Inflammatory Phenotypes in Patients With Severe Asthma Are Associated With Distinct Airway Microbiology

Inflammatory phenotypes in patients with severe
asthma are associated with distinct airway

microbiology
Steven L. Taylor, BSc,a,b* Lex E. X. Leong, PhD,a,b* Jocelyn M. Choo, PhD,a,b Steve Wesselingh, FRACP, PhD,a,b
Ian A. Yang, FRACP, PhD,c,d John W. Upham, FRACP, PhD,c,e Paul N. Reynolds, MBBS, PhD,f,g Sandra Hodge, PhD,f,g
Alan L. James, FRACP, PhD,h,i Christine Jenkins, MBBS, FRACP,j,k Matthew J. Peters, MD, FRACP,k,l
Melissa Baraket, FRACP, PhD,m,p Guy B. Marks, MBBS, PhD,m,n,p Peter G. Gibson, MBBS, FRACP,n,o
Jodie L. Simpson, PhD,oà and Geraint B. Rogers, PhDa,bà Adelaide, St Lucia, Chermside, Wooloongabba, Nedlands,
Crawley, Newtown, North Ryde, Concord, Liverpool, Glebe, Callaghan, and Sydney, Australia
GRAPHICAL ABSTRACT
Background: Asthma pathophysiology and treatment Objective: We aimed to characterize the airway microbiota in
responsiveness are predicted by inflammatory phenotype. patients with symptomatic stable asthma and relate composition
However, the relationship between airway microbiology and to airway inflammatory phenotype and other phenotypic
asthma phenotype is poorly understood. characteristics.
From athe South Australian Health and Medical Research Institute, Adelaide; bthe Disclosure of potential conflict of interest: I. A. Yang’s, J. W. Upham’s, and C. Jenkins’
SAHMRI Microbiome Research Laboratory, School of Medicine, Flinders University, institution received a grant from National Health & Medical Research Council
Adelaide; cthe School of Medicine, University of Queensland, St Lucia; dthe Depart- (NHMRC), Australia, for this work. P. N. Reynolds’ and J. L. Simpson’s institution
ment of Thoracic Medicine, The Prince Charles Hospital, Chermside; ethe Transla- received a grant from John Hunter Charitable Trust and the NHMRC CRE Severe
tional Research Institute, Princess Alexandra Hospital, Woolloongabba; fthe Asthma for this work. S. Hodge is employed by the University of Adelaide; her
Department of Thoracic Medicine, Royal Adelaide Hospital and Lung Research Lab- institution has grants with NHMRC; and she received royalties from the book Lung
oratory, Hanson Institute, Adelaide; gthe School of Medicine, University of Adelaide, Macrophages in Health and Disease. G. B. Marks’ institution received grants from As-
Adelaide; hthe Department of Pulmonary Physiology and Sleep Medicine, Sir Charles traZeneca and GlaxoSmithKline for other works. P. G. Gibson’s institution received a
Gairdner Hospital, Nedlands; ithe School of Medicine and Pharmacology, University grant from NHMRC for this work and has grants from the NHMRC, AstraZeneca, and
of Western Australia, Crawley; jRespiratory Trials, George Institute for Global Health, GlaxoSmithKline for other works; he has personally received payment for lectures
Newtown; kthe Australian School of Advanced Medicine, Macquarie University, from AstraZeneca, GlaxoSmithKline, and Novartis. The rest of the other authors
North Ryde; lthe Department of Thoracic Medicine, Concord General Hospital; declare that they have no relevant conflicts of interest.
m
the Respiratory Medicine Department and Ingham Institute, Liverpool Hospital; Received for publication October 16, 2016; revised February 28, 2017; accepted for pub-
n
the Woolcock Institute of Medical Research, Glebe; oRespiratory and Sleep Medi- lication March 15, 2017.
cine, Priority Research Centre for Healthy Lungs, University of Newcastle, Callaghan; Corresponding author: Geraint B. Rogers, PhD, School of Medicine, Flinders University,
and pthe South Western Sydney Clinical School, University of New South Wales, University Drive, Bedford Park, Adelaide SA 5042, Australia. E-mail: geraint.
Sydney. rogers@sahmri.com.
*These authors contributed equally to this work. 0091-6749
àThese authors contributed equally to this work. Ó 2017 The Authors. Published by Elsevier Inc. on behalf of the American Academy of
Supported by the Australian National Health and Medical Research Council (NHMRC); Allergy, Asthma & Immunology. This is an open access article under the CC BY-NC-
grant no. 569246) and a grant from the John Hunter Hospital Charitable Trust. ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
http://dx.doi.org/10.1016/j.jaci.2017.03.044
Downloaded for FK UMI 3 Makassar (fkumimks3@gmail.com) at ClinicalKey Global Flex Package Trial from ClinicalKey.com by Elsevier on November 07, 2017.
For personal use only. No other uses without permission. Copyright ©2017. Elsevier Inc. All rights reserved.
2 TAYLOR ET AL J ALLERGY CLIN IMMUNOL
nnn 2017
Methods: The microbial composition of induced sputum

specimens collected from adult patients screened for a Abbreviations used
multicenter randomized controlled trial was determined by ICS: Inhaled corticosteroid
using 16S rRNA gene sequencing. Inflammatory phenotypes OTU: Operational taxonomic unit
were defined by sputum neutrophil and eosinophil cell PERMANOVA: Permutational multiple ANOVA
qPCR: Quantitative PCR
proportions. Microbiota were defined by using a- and
SIMPER: Similarity of percentages
b-diversity measures, and interphenotype differences were
identified by using similarity of percentages, network analysis,
and taxon fold change. Phenotypic predictors of airway
microbiology were identified by using multivariate linear associated with the degree of airway hyperresponsiveness among
regression. patients with suboptimally controlled asthma,8 interphenotype dif-
Results: Microbiota composition was determined in 167 ferences in airway microbiology are likely to be clinically
participants and classified as eosinophilic (n 5 84), neutrophilic important.
(n 5 14), paucigranulocytic (n 5 60), or mixed neutrophilic- The relationships between asthma inflammatory phenotypes
eosinophilic (n 5 9) asthma phenotypes. Airway microbiology and airway microbiology are likely to be complex and bidirec-
was significantly less diverse (P 5 .022) and more dissimilar tional. Asthma phenotypes represent immunologic and physico-
(P 5 .005) in neutrophilic compared with eosinophilic chemical differences within the lower airways that are likely to be
participants. Sputum neutrophil proportions, but not eosinophil reflected through their selective effect on microbial growth and
proportions, correlated significantly with these diversity airway clearance in divergent lower airway microbiota.9 Where
measures (a-diversity: Spearman r 5 20.374, P < .001; these differences involve the increased abundance of particular
b-diversity: r 5 0.238, P 5 .002). Interphenotype differences respiratory pathogens or a depletion of commensal populations,
were characterized by a greater frequency of pathogenic taxa at they could contribute substantially to the course of airway disease
high relative abundance and reduced Streptococcus, Gemella, or risk of adverse treatment events, such as corticosteroid-
and Porphyromonas taxa relative abundance in patients with associated pneumonia.10 On the other hand, the characteristics
neutrophilic asthma. Multivariate regression confirmed that of airway microbiology, even in the absence of frank infection,
sputum neutrophil proportion was the strongest predictor of could influence the asthma inflammatory phenotype. Defining
microbiota composition. the relationships between inflammatory phenotype and lower
Conclusions: Neutrophilic asthma is associated with airway airway microbiota would inform our understanding of asthma
microbiology that is significantly different from that seen in pathophysiology and could help identify prognostic markers.
patients with other inflammatory phenotypes, particularly Several previous studies have reported differences in airway
eosinophilic asthma. Differences in microbiota composition microbiology in patients with eosinophilic and noneosinophilic
might influence the response to antimicrobial and steroid asthma and the existence of significant relationships between this
therapies and the risk of lung infection. (J Allergy Clin Immunol microbiota composition and clinical asthma measures.5,11-14
2017;nnn:nnn-nnn.) However, although providing important insight, these studies
have involved relatively small and heterogeneous patient cohorts.
Key words: Asthma, microbiome, neutrophil, eosinophil Our study, based on participants enrolled in the Asthma and Mac-
rolides: the Azithromycin Efficacy and Safety Study
(ACTRN12609000197235), was more than 3 times the size of
Evidence-based patient stratification is required to address the any study performed previously and focused on a well-defined
heterogeneity of asthma severity and treatment responsiveness. population of patients with severe but stable asthma, the majority
Asthma phenotype based on characteristics of airway inflamma- of whom were treated with inhaled corticosteroids (ICSs).
tion are increasingly recognized as an important prognostic Through application of a systematic approach to microbiota
indicator.1 In addition to an allergen-induced, TH2 lymphocyte, characterization, we aimed to assess whether asthma inflamma-
IL-5–mediated, eosinophilic inflammatory response, asthma can tory phenotypes were associated with substantially different
also occur in the absence of eosinophilic inflammation (termed lower airway bacteriology (herein referred to as microbiology),
noneosinophilic asthma).2,3 Indeed, based on relative numbers to identify bacterial taxa that discriminate among inflammatory
of sputum eosinophils and neutrophils, 4 inflammatory subtypes phenotypes, and to determine the contribution of patient and
have been described: eosinophilic asthma, neutrophilic asthma, clinical characteristics to variation in the composition of the
mixed granulocytic asthma, and paucigranulocytic asthma.2 bacterial component of the microbiome.
Unlike the relatively well-defined mechanisms that result in
eosinophilic airway inflammation, those leading to noneosino-
philic asthma, particularly neutrophilic asthma, remain relatively
poorly understood.4 Furthermore, although noneosinophilic phe- METHODS
notypes occur across the spectrum of asthma severity,1,2,5 they typi- Study population
cally respond poorly to corticosteroids.2,5 Inflammatory For full methodology, see the Methods section in this article’s Online Repos-
itory at www.jacionline.org. Analysis was performed on samples collected as
phenotypes have also been shown to differ with respect to airway
part of the baseline screening population from the clinical Azithromycin Effi-
microbiology. Compared with other patients with asthma, those cacy and Safety Study (ACTRN12609000197235). Participants were recruited
with neutrophilic asthma are more likely to have a potentially path- from 8 centers across Australia (detailed in the Methods section in this article’s
ogenic organism identified by means of either culture-based6 or Online Repository). Predefined inflammatory phenotype categories based on
culture-independent5 approaches and have reduced airway bacte- sputum cell counts relative to patient age were assigned, as published previously
rial diversity.7 Given that airway microbiota composition is and detailed in the Methods section in this article’s Online Repository.3,15
J ALLERGY CLIN IMMUNOL TAYLOR ET AL 3
VOLUME nnn, NUMBER nn
Briefly, the neutrophilic phenotype was defined as 61% or greater neutrophils 174 that underwent 16S rRNA gene amplicon sequencing, a
(neutrophil percentage cutoff dependent on age), the eosinophilic phenotype further 7 were excluded because of an insufficient sequence read
as 3% or greater eosinophils, the paucigranulocytic phenotype as 61% or less depth (see Table E1 in this article’s Online Repository at www.
neutrophils and 3% or less eosinophils, and the mixed granulocytic phenotype jacionline.org). The remaining 167 subjects were classified as
as 61% or greater neutrophils and 3% or greater eosinophils.
one of 4 inflammatory phenotypes based on previously described
sputum inflammatory cell count percentages: neutrophilic
DNA extraction, 16S rRNA gene amplicon (n 5 14), eosinophilic (n 5 84), paucigranulocytic (n 5 60), or
sequencing, and gene copy numbers mixed granulocytic (n 5 9).3,15 There was no significant differ-
DNA extraction was performed on 100-mL sputum aliquots by using a ence in age, sex distribution, atopy, smoking history, ICS dose,
combined physical, enzymatic, and heat-based cell lysis, followed by phenol Global Initiative for Asthma treatment step, or mean Asthma
chloroform extraction and DNA recovery with EZ-10 Spin columns (Bio Basic, Control Questionnaire 6 score between these phenotypic groups,
Ontario, Canada). The V1-3 hypervariable region of the bacterial 16S rRNA gene as assessed by using multiple comparison tests (Table I). Howev-
was amplified from sputum DNA by using the modified primers 27F and 519R. er, there were significant differences in lung function, as assessed
Amplicons were cleaned, indexed, and sequenced according to the Illumina by both FEV1 percent predicted (P 5 .035) and FEV1/forced vital
MiSeq 16S Metagenomic Sequencing Library Preparation Protocol (Illumina, capacity percentage (P 5 .013).
San Diego, Calif). 16S rRNA sequence data were processed, as previously
After quality filtering and chimera removal, 16S rRNA gene
described.16 Spurious operational taxonomic units (OTUs) were removed sys-
amplicon sequencing resulted in a median read depth of 12,792
tematically by using previous reports of common laboratory sequencing
contaminants.17 Bacterial burden was quantified with quantitative PCR
(quartile 1 and quartile 3, 8,060 and 16,595). Sequence data were
(qPCR) for the 16S rRNA gene.18 Detailed extraction, sequencing, and qPCR subsampled to a uniform depth of 1,732 reads based on rarefaction
protocols are described in the Methods section in this article’s Online Repository. curve asymptotes and Good’s coverage values. No significant
differences in total bacterial burden were found between inflam-
matory phenotypes (P 5.51, Kruskal-Wallis test; see Fig E1 in this
Diversity measurements and statistical analyses article’s Online Repository at www.jacionline.org).
Five a-diversity (within-sample variance) indices were used to test a variety
of parameters of within-patient taxon distribution: Faith’s phylogenetic diversity
(in which a higher value indicates a more phylogenetically diverse sample),
a-Diversity
Simpson’s and Pielou’s evenness indices (in which a higher value indicates a
Participants with neutrophilic asthma had significantly lower
more equitable distribution of taxa abundance), taxa richness (the total number
of taxa detected), and Shannon-Weiner diversity (a measure incorporating both
Faith’s phylogenetic scores (P 5 .022) than participants with
the number and equitability of detected taxa). b-Diversity (intersample eosinophilic asthma, which resembled those of patients with pau-
variance) was determined by using 2 approaches: weighted UniFrac similarity cigranulocytic asthma (Fig 1, A). Faith’s phylogenetic diversity
(which accounts for phylogenetic distance) and square root–transformed Bray- significantly correlated with the sputum neutrophil percentage
Curtis similarity (based on the relative abundance of taxa alone). Both a- and b- (r 5 20.374, P < .0001; Fig 1, B) but not with the sputum eosin-
diversity measures were calculated with either QIIME (version 1.8.0) or ophil percentage (r 5 0.146, P 5 .060; Fig 1, C). Analysis with a
PRIMER (version 6; PRIMER-E, Plymouth, United Kingdom) software. range of alternative a-diversity indices (taxa richness, Shannon-
Continuous data were tested for nonnormality, including skewness and Wiener index, Simpson index, and Pielou evenness; see Figs E2
kurtosis, by using the D’Agostino-Pearson omnibus test. Kruskal-Wallis 1-
and E3 in this article’s Online Repository at www.jacionline.
way ANOVAwith the Dunn post hoc test was used for multiple comparisons of
org) resulted in consistent findings in relation to phenotype,
nonnormally distributed data, the Mann-Whitney U test was used for pairwise
comparisons, the x2 test was used for categorical data, and the Spearman test
sputum neutrophil percentage, and sputum eosinophil percentage.
was used for correlations (GraphPad Prism, version 7.01; GraphPad Software, Together, these results demonstrate a significant relationship be-
La Jolla, Calif). tween airway microbiota composition and sputum neutrophilia
Multivariate linear regression was performed with Faith’s phylogenetic but not sputum eosinophilia.
diversity and UniFrac distance from centroid as 2 dependent variables
reflecting aspects of diversity (SPSS software, version 23.0; IBM, Armonk,
NY). Covariates were selected a priori and included in the model based on a b-Diversity
significant correlation with either dependent variable. CIs were obtained by Principal coordinate analysis of weighted UniFrac similarity
means of bootstrapping, resampling 1000 times. Covariates were tested for distance showed that neutrophilic samples were distinguished
collinearity by using variance inflation factors.
from other phenotypes along the first and second principal
coordinates, whereas the other phenotypes broadly clustered
Taxon dispersion together (Fig 2, A). Consistent with these observations, a permu-
Variation in microbiota composition between groups was assessed by using tational multiple ANOVA (PERMANOVA) test showed that
similarity of percentages (SIMPER) analysis in PRIMER and SparCC, from phenotype grouping contributed significantly to differences in mi-
which correlations (r > _ 20.25) and P values (P <
_ 0.25 or r < _ .01) were visual- crobial composition of the samples (P 5 .0004, pseudo-
ized by means of network analysis with Cytoscape (version 3.4.0). Pathogen F 5 3.997; see Table E2 in this article’s Online Repository at
overgrowth adjustment is described in the Methods section in this article’s On- www.jacionline.org). Pairwise PERMANOVA comparing the
line Repository. phenotype groups indicated that variance was attributed to the
neutrophilic versus eosinophilic (P 5 .0001, T 5 3.30) and
neutrophilic versus paucigranulocytic (P 5 .0015, T 5 2.52)
RESULTS groups (see Table E3 in this article’s Online Repository at
Clinical characteristics www.jacionline.org).
Induced sputum samples were obtained from 187 participants. Assessment of microbiota dispersion based on distance from
Of these, 13 were excluded because of poor sample quality. Of the centroid was consistent with PERMANOVA analysis, with the
nnn 2017
TABLE I. Clinical and inflammatory cell parameters of participants

Neutrophilic Eosinophilic Paucigranulocytic Mixed granulocytic P value
No. 14 84 60 9
Age (y), mean (SD) 59.8 (13.9) 57.2 (15.2) 55.8 (13.8) 56.8 (17.8) .817
Male sex, no. (%) 9 (64.3) 30 (35.7) 23 (38.3) 6 (66.7) .084
Atopic, no. (%) 11 (78.6) 62 (76.5), n 5 81 52 (86.7) 6 (66.7) .305
Previous smoker, no. (%) 6 (42.9) 32 (38.1) 17 (28.3) 4 (44.4) .491
Smoking pack years, median 22.0 (18.6, 27.5) 5.3 (1.2, 15.9) 5.0 (1.3, 30.0) 4.4 (0.9, 63.6) .379
(Q1, Q3)
Duration of asthma (y), median 33.8 (3.5, 48.3) 36.3 (19.7, 49.2) 33.7 (14.2, 54.4) 53.6 (32.6, 60.4) .205
(Q1, Q3)
FEV1 (% predicted), mean (SD) 70.3 (18.2) 70.0 (17.9) 78.7 (19.2)* 68.4 (17.3) .035
FVC (% predicted), mean (SD) 81.4 (11.6) 83.6 (16.4) 85.1 (15.0) 82.9 (11.7) .854
FEV1/FVC (%), mean (SD) 65.8 (14.5) 64.9 (12.2) 71.3 (12.2)* 63.9 (12.7) .013
ACQ6 score, mean (SD) 2.1 (1.2) 1.9 (0.9) 1.6 (0.8) 1.5 (0.6) .210
GINA treatment step .058
1 1 (7.1) 2 (2.5) 0 (0) 0 (0)
2 0 (0) 0 (0) 0 (0) 0 (0)
3 0 (0) 10 (12.4) 12 (20.3) 2 (22.2)
4 11 (78.6) 67 (82.7) 47 (80.0) 7 (77.8)
5 2 (14.3) 2 (2.5) 0 (0) 0 (0)
ICS dose (mg), median (Q1, Q3) 2000 (1280, 2000), n 5 13 1000 (800, 2000), n 5 82 1000 (800, 2000), n 5 59 1600 (1000, 2000) .254
Total cell count (3 106/mL 9.8 (7.6, 12.7)* 3.6 (1.9, 7.6) 3.74 (2.0, 7.6) 8.64 (5.22, 11.34) <.001
[Q1, Q3])
Viability (%) 89.2 (73.5, 93.0)* 69.8 (52.2, 80.7) 68.9 (55.7, 79.5) 90.0 (84.4, 93.7)* <.001
Neutrophils (%) 75.0 (68.80, 84.00)* 27.13 (14.38, 41.00) 34.13 (12.63, 49.25) 77.00 (71.25, 77.50)* <.001
Eosinophils (%) 0.63 (0.50, 1.00)* 6.88 (3.63, 18.13) 0.25 (0.00, 1.00)* 6.75 (3.25, 12.25) à <.001
Macrophages (%) 20.75 (14.50, 30.25)* 52.75 (38.25, 71.00) 53.34 (43.50, 75.63) 16.00 (15.50, 20.25)* <.001
Lymphocytes (%) 0.63 (0.25, 1.00) 1.13 (0.25, 2.25) 0.88 (0.25, 2.63) 0.25 (0.00, 0.38)* .018
Columnar epithelial cells (%) 0.25 (0.00, 1.25)* 2.17 (0.75, 6.00) 4.13 (1.21, 8.50) 0.50 (0.25, 1.50) <.001
Squamous cells (%) 1.11 (0.50, 3.85)* 5.99 (2.32, 16.23) 6.18 (2.92, 11.31) 3.38 (0.25, 8.05) .002
P values in the last column describe variance across 4 phenotypes.
ACQ6, Asthma Control Questionnaire 6; FVC, forced vital capacity; GINA, Global Initiative for Asthma; Q1, Q3, quartile 1, quartile 3.
*P < .05 versus eosinophilic asthma.
P < .05 versus paucigranulocytic asthma.
àP < .05 versus neutrophilic asthma.
samples from neutrophilic phenotype participants having signif- with a mixed phenotype (x2 5 25.5, P < .0001; Fig 3). Rela-
icantly higher distances from the centroid than samples from tionships between bacterial taxon relative abundance were
eosinophilic and paucigranulocytic participants (Fig 2, B). In further visualized by using network analysis (Fig 4 and see
keeping with a-diversity analyses, variance in distance from Fig E6 in this article’s Online Repository at www.jacionline.
centroid was associated with sputum neutrophil percentage rather org), revealing a bacterial community of taxa with positively
than sputum eosinophil percentage (see Fig E4 in this article’s correlated abundances in almost all cases. Most of these taxa
Online Repository at www.jacionline.org). b-Diversity analyses were more prevalent in eosinophilic samples than in neutro-
using a second distance measure, Bray-Curtis similarity, pro- philic samples. Haemophilus taxon, which had a mean abun-
duced consistent findings (see Fig E5 in this article’s Online Re- dance that was higher in neutrophilic samples, was the single
pository at www.jacionline.org). exception, negatively correlating with other members of the
sputum bacterial community.
Of the 13 discriminant taxa identified by SIMPER, Strepto-
Taxon distribution and network analysis coccus II (see the Methods section in this article’s Online Repos-
SIMPER analysis was used to rank taxa according to their itory at www.jacionline.org for classification), Gemella, Rothia,
contribution to intergroup variance in microbiota composition. and Porphyromonas taxa were significantly less abundant in
Thirteen taxa were identified, which cumulatively accounted neutrophilic than in eosinophilic and paucigranulocytic pheno-
for approximately 50% of total variance between neutrophilic types (Fig 5, A). Sputum neutrophil percentage positively corre-
and eosinophilic samples (see Table E4 in this article’s Online lated with the relative abundance of Moraxella taxon and
Repository at www.jacionline.org). Hierarchical cluster anal- negatively correlated with the relative abundance of Strepto-
ysis based on relative taxon abundance revealed that Moraxella coccus I, Gemella, and Porphyromonas taxa (Fig 5, B). In
and Haemophilus clustered separately from the other 11 taxa contrast, Haemophilus taxon negatively correlated with eosino-
(Fig 3). In the patients with neutrophilic asthma, Moraxella phil percentage, and Streptococcus I, Neisseria, and Gemella
and Haemophilus taxa exceeded 40% relative abundance in 6 taxa positively correlated with eosinophil percentage (see Fig
(42.9%) of 14 samples compared with only 1 (1.19%) of 84 E7 in this article’s Online Repository at www.jacionline.org).
patients with eosinophilic asthma, 7 (11.7%) of 60 patients Prevotella, Actinomyces, Leptotrichia, and Veillonella taxa,
with paucigranulocytic asthma, and 1 (11.1%) of 9 patients although identified by using SIMPER and represented by highly
FIG 1. Faith’s phylogenetic diversity is significantly associated with sputum neutrophilia but not
eosinophilia. A, Patients grouped by asthma phenotype. B, Neutrophil percentage. The dotted line at
61% neutrophils indicates the phenotype cutoff point. C, Eosinophil percentage. The dotted line at 3% eo-
sinophils indicates the phenotype cutoff point. Colors represent the asthma phenotype: blue is greater than
61% neutrophils, green is greater than 3% eosinophils, yellow is less than 61% neutrophils and less than 3%
eosinophils (paucigranulocytic), and purple is both greater than 61% neutrophils and greater than 3% eosin-
ophils (mixed). Statistical significance was assessed by using the Kruskal-Wallis 1-way ANOVA with the
Dunn post hoc test (Fig 1, A) or Spearman rank correlation (Fig 1, B and C).
connected nodes in the network analysis, neither differed between

airway inflammatory phenotypes nor correlated with sputum cell
counts.
Nondominant microbiome
We sought to establish whether differences in microbiota
composition between inflammatory phenotypes were explained
solely by overgrowth of opportunistic taxa (eg, Haemophilus and
Moraxella taxa) or whether differences existed even in the
absence of pathogen predominance. Two separate approaches
were used to investigate this: rescaling of relative abundance
data after exclusion of pathogen predominance and assessment
of ranked taxon fold change between neutrophilic and eosino-
philic groups based on nonsubsampled taxa counts. Rescaled rela-
tive abundance data remained significantly different between
inflammatory phenotypes (P 5 .0004, pseudo-F 5 2.38;
see Table E5 in this article’s Online Repository at www.
jacionline.org). Pairwise tests revealed significant differences be-
tween participants with neutrophilic and eosinophilic phenotypes
(P 5 .0001, T 5 2.31) and between participants with neutrophilic
and paucigranulocytic phenotypes (P 5 .0002, T 5 2.26; see
Table E5). Assessment of taxa count2ranked fold change sup-
ported these findings, with significant taxa count differences be-
tween participants with neutrophilic and eosinophilic
phenotypes (see Fig E8 in this article’s Online Repository at
www.jacionline.org).
Clinical and inflammatory associations with

microbiota composition
In univariate analysis Faith’s phylogenetic diversity signifi-
cantly inversely correlated with sputum neutrophil percentage,
age, and ICS dose and significantly positively correlated with
FEV1 percentage (Table II). Conversely, weighted UniFrac dis-
tance from centroid significantly positively correlated with
sputum neutrophil percentage and was significantly different FIG 2. Microbiota dispersion grouped by asthma phenotype. A, Principal
coordinate analysis. The first 2 principal coordinates are plotted on the x-
based on sex and atopy but not with age, ICS dose, FEV1 percent- and y-axes, respectively (representing 59.9% of the total variation). B, Dis-
age, or previous smoking status (Table II). In multivariate anal- tance from centroid. Statistical significance was assessed by using
ysis, sputum neutrophil percentage was the only variable that Kruskal-Wallis 1-way ANOVA with the Dunn post hoc test.
nnn 2017
FIG 3. Relative abundance of discriminant taxa among asthma phenotypes. The 13 taxa that collectively
contribute to approximately 50% of variance among phenotypes, as determined by using SIMPER analysis.
Clustering shows the similarity relationship of genera based on Bray-Curtis similarity distance and the
single linkage hierarchical clustering method. *Taxa assigned Actinomyces species uncultured bacteria.
#Taxa assigned Actinomyces species oral clone DR002.
FIG 4. Bacterial network analysis of the asthma cohort. Each edge represents a significant correlation
colored to indicate either positivity (blue) or negativity (red). Edge width and transparency are proportional
to the absolute value of the correlation coefficient. Node size is proportional to mean relative abundance.
Node hue is proportional to the difference in taxon relative abundance between the neutrophilic phenotype
group and the eosinophilic phenotype group. Correlations were performed with SparCC with a correlation
cutoff R value of greater than 0.25 or less than 20.25. *Actinomyces species uncultured bacteria. #Actino-
myces species oral clone DR002.
independently predicted both Faith’s phylogenetic diversity and (P 5 .018 [95% CI 5 1.3 to 8.4] and P 5 .039 [95%
weighted UniFrac distance from centroid (P 5 .002 [95% CI, CI 5 27.5 to 20.30], respectively).
20.07 to 20.02] and P < .001 [95% CI, 0.07 to 0.22], respec-
tively; Table III). Age and ICS dose both independently predicted
Faith’s diversity (P 5 .030 [95% CI 5 20.07 to 20.004] and DISCUSSION
P 5 .042 [95% CI 5 20.001 to 20.001], respectively), whereas To our knowledge, this is the largest study to date to assess
atopy and sex independently predicted distance from centroid predictors of the airway microbiota composition in asthmatic
FIG 5. Taxa distribution differs by sputum neutrophilia. A, Taxa that significantly differ by patient inflamma-
tory phenotype. B, Significant correlations between taxa and neutrophil percentages. Colors represent
asthma phenotype based on neutrophilia or eosinophilia: blue is greater than 61% neutrophils, green is
greater than 3% eosinophils, yellow is less than 61% neutrophils and less than 3% eosinophils (paucigranu-
locytic), and purple is both greater than 61% neutrophils and greater than 3% eosinophils (mixed). The
dotted line at 61% neutrophils indicates the phenotype cutoff point. Statistical significance was assessed
by using Kruskal-Wallis 1-way ANOVA with the Dunn post hoc test (Fig 5, A) and Spearman rank correlation
(Fig 5, B).
nnn 2017
TABLE II. Comparison of patients’ characteristics with a-diversity (Faith’s phylogenetic diversity) and b-diversity (weighted
UniFrac distance from centroid) assessed by using the Spearman or Mann-Whitney test
Neutrophil percentage Age ICS dose* FEV1 (% predicted) Atopyyz Sexy Ever smokedy
Faith’s diversity r 20.374 20.309 20.242 0.193

P value <0.001 <0.001 0.002 0.013 0.32 0.97 0.53
UniFrac distance r 0.24 0.015 0.096 0.034
P value 0.002 0.84 0.22 0.66 0.019 0.003 0.70
Spearman correlation coefficient (r) and probability values are as indicated.
*n 5 163.
Assessed by using the Mann-Whitney test.
àn 5 164.
TABLE III. Multivariate linear regression on a-diversity (Faith’s phenotypes, namely high abundance of Haemophilus and Morax-
phylogenetic diversity) and b-diversity (weighted UniFrac ella taxa in neutrophilic participants, supporting previous find-
distance from centroid) performed on 160 participants ings.5 This could be interpreted as simply an increased relative
B 95% CI P value abundance of airway pathogens in neutrophilic patients, reflecting
neutrophilic influx into the airways during subclinical lower
Faith’s phylogenetic diversity
airway infection, with a reciprocal decrease in the relative abun-
Neutrophil percentage 20.046 20.07 to 20.02 .002
Age 20.036 20.07 to 20.004 .030 dance of commensal taxa. However, a group of common airway
Sex 0.35 20.59 to 1.4 .49 taxa correlated negatively with sputum neutrophil percentages
Atopy 20.77 21.7 to 0.12 .10 (Gemella, Porphyromonas, and Streptococcus taxa) and, impor-
FEV1 (% predicted) 0.016 20.01 to 0.04 .23 tantly, even after controlling for overgrowth effects of Haemophi-
ICS dose 20.001 20.001 to 20.001 .042 lus and Moraxella taxa, significant differences in microbiota
UniFrac distance from centroid composition between neutrophilic and eosinophilic participants
Neutrophil percentage 0.14 0.07 to 0.22 <.001 were still observed. This finding suggests that 2 separate phenom-
Age 0.056 20.05 to 0.16 .30 ena contribute to microbial differences between inflammatory
Sex 23.9 27.5 to 20.30 .039
phenotypes: the effect of pathogen overgrowth and the selective
Atopy 4.8 1.3 to 8.4 .018
FEV1 (% predicted) 0.049 20.04 to 0.14 .27
pressure of airway inflammatory characteristics in the absence
ICS dose <0.001 20.001 to 0.001 .56 of infection. The latter could result in more broad-scale diver-
gence in composition between the neutrophilic and eosinophilic
subgroups, in turn contributing to an increased risk of lower
airway infection in neutrophilic patients through an increased
patients. Our primary comparisons were between asthma inflam- presence of opportunistic pathogens.
matory phenotypes, where we observed significant differences in This finding has clear implications for the clinical management
the composition of airway microbiota. These differences were of asthma, in which low-dose macrolide and ICS therapies have
largely between neutrophilic and eosinophilic phenotypes and been shown to influence overgrowth by opportunistic respiratory
reflected a reduced diversity and evenness of detectable bacterial pathogens and innate immune function, respectively.6,14,25
taxa in the neutrophilic participants. Reduced microbiota di- Furthermore, the relative lack of efficacy of ICSs in patients
versity has been reported after acute and chronic airway in- with noneosinophilic asthma2 might lead to use of higher doses
fections in asthmatic patients5,7 and in those with other compared with those used by eosinophilic patients. The combina-
respiratory disorders,19-21 as well as with the effects of exposure tion of underlying differences in airway microbiota (associated
to antibiotics.22,23 Importantly, none of the study participants re- with differences in inflammatory phenotype) and inefficacious
ported clinical features of respiratory tract infection or had anti- therapies being used at higher doses might contribute to reduced
biotic therapy during the preceding month. bacterial diversity,14 the high concentrations of Proteobacteria
We further assessed a-diversity metrics relative to continuous seen in the airways of neutrophilic patients,7 a greater propensity
neutrophil and eosinophil count data as an alternative to categor- for lung infection, and a further enhancement of the neutrophilic
ical inflammatory phenotypes. Significant correlations were phenotype.6
observed between sputum neutrophil percentages and each Although strong associations between the neutrophilic
assessed a-diversity metric, with no significant interactions phenotype and sputum microbiota composition were found,
between any diversity metric and sputum eosinophil percentages, associations between eosinophil counts and microbiota compo-
strongly suggesting that decreased microbiota richness, evenness, sition were minimal. This contrasts with previous studies
and diversity are associated with airway neutrophilia. Analysis of reporting increased Tropheryma taxon associated with eosino-
sputum microbiota b-diversity (intersample similarity) also philia7 and associations between bronchial biopsy eosinophil
demonstrated substantial differences between patients with count and bacterial composition11 and reduced bacterial burden
neutrophilic airway inflammation and those with other inflam- associated with type 2–high airway inflammation.26
matory phenotypes, which is consistent with the stochastic An important strength of our study was its involvement of a
overgrowth of complex commensal communities by individual large group of well-defined participants with poorly controlled
opportunistic pathogens.24 asthma who were taking regular inhaled therapy. The application
We identified bacterial taxa that contributed to observed of detailed induced sputum microbiota characterization from
differences in microbiota composition between inflammatory these participants then allowed us to assess the extent to which
clinical and inflammatory characteristics independently associ- Madzinga, Melissa McClean, Pamela Fung, Janet Shaw, Joanne McNamara,
ated with variations in airway microbiology by using multivariate Kevin Oreo, Peta Grayson, Robyn Jones, Tessa Bird, Kerrie Wade, Sara Baum,
linear regression analysis. Multivariate regression identified the Chaitali Patel, Michelle Towers, Tina Collins, Melanie Carroll, Alice Chen,
sputum neutrophil percentage as the strongest predictor of Chris Choo, Mary Vukovich, and Kerry Carson for their assistance.
microbiota variance. However, age, ICS dose, sex, and atopy
were also significant independent predictors; conversely, lung Key messages
function (as measured by FEV1 percentage) and smoking status
d Lower airway microbiology differs significantly between
were not.
patients with neutrophilic and eosinophilic asthma char-
Of particular interest was the finding that increasing age
acterized based on both increased frequency of patho-
predicted reduced microbiota a-diversity because age has
genic taxa and divergence in the wider airway microbiota.
been previously associated with microbiome composition in
patients with other chronic respiratory diseases. The airway d Sputum neutrophilia is the strongest predictor of airway
microbiome of patients with cystic fibrosis is strongly microbiota composition, with age, ICS dose, sex, and
affected by age,27 which is presumed to relate to the selective atopy also being independent predictors.
effects of increased antibiotic exposure over time and the d The clear relationships between airway microbiota
changing characteristics of the airway environment.9 It is composition, inflammatory phenotype, and clinical mea-
interesting to speculate that the relationship between neutro- sures suggest microbiota characterization could be a use-
philia and microbiota composition might reflect the effect ful contributor to the individualization of asthma
of age on neutrophilia,15 suggesting that a tendency toward treatments.
a neutrophilic phenotype and/or a susceptibility to opportu-
nistic airway infections increases with age in patients with
severe asthma.
It is important to recognize a number of limitations of our REFERENCES
study. Airway microbiology was assessed based on induced 1. Wenzel SE. Asthma phenotypes: the evolution from clinical to molecular ap-
sputum, which, although shown to provide reproducible proaches. Nat Med 2012;18:716-25.
inflammatory cell levels in patients with moderate-to-severe 2. Pavord ID, Brightling CE, Woltmann G, Wardlaw AJ. Non-eosinophilic corticoste-
roid unresponsive asthma. Lancet 1999;353:2213-4.
asthma,28,29 only provides an approximation of lower airway
3. Simpson JL, Scott R, Boyle MJ, Gibson PG. Inflammatory subtypes in asthma:
microbiology (in common with other lower airway sampling assessment and identification using induced sputum. Respirology 2006;11:54-61.
strategies).30 It is also important to note that induced sputum 4. Pelaia G, Vatrella A, Busceti MT, Gallelli L, Calabrese C, Terracciano R, et al.
levels of neutrophils and eosinophils can change Cellular mechanisms underlying eosinophilic and neutrophilic airway inflamma-
frequently31,32 and that the relationships between airway tion in asthma. Mediat Inflamm 2015;2015:879783.
5. Green BJ, Wiriyachaiporn S, Grainge C, Rogers GB, Kehagia V, Lau L, et al.
microbiology and inflammatory phenotype reported are Potentially pathogenic airway bacteria and neutrophilic inflammation in treatment
cross-sectional. Detailed longitudinal analysis is now required resistant severe asthma. PLoS One 2014;9:e100645.
to determine how these relationships change with time. 16S 6. Essilfie AT, Simpson JL, Dunkley ML, Morgan LC, Oliver BG, Gibson PG, et al. Com-
rRNA gene amplicon sequence data were subsampled to a bined Haemophilus influenzae respiratory infection and allergic airways disease
drives chronic infection and features of neutrophilic asthma. Thorax 2012;67:588-99.
level that allowed inclusion of the greatest number of subjects
7. Simpson JL, Daly J, Baines KJ, Yang IA, Upham JW, Reynolds PN, et al. Airway
while maintaining sufficient read depth to accurately describe dysbiosis: Haemophilus influenzae and Tropheryma in poorly controlled asthma.
microbiota composition. However, additional analysis with a Eur Respir J 2016;47:792-800.
greater read depth might identify rare taxa that contribute to 8. Huang YJ, Nelson CE, Brodie EL, Desantis TZ, Baek MS, Liu J, et al. Airway mi-
disease characteristics. Although multivariate regression iden- crobiota and bronchial hyperresponsiveness in patients with suboptimally
controlled asthma. J Allergy Clin Immunol 2011;127:372-81, e1-3.
tified sputum neutrophilia as an independent predictor of mi- 9. Rogers GB, Hoffman LR, Carroll MP, Bruce KD. Interpreting infective microbiota:
crobiota composition, the effects of variation in ICS dose the importance of an ecological perspective. Trends Microbiol 2013;21:271-6.
between phenotypes (although nonsignificant) should be 10. McKeever T, Harrison TW, Hubbard R, Shaw D. Inhaled corticosteroids and the risk
noted. Finally, although none of the study participants of pneumonia in people with asthma: a case-control study. Chest 2013;144:1788-94.
11. Huang YJ, Nariya S, Harris JM, Lynch SV, Choy DF, Arron JR, et al. The airway
received antibiotics in the month before recruitment, data on
microbiome in patients with severe asthma: associations with disease features and
less recent exposure were not available and could have a last- severity. J Allergy Clin Immunol 2015;136:874-84.
ing effect on the lower airway microbiome composition.22 12. Simpson JL, Carroll M, Yang IA, Reynolds PN, Hodge S, James AL, et al. Reduced
The clear relationship between airway inflammatory pheno- antiviral interferon production in poorly controlled asthma is associated with
type and the microbiota highlight the need for studies examining neutrophilic inflammation and high-dose inhaled corticosteroids. Chest 2016;
149:704-13.
whether asthma treatments should be individualized based on 13. Zhang Q, Cox M, Liang Z, Brinkmann F, Cardenas PA, Duff R, et al. Airway mi-
both inflammatory phenotype stratification and lower airway crobiota in severe asthma and relationship to asthma severity and phenotypes.
microbiology. There is now a clear need to investigate the extent PLoS One 2016;11:e0152724.
to which variations in the airway microbiota predict the risk of 14. Denner DR, Sangwan N, Becker JB, Hogarth DK, Oldham J, Castillo J, et al. Corti-
costeroid therapy and airflow obstruction influence the bronchial microbiome,
future asthma exacerbations and to determine whether airway
which is distinct from that of bronchoalveolar lavage in asthmatic airways.
microbiota characterization could be used as a basis for asthma J Allergy Clin Immunol 2016;137:1398-405.e3.
treatment selection. 15. Brooks CR, Gibson PG, Douwes J, Dalen CJV, Simpson JL. Relationship between
airway neutrophilia and ageing in asthmatics and non-asthmatics. Respirology
2013;18:857-65.
We thank Heather Powell, Catherine Delahunty, Kellie Fakes, Bridgette 16. Jervis-Bardy J, Leong LE, Marri S, Smith RJ, Choo JM, Smith-Vaughan HC, et al.
Donati, Michelle Gleeson, Erin Harvey, Calida Garside, Gabriele le Brocq, Deriving accurate microbiota profiles from human samples with low bacterial con-
Kelly Steel, Sandra Dowley, Amy Cashmore, Gloria Foxley, Michael Guo, I- tent through post-sequencing processing of Illumina MiSeq data. Microbiome
Chin Wu, Monique de Pedro, Kirsty Herewane, Miranda Ween, Fungai 2015;3:19.
nnn 2017
17. Salter SJ, Cox MJ, Turek EM, Calus ST, Cookson WO, Moffatt MF, et al. Reagent bronchiectasis: an analysis from the randomised, double-blind, placebo-controlled
and laboratory contamination can critically impact sequence-based microbiome an- BLESS trial. Lancet Respir Med 2014;2:988-96.
alyses. BMC Biol 2014;12:1-12. 26. Durack J, Lynch SV, Nariya S, Bhakta NR, Beigelman A, Castro M, et al. Features
18. Nadkarni MA, Martin FE, Jacques NA, Hunter N. Determination of bacterial load of the bronchial bacterial microbiome associated with atopy, asthma, and respon-
by real-time PCR using a broad-range (universal) probe and primers set. Microbi- siveness to inhaled corticosteroid treatment. J Allergy Clin Immunol 2016 [Epub
ology 2002;148:257-66. ahead of print].
19. Cuthbertson L, Rogers GB, Walker AW, Oliver A, Green LE, Daniels TW, et al. 27. Zhao J, Schloss PD, Kalikin LM, Carmody LA, Foster BK, Petrosino JF, et al.
Respiratory microbiota resistance and resilience to pulmonary exacerbation and Decade-long bacterial community dynamics in cystic fibrosis airways. Proc Natl
subsequent antimicrobial intervention. ISME J 2016;10:1081-91. Acad Sci U S A 2012;109:5809-14.
20. Hofstra JJ, Matamoros S, van de Pol MA, de Wever B, Tanck MW, Wendt-Knol H, 28. Bacci E, Cianchetti S, Carnevali S, Bartoli ML, Dente FL, Di Franco A, et al.
et al. Changes in microbiota during experimental human rhinovirus infection. BMC Induced sputum is a reproducible method to assess airway inflammation in asthma.
Infect Dis 2015;15:336. Mediat Inflamm 2002;11:293-8.
21. Pettigrew MM, Gent JF, Kong Y, Wade M, Gansebom S, Bramley AM, et al. As- 29. Rossall MR, Cadden PA, Molphy SD, Plumb J, Singh D. Repeatability of
sociation of sputum microbiota profiles with severity of community-acquired pneu- induced sputum measurements in moderate to severe asthma. Respir Med
monia in children. BMC Infect Dis 2016;16:317. 2014;108:1566-8.
22. Daniels TW, Rogers GB, Stressmann FA, van der Gast CJ, Bruce KD, Jones GR, 30. Rogers GB, Skelton S, Serisier DJ, van der Gast CJ, Bruce KD. Deter-
et al. Impact of antibiotic treatment for pulmonary exacerbations on bacterial diver- mining cystic fibrosis-affected lung microbiology: comparison of sponta-
sity in cystic fibrosis. J Cystic Fibros 2013;12:22-8. neous and serially induced sputum samples by use of terminal
23. Huang YJ, Sethi S, Murphy T, Nariya S, Boushey HA, Lynch SV. Airway micro- restriction fragment length polymorphism profiling. J Clin Microbiol
biome dynamics in exacerbations of chronic obstructive pulmonary disease. J Clin 2010;48:78-86.
Microbiol 2014;52:2813-23. 31. Hancox RJ, Cowan DC, Aldridge RE, Cowan JO, Palmay R, Williamson A, et al.
24. Tarabichi Y, Li K, Hu S, Nguyen C, Wang X, Elashoff D, et al. The administration Asthma phenotypes: Consistency of classification using induced sputum. Respirol-
of intranasal live attenuated influenza vaccine induces changes in the nasal micro- ogy 2012;17:461-6.
biota and nasal epithelium gene expression profiles. Microbiome 2015;3:74. 32. Suarez-Cuartin G, Crespo A, Mateus E, Torrejon M, Giner J, Belda A, et al. Vari-
25. Rogers GB, Bruce KD, Martin ML, Burr LD, Serisier DJ. The effect of long-term ability in asthma inflammatory phenotype in induced sputum. Frequency and
macrolide treatment on respiratory microbiota composition in non-cystic fibrosis causes. Arch Bronconeumol 2016;52:76-81.
J ALLERGY CLIN IMMUNOL TAYLOR ET AL 10.e1
METHODS pelleted by means of centrifugation at 13,000g for 10 minutes. After removal

of supernatant, 300 mL of Tris-EDTA solution (10 mmol/L Tris-HCl and
Exclusion criteria
1 mmol/L EDTA [pH 8.0]; Ambion, Thermo Fisher Scientific, Victoria,
Asthma diagnosis was established by using American Thoracic Society
Australia), 200 mg of silica/zirconium beads (1:1 of 0.1 mm and 1.0 mm; Bio-
guidelines based on current episodic respiratory symptoms, clinical diagnosis,
Spec Products, Bartlesville, Okla), and a single chrome bead (3.2 mm; Bio-
and evidence of variable airflow obstruction.E1 Participants with asthma were
Spec Products) were added to the tube containing the cell pellet. Samples
included if stable but symptomatic, despite being prescribed maintenance ICS
underwent bead beating at 6.5 m/s for 60 seconds in a FastPrep-24 Instrument
and long-acting bronchodilator treatment with an Asthma Control Question-
(MP Biomedicals, Burlingame, Calif). The homogenized sample was heated
naire 6 score of greater than 0.75.E2
to 908C for 5 minutes before being cooled on ice for 5 minutes. Lysozyme
Participants with an FEV1 of less than 40% of predicted value, current
(ROCHE, Thermo Fisher Scientific, Victoria, Australia) and lysostaphin
smokers, ex-smokers who had ceased smoking in the previous year, and those
(Sigma-Aldrich, St Louis, Mo) were then added to a final concentration of 2
with a recent (past 4 weeks) exacerbation or respiratory tract infection were
and 0.1 mg/mL, respectively, and samples were incubated at 378C for
excluded. Those with significant smoking-related air-space disease (ex-
1 hour. Proteinase K (Fermentas, Thermo Fisher Scientific, Victoria,
smokers with a >10 pack year history and a diffusing capacity/alveolar volume
Australia) and sodium dodecyl sulfate (Sigma-Aldrich) were then added to
of less than 70% of predicted value or those with a smoking history of greater
a final concentration of 1.2 mg/mL and 1.5% (wt/vol), respectively. After in-
than 10 pack years and exhaled carbon monoxide >10 ppm) were also
cubation at 30 minutes at 568C, 40 mL of 5 mol/L sodium chloride and 450 mL
excluded. This study was conducted in accordance with the amended Decla-
of phenol/chloroform/isoamyl alcohol (25:24:1; saline buffered at pH 8.0;
ration of Helsinki. Local institutional review boards approved the protocol,
Sigma-Aldrich) were added, and samples were vortexed for 30 seconds. The
and written informed consent was obtained from all participants.
aqueous organic layers were separated by means of centrifugation at
13,000g for 10 minutes, and 400 mL of the aqueous layer was transferred to
Institutional centers a new microfuge tube. DNA was recovered by using an EZ-10 Spin column
Sputum samples were collected from 8 Australian centers: Hunter Medical in accordance with the manufacturer’s instructions (Bio Basic, Markham, On-
Research Institute, Newcastle, Australia; the Prince Charles Hospital, Cherm- tario, Canada) after precipitation by addition of 10 mol/L ammonium acetate
side, Australia; Princess Alexandra Hospital, Woolloongabba, Australia; and 99% ethanol (Sigma Aldrich) in a 1:10 and 1:1 ratio with sample volume,
Royal Adelaide Hospital, Adelaide, Australia; Sir Charles Gairdner Hospital, respectively. DNA was eluted in 50 mL of UltraPure DNase/RNase-free
Nedlands, Australia; the Woolcock Institute of Medical Research, Glebe, distilled water (Gibco, Thermo Fisher Scientific, Victoria, Australia) and
Australia; Concord Repatriation General Hospital, Concord, Australia; and stored at 2808C before analysis.
Liverpool Hospital, Liverpool, Australia.
16S rRNA gene amplicon sequencing
Sample collection The V1-3 hypervariable region of the bacterial 16S rRNA gene was
All study participants attended a single visit that included assessment of amplified from sputum DNA by using modified primers 27F (59-
lung function, asthma symptoms, asthma-specific quality of life,E3 medication TCGTCGGCAGCGTCAGATGTGTATAAGAGACAGAGRGTTTGATCMT
use, and smoking status. Sputum induction with hypertonic saline (4.5%) was GGCTCAG-39) and 519R (59-GTCTCGTGGGCTCGGAGATGTGTATAA
performed, as described previously.E4 Sputum aliquots were stored at 2808C GAGACAGGTNTTACNGCGGCKGCTG-39), with Illumina adapter over-
for DNA extraction or dispersed by using dithiothreitol for sputum cell count hang sequences as indicated by underlining. Amplicons were generated,
assessment and inflammatory subtype determination.E5 cleaned, indexed, and sequenced according to the Illumina MiSeq 16S Meta-
genomic Sequencing Library Preparation Protocol (http://support.illumina.
Patient inflammatory phenotyping com/downloads/16s_metagenomic_sequencing_library_preparation.html)
Patients’ sputum was dispersed by using dithiothreitol, and inflammatory with certain modifications. Briefly, an initial PCR reaction contained at least
cells were counted as a percentage of total sputum cells. Inflammatory subtype 12.5 ng of DNA, 5 mL of forward primer (1 mmol/L), 5 mL of reverse primer
was determined, as described below. Neutrophilic cutoff values were age (1 mmol/L), and 12.5 mL of 23 KAPA HiFi Hotstart ReadyMix (KAPA Bio-
dependent, as described previously.E6,E7 systems, Wilmington, Mass) in a total volume of 25 mL. The PCR reaction was
Neutrophilic phenotype. Values for the neutrophilic phenotype performed on a Veriti 96-well Thermal Cycler (Life Technologies, Grans Is-
were as follows: neutrophil percentage (<20 years old), 75.57% or greater; land, NY) by using the following program: 95 8C for 3 minutes, followed by
neutrophil percentage (20-40 years old), 61.61% or greater; neutrophil 25 cycles of 95 8C for 30 seconds, 55 8C for 30 seconds, and 72 8C for 30 sec-
percentage (40-60 years old), 63.25% or greater; and neutrophil percentage onds and a final extension step at 72 8C for 5 minutes. Samples were multi-
(>60 years old), 67.25% or greater. plexed by using a dual-index approach with the Nextera XT Index kit
(Illumina), according to the manufacturer’s instructions. The final library
Eosinophilic phenotype. Values for the eosinophilic phenotype
was paired-end sequenced at 2 3 300 bp by using a MiSeq Reagent Kit v3
were as follows: eosinophil percentage, 3% or greater.
on the Illumina MiSeq platform. Sequencing was performed at the David R
Paucigranulocytic phenotype. Values for the paucigranulo-
Gunn Genomics Facility, South Australian Health and Medical Research
cytic phenotype were as follows: eosinophil percentage, 3% or less; neutrophil
Institute.
percentage (<20 years old), 75.57% or less; neutrophil percentage (20-40 years
old), 61.61% or less; neutrophil percentage (40-60 years old), 63.25% or less;
and neutrophil percentage (>60 years old), 67.25% and less. 16S rRNA gene qPCR
Mixed granulocytic phenotype. Values for the mixed granu- The approximate 16S rRNA gene copy number was assessed by using
locytic phenotype were as follows: eosinophil percentage, 3% or greater; qPCR with the 16S rRNA universal primers B331F (59-TCCTACGGGAGG-
neutrophil percentage (<20 years old), 75.57% or greater; neutrophil percent- CAGCAGT-39) and B797R (59-GGACTACCAGGGTATCTAATCCTGTT-
age (20-40 years old), 61.61% or greater; neutrophil percentage (40-60 years 39) and Platinum SYBR Green (Thermo Fisher Scientific), as previously
old), 63.25% or greater; and neutrophil percentage (>60 years old), 67.25% or described.E8 Reactions were performed in duplicates, and averages were
greater. taken. Sample total bacterial copy number was calculated per microliter of
DNA eluate against a standard curve of a known bacterial copy number.
DNA extraction
DNA extraction was performed on sputum sample aliquots of approxi- Sequence data processing
mately 100 mL. After the addition of 300 mL of PBS, samples were vortexed The Quantitative Insights Into Microbial Ecology (version 1.8.0)E9
for 10 seconds and placed on ice for 2 minutes. Bacterial cells were then software was used to analyze the 16S rRNA sequence generated from
10.e2 TAYLOR ET AL J ALLERGY CLIN IMMUNOL
nnn 2017
paired-end amplicon sequencing by using a bioinformatics pipeline, as previ- described previously.E20 PERMANOVA analyses were then performed on
ously described.E10 Briefly, barcoded forward and reverse sequencing reads the rescaled data.
were quality filtered and merged with Paired-End reAd mergeR (version Second, pairwise comparisons between neutrophilic and eosinophilic
0.9.6).E11 Chimeras were detected and filtered from the paired-end reads by samples were performed by using the phyloseq R packageE21 with the DE-
using USEARCH (version 6.1)E12 against the 97% clustered representative se- seq2E22 extension based on count data. P values were corrected by using the
quences from the Greengenes database (versus 13.8).E13 OTUs were assigned Benjamini-Hochberg false discovery rate procedure, and a corrected a value
to the reads by using an open reference approach with the UCLUST algorithm cutoff of less than 0.05 was used for inclusion.
(version 1.2.22q) against the SILVA database (release 111, July 2012),E14
which was clustered at 97% identity. Spurious OTUs were then removed sys- REFERENCES
tematically by using previous reports of common laboratory sequencing con- E1. Standards for the diagnosis and care of patients with chronic obstructive pulmo-
taminants.E15 A minimum subsampling depth of 1732 reads was then selected nary disease (COPD) and asthma. Am Rev Respir Dis 1987;136:225-44.
for all samples. E2. Juniper EF, O’Byrne PM, Guyatt GH, Ferrie PJ, King DR. Development and vali-
Where taxa assignment did not classify to the family or genus level, OTU dation of a questionnaire to measure asthma control. Eur Respir J 1999;14:902-7.
reference sequences (accounting for >99% of OTU reads) were separately E3. Juniper EF, Guyatt GH, Ferrie PJ, Griffith LE. Measuring quality of life in
asthma. Am Rev Respir Dis 1993;147:832-8.
aligned by using the SILVA Incremental Aligner (https://www.arb-silva.de/),
E4. Gibson PG, Wlodarczyk JW, Hensley MJ, Gleeson M, Henry RL, Cripps AW,
which uses SILVA, RDP, Greengenes, LTP, and EMBL sequence collections. et al. Epidemiological association of airway inflammation with asthma symptoms
If the alignments identified taxa to the genus level and at greater than 99% and airway hyperresponsiveness in childhood. Am J Respir Crit Care Med 1998;
similarity, they replaced the previous taxon assignment. This occurred for 158:36-41.
Streptococcus II, which was previously incorrectly assigned as Clostridiales; E5. Simpson JL, Phipps S, Baines KJ, Oreo KM, Gunawardhana L, Gibson PG.
Other;Other. Streptococcus I refers to the OTU cluster, which was assigned Elevated expression of the NLRP3 inflammasome in neutrophilic asthma. Eur
as Streptococcus taxon during initial assignment. Respir J 2014;43:1067-76.
E6. Brooks CR, Gibson PG, Douwes J, Dalen CJV, Simpson JL. Relationship be-
tween airway neutrophilia and ageing in asthmatics and non-asthmatics. Respir-
Diversity measurements and statistical analyses ology 2013;18:857-65.
The Bray-Curtis matrix was calculated based on sample-normalized, E7. Simpson JL, Scott R, Boyle MJ, Gibson PG. Inflammatory subtypes in asthma:
square root–transformed relative taxon abundance. Principal coordinate assessment and identification using induced sputum. Respirology 2006;11:54-61.
analysis was used to visualize clustering of samples based on their similarity E8. Nadkarni MA, Martin FE, Jacques NA, Hunter N. Determination of bacterial load
matrices, with PCO1 and PCO2 coordinates and group centroids plotted by by real-time PCR using a broad-range (universal) probe and primers set. Micro-
using the ggplot2 package of R statistical software.E16 Distance from centroid biology 2002;148:257-66.
E9. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman FD, Costello EK,
was calculated, as previously described, by using PRIMER.E17 PERMANO-
et al. QIIME allows analysis of high-throughput community sequencing data. Nat
VAE18 on the b-diversity matrices was used to test the null hypothesis of no Methods 2010;7:335-6.
difference among a priori–defined groups with the PERMANOVA 1 add-on E10. Jervis-Bardy J, Leong LE, Marri S, Smith RJ, Choo JM, Smith-Vaughan HC,
package for PRIMER. The test was computed by using unrestricted permuta- et al. Deriving accurate microbiota profiles from human samples with low bacte-
tion of raw data with 9999 random permutations at a significance level of .01. rial content through post-sequencing processing of Illumina MiSeq data. Micro-
biome 2015;3:19.
E11. Zhang J, Kobert K, Flouri T, Stamatakis A. PEAR: a fast and accurate Illumina
Taxon dispersion Paired-End reAd mergeR. Bioinformatics 2014;30:614-20.
Variation in microbiota composition at the genus level was assessed by E12. Edgar RC. Search and clustering orders of magnitude faster than BLAST. Bioin-
using multiple approaches. First, taxa that contributed to the overall variation formatics 2010;26:2460-1.
between asthma phenotypes were identified by using SIMPER analysis in E13. McDonald D, Price MN, Goodrich J, Nawrocki EP, DeSantis TZ, Probst A, et al.
PRIMER. Subsequently, the abundance of the 13 highest ranked taxa An improved Greengenes taxonomy with explicit ranks for ecological and evolu-
(accounting for 50% of the dissimilarity between neutrophilic and eosino- tionary analyses of bacteria and archaea. ISME J 2012;6:610-8.
philic groups) were used to generate a heat map with the ggplot2 package of R E14. Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA
statistical software.E16 Hierarchical clustering of the taxa was performed on ribosomal RNA gene database project: improved data processing and web-
based tools. Nucleic Acids Res 2013;41(Database issue):D590-6.
Bray-Curtis dissimilarity and clustered by using the single linkage method.
E15. Salter SJ, Cox MJ, Turek EM, Calus ST, Cookson WO, Moffatt MF, et al. Re-
Dominance of Haemophilus and Moraxella taxa was determined when the agent and laboratory contamination can critically impact sequence-based micro-
relative abundance of each taxa exceeded 40%. This cutoff was selected based biome analyses. BMC Biol 2014;12:1-12.
to the distribution of the relative abundance, where a clear distinction between E16. Wickham H. ggplot2: elegant graphics for data analysis. New York: Springer-Ver-
samples with greater than 40% and less than 40% was evident, suggesting lag; 2009. http://ggplot2.org.
overgrowth of these taxa. E17. Anderson MJ, Ellingsen KE, McArdle BH. Multivariate dispersion as a measure
Second, strong taxon-taxon correlations were identified by using of beta diversity. Ecol Lett 2006;9:683-93.
SparCC,E19 where absolute taxon abundances were bootstrapped 100 times E18. Anderson MJ. A new method for non-parametric multivariate analysis of vari-
to generate correlation P values. Networks were then generated from selected ance. Austral Ecol 2001;26:2-46.
E19. Friedman J, Alm EJ. Inferring correlation networks from genomic survey data.
correlations (r > _ 20.25), and P values (P <
_ 0.25 or r < _ .01) were generated by
PLoS Comp Biol 2012;8:e1002687.
using Cytoscape (version 3.4.0). E20. Rogers GB, van der Gast CJ, Serisier DJ. Predominant pathogen competition and
Two approaches were used to investigate the effect of pathogen overgrowth core microbiota divergence in chronic airway infection. ISME J 2015;9:217-25.
on microbiota composition. First, in samples in which Haemophilus or Mor- E21. McMurdie PJ, Holmes S. phyloseq: an R package for reproducible interactive
axella taxa were the dominant taxa and represented 40% or greater of total analysis and graphics of microbiome census data. PLoS One 2013;8:e61217.
reads, their relative abundance was adjusted to the mean value for the study E22. Love MI, Huber W, Anders S. Moderated estimation of fold change and disper-
cohort, and the remaining relative abundance measures were rescaled, as sion for RNA-seq data with DESeq2. Genome Biol 2014;15:550.
FIG E1. Bacterial burden, as assessed based on 16S rRNA gene copy
number. Bars show medians 6 95% CIs. Statistical significance was as-
sessed by using Kruskal-Wallis 1-way ANOVA with the Dunn post hoc
test. No significant difference between phenotypes was seen.
nnn 2017
FIG E2. a-Diversity measures among asthma phenotypes: A, taxa richness; B, Shannon-Wiener index; C,
Simpson evenness index; and D, Pielou evenness. Bars show medians 6 95% CIs. Statistical significance
was assessed by using Kruskal-Wallis 1-way ANOVA with the Dunn post hoc test.
FIG E3. Correlations between sputum neutrophil/eosinophil counts (as a percentage of total cell count) and
a-diversity measures. A, Neutrophil percentage, The dotted line at 61% neutrophils indicates phenotype cut-
off point. B, Eosinophil percentage, The dotted line at 3% eosinophils indicates phenotype cutoff point.
Colors represent asthma phenotype based on neutrophilia or eosinophilia: blue is greater than 61% neutro-
phils, green is greater than 3% eosinophils, yellow is less than 61% neutrophils and less than 3% eosinophils
(paucigranulocytic), and purple is both greater than 61% neutrophils and greater than 3% eosinophils
(mixed). Statistical significance was assessed by using Spearman rank correlation.
nnn 2017
FIG E4. Correlations between sputum neutrophil/eosinophil counts (as a percentage of total cell count) and
weighted UniFrac distance from centroid. A, Neutrophil percentage. The dotted line at 61% neutrophils in-
dicates phenotype cutoff point. B, Eosinophil percentage. The dotted line at 3% eosinophils indicates
phenotype cutoff point. Colors represent asthma phenotype based on neutrophilia or eosinophilia: blue
is greater than 61% neutrophils, green is greater than 3% eosinophils, yellow is less than 61% neutrophils
and less than 3% eosinophils (paucigranulocytic), and purple is both greater than 61% neutrophils and
greater than 3% eosinophils (mixed). Statistical significance was assessed by using Spearman rank
correlation.
FIG E5. A, Principal coordinate analysis of asthma phenotype groups based on Bray-Curtis similarity dis-
tances. The first 2 principal coordinates are plotted on the x- and y-axes, respectively (representing 36.5%
of total variation). B, Microbiota dispersion grouped by asthma phenotype. Distance from centroid was
calculated from the Bray-Curtis dissimilarity matrix. C and D, Correlations between sputum inflammatory
cell percentages and distance from centroid. Fig E5, C, Sputum neutrophil percentage versus Bray-Curtis
distance from centroid. Fig E5, D, Sputum eosinophil percentage vs Bray-Curtis distance from centroid.
nnn 2017
FIG E6. Bacterial network analysis, showing weight and color assigned to
edges and nodes.
FIG E7. Taxa that significantly correlated with eosinophil percentages. Colors represent asthma phenotype
based on neutrophilia or eosinophilia: blue is greater than 61% neutrophils, green is greater than 3%
eosinophils, yellow is less than 61% neutrophils and less than 3% eosinophils (paucigranulocytic), and
purple is both greater than 61% neutrophils and greater than 3% eosinophils (mixed). The dotted line at 3%
eosinophils indicates phenotype cutoff points. Statistical significance was assessed by using Spearman
rank correlation.
nnn 2017
FIG E8. Normalized log2 fold changes of nonrarefied taxa read counts that
significantly (P < .05) differed between neutrophilic and eosinophilic pheno-
types. Positive fold change indicates taxa with significantly higher counts in
neutrophilic participants, and negative fold change indicates taxa with
significantly higher counts in eosinophilic participants. This shows that
when Haemophilus or Moraxella taxa dominance do not influence data
(because of nonrarefied count data as opposed to relative abundance), mul-
tiple taxa remain significantly different between neutrophilic and eosino-
philic participants.
TABLE E1. 16S rRNA sequencing information

Median read count (Q1, Q3) 12,792 (8,060, 16,595)
Subsample depth 1,732

Samples excluded 7
Median Good coverage (Q1, Q3) 0.952 (0.942, 0.963)
Q1, Q3, Quartile 1, quartile 3.
nnn 2017
TABLE E2. PERMANOVA analysis testing: significance of variance of weighted UniFrac and Bray-Curtis distance of sputum
microbiota between asthma phenotypes (permutations 5 9999)
Matrix Source df SS MS Pseudo-F P value (perm)
Weighted UniFrac Phenotype 3 9,444.4 3,148.1 3.9969 .0004

Residual 163 128,390 787.64
Total 166 137,830
Bray-Curtis Phenotype 3 8,215.4 2,738.5 3.3694 .0001
Residual 163 132,480 812.74
Total 166 140,690
TABLE E3. Pairwise PERMANOVA analysis testing: significance of variance of weighted UniFrac and Bray-Curtis distance of
sputum microbiota between asthma phenotypes (permutations 5 9999)
Matrix Groups T P value (perm) Unique perms
Weighted UniFrac Pauci vs Neutro 2.52 <.01 9943

Pauci vs Eosino 1.28 .13 9928
Pauci vs Mixed 0.80 .62 9952
Neutro vs Eosino 3.30 <.0001 9924
Neutro vs Mixed 1.20 .21 9876
Eosino vs Mixed 1.19 .19 9939
Bray-Curtis Neutro vs Eosino 2.89 <.0001 9928
Neutro vs Pauci 2.43 <.001 9918
Eosino vs Pauci 1.15 .17 9921
Eosino, Eosinophilic; Neutro, neutrophilic; Pauci, paucigranulocytic.
nnn 2017
TABLE E4. SIMPER analysis comparing taxa relative abundances between neutrophilic and eosinophilic phenotype groups and
showing the 13 top contributing taxa, which collectively account for approximately 50% of variance between groups
Neutrophilic Eosinophilic
Taxa Av. Abund Av. Abund Av. Diss Contrib %
Haemophilus 0.41 0.2 4.49 8.89

Prevotella 0.23 0.33 2.61 5.16
Streptococcus II 0.13 0.27 2.58 5.11
Streptococcus I 0.45 0.51 2.34 4.63
Veillonella 0.2 0.22 1.98 3.91
Moraxella 0.13 0 1.88 3.72
Neisseria 0.11 0.19 1.65 3.26
Rothia 0.1 0.19 1.52 3
Actinomyces sp. uncultured bacterium 0.1 0.16 1.5 2.96
Gemella 0.09 0.17 1.26 2.49
Leptotrichia 0.08 0.11 1.2 2.38
Actinomyces species oral clone DR002 0.05 0.08 1.1 2.17
Porphyromonas 0.04 0.11 1.1 2.17
TABLE E5. PERMANOVA analysis (top) and pairwise PERMANOVA on nondominant microbiome (battle), showing Bray-Curtis
distance of sputum microbiota on the genera level grouped by asthma phenotype (permutations 5 9999)
Source df SS MS Pseudo-F P value (perm)
Phenotype 3 5,381 1,793.7 2.378 .0004

Residual 163 122,950 754.3
Total 166 128,330
Groups T P value (perm) Unique perms
Neutro vs Eosino 2.31 <.0001 9922

Neutro vs Pauci 2.26 <.001 9911
Eosino vs Pauci 0.91 .65 9904
Eosino, Eosinophilic; Neutro, neutrophilic; Pauci, paucigranulocytic.

Inflammatory Phenotypes in Patients With Severe Asthma Are Associated With Distinct Airway Microbiology

Uploaded by

Copyright:

Available Formats

Inflammatory Phenotypes in Patients With Severe Asthma Are Associated With Distinct Airway Microbiology

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Inflammatory Phenotypes in Patients With Severe Asthma Are Associated With Distinct Airway Microbiology

Uploaded by

Copyright:

Available Formats

Inflammatory phenotypes in patients with severe

asthma are associated with distinct airway

Methods: The microbial composition of induced sputum

TABLE I. Clinical and inflammatory cell parameters of participants

connected nodes in the network analysis, neither differed between

Clinical and inflammatory associations with

Faith’s diversity r 20.374 20.309 20.242 0.193

METHODS pelleted by means of centrifugation at 13,000g for 10 minutes. After removal

TABLE E1. 16S rRNA sequencing information

Subsample depth 1,732

Q1, Q3, Quartile 1, quartile 3.

Weighted UniFrac Phenotype 3 9,444.4 3,148.1 3.9969 .0004

Weighted UniFrac Pauci vs Neutro 2.52 <.01 9943

Haemophilus 0.41 0.2 4.49 8.89

Phenotype 3 5,381 1,793.7 2.378 .0004

Groups T P value (perm) Unique perms

Neutro vs Eosino 2.31 <.0001 9922

You might also like