- Email: [email protected]
Identification of airway mucosal type 2 inflammation by using clinical biomarkers in asthmatic patients
Accepted Manuscript Identification of Airway-Mucosal Type-2 inflammation by Clinical Biomarkers in Asthma Philip E. Silkoff, MD, Michel Laviolette, MD, Dave Singh, MD, J Mark FitzGerald, MD, Steven Kelsen, MD, Vibeke Backer, MD, Celeste M. Porsbjerg, MD, Pierre-Olivier Girodet, MD, Patrick Berger, MD, Joel N. Kline, MD, Geoffrey Chupp, MD, Vedrana S. Susulic, PhD, Elliot S. Barnathan, MD, Frédéric Baribaud, PhD, Matthew J. Loza, PhD PII:
S0091-6749(17)30007-6
DOI:
10.1016/j.jaci.2016.11.038
Reference:
YMAI 12573
To appear in:
Journal of Allergy and Clinical Immunology
Received Date: 25 February 2016 Revised Date:
6 November 2016
Accepted Date: 21 November 2016
Please cite this article as: Silkoff PE, Laviolette M, Singh D, FitzGerald JM, Kelsen S, Backer V, Porsbjerg CM, Girodet P-O, Berger P, Kline JN, Chupp G, Susulic VS, Barnathan ES, Baribaud F, Loza MJ, the Airways Disease Endotyping for Personalized Therapeutics (ADEPT) study investigators, Identification of Airway-Mucosal Type-2 inflammation by Clinical Biomarkers in Asthma, Journal of Allergy and Clinical Immunology (2017), doi: 10.1016/j.jaci.2016.11.038. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Identification of Airway-Mucosal Type-2 inflammation by Clinical Biomarkers in
2
Asthma
3 4 5 6
Philip E Silkoff MD1, Michel Laviolette MD2, Dave Singh MD3, J Mark FitzGerald MD4, Steven Kelsen MD 5 , Vibeke Backer MD 6 , Celeste M Porsbjerg MD 6, Pierre-Olivier Girodet MD 7 , Patrick Berger MD7 , Joel N Kline MD8, Geoffrey Chupp MD 9 , Vedrana S Susulic PhD1, Elli ot S Bar n ath a n M D 1 , Frédéric Baribaud PhD1, and Matthew J Loza PhD 1
7
& the Airways Disease Endotyping for Personalized Therapeutics (ADEPT) study investigators. Janssen Research & Development LLC, Spring House, USA
2
SC
1
Institut Universitaire de Cardiologie et Pneumologie de Québec (IUCPQ), 2725, Chemin Ste-Foy, Québec, Canada, G1V 4G5. Email: [email protected] 3
M AN U
Cen tre f o r R es p i ra tory Med i ci n e an d Al l ergy, the U ni versi ty o f Ma n chester, Medi ci nes E val u ati o n Un i t, Un i v ersi ty Ho s pi tal of Sou th Man ch es ter NHS Fou n da ti on Tru s t, So u thm o or R oa d , Ma nch es ter M2 3 9 Q Z , United Kingdom; Email: [email protected] 4
Institute for Heart and Lung Health, The Lung Centre, 7th Floor, Gordon and Leslie Diamond Health Care Centre, 2775 Laurel Street, Vancouver, B.C., Canada, V5Z 1M9. Email: [email protected] 5
Department of Thoracic Medicine and Surgery, Temple University School of Medicine, 3401 N. Broad St., Philadelphia PA 19140. Email: [email protected] 6
7
TE D
Respiratory Research Unit, Department of Respiratory Medicine, Bispebjerg University Hospital, Bispebjerg bakke 23, DK-2400, Copenhagen NV, Denmark. Email: [email protected], p o r s b j e r g @ d a d l n e t . d k Univ. Bordeaux, Centre de Recherche Cardio-Thoracique de Bordeaux, U1045, CIC 1401, F-33000 Bordeaux, France. Email: [email protected]; [email protected] 8
Division of Pulmonary, Critical Care, and Occupational Medicine, University of Iowa, W219B GH UIHC, 200 Hawkins Drive, Iowa City, IA, 52242, USA. Email: [email protected] 9Yale School of Medicine, TAC 441, 300 Cedar Street, New Haven, CT 06520, USA. Email:
EP
[email protected]
Corresponding Author: Philip E Silkoff 715 Bryn Mawr Avenue Penn Valley, PA, 19072, USA Tel: 6103109142; Fax 6106671698 Email: [email protected]
AC C
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
RI PT
1
The study was funded by Janssen R&D Inc. Spring House, PA, USA Author contributions: Conception and design: MJL, VSS, ESB, and PES Acquisition of data: IS, ML, DS, MF, SL, SK, AE, AL-S, GC, VB, CP, P-OG, PB, RL, JNK, MD, WJC, AH, SK, PC (the ADEPT investigators) Analysis and interpretation: All authors 1
ACCEPTED MANUSCRIPT
Drafting the manuscript for important intellectual content: MSL, and PES Approval of the final version: All authors Funding disclosure: Support was provided by Janssen Research & Development, LLC
44
Short Running Head: Clinical biomarkers and Type2 Endotyping in Asthma
45
Clinicaltrials.gov identifier: NCT01274507
AC C
EP
TE D
M AN U
SC
46
RI PT
41 42 43
2
ACCEPTED MANUSCRIPT
Abstract
48
Background and Objective: The Airways Disease Endotyping for Personalized Therapeutics
49
(ADEPT) study profiled mild, moderate and severe asthma, and non-atopic healthy controls. We
50
explored this dataset to define Type-2 inflammation based on airway-mucosal IL-13-driven gene
51
expression and how this related to clinically-accessible biomarkers.
52
Methods: IL-13-driven gene expression was evaluated in several human cell lines. We then
53
defined Type-2 status in 25 healthy subjects, 28 mild, 29 moderate, and 26 severe asthmatics,
54
based on airway-mucosal expression of 1) CC-motif chemokine ligand (CCL)-26, (the most
55
differentially expressed gene), 2) periostin, or 3) a multi-gene IL-13 in-vitro signature (IVS).
56
Clinically accessible biomarkers included fractional exhaled nitric oxide (FENO), blood
57
eosinophils (bEOS), serum CCL26, and serum CCL17.
58
Results: Expression of airway-mucosal-CCL26, periostin, and IL-13-IVS all provided segregation
59
into Type-2-high and -low asthmatics, but in the ADEPT population, CCL26 was optimal. All
60
airway-mucosal-CCL26-high subjects with moderate-severe asthma were FENO-high (≥35 ppb)
61
and/or blood eosinophils-high (≥300cells/mm3), compared to a minority (36%) of airway-
62
mucosal- CCL26-low subjects. A combination of FENO, blood eosinophils, serum CCL17 and
63
CCL26 had 100% positive-predictive-value and 87% negative-predictive-value for airway-
64
mucosal-CCL26-high status. Clinical variables did not differ between Type-2 high and –low
65
subjects. Eosinophilic inflammation was associated with, but not limited to, airway-mucosal
66
Type-2 gene expression.
67
Conclusion: A panel of clinical biomarkers accurately classified Type-2 status based on airway-
68
mucosa CCL26, perisotin or IL-13-IVS gene expression. Use of FENO, blood eosinophils and
AC C
EP
TE D
M AN U
SC
RI PT
47
3
ACCEPTED MANUSCRIPT
serum markers e.g. CCL26, CCL17 in combination may allow patient selection for novel Type-2
70
therapeutics.
71
Abstract Word count: 241
72
Key Messages:
RI PT
69
•
Type-2 inflammation phenotype has emerged as the target of novel asthma therapies.
74
•
Airway-mucosal gene expression of CCL26, periostin or a multigene IL-13 in vitro
SC
73
signature were used to segregate Type-2 high from Type-2 low subjects. In our dataset,
76
airway-mucosal-CCL26 was optimal.
77
•
M AN U
75
Combinations of clinical biomarkers, FENO, blood eosinophils, serum CCL17 and CCL26 optimally identified Type-2 inflammation based on airway-mucosal CCL26 with similar
79
findings for periostin and IL-13-IVS.
TE D
78
Capsule summary
81
A panel of clinically-accessible biomarkers (FENO, blood eosinophils, serum CCL17 and CCL26)
82
identified asthma patients with an airway ‘Type-2-high’ or ‘Type-2-low’ phenotype based on
83
expression of airway-mucosal-CCL26, periostin, or an IL-13 multigene signature. This
84
combination may prove useful in selecting patients for novel therapies targeting Type-2
85
inflammation.
86
Key words: asthma, Type-2 inflammation, phenotypes, airway-mucosal gene expression, biomarkers
AC C
EP
80
87 4
ACCEPTED MANUSCRIPT
List of Abbreviations Asthma Control Questionnaire 7
ADEPT
Airways Disease Endotyping for Personalized Therapeutics
AHR
airway hyper-responsiveness
AQLQ
Asthma Quality of Life Questionnaire
bEOS
blood eosinophils
BDR
bronchodilator reversibility
CLCA1
calcium-activated chloride channel regulator 1
CCL
C-C motif chemokine ligand
FDR
false discovery rate
FENO
Fractional concentration of exhaled nitric oxide
FEV1
forced expired volume in 1 second
GSVA
gene set variation analysis
ICS
inhaled corticosteroids
Ig
immunoglobulin
IL
Interleukin
IL-4R
interleukin 4 receptor
IVS
in-vitro signature
mAb
monoclonal antibody
NPV
negative predictive values
PC20
POSTN ppb
EP
TE D
M AN U
SC
RI PT
ACQ7
AC C
88
The provocative concentration of methacholine resulting in a 20% or greater fall in the forced expired volume in 1 second periostin
parts per billion
PPV
positive predictive value
Pre-BD
pre-bronchodilator
RNA
ribonucleic acid
5
ACCEPTED MANUSCRIPT
s
serum
SD
standard deviation
SERPINB2
Serpin peptidase inhibitor, clade B, member 2 (also known as plasminogen activator inhibitor-2)
sIgE
serum immunoglobulin E
spEOS
sputum eosinophils
spNEU
sputum neutrophils
TGF
transforming growth factor
TNF
tumor necrosis factor
AC C
EP
TE D
M AN U
SC
RI PT
89
6
ACCEPTED MANUSCRIPT
INTRODUCTION
91
Asthma is a heterogeneous disease, characterized by chronic airway inflammation 1. While
92
there are many proposed asthma phenotypes, the underlying phenotypes are poorly
93
understood, hard to identify, and tailored anti-asthma treatment an even further challenge.
94
Type-2 inflammation in asthma (also termed T-helper 2 inflammation) refers to that driven by
95
Type-2 cytokines e.g. IL-4, IL-5 and IL-13 secreted by CD4+ T helper 2 cells as recently reviewed
96
2
97
and multiple gene signatures in airway brushings and sputum inflammatory cells have recently
98
been used to define Type-2-high or –low status 4-6.
99
The most effective therapies for asthma are anti-inflammatory, including inhaled
SC
RI PT
90
M AN U
. Type-2 innate lymphoid cells have also been proposed to secrete the same cytokines 3. Single
corticosteroids (ICS), and monoclonal antibodies (mAbs), but these do not work in all patients.
101
For example, the response to ICS is associated with a high fractional exhaled nitric oxide (FENO)
102
7
103
5
104
(anti-IL-4 receptor) 10, mepolizumab and reslizumab (anti-IL-5) 11, 12, or lebrikizumab (anti-IL-13)
105
13
106
biomarkers including blood eosinophils (bEOS), spEOS, FENO, or serum periostin9-13.
107
Notably, despite these patient-selection strategies, severe asthma is often incompletely
108
responsive to these and other currently-available therapies 14, 15. This may be due to difficulty
109
in defining phenotypes including the Type-2 phenotype leading to selection of inappropriate
110
therapy, or additional mechanisms, alone or in addition to Type-2 inflammation, that are
TE D
100
, high sputum eosinophils (spEOS) 8 or a 3-gene ‘Type-2-high’ signature in epithelial brushings
EP
. Similarly, for the following mAbs, omalizumab (anti-immunoglobulin E (IgE), 9, dupilumab
AC C
, optimal efficacy is seen in severe asthma with Type-2 inflammation identified with
7
ACCEPTED MANUSCRIPT
driving pathology. For example, in a recent report for lebrikizumab, even periostin-high subjects
112
had a 95% CL range for improvement in FEV1 between 1.0-15.4% 13 illustrating that single
113
biomarkers cannot capture complex biological networks.
114
The previously reported ADEPT study 16, allowed us to evaluate airway-mucosal gene
115
expression associated with Type-2 inflammation, including periostin, which has been associated
116
with enhanced responses to the anti-IL-13 mAb, lebrikizumab 13, CCL26, which in our hands was
117
the most highly expressed gene in IL-13 treated human cells lines, and a multigene in vitro
118
signature from the same cell lines (IL-13-IVS). We also ascertained if Type-2 gene expression
119
profiles for CCL26, IL-13-IVS and periostin could be classified using clinically accessible
120
biomarkers, namely FENO, bEOS, and less-commonly used serum markers, namely CCL17/TARC,
121
and CCL26/eotaxin-3, which are Type-2 chemokines reliably measured in serum.
SC
M AN U
TE D EP AC C
122
RI PT
111
8
ACCEPTED MANUSCRIPT
Methods
124
Study design and population
125
The ADEPT study design and population (clinicaltrials.gov registration number NCT01274507)
126
have been described in detail
127
medications, with a pre-bronchodilator [pre-BD] forced expired volume in 1 second [FEV1]>80%
128
predicted), moderate (low-moderate dose ICS, pre-BD FEV1 60-<80% predicted), and severe
129
(high-dose ICS, pre-BD FEV1 50 -<80% predicted) asthma subjects, were enrolled. Healthy
130
subjects and a subset of asthmatic subjects underwent bronchoscopy. All ADEPT subjects with
131
good quality airway-mucosal tissue from biopsies were included in the biopsy assessment
132
details below. Clinical assessments included spirometry, bronchodilator reversibility (BDR),
133
methacholine airway hyper-responsiveness (AHR), the Asthma Control Questionnaire (ACQ7) 17,
134
and the Asthma Quality of Life Questionnaire (AQLQ)
135
specific institutional review boards and all subjects signed an informed consent form. Healthy
136
and asthmatic subjects attended screening, baseline and bronchoscopy visits, while asthmatic
137
subjects also attended biomarker visits at 3, 6 and 12 months.
138
Biomarker assessments
139
Bioanalysis methods are detailed in the Online Supplement section E1. Briefly, gene expression
140
in RNAlater®-preserved biopsy samples was analyzed by microarray (HG-U133+PM platform,
141
Affymetrix, Santa Clara, CA) and histology of biopsies was performed by Pantomics, Inc.
142
(Richmond, CA). Induced sputum was collected, processed via the plug selection method 19, and
RI PT
123
16
18
. The study was approved by site-
AC C
EP
TE D
M AN U
SC
. In brief, non-atopic healthy subjects, mild (no controller
9
ACCEPTED MANUSCRIPT
analyzed for differential cell counts. Biomarker analyses focused on molecular Type-2-activity
144
status in biopsies as they relate to asthma characteristics, and clinical biomarkers.
145
Definition of Type-2 status in biopsies
146
‘Type-2-high’ vs. ‘Type-2-low’ phenotypes were defined a priori based on inferred IL-13 activity
147
in airway-mucosa biopsies from asthma subjects compared to healthy controls using gene
148
expression of CCL26, periostin or a multigene in-house IL-13-IVS (see Online Supplement
149
section E1 and Table E1). IL-13 activity was selected as it is the most broadly-expressed Type-2
150
cytokine produced by Th2 cells 20 and also Type-2 innate lymphoid cells 21. CCL26 mediates
151
eosinophil infiltration into the airway and is the most highly-expressed selective IL-13 driven
152
gene 22-24. The chemokine CCL17 is induced by IL-13 24, 25 and is decreased by anti-IL-13 and
153
anti-IL-4R mAb therapeutics 10, 13, 26. For CCL26, the highest signals in the healthy controls were
154
just below the limit of reliable quantification for the microarray, so this limit of quantification
155
(log2 intensity of 5.0) was set as the threshold for airway-CCL26-high versus -low status. For
156
periostin and IL-13-IVS, Type-2-high status was defined as gene expression (or enrichment)
157
beyond the 95th percentile of the healthy controls.
158
Clinical biomarkers evaluated for their association to Type-2 high or low status by airway IL-13-
159
driven gene expression included FENO, bEOS, spEOS, serum IgE, and 2 serum Type-2-associated
160
biomarkers (serum (s)CCL17 and sCCL26). Details of the assays can be found in the Online
161
Supplement Section E1. Serum CCL17 and sCCL26 were measured repeatedly over 12 months
162
to assess stability.
163
Statistical analysis
AC C
EP
TE D
M AN U
SC
RI PT
143
10
ACCEPTED MANUSCRIPT
Healthy subjects (n=25) and non-ICS-treated mild asthma subjects (n=28) were analyzed
165
separately, while ICS-treated moderate (n=29) and severe (n=26) asthma cohorts were pooled.
166
We focused analyses on the moderate-severe asthmatics as they have unmet clinical need
167
despite ICS treatment. Separate analyses on the moderate and severe cohorts were not
168
undertaken due to limited statistical power.
169
Statistical analyses used OmicSoft ArrayStudio v7 (Cary, NC; www.omicsoft.com). For data with
170
log-normal distributions, logarithmic transformations were performed. Group comparisons of
171
gene and protein expression levels (log2 –transformed) were performed using General Linear
172
Models. For variables with log-normal distributions, geometric mean and asymmetric standard
173
deviations were estimated from the log2-scaled mean and standard deviation: geometric mean
174
= 2^(mean); asymmetric upper standard deviation = 2^(mean + SD) – 2^(mean); asymmetric
175
lower standard deviation = 2^(mean – SD) – 2^(mean). Significance of differences in
176
proportions between 2 categorical variables was tested using Fisher’s exact test. Nominal p-
177
values are reported, with multiple-testing adjusted FDR values below 0.05 indicated.
178
Classification model statistics were obtained from logistic regression and receiver operating
179
characteristic analyses (NCSS v8, www.NCSS.com).
180
Box-and-whisker plot representations of distributions in figures show the median and
181
interquartile range (box), minimum and maximum range (whiskers), mean (‘+’ symbol), and
182
diamond symbols the values for each individual subject/sample.
AC C
EP
TE D
M AN U
SC
RI PT
164
183
184
Results 11
ACCEPTED MANUSCRIPT
Airway-mucosal CCL26, periostin, and IL-13-IVS gene expression compared for classification of
186
Type-2 inflammation
187
Figure 1 displays segregation into high and low Type-2 status of asthmatic from healthy
188
subjects based on airway-mucosal gene expression of CCL26 (panel A), periostin (panel B), or IL-
189
13-IVS enrichment (panel C). Airway-mucosal-CCL26-high asthmatics were mostly high for
190
airway periostin (90%) and airway IL-13-IVS (83%) expression (all asthmatic subjects combined),
191
a finding unimpacted by asthma severity (Table 1; Online Supplement Figure E1). Lower
192
proportions of airway periostin-high asthmatics were high for airway CCL26 (62%) or airway IL-
193
13-IVS (67%) expression for all asthmatic subjects combined, but the concurrence was higher in
194
mild compared to moderate-severe asthma (Table 1; Online Supplement Figure E1). Similarly,
195
lower proportions of airway IL-13-IVS high asthmatics were high for airway CCL26 (62%) or
196
airway periostin (67%) expression for all asthmatic subjects, with higher concurrence in mild
197
compared to moderate-severe asthmatic subjects. (Table 1; Online Supplement Figure E1).
198
These observations drove selection of airway-mucosal-CCL26 expression as the primary anchor
199
for Type-2 status, although periostin and IL-13-IVS also demonstrated utility for segregation of
200
asthma from healthy control subjects as discussed in Online Supplement Section E2.
201
Demographics by airway-mucosal-CCL26 status
202
Table 2 presents demographic characteristics for mild and moderate asthma by airway-mucosal
203
-CCL26-high and –low status. There were no significant differences by CCL26 status. Similar
204
proportions of CCL26 -high (3/13=23%) and -low (7/42=17%) status subjects used leukotriene
205
inhibitors (data now shown).
AC C
EP
TE D
M AN U
SC
RI PT
185
12
ACCEPTED MANUSCRIPT
Asthma disease characteristics and clinical biomarkers by airway-CCL26 status
207
There were decreasing proportions of airway-CCL26-high subjects in the mild (N=16/28, 57%),
208
moderate (N=8/29, 28%) and severe (N=5/26, 19%) asthma cohorts.
209
Disease characteristics including airflow obstruction were not associated with airway-CCL26
210
status (Table 3) except that CCL26-high status was associated with nominally significantly
211
increased AHR in mild asthma (p=0.045). In contrast, airway-mucosal-CCL26 status was strongly
212
associated with multiple Type-2 clinical biomarkers. Thus, FENO (all asthma severities), and
213
bEOS (mild-moderate asthma), were significantly elevated (p<0.05) in the airway-mucosal-
214
CCL26-high compared to the -low group (Table 3, and Figure 2 A, B). Serum IgE was highly
215
elevated in the airway-mucosal-CCL26-high moderate-severe asthma group (N=13) compared
216
with healthy controls (FDR<0.0001, p<0.0001, fold=24) and the airway-mucosal-CCL26-low
217
group (n=41, FDR=0.091, p=0.018, fold=2.8), though sIgE was also elevated in the airway-
218
mucosal-CCL26-low group compared to healthy controls (FDR<0.0001, p<0.0001, fold=8.6)
219
(Figure 3A). Serum CCL17 levels were significantly elevated in airway-mucosal-CCL26-high
220
compared with the -CCL26-low group (p = 0.0039, 2.17-fold/Type-2-low) and with healthy
221
controls (p = 0.018, 2.02-fold/healthy controls) for moderate-severe but not mild asthma
222
subjects (p=0.29 and 0.90, respectively; see Figure 3B). Similarly, serum CCL26 levels were
223
significantly elevated in airway-mucosal-CCL26-high compared with the -CCL26-low group (p =
224
0.0027, 7.35-fold/Type-2-low) and compared with healthy controls (p = 0.029, 4.67-fold/healthy
225
controls) for moderate-severe but not mild asthma subjects (p=0.21 and 0.67, respectively; see
226
Figure 3C).
AC C
EP
TE D
M AN U
SC
RI PT
206
13
ACCEPTED MANUSCRIPT
Categorical analysis of clinical biomarkers by airway-mucosal CCL26 status
228
Airway-mucosal-CCL26 status was compared with clinical biomarkers categorized using the
229
following cut-offs: ≥35ppb for FENO, a previously described cutoff 27-30; ≥300 cells/mm3 for
230
bEOS, a cutoff used for novel “Type-2/eosinophilic” therapeutics 31; and by the 95th percentile
231
of the healthy control cohort for sIgE (Table 4). Serum CCL17-high and sCCL26-high status was
232
based on the 75th percentile of the healthy controls rather than the 95th centile because of the
233
wide distributions in healthy controls.
234
A majority of airway-CCL26-high moderate-severe asthma subjects were FENO-high (69%),
235
compared to 24% of airway-CCL26-low subjects (p=0.0063). Similarly, 77% airway-CCL26 high
236
subjects were bEOS-high, compared to 24% of airway-CCL26-low subjects (p=0.0009). All
237
airway-CCL26-high subjects were FENO-high or bEOS-high but only 55% were FENO-high and
238
bEOS-high. In contrast, for airway-CCL26-low subjects, 36% were FENO-high or bEOS-high but
239
few (12%) were FENO-high and bEOS-high (p<0.0001 and 0.014, respectively), for airway-CCL26
240
high vs. low. Of note, 92% of airway-CCL26-high subjects were sIgE-high, but 71% of airway-
241
CCL26-low subjects were also sIgE-high (p=0.15). For moderate-severe asthma, 11/21 (52%)
242
were sCCL17-high but only 2/34 (6%) CCL17-low asthmatics were airway-CCL26 high.
243
Additionally, 7/11 (64%) were sCCL26-high but only 5/42 (12%) sCCL26-low asthmatics were
244
airway-CCL26-high.
245
Similar associations for mild asthma (see Table 3 and Online Supplement Section E2) were
246
observed as for moderate-severe asthma except that both the airway-CCL26-high and -low
247
groups were mostly low for bEOS in mild asthma (75% and 92% of respective groups).
AC C
EP
TE D
M AN U
SC
RI PT
227
14
ACCEPTED MANUSCRIPT
In general, the patterns of categorical clinical biomarkers reported for airway-mucosal-CCL26
249
expression were similar for periostin (Online Supplement Table E2) and IL-13-IVS expression
250
(Online Supplement Table E3), although the between-group differences were less pronounced.
251
For example, when defining airway-mucosal-Type-2-high status by CCL26, periostin, and IL-13-
252
IVS, 100%, 80%, and 75% of airway-Type-2-high moderate-severe asthmatics were FENO-high
253
or bEOS-high, but similar proportions of airway-Type-2-low moderate-severe asthmatics were
254
FENO-high or bEOS-high (36%, 34%, and 32%) respectively. These observations reinforce our
255
selection of airway-mucosal-CCL26 expression as the primary anchor for Type-2 status.
256
Airway eosinophilia by airway-CCL26 status
257
Thirty-nine moderate-severe asthmatic subjects had acceptable sputum quality, 8 of whom
258
were airway-CCL26-high and comparisons were made with commonly-used cutoffs, namely,
259
sputum eosinophils (spEOS) ≥3% and sputum neutrophils (spPMN) ≥60% status (see Table 4
260
and Figure 2C). Most (88%) of airway-CCL26-high subjects but also 45% of airway-mucosal-
261
CCL26-low subjects had spEOS≥3% (p=0.049). Those with neutrophilic or paucigranulocytic
262
patterns were predominantly airway-mucosal-CCL26-low (86% and 100%, respectively).
263
Submucosal eosinophil intensity (histology) was available for 45 moderate-severe asthma
264
subjects, 13 of whom were airway-mucosal-CCL26-high. Eosinophil intensity was significantly
265
higher in airway-mucosal-CCL26-high compared to –low (p=0.025) moderate-severe asthmatic
266
subjects and to healthy controls (p=0.0039), with similar associations for mild asthma (p=0.024
267
and 0.0002, respectively) (Figure 2D).
268
Classification of airway-mucosal-CCL26 Type-2 status using clinical biomarkers
AC C
EP
TE D
M AN U
SC
RI PT
248
15
ACCEPTED MANUSCRIPT
Pre-specified thresholds for high/low status of FENO, bEOS, sCCL17, and sCCL26 were employed
270
to test classification of airway-CCL26-high status (Table 5) (optimized thresholds from modeling
271
reported in Online Supplement Table E8). Positive predictive values (PPV) and negative
272
predictive values (NPV) were estimated based on the prevalences observed in the ADEPT study
273
population. The actual predictive values for a given population would be dependent on the
274
prevalence of CCL26-high status in that population.
275
FENO-high, bEOS-high, and the criterion ‘FENO-high OR bEOS-high’ all had poor PPV of ≤50%
276
but good NPV (≥89%), with the ‘FENO-high OR bEOS-high’ classification having the best NPV
277
(100%). The criteria of ‘FENO-high AND bEOS-high’ had only marginally better PPV (55%) but
278
with reduced NPV (84%). Serum CCL17-high or sCCL26-high gave modestly improved PPV (62%
279
and 64%, respectively) while maintaining a good NPV (92% and 88%, respectively).
280
The combination of ‘FENO-high OR bEOS-high’ AND sCCL17-high modestly improved PPV (66%)
281
while maintaining high NPV (97%). The criteria ‘FENO-high OR bEOS-high’ AND sCCL26-high
282
resulted in much-improved PPV (88%) while maintaining good NPV (89%). Finally, the
283
combination of sCCL17-high AND sCCL26-high AND ‘FENO-high OR bEOS-high’ criteria resulted
284
in the best PPV (100%) while maintaining good NPV (87%). Therefore, depending on the
285
application, one of these 3 combination models would be optimal, depending on whether high
286
NPV, high PPV, or a balance of NPV and PPV is required.
AC C
EP
TE D
M AN U
SC
RI PT
269
287
288
16
ACCEPTED MANUSCRIPT
289
Discussion
291
Based on the ADEPT dataset, we report that airway-mucosal expression of CCL26 was a robust
292
discriminator of Type-2 inflammation from healthy non-atopic subjects. Furthermore, airway-
293
mucosal-CCL26 expression was best identified using a panel of clinical biomarkers including
294
FENO, bEOS, and 2 novel markers, sCCL17 and sCCL26.
295
Airway-mucosal-CCL26, compared to periostin and IL-13-IVS, expression provided the most
296
robust segregation for airway-mucosa Type-2 activity status in our dataset. The airway-mucosa
297
CCL26 Type-2 phenotype also had the best concordance with airway eosinophilic inflammation,
298
including 100% of airway-mucosa-CCL26-high moderate-severe asthmatics being ‘FENO-high or
299
bEOS-high’ and 87.5% being sputum eosinophil-high, compared to using periostin or IL-13-IVS
300
as anchors. Similar low proportions of Type-2-low status moderate-severe asthmatics
301
regardless of the airway-mucosa-Type-2 anchor tested were high for eosinophilic inflammation.
302
Taken together, these observations affirm our preference for airway-mucosa-CCL26 expression
303
as the anchor for Type-2 status. Similar findings pertain to the mild asthmatic subjects.
304
Notwithstanding this preference, others have confirmed the substantial utility of periostin, and
305
other genes including calcium-activated chloride channel regulator 1 (CLCA1) and SERPINB2
306
(neither of which had quantifiable expression levels in the ADEPT airway mucosa), as anchors
307
for Type-2 status 4, 32, which are not inconsistent with the results presented in this report.
AC C
EP
TE D
M AN U
SC
RI PT
290
17
ACCEPTED MANUSCRIPT
Both CCL26 and periostin expression are induced by IL-13 in multiple cell types present in
309
endobronchial biopsies, including epithelial cells, airway smooth muscle cells, and fibroblasts,
310
as well as macrophages for CCL26, making them ideally suited for evaluation in biopsies (23, 11,
311
24, 33
312
CCL11 and CCL24 34,and in one report, IL-13 induced greater expression of CCL26 in bronchial
313
epithelial cells and higher supernatant CCL26 levels from severe compared to mild asthma
314
suggesting a role for CCL26 in the sustained inflammation in severe eosinophilic asthma 22. The
315
less-consistent performance of periostin may be a consequence of transforming growth factor
316
(TGF)-β being a major regulator in addition to IL-4/IL-13 35, 36. The lesser performance of the IL-
317
13-IVS may be related to the inclusion of genes not specific to IL-13 activity, compared to
318
specificity of CCL26 per se.
319
The majority of mild, but only a minority of moderate-severe asthmatics, was airway-mucosal-
320
CCL26-high, likely related to ICS treatment in the latter. Persistence of Type-2-high status
321
despite ICS treatment could be driven by steroid non-responsiveness (or non-adherence),
322
supported by the significant up-regulation of the steroid-response gene FKBP5 in Type-2-low,
323
but not Type-2-high, moderate-severe asthmatics (data not shown). Consistent with this
324
contention, increased CCL26 protein expression has been found in bronchial epithelium of
325
subjects with severe asthma selected for persistent high spEOS counts despite high doses of
326
corticosteroid 22.
327
In general, Type-2 status was not associated with clinical asthma characteristics, perhaps
328
related to the milder asthma population in ADEPT with few refractory asthmatic subjects. In
RI PT
308
AC C
EP
TE D
M AN U
SC
. CCL26 is a more effective chemoattractant for eosinophils than other chemokines e.g.
18
ACCEPTED MANUSCRIPT
contrast, there were marked differences between airway-mucosal-CCL26-high and -low status
330
for eosinophilia-associated clinical biomarkers, including FENO, bEOS, and spEOS, consistent
331
with the biology of CCL26, a potent eosinophil chemoattractant in asthmatics 34. However, only
332
about a third of sputum ‘eosinophilic’ moderate-severe asthmatics were airway-mucosal-
333
CCL26-high (Table 4). This finding is consistent with a previous report by Choy et al. of only a
334
fraction of sputum ‘eosinophilic’ mild asthmatics having a ‘Th2’ phenotype (evaluated by
335
periostin, SERPINB2, and CLCA1 expression in endobronchial brushings) 37. Type-2 activity,
336
evidenced by enriched IL-4/IL-13 downstream activity, would also be expected to coincide with
337
IL-5 activity, which leads to generation of eosinophils. This is consistent with airway-mucosal-
338
CCL26-high asthmatics having elevated airway eosinophilic inflammation in both airway
339
submucosa and sputum. Of note, eosinophilia in blood, sputum and airway wall, and FENO may
340
be discordant from each other perhaps due different mechanisms affecting these distinct
341
compartments e.g., IL-5 driving blood eosinophils and IL-4/IL-13 elevating FENO by local action
342
on the airway mucosa 38 and bEOS and FENO associate independently with wheezing episodes
343
39
344
reflect the transient nature of Type-2-high status that was inhibited by ICS or had
345
spontaneously resolved. However, despite CCL26 being a potent inducer of airway eosinophilia
346
34
347
e.g. from CCL11 (eotaxin-1) induced by non-Type-2 mediators, such as tumor necrosis factor
348
(TNF)-α and IL-1β 40-42.
349
Serum CCL17, an IL-13-induced chemoattractant for Th2 cells 25, 43 that is reduced by anti-IL-13
350
and anti-IL-4R therapeutics 10, 13, 26, was significantly higher in airway-mucosal-CCL26 high
TE D
M AN U
SC
RI PT
329
AC C
EP
. The presence of eosinophilia as well as atopy/elevated IgE with Type-2-low status could
, recruitment of eosinophils to the airways could also occur independently of Type-2 activity,
19
ACCEPTED MANUSCRIPT
moderate-severe asthmatics, despite ICS treatment. Serum CCL26 was also significantly
352
elevated in airway-mucosal-CCL26 high moderate-severe asthmatics. CCL26 could come from
353
leakage from the airways, where transcriptional activity defines the airway-mucosal-CCL26-high
354
phenotype. Larose et al. 22 showed a correlation between CCL26 levels and spEOS counts and
355
suggested that bronchial epithelial cells are the dominant source of airway CCL26 that drives
356
eosinophils into the airway lumen. Accordingly, CCL26 gene expression in endobronchial
357
brushings samples from ADEPT was highly elevated selectively in the airway-mucosa-CCL26-
358
high asthma group (data not shown). In one report, CCL26 gene expression in epithelial
359
brushings was also highest in severe asthma, associated with higher sputum eosinophils, lower
360
FEV1, and more frequent exacerbations 44.
361
Knowing whether eosinophilic inflammation, in the absence of Type-2-high status, is pathogenic
362
will be important in determining which asthma subpopulations would benefit more from IL-
363
4/IL-13 vs. IL-5 targeted therapeutics. The efficacy of IL-4/IL-13-targeted therapeutics to impact
364
eosinophilic inflammation is suggested by their capacity to reduce FENO as reported for
365
lebrikizumab 13, although FENO is not an absolute indicator of eosinophilic activity. Impact on
366
spEOS by IL-4/IL-13-targeted therapeutics has not been reported. The efficacy of IL-5 targeted
367
therapeutics in reducing Type-2-associated pathology is unclear. Evidence of reductions in sIgE
368
and CCL17, as observed for the IL-4/IL-13 targeted therapeutics, would provide the clearest
369
evidence for such, but has not been reported.
370
As the majority of effective anti-inflammatory agents work best in a Type-2-high population e.g.
371
ICS 4, and anti-IL-13 antibodies 13, defining Type-2 status in patients using clinically accessible
AC C
EP
TE D
M AN U
SC
RI PT
351
20
ACCEPTED MANUSCRIPT
biomarker will be important. Identification of airway-mucosa-CCL26-high subjects in moderate-
373
severe asthma was best achieved based on combinations of FENO-high, blood EOS-high,
374
sCCL17-high, and sCCL26-high criteria. Classification based on ‘FENO-high or bEOS-high’ could
375
be readily implemented in screening of asthma patients as these biomarkers can be routinely
376
measured at present. Although having perfect NPV for classifying airway-mucosal-CCL26-high
377
status, the PPV (47%) was poor, limiting utility of this classification for a co-diagnostic test for
378
Type-2-high phenotype. However, the classification could still be useful for enriching a study
379
population for a Type-2-low phenotype.
380
Adding sCCL17-high to the classification with ‘FENO-high or bEOS-high’ maintained an optimal
381
NPV (97%) while moderately improving the PPV to 66%. This classification has the potential to
382
be implemented as a co-diagnostic tool to triage patients likely to not benefit from of Type-2-
383
targeting therapeutic while moderately enriching for those who would likely benefit.
384
Adding sCCL26-high to the classification with ‘FENO-high or bEOS-high’ provided more of a
385
balance between NPV (89%) and PPV (88%). The full model including ‘FENO-high or bEOS-high’,
386
sCCL17-high, and sCCL26-high criteria had the best PPV (100%) with a modestly reduced NPV of
387
87%. These last 2 classification models may be useful to prospectively select patients for Type-
388
2-targeted therapeutic clinical studies, highly enriching for patients more likely to respond.
389
However, these models have low sensitivities that would result in about half of Type-2-high
390
patients being called Type-2-low, which may be tolerable for proof-of-concept studies but not
391
for application as a co-diagnostic test.
AC C
EP
TE D
M AN U
SC
RI PT
372
21
ACCEPTED MANUSCRIPT
These classification models will need to be confirmed in independent study populations, with
393
ultimate validation coming from interventional clinical trials. The generalizability of our findings
394
is unknown as the ADEPT study did not include smokers, subjects with high BMI, and those
395
requiring chronic oral corticosteroids. Because of the limited sample size of the Type-2-high
396
asthma subpopulation, training-test set confirmation approaches could not be employed. To
397
limit risks of over-modeling, thresholds for categorization of the accessible biomarkers were
398
pre-defined before modeling, although the optimal modeled thresholds performed similarly. In
399
post hoc analyses, logistic regression modeling with cross-validation (in full study population,
400
without randomized training and test sets) resulted in an optimal model where the weights for
401
each categorical predictor (‘FENO or bEOS’, sCCL17, sCCL26) or continuous values for each
402
predictor resulting in comparable model performance compared to the simple Boolean models
403
presented in Table 5, with stronger performance with discrete predictors compared to the
404
respective continuous values (Table E9 in Online Repository). The ultimate cut-offs for the
405
predictors would need to be retrospectively established from interventional clinical response
406
data, with current thresholds providing for an enrichment to increase probability of success for
407
a therapeutic in the absence of available clinical response data. The stability of the airway-
408
mucosal Type-2 phenotype over time has not been directly evaluated. The consistency of
409
sCCL26 levels reported here suggests that the airway-mucosal Type-2 phenotype is a stable
410
phenotype.
411
Limitations of this study include the absence of smoking, oral-corticosteroid dependent, and
412
morbidly obese asthmatic subjects as described previously 16. Furthermore, we did not recruit
413
an atopic healthy subject population that might have provided an additional group for
AC C
EP
TE D
M AN U
SC
RI PT
392
22
ACCEPTED MANUSCRIPT
comparison purposes. Finally, we did not have a reliable serum periostin assay available to
415
enable comparison of this to serum CCL17 and serum CCL26. However, as reported by Choy et
416
al., serum periostin was not significantly higher in mild asthmatics with an airway-mucosal ‘Th2’
417
phenotype than those with a ‘non-Th2’ phenotype, but rather elevated only in those with an
418
airway-mucosal ‘Th17’ phenotype 37.
419
In conclusion, airway-mucosal-CCL26 gene expression, representing airway-mucosal IL-13
420
activity, was used to optimally define an airway-mucosal Type-2-high phenotype, which is a
421
subgroup of a broader eosinophilic phenotype. Combinations of the clinically accessible
422
biomarkers FENO, bEOS, sCCL17, and sCCL26 best identified the airway-mucosal-CCL26-high
423
phenotype and promises to help select patients for Type-2 inflammation-targeted therapies.
424
Type-2-low asthma, a group with high unmet need and few therapeutic options, can be reliably
425
identified using the same biomarkers.
SC
M AN U
TE D EP AC C
426
RI PT
414
23
ACCEPTED MANUSCRIPT
Acknowledgments
428
The ADEPT investigators
429 430
I Strambu1, S Lam 2, A Eich3, A Ludwig-Sengpiel 4, R Leigh5, M Dransfield6, W Calhoun7, A Hussaini8, a n d P C hanez 9
1Arensia Exploratory Medicine, Sos. Viilor 90, Bucharest 050159, Romania. Email: [email protected]
2Institute for Heart and Lung Health, The Lung Centre, 7th Floor, Gordon and Leslie Diamond Health Care Centre, 2775 Laurel Street, Vancouver, B.C., Canada, V5Z 1M9. Email: [email protected] 3IKF Pneumologie Frankfurt, Institut für klinische Forschung Pneumologie, Clinical Research Centre Respiratory
SC
Diseases, Schaumainkai 101-103, Stresemannallee 360596, Frankfurt, Germany. Email: [email protected] 4KLB Gesundheitsforschung Lübeck GmbH, Sandstr. 18, 23552 Lübeck, Germany. Email: [email protected]
5Cumming Scholl of Medicine, University of Calgary, 3280 Hospital Drive NW, Calgary, AB T2N 4Z6, Canada. Email:
M AN U
[email protected]
6Division of Pulmonary, Allergy and Critical Care Medicine, University of Alabama at Birmingham & Birmingham VA Medical Center, 422 THT, 1900 University Blvd, Birmingham, AL 35294, USA. Email: [email protected]
74.116 John Sealy Annex, University of Texas Medical Branch, 301 University Blvd, Galveston, TX, 77555-0568, USA: Email: [email protected] 8Parexel International, Shelton Simmons (MD), 3001 S Hanover St #7, Brooklyn, MD 21225, USA. Email: [email protected]
9 Pneumologie, Aix Marseilles University, APHM/ INSERM U1067, Chemin des Bourellys 13015, Marseille, France. Email: [email protected]
TE D
431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451
RI PT
427
The following Janssen personnel contributed significantly to the success of ADEPT: Debra
453
Alvarez, Jennifer Campos, Robert Gordon, Keith Lasher, Francisco Leon, Hongjuan Liu, Jennifer
454
Montello, Nancy Peffer, Kevin Petty, Filza Potapova, and Dipti Shah.
AC C
EP
452
24
ACCEPTED MANUSCRIPT
455
References
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
1.
7. 8. 9.
10. 11. 12. 13. 14.
15.
16.
17. 18. 19.
RI PT
SC
6.
M AN U
5.
TE D
4.
EP
3.
AC C
2.
From the Global Strategy for Asthma Management and Prevention, Global Initiative for Asthma (GINA) 2015. 2015.] Available from http://www.ginasthma.org/. Fahy JV. Type 2 inflammation in asthma--present in most, absent in many. Nat Rev Immunol 2015; 15:57-65. Karta MR, Broide DH, Doherty TA. Insights into Group 2 Innate Lymphoid Cells in Human Airway Disease. Curr Allergy Asthma Rep 2016; 16:8. Woodruff PG, Modrek B, Choy DF, Jia G, Abbas AR, Ellwanger A, et al. T-helper type 2-driven inflammation defines major subphenotypes of asthma. Am J Respir Crit Care Med 2009; 180:388-95. Woodruff PG, Boushey HA, Dolganov GM, Barker CS, Yang YH, Donnelly S, et al. Genome-wide profiling identifies epithelial cell genes associated with asthma and with treatment response to corticosteroids. Proc Natl Acad Sci U S A 2007; 104:15858-63. Peters MC, Mekonnen ZK, Yuan S, Bhakta NR, Woodruff PG, Fahy JV. Measures of gene expression in sputum cells can identify TH2-high and TH2-low subtypes of asthma. J Allergy Clin Immunol 2014; 133:388-94. Little SA, Chalmers GW, MacLeod KJ, McSharry C, Thomson NC. Non-invasive markers of airway inflammation as predictors of oral steroid responsiveness in asthma. Thorax 2000; 55:232-4. Jang AS, Lee JH, Park SW, Lee YM, Uh ST, Kim YH, et al. Factors influencing the responsiveness to inhaled glucocorticoids of patients with moderate-to-severe asthma. Chest 2005; 128:1140-5. Hanania NA, Wenzel S, Rosen K, Hsieh HJ, Mosesova S, Choy DF, et al. Exploring the effects of omalizumab in allergic asthma: an analysis of biomarkers in the EXTRA study. Am J Respir Crit Care Med 2013; 187:804-11. Wenzel S, Ford L, Pearlman D, Spector S, Sher L, Skobieranda F, et al. Dupilumab in persistent asthma with elevated eosinophil levels. N Engl J Med 2013; 368:2455-66. Bel EH, Wenzel SE, Thompson PJ, Prazma CM, Keene ON, Yancey SW, et al. Oral glucocorticoidsparing effect of mepolizumab in eosinophilic asthma. N Engl J Med 2014; 371:1189-97. Ortega HG, Liu MC, Pavord ID, Brusselle GG, FitzGerald JM, Chetta A, et al. Mepolizumab treatment in patients with severe eosinophilic asthma. N Engl J Med 2014; 371:1198-207. Corren J, Lemanske RF, Hanania NA, Korenblat PE, Parsey MV, Arron JR, et al. Lebrikizumab treatment in adults with asthma. N Engl J Med 2011; 365:1088-98. Proceedings of the ATS workshop on refractory asthma: current understanding, recommendations, and unanswered questions. American Thoracic Society. Am J Respir Crit Care Med 2000; 162:2341-51. Bel EH, Sousa A, Fleming L, Bush A, Chung KF, Versnel J, et al. Diagnosis and definition of severe refractory asthma: an international consensus statement from the Innovative Medicine Initiative (IMI). Thorax 2011; 66:910-7. Silkoff PE, Strambu I, Laviolette M, Singh D, FitzGerald JM, Lam S, et al. Asthma characteristics and biomarkers from the Airways Disease Endotyping for Personalized Therapeutics (ADEPT) longitudinal profiling study. Respir Res 2015; 16:142. Juniper EF, O'Byrne PM, Guyatt GH, Ferrie PJ, King DR. Development and validation of a questionnaire to measure asthma control. Eur Respir J 1999; 14:902-7. Juniper EF, Buist AS, Cox FM, Ferrie PJ, King DR. Validation of a standardized version of the Asthma Quality of Life Questionnaire. Chest 1999; 115:1265-70. Kelly MM, Efthimiadis A, Hargreave FE. Induced sputum : selection method. Methods Mol Med 2001; 56:77-91. 25
ACCEPTED MANUSCRIPT
25.
26.
27.
28.
29. 30. 31.
32.
33.
34.
35.
36.
RI PT
24.
SC
23.
M AN U
22.
TE D
21.
Loza MJ, Perussia B. Peripheral immature CD2-/low T cell development from type 2 to type 1 cytokine production. Journal of Immunology 2002; 169:3061-8. Loza MJ, Perussia B. Final steps of natural killer cell maturation: a model for type 1-type 2 differentiation? Nat Immunol 2001; 2:917-24. Larose MC, Chakir J, Archambault AS, Joubert P, Provost V, Laviolette M, et al. Correlation between CCL26 production by human bronchial epithelial cells and airway eosinophils: Involvement in patients with severe eosinophilic asthma. J Allergy Clin Immunol 2015. Komiya A, Nagase H, Yamada H, Sekiya T, Yamaguchi M, Sano Y, et al. Concerted expression of eotaxin-1, eotaxin-2, and eotaxin-3 in human bronchial epithelial cells. Cell Immunol 2003; 225:91-100. Syed F, Huang CC, Li K, Liu V, Shang T, Amegadzie BY, et al. Identification of interleukin-13 related biomarkers using peripheral blood mononuclear cells. Biomarkers 2007; 12:414-23. Nomura T, Terada N, Kim WJ, Nakano K, Fukuda Y, Wakita A, et al. Interleukin-13 induces thymus and activation-regulated chemokine (CCL17) in human peripheral blood mononuclear cells. Cytokine 2002; 20:49-55. van Hartingsveldt B, Nnane IP, Bouman-Thio E, Loza MJ, Piantone A, Davis HM, et al. Safety, tolerability and pharmacokinetics of a human anti-interleukin-13 monoclonal antibody (CNTO 5825) in an ascending single-dose first-in-human study. Br J Clin Pharmacol 2013; 75:1289-98. Alvarez-Gutierrez FJ, Medina-Gallardo JF, Perez-Navarro P, Martin-Villasclaras JJ, Martin Etchegoren B, Romero-Romero B, et al. [Comparison of the Asthma Control Test (ACT) with lung function, levels of exhaled nitric oxide and control according to the Global Initiative for Asthma (GINA)]. Arch Bronconeumol 2010; 46:370-7. Barros R, Moreira A, Fonseca J, de Oliveira JF, Delgado L, Castel-Branco MG, et al. Adherence to the Mediterranean diet and fresh fruit intake are associated with improved asthma control. Allergy 2008; 63:917-23. Cowan DC, Hewitt RS, Cowan JO, Palmay R, Williamson A, Lucas SJ, et al. Exercise-induced wheeze: Fraction of exhaled nitric oxide-directed management. Respirology 2010; 15:683-90. Michils A, Baldassarre S, Van Muylem A. Exhaled nitric oxide and asthma control: a longitudinal study in unselected patients. Eur Respir J 2008; 31:539-46. Castro M, Wenzel SE, Bleecker ER, Pizzichini E, Kuna P, Busse WW, et al. Benralizumab, an antiinterleukin 5 receptor alpha monoclonal antibody, versus placebo for uncontrolled eosinophilic asthma: a phase 2b randomised dose-ranging study. Lancet Respir Med 2014; 2:879-90. Peters MC, Mekonnen ZK, Yuan S, Bhakta NR, Woodruff PG, Fahy JV. Measures of gene expression in sputum cells can identify TH2-high and TH2-low subtypes of asthma. Journal of Allergy and Clinical Immunology 2014; 133:388-94. Choy DF, Modrek B, Abbas AR, Kummerfeld S, Clark HF, Wu LC, et al. Gene expression patterns of Th2 inflammation and intercellular communication in asthmatic airways. J Immunol 2011; 186:1861-9. Provost V, Larose MC, Langlois A, Rola-Pleszczynski M, Flamand N, Laviolette M. CCL26/eotaxin3 is more effective to induce the migration of eosinophils of asthmatics than CCL11/eotaxin-1 and CCL24/eotaxin-2. J Leukoc Biol 2013; 94:213-22. Horiuchi K, Amizuka N, Takeshita S, Takamatsu H, Katsuura M, Ozawa H, et al. Identification and characterization of a novel protein, periostin, with restricted expression to periosteum and periodontal ligament and increased expression by transforming growth factor beta. J Bone Miner Res 1999; 14:1239-49. Li G, Oparil S, Sanders JM, Zhang L, Dai M, Chen LB, et al. Phosphatidylinositol-3-kinase signaling mediates vascular smooth muscle cell expression of periostin in vivo and in vitro. Atherosclerosis 2006; 188:292-300.
EP
20.
AC C
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548
26
ACCEPTED MANUSCRIPT
39.
40.
41.
42.
43.
44.
RI PT
38.
Choy DF, Hart KM, Borthwick LA, Shikotra A, Nagarkar DR, Siddiqui S, et al. TH2 and TH17 inflammatory pathways are reciprocally regulated in asthma. Sci Transl Med 2015; 7:301ra129. Alving K, Malinovschi, A. Basic aspects of exhaled nitric oxide. In: European Respiratory Monograph. United Kingdom: European Respiratory Publications; 2010. p. 1-31. Malinovschi A, Fonseca JA, Jacinto T, Alving K, Janson C. Exhaled nitric oxide levels and blood eosinophil counts independently associate with wheeze and asthma events in National Health and Nutrition Examination Survey subjects. J Allergy Clin Immunol 2013; 132:821-7 e1-5. Lilly CM, Nakamura H, Kesselman H, Nagler-Anderson C, Asano K, Garcia-Zepeda EA, et al. Expression of eotaxin by human lung epithelial cells: induction by cytokines and inhibition by glucocorticoids. J Clin Invest 1997; 99:1767-73. Teran LM, Mochizuki M, Bartels J, Valencia EL, Nakajima T, Hirai K, et al. Th1- and Th2-type cytokines regulate the expression and production of eotaxin and RANTES by human lung fibroblasts. Am J Respir Cell Mol Biol 1999; 20:777-86. Ghaffar O, Hamid Q, Renzi PM, Allakhverdi Z, Molet S, Hogg JC, et al. Constitutive and cytokinestimulated expression of eotaxin by human airway smooth muscle cells. Am J Respir Crit Care Med 1999; 159:1933-42. Imai T, Baba M, Nishimura M, Kakizaki M, Takagi S, Yoshie O. The T cell-directed CC chemokine TARC is a highly specific biological ligand for CC chemokine receptor 4. J Biol Chem 1997; 272:15036-42. Coleman JM, Naik C, Holguin F, Ray A, Ray P, Trudeau JB, et al. Epithelial eotaxin-2 and eotaxin-3 expression: relation to asthma severity, luminal eosinophilia and age at onset. Thorax 2012; 67:1061-6.
SC
37.
M AN U
549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570
AC C
EP
TE D
571
27
ACCEPTED MANUSCRIPT
Figure legends Figure legends
SC
RI PT
Figure 1. Airway-mucosal expression-based categorization of Type-2 phenotype. Airwaymucosal CCL26 (A) and periostin (B) gene expression (log2 intensity) and IL-13-IVS enrichment (GSVA scores) (C) presented in the healthy or mild and moderate-severe asthma cohorts, stratified by airway-mucosa-Type-2 high vs. low groups (x-axis top label) defined by CCL26 (A), periostin (B), or IL-13-IVS (C). Horizontal solid line indicates threshold for high.
M AN U
Figure 2. Top-associated clinical and biomarker variables to airway-mucosa-CCL26 status. Screening/baseline values for (A) FENO, (B) blood eosinophils, (C) sputum eosinophils, and (D) submucosal eosinophils are shown for subjects in the mild, moderate/severe asthma groups, stratified by airway-mucosa-CCL26 high vs. low status (x-axis top labels). * P<0.05 for airwaymucosa-CCL26 high vs. low status within mild or moderate-severe asthma population.
TE D
Figure 3. Type-2-associated serum biomarkers and airway-mucosa-CCL26 status.
AC C
EP
Baseline values of serum IgE (A), serum CCL17 (B) and serum CCL26 (C) are shown for subjects in the mild, moderate/severe asthma groups (x-axis bottom labels), stratified by airwaymucosa-CCL26 high vs. low status (x-axis top labels). * P<0.05 for airway-mucosa-CCL26 high vs. low status within mild or moderate-severe asthma population.
28
ACCEPTED MANUSCRIPT
Tables
airway-mucosal CCL26-high
Mild asthma airway-mucosal periostin-high
airway-mucosal IL-13-IVS-high
airway-mucosal CCL26-high
100% *
94%
81%
airway-mucosal CCL26-low
0%
33%
17%
airway-mucosal periostin-high
79%
100%
68%
airway-mucosal periostin-low
11%
0%
22%
airway-mucosal IL-13-IVS-high
87%
87%
airway-mucosal IL-13-IVS-low
23%
46%
Moderate-Severe asthma airway-mucosal airway-mucosal airway-mucosal CCL26-high periostin-high IL-13-IVS-high 85%
85%
0%
21%
31%
55%
100%
65%
6%
0%
31%
100%
46%
54%
100%
0%
6%
23%
0%
SC
100%
M AN U
% of group:
RI PT
Table 1. Segregation of airway-mucosal Type-2 high versus -low by airway-mucosal CCL26, periostin, and IL-13-IVS gene expression
AC C
EP
TE D
* The percentage of mild or moderate-severe asthma subjects within the airway group indicated in the row label that have airwaymucosal-high status based on biomarker indicated in column label
29
ACCEPTED MANUSCRIPT
Table 2. Demographics by airway-mucosal- CCL26 status for mild and moderate-severe asthma
P*
Moderate-severe asthma airway-mucosal-CCL26 airway-mucosal-CCL26 high (n=13) low (n=42)
P*
Age, years
27.0 ± 7.6 (21.0 - 49.0)
30.4 ± 12.2 (18.0 - 54.0)
0.37
42.8 ± 9.4 (26.0 - 54.0)
41.2 ± 11.3 (18.0 - 55.0)
0.66
BMI
24.6 ± 3.0 (19.6 - 29.6)
24.1 ± 3.4 (20.2 - 30.1)
0.69
26.9 ± 2.9 (21.4 - 31.2)
26.5 ± 3.8 (19.1 - 32.1)
0.71
Duration of asthma, years
16.6 ± 6.9 (7.1 - 31.6)
15.7 ± 8.7 (0.7 - 32.9)
0.76
21.6 ± 10.7 (3.6 - 39.5)
19.2 ± 12.8 (0.3 - 47.5)
0.55
8/16 (50%)
3/12 (25%)
0.25
8/13 (62%)
16/42 (38%)
0.20
White
14/16 (88%)
12/12 (100%)
10/13 (77%)
31/42 (74%)
Black
0/16 (0%)
0/12 (0%)
2/13 (15%)
8/42 (19%)
Other
2/16 (13%)
0/12 (0%)
1/13 (8%)
3/42 (7%)
N/total (% of total) Gender, male Race
RI PT
Mild asthma Airway-mucosal-CCL26 airway-mucosal-CCL26 high (n=16) low (n=12)
SC
Mean ± SD (range)
M AN U
0.50
1.00
AC C
EP
TE D
* P-value for comparison between airway-mucosal-CCL26-high and –low groups within mild asthma or moderate-severe asthma populations
30
ACCEPTED MANUSCRIPT
Table 3: Asthma disease characteristics and clinical biomarkers by airway-mucosal-CCL26 status Mild asthma
Moderate-severe asthma
airway-mucosal-CCL26- airway-mucosal-CCL26 phigh low value†
airway-mucosal-CCL26 high
airway-mucosal-CCL26 low
pvalue†
RI PT
Mean ± SD (range) [N]
93.6 ± 11.7 [15]
90.8 ± 14.3 [12]
0.577
70.2 ± 10.3 [13]
71.2 ± 11.6 [41]
0.785
FEV1, Post-BD (PN)
102.1 ± 9.1 [15]
97.9 ± 11.2 [12]
0.293
83.0 ± 8.9 [13]
80.8 ± 13.8 [41]
0.598
BDR, % change FEV1
6.25 ± 8.54 [15]
6.33 ± 3.25 [12]
0.978
19.79 ± 15.86 [13]
14.63 ± 10.96 [40]
0.195
ACQ
0.6 ± 0.5 [16]
0.9 ± 0.6 [12]
0.176
1.7 ± 0.8 [13]
1.5 ± 0.9 [42]
0.549
AQLQ
6.2 ± 0.7 [16]
5.9 ± 0.9 [12]
0.245
5.6 ± 0.8 [13]
5.5 ± 1.3 [42]
0.637
M AN U
SC
FEV1, Pre-BD (PN)
0.84 +5.33/-0.73 [16]
3.33 +8.18/-2.37 [12] 0.045
0.32 +1.77/-0.27 [10]
0.95 +4.76/-0.79 [35]
0.102
FENO (ppb) *
52.8 +44.0/-24.0 [16]
27.4 +17.7/-10.8 [12] 0.0051
56.8 +56.7/-28.4 [13]
24.9 +23.7/-12.1 [41]
0.0003
bEOS (1000/mm3) *
0.21 +0.19/-0.10 [16]
0.12 +0.13/-0.06 [12] 0.051
0.36 +0.22/-0.14 [13]
0.19 +0.20/-0.10 [42]
0.0032
sIgE (RFU) *
17.1 +33.2/-11.3 [16]
0.032
20.0 +27.4/-11.6 [13]
7.8 +25.9/-6.0 [41]
0.032
0.14
398 +299/-171 [13]
169 +236/-98 [42]
0.0006
TE D
PC20 (mg/ml) *
EP
5.4 +21.8/-4.3 [12]
169 +226/-97 [16]
117 +103/-55 [12]
sCCL26 (pg/ml) *
117 +732/-101 [16]
42 +127/-32 [9]
0.185
416 +3301/-369 [12]
57 +290/-48 [41]
0.0023
SpEOS (% WBC) *
1.2 +8.4/-1.1 [12]
0.5 +1.1/-0.3 [7]
0.291
6.8 +12.9/-4.5 [8]
2.2 +12.4/-1.9 [31]
0.117
SpNEU (% WBC)
45.6 ± 35.6 [12]
52.1 ± 41.1 [7]
0.561
50.8 ± 35.1 [8]
46.9 ± 25.4 [31]
0.893
AC C
sCCL17 (pg/ml) *
* Geometric mean ± asymmetric standard deviation [N], p-value calculated from log-transformed data. † P-value for comparison between airway-CCL26-high and –low groups within mild asthma or moderate-severe asthma populations; bolded when FDR<0.05. 31
Mild asthma Type-2 low (CCL26-) (n=12) 1 (8%) [20%]
P-value* 0.36
Type-2 high (CCL26+) (n=13) 6 (46%) [55%]
12 (75%) [75%]
4 (33%) [25%]
0.053
13 (100%) [46%]
4 (25%) [33%]
8 (67%) [67%]
FENO ≥ 35 ppb FENO < 35 ppb
12 (75%) [75%] 4 (25%) [33%]
4 (33%) [25%] 8 (67%) [67%]
0.053
Blood EOS ≥ 300/mm3 Blood EOS < 300/mm3
4 (25%) [80%] 12 (75%) [52%]
1 (8%) [20%] 11 (92%) [48%]
0.36
Serum IgE high Serum IgE low
15 (94%) [71%] 1 (6%) [14%]
6 (50%) [29%] 6 (50%) [86%]
Atopy (Phadiatop): Positive Atopy (Phadiatop): Negative
16 (100%) [70%] 0 (0%) [0%]
Serum CCL17 ≥75th%ile Serum CCL17 <75th%ile
3 (100%) [100%] 13 (0%) [54%]
Serum CCL26 ≥75th%ile Serum CCL26 <75th%ile
4 (100%) [100%] 12 (0%) [52%]
Sputum EOS ≥ 3% Sputum EOS < 3%
4 (33%) [100%] 8 (67%) [53%]
Sputum NEU ≥ 65% Sputum NEU < 65% Paucigranulocytic sputum Neutrophilic sputum Mixed granulocytic sputum Eosinophilic sputum
Moderate-Severe asthma Type-2 low (CCL26-) (n=42) 5 (12%) [45%] 15 (36%) [54%]
P-value* 0.014 0.000040
27 (64%) [100%]
9 (69%) [47%] 4 (31%) [11%]
10 (24%) [53%] 31 (76%) [89%]
0.0063
10 (77%) [50%] 3 (23%) [9%]
10 (24%) [50%] 32 (76%) [91%]
0.00090
0.023
12 (92%) [29%] 1 (8%) [8%]
29 (71%) [71%] 12 (29%) [92%]
0.15
7 (64%) [30%] 4 (36%) [100%]
0.019
11 (85%) [26%] 2 (15%) [15%]
31 (74%) [74%] 11 (26%) [85%]
0.71
0 (64%) [0%] 11 (36%) [46%]
0.25
11 (85%) [52%] 1 (15%) [3%]
10 (74%) [48%] 32 (26%) [97%]
0.00004
0 (64%) [0%] 11 (36%) [48%]
0.12
7 (85%) [64%] 5 (15%) [12%]
4 (74%) [36%] 38 (26%) [88%]
0.0010
0 (0%) [0%] 7 (100%) [47%]
0.25
7 (87.5%) [33%] 1 (12.5%) [6%]
14 (45%) [67%] 17 (55%) [94%]
0.049
2 (33%) [33%] 10 (67%) [77%]
4 (0%) [67%] 3 (100%) [23%]
0.13
3 (88%) [25%] 5 (12%) [19%]
9 (45%) [75%] 22 (55%) [81%]
0.68
6 (50%) [67%] 2 (17%) [33%] 0 (0%) [na] 4 (33%) [100%]
3 (43%) [33%] 4 (57%) [67%] 0 (0%) [na] 0 (0%) [0%]
0.10
0 (0%) [0%] 1 (12.5%) [14%] 2 (25%) [40%] 5 (62.5%) [31%]
11 (35.5%) [100%] 6 (19%) [86%] 3 (10%) [60%] 11 (35.5%) [69%]
0.15
M AN U
SC
0 (0%) [0%]
TE D
FENO ≥ 35 ppb and Blood Eos ≥ 300/mm3 FENO ≥ 35 ppb or Blood Eos ≥ 300/mm3 Neither FENO ≥ 35 ppb nor Blood Eos ≥ 300/mm3
EP
N (% of Type-2 group) [% of row group]
RI PT
Type-2 high (CCL26+) (n=16) 4 (25%) [80%]
AC C
ACCEPTED MANUSCRIPT Table 4. Comparison of airway-mucosal-Type-2-CCL26 status and categorical biomarkers (high or low)
* P-value (FDR) for comparison between airway-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test, or chi-square test for sputum EOS-PMN subgroups); bolded when FDR<0.05.
32
ACCEPTED MANUSCRIPT
Table 5 airway-mucosal-CCL26 diagnostic characteristics of accessible biomarkers in moderate-severe asthma airway-mucosal-CCL26-high classifiers*
AUC
PPV
NPV
100% 46% 69% 77% 76% 93%
63% 88% 76% 76% 85% 85%
0.82 0.67 0.72 0.77 0.80 0.89
46% 55% 48% 50% 62% 66%
100% 84% 89% 91% 92% 97%
3
58% 58%
90% 98%
3
50% 50%
93% 100%
FENO≥35ppb bEOS ≥300/mm
3
sCCL17-high 3
(FENO≥35ppb or bEOS ≥300/mm ) AND sCCL17-high sCCL26-high (FENO≥35ppb or bEOS ≥300/mm ) AND sCCL26-high sCCL26-high AND sCCL17-high (FENO≥35ppb or bEOS ≥300/mm ) AND sCCL26-high AND sCCL17-high
SC
FENO≥35ppb and bEOS ≥300/mm
3
M AN U
3
Prevalence†
RI PT
Specificity
24% 24% 24% 24% 24% 24%
0.74 0.77
64% 88%
88% 89%
23% 23%
0.71 0.75
67% 100%
86% 87%
23% 23%
TE D
FENO≥35ppb or bEOS ≥300/mm
Sensitivity
* Logistic regression model statistics for classification of moderate-severe asthmatics to airway-mucosal-CCL26-high status: AUC, area-undercurve for receiver operating characteristic; NPV, negative predictive value; PPV, positive predictive value.
AC C
EP
† Prevalence based on the proportion of airway-mucosal-CCL26-high subjects in the ADEPT moderate-severe asthma group, with PPV and NPV statistics based on this prevalence. ADEPT is a cross-sectional study that has rigid enrollment criteria, therefore the prevalence in other asthma study populations may vary based on recruitment criteria.
33
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Identification of Airway-Mucosal Type-2 inflammation by Clinical Biomarkers in Asthma Philip E Silkoff, Michel Laviolette, Dave Singh, J Mark FitzGerald, Steven Kelsen, Vibeke
RI PT
Backer, Celeste M Porsbjerg, Pierre-Olivier Girodet, Patrick Berger, Joel N Kline, Geoffrey
& the ADEPT investigators
Online Data Supplement
M AN U
E1. Biomarker matrices and assessments
SC
Chupp, Vedrana S Susulic, Elliot S Barnathan, Frédéric Baribaud, and Matt J Loza
Bronchoscopy sampling
Up to 4 biopsies were immediately preserved in RNAlater ® solution and then maintained at -70°C. For 2 biopsies per subject, RNA was extracted using Qiagen miRNeasy kit (Qiagen; Germantown, MD). RNA quality was assessed by Caliper Life
TE D
Sciences' LabChip System, with 11 of 119 samples processed having insufficient or too poor quality RNA for further processing. 108 RNA samples were amplified with NuGen ovation pico WTA kit (NuGen Technologies; San Carlos, CA). The cDNA was
EP
analyzed using the Affymetrix HG-U133+PM microarray platform (Affymetrix, Santa Clara, CA). CEL files were normalized, assessed for quality control to exclude technical outliers (chip image analysis, Affymetrix GeneChip QC, RNA degradation
AC C
analysis, distribution analysis, principal components analysis, and correlation analysis), and re-normalized using the robust multi-array (RMA) method. The log2normalized data matrix was imported into OmicSoft ArrayStudio software (Cary, NC; www.omicsoft.com) for subsequent analysis. No samples failed the quality control metrics. Batch effects from the two RNA processing sets were observed, with the batch effect adjusted in the data matrices using linear modeling of batch (as random factor) and cohort. A log2-intensity threshold of 5.5 for biopsies was established as the limit of reliable quantification based on the 90th percentile signal
ACCEPTED MANUSCRIPT
of merged nonspecific probesets distribution in the array and by the inflection point of maximum variance with decreasing signal in a standard deviation vs. mean intensity plot across all probesets. Probesets with mean log2 intensity above this
RI PT
threshold in at least one of the 4 study cohorts were considered quantifiable and included in subsequent analyses (24033 probesets). Group comparisons of gene and protein expression (log2 –transformed) were performed using General Linear
Models, adjusting for age and gender. For microarray experiments, significance for
SC
association with a Type-2 group was generally defined as passing a False Discovery
Rate (FDR, Benjamini-Hochberg method) < 0.05 vs. healthy controls and Type-2 high vs. low groups, with > 2-fold estimate (ratio of least-square means of the two
M AN U
comparison groups) for each comparison. Nominally significant associations are reported for comparisons with p-value < 0.05 and >2-fold estimate, reported to provide additional context because of the inflated false-negative rate from FDRadjustment.
Histological evaluation of airway-mucosal biopsies was performed centrally by
TE D
Pantomics Inc. (Richmond, CA; www.pantomics.com). Immediately after collection, the tissues were fixed in 10% neutral buffered formalin and stored at room temperature in 70% ethanol until transfer to Pantomics Inc. Tissue samples were processed in a Leica
EP
TP1020 Processor and paraffin-embedded in a Leica Embedding Center following manufacturer’s instructions. For each paraffin-embedded sample, a section was cut and H&E-stained, which was then assessed and marked under a microscope. The marked
AC C
areas were then punched and transferred to a recipient paraffin block using a stereomicroscope and the Beecher TMA instrument (Beecher Instruments, Sun Prairie, WI). Three tissue microarrays (TMA), each containing 69 (1.5mm) cores were constructed. H&E staining was performed using an Autostainer (Thermo Scientific™ Lab Vision™ Autostainer 360-2D) following manufacturer’s instruction. General histological characteristics and specifically the intensity of submucosal eosinophil infiltration were interpreted and scored on a scale from 0 – 3 by a staff pathologist.
ACCEPTED MANUSCRIPT
Induced sputum Only samples with squamous cell content ≤30% evaluated from cytospin slides evaluated by a central laboratory were included in the analyses. A significant proportion
RI PT
of subjects had only a screening or only a baseline sample available that passed quality control standards (e.g., for cytospin slides, 105/189 possible subjects had acceptable readings at screening, 85 acceptable at baseline, with 128 subjects having either a screening and/or baseline read acceptable). Therefore, the mean (differential cell
SC
counts) or geometric mean (analyte, gene expression measurements) of screening and
M AN U
baseline measurements was used for subsequent analyses.
Induction procedure: All study participants underwent sputum induction during screening to fulfil inclusion criteria and again at the baseline Visit. Asthmatic subjects only had a 3rd sputum induction at the 6 month biomarker Visit. Sputum was induced for 21 minutes divided into three 7-minute sessions of nebulization each followed by a 3
TE D
step cleansing procedure and a focused cough attempt. An aerosol of hypertonic saline (in increasing concentrations of 3, 4, and 5%) was generated by an ultrasonic nebulizer for inhalation by subjects with a post-bronchodilator pre-induction FEV1 of ≥60% predicted; for those with FEV1 ≥50-<60% predicted, induction was performed with
EP
normal/isotonic saline (0.9%). Subjects with a post-bronchodilator FEV1 <50% predicted
AC C
did not undergo induction.
Sputum Processing: The plug selection method was used for this study in all participants 1
. A plug weight of at least 50mg and squamous cell percentages≤20% (evaluated at
local processing laboratory from hemocytometer counting) were required for enrollment into the study during screening. Sputum plugs were treated with dithiothreitol to disperse mucus before collecting supernatant (immediately frozen at ≤ 20°C) and cells for cytospin slide generation and cell pellet RNA preservation in RNAlater®. Samples were further required to have ≤30% squamous cells determined from centrally read cytospin slides to pass quality control metrics.
ACCEPTED MANUSCRIPT
Serum analytical methods Serum was collected using standard Serum Separation Tubes, and frozen within 30 minutes. One of these frozen aliquots, without intermediate freeze-thaw cycles, was
RI PT
provided for quantification of 1129 serum analytes using the SomaScan v3 platform (SomaLogic, Boulder, CO; www.somalogic.com). Serum analyte levels were reported by
Somalogic as relative fluorescence units, cross-plate calibrated, and median normalized. Analyte levels are presented as the log2 ratio to the geometric means of the healthy
SC
controls population for further analysis. Results for serum total immunoglobulin E (IgE)
are presented from this panel, defining high IgE levels as those above the 95th percentile
M AN U
of the healthy controls distribution. In previous evaluations of the platform in asthmatics and healthy controls, IgE measurements highly correlated (Pearson’s correlation coefficient r > 0.9) with those obtained from standard ELISA-based assays (data not shown). 1129 analytes were included in this panel but were not generally considered in the current manuscript given the exploratory nature of the panel.
TE D
Classical Type-2 cytokines including IL-4, IL-5, and IL-13 (lower limit of quantification of 0.13, 1.50, and 1.57 pg/ml, respectively) were specifically measured in serum using the high-sensitivity Meso Scale Discovery (MSD)® electrochemiluminescence platform (Meso Scale Discovery; Rockville, MD). For baseline samples, the Type-2-associated
EP
cytokine CCL17 (TARC) was measured by ELISA (R&D Systems, Minneapolis, MN), with the lower limit of quantification determined at 16 pg/ml. CCL17 (for month 3, 6, and 12
AC C
samples) (RayBio© kit ELH-TARC) and CCL26 (eotaxin-3) (RayBio© kit ELH-Eotaxin3) was measured by ELISA at Cirquest Labs, LLC (Memphis, TN), with the lower limit of quantification determined at 5 pg/ml and 28.7 pg/ml, respectively. Rationale underlying Definition of Type-2 status in airway mucosa. Type-2 high/low status was defined as inferred IL-13 activity in airway-mucosa from asthma subjects compared to non-atopic healthy controls, evaluated by gene expression for CCL26 (eotaxin-3), biopsy periostin (periostin) gene expression, or IL-13 in-vitro gene signature enrichment (IL-13- IVS). CCL26 (chemokine (C-C motif) ligand 26), was
ACCEPTED MANUSCRIPT
selected as the primary Type-2 indicator because it was the gene most highly and consistently induced in vitro by IL-13 in multiple airway resident non-hematopoietic cell types and macrophages (unpublished observations and 2, 3, and was also the most highly differentially expressed gene in biopsies from the ‘Type-2-high’ group defined by
RI PT
Woodruff et al4, and strongly correlated to IL-5 and IL-13 gene expression assessed by qPCR5. CCL26 is tightly regulated by IL-4 and IL-13 and is not directly induced by other
cytokines, such as TNFα, IL-6, interferons (IFNα, IFNɣ) and by tumor growth factor-beta
(TGF-β) (unpublished results). CCL26 was preferred a priori over periostin because of the
SC
co-regulation of periostin, but not CCL26, by TGF-β) (see references6, 7, and unpublished observations), and the constitutive expression of periostin, but not CCL26, observed in
M AN U
healthy controls5, 8. The IL-13-IVS was evaluated to provide a composite score across a broader range of genes inducible by IL-13 rather than using a single gene. However, this approach has the limitation that genes included in the signature may not be specific for IL-13, compared to the specificity of CCL26.
For CCL26, the highest signals in the non-atopic healthy control cohort were just below
TE D
the limit of reliable quantification for the microarray, so this limit of quantification (log2 intensity of 5.0) was set as the threshold for airway-mucosal-CCL26-high versus -low status. For periostin and IL-13-IVS, Type-2-high status was defined as gene expression
EP
(or enrichment) beyond the 95th percentile of the non-atopic healthy control cohort. IL-13 in vitro stimulation signature
AC C
The IL-13 in vitro signature (IVS) was established from the genes induced by IL-13 in airway-liquid interface cultures of airway-mucosal epithelial cells, selecting genes commonly induced across 3 independent sets of experiments (genes in signature listed in Table E1): •
Experiment 1: Bronchial epithelial cultures were established from bronchial
epithelia from fatal asthma cases (n=6) and normal controls from fatal accidents (n=6). After establishment of air-liquid interface, cultures were stimulated with IL-4 and IL-13
ACCEPTED MANUSCRIPT
in combination, or vehicle, for 14-days. Genes passing a filter of FDR<0.05 and 2fold/vehicle for IL-4/IL-13 vs. vehicle comparison in either asthma or control groups were selected. Experiment 2: Bronchial epithelial cultures were established from bronchial
RI PT
•
epithelia obtained from endobronchial biopsies from asthmatics (n=6) or healthy
controls (n=3). After establishment of air-liquid interface, cultures were stimulated with IL-13 at 5 ng/ml or 10 ng/ml or with vehicle for 7, 14, or 21 days. Genes passing a filter
SC
of p<0.05 and 2-fold/vehicle for IL-13 vs. vehicle comparisons for both doses of IL-13 and for at least 2 time points, and common to both asthma and control groups were selected.
Experiment 3: Bronchial epithelial cultures were established from bronchial
M AN U
•
epithelia from fatal asthma cases (n=10). After establishment of air-liquid interface, cultures were stimulated with IL-13 or vehicle for 6-hours, 24-hours, 7-days, or 14-days. Genes passing a filter of FDR<0.05 and 2-fold/vehicle for IL-13 vs. vehicle comparison for at least 2 consecutive time points were selected.
TE D
Enrichment was evaluated on a per-subject basis using the R-Bioconductor package Gene Set Variation Analysis (GSVA, v 1.14.19), providing enrichment scores for each
RESULTS
EP
subject representing IL-13 pathway activity.
AC C
E2. Biomarker Analysis
108 acceptable airway mucosal biopsies were obtained for analysis (healthy controls n=25; mild n=28, moderate n=29, severe asthma n=26). 128 subjects had acceptable sputum samples available at screening or baseline, of which 74 subjects had matching airway biopsy microarray data (healthy controls n=16; mild n=19, moderate n=18, and severe asthma n=21). 188 subjects had serum analytes measured at baseline, of which 108 subjects had matching airway biopsy microarray data (healthy n=25; mild n=28, moderate n=29, severe asthma n=26).
ACCEPTED MANUSCRIPT
The mild asthma cohort was analyzed separately from the moderate and severe (moderate-severe) asthma cohorts because the mild asthma patients were not taking ICS, which could be influential on the biomarker profiles including airway-mucosa
RI PT
expression of Type-2-associated genes (as demonstrated by Woodruff et al8. Moderatesevere asthmatics were the focus of the analyses because they have unmet clinical need (reduced lung function despite ICS treatment), and those presenting with the Type-2 phenotype would be considered to have a persistent Type-2 phenotype (that is,
SC
persistent despite ICS treatment). Separate analyses on the moderate and severe cohorts were generally not undertaken because the sample size for Type-2-high
subgroups within each cohort would have resulted in substantially limited statistical
M AN U
power to observe significance of true differences.
Relationship between clinical biomarkers and biopsy Type-2 status by CCL26, periostin, or IL-13-IVS: further analyses Moderate-Severe Asthma
TE D
In the moderate-severe asthma cohorts, all airway-mucosal-CCL26-Type-2-high subjects had either high FENO (>=35 ppb) or high blood eosinophils (>=300/mm3), whereas 80% of airway-mucosal-periostin-high and 75% of airway-mucosal-IL-13-IVS-high subjects had high FENO or bEOS (for CCL26, Table 2 in main report; for periostin and IL-13-IVS,
EP
Table E2, and Table E3, respectively. Similar proportions of Type-2-low subjects (32 -
AC C
36%) using the 3 Type-2 group definitions had either high FENO or high bEOS. When defining airway-mucosa-high status by CCL26, periostin, and IL-13-IVS, 100%, 80%, and 75% of Type-2-high moderate-severe asthmatics were FENO-high and/or bEOS-high, respectively. Similar proportions of Type-2-low moderate-severe asthmatics were high for either FENO or bEOS (36%, 34%, and 32%, respectively). Mild asthma For mild asthma (Table 2 in main report), similar to moderate-severe asthma, the airway-mucosal-CCL26-high group was mostly FENO high (>=35ppb) (75% of high group)
ACCEPTED MANUSCRIPT
and the airway-mucosal-CCL26-low group had a majority that was FENO low (<35ppb) (67% of low group). However, for mild asthma both the airway-mucosal-CCL26-high and -low groups were mostly low for bEOS (75% and 92% of respective groups). Almost all of
RI PT
the mild asthma airway-mucosal-CCL26-high subjects (96%) had high serum IgE, compared to 50% of mild asthma airway-mucosal-CCL26-low subjects. Atopic status was similarly distributed across airway-mucosal-CCL26-high and -low groups as was serum IgE.
SC
Associations of airway gene expression with airway-mucosal-CCL26 status
Genes associated specifically with airway-mucosal-CCL26-high status in moderate-
M AN U
severe asthma, were cystatin-1, periostin, and intelectin-1 (>2-fold with FDR<0.05 vs. healthy controls for each; >2-fold with FDR=0.0013, 0.0008, and 0.18, respectively, vs. airway-CCL26-low). See the full list of probe sets for mild and moderate asthma (Tables E4 and E5).
In mild asthma (non-steroid-treated), similar to moderate-severe asthma, elevated
TE D
periostin, CST1, and ITLN1 biopsy gene expression was also specifically associated with the airway-mucosal-CCL26-high phenotype. In addition to these genes, CST4 (cystatin-4) and TPSAB1 (mast cell tryptase) were also specifically elevated in airway-mucosalCCL26-high mild asthma. CST4 was only modestly elevated in airway-mucosal-CCL26-
EP
high moderate-severe asthma compared to healthy controls (p=0.0001, 1.53-fold) and the airway-mucosal-CCL26-low group (p=0.0017, 1.37-fold), whereas TPSAB1 didn’t
AC C
even have a trend for being elevated in airway-mucosal-CCL26-high moderate-severe asthma. Rather, TPSAB1 was significantly decreased in airway-mucosal-CCL26-low moderate-severe asthma (p=0.0041, 1.58-fold) but not in airway-mucosal-CCL26-low mild asthma (p=0.12, +1.40-fold). Table E5 reports the comparison statistics for all probe sets associated with airway-mucosal-CCL26-high and -low mild asthma compared to healthy controls, passing significance criteria of p<0.05 and >2-fold. Statistical differences between Type-2-CCL26 groups in the mild asthma cohort could not be evaluated because of the 8 mild asthmatics with EPBR microarray data, 6 were
ACCEPTED MANUSCRIPT
airway-mucosal-CCL26-high but only 2 airway-mucosal-CCL26-low. However, for the Type-2-CCL26-high mild asthma subjects, periostin, CST1, CST4, and TPSAB1 were significantly elevated compared to healthy controls (p<0.05, 2-fold), as observed for the
RI PT
airway-mucosal-CCL26-high moderate-severe asthma group. ITLN1 was the mostly highly over-expressed gene in the airway-mucosal-CCL26-high mild asthma group,
consistent with elevation of this gene in biopsies from airway-mucosal-CCL26-high
moderate-severe asthma compared to healthy controls and the airway-mucosal-CCL26-
SC
low group. These 5 genes were also expressed significantly higher in biopsies from the airway-mucosal-CCL26-high mild asthma subjects compared to healthy controls and
airway-mucosal-CCL26-low mild asthma subjects. Airway mucosal brushings expression
M AN U
of SERPINB2 (one of the 3 ‘Type-2-high’ defining genes, along with periostin and CLCA1, from Woodruff et al4) was also elevated in airway-mucosal-CCL26-high mild asthma group. SERPINB2 was also expressed significantly higher in airway-mucosal-CCL26-high compared to -low moderate-severe asthma (p=0.020, 3.23-fold), but not compared to
E3. Induced Sputum
TE D
healthy controls (p=0.23).
Airway-mucosal-periostin and- IL-13-IVS status and induced sputum cellular profiles
EP
Using biopsy periostin gene expression and IL-13-IVS enrichment as alternate identifiers for Type-2-high and Type-2-low status, further analyses for sputum cell counts are presented in Table E6 (periostin) and Table E7 (IL-13-IVS). Lower proportions of airway-
AC C
mucosal-periostin high (69%) and airway-mucosal-IL-13-IVS high (65%) moderate-severe asthmatics were spEOS high compared to that for airway-mucosal-CCL26 high status (88%). Similar proportions of moderate-severe asthmatics were spEOS-high (45-46%) in the airway mucosa-Type-2 low groups defined by CCL26, periostin, and IL-13-IVS (p=0.049, 0.31, and 0.33, respectively, for Type-2 -high vs. -low). These observations further reinforce the application of CCL26 as the primary discriminator for airwaymucosal-Type-2 status in this report.
ACCEPTED MANUSCRIPT
When defining airway-mucosal-Type-2 high status by CCL26, periostin, and IL-13-IVS, 87.5%, 69%, and 65% of Type-2-high moderate-severe asthmatics were sputum EOShigh (i.e. ≥3% of leukocytes), respectively. Similar proportions of Type-2-low moderate-
RI PT
severe asthmatics were sputum EOS high (45%, 46%, and 45%, respectively). See Tables
SC
E6 and E7.
M AN U
E4. Longitudinal Stability of serum CCL26 and CCL17
Serum CCL26 was remarkably stable at repeated visits over 12 months. Figure E2 shows the concentrations of sCCL26 at Months 3, 6, and 12, stratified by baseline visit high/low status. Only 4 of 94 samples had a change from baseline in sCCL26-high/-low status (96% concordance). Two airway-mucosa-CCL26-high subjects who were sCCL26-high at baseline were transiently sCCL26-low for one post-baseline visit but sCCL26-high for the
TE D
other visits. One airway-mucosa-CCL26-high subject who was sCCL26-low at baseline was sCCL26-high at 3 and 12 but low at month 6. Because of limitations of sample availability, baseline samples were analyzed at different times than the months 3-12. Nevertheless, the correlations for sCCL26 between visits were very strong (r>0.96) for all
EP
pairwise comparisons, thus demonstrating the robustness of the detection method.
AC C
Baseline samples for sCCL17 were measured using a different assay system (R&D Systems) than for 3, 6 and 12 samples (RayBio, qualified and run by Cirquest Labs). For sCCL17, the correlations between visits (excluding baseline visit, which was a different assay) were only moderate (Spearman r = 0.7 – 0.8). Additionally, there was poor correlation of baseline sCCL17 with the measurements at months 3-12 (r= 0.20 – 0.48). However, baseline sCCL17 measurements correlated better with sCCL26 measurements (r = 0.5 to 0.6) than it did with sCCL17 measurements at months 3-12. Therefore, it is likely that the R&D Systems assay run for baseline samples was more reliable than the
ACCEPTED MANUSCRIPT
RayBio assay and may explain reduced stability of serum-CCL17-high/-low status assessed at baseline over time.
E5. Correlations between Type-2 inflammation biomarkers and
RI PT
eosinophilic phenotypic markers
Figure E3 presents the correlations between Type-2 inflammation biomarkers and eosinophilic phenotypic markers, for mild and moderate-severe asthma. The correlations of airway type 2
SC
inflammation with eosinophilic and other type 2 inflammation biomarkers are similar when using airway CCL26 expression compared to airway periostin expression and IL-13-IVS scores,
M AN U
albeit generally modestly stronger for CCL26.
Interestingly, for the correlation of submucosal eosinophil density vs. serum CCL26, the correlation is inverse for mild asthma (r= -0.57) but modestly positive for moderate-severe asthma (r=0.36). This is in contrast to airway-mucosal-CCL26 expression vs. submucosal eosinophil density, which are positively correlated in both mild (r=0.47) and moderate-severe (r=0.36) asthma. This is consistent with significant correlation of airway CCL26 expression and
TE D
serum CCL26 levels only in mod-severe asthma (r=0.41) but not mild asthma (r=0.13). Similar correlations are also observed with FENO and submucosal (and sputum) eosinophils, with significant positive correlations observed only in moderate-severe but not mild asthma.
EP
However, airway CCL26 expression correlates with submucosal eosinophils in both mild and mod-severe asthma, with a stronger correlation in mild asthma.
AC C
A possible mechanism to explain this difference is that in mild asthma, leakage out of the submucosa, whether for eosinophils in the airway lumen or soluble mediators like CCL26 into the circulation, may be more variable than in moderate-severe asthma with a persistent inflammatory phenotype (despite inhaled steroids). Whatever the mechanisms, there are clear differences in mild (untreated) asthma and moderate-severe asthma with a persistent type 2 inflammatory phenotype limiting the ability to classify airway type inflammation in mild asthma using non-invasive biomarkers (e.g., FENO, serum proteins).
RI PT
ACCEPTED MANUSCRIPT
GGH
SIDT1
ALOX15
CISH
HS3ST1
SLC26A4
BCL2L15
CST1
LRRC31
SLC39A8
C1QTNF1
CST2
NTRK1
SLC5A1
CA2
CST4
OBFC2A
SOCS1
CCBL1
CTSC
PCSK6
SUSD2
CCL26
DPP4
POSTN
USP54
CD274
FAM26E
SERPINB4
CD44
FETUB
SH2D1B
a
Genes significantly induced by IL-13 commonly in each of 3
AC C
EP
independent experiments of bronchial epithelial air-liquid interface culture
M AN U
CDH26
TE D
ADAMTS9
SC
Table E1. IL-13 in vitro stimulation (IVS) signature a
ACCEPTED MANUSCRIPT
N (% of Type-2 group) [% of
moderate-severe asthma
airway-mucosa-
airway-mucosa-
periostin-high
periostin-low
(n=19)
(n=9)
4 (21%) [80%]
1 (11%) [20%]
13 (68%) [81%]
3 (33%) [19%]
6 (32%) [50%]
6 (67%) [50%]
FENO ≥ 35 ppb
13 (68%) [81%]
3 (33%) [19%]
FENO < 35 ppb
6 (32%) [50%]
6 (67%) [50%]
Blood EOS ≥ 300/mm3
4 (21%) [80%]
1 (11%) [20%]
row group]
P-valuea
FENO ≥ 35 ppb and Blood 3
FENO ≥ 35 ppb or Blood Eos ≥ 300/mm
3
3
Blood EOS < 300/mm
3
15 (79%) [65%]
EP
Blood Eos ≥ 300/mm
0.1139
1.0000
8 (89%) [35%]
16 (84%) [76%]
5 (56%) [24%]
Serum IgE low
3 (16%) [43%]
4 (44%) [57%]
Atopy (Phadiatop): Positive
17 (89%) [74%]
6 (75%) [26%]
Atopy (Phadiatop): Negative
2 (11%) [50%]
2 (25%) [50%]
AC C
Serum IgE high
a
0.1139
TE D
Neither FENO ≥ 35 ppb nor
0.3553
airway-mucosa-
periostin-high
periostin-low
0.1652
0.5583
P-value
(n=20)
(n=35)
6 (30%) [55%]
5 (15%) [45%]
0.0139
16 (80%) [57%]
12 (34%) [43%]
0.0018
4 (20%) [15%]
23 (66%) [85%]
10 (50%) [53%]
9 (26%) [47%]
10 (50%) [29%]
25 (74%) [71%]
12 (60%) [60%]
8 (23%) [40%]
8 (40%) [23%]
27 (77%) [77%]
19 (95%) [46%]
22 (65%) [54%]
1 (5%) [8%]
12 (35%) [92%]
18 (90%) [43%]
24 (69%) [57%]
2 (10%) [15%]
11 (31%) [85%]
M AN U
Eos ≥ 300/mm
airway-mucosa-
SC
Mild asthma
RI PT
Table E2. Comparison of airway-mucosal-periostin status and clinical biomarkers
P-value for comparison between airway-mucosa-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test)
0.1390
0.0089
0.0188
0.1023
ACCEPTED MANUSCRIPT
Table E3. Comparison of airway-mucosal Type-2-IL-13-IVS status and clinical biomarkers
airway-mucosal-
airway-mucosal-
airway-mucosal-IL-
row group]
IL-13-IVS-low
P-value
a
13-IVS-high (n=15)
(n=24)
FENO ≥ 35 ppb and Blood 4 (27%) [80%]
1 (8%) [20%]
0.3553
11 (73%) [69%]
5 (38%) [31%]
0.1248
4 (27%) [33%]
8 (62%) [67%]
FENO ≥ 35 ppb
11 (73%) [69%]
5 (38%) [31%]
FENO < 35 ppb
4 (27%) [33%]
8 (62%) [67%]
3
3
Blood Eos ≥ 300/mm3
3
11 (73%) [92%]
1 (8%) [8%]
Blood EOS < 300/mm
3
4 (27%) [25%]
12 (92%) [75%]
Serum IgE low
2 (13%) [29%]
8 (62%) [38%]
Atopy (Phadiatop): Positive
15 (100%) [65%]
8 (67%) [35%]
Atopy (Phadiatop): Negative
0 (0%) [0%]
4 (33%) [100%]
a
0.3333
0.1977
5 (38%) [71%]
AC C
13 (87%) [62%]
EP
Blood EOS ≥ 300/mm
Serum IgE high
0.1248
TE D
Neither FENO ≥ 35 ppb nor
6 (25%) [55%]
5 (16%) [45%]
0.0139
18 (75%) [64%]
10 (32%) [36%]
0.0026
6 (25%) [22%]
21 (68%) [78%]
12 (50%) [63%]
7 (23%) [37%]
12 (50%) [34%]
23 (77%) [66%]
12 (50%) [60%]
8 (26%) [40%]
12 (50%) [34%]
23 (74%) [66%]
21 (88%) [51%]
20 (67%) [49%]
3 (12%) [23%]
10 (33%) [77%]
21 (88%) [50%]
21 (68%) [50%]
3 (12%) [23%]
10 (32%) [77%]
M AN U
FENO ≥ 35 ppb or Blood Eos ≥ 300/mm
P-value
13-IVS-low (n=31)
(n=13)
Eos ≥ 300/mm
airway-mucosal-IL-
IL-13-IVS-high
SC
N (% of Type-2 group) [% of
moderate-severe asthma
RI PT
Mild asthma
0.0282
P-value for comparison between airway-mucosal-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test)
0.0506
0.0914
0.111
0.1110
ACCEPTED MANUSCRIPT
Table E4. Moderate-severe asthma airway-mucosal-Type-2-CCL26-associated gene expression
asthma vs. Healthy
asthma vs. Healthy
Low (moderate-severe asthma)
FoldChange
P-value
RI PT
Gene Symbol
Type-2-CCL26 High vs. Type-2-CCL26
FDR-BH
FoldChange
P-value
FDR-BH
FoldChange
P-value
FDR-BH
4.01
1.1E-22
2.6E-18
4.00
3.5E-25
8.5E-21
5.84
1.8E-09
8.65E-06
3.77
2.1E-07
0.0013
5.32
5.1E-12
4.1E-08
2.95
1.1E-07
0.0008
Type-2-CCL26 high moderate-severe asthma vs. Healthy (p<0.05, >2-fold) 223710_PM_at
CCL26
1.00
0.9748
0.9994
206224_PM_at
CST1
1.55
0.0277
0.548
210809_PM_s_at
POSTN
1.80
0.0003
0.0673
1555778_PM_a_at
POSTN
1.87
0.0003
223597_PM_at
ITLN1
1.36
0.0961
228481_PM_at
---
1.25
0.0044
201884_PM_at
CEACAM5
1.74
0.0148
225720_PM_at
SYNPO2
1.65
0.0505
225895_PM_at
SYNPO2
1.70
0.0384
232119_PM_at
SYNPO2
1.61
211430_PM_s_at
IGH@
-2.64
217036_PM_at
LOC100293679
-2.01
206326_PM_at
GRP
-1.99
217179_PM_x_at
---
219159_PM_s_at
SLAMF7
222102_PM_at
GSTA3
207430_PM_s_at
MSMB
210297_PM_s_at
MSMB
SC
a
Type-2-CCL26 High moderate-severe
M AN U
Probe Set ID
Type-2-CCL26 Low moderate-severe
4.96
1.7E-10
9.9E-07
2.66
3.5E-06
0.0168
0.7447
3.48
2.1E-06
0.0072
2.56
5.2E-05
0.1845
0.2811
2.47
1.2E-13
1.4E-09
1.98
7.4E-11
8.9E-07
0.4531
2.50
0.0027
0.5019
1.43
0.1799
0.9997
0.6524
2.15
0.0253
0.7389
1.30
0.3879
0.9997
0.598
2.18
0.0215
0.7326
1.29
0.4028
0.9997
TE D
0.0693
0.5142
2.05
0.0093
0.6183
1.27
0.3211
0.9997
0.0012
0.1578
-2.40
0.0243
0.7361
1.10
0.7837
0.9997
0.0096
0.402
-2.03
0.046
0.7865
-1.01
0.9801
0.9997
6.2E-06
0.0091
-2.05
0.0003
0.2092
-1.03
0.8573
0.9997
-1.81
0.0077
0.3625
-2.05
0.0148
0.6724
-1.13
0.6372
0.9997
-1.82
0.0036
0.2673
-2.43
0.0012
0.3531
-1.33
0.2299
0.9997
-1.42
0.0263
0.543
-2.07
0.0007
0.3120
-1.45
0.0467
0.9946
-1.14
0.5268
0.9677
-2.03
0.0127
0.6488
-1.78
0.0243
0.9318
-1.14
0.5483
0.9694
-2.14
0.0095
0.6274
-1.88
0.0168
0.9134
-1.27
0.4143
1.65
0.0579
AC C
EP
0.0216
Type-2-CCL26 low moderate-severe asthma vs. Healthy (p<0.05, >2-fold) 206291_PM_at
NTS
-2.09
0.0011
0.1550
0.9691
0.9997
ACCEPTED MANUSCRIPT
a
Gene Symbol
Type-2-CCL26 High moderate-severe
Type-2-CCL26 High vs. Type-2-CCL26
asthma vs. Healthy
asthma vs. Healthy
Low (moderate-severe asthma)
P-value
FDR-BH
FoldChange
P-value
FDR-BH
FoldChange
P-value
FDR-BH
IGK@
-2.01
0.0152
0.4571
-1.78
0.1262
0.8757
1.13
0.7167
0.9997
202988_PM_s_at
RGS1
-2.20
1.3E-05
0.0140
-1.97
0.0032
0.5019
1.11
0.5974
0.9997
211430_PM_s_at
IGH@
-2.64
0.0012
0.1578
-2.40
0.0243
0.7361
1.10
0.7837
0.9997
211645_PM_x_at
---
-2.07
0.0139
0.4424
-1.96
0.0830
0.8412
1.06
0.8735
0.9997
217036_PM_at
LOC100293679
-2.01
0.0096
0.4020
-2.03
0.0460
0.7865
-1.01
0.9801
0.9997
205792_PM_at
WISP2
2.04
6.4E-06
0.0091
1.99
0.0007
0.3120
-1.03
0.8711
0.9997
202376_PM_at
SERPINA3
2.09
0.0104
0.4087
1.62
0.2001
0.9213
-1.29
0.4459
0.9997
224840_PM_at
FKBP5
2.51
2.4E-05
0.0208
1.82
0.0301
0.7524
-1.38
0.1910
0.9997
1556069_PM_s_at
HIF3A
2.39
0.0003
0.0737
1.63
0.1149
0.8659
-1.47
0.1673
0.9997
204560_PM_at
FKBP5
2.28
7.4E-07
0.0044
1.36
0.1301
0.8779
-1.67
0.0060
0.8506
224856_PM_at
FKBP5
2.98
9.4E-07
0.0045
1.75
0.0427
0.7840
-1.70
0.0319
0.9473
222124_PM_at
HIF3A
2.74
9.9E-05
0.0405
1.47
0.2379
0.9325
-1.86
0.0370
0.9665
244697_PM_at
---
2.43
1.28
0.4181
0.9691
-1.89
0.0236
0.9318
a
0.0003
M AN U
SC
216576_PM_x_at
TE D
FoldChange
RI PT
Probe Set ID
Type-2-CCL26 Low moderate-severe
0.0673
For the airway-mucosa biospy microarray dataset, probe sets passing the significance criteria for the indicated comparisons in section headers are displayed with the fold-change,
AC C
EP
nominal p-value, and FDR-BH for the comparisons indicated for each of the comparisons listed in the column headers
ACCEPTED MANUSCRIPT
Table E5. Mild asthma Type-2-CCL26-associated gene expression in airway-mucosa airway-mucosa -CCL26 High vs. airway-mucosa -CCL26 High (mild
asthma) vs. Healthy
asthma) vs. Healthy
Gene Symbol
FoldChange
P-value
FDR-BH
RI PT
Probe Set ID
airway-mucosa-CCL26 Low (mild
FoldChange
SC
Type-2-CCL26 high mild asthma vs. Healthy (p<0.05, >2-fold)
P-value
airway-mucosa -CCL26 Low (mild asthma)
FDR-BH
FoldChange
P-value
FDR-BH
4.59
2.8E-12
6.8E-08
3.83
5.8E-09
0.0001
2.85
5.7E-10
4.5E-06
2.90
1.5E-08
0.0002
0.9899
7.90
2.2E-10
2.7E-06
7.12
3.4E-08
0.0003
0.9956
2.05
6.0E-08
2.0E-04
2.09
7.4E-07
0.0045
0.9813
7.39
2.9E-09
1.2E-05
5.16
4.8E-06
0.023
0.9813
7.82
1.8E-09
8.4E-06
4.85
1.1E-05
0.0432
0.9813
5.47
7.5E-10
4.5E-06
3.17
4.5E-05
0.1342
0.9813
2.05
0.0001
0.1304
1.70
0.0112
0.9992
CCL26
1.20
0.3234
0.9813
228481_PM_at
---
-1.02
0.8974
0.9959
206224_PM_at
CST1
1.11
0.7146
206994_PM_at
CST4
-1.02
0.8745
210809_PM_s_at
POSTN
1.43
0.2431
1555778_PM_a_at
POSTN
1.61
0.1243
223597_PM_at
ITLN1
1.73
0.0303
207741_PM_x_at
TPSAB1
1.21
0.3179
207134_PM_x_at
TPSAB1 /// TPSB2
1.33
217023_PM_x_at
TPSAB1
1.34
210084_PM_x_at
TPSAB1
1.33
205683_PM_x_at
TPSAB1
1.40
201884_PM_at
CEACAM5
201058_PM_s_at
MYL9
1558438_PM_a_at
IGHA1
216474_PM_x_at
TPSAB1 /// TPSB2
205624_PM_at
CPA3
225721_PM_at
SYNPO2
228728_PM_at 204938_PM_s_at
TE D
M AN U
223710_PM_at
0.9813
2.24
0.0002
0.1645
1.69
0.0296
0.9992
0.1501
0.9813
2.08
0.0002
0.1645
1.55
0.0433
0.9992
0.1816
0.9813
2.08
0.0003
0.204
1.57
0.0448
0.9992
0.1227
0.9813
2.19
0.0002
0.1645
1.56
0.0549
0.9992
-1.01
0.9781
0.9997
2.01
0.0309
0.6606
2.03
0.0572
0.9992
1.23
0.4822
0.9813
2.26
0.004
0.5704
1.83
0.0581
0.9992
1.48
0.0666
0.9813
2.22
0.0001
0.1203
1.50
0.0678
0.9992
1.42
0.1591
0.9813
2.25
0.0007
0.3161
1.58
0.0845
0.9992
1.87
0.0172
0.9813
2.83
3.92E-05
0.0857
1.52
0.1239
0.9992
1.21
0.5651
0.9813
2.04
0.0205
0.6561
1.69
0.1351
0.9992
C7orf58
1.28
0.4399
0.9813
2.09
0.0124
0.6464
1.64
0.1389
0.9992
PLN
1.28
0.4081
0.9813
2.01
0.0123
0.646
1.57
0.152
0.9992
AC C
EP
0.2084
ACCEPTED MANUSCRIPT
airway-mucosa -CCL26 High vs. airway-mucosa-CCL26 Low (mild
airway-mucosa -CCL26 High (mild
asthma) vs. Healthy
asthma) vs. Healthy
airway-mucosa -CCL26 Low (mild Gene Symbol
asthma) FoldChange
P-value
FDR-BH
FoldChange
P-value
FDR-BH
FoldChange
P-value
FDR-BH
0.0176
0.6505
1.82
0.1815
0.9992
0.0107
0.6315
1.84
0.1884
0.9992
0.0261
0.6561
1.57
0.2522
0.9992
SORBS1
1.41
0.4228
0.9813
2.56
202274_PM_at
ACTG2
1.55
0.3177
0.9813
2.86
227843_PM_at
NDE1
1.38
0.3899
0.9813
2.16
207961_PM_x_at
MYH11
1.47
0.2741
0.9813
201497_PM_x_at
MYH11
1.45
0.2897
0.9813
225720_PM_at
SYNPO2
1.56
0.2353
0.9813
209071_PM_s_at
RGS5
1.48
0.1939
225895_PM_at
SYNPO2
1.43
0.3510
200795_PM_at
SPARCL1
1.47
0.2192
200974_PM_at
ACTA2
1.56
0.2152
211340_PM_s_at
MCAM
1.71
0.1033
201540_PM_at
FHL1
1.52
212077_PM_at
CALD1
1.69
209189_PM_at
FOS
-1.12
206336_PM_at
CXCL6
220542_PM_s_at
PLUNC
SC
218087_PM_s_at
0.0182
0.6505
1.48
0.294
0.9992
2.12
0.0204
0.6561
1.46
0.2969
0.9992
2.35
0.0147
0.6505
1.50
0.3014
0.9992
0.9813
2.04
0.0112
0.6347
1.38
0.3097
0.9992
0.9813
2.11
0.0342
0.6606
1.48
0.3279
0.9992
0.9813
2.02
0.0162
0.6505
1.37
0.3374
0.9992
0.9813
2.16
0.0209
0.6561
1.38
0.389
0.9992
0.9813
2.21
0.0092
0.6205
1.30
0.4476
0.9992
TE D
M AN U
2.18
0.9813
2.01
0.0347
0.6606
1.33
0.4507
0.9992
0.1428
0.9813
2.10
0.0252
0.6561
1.24
0.5653
0.9992
0.6947
0.9886
-2.04
0.0096
0.6209
-1.82
0.0563
0.9992
-1.31
0.2771
0.9813
-2.13
0.0013
0.3646
-1.63
0.0619
0.9992
-1.35
0.5507
0.9813
-3.46
0.0094
0.6205
-2.55
0.0837
0.9992
EP
0.2457
AC C
a
RI PT
Probe Set ID
For the airway-mucosa biospy microarray dataset, probe sets passing the significance criteria of p<0.05 and 2-fold for mild asthma Type-2-CCL26 high vs. healthy comparison
are displayed with the fold-change, nominal p-value, and FDR-BH for each of the comparisons listed in the column headers
ACCEPTED MANUSCRIPT
Table E6. Comparison of airway-mucosa -periostin status and sputum inflammatory cells
N (% of Type-2 group) [% of row group]
airway-mucosal-
airway-mucosal-
periostin-high
periostin-low
(n=14)
(n=5)
P-value
airway-mucosal-
airway-mucosal-
periostin-high
periostin-low
(n=13)
(n=26)
9 (69%) [43%]
12 (46%) [57%]
4 (31%) [22%]
14 (54%) [78%]
0 (0%) [0%]
Sputum EOS < 3%
10 (71%) [67%]
5 (100%) [33%]
6 (43%) [67%]
3 (60%) [33%]
2 (15%) [18%]
9 (35%) [82%]
4 (29%) [67%]
2 (40%) [33%]
2 (15%) [29%]
5 (19%) [71%]
0 (0%) [na]
0 (0%) [na]
2 (15%) [40%]
3 (12%) [60%]
4 (29%) [100%]
0 (0%) [0%]
7 (54%) [44%]
9 (35%) [56%]
M AN U
PMN<60%
Sputum Mixed
TE D
Gran.,PMN≥60%, EOS≥3% Sputum EOS≥3%, PMN<60%
SC
4 (29%) [100%]
Sputum PMN ≥60% EOS<3%
0.530
a
Sputum EOS ≥ 3%
Sputum Pauci EOS<3%,
EP
P-value for comparison between airway-mucosal-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test)
AC C
a
moderate-severe asthma
RI PT
Mild asthma
P-value
0.307
ACCEPTED MANUSCRIPT
Table E7. Comparison of airway-mucosal-IL-13-IVS status and sputum inflammatory cells
N (% of Type-2 group) [% of row group]
airway-mucosal-
airway-mucosal-
IL-13-IVS-high
IL-13-IVS-low
(n=11)
(n=8)
P-value
airway-mucosal-
airway-mucosal-
IL-13-IVS-high
IL-13-IVS-low
(n=17)
(n=22)
11 (65%) [52%]
10 (45%) [48%]
6 (35%) [33%]
12 (55%) [67%]
0 (0%) [0%]
Sputum EOS < 3%
7 (64%) [47%]
8 (100%) [53%]
4 (36%) [44%]
5 (63%) [56%]
4 (24%) [36%]
7 (32%) [64%]
3 (27%) [50%]
3 (38%) [50%]
2 (12%) [29%]
5 (23%) [71%]
0 (0%) [na]
0 (0%) [na]
3 (18%) [60%]
2 (9%) [40%]
4 (36%) [100%]
0 (0%) [0%]
8 (47%) [50%]
8 (36%) [50%]
M AN U
PMN<60%
Sputum Mixed
TE D
Gran.,PMN≥60%, EOS≥3% Sputum EOS≥3%, PMN<60%
SC
4 (36%) [100%]
Sputum PMN ≥60% EOS<3%
0.103
a
Sputum EOS ≥ 3%
Sputum Pauci EOS<3%,
EP
P-value for comparison between airway-mucosal-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test)
AC C
a
moderate-severe asthma
RI PT
Mild asthma
P-value
0.334
ACCEPTED MANUSCRIPT
Table E8. Pre-specified vs. modeled optimal thresholds for classifiers of airway-mucosal-CCL26-high status Pre-specified threshold
a
Optimal threshold
b
c
blood eosinophils, cells/ul
Sensitivity
Specificity
PPV
NPV
Threshold
35
69%
76%
48%
89%
38
300
77%
76%
50%
91%
309 304
serum CCL17, pg/ml
d
285
85%
76%
52%
94%
serum CCL26, pg/ml
e
775
58%
90%
64%
88%
Specificity
PPV
NPV
69%
76%
48%
89%
24.1%
77%
76%
50%
91%
24.1%
85%
81%
58%
94%
24.1%
93%
70%
88%
22.6%
1595
58%
M AN U
a
c
Sensitivity
SC
FENO, ppb
Prevalence Threshold
RI PT
airway-mucosal-CCL26-high classifier
Model statistics for classification of airway-mucosa-CCL26-high status was based on pre-specified thresholds for the indicated classifer. FENO and bEOS
thresholds were pre-specified based on commonly used cut-offs; serum CCL17 and CCL26 thresholds were set as the 75th %ile of the healthy control distribution. b
Model statistics for classification of airway-mucosal-CCL26-high status was based on optimal thresholds established from ROC analyses, selecting the
TE D
thresholds for the classifiers that yielded the maximum for the sum of sensitivity and specificity.
FENO-high/-low and bEOS-high/-low assignments were equivalent using the pre-specified vs. optimal thresholds.
d
2 airway-mucosla-CCL26-low subjects assigned as 'serum CCL17-high' with the pre-specified threshold considered 'serum CCL17-low' with the optimal
threshold.
1 airway-mucosal-CCL26-low subject assigned as 'serum CCL26-high' with the pre-specified threshold considered 'serum CCL26-low' with the optimal
threshold.
AC C
e
EP
c
ACCEPTED MANUSCRIPT
Table E9. Airway-mucosal-CCL26 diagnostic characteristics of accessible biomarkers in moderate-severe asthma, multi-predictor models
Multi-predictor, discrete
Sensitivity
Specificity
AUC
sCCL26-high, sCCL17-high, (FENO≥35ppb or bEOS ≥300/mm3)
92%
93%
0.97
sCCL26 (log), sCCL17 (log), FENO (log), bEOS (log)
50%
93%
0.92
PPV
NPV
Prevalence†
79%
98%
23%
67%
86%
23%
SC
Multi-predictor, continuous
Classifiers
RI PT
airway-mucosal-CCL26-high classification model *
M AN U
* Logistic regression model statistics for classification of moderate-severe asthmatics to airway-mucosal-CCL26-high status, with 10-fold crossvalidation (on full population) to select model parameters and statistics based on a 50% probability cut-off: AUC, area-under-curve for receiver operating characteristic; NPV, negative predictive value; PPV, positive predictive value. † Prevalence based on the proporaon of airway-mucosal-CCL26-high subjects in the ADEPT moderate-severe asthma group, with PPV and NPV statistics based on this prevalence. ADEPT is a cross-sectional study that has rigid enrollment criteria, therefore the prevalence in other asthma
AC C
EP
TE D
study populations may vary based on recruitment criteria.
ACCEPTED MANUSCRIPT
Figure Legends Figure E1. Segregation of Type-2 high versus low by airway-mucosal CCL26, periostin, and IL-13-IVS gene expression.
RI PT
Airway mucosal biopsy gene expression is presented for CCL26, and periostin as log2 intensity from microarray experiments and for IL-13-IVS as GSVA enrichment score (GSVA ES), in the healthy nonatopic, mild, moderate and severe asthma cohorts, stratified by Type-2 high (red) vs. low (blue) groups defined by (A) CCL26 expression, (B) periostin expression, and (C) IL-13-IVS enrichment. The highlighted
Figure E2. Longitudinal evaluation of serum CCL26 levels.
SC
horizontal (y-axis) line indicates the threshold for Type-2 high vs. low status of the plotted biomarker.
M AN U
Serum CCL26 levels (y-axis) are shown for asthma subjects (all severities), stratified by visit time point and by baseline sCCL26 category (high, above or low, below the 75th percentile of healthy control population). Filled and open symbols respectively indicate samples above and below the threshold at 775 pg/ml for high and low sCCL26 status, based on 75th percentile of healthy control population from baseline analysis. Cramer’s V statistic for concordance of serum CCL26 categories at baseline vs. 3, 6,
TE D
and 12 month visits were 0.87, 1.00, and 0.87, respectively.
Figure E3. Correlations between Type-2 inflammation and eosinophilic phenotype biomarkers. The Spearman’s correlation coefficient (RSp) and associated p-values are shown in the bottom-left and topright halves, respectively, of tables for (A) mild asthma and (B) moderate-severe asthma groups for the
EP
indicated biomarkers. Correlation coefficient values are shaded as indicated in key to right of tables. Pvalues <0.00143 pass the Bonferroni family-wise error rate of 0.05 for 35 pair-wise tests and are
AC C
indicated in bold.
Figure E4. Correlations between Type-2 inflammation and eosinophilic phenotype biomarkers (samples sizes).
Sample sizes for the pair-wise Spearman’s correlation tests reported in Figure E3, for mild asthma (topright half) and moderate-severe asthma (bottom-left half).
RI PT
ACCEPTED MANUSCRIPT
Supplement References
5.
6.
7.
8.
9.
SC
M AN U
4.
TE D
3.
EP
2.
Kelly MM, Efthimiadis A, Hargreave FE. Induced sputum : selection method. Methods Mol Med 2001; 56:77-91. Komiya A, Nagase H, Yamada H, Sekiya T, Yamaguchi M, Sano Y, et al. Concerted expression of eotaxin-1, eotaxin-2, and eotaxin-3 in human bronchial epithelial cells. Cell Immunol 2003; 225:91-100. Syed F, Huang CC, Li K, Liu V, Shang T, Amegadzie BY, et al. Identification of interleukin-13 related biomarkers using peripheral blood mononuclear cells. Biomarkers 2007; 12:414-23. Woodruff PG, Modrek B, Choy DF, Jia G, Abbas AR, Ellwanger A, et al. T-helper type 2-driven inflammation defines major subphenotypes of asthma. Am J Respir Crit Care Med 2009; 180:388-95. Choy DF, Modrek B, Abbas AR, Kummerfeld S, Clark HF, Wu LC, et al. Gene expression patterns of Th2 inflammation and intercellular communication in asthmatic airways. J Immunol 2011; 186:1861-9. Li G, Oparil S, Sanders JM, Zhang L, Dai M, Chen LB, et al. Phosphatidylinositol-3-kinase signaling mediates vascular smooth muscle cell expression of periostin in vivo and in vitro. Atherosclerosis 2006; 188:292-300. Horiuchi K, Amizuka N, Takeshita S, Takamatsu H, Katsuura M, Ozawa H, et al. Identification and characterization of a novel protein, periostin, with restricted expression to periosteum and periodontal ligament and increased expression by transforming growth factor beta. J Bone Miner Res 1999; 14:1239-49. Woodruff PG, Boushey HA, Dolganov GM, Barker CS, Yang YH, Donnelly S, et al. Genome-wide profiling identifies epithelial cell genes associated with asthma and with treatment response to corticosteroids. Proc Natl Acad Sci U S A 2007; 104:15858-63. Hanzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNAseq data. BMC Bioinformatics 2013; 14:7.
AC C
1.
ACCEPTED MANUSCRIPT
Figure E1. Segregation of Type-2 high versus low by airway-mucosaL CCL26, periostin, and IL-13-IVS gene expression.
Healthy Mild
RI PT
POSTN, log2log2 intensity EBBX-POSTN, intensity
IL-13 IVS,IVS, enrichment EBBX-IL-13 enrichmentscore score
A
CCL26:
Cohort:
Cohort:
M AN U
CCL26, log2 intensity
IL-13 IVS, enrichment score
B
POSTN:
POSTN:
TE D
AC C
EP
CCL26, log2 intensity IL13 IVS: Cohort:
POSTN, log2 intensity
Cohort:
Cohort:
C
SC
CCL26:
IL13 IVS: Cohort:
Moderate Severe
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Figure E2. Longitudinal evaluation of serum CCL26 levels.
ACCEPTED MANUSCRIPT
Figure E3. Correlations between Type-2 inflammation and eosinophilic phenotype biomarkers
serum IgE
serum CCL26
serum CCL17
RI PT
Blood Eos.
Sputum Eos.
FENO
Submucosal Eos.
RSp <0 0.00 0.15 0.30 0.45 0.60 0.75 0.90 1.00
serum IgE
serum CCL26
serum CCL17
SC Blood Eos.
Sputum Eos.
Submucosal Eos.
EBBX IL13 IVS
TE D
0.66 0.55 0.46 0.36 0.25 0.41 0.50 0.41 0.31
AC C
EBBX CCL26 EBBX POSTN EBBX IL13 IVS FENO Submucosal Eos. Sputum Eos. Blood Eos. serum CCL17 serum CCL26 serum IgE
<0.0001 <0.0001 0.0005 0.0149 0.1253 0.0020 0.0001 0.0021 0.0239 0.0002 <0.0001 0.0455 0.0516 0.0005 0.0010 0.0668 0.0208 0.48 0.0002 0.0104 0.1003 0.0421 0.0129 0.0835 0.1330 0.54 0.49 <0.0001 <0.0001 0.0002 0.0143 0.1738 0.0007 0.30 0.38 0.58 0.0834 0.0153 0.0099 0.0181 0.2484 0.31 0.27 0.53 0.32 0.0009 0.3865 0.8188 0.0656 0.45 0.28 0.36 0.36 0.37 0.1420 0.0762 0.0156 0.44 0.34 0.24 0.38 0.10 0.15 <0.0001 0.9040 0.25 0.24 0.14 0.36 0.03 0.18 0.55 0.5160 0.31 0.21 0.33 0.18 0.21 0.24 0.01 0.07
EP
Correlation (RSp)
EBBX CCL26
P-value
EBBX POSTN
B. Moderate-severe asthma
M AN U
0.85 0.69 0.41 0.47 0.05 0.36 0.37 0.13 0.27
<0.0001 <0.0001 0.0305 0.0346 0.8543 0.0635 0.0609 0.5512 0.1597 0.0001 0.0435 0.2179 0.3212 0.0706 0.0062 0.6190 0.0610 0.67 0.0017 0.1236 0.1936 0.1624 0.0977 0.9574 0.0598 0.38 0.57 0.2413 0.1478 0.1475 0.3969 0.2115 0.2969 0.29 0.36 0.27 0.6667 0.0600 0.7240 0.0115 0.1959 0.24 0.31 0.26 -0.12 0.1062 0.7399 0.4531 0.3118 0.35 0.27 0.20 0.42 0.29 0.7787 0.7873 0.2023 0.51 0.33 0.12 -0.08 -0.06 0.04 0.0002 0.8626 0.10 -0.01 0.18 -0.57 -0.14 -0.04 0.51 0.9788 0.36 0.36 0.15 0.29 0.18 0.18 0.02 0.00
FENO
EBBX CCL26 EBBX POSTN EBBX IL13 IVS FENO Submucosal Eos. Sputum Eos. Blood Eos. serum CCL17 serum CCL26 serum IgE
EBBX IL13 IVS
Correlation (RSp)
EBBX CCL26
P-value
EBBX POSTN
A. Mild asthma
RSp <0 0.00 0.15 0.30 0.45 0.60 0.75 0.90 1.00
ACCEPTED MANUSCRIPT
EP AC C
31 46 44 43 45
25 25 25 48 19 31 48 48
101 98 102
serum IgE
27 27 27 51 20 32 51
RI PT
Blood Eos.
Sputum Eos. 78 77 74 76
28 28 28 52 21 32
serum CCL26
45 78 102 101 98 102
19 19 19 32 16
serum CCL17
54 45 39 55 54 53 54
20 20 20 21
SC
55 54 45 39 55 54 53 54
28 28 28
Submucosal Eos.
28 28
M AN U
28
FENO
55 55 54 45 39 55 54 53 54
EBBX IL13 IVS
EBBX CCL26 EBBX POSTN EBBX IL13 IVS FENO Submucosal Eos. Sputum Eos. Blood Eos. serum CCL17 serum CCL26 serum IgE
TE D
ModerateSevere
EBBX CCL26
Mild
EBBX POSTN
Figure E4. Correlations between Type-2 inflammation and eosinophilic phenotype biomarkers (sample sizes)
100 102
99
28 28 28 52 21 32 52 51 48
S0091-6749(17)30007-6
DOI:
10.1016/j.jaci.2016.11.038
Reference:
YMAI 12573
To appear in:
Journal of Allergy and Clinical Immunology
Received Date: 25 February 2016 Revised Date:
6 November 2016
Accepted Date: 21 November 2016
Please cite this article as: Silkoff PE, Laviolette M, Singh D, FitzGerald JM, Kelsen S, Backer V, Porsbjerg CM, Girodet P-O, Berger P, Kline JN, Chupp G, Susulic VS, Barnathan ES, Baribaud F, Loza MJ, the Airways Disease Endotyping for Personalized Therapeutics (ADEPT) study investigators, Identification of Airway-Mucosal Type-2 inflammation by Clinical Biomarkers in Asthma, Journal of Allergy and Clinical Immunology (2017), doi: 10.1016/j.jaci.2016.11.038. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
ACCEPTED MANUSCRIPT
Identification of Airway-Mucosal Type-2 inflammation by Clinical Biomarkers in
2
Asthma
3 4 5 6
Philip E Silkoff MD1, Michel Laviolette MD2, Dave Singh MD3, J Mark FitzGerald MD4, Steven Kelsen MD 5 , Vibeke Backer MD 6 , Celeste M Porsbjerg MD 6, Pierre-Olivier Girodet MD 7 , Patrick Berger MD7 , Joel N Kline MD8, Geoffrey Chupp MD 9 , Vedrana S Susulic PhD1, Elli ot S Bar n ath a n M D 1 , Frédéric Baribaud PhD1, and Matthew J Loza PhD 1
7
& the Airways Disease Endotyping for Personalized Therapeutics (ADEPT) study investigators. Janssen Research & Development LLC, Spring House, USA
2
SC
1
Institut Universitaire de Cardiologie et Pneumologie de Québec (IUCPQ), 2725, Chemin Ste-Foy, Québec, Canada, G1V 4G5. Email: [email protected] 3
M AN U
Cen tre f o r R es p i ra tory Med i ci n e an d Al l ergy, the U ni versi ty o f Ma n chester, Medi ci nes E val u ati o n Un i t, Un i v ersi ty Ho s pi tal of Sou th Man ch es ter NHS Fou n da ti on Tru s t, So u thm o or R oa d , Ma nch es ter M2 3 9 Q Z , United Kingdom; Email: [email protected] 4
Institute for Heart and Lung Health, The Lung Centre, 7th Floor, Gordon and Leslie Diamond Health Care Centre, 2775 Laurel Street, Vancouver, B.C., Canada, V5Z 1M9. Email: [email protected] 5
Department of Thoracic Medicine and Surgery, Temple University School of Medicine, 3401 N. Broad St., Philadelphia PA 19140. Email: [email protected] 6
7
TE D
Respiratory Research Unit, Department of Respiratory Medicine, Bispebjerg University Hospital, Bispebjerg bakke 23, DK-2400, Copenhagen NV, Denmark. Email: [email protected], p o r s b j e r g @ d a d l n e t . d k Univ. Bordeaux, Centre de Recherche Cardio-Thoracique de Bordeaux, U1045, CIC 1401, F-33000 Bordeaux, France. Email: [email protected]; [email protected] 8
Division of Pulmonary, Critical Care, and Occupational Medicine, University of Iowa, W219B GH UIHC, 200 Hawkins Drive, Iowa City, IA, 52242, USA. Email: [email protected] 9Yale School of Medicine, TAC 441, 300 Cedar Street, New Haven, CT 06520, USA. Email:
EP
[email protected]
Corresponding Author: Philip E Silkoff 715 Bryn Mawr Avenue Penn Valley, PA, 19072, USA Tel: 6103109142; Fax 6106671698 Email: [email protected]
AC C
8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40
RI PT
1
The study was funded by Janssen R&D Inc. Spring House, PA, USA Author contributions: Conception and design: MJL, VSS, ESB, and PES Acquisition of data: IS, ML, DS, MF, SL, SK, AE, AL-S, GC, VB, CP, P-OG, PB, RL, JNK, MD, WJC, AH, SK, PC (the ADEPT investigators) Analysis and interpretation: All authors 1
ACCEPTED MANUSCRIPT
Drafting the manuscript for important intellectual content: MSL, and PES Approval of the final version: All authors Funding disclosure: Support was provided by Janssen Research & Development, LLC
44
Short Running Head: Clinical biomarkers and Type2 Endotyping in Asthma
45
Clinicaltrials.gov identifier: NCT01274507
AC C
EP
TE D
M AN U
SC
46
RI PT
41 42 43
2
ACCEPTED MANUSCRIPT
Abstract
48
Background and Objective: The Airways Disease Endotyping for Personalized Therapeutics
49
(ADEPT) study profiled mild, moderate and severe asthma, and non-atopic healthy controls. We
50
explored this dataset to define Type-2 inflammation based on airway-mucosal IL-13-driven gene
51
expression and how this related to clinically-accessible biomarkers.
52
Methods: IL-13-driven gene expression was evaluated in several human cell lines. We then
53
defined Type-2 status in 25 healthy subjects, 28 mild, 29 moderate, and 26 severe asthmatics,
54
based on airway-mucosal expression of 1) CC-motif chemokine ligand (CCL)-26, (the most
55
differentially expressed gene), 2) periostin, or 3) a multi-gene IL-13 in-vitro signature (IVS).
56
Clinically accessible biomarkers included fractional exhaled nitric oxide (FENO), blood
57
eosinophils (bEOS), serum CCL26, and serum CCL17.
58
Results: Expression of airway-mucosal-CCL26, periostin, and IL-13-IVS all provided segregation
59
into Type-2-high and -low asthmatics, but in the ADEPT population, CCL26 was optimal. All
60
airway-mucosal-CCL26-high subjects with moderate-severe asthma were FENO-high (≥35 ppb)
61
and/or blood eosinophils-high (≥300cells/mm3), compared to a minority (36%) of airway-
62
mucosal- CCL26-low subjects. A combination of FENO, blood eosinophils, serum CCL17 and
63
CCL26 had 100% positive-predictive-value and 87% negative-predictive-value for airway-
64
mucosal-CCL26-high status. Clinical variables did not differ between Type-2 high and –low
65
subjects. Eosinophilic inflammation was associated with, but not limited to, airway-mucosal
66
Type-2 gene expression.
67
Conclusion: A panel of clinical biomarkers accurately classified Type-2 status based on airway-
68
mucosa CCL26, perisotin or IL-13-IVS gene expression. Use of FENO, blood eosinophils and
AC C
EP
TE D
M AN U
SC
RI PT
47
3
ACCEPTED MANUSCRIPT
serum markers e.g. CCL26, CCL17 in combination may allow patient selection for novel Type-2
70
therapeutics.
71
Abstract Word count: 241
72
Key Messages:
RI PT
69
•
Type-2 inflammation phenotype has emerged as the target of novel asthma therapies.
74
•
Airway-mucosal gene expression of CCL26, periostin or a multigene IL-13 in vitro
SC
73
signature were used to segregate Type-2 high from Type-2 low subjects. In our dataset,
76
airway-mucosal-CCL26 was optimal.
77
•
M AN U
75
Combinations of clinical biomarkers, FENO, blood eosinophils, serum CCL17 and CCL26 optimally identified Type-2 inflammation based on airway-mucosal CCL26 with similar
79
findings for periostin and IL-13-IVS.
TE D
78
Capsule summary
81
A panel of clinically-accessible biomarkers (FENO, blood eosinophils, serum CCL17 and CCL26)
82
identified asthma patients with an airway ‘Type-2-high’ or ‘Type-2-low’ phenotype based on
83
expression of airway-mucosal-CCL26, periostin, or an IL-13 multigene signature. This
84
combination may prove useful in selecting patients for novel therapies targeting Type-2
85
inflammation.
86
Key words: asthma, Type-2 inflammation, phenotypes, airway-mucosal gene expression, biomarkers
AC C
EP
80
87 4
ACCEPTED MANUSCRIPT
List of Abbreviations Asthma Control Questionnaire 7
ADEPT
Airways Disease Endotyping for Personalized Therapeutics
AHR
airway hyper-responsiveness
AQLQ
Asthma Quality of Life Questionnaire
bEOS
blood eosinophils
BDR
bronchodilator reversibility
CLCA1
calcium-activated chloride channel regulator 1
CCL
C-C motif chemokine ligand
FDR
false discovery rate
FENO
Fractional concentration of exhaled nitric oxide
FEV1
forced expired volume in 1 second
GSVA
gene set variation analysis
ICS
inhaled corticosteroids
Ig
immunoglobulin
IL
Interleukin
IL-4R
interleukin 4 receptor
IVS
in-vitro signature
mAb
monoclonal antibody
NPV
negative predictive values
PC20
POSTN ppb
EP
TE D
M AN U
SC
RI PT
ACQ7
AC C
88
The provocative concentration of methacholine resulting in a 20% or greater fall in the forced expired volume in 1 second periostin
parts per billion
PPV
positive predictive value
Pre-BD
pre-bronchodilator
RNA
ribonucleic acid
5
ACCEPTED MANUSCRIPT
s
serum
SD
standard deviation
SERPINB2
Serpin peptidase inhibitor, clade B, member 2 (also known as plasminogen activator inhibitor-2)
sIgE
serum immunoglobulin E
spEOS
sputum eosinophils
spNEU
sputum neutrophils
TGF
transforming growth factor
TNF
tumor necrosis factor
AC C
EP
TE D
M AN U
SC
RI PT
89
6
ACCEPTED MANUSCRIPT
INTRODUCTION
91
Asthma is a heterogeneous disease, characterized by chronic airway inflammation 1. While
92
there are many proposed asthma phenotypes, the underlying phenotypes are poorly
93
understood, hard to identify, and tailored anti-asthma treatment an even further challenge.
94
Type-2 inflammation in asthma (also termed T-helper 2 inflammation) refers to that driven by
95
Type-2 cytokines e.g. IL-4, IL-5 and IL-13 secreted by CD4+ T helper 2 cells as recently reviewed
96
2
97
and multiple gene signatures in airway brushings and sputum inflammatory cells have recently
98
been used to define Type-2-high or –low status 4-6.
99
The most effective therapies for asthma are anti-inflammatory, including inhaled
SC
RI PT
90
M AN U
. Type-2 innate lymphoid cells have also been proposed to secrete the same cytokines 3. Single
corticosteroids (ICS), and monoclonal antibodies (mAbs), but these do not work in all patients.
101
For example, the response to ICS is associated with a high fractional exhaled nitric oxide (FENO)
102
7
103
5
104
(anti-IL-4 receptor) 10, mepolizumab and reslizumab (anti-IL-5) 11, 12, or lebrikizumab (anti-IL-13)
105
13
106
biomarkers including blood eosinophils (bEOS), spEOS, FENO, or serum periostin9-13.
107
Notably, despite these patient-selection strategies, severe asthma is often incompletely
108
responsive to these and other currently-available therapies 14, 15. This may be due to difficulty
109
in defining phenotypes including the Type-2 phenotype leading to selection of inappropriate
110
therapy, or additional mechanisms, alone or in addition to Type-2 inflammation, that are
TE D
100
, high sputum eosinophils (spEOS) 8 or a 3-gene ‘Type-2-high’ signature in epithelial brushings
EP
. Similarly, for the following mAbs, omalizumab (anti-immunoglobulin E (IgE), 9, dupilumab
AC C
, optimal efficacy is seen in severe asthma with Type-2 inflammation identified with
7
ACCEPTED MANUSCRIPT
driving pathology. For example, in a recent report for lebrikizumab, even periostin-high subjects
112
had a 95% CL range for improvement in FEV1 between 1.0-15.4% 13 illustrating that single
113
biomarkers cannot capture complex biological networks.
114
The previously reported ADEPT study 16, allowed us to evaluate airway-mucosal gene
115
expression associated with Type-2 inflammation, including periostin, which has been associated
116
with enhanced responses to the anti-IL-13 mAb, lebrikizumab 13, CCL26, which in our hands was
117
the most highly expressed gene in IL-13 treated human cells lines, and a multigene in vitro
118
signature from the same cell lines (IL-13-IVS). We also ascertained if Type-2 gene expression
119
profiles for CCL26, IL-13-IVS and periostin could be classified using clinically accessible
120
biomarkers, namely FENO, bEOS, and less-commonly used serum markers, namely CCL17/TARC,
121
and CCL26/eotaxin-3, which are Type-2 chemokines reliably measured in serum.
SC
M AN U
TE D EP AC C
122
RI PT
111
8
ACCEPTED MANUSCRIPT
Methods
124
Study design and population
125
The ADEPT study design and population (clinicaltrials.gov registration number NCT01274507)
126
have been described in detail
127
medications, with a pre-bronchodilator [pre-BD] forced expired volume in 1 second [FEV1]>80%
128
predicted), moderate (low-moderate dose ICS, pre-BD FEV1 60-<80% predicted), and severe
129
(high-dose ICS, pre-BD FEV1 50 -<80% predicted) asthma subjects, were enrolled. Healthy
130
subjects and a subset of asthmatic subjects underwent bronchoscopy. All ADEPT subjects with
131
good quality airway-mucosal tissue from biopsies were included in the biopsy assessment
132
details below. Clinical assessments included spirometry, bronchodilator reversibility (BDR),
133
methacholine airway hyper-responsiveness (AHR), the Asthma Control Questionnaire (ACQ7) 17,
134
and the Asthma Quality of Life Questionnaire (AQLQ)
135
specific institutional review boards and all subjects signed an informed consent form. Healthy
136
and asthmatic subjects attended screening, baseline and bronchoscopy visits, while asthmatic
137
subjects also attended biomarker visits at 3, 6 and 12 months.
138
Biomarker assessments
139
Bioanalysis methods are detailed in the Online Supplement section E1. Briefly, gene expression
140
in RNAlater®-preserved biopsy samples was analyzed by microarray (HG-U133+PM platform,
141
Affymetrix, Santa Clara, CA) and histology of biopsies was performed by Pantomics, Inc.
142
(Richmond, CA). Induced sputum was collected, processed via the plug selection method 19, and
RI PT
123
16
18
. The study was approved by site-
AC C
EP
TE D
M AN U
SC
. In brief, non-atopic healthy subjects, mild (no controller
9
ACCEPTED MANUSCRIPT
analyzed for differential cell counts. Biomarker analyses focused on molecular Type-2-activity
144
status in biopsies as they relate to asthma characteristics, and clinical biomarkers.
145
Definition of Type-2 status in biopsies
146
‘Type-2-high’ vs. ‘Type-2-low’ phenotypes were defined a priori based on inferred IL-13 activity
147
in airway-mucosa biopsies from asthma subjects compared to healthy controls using gene
148
expression of CCL26, periostin or a multigene in-house IL-13-IVS (see Online Supplement
149
section E1 and Table E1). IL-13 activity was selected as it is the most broadly-expressed Type-2
150
cytokine produced by Th2 cells 20 and also Type-2 innate lymphoid cells 21. CCL26 mediates
151
eosinophil infiltration into the airway and is the most highly-expressed selective IL-13 driven
152
gene 22-24. The chemokine CCL17 is induced by IL-13 24, 25 and is decreased by anti-IL-13 and
153
anti-IL-4R mAb therapeutics 10, 13, 26. For CCL26, the highest signals in the healthy controls were
154
just below the limit of reliable quantification for the microarray, so this limit of quantification
155
(log2 intensity of 5.0) was set as the threshold for airway-CCL26-high versus -low status. For
156
periostin and IL-13-IVS, Type-2-high status was defined as gene expression (or enrichment)
157
beyond the 95th percentile of the healthy controls.
158
Clinical biomarkers evaluated for their association to Type-2 high or low status by airway IL-13-
159
driven gene expression included FENO, bEOS, spEOS, serum IgE, and 2 serum Type-2-associated
160
biomarkers (serum (s)CCL17 and sCCL26). Details of the assays can be found in the Online
161
Supplement Section E1. Serum CCL17 and sCCL26 were measured repeatedly over 12 months
162
to assess stability.
163
Statistical analysis
AC C
EP
TE D
M AN U
SC
RI PT
143
10
ACCEPTED MANUSCRIPT
Healthy subjects (n=25) and non-ICS-treated mild asthma subjects (n=28) were analyzed
165
separately, while ICS-treated moderate (n=29) and severe (n=26) asthma cohorts were pooled.
166
We focused analyses on the moderate-severe asthmatics as they have unmet clinical need
167
despite ICS treatment. Separate analyses on the moderate and severe cohorts were not
168
undertaken due to limited statistical power.
169
Statistical analyses used OmicSoft ArrayStudio v7 (Cary, NC; www.omicsoft.com). For data with
170
log-normal distributions, logarithmic transformations were performed. Group comparisons of
171
gene and protein expression levels (log2 –transformed) were performed using General Linear
172
Models. For variables with log-normal distributions, geometric mean and asymmetric standard
173
deviations were estimated from the log2-scaled mean and standard deviation: geometric mean
174
= 2^(mean); asymmetric upper standard deviation = 2^(mean + SD) – 2^(mean); asymmetric
175
lower standard deviation = 2^(mean – SD) – 2^(mean). Significance of differences in
176
proportions between 2 categorical variables was tested using Fisher’s exact test. Nominal p-
177
values are reported, with multiple-testing adjusted FDR values below 0.05 indicated.
178
Classification model statistics were obtained from logistic regression and receiver operating
179
characteristic analyses (NCSS v8, www.NCSS.com).
180
Box-and-whisker plot representations of distributions in figures show the median and
181
interquartile range (box), minimum and maximum range (whiskers), mean (‘+’ symbol), and
182
diamond symbols the values for each individual subject/sample.
AC C
EP
TE D
M AN U
SC
RI PT
164
183
184
Results 11
ACCEPTED MANUSCRIPT
Airway-mucosal CCL26, periostin, and IL-13-IVS gene expression compared for classification of
186
Type-2 inflammation
187
Figure 1 displays segregation into high and low Type-2 status of asthmatic from healthy
188
subjects based on airway-mucosal gene expression of CCL26 (panel A), periostin (panel B), or IL-
189
13-IVS enrichment (panel C). Airway-mucosal-CCL26-high asthmatics were mostly high for
190
airway periostin (90%) and airway IL-13-IVS (83%) expression (all asthmatic subjects combined),
191
a finding unimpacted by asthma severity (Table 1; Online Supplement Figure E1). Lower
192
proportions of airway periostin-high asthmatics were high for airway CCL26 (62%) or airway IL-
193
13-IVS (67%) expression for all asthmatic subjects combined, but the concurrence was higher in
194
mild compared to moderate-severe asthma (Table 1; Online Supplement Figure E1). Similarly,
195
lower proportions of airway IL-13-IVS high asthmatics were high for airway CCL26 (62%) or
196
airway periostin (67%) expression for all asthmatic subjects, with higher concurrence in mild
197
compared to moderate-severe asthmatic subjects. (Table 1; Online Supplement Figure E1).
198
These observations drove selection of airway-mucosal-CCL26 expression as the primary anchor
199
for Type-2 status, although periostin and IL-13-IVS also demonstrated utility for segregation of
200
asthma from healthy control subjects as discussed in Online Supplement Section E2.
201
Demographics by airway-mucosal-CCL26 status
202
Table 2 presents demographic characteristics for mild and moderate asthma by airway-mucosal
203
-CCL26-high and –low status. There were no significant differences by CCL26 status. Similar
204
proportions of CCL26 -high (3/13=23%) and -low (7/42=17%) status subjects used leukotriene
205
inhibitors (data now shown).
AC C
EP
TE D
M AN U
SC
RI PT
185
12
ACCEPTED MANUSCRIPT
Asthma disease characteristics and clinical biomarkers by airway-CCL26 status
207
There were decreasing proportions of airway-CCL26-high subjects in the mild (N=16/28, 57%),
208
moderate (N=8/29, 28%) and severe (N=5/26, 19%) asthma cohorts.
209
Disease characteristics including airflow obstruction were not associated with airway-CCL26
210
status (Table 3) except that CCL26-high status was associated with nominally significantly
211
increased AHR in mild asthma (p=0.045). In contrast, airway-mucosal-CCL26 status was strongly
212
associated with multiple Type-2 clinical biomarkers. Thus, FENO (all asthma severities), and
213
bEOS (mild-moderate asthma), were significantly elevated (p<0.05) in the airway-mucosal-
214
CCL26-high compared to the -low group (Table 3, and Figure 2 A, B). Serum IgE was highly
215
elevated in the airway-mucosal-CCL26-high moderate-severe asthma group (N=13) compared
216
with healthy controls (FDR<0.0001, p<0.0001, fold=24) and the airway-mucosal-CCL26-low
217
group (n=41, FDR=0.091, p=0.018, fold=2.8), though sIgE was also elevated in the airway-
218
mucosal-CCL26-low group compared to healthy controls (FDR<0.0001, p<0.0001, fold=8.6)
219
(Figure 3A). Serum CCL17 levels were significantly elevated in airway-mucosal-CCL26-high
220
compared with the -CCL26-low group (p = 0.0039, 2.17-fold/Type-2-low) and with healthy
221
controls (p = 0.018, 2.02-fold/healthy controls) for moderate-severe but not mild asthma
222
subjects (p=0.29 and 0.90, respectively; see Figure 3B). Similarly, serum CCL26 levels were
223
significantly elevated in airway-mucosal-CCL26-high compared with the -CCL26-low group (p =
224
0.0027, 7.35-fold/Type-2-low) and compared with healthy controls (p = 0.029, 4.67-fold/healthy
225
controls) for moderate-severe but not mild asthma subjects (p=0.21 and 0.67, respectively; see
226
Figure 3C).
AC C
EP
TE D
M AN U
SC
RI PT
206
13
ACCEPTED MANUSCRIPT
Categorical analysis of clinical biomarkers by airway-mucosal CCL26 status
228
Airway-mucosal-CCL26 status was compared with clinical biomarkers categorized using the
229
following cut-offs: ≥35ppb for FENO, a previously described cutoff 27-30; ≥300 cells/mm3 for
230
bEOS, a cutoff used for novel “Type-2/eosinophilic” therapeutics 31; and by the 95th percentile
231
of the healthy control cohort for sIgE (Table 4). Serum CCL17-high and sCCL26-high status was
232
based on the 75th percentile of the healthy controls rather than the 95th centile because of the
233
wide distributions in healthy controls.
234
A majority of airway-CCL26-high moderate-severe asthma subjects were FENO-high (69%),
235
compared to 24% of airway-CCL26-low subjects (p=0.0063). Similarly, 77% airway-CCL26 high
236
subjects were bEOS-high, compared to 24% of airway-CCL26-low subjects (p=0.0009). All
237
airway-CCL26-high subjects were FENO-high or bEOS-high but only 55% were FENO-high and
238
bEOS-high. In contrast, for airway-CCL26-low subjects, 36% were FENO-high or bEOS-high but
239
few (12%) were FENO-high and bEOS-high (p<0.0001 and 0.014, respectively), for airway-CCL26
240
high vs. low. Of note, 92% of airway-CCL26-high subjects were sIgE-high, but 71% of airway-
241
CCL26-low subjects were also sIgE-high (p=0.15). For moderate-severe asthma, 11/21 (52%)
242
were sCCL17-high but only 2/34 (6%) CCL17-low asthmatics were airway-CCL26 high.
243
Additionally, 7/11 (64%) were sCCL26-high but only 5/42 (12%) sCCL26-low asthmatics were
244
airway-CCL26-high.
245
Similar associations for mild asthma (see Table 3 and Online Supplement Section E2) were
246
observed as for moderate-severe asthma except that both the airway-CCL26-high and -low
247
groups were mostly low for bEOS in mild asthma (75% and 92% of respective groups).
AC C
EP
TE D
M AN U
SC
RI PT
227
14
ACCEPTED MANUSCRIPT
In general, the patterns of categorical clinical biomarkers reported for airway-mucosal-CCL26
249
expression were similar for periostin (Online Supplement Table E2) and IL-13-IVS expression
250
(Online Supplement Table E3), although the between-group differences were less pronounced.
251
For example, when defining airway-mucosal-Type-2-high status by CCL26, periostin, and IL-13-
252
IVS, 100%, 80%, and 75% of airway-Type-2-high moderate-severe asthmatics were FENO-high
253
or bEOS-high, but similar proportions of airway-Type-2-low moderate-severe asthmatics were
254
FENO-high or bEOS-high (36%, 34%, and 32%) respectively. These observations reinforce our
255
selection of airway-mucosal-CCL26 expression as the primary anchor for Type-2 status.
256
Airway eosinophilia by airway-CCL26 status
257
Thirty-nine moderate-severe asthmatic subjects had acceptable sputum quality, 8 of whom
258
were airway-CCL26-high and comparisons were made with commonly-used cutoffs, namely,
259
sputum eosinophils (spEOS) ≥3% and sputum neutrophils (spPMN) ≥60% status (see Table 4
260
and Figure 2C). Most (88%) of airway-CCL26-high subjects but also 45% of airway-mucosal-
261
CCL26-low subjects had spEOS≥3% (p=0.049). Those with neutrophilic or paucigranulocytic
262
patterns were predominantly airway-mucosal-CCL26-low (86% and 100%, respectively).
263
Submucosal eosinophil intensity (histology) was available for 45 moderate-severe asthma
264
subjects, 13 of whom were airway-mucosal-CCL26-high. Eosinophil intensity was significantly
265
higher in airway-mucosal-CCL26-high compared to –low (p=0.025) moderate-severe asthmatic
266
subjects and to healthy controls (p=0.0039), with similar associations for mild asthma (p=0.024
267
and 0.0002, respectively) (Figure 2D).
268
Classification of airway-mucosal-CCL26 Type-2 status using clinical biomarkers
AC C
EP
TE D
M AN U
SC
RI PT
248
15
ACCEPTED MANUSCRIPT
Pre-specified thresholds for high/low status of FENO, bEOS, sCCL17, and sCCL26 were employed
270
to test classification of airway-CCL26-high status (Table 5) (optimized thresholds from modeling
271
reported in Online Supplement Table E8). Positive predictive values (PPV) and negative
272
predictive values (NPV) were estimated based on the prevalences observed in the ADEPT study
273
population. The actual predictive values for a given population would be dependent on the
274
prevalence of CCL26-high status in that population.
275
FENO-high, bEOS-high, and the criterion ‘FENO-high OR bEOS-high’ all had poor PPV of ≤50%
276
but good NPV (≥89%), with the ‘FENO-high OR bEOS-high’ classification having the best NPV
277
(100%). The criteria of ‘FENO-high AND bEOS-high’ had only marginally better PPV (55%) but
278
with reduced NPV (84%). Serum CCL17-high or sCCL26-high gave modestly improved PPV (62%
279
and 64%, respectively) while maintaining a good NPV (92% and 88%, respectively).
280
The combination of ‘FENO-high OR bEOS-high’ AND sCCL17-high modestly improved PPV (66%)
281
while maintaining high NPV (97%). The criteria ‘FENO-high OR bEOS-high’ AND sCCL26-high
282
resulted in much-improved PPV (88%) while maintaining good NPV (89%). Finally, the
283
combination of sCCL17-high AND sCCL26-high AND ‘FENO-high OR bEOS-high’ criteria resulted
284
in the best PPV (100%) while maintaining good NPV (87%). Therefore, depending on the
285
application, one of these 3 combination models would be optimal, depending on whether high
286
NPV, high PPV, or a balance of NPV and PPV is required.
AC C
EP
TE D
M AN U
SC
RI PT
269
287
288
16
ACCEPTED MANUSCRIPT
289
Discussion
291
Based on the ADEPT dataset, we report that airway-mucosal expression of CCL26 was a robust
292
discriminator of Type-2 inflammation from healthy non-atopic subjects. Furthermore, airway-
293
mucosal-CCL26 expression was best identified using a panel of clinical biomarkers including
294
FENO, bEOS, and 2 novel markers, sCCL17 and sCCL26.
295
Airway-mucosal-CCL26, compared to periostin and IL-13-IVS, expression provided the most
296
robust segregation for airway-mucosa Type-2 activity status in our dataset. The airway-mucosa
297
CCL26 Type-2 phenotype also had the best concordance with airway eosinophilic inflammation,
298
including 100% of airway-mucosa-CCL26-high moderate-severe asthmatics being ‘FENO-high or
299
bEOS-high’ and 87.5% being sputum eosinophil-high, compared to using periostin or IL-13-IVS
300
as anchors. Similar low proportions of Type-2-low status moderate-severe asthmatics
301
regardless of the airway-mucosa-Type-2 anchor tested were high for eosinophilic inflammation.
302
Taken together, these observations affirm our preference for airway-mucosa-CCL26 expression
303
as the anchor for Type-2 status. Similar findings pertain to the mild asthmatic subjects.
304
Notwithstanding this preference, others have confirmed the substantial utility of periostin, and
305
other genes including calcium-activated chloride channel regulator 1 (CLCA1) and SERPINB2
306
(neither of which had quantifiable expression levels in the ADEPT airway mucosa), as anchors
307
for Type-2 status 4, 32, which are not inconsistent with the results presented in this report.
AC C
EP
TE D
M AN U
SC
RI PT
290
17
ACCEPTED MANUSCRIPT
Both CCL26 and periostin expression are induced by IL-13 in multiple cell types present in
309
endobronchial biopsies, including epithelial cells, airway smooth muscle cells, and fibroblasts,
310
as well as macrophages for CCL26, making them ideally suited for evaluation in biopsies (23, 11,
311
24, 33
312
CCL11 and CCL24 34,and in one report, IL-13 induced greater expression of CCL26 in bronchial
313
epithelial cells and higher supernatant CCL26 levels from severe compared to mild asthma
314
suggesting a role for CCL26 in the sustained inflammation in severe eosinophilic asthma 22. The
315
less-consistent performance of periostin may be a consequence of transforming growth factor
316
(TGF)-β being a major regulator in addition to IL-4/IL-13 35, 36. The lesser performance of the IL-
317
13-IVS may be related to the inclusion of genes not specific to IL-13 activity, compared to
318
specificity of CCL26 per se.
319
The majority of mild, but only a minority of moderate-severe asthmatics, was airway-mucosal-
320
CCL26-high, likely related to ICS treatment in the latter. Persistence of Type-2-high status
321
despite ICS treatment could be driven by steroid non-responsiveness (or non-adherence),
322
supported by the significant up-regulation of the steroid-response gene FKBP5 in Type-2-low,
323
but not Type-2-high, moderate-severe asthmatics (data not shown). Consistent with this
324
contention, increased CCL26 protein expression has been found in bronchial epithelium of
325
subjects with severe asthma selected for persistent high spEOS counts despite high doses of
326
corticosteroid 22.
327
In general, Type-2 status was not associated with clinical asthma characteristics, perhaps
328
related to the milder asthma population in ADEPT with few refractory asthmatic subjects. In
RI PT
308
AC C
EP
TE D
M AN U
SC
. CCL26 is a more effective chemoattractant for eosinophils than other chemokines e.g.
18
ACCEPTED MANUSCRIPT
contrast, there were marked differences between airway-mucosal-CCL26-high and -low status
330
for eosinophilia-associated clinical biomarkers, including FENO, bEOS, and spEOS, consistent
331
with the biology of CCL26, a potent eosinophil chemoattractant in asthmatics 34. However, only
332
about a third of sputum ‘eosinophilic’ moderate-severe asthmatics were airway-mucosal-
333
CCL26-high (Table 4). This finding is consistent with a previous report by Choy et al. of only a
334
fraction of sputum ‘eosinophilic’ mild asthmatics having a ‘Th2’ phenotype (evaluated by
335
periostin, SERPINB2, and CLCA1 expression in endobronchial brushings) 37. Type-2 activity,
336
evidenced by enriched IL-4/IL-13 downstream activity, would also be expected to coincide with
337
IL-5 activity, which leads to generation of eosinophils. This is consistent with airway-mucosal-
338
CCL26-high asthmatics having elevated airway eosinophilic inflammation in both airway
339
submucosa and sputum. Of note, eosinophilia in blood, sputum and airway wall, and FENO may
340
be discordant from each other perhaps due different mechanisms affecting these distinct
341
compartments e.g., IL-5 driving blood eosinophils and IL-4/IL-13 elevating FENO by local action
342
on the airway mucosa 38 and bEOS and FENO associate independently with wheezing episodes
343
39
344
reflect the transient nature of Type-2-high status that was inhibited by ICS or had
345
spontaneously resolved. However, despite CCL26 being a potent inducer of airway eosinophilia
346
34
347
e.g. from CCL11 (eotaxin-1) induced by non-Type-2 mediators, such as tumor necrosis factor
348
(TNF)-α and IL-1β 40-42.
349
Serum CCL17, an IL-13-induced chemoattractant for Th2 cells 25, 43 that is reduced by anti-IL-13
350
and anti-IL-4R therapeutics 10, 13, 26, was significantly higher in airway-mucosal-CCL26 high
TE D
M AN U
SC
RI PT
329
AC C
EP
. The presence of eosinophilia as well as atopy/elevated IgE with Type-2-low status could
, recruitment of eosinophils to the airways could also occur independently of Type-2 activity,
19
ACCEPTED MANUSCRIPT
moderate-severe asthmatics, despite ICS treatment. Serum CCL26 was also significantly
352
elevated in airway-mucosal-CCL26 high moderate-severe asthmatics. CCL26 could come from
353
leakage from the airways, where transcriptional activity defines the airway-mucosal-CCL26-high
354
phenotype. Larose et al. 22 showed a correlation between CCL26 levels and spEOS counts and
355
suggested that bronchial epithelial cells are the dominant source of airway CCL26 that drives
356
eosinophils into the airway lumen. Accordingly, CCL26 gene expression in endobronchial
357
brushings samples from ADEPT was highly elevated selectively in the airway-mucosa-CCL26-
358
high asthma group (data not shown). In one report, CCL26 gene expression in epithelial
359
brushings was also highest in severe asthma, associated with higher sputum eosinophils, lower
360
FEV1, and more frequent exacerbations 44.
361
Knowing whether eosinophilic inflammation, in the absence of Type-2-high status, is pathogenic
362
will be important in determining which asthma subpopulations would benefit more from IL-
363
4/IL-13 vs. IL-5 targeted therapeutics. The efficacy of IL-4/IL-13-targeted therapeutics to impact
364
eosinophilic inflammation is suggested by their capacity to reduce FENO as reported for
365
lebrikizumab 13, although FENO is not an absolute indicator of eosinophilic activity. Impact on
366
spEOS by IL-4/IL-13-targeted therapeutics has not been reported. The efficacy of IL-5 targeted
367
therapeutics in reducing Type-2-associated pathology is unclear. Evidence of reductions in sIgE
368
and CCL17, as observed for the IL-4/IL-13 targeted therapeutics, would provide the clearest
369
evidence for such, but has not been reported.
370
As the majority of effective anti-inflammatory agents work best in a Type-2-high population e.g.
371
ICS 4, and anti-IL-13 antibodies 13, defining Type-2 status in patients using clinically accessible
AC C
EP
TE D
M AN U
SC
RI PT
351
20
ACCEPTED MANUSCRIPT
biomarker will be important. Identification of airway-mucosa-CCL26-high subjects in moderate-
373
severe asthma was best achieved based on combinations of FENO-high, blood EOS-high,
374
sCCL17-high, and sCCL26-high criteria. Classification based on ‘FENO-high or bEOS-high’ could
375
be readily implemented in screening of asthma patients as these biomarkers can be routinely
376
measured at present. Although having perfect NPV for classifying airway-mucosal-CCL26-high
377
status, the PPV (47%) was poor, limiting utility of this classification for a co-diagnostic test for
378
Type-2-high phenotype. However, the classification could still be useful for enriching a study
379
population for a Type-2-low phenotype.
380
Adding sCCL17-high to the classification with ‘FENO-high or bEOS-high’ maintained an optimal
381
NPV (97%) while moderately improving the PPV to 66%. This classification has the potential to
382
be implemented as a co-diagnostic tool to triage patients likely to not benefit from of Type-2-
383
targeting therapeutic while moderately enriching for those who would likely benefit.
384
Adding sCCL26-high to the classification with ‘FENO-high or bEOS-high’ provided more of a
385
balance between NPV (89%) and PPV (88%). The full model including ‘FENO-high or bEOS-high’,
386
sCCL17-high, and sCCL26-high criteria had the best PPV (100%) with a modestly reduced NPV of
387
87%. These last 2 classification models may be useful to prospectively select patients for Type-
388
2-targeted therapeutic clinical studies, highly enriching for patients more likely to respond.
389
However, these models have low sensitivities that would result in about half of Type-2-high
390
patients being called Type-2-low, which may be tolerable for proof-of-concept studies but not
391
for application as a co-diagnostic test.
AC C
EP
TE D
M AN U
SC
RI PT
372
21
ACCEPTED MANUSCRIPT
These classification models will need to be confirmed in independent study populations, with
393
ultimate validation coming from interventional clinical trials. The generalizability of our findings
394
is unknown as the ADEPT study did not include smokers, subjects with high BMI, and those
395
requiring chronic oral corticosteroids. Because of the limited sample size of the Type-2-high
396
asthma subpopulation, training-test set confirmation approaches could not be employed. To
397
limit risks of over-modeling, thresholds for categorization of the accessible biomarkers were
398
pre-defined before modeling, although the optimal modeled thresholds performed similarly. In
399
post hoc analyses, logistic regression modeling with cross-validation (in full study population,
400
without randomized training and test sets) resulted in an optimal model where the weights for
401
each categorical predictor (‘FENO or bEOS’, sCCL17, sCCL26) or continuous values for each
402
predictor resulting in comparable model performance compared to the simple Boolean models
403
presented in Table 5, with stronger performance with discrete predictors compared to the
404
respective continuous values (Table E9 in Online Repository). The ultimate cut-offs for the
405
predictors would need to be retrospectively established from interventional clinical response
406
data, with current thresholds providing for an enrichment to increase probability of success for
407
a therapeutic in the absence of available clinical response data. The stability of the airway-
408
mucosal Type-2 phenotype over time has not been directly evaluated. The consistency of
409
sCCL26 levels reported here suggests that the airway-mucosal Type-2 phenotype is a stable
410
phenotype.
411
Limitations of this study include the absence of smoking, oral-corticosteroid dependent, and
412
morbidly obese asthmatic subjects as described previously 16. Furthermore, we did not recruit
413
an atopic healthy subject population that might have provided an additional group for
AC C
EP
TE D
M AN U
SC
RI PT
392
22
ACCEPTED MANUSCRIPT
comparison purposes. Finally, we did not have a reliable serum periostin assay available to
415
enable comparison of this to serum CCL17 and serum CCL26. However, as reported by Choy et
416
al., serum periostin was not significantly higher in mild asthmatics with an airway-mucosal ‘Th2’
417
phenotype than those with a ‘non-Th2’ phenotype, but rather elevated only in those with an
418
airway-mucosal ‘Th17’ phenotype 37.
419
In conclusion, airway-mucosal-CCL26 gene expression, representing airway-mucosal IL-13
420
activity, was used to optimally define an airway-mucosal Type-2-high phenotype, which is a
421
subgroup of a broader eosinophilic phenotype. Combinations of the clinically accessible
422
biomarkers FENO, bEOS, sCCL17, and sCCL26 best identified the airway-mucosal-CCL26-high
423
phenotype and promises to help select patients for Type-2 inflammation-targeted therapies.
424
Type-2-low asthma, a group with high unmet need and few therapeutic options, can be reliably
425
identified using the same biomarkers.
SC
M AN U
TE D EP AC C
426
RI PT
414
23
ACCEPTED MANUSCRIPT
Acknowledgments
428
The ADEPT investigators
429 430
I Strambu1, S Lam 2, A Eich3, A Ludwig-Sengpiel 4, R Leigh5, M Dransfield6, W Calhoun7, A Hussaini8, a n d P C hanez 9
1Arensia Exploratory Medicine, Sos. Viilor 90, Bucharest 050159, Romania. Email: [email protected]
2Institute for Heart and Lung Health, The Lung Centre, 7th Floor, Gordon and Leslie Diamond Health Care Centre, 2775 Laurel Street, Vancouver, B.C., Canada, V5Z 1M9. Email: [email protected] 3IKF Pneumologie Frankfurt, Institut für klinische Forschung Pneumologie, Clinical Research Centre Respiratory
SC
Diseases, Schaumainkai 101-103, Stresemannallee 360596, Frankfurt, Germany. Email: [email protected] 4KLB Gesundheitsforschung Lübeck GmbH, Sandstr. 18, 23552 Lübeck, Germany. Email: [email protected]
5Cumming Scholl of Medicine, University of Calgary, 3280 Hospital Drive NW, Calgary, AB T2N 4Z6, Canada. Email:
M AN U
[email protected]
6Division of Pulmonary, Allergy and Critical Care Medicine, University of Alabama at Birmingham & Birmingham VA Medical Center, 422 THT, 1900 University Blvd, Birmingham, AL 35294, USA. Email: [email protected]
74.116 John Sealy Annex, University of Texas Medical Branch, 301 University Blvd, Galveston, TX, 77555-0568, USA: Email: [email protected] 8Parexel International, Shelton Simmons (MD), 3001 S Hanover St #7, Brooklyn, MD 21225, USA. Email: [email protected]
9 Pneumologie, Aix Marseilles University, APHM/ INSERM U1067, Chemin des Bourellys 13015, Marseille, France. Email: [email protected]
TE D
431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451
RI PT
427
The following Janssen personnel contributed significantly to the success of ADEPT: Debra
453
Alvarez, Jennifer Campos, Robert Gordon, Keith Lasher, Francisco Leon, Hongjuan Liu, Jennifer
454
Montello, Nancy Peffer, Kevin Petty, Filza Potapova, and Dipti Shah.
AC C
EP
452
24
ACCEPTED MANUSCRIPT
455
References
456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500
1.
7. 8. 9.
10. 11. 12. 13. 14.
15.
16.
17. 18. 19.
RI PT
SC
6.
M AN U
5.
TE D
4.
EP
3.
AC C
2.
From the Global Strategy for Asthma Management and Prevention, Global Initiative for Asthma (GINA) 2015. 2015.] Available from http://www.ginasthma.org/. Fahy JV. Type 2 inflammation in asthma--present in most, absent in many. Nat Rev Immunol 2015; 15:57-65. Karta MR, Broide DH, Doherty TA. Insights into Group 2 Innate Lymphoid Cells in Human Airway Disease. Curr Allergy Asthma Rep 2016; 16:8. Woodruff PG, Modrek B, Choy DF, Jia G, Abbas AR, Ellwanger A, et al. T-helper type 2-driven inflammation defines major subphenotypes of asthma. Am J Respir Crit Care Med 2009; 180:388-95. Woodruff PG, Boushey HA, Dolganov GM, Barker CS, Yang YH, Donnelly S, et al. Genome-wide profiling identifies epithelial cell genes associated with asthma and with treatment response to corticosteroids. Proc Natl Acad Sci U S A 2007; 104:15858-63. Peters MC, Mekonnen ZK, Yuan S, Bhakta NR, Woodruff PG, Fahy JV. Measures of gene expression in sputum cells can identify TH2-high and TH2-low subtypes of asthma. J Allergy Clin Immunol 2014; 133:388-94. Little SA, Chalmers GW, MacLeod KJ, McSharry C, Thomson NC. Non-invasive markers of airway inflammation as predictors of oral steroid responsiveness in asthma. Thorax 2000; 55:232-4. Jang AS, Lee JH, Park SW, Lee YM, Uh ST, Kim YH, et al. Factors influencing the responsiveness to inhaled glucocorticoids of patients with moderate-to-severe asthma. Chest 2005; 128:1140-5. Hanania NA, Wenzel S, Rosen K, Hsieh HJ, Mosesova S, Choy DF, et al. Exploring the effects of omalizumab in allergic asthma: an analysis of biomarkers in the EXTRA study. Am J Respir Crit Care Med 2013; 187:804-11. Wenzel S, Ford L, Pearlman D, Spector S, Sher L, Skobieranda F, et al. Dupilumab in persistent asthma with elevated eosinophil levels. N Engl J Med 2013; 368:2455-66. Bel EH, Wenzel SE, Thompson PJ, Prazma CM, Keene ON, Yancey SW, et al. Oral glucocorticoidsparing effect of mepolizumab in eosinophilic asthma. N Engl J Med 2014; 371:1189-97. Ortega HG, Liu MC, Pavord ID, Brusselle GG, FitzGerald JM, Chetta A, et al. Mepolizumab treatment in patients with severe eosinophilic asthma. N Engl J Med 2014; 371:1198-207. Corren J, Lemanske RF, Hanania NA, Korenblat PE, Parsey MV, Arron JR, et al. Lebrikizumab treatment in adults with asthma. N Engl J Med 2011; 365:1088-98. Proceedings of the ATS workshop on refractory asthma: current understanding, recommendations, and unanswered questions. American Thoracic Society. Am J Respir Crit Care Med 2000; 162:2341-51. Bel EH, Sousa A, Fleming L, Bush A, Chung KF, Versnel J, et al. Diagnosis and definition of severe refractory asthma: an international consensus statement from the Innovative Medicine Initiative (IMI). Thorax 2011; 66:910-7. Silkoff PE, Strambu I, Laviolette M, Singh D, FitzGerald JM, Lam S, et al. Asthma characteristics and biomarkers from the Airways Disease Endotyping for Personalized Therapeutics (ADEPT) longitudinal profiling study. Respir Res 2015; 16:142. Juniper EF, O'Byrne PM, Guyatt GH, Ferrie PJ, King DR. Development and validation of a questionnaire to measure asthma control. Eur Respir J 1999; 14:902-7. Juniper EF, Buist AS, Cox FM, Ferrie PJ, King DR. Validation of a standardized version of the Asthma Quality of Life Questionnaire. Chest 1999; 115:1265-70. Kelly MM, Efthimiadis A, Hargreave FE. Induced sputum : selection method. Methods Mol Med 2001; 56:77-91. 25
ACCEPTED MANUSCRIPT
25.
26.
27.
28.
29. 30. 31.
32.
33.
34.
35.
36.
RI PT
24.
SC
23.
M AN U
22.
TE D
21.
Loza MJ, Perussia B. Peripheral immature CD2-/low T cell development from type 2 to type 1 cytokine production. Journal of Immunology 2002; 169:3061-8. Loza MJ, Perussia B. Final steps of natural killer cell maturation: a model for type 1-type 2 differentiation? Nat Immunol 2001; 2:917-24. Larose MC, Chakir J, Archambault AS, Joubert P, Provost V, Laviolette M, et al. Correlation between CCL26 production by human bronchial epithelial cells and airway eosinophils: Involvement in patients with severe eosinophilic asthma. J Allergy Clin Immunol 2015. Komiya A, Nagase H, Yamada H, Sekiya T, Yamaguchi M, Sano Y, et al. Concerted expression of eotaxin-1, eotaxin-2, and eotaxin-3 in human bronchial epithelial cells. Cell Immunol 2003; 225:91-100. Syed F, Huang CC, Li K, Liu V, Shang T, Amegadzie BY, et al. Identification of interleukin-13 related biomarkers using peripheral blood mononuclear cells. Biomarkers 2007; 12:414-23. Nomura T, Terada N, Kim WJ, Nakano K, Fukuda Y, Wakita A, et al. Interleukin-13 induces thymus and activation-regulated chemokine (CCL17) in human peripheral blood mononuclear cells. Cytokine 2002; 20:49-55. van Hartingsveldt B, Nnane IP, Bouman-Thio E, Loza MJ, Piantone A, Davis HM, et al. Safety, tolerability and pharmacokinetics of a human anti-interleukin-13 monoclonal antibody (CNTO 5825) in an ascending single-dose first-in-human study. Br J Clin Pharmacol 2013; 75:1289-98. Alvarez-Gutierrez FJ, Medina-Gallardo JF, Perez-Navarro P, Martin-Villasclaras JJ, Martin Etchegoren B, Romero-Romero B, et al. [Comparison of the Asthma Control Test (ACT) with lung function, levels of exhaled nitric oxide and control according to the Global Initiative for Asthma (GINA)]. Arch Bronconeumol 2010; 46:370-7. Barros R, Moreira A, Fonseca J, de Oliveira JF, Delgado L, Castel-Branco MG, et al. Adherence to the Mediterranean diet and fresh fruit intake are associated with improved asthma control. Allergy 2008; 63:917-23. Cowan DC, Hewitt RS, Cowan JO, Palmay R, Williamson A, Lucas SJ, et al. Exercise-induced wheeze: Fraction of exhaled nitric oxide-directed management. Respirology 2010; 15:683-90. Michils A, Baldassarre S, Van Muylem A. Exhaled nitric oxide and asthma control: a longitudinal study in unselected patients. Eur Respir J 2008; 31:539-46. Castro M, Wenzel SE, Bleecker ER, Pizzichini E, Kuna P, Busse WW, et al. Benralizumab, an antiinterleukin 5 receptor alpha monoclonal antibody, versus placebo for uncontrolled eosinophilic asthma: a phase 2b randomised dose-ranging study. Lancet Respir Med 2014; 2:879-90. Peters MC, Mekonnen ZK, Yuan S, Bhakta NR, Woodruff PG, Fahy JV. Measures of gene expression in sputum cells can identify TH2-high and TH2-low subtypes of asthma. Journal of Allergy and Clinical Immunology 2014; 133:388-94. Choy DF, Modrek B, Abbas AR, Kummerfeld S, Clark HF, Wu LC, et al. Gene expression patterns of Th2 inflammation and intercellular communication in asthmatic airways. J Immunol 2011; 186:1861-9. Provost V, Larose MC, Langlois A, Rola-Pleszczynski M, Flamand N, Laviolette M. CCL26/eotaxin3 is more effective to induce the migration of eosinophils of asthmatics than CCL11/eotaxin-1 and CCL24/eotaxin-2. J Leukoc Biol 2013; 94:213-22. Horiuchi K, Amizuka N, Takeshita S, Takamatsu H, Katsuura M, Ozawa H, et al. Identification and characterization of a novel protein, periostin, with restricted expression to periosteum and periodontal ligament and increased expression by transforming growth factor beta. J Bone Miner Res 1999; 14:1239-49. Li G, Oparil S, Sanders JM, Zhang L, Dai M, Chen LB, et al. Phosphatidylinositol-3-kinase signaling mediates vascular smooth muscle cell expression of periostin in vivo and in vitro. Atherosclerosis 2006; 188:292-300.
EP
20.
AC C
501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548
26
ACCEPTED MANUSCRIPT
39.
40.
41.
42.
43.
44.
RI PT
38.
Choy DF, Hart KM, Borthwick LA, Shikotra A, Nagarkar DR, Siddiqui S, et al. TH2 and TH17 inflammatory pathways are reciprocally regulated in asthma. Sci Transl Med 2015; 7:301ra129. Alving K, Malinovschi, A. Basic aspects of exhaled nitric oxide. In: European Respiratory Monograph. United Kingdom: European Respiratory Publications; 2010. p. 1-31. Malinovschi A, Fonseca JA, Jacinto T, Alving K, Janson C. Exhaled nitric oxide levels and blood eosinophil counts independently associate with wheeze and asthma events in National Health and Nutrition Examination Survey subjects. J Allergy Clin Immunol 2013; 132:821-7 e1-5. Lilly CM, Nakamura H, Kesselman H, Nagler-Anderson C, Asano K, Garcia-Zepeda EA, et al. Expression of eotaxin by human lung epithelial cells: induction by cytokines and inhibition by glucocorticoids. J Clin Invest 1997; 99:1767-73. Teran LM, Mochizuki M, Bartels J, Valencia EL, Nakajima T, Hirai K, et al. Th1- and Th2-type cytokines regulate the expression and production of eotaxin and RANTES by human lung fibroblasts. Am J Respir Cell Mol Biol 1999; 20:777-86. Ghaffar O, Hamid Q, Renzi PM, Allakhverdi Z, Molet S, Hogg JC, et al. Constitutive and cytokinestimulated expression of eotaxin by human airway smooth muscle cells. Am J Respir Crit Care Med 1999; 159:1933-42. Imai T, Baba M, Nishimura M, Kakizaki M, Takagi S, Yoshie O. The T cell-directed CC chemokine TARC is a highly specific biological ligand for CC chemokine receptor 4. J Biol Chem 1997; 272:15036-42. Coleman JM, Naik C, Holguin F, Ray A, Ray P, Trudeau JB, et al. Epithelial eotaxin-2 and eotaxin-3 expression: relation to asthma severity, luminal eosinophilia and age at onset. Thorax 2012; 67:1061-6.
SC
37.
M AN U
549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570
AC C
EP
TE D
571
27
ACCEPTED MANUSCRIPT
Figure legends Figure legends
SC
RI PT
Figure 1. Airway-mucosal expression-based categorization of Type-2 phenotype. Airwaymucosal CCL26 (A) and periostin (B) gene expression (log2 intensity) and IL-13-IVS enrichment (GSVA scores) (C) presented in the healthy or mild and moderate-severe asthma cohorts, stratified by airway-mucosa-Type-2 high vs. low groups (x-axis top label) defined by CCL26 (A), periostin (B), or IL-13-IVS (C). Horizontal solid line indicates threshold for high.
M AN U
Figure 2. Top-associated clinical and biomarker variables to airway-mucosa-CCL26 status. Screening/baseline values for (A) FENO, (B) blood eosinophils, (C) sputum eosinophils, and (D) submucosal eosinophils are shown for subjects in the mild, moderate/severe asthma groups, stratified by airway-mucosa-CCL26 high vs. low status (x-axis top labels). * P<0.05 for airwaymucosa-CCL26 high vs. low status within mild or moderate-severe asthma population.
TE D
Figure 3. Type-2-associated serum biomarkers and airway-mucosa-CCL26 status.
AC C
EP
Baseline values of serum IgE (A), serum CCL17 (B) and serum CCL26 (C) are shown for subjects in the mild, moderate/severe asthma groups (x-axis bottom labels), stratified by airwaymucosa-CCL26 high vs. low status (x-axis top labels). * P<0.05 for airway-mucosa-CCL26 high vs. low status within mild or moderate-severe asthma population.
28
ACCEPTED MANUSCRIPT
Tables
airway-mucosal CCL26-high
Mild asthma airway-mucosal periostin-high
airway-mucosal IL-13-IVS-high
airway-mucosal CCL26-high
100% *
94%
81%
airway-mucosal CCL26-low
0%
33%
17%
airway-mucosal periostin-high
79%
100%
68%
airway-mucosal periostin-low
11%
0%
22%
airway-mucosal IL-13-IVS-high
87%
87%
airway-mucosal IL-13-IVS-low
23%
46%
Moderate-Severe asthma airway-mucosal airway-mucosal airway-mucosal CCL26-high periostin-high IL-13-IVS-high 85%
85%
0%
21%
31%
55%
100%
65%
6%
0%
31%
100%
46%
54%
100%
0%
6%
23%
0%
SC
100%
M AN U
% of group:
RI PT
Table 1. Segregation of airway-mucosal Type-2 high versus -low by airway-mucosal CCL26, periostin, and IL-13-IVS gene expression
AC C
EP
TE D
* The percentage of mild or moderate-severe asthma subjects within the airway group indicated in the row label that have airwaymucosal-high status based on biomarker indicated in column label
29
ACCEPTED MANUSCRIPT
Table 2. Demographics by airway-mucosal- CCL26 status for mild and moderate-severe asthma
P*
Moderate-severe asthma airway-mucosal-CCL26 airway-mucosal-CCL26 high (n=13) low (n=42)
P*
Age, years
27.0 ± 7.6 (21.0 - 49.0)
30.4 ± 12.2 (18.0 - 54.0)
0.37
42.8 ± 9.4 (26.0 - 54.0)
41.2 ± 11.3 (18.0 - 55.0)
0.66
BMI
24.6 ± 3.0 (19.6 - 29.6)
24.1 ± 3.4 (20.2 - 30.1)
0.69
26.9 ± 2.9 (21.4 - 31.2)
26.5 ± 3.8 (19.1 - 32.1)
0.71
Duration of asthma, years
16.6 ± 6.9 (7.1 - 31.6)
15.7 ± 8.7 (0.7 - 32.9)
0.76
21.6 ± 10.7 (3.6 - 39.5)
19.2 ± 12.8 (0.3 - 47.5)
0.55
8/16 (50%)
3/12 (25%)
0.25
8/13 (62%)
16/42 (38%)
0.20
White
14/16 (88%)
12/12 (100%)
10/13 (77%)
31/42 (74%)
Black
0/16 (0%)
0/12 (0%)
2/13 (15%)
8/42 (19%)
Other
2/16 (13%)
0/12 (0%)
1/13 (8%)
3/42 (7%)
N/total (% of total) Gender, male Race
RI PT
Mild asthma Airway-mucosal-CCL26 airway-mucosal-CCL26 high (n=16) low (n=12)
SC
Mean ± SD (range)
M AN U
0.50
1.00
AC C
EP
TE D
* P-value for comparison between airway-mucosal-CCL26-high and –low groups within mild asthma or moderate-severe asthma populations
30
ACCEPTED MANUSCRIPT
Table 3: Asthma disease characteristics and clinical biomarkers by airway-mucosal-CCL26 status Mild asthma
Moderate-severe asthma
airway-mucosal-CCL26- airway-mucosal-CCL26 phigh low value†
airway-mucosal-CCL26 high
airway-mucosal-CCL26 low
pvalue†
RI PT
Mean ± SD (range) [N]
93.6 ± 11.7 [15]
90.8 ± 14.3 [12]
0.577
70.2 ± 10.3 [13]
71.2 ± 11.6 [41]
0.785
FEV1, Post-BD (PN)
102.1 ± 9.1 [15]
97.9 ± 11.2 [12]
0.293
83.0 ± 8.9 [13]
80.8 ± 13.8 [41]
0.598
BDR, % change FEV1
6.25 ± 8.54 [15]
6.33 ± 3.25 [12]
0.978
19.79 ± 15.86 [13]
14.63 ± 10.96 [40]
0.195
ACQ
0.6 ± 0.5 [16]
0.9 ± 0.6 [12]
0.176
1.7 ± 0.8 [13]
1.5 ± 0.9 [42]
0.549
AQLQ
6.2 ± 0.7 [16]
5.9 ± 0.9 [12]
0.245
5.6 ± 0.8 [13]
5.5 ± 1.3 [42]
0.637
M AN U
SC
FEV1, Pre-BD (PN)
0.84 +5.33/-0.73 [16]
3.33 +8.18/-2.37 [12] 0.045
0.32 +1.77/-0.27 [10]
0.95 +4.76/-0.79 [35]
0.102
FENO (ppb) *
52.8 +44.0/-24.0 [16]
27.4 +17.7/-10.8 [12] 0.0051
56.8 +56.7/-28.4 [13]
24.9 +23.7/-12.1 [41]
0.0003
bEOS (1000/mm3) *
0.21 +0.19/-0.10 [16]
0.12 +0.13/-0.06 [12] 0.051
0.36 +0.22/-0.14 [13]
0.19 +0.20/-0.10 [42]
0.0032
sIgE (RFU) *
17.1 +33.2/-11.3 [16]
0.032
20.0 +27.4/-11.6 [13]
7.8 +25.9/-6.0 [41]
0.032
0.14
398 +299/-171 [13]
169 +236/-98 [42]
0.0006
TE D
PC20 (mg/ml) *
EP
5.4 +21.8/-4.3 [12]
169 +226/-97 [16]
117 +103/-55 [12]
sCCL26 (pg/ml) *
117 +732/-101 [16]
42 +127/-32 [9]
0.185
416 +3301/-369 [12]
57 +290/-48 [41]
0.0023
SpEOS (% WBC) *
1.2 +8.4/-1.1 [12]
0.5 +1.1/-0.3 [7]
0.291
6.8 +12.9/-4.5 [8]
2.2 +12.4/-1.9 [31]
0.117
SpNEU (% WBC)
45.6 ± 35.6 [12]
52.1 ± 41.1 [7]
0.561
50.8 ± 35.1 [8]
46.9 ± 25.4 [31]
0.893
AC C
sCCL17 (pg/ml) *
* Geometric mean ± asymmetric standard deviation [N], p-value calculated from log-transformed data. † P-value for comparison between airway-CCL26-high and –low groups within mild asthma or moderate-severe asthma populations; bolded when FDR<0.05. 31
Mild asthma Type-2 low (CCL26-) (n=12) 1 (8%) [20%]
P-value* 0.36
Type-2 high (CCL26+) (n=13) 6 (46%) [55%]
12 (75%) [75%]
4 (33%) [25%]
0.053
13 (100%) [46%]
4 (25%) [33%]
8 (67%) [67%]
FENO ≥ 35 ppb FENO < 35 ppb
12 (75%) [75%] 4 (25%) [33%]
4 (33%) [25%] 8 (67%) [67%]
0.053
Blood EOS ≥ 300/mm3 Blood EOS < 300/mm3
4 (25%) [80%] 12 (75%) [52%]
1 (8%) [20%] 11 (92%) [48%]
0.36
Serum IgE high Serum IgE low
15 (94%) [71%] 1 (6%) [14%]
6 (50%) [29%] 6 (50%) [86%]
Atopy (Phadiatop): Positive Atopy (Phadiatop): Negative
16 (100%) [70%] 0 (0%) [0%]
Serum CCL17 ≥75th%ile Serum CCL17 <75th%ile
3 (100%) [100%] 13 (0%) [54%]
Serum CCL26 ≥75th%ile Serum CCL26 <75th%ile
4 (100%) [100%] 12 (0%) [52%]
Sputum EOS ≥ 3% Sputum EOS < 3%
4 (33%) [100%] 8 (67%) [53%]
Sputum NEU ≥ 65% Sputum NEU < 65% Paucigranulocytic sputum Neutrophilic sputum Mixed granulocytic sputum Eosinophilic sputum
Moderate-Severe asthma Type-2 low (CCL26-) (n=42) 5 (12%) [45%] 15 (36%) [54%]
P-value* 0.014 0.000040
27 (64%) [100%]
9 (69%) [47%] 4 (31%) [11%]
10 (24%) [53%] 31 (76%) [89%]
0.0063
10 (77%) [50%] 3 (23%) [9%]
10 (24%) [50%] 32 (76%) [91%]
0.00090
0.023
12 (92%) [29%] 1 (8%) [8%]
29 (71%) [71%] 12 (29%) [92%]
0.15
7 (64%) [30%] 4 (36%) [100%]
0.019
11 (85%) [26%] 2 (15%) [15%]
31 (74%) [74%] 11 (26%) [85%]
0.71
0 (64%) [0%] 11 (36%) [46%]
0.25
11 (85%) [52%] 1 (15%) [3%]
10 (74%) [48%] 32 (26%) [97%]
0.00004
0 (64%) [0%] 11 (36%) [48%]
0.12
7 (85%) [64%] 5 (15%) [12%]
4 (74%) [36%] 38 (26%) [88%]
0.0010
0 (0%) [0%] 7 (100%) [47%]
0.25
7 (87.5%) [33%] 1 (12.5%) [6%]
14 (45%) [67%] 17 (55%) [94%]
0.049
2 (33%) [33%] 10 (67%) [77%]
4 (0%) [67%] 3 (100%) [23%]
0.13
3 (88%) [25%] 5 (12%) [19%]
9 (45%) [75%] 22 (55%) [81%]
0.68
6 (50%) [67%] 2 (17%) [33%] 0 (0%) [na] 4 (33%) [100%]
3 (43%) [33%] 4 (57%) [67%] 0 (0%) [na] 0 (0%) [0%]
0.10
0 (0%) [0%] 1 (12.5%) [14%] 2 (25%) [40%] 5 (62.5%) [31%]
11 (35.5%) [100%] 6 (19%) [86%] 3 (10%) [60%] 11 (35.5%) [69%]
0.15
M AN U
SC
0 (0%) [0%]
TE D
FENO ≥ 35 ppb and Blood Eos ≥ 300/mm3 FENO ≥ 35 ppb or Blood Eos ≥ 300/mm3 Neither FENO ≥ 35 ppb nor Blood Eos ≥ 300/mm3
EP
N (% of Type-2 group) [% of row group]
RI PT
Type-2 high (CCL26+) (n=16) 4 (25%) [80%]
AC C
ACCEPTED MANUSCRIPT Table 4. Comparison of airway-mucosal-Type-2-CCL26 status and categorical biomarkers (high or low)
* P-value (FDR) for comparison between airway-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test, or chi-square test for sputum EOS-PMN subgroups); bolded when FDR<0.05.
32
ACCEPTED MANUSCRIPT
Table 5 airway-mucosal-CCL26 diagnostic characteristics of accessible biomarkers in moderate-severe asthma airway-mucosal-CCL26-high classifiers*
AUC
PPV
NPV
100% 46% 69% 77% 76% 93%
63% 88% 76% 76% 85% 85%
0.82 0.67 0.72 0.77 0.80 0.89
46% 55% 48% 50% 62% 66%
100% 84% 89% 91% 92% 97%
3
58% 58%
90% 98%
3
50% 50%
93% 100%
FENO≥35ppb bEOS ≥300/mm
3
sCCL17-high 3
(FENO≥35ppb or bEOS ≥300/mm ) AND sCCL17-high sCCL26-high (FENO≥35ppb or bEOS ≥300/mm ) AND sCCL26-high sCCL26-high AND sCCL17-high (FENO≥35ppb or bEOS ≥300/mm ) AND sCCL26-high AND sCCL17-high
SC
FENO≥35ppb and bEOS ≥300/mm
3
M AN U
3
Prevalence†
RI PT
Specificity
24% 24% 24% 24% 24% 24%
0.74 0.77
64% 88%
88% 89%
23% 23%
0.71 0.75
67% 100%
86% 87%
23% 23%
TE D
FENO≥35ppb or bEOS ≥300/mm
Sensitivity
* Logistic regression model statistics for classification of moderate-severe asthmatics to airway-mucosal-CCL26-high status: AUC, area-undercurve for receiver operating characteristic; NPV, negative predictive value; PPV, positive predictive value.
AC C
EP
† Prevalence based on the proportion of airway-mucosal-CCL26-high subjects in the ADEPT moderate-severe asthma group, with PPV and NPV statistics based on this prevalence. ADEPT is a cross-sectional study that has rigid enrollment criteria, therefore the prevalence in other asthma study populations may vary based on recruitment criteria.
33
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
ACCEPTED MANUSCRIPT
ACCEPTED MANUSCRIPT
Identification of Airway-Mucosal Type-2 inflammation by Clinical Biomarkers in Asthma Philip E Silkoff, Michel Laviolette, Dave Singh, J Mark FitzGerald, Steven Kelsen, Vibeke
RI PT
Backer, Celeste M Porsbjerg, Pierre-Olivier Girodet, Patrick Berger, Joel N Kline, Geoffrey
& the ADEPT investigators
Online Data Supplement
M AN U
E1. Biomarker matrices and assessments
SC
Chupp, Vedrana S Susulic, Elliot S Barnathan, Frédéric Baribaud, and Matt J Loza
Bronchoscopy sampling
Up to 4 biopsies were immediately preserved in RNAlater ® solution and then maintained at -70°C. For 2 biopsies per subject, RNA was extracted using Qiagen miRNeasy kit (Qiagen; Germantown, MD). RNA quality was assessed by Caliper Life
TE D
Sciences' LabChip System, with 11 of 119 samples processed having insufficient or too poor quality RNA for further processing. 108 RNA samples were amplified with NuGen ovation pico WTA kit (NuGen Technologies; San Carlos, CA). The cDNA was
EP
analyzed using the Affymetrix HG-U133+PM microarray platform (Affymetrix, Santa Clara, CA). CEL files were normalized, assessed for quality control to exclude technical outliers (chip image analysis, Affymetrix GeneChip QC, RNA degradation
AC C
analysis, distribution analysis, principal components analysis, and correlation analysis), and re-normalized using the robust multi-array (RMA) method. The log2normalized data matrix was imported into OmicSoft ArrayStudio software (Cary, NC; www.omicsoft.com) for subsequent analysis. No samples failed the quality control metrics. Batch effects from the two RNA processing sets were observed, with the batch effect adjusted in the data matrices using linear modeling of batch (as random factor) and cohort. A log2-intensity threshold of 5.5 for biopsies was established as the limit of reliable quantification based on the 90th percentile signal
ACCEPTED MANUSCRIPT
of merged nonspecific probesets distribution in the array and by the inflection point of maximum variance with decreasing signal in a standard deviation vs. mean intensity plot across all probesets. Probesets with mean log2 intensity above this
RI PT
threshold in at least one of the 4 study cohorts were considered quantifiable and included in subsequent analyses (24033 probesets). Group comparisons of gene and protein expression (log2 –transformed) were performed using General Linear
Models, adjusting for age and gender. For microarray experiments, significance for
SC
association with a Type-2 group was generally defined as passing a False Discovery
Rate (FDR, Benjamini-Hochberg method) < 0.05 vs. healthy controls and Type-2 high vs. low groups, with > 2-fold estimate (ratio of least-square means of the two
M AN U
comparison groups) for each comparison. Nominally significant associations are reported for comparisons with p-value < 0.05 and >2-fold estimate, reported to provide additional context because of the inflated false-negative rate from FDRadjustment.
Histological evaluation of airway-mucosal biopsies was performed centrally by
TE D
Pantomics Inc. (Richmond, CA; www.pantomics.com). Immediately after collection, the tissues were fixed in 10% neutral buffered formalin and stored at room temperature in 70% ethanol until transfer to Pantomics Inc. Tissue samples were processed in a Leica
EP
TP1020 Processor and paraffin-embedded in a Leica Embedding Center following manufacturer’s instructions. For each paraffin-embedded sample, a section was cut and H&E-stained, which was then assessed and marked under a microscope. The marked
AC C
areas were then punched and transferred to a recipient paraffin block using a stereomicroscope and the Beecher TMA instrument (Beecher Instruments, Sun Prairie, WI). Three tissue microarrays (TMA), each containing 69 (1.5mm) cores were constructed. H&E staining was performed using an Autostainer (Thermo Scientific™ Lab Vision™ Autostainer 360-2D) following manufacturer’s instruction. General histological characteristics and specifically the intensity of submucosal eosinophil infiltration were interpreted and scored on a scale from 0 – 3 by a staff pathologist.
ACCEPTED MANUSCRIPT
Induced sputum Only samples with squamous cell content ≤30% evaluated from cytospin slides evaluated by a central laboratory were included in the analyses. A significant proportion
RI PT
of subjects had only a screening or only a baseline sample available that passed quality control standards (e.g., for cytospin slides, 105/189 possible subjects had acceptable readings at screening, 85 acceptable at baseline, with 128 subjects having either a screening and/or baseline read acceptable). Therefore, the mean (differential cell
SC
counts) or geometric mean (analyte, gene expression measurements) of screening and
M AN U
baseline measurements was used for subsequent analyses.
Induction procedure: All study participants underwent sputum induction during screening to fulfil inclusion criteria and again at the baseline Visit. Asthmatic subjects only had a 3rd sputum induction at the 6 month biomarker Visit. Sputum was induced for 21 minutes divided into three 7-minute sessions of nebulization each followed by a 3
TE D
step cleansing procedure and a focused cough attempt. An aerosol of hypertonic saline (in increasing concentrations of 3, 4, and 5%) was generated by an ultrasonic nebulizer for inhalation by subjects with a post-bronchodilator pre-induction FEV1 of ≥60% predicted; for those with FEV1 ≥50-<60% predicted, induction was performed with
EP
normal/isotonic saline (0.9%). Subjects with a post-bronchodilator FEV1 <50% predicted
AC C
did not undergo induction.
Sputum Processing: The plug selection method was used for this study in all participants 1
. A plug weight of at least 50mg and squamous cell percentages≤20% (evaluated at
local processing laboratory from hemocytometer counting) were required for enrollment into the study during screening. Sputum plugs were treated with dithiothreitol to disperse mucus before collecting supernatant (immediately frozen at ≤ 20°C) and cells for cytospin slide generation and cell pellet RNA preservation in RNAlater®. Samples were further required to have ≤30% squamous cells determined from centrally read cytospin slides to pass quality control metrics.
ACCEPTED MANUSCRIPT
Serum analytical methods Serum was collected using standard Serum Separation Tubes, and frozen within 30 minutes. One of these frozen aliquots, without intermediate freeze-thaw cycles, was
RI PT
provided for quantification of 1129 serum analytes using the SomaScan v3 platform (SomaLogic, Boulder, CO; www.somalogic.com). Serum analyte levels were reported by
Somalogic as relative fluorescence units, cross-plate calibrated, and median normalized. Analyte levels are presented as the log2 ratio to the geometric means of the healthy
SC
controls population for further analysis. Results for serum total immunoglobulin E (IgE)
are presented from this panel, defining high IgE levels as those above the 95th percentile
M AN U
of the healthy controls distribution. In previous evaluations of the platform in asthmatics and healthy controls, IgE measurements highly correlated (Pearson’s correlation coefficient r > 0.9) with those obtained from standard ELISA-based assays (data not shown). 1129 analytes were included in this panel but were not generally considered in the current manuscript given the exploratory nature of the panel.
TE D
Classical Type-2 cytokines including IL-4, IL-5, and IL-13 (lower limit of quantification of 0.13, 1.50, and 1.57 pg/ml, respectively) were specifically measured in serum using the high-sensitivity Meso Scale Discovery (MSD)® electrochemiluminescence platform (Meso Scale Discovery; Rockville, MD). For baseline samples, the Type-2-associated
EP
cytokine CCL17 (TARC) was measured by ELISA (R&D Systems, Minneapolis, MN), with the lower limit of quantification determined at 16 pg/ml. CCL17 (for month 3, 6, and 12
AC C
samples) (RayBio© kit ELH-TARC) and CCL26 (eotaxin-3) (RayBio© kit ELH-Eotaxin3) was measured by ELISA at Cirquest Labs, LLC (Memphis, TN), with the lower limit of quantification determined at 5 pg/ml and 28.7 pg/ml, respectively. Rationale underlying Definition of Type-2 status in airway mucosa. Type-2 high/low status was defined as inferred IL-13 activity in airway-mucosa from asthma subjects compared to non-atopic healthy controls, evaluated by gene expression for CCL26 (eotaxin-3), biopsy periostin (periostin) gene expression, or IL-13 in-vitro gene signature enrichment (IL-13- IVS). CCL26 (chemokine (C-C motif) ligand 26), was
ACCEPTED MANUSCRIPT
selected as the primary Type-2 indicator because it was the gene most highly and consistently induced in vitro by IL-13 in multiple airway resident non-hematopoietic cell types and macrophages (unpublished observations and 2, 3, and was also the most highly differentially expressed gene in biopsies from the ‘Type-2-high’ group defined by
RI PT
Woodruff et al4, and strongly correlated to IL-5 and IL-13 gene expression assessed by qPCR5. CCL26 is tightly regulated by IL-4 and IL-13 and is not directly induced by other
cytokines, such as TNFα, IL-6, interferons (IFNα, IFNɣ) and by tumor growth factor-beta
(TGF-β) (unpublished results). CCL26 was preferred a priori over periostin because of the
SC
co-regulation of periostin, but not CCL26, by TGF-β) (see references6, 7, and unpublished observations), and the constitutive expression of periostin, but not CCL26, observed in
M AN U
healthy controls5, 8. The IL-13-IVS was evaluated to provide a composite score across a broader range of genes inducible by IL-13 rather than using a single gene. However, this approach has the limitation that genes included in the signature may not be specific for IL-13, compared to the specificity of CCL26.
For CCL26, the highest signals in the non-atopic healthy control cohort were just below
TE D
the limit of reliable quantification for the microarray, so this limit of quantification (log2 intensity of 5.0) was set as the threshold for airway-mucosal-CCL26-high versus -low status. For periostin and IL-13-IVS, Type-2-high status was defined as gene expression
EP
(or enrichment) beyond the 95th percentile of the non-atopic healthy control cohort. IL-13 in vitro stimulation signature
AC C
The IL-13 in vitro signature (IVS) was established from the genes induced by IL-13 in airway-liquid interface cultures of airway-mucosal epithelial cells, selecting genes commonly induced across 3 independent sets of experiments (genes in signature listed in Table E1): •
Experiment 1: Bronchial epithelial cultures were established from bronchial
epithelia from fatal asthma cases (n=6) and normal controls from fatal accidents (n=6). After establishment of air-liquid interface, cultures were stimulated with IL-4 and IL-13
ACCEPTED MANUSCRIPT
in combination, or vehicle, for 14-days. Genes passing a filter of FDR<0.05 and 2fold/vehicle for IL-4/IL-13 vs. vehicle comparison in either asthma or control groups were selected. Experiment 2: Bronchial epithelial cultures were established from bronchial
RI PT
•
epithelia obtained from endobronchial biopsies from asthmatics (n=6) or healthy
controls (n=3). After establishment of air-liquid interface, cultures were stimulated with IL-13 at 5 ng/ml or 10 ng/ml or with vehicle for 7, 14, or 21 days. Genes passing a filter
SC
of p<0.05 and 2-fold/vehicle for IL-13 vs. vehicle comparisons for both doses of IL-13 and for at least 2 time points, and common to both asthma and control groups were selected.
Experiment 3: Bronchial epithelial cultures were established from bronchial
M AN U
•
epithelia from fatal asthma cases (n=10). After establishment of air-liquid interface, cultures were stimulated with IL-13 or vehicle for 6-hours, 24-hours, 7-days, or 14-days. Genes passing a filter of FDR<0.05 and 2-fold/vehicle for IL-13 vs. vehicle comparison for at least 2 consecutive time points were selected.
TE D
Enrichment was evaluated on a per-subject basis using the R-Bioconductor package Gene Set Variation Analysis (GSVA, v 1.14.19), providing enrichment scores for each
RESULTS
EP
subject representing IL-13 pathway activity.
AC C
E2. Biomarker Analysis
108 acceptable airway mucosal biopsies were obtained for analysis (healthy controls n=25; mild n=28, moderate n=29, severe asthma n=26). 128 subjects had acceptable sputum samples available at screening or baseline, of which 74 subjects had matching airway biopsy microarray data (healthy controls n=16; mild n=19, moderate n=18, and severe asthma n=21). 188 subjects had serum analytes measured at baseline, of which 108 subjects had matching airway biopsy microarray data (healthy n=25; mild n=28, moderate n=29, severe asthma n=26).
ACCEPTED MANUSCRIPT
The mild asthma cohort was analyzed separately from the moderate and severe (moderate-severe) asthma cohorts because the mild asthma patients were not taking ICS, which could be influential on the biomarker profiles including airway-mucosa
RI PT
expression of Type-2-associated genes (as demonstrated by Woodruff et al8. Moderatesevere asthmatics were the focus of the analyses because they have unmet clinical need (reduced lung function despite ICS treatment), and those presenting with the Type-2 phenotype would be considered to have a persistent Type-2 phenotype (that is,
SC
persistent despite ICS treatment). Separate analyses on the moderate and severe cohorts were generally not undertaken because the sample size for Type-2-high
subgroups within each cohort would have resulted in substantially limited statistical
M AN U
power to observe significance of true differences.
Relationship between clinical biomarkers and biopsy Type-2 status by CCL26, periostin, or IL-13-IVS: further analyses Moderate-Severe Asthma
TE D
In the moderate-severe asthma cohorts, all airway-mucosal-CCL26-Type-2-high subjects had either high FENO (>=35 ppb) or high blood eosinophils (>=300/mm3), whereas 80% of airway-mucosal-periostin-high and 75% of airway-mucosal-IL-13-IVS-high subjects had high FENO or bEOS (for CCL26, Table 2 in main report; for periostin and IL-13-IVS,
EP
Table E2, and Table E3, respectively. Similar proportions of Type-2-low subjects (32 -
AC C
36%) using the 3 Type-2 group definitions had either high FENO or high bEOS. When defining airway-mucosa-high status by CCL26, periostin, and IL-13-IVS, 100%, 80%, and 75% of Type-2-high moderate-severe asthmatics were FENO-high and/or bEOS-high, respectively. Similar proportions of Type-2-low moderate-severe asthmatics were high for either FENO or bEOS (36%, 34%, and 32%, respectively). Mild asthma For mild asthma (Table 2 in main report), similar to moderate-severe asthma, the airway-mucosal-CCL26-high group was mostly FENO high (>=35ppb) (75% of high group)
ACCEPTED MANUSCRIPT
and the airway-mucosal-CCL26-low group had a majority that was FENO low (<35ppb) (67% of low group). However, for mild asthma both the airway-mucosal-CCL26-high and -low groups were mostly low for bEOS (75% and 92% of respective groups). Almost all of
RI PT
the mild asthma airway-mucosal-CCL26-high subjects (96%) had high serum IgE, compared to 50% of mild asthma airway-mucosal-CCL26-low subjects. Atopic status was similarly distributed across airway-mucosal-CCL26-high and -low groups as was serum IgE.
SC
Associations of airway gene expression with airway-mucosal-CCL26 status
Genes associated specifically with airway-mucosal-CCL26-high status in moderate-
M AN U
severe asthma, were cystatin-1, periostin, and intelectin-1 (>2-fold with FDR<0.05 vs. healthy controls for each; >2-fold with FDR=0.0013, 0.0008, and 0.18, respectively, vs. airway-CCL26-low). See the full list of probe sets for mild and moderate asthma (Tables E4 and E5).
In mild asthma (non-steroid-treated), similar to moderate-severe asthma, elevated
TE D
periostin, CST1, and ITLN1 biopsy gene expression was also specifically associated with the airway-mucosal-CCL26-high phenotype. In addition to these genes, CST4 (cystatin-4) and TPSAB1 (mast cell tryptase) were also specifically elevated in airway-mucosalCCL26-high mild asthma. CST4 was only modestly elevated in airway-mucosal-CCL26-
EP
high moderate-severe asthma compared to healthy controls (p=0.0001, 1.53-fold) and the airway-mucosal-CCL26-low group (p=0.0017, 1.37-fold), whereas TPSAB1 didn’t
AC C
even have a trend for being elevated in airway-mucosal-CCL26-high moderate-severe asthma. Rather, TPSAB1 was significantly decreased in airway-mucosal-CCL26-low moderate-severe asthma (p=0.0041, 1.58-fold) but not in airway-mucosal-CCL26-low mild asthma (p=0.12, +1.40-fold). Table E5 reports the comparison statistics for all probe sets associated with airway-mucosal-CCL26-high and -low mild asthma compared to healthy controls, passing significance criteria of p<0.05 and >2-fold. Statistical differences between Type-2-CCL26 groups in the mild asthma cohort could not be evaluated because of the 8 mild asthmatics with EPBR microarray data, 6 were
ACCEPTED MANUSCRIPT
airway-mucosal-CCL26-high but only 2 airway-mucosal-CCL26-low. However, for the Type-2-CCL26-high mild asthma subjects, periostin, CST1, CST4, and TPSAB1 were significantly elevated compared to healthy controls (p<0.05, 2-fold), as observed for the
RI PT
airway-mucosal-CCL26-high moderate-severe asthma group. ITLN1 was the mostly highly over-expressed gene in the airway-mucosal-CCL26-high mild asthma group,
consistent with elevation of this gene in biopsies from airway-mucosal-CCL26-high
moderate-severe asthma compared to healthy controls and the airway-mucosal-CCL26-
SC
low group. These 5 genes were also expressed significantly higher in biopsies from the airway-mucosal-CCL26-high mild asthma subjects compared to healthy controls and
airway-mucosal-CCL26-low mild asthma subjects. Airway mucosal brushings expression
M AN U
of SERPINB2 (one of the 3 ‘Type-2-high’ defining genes, along with periostin and CLCA1, from Woodruff et al4) was also elevated in airway-mucosal-CCL26-high mild asthma group. SERPINB2 was also expressed significantly higher in airway-mucosal-CCL26-high compared to -low moderate-severe asthma (p=0.020, 3.23-fold), but not compared to
E3. Induced Sputum
TE D
healthy controls (p=0.23).
Airway-mucosal-periostin and- IL-13-IVS status and induced sputum cellular profiles
EP
Using biopsy periostin gene expression and IL-13-IVS enrichment as alternate identifiers for Type-2-high and Type-2-low status, further analyses for sputum cell counts are presented in Table E6 (periostin) and Table E7 (IL-13-IVS). Lower proportions of airway-
AC C
mucosal-periostin high (69%) and airway-mucosal-IL-13-IVS high (65%) moderate-severe asthmatics were spEOS high compared to that for airway-mucosal-CCL26 high status (88%). Similar proportions of moderate-severe asthmatics were spEOS-high (45-46%) in the airway mucosa-Type-2 low groups defined by CCL26, periostin, and IL-13-IVS (p=0.049, 0.31, and 0.33, respectively, for Type-2 -high vs. -low). These observations further reinforce the application of CCL26 as the primary discriminator for airwaymucosal-Type-2 status in this report.
ACCEPTED MANUSCRIPT
When defining airway-mucosal-Type-2 high status by CCL26, periostin, and IL-13-IVS, 87.5%, 69%, and 65% of Type-2-high moderate-severe asthmatics were sputum EOShigh (i.e. ≥3% of leukocytes), respectively. Similar proportions of Type-2-low moderate-
RI PT
severe asthmatics were sputum EOS high (45%, 46%, and 45%, respectively). See Tables
SC
E6 and E7.
M AN U
E4. Longitudinal Stability of serum CCL26 and CCL17
Serum CCL26 was remarkably stable at repeated visits over 12 months. Figure E2 shows the concentrations of sCCL26 at Months 3, 6, and 12, stratified by baseline visit high/low status. Only 4 of 94 samples had a change from baseline in sCCL26-high/-low status (96% concordance). Two airway-mucosa-CCL26-high subjects who were sCCL26-high at baseline were transiently sCCL26-low for one post-baseline visit but sCCL26-high for the
TE D
other visits. One airway-mucosa-CCL26-high subject who was sCCL26-low at baseline was sCCL26-high at 3 and 12 but low at month 6. Because of limitations of sample availability, baseline samples were analyzed at different times than the months 3-12. Nevertheless, the correlations for sCCL26 between visits were very strong (r>0.96) for all
EP
pairwise comparisons, thus demonstrating the robustness of the detection method.
AC C
Baseline samples for sCCL17 were measured using a different assay system (R&D Systems) than for 3, 6 and 12 samples (RayBio, qualified and run by Cirquest Labs). For sCCL17, the correlations between visits (excluding baseline visit, which was a different assay) were only moderate (Spearman r = 0.7 – 0.8). Additionally, there was poor correlation of baseline sCCL17 with the measurements at months 3-12 (r= 0.20 – 0.48). However, baseline sCCL17 measurements correlated better with sCCL26 measurements (r = 0.5 to 0.6) than it did with sCCL17 measurements at months 3-12. Therefore, it is likely that the R&D Systems assay run for baseline samples was more reliable than the
ACCEPTED MANUSCRIPT
RayBio assay and may explain reduced stability of serum-CCL17-high/-low status assessed at baseline over time.
E5. Correlations between Type-2 inflammation biomarkers and
RI PT
eosinophilic phenotypic markers
Figure E3 presents the correlations between Type-2 inflammation biomarkers and eosinophilic phenotypic markers, for mild and moderate-severe asthma. The correlations of airway type 2
SC
inflammation with eosinophilic and other type 2 inflammation biomarkers are similar when using airway CCL26 expression compared to airway periostin expression and IL-13-IVS scores,
M AN U
albeit generally modestly stronger for CCL26.
Interestingly, for the correlation of submucosal eosinophil density vs. serum CCL26, the correlation is inverse for mild asthma (r= -0.57) but modestly positive for moderate-severe asthma (r=0.36). This is in contrast to airway-mucosal-CCL26 expression vs. submucosal eosinophil density, which are positively correlated in both mild (r=0.47) and moderate-severe (r=0.36) asthma. This is consistent with significant correlation of airway CCL26 expression and
TE D
serum CCL26 levels only in mod-severe asthma (r=0.41) but not mild asthma (r=0.13). Similar correlations are also observed with FENO and submucosal (and sputum) eosinophils, with significant positive correlations observed only in moderate-severe but not mild asthma.
EP
However, airway CCL26 expression correlates with submucosal eosinophils in both mild and mod-severe asthma, with a stronger correlation in mild asthma.
AC C
A possible mechanism to explain this difference is that in mild asthma, leakage out of the submucosa, whether for eosinophils in the airway lumen or soluble mediators like CCL26 into the circulation, may be more variable than in moderate-severe asthma with a persistent inflammatory phenotype (despite inhaled steroids). Whatever the mechanisms, there are clear differences in mild (untreated) asthma and moderate-severe asthma with a persistent type 2 inflammatory phenotype limiting the ability to classify airway type inflammation in mild asthma using non-invasive biomarkers (e.g., FENO, serum proteins).
RI PT
ACCEPTED MANUSCRIPT
GGH
SIDT1
ALOX15
CISH
HS3ST1
SLC26A4
BCL2L15
CST1
LRRC31
SLC39A8
C1QTNF1
CST2
NTRK1
SLC5A1
CA2
CST4
OBFC2A
SOCS1
CCBL1
CTSC
PCSK6
SUSD2
CCL26
DPP4
POSTN
USP54
CD274
FAM26E
SERPINB4
CD44
FETUB
SH2D1B
a
Genes significantly induced by IL-13 commonly in each of 3
AC C
EP
independent experiments of bronchial epithelial air-liquid interface culture
M AN U
CDH26
TE D
ADAMTS9
SC
Table E1. IL-13 in vitro stimulation (IVS) signature a
ACCEPTED MANUSCRIPT
N (% of Type-2 group) [% of
moderate-severe asthma
airway-mucosa-
airway-mucosa-
periostin-high
periostin-low
(n=19)
(n=9)
4 (21%) [80%]
1 (11%) [20%]
13 (68%) [81%]
3 (33%) [19%]
6 (32%) [50%]
6 (67%) [50%]
FENO ≥ 35 ppb
13 (68%) [81%]
3 (33%) [19%]
FENO < 35 ppb
6 (32%) [50%]
6 (67%) [50%]
Blood EOS ≥ 300/mm3
4 (21%) [80%]
1 (11%) [20%]
row group]
P-valuea
FENO ≥ 35 ppb and Blood 3
FENO ≥ 35 ppb or Blood Eos ≥ 300/mm
3
3
Blood EOS < 300/mm
3
15 (79%) [65%]
EP
Blood Eos ≥ 300/mm
0.1139
1.0000
8 (89%) [35%]
16 (84%) [76%]
5 (56%) [24%]
Serum IgE low
3 (16%) [43%]
4 (44%) [57%]
Atopy (Phadiatop): Positive
17 (89%) [74%]
6 (75%) [26%]
Atopy (Phadiatop): Negative
2 (11%) [50%]
2 (25%) [50%]
AC C
Serum IgE high
a
0.1139
TE D
Neither FENO ≥ 35 ppb nor
0.3553
airway-mucosa-
periostin-high
periostin-low
0.1652
0.5583
P-value
(n=20)
(n=35)
6 (30%) [55%]
5 (15%) [45%]
0.0139
16 (80%) [57%]
12 (34%) [43%]
0.0018
4 (20%) [15%]
23 (66%) [85%]
10 (50%) [53%]
9 (26%) [47%]
10 (50%) [29%]
25 (74%) [71%]
12 (60%) [60%]
8 (23%) [40%]
8 (40%) [23%]
27 (77%) [77%]
19 (95%) [46%]
22 (65%) [54%]
1 (5%) [8%]
12 (35%) [92%]
18 (90%) [43%]
24 (69%) [57%]
2 (10%) [15%]
11 (31%) [85%]
M AN U
Eos ≥ 300/mm
airway-mucosa-
SC
Mild asthma
RI PT
Table E2. Comparison of airway-mucosal-periostin status and clinical biomarkers
P-value for comparison between airway-mucosa-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test)
0.1390
0.0089
0.0188
0.1023
ACCEPTED MANUSCRIPT
Table E3. Comparison of airway-mucosal Type-2-IL-13-IVS status and clinical biomarkers
airway-mucosal-
airway-mucosal-
airway-mucosal-IL-
row group]
IL-13-IVS-low
P-value
a
13-IVS-high (n=15)
(n=24)
FENO ≥ 35 ppb and Blood 4 (27%) [80%]
1 (8%) [20%]
0.3553
11 (73%) [69%]
5 (38%) [31%]
0.1248
4 (27%) [33%]
8 (62%) [67%]
FENO ≥ 35 ppb
11 (73%) [69%]
5 (38%) [31%]
FENO < 35 ppb
4 (27%) [33%]
8 (62%) [67%]
3
3
Blood Eos ≥ 300/mm3
3
11 (73%) [92%]
1 (8%) [8%]
Blood EOS < 300/mm
3
4 (27%) [25%]
12 (92%) [75%]
Serum IgE low
2 (13%) [29%]
8 (62%) [38%]
Atopy (Phadiatop): Positive
15 (100%) [65%]
8 (67%) [35%]
Atopy (Phadiatop): Negative
0 (0%) [0%]
4 (33%) [100%]
a
0.3333
0.1977
5 (38%) [71%]
AC C
13 (87%) [62%]
EP
Blood EOS ≥ 300/mm
Serum IgE high
0.1248
TE D
Neither FENO ≥ 35 ppb nor
6 (25%) [55%]
5 (16%) [45%]
0.0139
18 (75%) [64%]
10 (32%) [36%]
0.0026
6 (25%) [22%]
21 (68%) [78%]
12 (50%) [63%]
7 (23%) [37%]
12 (50%) [34%]
23 (77%) [66%]
12 (50%) [60%]
8 (26%) [40%]
12 (50%) [34%]
23 (74%) [66%]
21 (88%) [51%]
20 (67%) [49%]
3 (12%) [23%]
10 (33%) [77%]
21 (88%) [50%]
21 (68%) [50%]
3 (12%) [23%]
10 (32%) [77%]
M AN U
FENO ≥ 35 ppb or Blood Eos ≥ 300/mm
P-value
13-IVS-low (n=31)
(n=13)
Eos ≥ 300/mm
airway-mucosal-IL-
IL-13-IVS-high
SC
N (% of Type-2 group) [% of
moderate-severe asthma
RI PT
Mild asthma
0.0282
P-value for comparison between airway-mucosal-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test)
0.0506
0.0914
0.111
0.1110
ACCEPTED MANUSCRIPT
Table E4. Moderate-severe asthma airway-mucosal-Type-2-CCL26-associated gene expression
asthma vs. Healthy
asthma vs. Healthy
Low (moderate-severe asthma)
FoldChange
P-value
RI PT
Gene Symbol
Type-2-CCL26 High vs. Type-2-CCL26
FDR-BH
FoldChange
P-value
FDR-BH
FoldChange
P-value
FDR-BH
4.01
1.1E-22
2.6E-18
4.00
3.5E-25
8.5E-21
5.84
1.8E-09
8.65E-06
3.77
2.1E-07
0.0013
5.32
5.1E-12
4.1E-08
2.95
1.1E-07
0.0008
Type-2-CCL26 high moderate-severe asthma vs. Healthy (p<0.05, >2-fold) 223710_PM_at
CCL26
1.00
0.9748
0.9994
206224_PM_at
CST1
1.55
0.0277
0.548
210809_PM_s_at
POSTN
1.80
0.0003
0.0673
1555778_PM_a_at
POSTN
1.87
0.0003
223597_PM_at
ITLN1
1.36
0.0961
228481_PM_at
---
1.25
0.0044
201884_PM_at
CEACAM5
1.74
0.0148
225720_PM_at
SYNPO2
1.65
0.0505
225895_PM_at
SYNPO2
1.70
0.0384
232119_PM_at
SYNPO2
1.61
211430_PM_s_at
IGH@
-2.64
217036_PM_at
LOC100293679
-2.01
206326_PM_at
GRP
-1.99
217179_PM_x_at
---
219159_PM_s_at
SLAMF7
222102_PM_at
GSTA3
207430_PM_s_at
MSMB
210297_PM_s_at
MSMB
SC
a
Type-2-CCL26 High moderate-severe
M AN U
Probe Set ID
Type-2-CCL26 Low moderate-severe
4.96
1.7E-10
9.9E-07
2.66
3.5E-06
0.0168
0.7447
3.48
2.1E-06
0.0072
2.56
5.2E-05
0.1845
0.2811
2.47
1.2E-13
1.4E-09
1.98
7.4E-11
8.9E-07
0.4531
2.50
0.0027
0.5019
1.43
0.1799
0.9997
0.6524
2.15
0.0253
0.7389
1.30
0.3879
0.9997
0.598
2.18
0.0215
0.7326
1.29
0.4028
0.9997
TE D
0.0693
0.5142
2.05
0.0093
0.6183
1.27
0.3211
0.9997
0.0012
0.1578
-2.40
0.0243
0.7361
1.10
0.7837
0.9997
0.0096
0.402
-2.03
0.046
0.7865
-1.01
0.9801
0.9997
6.2E-06
0.0091
-2.05
0.0003
0.2092
-1.03
0.8573
0.9997
-1.81
0.0077
0.3625
-2.05
0.0148
0.6724
-1.13
0.6372
0.9997
-1.82
0.0036
0.2673
-2.43
0.0012
0.3531
-1.33
0.2299
0.9997
-1.42
0.0263
0.543
-2.07
0.0007
0.3120
-1.45
0.0467
0.9946
-1.14
0.5268
0.9677
-2.03
0.0127
0.6488
-1.78
0.0243
0.9318
-1.14
0.5483
0.9694
-2.14
0.0095
0.6274
-1.88
0.0168
0.9134
-1.27
0.4143
1.65
0.0579
AC C
EP
0.0216
Type-2-CCL26 low moderate-severe asthma vs. Healthy (p<0.05, >2-fold) 206291_PM_at
NTS
-2.09
0.0011
0.1550
0.9691
0.9997
ACCEPTED MANUSCRIPT
a
Gene Symbol
Type-2-CCL26 High moderate-severe
Type-2-CCL26 High vs. Type-2-CCL26
asthma vs. Healthy
asthma vs. Healthy
Low (moderate-severe asthma)
P-value
FDR-BH
FoldChange
P-value
FDR-BH
FoldChange
P-value
FDR-BH
IGK@
-2.01
0.0152
0.4571
-1.78
0.1262
0.8757
1.13
0.7167
0.9997
202988_PM_s_at
RGS1
-2.20
1.3E-05
0.0140
-1.97
0.0032
0.5019
1.11
0.5974
0.9997
211430_PM_s_at
IGH@
-2.64
0.0012
0.1578
-2.40
0.0243
0.7361
1.10
0.7837
0.9997
211645_PM_x_at
---
-2.07
0.0139
0.4424
-1.96
0.0830
0.8412
1.06
0.8735
0.9997
217036_PM_at
LOC100293679
-2.01
0.0096
0.4020
-2.03
0.0460
0.7865
-1.01
0.9801
0.9997
205792_PM_at
WISP2
2.04
6.4E-06
0.0091
1.99
0.0007
0.3120
-1.03
0.8711
0.9997
202376_PM_at
SERPINA3
2.09
0.0104
0.4087
1.62
0.2001
0.9213
-1.29
0.4459
0.9997
224840_PM_at
FKBP5
2.51
2.4E-05
0.0208
1.82
0.0301
0.7524
-1.38
0.1910
0.9997
1556069_PM_s_at
HIF3A
2.39
0.0003
0.0737
1.63
0.1149
0.8659
-1.47
0.1673
0.9997
204560_PM_at
FKBP5
2.28
7.4E-07
0.0044
1.36
0.1301
0.8779
-1.67
0.0060
0.8506
224856_PM_at
FKBP5
2.98
9.4E-07
0.0045
1.75
0.0427
0.7840
-1.70
0.0319
0.9473
222124_PM_at
HIF3A
2.74
9.9E-05
0.0405
1.47
0.2379
0.9325
-1.86
0.0370
0.9665
244697_PM_at
---
2.43
1.28
0.4181
0.9691
-1.89
0.0236
0.9318
a
0.0003
M AN U
SC
216576_PM_x_at
TE D
FoldChange
RI PT
Probe Set ID
Type-2-CCL26 Low moderate-severe
0.0673
For the airway-mucosa biospy microarray dataset, probe sets passing the significance criteria for the indicated comparisons in section headers are displayed with the fold-change,
AC C
EP
nominal p-value, and FDR-BH for the comparisons indicated for each of the comparisons listed in the column headers
ACCEPTED MANUSCRIPT
Table E5. Mild asthma Type-2-CCL26-associated gene expression in airway-mucosa airway-mucosa -CCL26 High vs. airway-mucosa -CCL26 High (mild
asthma) vs. Healthy
asthma) vs. Healthy
Gene Symbol
FoldChange
P-value
FDR-BH
RI PT
Probe Set ID
airway-mucosa-CCL26 Low (mild
FoldChange
SC
Type-2-CCL26 high mild asthma vs. Healthy (p<0.05, >2-fold)
P-value
airway-mucosa -CCL26 Low (mild asthma)
FDR-BH
FoldChange
P-value
FDR-BH
4.59
2.8E-12
6.8E-08
3.83
5.8E-09
0.0001
2.85
5.7E-10
4.5E-06
2.90
1.5E-08
0.0002
0.9899
7.90
2.2E-10
2.7E-06
7.12
3.4E-08
0.0003
0.9956
2.05
6.0E-08
2.0E-04
2.09
7.4E-07
0.0045
0.9813
7.39
2.9E-09
1.2E-05
5.16
4.8E-06
0.023
0.9813
7.82
1.8E-09
8.4E-06
4.85
1.1E-05
0.0432
0.9813
5.47
7.5E-10
4.5E-06
3.17
4.5E-05
0.1342
0.9813
2.05
0.0001
0.1304
1.70
0.0112
0.9992
CCL26
1.20
0.3234
0.9813
228481_PM_at
---
-1.02
0.8974
0.9959
206224_PM_at
CST1
1.11
0.7146
206994_PM_at
CST4
-1.02
0.8745
210809_PM_s_at
POSTN
1.43
0.2431
1555778_PM_a_at
POSTN
1.61
0.1243
223597_PM_at
ITLN1
1.73
0.0303
207741_PM_x_at
TPSAB1
1.21
0.3179
207134_PM_x_at
TPSAB1 /// TPSB2
1.33
217023_PM_x_at
TPSAB1
1.34
210084_PM_x_at
TPSAB1
1.33
205683_PM_x_at
TPSAB1
1.40
201884_PM_at
CEACAM5
201058_PM_s_at
MYL9
1558438_PM_a_at
IGHA1
216474_PM_x_at
TPSAB1 /// TPSB2
205624_PM_at
CPA3
225721_PM_at
SYNPO2
228728_PM_at 204938_PM_s_at
TE D
M AN U
223710_PM_at
0.9813
2.24
0.0002
0.1645
1.69
0.0296
0.9992
0.1501
0.9813
2.08
0.0002
0.1645
1.55
0.0433
0.9992
0.1816
0.9813
2.08
0.0003
0.204
1.57
0.0448
0.9992
0.1227
0.9813
2.19
0.0002
0.1645
1.56
0.0549
0.9992
-1.01
0.9781
0.9997
2.01
0.0309
0.6606
2.03
0.0572
0.9992
1.23
0.4822
0.9813
2.26
0.004
0.5704
1.83
0.0581
0.9992
1.48
0.0666
0.9813
2.22
0.0001
0.1203
1.50
0.0678
0.9992
1.42
0.1591
0.9813
2.25
0.0007
0.3161
1.58
0.0845
0.9992
1.87
0.0172
0.9813
2.83
3.92E-05
0.0857
1.52
0.1239
0.9992
1.21
0.5651
0.9813
2.04
0.0205
0.6561
1.69
0.1351
0.9992
C7orf58
1.28
0.4399
0.9813
2.09
0.0124
0.6464
1.64
0.1389
0.9992
PLN
1.28
0.4081
0.9813
2.01
0.0123
0.646
1.57
0.152
0.9992
AC C
EP
0.2084
ACCEPTED MANUSCRIPT
airway-mucosa -CCL26 High vs. airway-mucosa-CCL26 Low (mild
airway-mucosa -CCL26 High (mild
asthma) vs. Healthy
asthma) vs. Healthy
airway-mucosa -CCL26 Low (mild Gene Symbol
asthma) FoldChange
P-value
FDR-BH
FoldChange
P-value
FDR-BH
FoldChange
P-value
FDR-BH
0.0176
0.6505
1.82
0.1815
0.9992
0.0107
0.6315
1.84
0.1884
0.9992
0.0261
0.6561
1.57
0.2522
0.9992
SORBS1
1.41
0.4228
0.9813
2.56
202274_PM_at
ACTG2
1.55
0.3177
0.9813
2.86
227843_PM_at
NDE1
1.38
0.3899
0.9813
2.16
207961_PM_x_at
MYH11
1.47
0.2741
0.9813
201497_PM_x_at
MYH11
1.45
0.2897
0.9813
225720_PM_at
SYNPO2
1.56
0.2353
0.9813
209071_PM_s_at
RGS5
1.48
0.1939
225895_PM_at
SYNPO2
1.43
0.3510
200795_PM_at
SPARCL1
1.47
0.2192
200974_PM_at
ACTA2
1.56
0.2152
211340_PM_s_at
MCAM
1.71
0.1033
201540_PM_at
FHL1
1.52
212077_PM_at
CALD1
1.69
209189_PM_at
FOS
-1.12
206336_PM_at
CXCL6
220542_PM_s_at
PLUNC
SC
218087_PM_s_at
0.0182
0.6505
1.48
0.294
0.9992
2.12
0.0204
0.6561
1.46
0.2969
0.9992
2.35
0.0147
0.6505
1.50
0.3014
0.9992
0.9813
2.04
0.0112
0.6347
1.38
0.3097
0.9992
0.9813
2.11
0.0342
0.6606
1.48
0.3279
0.9992
0.9813
2.02
0.0162
0.6505
1.37
0.3374
0.9992
0.9813
2.16
0.0209
0.6561
1.38
0.389
0.9992
0.9813
2.21
0.0092
0.6205
1.30
0.4476
0.9992
TE D
M AN U
2.18
0.9813
2.01
0.0347
0.6606
1.33
0.4507
0.9992
0.1428
0.9813
2.10
0.0252
0.6561
1.24
0.5653
0.9992
0.6947
0.9886
-2.04
0.0096
0.6209
-1.82
0.0563
0.9992
-1.31
0.2771
0.9813
-2.13
0.0013
0.3646
-1.63
0.0619
0.9992
-1.35
0.5507
0.9813
-3.46
0.0094
0.6205
-2.55
0.0837
0.9992
EP
0.2457
AC C
a
RI PT
Probe Set ID
For the airway-mucosa biospy microarray dataset, probe sets passing the significance criteria of p<0.05 and 2-fold for mild asthma Type-2-CCL26 high vs. healthy comparison
are displayed with the fold-change, nominal p-value, and FDR-BH for each of the comparisons listed in the column headers
ACCEPTED MANUSCRIPT
Table E6. Comparison of airway-mucosa -periostin status and sputum inflammatory cells
N (% of Type-2 group) [% of row group]
airway-mucosal-
airway-mucosal-
periostin-high
periostin-low
(n=14)
(n=5)
P-value
airway-mucosal-
airway-mucosal-
periostin-high
periostin-low
(n=13)
(n=26)
9 (69%) [43%]
12 (46%) [57%]
4 (31%) [22%]
14 (54%) [78%]
0 (0%) [0%]
Sputum EOS < 3%
10 (71%) [67%]
5 (100%) [33%]
6 (43%) [67%]
3 (60%) [33%]
2 (15%) [18%]
9 (35%) [82%]
4 (29%) [67%]
2 (40%) [33%]
2 (15%) [29%]
5 (19%) [71%]
0 (0%) [na]
0 (0%) [na]
2 (15%) [40%]
3 (12%) [60%]
4 (29%) [100%]
0 (0%) [0%]
7 (54%) [44%]
9 (35%) [56%]
M AN U
PMN<60%
Sputum Mixed
TE D
Gran.,PMN≥60%, EOS≥3% Sputum EOS≥3%, PMN<60%
SC
4 (29%) [100%]
Sputum PMN ≥60% EOS<3%
0.530
a
Sputum EOS ≥ 3%
Sputum Pauci EOS<3%,
EP
P-value for comparison between airway-mucosal-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test)
AC C
a
moderate-severe asthma
RI PT
Mild asthma
P-value
0.307
ACCEPTED MANUSCRIPT
Table E7. Comparison of airway-mucosal-IL-13-IVS status and sputum inflammatory cells
N (% of Type-2 group) [% of row group]
airway-mucosal-
airway-mucosal-
IL-13-IVS-high
IL-13-IVS-low
(n=11)
(n=8)
P-value
airway-mucosal-
airway-mucosal-
IL-13-IVS-high
IL-13-IVS-low
(n=17)
(n=22)
11 (65%) [52%]
10 (45%) [48%]
6 (35%) [33%]
12 (55%) [67%]
0 (0%) [0%]
Sputum EOS < 3%
7 (64%) [47%]
8 (100%) [53%]
4 (36%) [44%]
5 (63%) [56%]
4 (24%) [36%]
7 (32%) [64%]
3 (27%) [50%]
3 (38%) [50%]
2 (12%) [29%]
5 (23%) [71%]
0 (0%) [na]
0 (0%) [na]
3 (18%) [60%]
2 (9%) [40%]
4 (36%) [100%]
0 (0%) [0%]
8 (47%) [50%]
8 (36%) [50%]
M AN U
PMN<60%
Sputum Mixed
TE D
Gran.,PMN≥60%, EOS≥3% Sputum EOS≥3%, PMN<60%
SC
4 (36%) [100%]
Sputum PMN ≥60% EOS<3%
0.103
a
Sputum EOS ≥ 3%
Sputum Pauci EOS<3%,
EP
P-value for comparison between airway-mucosal-CCL26 status (high/low) vs. row variable subgroups (Fisher’s exact test)
AC C
a
moderate-severe asthma
RI PT
Mild asthma
P-value
0.334
ACCEPTED MANUSCRIPT
Table E8. Pre-specified vs. modeled optimal thresholds for classifiers of airway-mucosal-CCL26-high status Pre-specified threshold
a
Optimal threshold
b
c
blood eosinophils, cells/ul
Sensitivity
Specificity
PPV
NPV
Threshold
35
69%
76%
48%
89%
38
300
77%
76%
50%
91%
309 304
serum CCL17, pg/ml
d
285
85%
76%
52%
94%
serum CCL26, pg/ml
e
775
58%
90%
64%
88%
Specificity
PPV
NPV
69%
76%
48%
89%
24.1%
77%
76%
50%
91%
24.1%
85%
81%
58%
94%
24.1%
93%
70%
88%
22.6%
1595
58%
M AN U
a
c
Sensitivity
SC
FENO, ppb
Prevalence Threshold
RI PT
airway-mucosal-CCL26-high classifier
Model statistics for classification of airway-mucosa-CCL26-high status was based on pre-specified thresholds for the indicated classifer. FENO and bEOS
thresholds were pre-specified based on commonly used cut-offs; serum CCL17 and CCL26 thresholds were set as the 75th %ile of the healthy control distribution. b
Model statistics for classification of airway-mucosal-CCL26-high status was based on optimal thresholds established from ROC analyses, selecting the
TE D
thresholds for the classifiers that yielded the maximum for the sum of sensitivity and specificity.
FENO-high/-low and bEOS-high/-low assignments were equivalent using the pre-specified vs. optimal thresholds.
d
2 airway-mucosla-CCL26-low subjects assigned as 'serum CCL17-high' with the pre-specified threshold considered 'serum CCL17-low' with the optimal
threshold.
1 airway-mucosal-CCL26-low subject assigned as 'serum CCL26-high' with the pre-specified threshold considered 'serum CCL26-low' with the optimal
threshold.
AC C
e
EP
c
ACCEPTED MANUSCRIPT
Table E9. Airway-mucosal-CCL26 diagnostic characteristics of accessible biomarkers in moderate-severe asthma, multi-predictor models
Multi-predictor, discrete
Sensitivity
Specificity
AUC
sCCL26-high, sCCL17-high, (FENO≥35ppb or bEOS ≥300/mm3)
92%
93%
0.97
sCCL26 (log), sCCL17 (log), FENO (log), bEOS (log)
50%
93%
0.92
PPV
NPV
Prevalence†
79%
98%
23%
67%
86%
23%
SC
Multi-predictor, continuous
Classifiers
RI PT
airway-mucosal-CCL26-high classification model *
M AN U
* Logistic regression model statistics for classification of moderate-severe asthmatics to airway-mucosal-CCL26-high status, with 10-fold crossvalidation (on full population) to select model parameters and statistics based on a 50% probability cut-off: AUC, area-under-curve for receiver operating characteristic; NPV, negative predictive value; PPV, positive predictive value. † Prevalence based on the proporaon of airway-mucosal-CCL26-high subjects in the ADEPT moderate-severe asthma group, with PPV and NPV statistics based on this prevalence. ADEPT is a cross-sectional study that has rigid enrollment criteria, therefore the prevalence in other asthma
AC C
EP
TE D
study populations may vary based on recruitment criteria.
ACCEPTED MANUSCRIPT
Figure Legends Figure E1. Segregation of Type-2 high versus low by airway-mucosal CCL26, periostin, and IL-13-IVS gene expression.
RI PT
Airway mucosal biopsy gene expression is presented for CCL26, and periostin as log2 intensity from microarray experiments and for IL-13-IVS as GSVA enrichment score (GSVA ES), in the healthy nonatopic, mild, moderate and severe asthma cohorts, stratified by Type-2 high (red) vs. low (blue) groups defined by (A) CCL26 expression, (B) periostin expression, and (C) IL-13-IVS enrichment. The highlighted
Figure E2. Longitudinal evaluation of serum CCL26 levels.
SC
horizontal (y-axis) line indicates the threshold for Type-2 high vs. low status of the plotted biomarker.
M AN U
Serum CCL26 levels (y-axis) are shown for asthma subjects (all severities), stratified by visit time point and by baseline sCCL26 category (high, above or low, below the 75th percentile of healthy control population). Filled and open symbols respectively indicate samples above and below the threshold at 775 pg/ml for high and low sCCL26 status, based on 75th percentile of healthy control population from baseline analysis. Cramer’s V statistic for concordance of serum CCL26 categories at baseline vs. 3, 6,
TE D
and 12 month visits were 0.87, 1.00, and 0.87, respectively.
Figure E3. Correlations between Type-2 inflammation and eosinophilic phenotype biomarkers. The Spearman’s correlation coefficient (RSp) and associated p-values are shown in the bottom-left and topright halves, respectively, of tables for (A) mild asthma and (B) moderate-severe asthma groups for the
EP
indicated biomarkers. Correlation coefficient values are shaded as indicated in key to right of tables. Pvalues <0.00143 pass the Bonferroni family-wise error rate of 0.05 for 35 pair-wise tests and are
AC C
indicated in bold.
Figure E4. Correlations between Type-2 inflammation and eosinophilic phenotype biomarkers (samples sizes).
Sample sizes for the pair-wise Spearman’s correlation tests reported in Figure E3, for mild asthma (topright half) and moderate-severe asthma (bottom-left half).
RI PT
ACCEPTED MANUSCRIPT
Supplement References
5.
6.
7.
8.
9.
SC
M AN U
4.
TE D
3.
EP
2.
Kelly MM, Efthimiadis A, Hargreave FE. Induced sputum : selection method. Methods Mol Med 2001; 56:77-91. Komiya A, Nagase H, Yamada H, Sekiya T, Yamaguchi M, Sano Y, et al. Concerted expression of eotaxin-1, eotaxin-2, and eotaxin-3 in human bronchial epithelial cells. Cell Immunol 2003; 225:91-100. Syed F, Huang CC, Li K, Liu V, Shang T, Amegadzie BY, et al. Identification of interleukin-13 related biomarkers using peripheral blood mononuclear cells. Biomarkers 2007; 12:414-23. Woodruff PG, Modrek B, Choy DF, Jia G, Abbas AR, Ellwanger A, et al. T-helper type 2-driven inflammation defines major subphenotypes of asthma. Am J Respir Crit Care Med 2009; 180:388-95. Choy DF, Modrek B, Abbas AR, Kummerfeld S, Clark HF, Wu LC, et al. Gene expression patterns of Th2 inflammation and intercellular communication in asthmatic airways. J Immunol 2011; 186:1861-9. Li G, Oparil S, Sanders JM, Zhang L, Dai M, Chen LB, et al. Phosphatidylinositol-3-kinase signaling mediates vascular smooth muscle cell expression of periostin in vivo and in vitro. Atherosclerosis 2006; 188:292-300. Horiuchi K, Amizuka N, Takeshita S, Takamatsu H, Katsuura M, Ozawa H, et al. Identification and characterization of a novel protein, periostin, with restricted expression to periosteum and periodontal ligament and increased expression by transforming growth factor beta. J Bone Miner Res 1999; 14:1239-49. Woodruff PG, Boushey HA, Dolganov GM, Barker CS, Yang YH, Donnelly S, et al. Genome-wide profiling identifies epithelial cell genes associated with asthma and with treatment response to corticosteroids. Proc Natl Acad Sci U S A 2007; 104:15858-63. Hanzelmann S, Castelo R, Guinney J. GSVA: gene set variation analysis for microarray and RNAseq data. BMC Bioinformatics 2013; 14:7.
AC C
1.
ACCEPTED MANUSCRIPT
Figure E1. Segregation of Type-2 high versus low by airway-mucosaL CCL26, periostin, and IL-13-IVS gene expression.
Healthy Mild
RI PT
POSTN, log2log2 intensity EBBX-POSTN, intensity
IL-13 IVS,IVS, enrichment EBBX-IL-13 enrichmentscore score
A
CCL26:
Cohort:
Cohort:
M AN U
CCL26, log2 intensity
IL-13 IVS, enrichment score
B
POSTN:
POSTN:
TE D
AC C
EP
CCL26, log2 intensity IL13 IVS: Cohort:
POSTN, log2 intensity
Cohort:
Cohort:
C
SC
CCL26:
IL13 IVS: Cohort:
Moderate Severe
ACCEPTED MANUSCRIPT
AC C
EP
TE D
M AN U
SC
RI PT
Figure E2. Longitudinal evaluation of serum CCL26 levels.
ACCEPTED MANUSCRIPT
Figure E3. Correlations between Type-2 inflammation and eosinophilic phenotype biomarkers
serum IgE
serum CCL26
serum CCL17
RI PT
Blood Eos.
Sputum Eos.
FENO
Submucosal Eos.
RSp <0 0.00 0.15 0.30 0.45 0.60 0.75 0.90 1.00
serum IgE
serum CCL26
serum CCL17
SC Blood Eos.
Sputum Eos.
Submucosal Eos.
EBBX IL13 IVS
TE D
0.66 0.55 0.46 0.36 0.25 0.41 0.50 0.41 0.31
AC C
EBBX CCL26 EBBX POSTN EBBX IL13 IVS FENO Submucosal Eos. Sputum Eos. Blood Eos. serum CCL17 serum CCL26 serum IgE
<0.0001 <0.0001 0.0005 0.0149 0.1253 0.0020 0.0001 0.0021 0.0239 0.0002 <0.0001 0.0455 0.0516 0.0005 0.0010 0.0668 0.0208 0.48 0.0002 0.0104 0.1003 0.0421 0.0129 0.0835 0.1330 0.54 0.49 <0.0001 <0.0001 0.0002 0.0143 0.1738 0.0007 0.30 0.38 0.58 0.0834 0.0153 0.0099 0.0181 0.2484 0.31 0.27 0.53 0.32 0.0009 0.3865 0.8188 0.0656 0.45 0.28 0.36 0.36 0.37 0.1420 0.0762 0.0156 0.44 0.34 0.24 0.38 0.10 0.15 <0.0001 0.9040 0.25 0.24 0.14 0.36 0.03 0.18 0.55 0.5160 0.31 0.21 0.33 0.18 0.21 0.24 0.01 0.07
EP
Correlation (RSp)
EBBX CCL26
P-value
EBBX POSTN
B. Moderate-severe asthma
M AN U
0.85 0.69 0.41 0.47 0.05 0.36 0.37 0.13 0.27
<0.0001 <0.0001 0.0305 0.0346 0.8543 0.0635 0.0609 0.5512 0.1597 0.0001 0.0435 0.2179 0.3212 0.0706 0.0062 0.6190 0.0610 0.67 0.0017 0.1236 0.1936 0.1624 0.0977 0.9574 0.0598 0.38 0.57 0.2413 0.1478 0.1475 0.3969 0.2115 0.2969 0.29 0.36 0.27 0.6667 0.0600 0.7240 0.0115 0.1959 0.24 0.31 0.26 -0.12 0.1062 0.7399 0.4531 0.3118 0.35 0.27 0.20 0.42 0.29 0.7787 0.7873 0.2023 0.51 0.33 0.12 -0.08 -0.06 0.04 0.0002 0.8626 0.10 -0.01 0.18 -0.57 -0.14 -0.04 0.51 0.9788 0.36 0.36 0.15 0.29 0.18 0.18 0.02 0.00
FENO
EBBX CCL26 EBBX POSTN EBBX IL13 IVS FENO Submucosal Eos. Sputum Eos. Blood Eos. serum CCL17 serum CCL26 serum IgE
EBBX IL13 IVS
Correlation (RSp)
EBBX CCL26
P-value
EBBX POSTN
A. Mild asthma
RSp <0 0.00 0.15 0.30 0.45 0.60 0.75 0.90 1.00
ACCEPTED MANUSCRIPT
EP AC C
31 46 44 43 45
25 25 25 48 19 31 48 48
101 98 102
serum IgE
27 27 27 51 20 32 51
RI PT
Blood Eos.
Sputum Eos. 78 77 74 76
28 28 28 52 21 32
serum CCL26
45 78 102 101 98 102
19 19 19 32 16
serum CCL17
54 45 39 55 54 53 54
20 20 20 21
SC
55 54 45 39 55 54 53 54
28 28 28
Submucosal Eos.
28 28
M AN U
28
FENO
55 55 54 45 39 55 54 53 54
EBBX IL13 IVS
EBBX CCL26 EBBX POSTN EBBX IL13 IVS FENO Submucosal Eos. Sputum Eos. Blood Eos. serum CCL17 serum CCL26 serum IgE
TE D
ModerateSevere
EBBX CCL26
Mild
EBBX POSTN
Figure E4. Correlations between Type-2 inflammation and eosinophilic phenotype biomarkers (sample sizes)
100 102
99
28 28 28 52 21 32 52 51 48