects (N 5 242; Table E1) who met American Thoracic Society criteria for diagnosis of asthma and were currently nonsmoking (less than 5 pack-years past use) underwent sputum induction, or if safety criteria were not met (postbronchodilator FEV1 % predicted >60% for induction), collection of a spontaneous sputum sample. Subjects were divided into groups based on 2 stratifications: (1) sputum percentage of eosinophils and percentage of neutrophils, or (2) disease severity by lung function criteriaE1 and current corticosteroid dosage. For stratification of subjects by sputum percentage of eosinophils and percentage of neutrophils, other groups have chosen various levels for eosinophils (1% to 3%) and neutrophils (61%).E2-E5 We considered the median levels for eosinophils and neutrophils in our severe group, 1.7% and 39.9%, respectively, to split those with severe asthma into approximately equal groups. For comparison with earlier reports, a slightly higher 2% was chosen for eosinophils. An earlier reportE5 split neutrophils at 61% neutrophils, which placed 75% of our severe group into the low neutrophil categories and _2% eosinophils and > _40% neutrophils was considered too restrictive. Thus, > were chosen for stratification of subjects according to sputum granulocyte percentages. Despite the somewhat different thresholds, many of our groups’ characteristics are similar to previously described subgroup characteristics,E5 although with larger numbers, differences became significant in our analyses. The demographics and clinical characteristics of subjects in these groups are presented in Table I. Nine (of 175 subjects with acceptable sputum) were between the ages of 12 and 18 years; 2 of these (1 severe and 1 nonsevere treated with corticosteroids) were analyzed by microarray. Thus, the majority of sputum samples analyzed for cell counts and ELISAs (95%) and microarrays (94%) were from adults ranging in age from 18 to 65 years.

Sputum induction The method of sputum induction was adopted from the Asthma Clinical Research Network method described in detail.E6 Briefly, subjects performed baseline spirometry (the best of 3 maneuvers), inhalation of 360 mg short-acting bronchodilator (albuterol) and repeat spirometry 15 minutes after bronchodilator. If the FEV1 percent predicted was greater than 60% (safety criterion), sputum induction was performed with inhalation of 3% buffered saline solution for 12 minutes; if not, a spontaneous sample was collected (N 5 17; 7%). Lung function was checked at 4 and 8 minutes during induction, and 0, 15, and 30 minutes postinduction. Saliva was separately collected, before independent expectoration of sputum.E7 Induction was stopped if FEV1 fell below 80% of postbronchodilator preinduction FEV1 percent predicted, and the subject was treated with bronchodilators until the FEV1 returned to within 10% of postalbuterol FEV1, at which time the subject was discharged. Safety protocol steps were followed for subjects experiencing more severe bronchoconstriction. Only 3 subjects, all nonsevere subjects with moderate asthma, received additional albuterol (4 puffs or less by inhaler) during 1 hour postinduction, returned to within 10% of postalbuterol FEV1, and did not require further treatment. Sputum was transferred to the laboratory for immediate processing.

Inflammation microarrays and analysis Total protein in sputum supernatants was determined by micro BCA assay (Thermo Scientific, Rockford, Ill) according to the manufacturer’s instructions before incubation with microarrays. Median sputum supernatant total protein levels ranged from 1.9 to 2.3 mg/mL across the asthma severity groups (P 5 .3) and 1.9 to 2.8 mg/mL across subjects grouped by percent eosinophils and percent neutrophils (P 5 .13). Inflammation protein microarrays (Human Cytokine Antibody Array C Series 1000; RayBiotech, Norcross, Ga) were performed with sputum supernatant diluted to 1 mg total protein in blocking buffer according to the manufacturer’s recommendations with these modifications: 4-hour incubation with sputum supernatant sample, overnight incubation with detection antibody mix at 4 8C, and 4-hour incubation with streptavidin complex. Washes between steps followed the manufacturer’s protocol. Enhanced chemiluminescence was captured and analyzed with a Kodak 2000MM or 4000MM Image Analysis Station (Carestream Molecular Imaging, Woodbridge, Conn). Densities of duplicate reactions on each membrane were background-subtracted, averaged for the 2 target sites/membrane, and normalized to positive and negative controls on each membrane. The sputum samples examined in this screening were selected from subjects with asthma having high eosinophil percentages, high neutrophil percentages, both high or both low granulocyte percentages, patients with severe asthma, patients with nonsevere asthma with corticosteroid therapy, and patients with nonsevere asthma without corticosteroid therapy. Thus, the sputa screened by microarray spanned the observed range of sputum cellular patterns, asthma severity, and corticosteroid use. Assessment of proteins is limited by the volume of sputum obtained. In addition, protein identification can be disrupted by dithiothreitol useE9,E10 and

1036.e2 HASTIE ET AL

proteolysis.E11 Paired comparison of microarray results from dialyzed sputum supernatant and from autologous sputum supernatant without dialysis was made on 4 subjects’ sputum samples. Sputum supernatants with sufficient volume were split into 2 aliquots, and 1 was dialyzed against 3 changes (1/100 ratio of sample/buffer solution each) of sterile saline using Slide-A-Lyzer cassettes (3.5-kd molecular weight cutoff, 0.5 to 3-mL volume, Thermo Scientific, Rockford, Ill). After dialysis, the protein concentration was determined on both dialyzed and undialyzed aliquots of each sample supernatant. After adjustment by dilution to 1 mg/mL total protein with buffer, each aliquot was separately analyzed according to the microarray instructions with our modifications as before. An additional 5 subjects’ sputum supernatants of sufficient volume were split into 2 aliquots, 1 untreated and 1 treated with protease inhibitor cocktail (1/100 dilution of reconstituted cOmplete Protease Inhibitor Cocktail tablets; Roche Applied Science, Indianapolis, Ind). Both protease inhibitor–treated and untreated aliquots (adjusted to 1 mg/mL total protein) were separately analyzed on microarrays as described for comparison of differences attributable to protease inhibition. Likewise, 6 subjects sputum supernatant samples were divided into 2 aliquots, 1 assessed without further addition and the other spiked with 4 recombinant human proteins selected from the protein targets on the array (IL-1b, IL-4, IL-13, and hepatocyte growth factor [HGF]) to validate microarray recognition of known proteins. All steps were identical to the previously described procedure.

ELISAs Quantitation of sputum proteins, identified from comparisons of high versus low percent eosinophils and percent neutrophils, or of severe versus nonsevere asthma, were obtained by specific ELISA according to the manufacturers’ recommendations (R & D Systems for BDNF, BLC/ CXCL13, bone morphogenic protein [BMP]–4, EGF, IL-1b, IL-8, IL-13, IFN-g, LIGHT/TNFSF14, MIP-3a/CCL20, and TNF-a; RayBiotech for PARC/CCL18 and Eotaxin 2/CCL24). All standard proteins were tested for recovery of a known amount after spiking into 4 or more separate normal subjects’ sputum supernatant samples containing dithiothreitol. The amount of standard added was selected to be in the lower 1/3 range of the standards, and aliquots of each sputum sample were tested without and with spiked standards. Sputum supernatants of healthy subjects were used to minimize elevated background levels potentially present in asthmatic sputum. The average recovery yields were as follows: BDNF, 103%; BMP-4, 100%; EGF, 102%; CXCL13, 102%; IFN-g, 98%; IL-1b, 110%; IL-8, 100%; IL-13, 146%; LIGHT, 143%; MIP-3a, 97%, and TNF-a, 94%. Separately, aliquots of the standards were assayed without or with dithiothreitol added at a concentration equivalent to the level in the sputum supernatants.

Western blots of sputum cell lysates Sputum cell pellets (1 3 106 cells) were resuspended in 200 mL lysis buffer (0.5% sodium deoxycholate, 1% sodium dodecyl sulfate, 1% nonylphenoxylpolyethoxylethanol [NP40], 0.1 mg/mL phenylmethanesulfonyl fluoride [PMSF], 0.05 mg/mL aprotinin, and 1 mmol/L sodium orthovanadate). Proteins separated on 6% to 15% polyacrylamide gradient gels were transferred to supported nitrocellulose membranes. Membranes were blocked in nonfat milk and probed with mouse monoclonal primary antibodies specific for human BDNF and IL-1b (R&D Systems Inc, Minneapolis, Minn), biotinconjugated antimouse secondary antibody, horseradish peroxidase–conjugated streptavidin, and PicoWest substrate (Thermo Scientific, Rockford, Ill) with appropriate washes between each step. The enhanced chemiluminescent signal was captured on x-ray film and analyzed by a Kodak 2000MM or 4000MM Image Analysis Station (Carestream Molecular Imaging, Woodbridge, Conn) for band densities.

Statistics Demographic and ELISA data are presented as means 6 SEMs for parametric tests or as medians (25% to 75% quartiles) for nonparametric tests as indicated. Measures not meeting normal distribution for the Kolmogorov-

J ALLERGY CLIN IMMUNOL MAY 2010

Smirnov test were transformed, generally as log or square root, for parametric tests; otherwise, nonparametric tests were run. A 0 in cell differentials was replaced with a value ½ of the lowest observed; for example, 0% eosinophils became 0.05% for log transformation.E12 Continuous variables were analyzed by using ANOVA or Kruskal-Wallis for 3 or more groups (SAS version 9.2, SAS Institute Inc, Cary, NC, or Sigmastat version 2.03, Systat Software, San Jose, Calif). The Student t test or Mann-Whitney tests were used for analysis of 2 groups. x2 Tests were used to analyze categorical variables. Multiple linear regression and general linear models examined associations between measurements of sputum cell counts and inflammatory mediators or lung function measurements. Bonferroni correction was applied to variables with a P value < .05 to determine significance. Microarray density data were log2-transformed and analyzed by SAM.E13 A score of >1, a false-positive discovery rate of less than 5%, and a greater than 2-fold change between groups was our criterion for accepting differences in protein levels as significant. Although the inflammatory protein microarray examines 120 proteins, the SAM program is specifically designed to address the problem of chance identification in even larger microarray datasets.E13 The greater than 2-fold change is a more rigorous criterion and reduces the number of proteins or genes that have been accepted in other studies,E14 but was deemed necessary because of limited sputum supernatant available for ELISA testing. Sample size for microarrays was calculated from previous work on cytokine ELISA assessments of bronchoalveolar lavage supernatants in patients with asthma.E15 SD may range from ½ to 3/4 of the mean. The ½ value with a power of 0.8 and an a of 0.05 calculates a sample size of 6; the 3/4 value with the same power and a values calculates a sample size of 10. Thus, subject numbers of 6 to 12 for microarray analyses were appropriate.

RESULTS Subject and sputum cell characteristics stratified by severity of asthma The physiological and clinical characteristics of all enrolled subjects with asthma are presented in Table E1 with groups stratified by lung function and current corticosteroid use: nonsevere, no ICSs, and FEV1 >80% predicted (N 5 59); nonsevere, on low-dose or medium-dose ICSs, and FEV1 >80% predicted (N 5 50); nonsevere, no ICSs, and FEV1 <80% predicted (N 5 40); nonsevere, low-dose or medium-dose ICSs, and FEV1 <80% predicted (N 5 45); and severe, high-dose ICS or near continuous oral steroid therapy, plus at least 2 secondary criteriaE1 (N 5 48). Reports by othersE8,E16 that ICS use can modulate cellular inflammation in asthma support the initial separation of corticosteroid-treated and untreated nonsevere groups within mild asthma (FEV1 % predicted > 80%) and within moderate asthma (FEV1 % predicted < 80%). However, no significant differences were found between subjects treated with steroids and those untreated at similar levels of asthma severity, either FEV1 % predicted <80% or FEV1 % predicted >80%. Physiological and clinical characteristics of subjects in nonsevere groups with FEV1 % predicted >80% (combined with or without ICS), nonsevere groups with FEV1 % predicted <80% (combined with or without ICS), and the severe group were similar to mild, moderate, and severe asthma groups, respectively, in the larger cohort of SARP.E1 Sputum total cell counts, WBC counts, and cell differentials are presented according to asthma severity in Table E2. Univariate analyses across these found no significant differences in cell percentages, and only eosinophil count per milliliter was significant after correction for multiple testing, but this was a result of the highest eosinophil count in the nonsevere group with FEV1 % predicted <80%, treated with ICS. Highest numbers in most of the WBC types occurred in nonsevere, FEV1 % predicted <80%,

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

treated with ICS, or for percent lymphocytes and percent neutrophils in the severe group, which did not differ from nonsevere, FEV1 % predicted <80%, treated with ICS. We examined sputum cell constituents for differences caused by corticosteroid use in subject groups with similar lung function values (FEV1 % predicted >80% or FEV1 % predicted <80%; Table E2) but found no differences secondary to corticosteroid therapy.

Data from stratification of microarray data based on asthma severity phenotypes Comparisons of microarray data for subjects with nonsevere asthma without and with ICS treatment (N 5 12 and N 5 12, respectively) with subjects with severe asthma (N 5 12) were (1) all nonsevere with subjects with severe asthma, (2) nonsevere subjects receiving corticosteroids with severe subjects, and (3) subjects with nonsevere asthma receiving corticosteroids with only mild lung function decrements (FEV1 % predicted > 80%) with severe subjects. These 3 comparisons yielded 9, 6, and 33 proteins, respectively, of significant, >2fold increases in sputa of severe subjects. Three proteins (underlined in Table E3) matched in all 3 comparisons: BDNF, BLC/ CXCL13, and EGF. The greatest number of significantly increased proteins (N 5 33) occurred between the sputum of the severe group compared with the sputum of the nonsevere group, FEV1 % predicted >80% and treated with ICSs. Eight proteins, also observed in the 2 comparisons with all nonsevere subjects, or only those nonsevere subjects on corticosteroid treatment, are listed in Table E3. The remaining 25 proteins are listed in Table E4. Those marked with asterisks were also found significantly increased in the comparison of severe to all nonsevere subjects (comparison 1) but did not reach the criterion of >2-fold. ELISAs on selected inflammatory mediators stratified by asthma severity The highest levels of the mediators assessed by ELISA and stratified by asthma severity were generally observed in the subjects with severe asthma, but no mediator reached significance in comparison with nonsevere groups (Table E5). Multiple linear regression analyses of sputum cell counts with measures of lung function Lung function measurements, FEV1 % predicted, FVC % predicted, and ratio of FEV1/FVC were examined for association with sputum cell constituents by multiple linear regression. As previously reported,E16 FEV1 % predicted was negatively associated with sputum eosinophil count per milliliter, but in addition, the model found a negative association with bronchial epithelial cell and a positive association with macrophage count per milliliter (overall R 5 0.35 and P < .001; individual coefficients and P values: eosinophils, –3.44 and .004; bronchial epithelial cells, –4.36 and .03; macrophages, 8.31 and .045, respectively). There was no significant association of any sputum cell constituent in a model with FVC % predicted (P 5 .38). As expected from these separate results, the baseline ratio of FEV1/FVC was significantly associated with the same sputum cell types (overall R 5 0.41 and P < .001; individual coefficients and P values: eosinophils, –0.025 and <.001; bronchial epithelial cells, –0.035 and .003; macrophages, 0.067 and .006, respectively). These novel, observed associations of

HASTIE ET AL 1036.e3

bronchial epithelial cells with FEV1 % predicted and with FEV1/FVC are shown in Fig E1 and may reflect worsening lung function relative to ongoing damage and loss of airway epithelium.

Correlation of sputum mediators with lung function measures, FeNO, and atopy Lung function measurements, FEV1 % predicted, FVC % predicted, ratio of FEV1/FVC, FeNO and serum IgE (both as log values), and number of positive skin tests were examined for association with mediator ELISA results (Table E6). IgE and number of positive skin tests, as measures of atopy, did not show significant associations with any mediator assessed. Baseline FEV1 % predicted was negatively associated with LIGHT/ TNFSF14, and baseline FVC % predicted was negatively associated with IL-8 and PARC/CCL18. It is not surprising that LIGHT/ TNFSF14 and PARC/CCL18 show association with baseline FEV1 % predicted and baseline FVC % predicted, respectively; these 2 mediators were the only ones to have any association with age, which is definitely significantly associated with these lung function measures. Only TNF-a was positively associated with FeNO. Except for the positive association of TNF-a with increasing FeNO, all other mediators showed negative correlations with clinical characteristics. Sputum cell counts per gram of subjects with asthma stratified by sputum granulocyte percentages Total cell count is determined independently from sputum cell differential percentages obtained from cytospin slide preparations. Count per gram for specific sputum cell types was subsequently calculated directly from total cell count per gram for bronchial epithelial cells and from WBC count per gram for macrophages, lymphocytes, neutrophils, and eosinophils. We examined whether specific sputum cell count per gram differed for subjects stratified by percent eosinophils and percent neutrophils. The total cell and WBC count per gram were significantly _ .001 for each) in the group with both > _2% eosinoincreased (P < _40% neutrophils (Table E7). The > _2% eosinophils 1 phils 1 > > _40% neutrophils group also had the highest eosinophils and neutrophils count per gram (both < .001). These results are consistent, although not identical, to the recent report by Simpson et alE5 _1% eosinophils stratifying 93 subjects on the basis of <1% or > _61% neutrophils. 1 <61% or > Paired microarray analyses for autologous dialyzed and undialyzed sputum supernatants Dialysis and protease inhibition of sputum supernatants containing dithiothreitol reportedly optimizes the detection of chemokines and cytokines.E17 On average, dialysis increased and decreased approximately equal numbers of proteins (N 5 15-16) in our microarray results. However, the same proteins were not consistently affected in each of the subjects. Only 3 proteins showed consistent alteration in all subjects analyzed: IL-4 and neutrophil activating peptide-2 (NAP2/CXCL7) were increased >50%, but <100% by dialysis, and GCSF was decreased >100% by dialysis. Thus, removal of dithiothreitol by dialysis altered detection of very few proteins by our criteria required for significance: at least a 2-fold change, scored >1, and a false-positive rate <5%.

1036.e4 HASTIE ET AL

Paired analyses of sputum supernatants treated with protease inhibitor compared with untreated showed decreased levels of mediators in the supernatant treated with a cocktail of protease inhibitors. For those proteins, only 10 of 120 of which showed an increase, the increase averaged only 14% in the protease-treated supernatant. Addition of protease inhibitors did not appear to improve detection of proteins substantially for the microarrays and was not further used. Paired analyses of sputum supernatants were also examined for detection of spiked known proteins: IL-1b, IL-4, IL-13, and HGF. All 4 were successfully identified through SAM analysis as significant and >2-fold increased in the spiked samples compared with the untreated aliquot.

Western blots of sputum cell lysates probed for mediator presence Sputum cell lysates were analyzed by Western blots of SDS– polyacrylamide gradient gels for BDNF and IL-1b. Mean band densities for these proteins in subjects with <2% eosinophils 1 _40% neutro<40% neutrophils (Neither), <2% eosinophils 1 > _2% eosinophils 1 <40% neutrophils (Eos), and phils (Neu), > > _2% eosinophils 1 > _40% neutrophils (Eos1Neu) are shown in Fig E2. Both BDNF and IL-1b in cell lysates were at higher levels in the group with elevated Eos1Neu than in the other groups but did not reach significance.

REFERENCES E1. Moore WC, Bleecker ER, Curran-Everett D, Erzurum SC, Ameredes BT, Bacharier L, et al. Characterization of the severe asthma phenotype by the NHLBI Severe Asthma Research Program. J Allergy Clin Immunol 2007;119: 405-13. E2. Gibson PG, Simpson JL, Saltos N. Heterogeneity of airway inflammation in persistent asthma: evidence of neutrophilic inflammation and increased sputum interleukin-8. Chest 2001;119:1329-36. E3. Green RH, Brightling CE, Woltmann G, Parker D, Wardlaw AJ, Pavord ID. Analysis of induced sputum in adults with asthma: identification of subgroup with isolated sputum neutrophilia and poor response to inhaled corticosteroids. Thorax 2002;57:875-9.

J ALLERGY CLIN IMMUNOL MAY 2010

E4. Lemiere C, Ernst P, Olivenstein R, Yamauchi Y, Govindaraju K, Ludwig MS, et al. Airway inflammation assessed by invasive and noninvasive means in severe asthma: eosinophilic and noneosinophilic phenotypes. J Allergy Clin Immunol 2006;118:1033-9. E5. Simpson JL, Scott R, Boyle MJ, Gibson PG. Inflammatory subtypes in asthma: assessment and identification using induced sputum. Respirology 2006;11: 54-61. E6. Fahy JV, Boushey HA, Lazarus SC, Mauger EA, Cherniack RM, Chinchilli VM, et al. Safety and reproducibility of sputum induction in asthmatic subjects in a multicenter study. Am J Respir Crit Care Med 2001;163:1470-5. E7. Gershman NH, Wong HH, Liu JT, Mahlmeister MJ, Fahy JV. Comparison of two methods of collecting induced sputum in asthmatic subjects. Eur Respir J 1996;9: 2448-53. E8. Woodruff PG, Khashayar R, Lazarus SC, Janson S, Avila P, Boushey HA, et al. Relationship between airway inflammation, hyperresponsiveness, and obstruction in asthma. J Allergy Clin Immunol 2001;108:753-8. E9. Amishima M, Munakata M, Nasuhara Y, Sato A, Kakahashi T, Homma Y, et al. Expression of epidermal growth factor and epidermal growth factor receptor immunoreactivity in the asthmatic human airway. Am J Respir Crit Care Med 1998; 157:1907-12. E10. Hadjicharalambous C, Dent G, May RD, Handy RL, Anderson IK, Davies DE, et al. Measurement of eotaxin (CCL11) in induced sputum supernatants: validation and detection in asthma. J Allergy Clin Immunol 2004;113:657-62. E11. Berry MA, Parker D, Neale N, Woodman L, Morgan A, Monk P, et al. Sputum and bronchial submucosal IL-13 expression in asthma and eosinophilic bronchitis. J Allergy Clin Immunol 2004;114:1106-9. E12. Berry MA, Shaw DE, Green RH, Brightling CE, Wardlaw AJ, Pavord ID. The use of exhaled nitric oxide concentration to identify eosinophilic airway inflammation: an observational study in adults with asthma. Clin Exp Allergy 2005;35: 1175-9. E13. Tusher V, Tibshirani R, Chu C. Significance analysis of microarrays applied to ionizing radiation response. Proc Nat Acad Sci U S A 2001;98:5116-21. E14. Lewis CC, Yang JYH, Huang X, Banerjee SK, Blackburn MR, Baluk P, et al. Disease-specific gene expression profiling in multiple models of lung disease. Am J Respir Crit Care Med 2008;177:376-87. E15. Batra V, Musani AI, Hastie AT, Khurana S, Carpenter KA, Zangrilli JG, et al. Bronchoalveolar lavage fluid concentration of transforming growth factor (TGF)-b1, TGF-b2, interleukin (IL)-4 and IL-13 after segmental allergen challenge and their effects on a-smooth muscle actin and collagen III synthesis by primary human lung fibroblasts. Clin Exp Allergy 2004;34:437-44. E16. Bartoli ML, Bacci E, Carnevali S, Cianchetti S, Dente FL, DiFranco A, et al. Clinical assessment of asthma severity partially corresponds to sputum eosinophilic airway inflammation. Respir Med 2004;98:184-93. E17. Erin EM, Jenkins GR, Kon OM, Zacharasiewicz AS, Nicholson GC, Neighbour H, et al. Optimized dialysis and protease inhibition of sputum dithiothreitol supernatants. Am J Respir Crit Care Med 2008;177:132-41.

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

HASTIE ET AL 1036.e5

FIG E1. Sputum epithelial cell count per milliliter plotted against FEV1 % predicted (bottom) and FEV1/FVC (top). Open diamond, Nonsevere, no ICSs, FEV1 >80% predicted; closed diamond, nonsevere, ICS-treated, FEV1 >80% predicted; open triangle, nonsevere, no ICSs, FEV1 <80% predicted; closed triangle, nonsevere, ICS-treated, FEV1 <80% predicted; and closed square, severe.E1 Bronchial epithelial cells negatively associated with FEV1 % predicted (P 5 .03) and ratio FEV1/FVC (P 5 .003), suggesting epithelial damage relates to worsening lung function.

1036.e6 HASTIE ET AL

FIG E2. Average Western blot band density for BDNF (top) and IL-1b (bottom) observed in sputum cell lysates for subjects grouped by <2% eosino_40% neutrophils phils 1 <40% neutrophils (Neither), <2% eosinophils 1 > _2% eosinophils 1 <40% neutrophils (Eos), or > _2% eosinophils 1 (Neu), > > _40%neutrophils (Eos1Neu). Higher levels of BDNF and IL-1b occurred in sputum cell lysates from subjects with both granulocytes elevated (Eos1Neu), although they were not significantly different.

J ALLERGY CLIN IMMUNOL MAY 2010

HASTIE ET AL 1036.e7

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

TABLE E1. Demographics and clinical characteristics (means 6 SDs) of Wake Forest University SARP subjects providing sputum samples stratified by lung function and corticosteroid use

N 5 242 Age (y) Age at onset (y) Duration (y) Sex, female* Race, % white/% black/% other* BMI Lung function FEV1 % FVC % FEV1/FVC % Reversibility % (2 puffs) Change FEV1 % (maximum – baseline) Log PC20 (mg/mL) Log FeNO Frequency of positive skin test* No. of positive skin tests Serum Eos % Log total IgE Asthma control and health care use Daily use of b-agonist* Symptoms worse if steroids reduced* >1 urgent health care visit in past year* >3 oral steroid bursts per year* Ever visit ED for breathing problems* Ever spent night in hospital for breathing problems* Ever admitted to ICU for asthma*

Nonsevere FEV1 >80% predicted no ICS

Nonsevere FEV1 >80% predicted ICS

Nonsevere FEV1 <80% predicted no ICS

Nonsevere FEV1 <80% predicted ICS

Severe high-dose ICS or oral

ANOVA p for 5-group analysis

59 33 6 12 16 6 13 17 6 13 81% 59/36/5

50 35 6 14 16 6 15 18 6 13 86% 60/32/8

40 35 6 12 11 6 11 25 6 13 70% 53/45/2

45 39 6 12 19 6 16 20 6 12 67% 58/40/2

48 44 6 12 20 6 18 24 6 16 65% 77/23/0

<.0001 .022 .010 .05 .19

28 6 7

30 6 8

29 6 8

35 6 12

33 6 8

.0004

93 99 .79 7.6 7.4

6 6 6 6 6

9 11 .05 6.3 5.1

92 98 .79 7.4 9.1

6 6 6 6 6

10 12 08 7.1 6.2

68 83 .69 16.7 16

6 6 6 6 6

10 13 .10 15.0 11.7

67 80 .69 13.9 12.7

6 6 6 6 6

11 12 .09 13.9 8.4

64 77 .66 15.7 12.8

6 24 6 21 6 .13 616.2 6 9.7

<.0001 <.0001 <.0001 <.0001 <.0001

.42 6 .78 n 5 56 1.5 6 0.4 90%

.34 6 .9 n 5 47 1.4 6 0.4 78%

–.03 6 .69 n 5 33 1.4 6 0.4 95%

.04 6 .74 n 5 38 1.5 6 0.3 93%

.60 6 .84 n 5 25 1.4 6 0.4 64%

.50 .01

3 (2-6) 3.2 6 2.4 2.0 6 0.6

4 (1-7) 3.2 6 2.6 2.0 6 0.7

4.5(3-7.5) 3.3 6 2.0 2.1 6 0.7

4 (2-6) 4.2 6 2.7 2.1 6 0.6

2 (0-6) 4.0 6 3.1 2.0 6 0.6

.33 .23 .88

25% NA

30% 69%

55% NA

31% 74%

56% 81%

<.001 <.001

15%

24%

32%

32%

63%

<.001

3%

12%

18%

6%

44%

<.001

54%

57%

71%

68%

88%

.026

15%

29%

46%

32%

56%

.007

5%

5%

18%

12%

28%

.025

P values meeting Bonferroni correction for significance in boldface. BMI, Body mass index; ED, emergency department; Eos, eosinophils; ICU, intensive care unit. *x2 Analysis; % is frequency answering ‘‘yes.’’ NA, Not applicable because not on steroid therapy.

.007

1036.e8 HASTIE ET AL

J ALLERGY CLIN IMMUNOL MAY 2010

TABLE E2. Cell characteristics* for sputum samples with <80% squamous cells in subjects stratified by lung function and corticosteroid use Phenotype groups

N 5 175 Total cells 3 106/g WBC 3 106/g %WBC %Bronchial Epithelial % Squamous % Mac % Lym % Neu % Eos Bronchial Epithelial 3 104/mL Mac 3 104/mL Lym 3 104/mL Neu 3 104/mL Eos 3 104/mL

Nonsevere FEV1 >80% predicted No ICS

Nonsevere FEV1 >80% predicted ICS

Nonsevere FEV1 <80% predicted No ICS

Nonsevere FEV1 <80% predicted ICS

Severe High-dose ICS or oral

P value for 5 phenotype groups

39 2.6 6 0.3 1.7 6 0.2 57 6 3.9 2.6 6 0.3 40 6 3.8 63 6 4.0 1.9 6 0.3 33 6 4.0 0.4 (0.2-2.6) 4.0 (2.5-7.2)

42 2.5 6 0.2 1.6 6 0.2 58 6 3.4 3.8 6 0.7 39 6 3.4 60 6 3.3 2.5 6 0.3 34 6 3.3 0.5 (0.1-1.4) 6.4 (2.3-10)

28 4.1 6 1.1 3.2 6 1.1 59 6 5.0 6.6 6 1.2 34 6 4.6 56 6 4.9 2.0 6 0.4 40 6 5.3 1.25 (0.3-4.2) 10 (2.4-24)

34 7.0 6 2.5 5.5 6 2.4 59 6 4.1 6.3 6 1.3 35 6 3.7 54 6 4.5 2.3 6 0.3 38 6 4.3 2.2 (0.7-5.6) 10 (3.9-25)

32 5.4 6 1.4 4.2 6 1.4 62 6 4.4 4.5 6 0.8 33 6 4.3 53 6 4.9 2.6 6 0.5 41 6 4.9 1.7 (0.1-4.1) 9.0 (4.6-20)

.023 .073 .93 .034 .63 .40 .58 .62 .012 .006

90 6 13 1.3 (0.7-5.8) 34 (7-92) 0.6 (0.2-2.2)

87 6 10 2.4 (0.5-6.2) 48 (7-99) 0.5 (0.1-2.9)

84 6 10 2.4 (0.7-4.3) 35 (15-119) 1.9 (0.5-5.8)

2116 114 3.1 (1.5-5.9) 57 (16-152) 3.4 (0.8-14)

127 6 24 5.9 (0.6-9.7) 48 (21-172) 3.1 (0.1-9)

.70 .12 .44 .003

Univariate P values are given, and those meeting Bonferroni correction for significance are presented in boldface. Eos, Eosinophils; Lym, lymphocytes; Neu, neutrophils. *Results from ANOVA are presented as means 6 SEMs; results from Kruskal-Wallis ANOVA are presented as medians (25% to 75%).

HASTIE ET AL 1036.e9

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

TABLE E3. Three comparisons of inflammatory protein microarray data in subjects stratified by severity of asthma 2. Nonsevere groups 1 ICS vs severe group N 5 12 vs N 5 12 False-positive rate 5 0%

1. All nonsevere groups vs severe group N 5 24 vs N 5 12 False-positive rate 5 0% Proteins

Angiogenin BDNF BLC/CXCL13 BMP-4 EGF Eotaxin-2/CCL24 IL-10 MCP-3 MIP-1d MIP-3a/CCL20 PARC/CCL18

Criteria failurey

Increase for severe fold change

Criteria failurey

2.1 2.3 2.6 2.0 2.2 2.4 2.2

b

2.1 2.1

NS NS

Criteria failurey

b 2.1 2.1 NS 3.0

2.0

Increase for severe fold change

2.0 2.6 2.5 2.2 2.1

2.5 2.4

a, b

b

Increase for severe fold change

3. *Nonsevere, FEV1 % predicted >80% 1I CS, vs severe group N 5 6 vs N 5 12 False-positive rate 5 0% to 3%

a, b a, b a 2.6 2.8 4.8

Underlined proteins met criteria for significance in all comparisons. *Remaining proteins increased >2-fold in this comparison only between severe and nonsevere asthma (FEV1 % predicted >80%, treated with corticosteroids) are reported in Table E4.  NS, Not significant (q>5%); a, scored <1; b, fold increase <2.

1036.e10 HASTIE ET AL

J ALLERGY CLIN IMMUNOL MAY 2010

TABLE E4. Proteins increased >2-fold in severe compared with nonsevere asthma, FEV1 % predicted >80% and treated with low _1, false discovery rate <5%), but to moderate ICS doses (scored > not observed for severe asthma compared with all nonsevere, or with nonsevere, FEV1 % predicted <80%, treated with moderate ICS doses Protein

CK b8-1 CNTF Eotaxin 3* FGF-6 Fractalkine GCSF HCC-4 HGF ICAM-1 IL-12 p40 IL-1b IL-2 Ra IL-6*

Fold change

Protein

Fold change

2.5 2.0 2.2 2.3 2.0 2.5 2.3 2.9 2.1 2.4 3.5 2.8 3.0

IL-6R IL-8 MIG* MIP-1b NT-3* Osteoprotegerin sTNF RI TGF-b3 TIMP-2 TRAIL R3 TRAIL R4 uPAR

2.7 2.1 2.8 6.3 2.4 2.1 3.1 2.0 2.4 3.3 2.4 2.6

CK b8-1, Creatine kinase b8-1; CNTF, ciliary neurotrophic factor; FGF-6, fibroblast growth factor-6; GCSF, granulocyte colony stimulating factor; HCC-4, hemofiltrate CC-chemokine-4; ICAM-1, intercellular adhesion molecule-1; IL, interleukins; MIG, monokine induced by IFNg; MIP-1b, macrophage inflammatory protein-1b; NT-3, neurotrophin-3; sTNF RI, soluble tumor necrosis factor receptor I; TGF-b3, transforming growth factor b-3; TIMP-2, tissue inhibitor of metalloproteinase-2; TRAIL R3 and TRAIL R4, tumor necrosis factor-related apoptosis-inducing ligand, TRAIL receptor 3 and 4; uPAR, urokinase-type plasminogen activator receptor. *These proteins were increased in severe compared with all nonsevere asthma but did not reach the >2-fold threshold.

HASTIE ET AL 1036.e11

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

TABLE E5. Sputum supernatant mediator levels stratified by phenotype group

BDNF pg/mL BLC/CXCL13 pg/mL BMP-4 pg/mL EGF pg/mL Eotaxin 2/ CCL24 pg/mL IFN-g pg/mL IL-1b pg/mL IL-8 ng/mL IL-13 pg/mL LIGHT/ TNFSF14 pg/mL MIP-3a/ CCL20 pg/mL PARC/CCL18 pg/mL TNF-a pg/mL

Nonsevere FEV1 >80% predicted No ICS

Nonsevere FEV1 >80% predicted ICS

Nonsevere FEV1 <80% predicted No ICS

Nonsevere FEV1 <80% predicted ICS

Severe High-dose ICS or oral

P value for severity Nonsevere vs severe* t test or Mann Whitney test

11 (6-17) 146 6 45 3.8 6 0.8 170 (52-229) 0.92 (0.01-4.8) 58 (37-131) 97 6 15 1.5 6 0.1 77 6 16 27 (9-52) 441 6 92 5.7 (0.06-21) 0.01 (0.01-1.1)

16 (10-23) 143 6 32 3.4 6 1.1 182 (125.-229) 1.12 (0.01-2.8) 67 (35-130) 139 6 28 1.8 6 0.2 94 6 18 57 (12-95) 487 6 71 7.6 (3.7-14) 0.34 (0.01-2.2)

14 (9-20) 116 6 41 2.7 6 0.7 155 (116-195) 1.76 (0.01-5.4) 126 (45-155) 151 6 37 1.7 6 0.2 74 6 15 52 (42-114) 570 6 113 6.2 (2.2-23) 1.2 (0.01-2.3)

17 (9-31) 116 6 36 1.2 6 0.2 187 (136-213) 2.84 (0.48-14.7) 50 (13-170) 142 6 30 1.8 6 0.2 57 6 11 58 (24-124) 506 6 75 16.2 (3.2-39) 0.63 (0.05-1.7)

19 (12-47) 159 6 49 4.7 6 1.9 213 (137-264) 2.14 (0.01-8.2) 71 (20-140) 245 6 97 2.1 6 0.2 86 6 20 120 (18-199) 597 6 81 22.5 (4.5-47) 1.78 (0.11-3.1)

.07 .14 .94 .06 .42 .87 .50 .08 .77 .07 .19 .06 .03

No mediator reached significant difference between severe and nonsevere groups after Bonferroni correction. *For 2 groups, either t test (means 6 SEMs) or Mann-Whitney (median [quartiles]), (SAS, SAS Institute Inc, Cary, NC).

1036.e12 HASTIE ET AL

J ALLERGY CLIN IMMUNOL MAY 2010

TABLE E6. Correlation of mediators with lung function (baseline FEV1 % predicted, baseline FVC % predicted, baseline FEV1/FVC ratio) and FeNO (log) Clinical characteristic Cytokine or growth factor

Baseline Baseline FEV1 % FVC % predicted* predicted*

BMP-4 EGF

R 5 –0.20 R 5 –0.22 P 5 .048 P 5 .026 R 5 –0.19 R 5 –0.23 P 5 .033 P 5 .009 R 5 –0.20 P 5 .0395

Eotaxin 2 IL-1b IL-8

R 5 –0.28 R 5 –0.24 P 5 .004 P 5 .014 R 5 –0.27 R 5 –0.31 P 5 .005 P 5 .002

IL-13 IFN-g LIGHT/TNFSF14 MIP-3aCCL20 PARC/CCL18 TNF-a

FeNO*

R 5 –0.18 P 5 .032

BDNF BLC/CXCL13

Baseline FEV1/FVC (ratio)*

R 5 0.20 P 5 .05 R 5 –0.32 R 5 –0.29 R 5 –0.28 P 5 .002 P 5 .006 P 5 .008 R 5 –0.23 R 5 –0.22 P 5 .021 P 5 .025 R 5 –0.27 P 5 .002 R 5 0.34 P < .001

*Values under lung function measures or FeNO are R and P value for association with each mediator. A blank indicates the correlation (either Pearson, if parametric, or Spearman, if nonparametric) was not significant. Only boldface R and P values meet significance after Bonferroni correction.

HASTIE ET AL 1036.e13

J ALLERGY CLIN IMMUNOL VOLUME 125, NUMBER 5

TABLE E7. Sputum cell counts per gram stratified by sputum percent eosinophils 1 percent neutrophils

N 5 175 Total cells 3 106/g WBC 3 106/g Bronchial Epithelial 3 104/mL Mac 3 104/mL Lym 3 104/mL Neu 3 104/mL Eos 3 104/mL

<2% Eos <40% Neu

<2% Eos _40% Neu >

_2% Eos > <40% Neu

_2% Eos > _40% Neu >

P value 4-group ANOVA

63 2.5 6 0.2 1.4 6 0.2 11.6 6 2.0

50 5.0 6 0.8 4.1 6 0.8 8.0 6 1.2

42 2.7 6 0.3 1.5 6 0.2 19.3 6 4.9

20 10.5 6 4.4 9.3 6 4.2 23.4 6 11.7

<.001 <.001 .024

101 6 10 1.2 (0.5-5.4) 33 6 5.4 0.3 (0.03-1.03)

87 6 8.7 3.3 (1.3-6.4) 310 6 76 0.6 (0.2-2.1)

93 6 13 2.3 (0.6-6.2) 41 6 9.5 6.5 (2.7-16.6)

304 6 195 4.3 (3.0-11) 574 6 243 12 (5.2-24)

.68 .015 <.001 <.001

P values meeting Bonferroni correction for significance are presented in boldface. Eos, Eosinophils; Lym, lymphocytes; Neu, neutrophils. *Results from ANOVA are presented as means 6 SEMs; results from Kruskal-Wallis ANOVA are presented as median (25% to 75%).