Research

JAMA | Original Investigation

Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes

in Youth
Molin Yue, MS; Kristina Gaietto, MD, MPH; Yueh Ying Han, PhD; Franziska J. Rosser, MD, MPH; Zhongli Xu, BS;
Christopher Qoyawayma, BSE; Edna Acosta-Perez, PhD; Glorisa Canino, PhD; Erick Forno, MD, MPH;
Wei Chen, PhD; Juan C. Celedón, MD, DrPH

Editorial
IMPORTANCE T helper 2 (T2) cells and T helper 17 (T17) cells are CD4+ T cell subtypes involved Multimedia
in asthma. Characterizing asthma endotypes based on these cell types in diverse groups is
important for developing effective therapies for youths with asthma. Supplemental content

OBJECTIVE To identify asthma endotypes in school-aged youths aged 6 to 20 years by

examining the distribution and characteristics of transcriptomic profiles in nasal epithelium.

DESIGN, SETTING, AND PARTICIPANTS Cross-sectional analysis of nasal epithelial samples

from 3 studies of youths with asthma aged 6 to 20 years: Stress and Treatment Response
in Puerto Rican and African American Children with Asthma (STAR; N = 156),
Epigenetic Variation and Childhood Asthma in Puerto Ricans (EVA-PR; N = 237), and
Vitamin D Kids Asthma (VDKA; N = 66).

MAIN OUTCOMES AND MEASURES The primary outcome was nasal epithelial transcription
profiles of 3 T2 and 5 T17 pathway genes. Clinical characteristics, total and allergen-specific
immunoglobulin E (IgE), blood eosinophils, and lung function were compared across profiles
in all studies.

RESULTS Mean ages for STAR, EVA-PR, and VDKA participants were 14.2, 15.4, and 10.3 years,
respectively. The percentage of female participants ranged from 41% to 53.2% across studies.
The predominant race or ethnicity was Puerto Rican in EVA-PR (100%) and Black or
African American in STAR (71.8%) and VDKA (57.6%). Three transcriptomic profiles were
identified: high T2 expression (T2HIGH), high T17 expression (T17HIGH), and low expression of
both pathways (T2LOW/T17LOW). Across studies, T2HIGH was present in 23% to 29% of
participants, T17HIGH in 35% to 47%, and T2LOW/T17LOW in 30% to 38%. In each study,
median total IgE and blood eosinophils for the T2HIGH profile was higher than for the T2LOW
profiles (IgE, 584-869 vs 105-382 IU/mL; eosinophils, 343-560 vs 164-413 cells/mL).
Of the participants in all profiles, at least 50% had 1 or more positive allergen-specific IgEs.
A differential expression meta-analysis identified 3516 and 2494 differentially expressed
genes for the T2HIGH and T17HIGH profiles, respectively. The T17HIGH profile was associated
with interleukin 17 and neutrophil signaling pathways and the T2HIGH profile was associated
with interleukin 13 signaling pathways.

CONCLUSIONS AND RELEVANCE Nasal transcriptomic profiles consistent with T2-high, Author Affiliations: Division of
T17-high, and T2-low/T17-low endotypes occurred in similar proportions across 3 studies of Pediatric Pulmonary Medicine, UPMC
Children’s Hospital of Pittsburgh,
predominantly racially and ethnically minoritized youths with asthma. Most participants University of Pittsburgh, Pittsburgh,
had T2-low asthma endotypes and sensitization to 1 or more allergens was common among Pennsylvania (Yue, Gaietto, Han,
these endotypes. Rosser, Xu, Qoyawayma, Forno, Chen,
Celedón); Department of Biostatistics
and Health Data Science, School of
Public Health, University of
Pittsburgh, Pittsburgh, Pennsylvania
(Yue, Chen); Behavioral Sciences
Research Institute, San Juan,
Puerto Rico (Acosta-Perez, Canino);
Department of Pediatrics, Medical
Science Campus, University of Puerto
Rico, San Juan (Canino).
Corresponding Author: Juan C.
Celedón, MD, DrPH, Division of
Pulmonary Medicine, UPMC
Children’s Hospital of Pittsburgh,
JAMA. doi:10.1001/jama.2024.22684 4401 Penn Ave, Pittsburgh, PA 15224
Published online January 2, 2025. ([email protected]).

(Reprinted) E1
© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Research Original Investigation Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes

A
sthma is the most common chronic respiratory illness
among children worldwide. 1 In the US, childhood Key Points
asthma disproportionately affects racially and ethni-
Questions Can nasal epithelial gene expression be used to
cally minoritized groups,2 with higher rates of emergency de- identify asthma endotypes in youths aged 6 to 20 years, and
partment and urgent care visits for asthma in Puerto Rican which asthma endotypes are most common?
(23.5%) and non-Hispanic Black (26.6%) children than non-
Findings Cluster analysis of 8 signature genes in nasal epithelium
Hispanic White youths aged 6 to 20 years (12.1%).3
from 3 studies of predominantly racially and ethnically minoritized
Asthma is a heterogeneous disease. Studies of gene ex- youths with asthma revealed 3 mutually exclusive profiles
pression in bronchial epithelium from adults have identified corresponding to T helper 2 (T2)–high, T helper 17 (T17)–high, and
molecular mechanisms or endotypes of asthma.4-6 T helper 2 T2-low/T17-low asthma endotypes, which were identified in similar
(T2)–high asthma is characterized by enhanced T2 immune re- proportions across the 3 studies, despite differences in racial and
sponses with high serum levels of T2 cytokines (interleukin ethnic composition and geographic location. T2-high asthma was
the least common asthma subtype in all 3 studies (23%-29%).
[IL]-4, IL-5, and IL-13) and elevated total immunoglobulin E
(IgE) and airway eosinophils. Several monoclonal antibodies Meaning Nasal transcriptomic profiles were similar across 3
are currently used to treat T2-high asthma, with medications studies of mostly Puerto Rican and African American youths with
selected according to T2 biomarkers, such as blood eosino- asthma and revealed predominantly T2-low asthma endotypes.
phils and total IgE.7
Far less is known about identifying and treating non–T2-
high asthma endotypes, collectively referred to as T2-low study of response to inhaled corticosteroids (ICS) in youths
asthma8,9 and including T17-high asthma and T2-low/T17- aged 8 to 20 years recruited at the University of Puerto Rico
low asthma. Although T17-high asthma is characterized by neu- Medical Center and UPMC Children’s Hospital of Pittsburgh
trophilic airway inflammation and high IL-17 and IL-22 se- from June 5, 2018, to May 16, 2022. Eligibility criteria in-
rum levels,10,11 T2-low/T17-low, or paucigranulocytic, asthma cluded having at least 3 Puerto Rican or at least 3 African Ameri-
lacks eosinophilic or neutrophilic airway inflammation and is can grandparents (per parental report); physician-diagnosed
poorly understood.8 mild to moderate persistent asthma; no use of nasal, inhaled,
T2-high asthma has often been assumed to be the most or systemic corticosteroids for at least 4 weeks; no current
common asthma endotype12-15 because asthma with atopy smoking and no former smoking history of 5 or more pack-
(sensitization to ≥1 allergens) has been equated with T2-high years; and no upper respiratory infection (URI) for at least 4
asthma. This assumption has recently been challenged by epi- weeks. At a baseline visit, participants completed question-
demiologic studies using biomarkers such as eosinophils and naires on demographics and respiratory health, performed spi-
total IgE.8 Additionally, studies of bronchial epithelial gene ex- rometry, had blood samples tested for total and allergen-
pression in predominantly non-Hispanic White adults with specific IgEs and eosinophil count, and underwent collection
moderate to severe asthma have found that only 14% to 27% of nasal specimens. After instilling lidocaine 1%, nasal samples
of individuals had upregulated expression of T2 genes.5,6 were collected by scraping each inferior turbinate using 4
Few studies have been published about the frequency of curettes, with RNA from these samples sequenced as previ-
asthma endotypes in youth, including underserved groups, due ously described.17 After nasal sampling, participants were
to the difficulty of conducting bronchoscopies in children. treated with daily medium-dose inhaled mometasone for
A promising approach to characterizing transcriptomic pro- 6 weeks, with adherence measured using a dose counter.
files of asthma endotypes in youth is nasal sampling, given its Spirometry was conducted and fractional exhaled nitric ox-
safety, feasibility, and evidence of high correlation between ide (FeNO) measurements were taken at baseline and after
nasal and bronchial epithelial expression in children.16 6 weeks of ICS.
Based on recent studies of bronchial epithelial transcrip-
tomics in adults, this study hypothesized that T2-low asthma EVA-PR
endotypes are more common than T2-high asthma endotypes The Epigenetic Variation and Childhood Asthma in Puerto
in youths, particularly those in racially and ethnically minori- Ricans (EVA-PR) study was a case-control study of asthma in
tized groups with high asthma burden. To test this hypothesis, Puerto Rican youths aged 9 to 20 years.17 All participants had
this study examined nasal epithelial expression of 8 T2 and T17 4 Puerto Rican grandparents (per parental report), currently
signature genes5 in 3 studies of predominantly Puerto Rican and did not smoke or previously smoked with 5 or fewer pack-
Black youths with asthma to identify transcriptomic profiles of years of smoking history, had not received systemic cortico-
asthma endotypes and their associated characteristics. steroids, and had no URIs for at least 4 weeks. Participants were
recruited from February 12, 2014, to May 8, 2017, in San Juan
and Caguas, Puerto Rico, as previously described17; only par-
ticipants with asthma are included in this analysis (eMethods
Methods in Supplement 1). Participants completed questionnaires on
Study Populations demographics and respiratory health, performed spirometry,
STAR had blood samples tested for total and allergen-specific IgEs
The Stress and Treatment Response in Puerto Rican and African and eosinophil count, and underwent collection of nasal
American Children with Asthma (STAR) study was a 6-week samples for gene expression, as in STAR.

E2 JAMA Published online January 2, 2025 (Reprinted) jama.com

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes Original Investigation Research

VDKA We compared demographic and clinical characteristics,

The Vitamin D Kids Asthma (VDKA) study was a 48-week mul- biomarkers, and lung function across the 3 profiles in each
ticenter, randomized, double-blind, placebo-controlled trial of study, using analysis of variance or the Kruskal-Wallis test
vitamin D3 for severe asthma exacerbations in youths aged 6 for continuous variables and χ2 tests for categorical variables.
to 16 years, recruited from February 28, 2016, to September In STAR, we compared the characteristics of participants
20, 2019.18 An ancillary study for nasal sampling was con- with and without ICS responsiveness (defined as ≥5% incre-
ducted at the Pittsburgh site. Eligible children had mild per- ment from baseline forced expiratory volume in 1 second af-
sistent asthma with 1 or more severe asthma exacerbations in ter ICS treatment21) across profiles using t tests or Wilcoxon
the previous year and a serum vitamin D level of less than rank-sum tests for continuous variables and χ2 tests for cat-
30 ng/mL, had not used systemic corticosteroids and had no egorical variables.
URIs for at least 4 weeks prior to nasal sampling, and cur- Next, we assessed the predictive capacity of biomarkers
rently did not smoke or previously smoked with 5 or fewer for the T2HIGH profile in STAR. Receiver operating character-
pack-years of smoking history. Prior to randomization, par- istic curves and values for the corresponding area under the
ticipants completed questionnaires on demographics and re- curve were generated using the plotROC library,22 and opti-
spiratory health, performed spirometry, had blood testing for mal cutoff points were determined using the Youden Index.23
total and allergen-specific IgEs and eosinophil count, and un- For simplicity and applicability in the clinical setting, a deci-
derwent collection of nasal samples for gene expression stud- sion tree machine learning method24 was implemented to in-
ies, as in STAR. Race and ethnicity were assessed to ensure in- tegrate multiple biomarkers (FeNO, total IgE, and eosino-
clusion of diverse participants. Parents of participants selected phils) to more accurately predict the T2HIGH profile. The method
1 option from the following categories: American Indian or was initially implemented in STAR and subsequently applied
Alaska Native, Asian, Black or African American, Multiracial to EVA-PR and VDKA (using predicted FeNO estimates in those
(defined as checking more than 1 category), White, or Other. studies) (eMethods in Supplement 1).
To validate the clusters’ biologic relevance, we con-
Study Approvals and Consent ducted a differential expression meta-analysis of partici-
STAR and EVA-PR were approved by the institutional review pants with a consensus profile using multivariable negative
boards of the University of Puerto Rico and the University of binomial regression in DESeq225 to assess for upregulation of
Pittsburgh. VDKA was approved by the institutional review other T2 pathway genes in the T2HIGH cluster and upregula-
boards of all participating institutions. In all 3 studies, writ- tion of other T17 pathway genes in the T17HIGH cluster. En-
ten parental consent and assent were obtained from partici- riched canonical pathways by upregulated differentially ex-
pants younger than 18 years and written consent was ob- pressed genes for the T2HIGH and T17HIGH transcriptomic
tained from participants 18 years or older. profiles were identified by ingenuity pathway analysis.26 We
performed consensus weighted gene coexpression network
Statistical Analysis analysis 27 to identify gene modules followed by gene
In all studies, expression of 3 IL-13–inducible or T2 signature ontology28 pathway enrichment analysis. Based on the gene
genes (periostin [POSTN], serpin family B member 2 [SERPINB2], modules we identified and significant genes in the differen-
and chloride channel, calcium-activated, family member 1 tial expression analysis (false discovery rate–adjusted P < .01),
[CLCA1])4 and 5 IL-17–inducible or T17 (C-X-C motif chemo- we built an expanded signature gene panel.
kine ligand 1 [CXCL1], CXCL2, CXCL3, colony-stimulating fac- This report follows the STROBE reporting guidelines for
tor 3 [CSF3], and IL-8)5 signature genes (used in a study of bron- cross-sectional studies. All analyses were conducted using
chial epithelial gene expression in adults with asthma6) was R version 4.3.0 (R Foundation).
normalized to transcript per million values and then trans-
formed to a logarithmic scale (log2) after adding a small con-
stant value (0.1).
K-means clustering19 analysis was performed across all
Results
samples in each study based on expression of the selected T2 Figure 1 shows the flowchart for selection of participants in
and T17 signature genes, with the optimal number of clusters this analysis (STAR, N = 156; EVA-PR, N = 237; and VDKA,
(k = 3) determined by the NbClust R package version 3.0.1 N = 66). The Table shows the characteristics of youths with
(R Foundation)20 in all cohorts and annotated according to asthma in the 3 studies. All EVA-PR participants were Puerto
the T2 panel 3-gene mean and T17 panel 5-gene mean into Rican, while more than half of STAR and VDKA participants
T2HIGH/T17LOW (hereafter, T2HIGH), T2LOW/T17HIGH (hereafter, were African American or non-Hispanic Black. VDKA partici-
T17HIGH), and T2LOW/T17LOW profiles. To validate the k-means pants were younger (mean age, 10.3 years) than those in
clustering transcriptomic profile assignments, hierarchical, STAR (mean age, 14.2 years) and EVA-PR (mean age, 15.4
k-medoid, Gaussian mixture models, and spectral clustering years), while the proportion of female participants in STAR
analyses were performed across all samples in each study (53.2%) was higher than that in EVA-PR (43%) or VDKA (41%).
based on expression of the same T2 and T17 genes. Partici- A total of 45.6% of participants in EVA-PR, 56.1% in VDKA,
pants with the same transcriptomic profile assignment across and 65.4% in STAR had overweight or obesity. Across the 3
all 5 clustering approaches were deemed to have a consen- studies, 17% to 38% had private or employer-based health
sus profile. insurance; 62% to 74% had medical assistance, Medicaid, or

jama.com (Reprinted) JAMA Published online January 2, 2025 E3

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Research Original Investigation Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes

Figure 1. Selection of Participants in a Study of Nasal Epithelial Transcriptomics of Asthma in Youths Aged 6 to 20 Years

STAR EVA-PR VDKA

249 Puerto Rican or African American youths 1111 Puerto Rican households with ≥1 youths 115 Youths aged 6-16 y with mild persistent
aged 8-20 y with mild to moderate aged 9-20 y assessed for eligibility asthma and vitamin D insufficiency
persistent asthma assessed for eligibility enrolled in Pittsburgh

391 Excluded
63 Excluded 180 Could not be reached
36 Not eligible 121 Moved away
13 Lost to follow-up 66 Missing contact information
24 Wrong age range 22 Refused to participate in
5 Refused to participate ancillary study
5 Could not complete PFTs
3 Had no bronchodilator response
or airway hyperresponsiveness 720 Eligible households
1 Withdrawal

82 Not interviewed (recruitment

goal reached) 93 Enrolled in ancillary study
186 Enrolled and administered
inhaled corticosteroids
638 Invited to participate (1 per household),
including 298 with asthma
17 Lost to follow-up
4 Refused blood draw or PFTs 27 Missing transcriptomic data
95 Refused to participate

165 Completed the 6-wk study 543 Participants enrolled

66 Included in main analyses with

306 Excluded from analysis transcriptomic cluster assignment
9 Missing transcriptomic data
274 Without asthma (controls)
32 Missing transcriptomic data

156 Included in main analyses with

transcriptomic cluster assignment 237 Included in main analyses with
transcriptomic cluster assignment 14 Transcriptomic profile not the
same in all 5 clustering methods
27 Transcriptomic profile not the
same in all 5 clustering methods 60 Transcriptomic profile not the
same in all 5 clustering methods

129 Included in meta-analysesa with same 177 Included in meta-analysesa with same 52 Included in meta-analysesa with same
transcriptomic profile assignment transcriptomic cluster assignment transcriptomic cluster assignment
across 5 clustering methods across 5 clustering methods across 5 clustering methods

a
EVA-PR indicates Epigenetic Variation and Childhood Asthma in Puerto Ricans A differential expression meta-analysis and consensus-weighted gene
study; PFTs, pulmonary function tests; STAR, Stress and Treatment Response coexpression analysis were conducted in participants with a consensus profile
in Puerto Rican and African American Children with Asthma study; and (those whose transcriptomic profile assignment was the same using all 5
VDKA, Vitamin D Kids Asthma study. clustering approaches).

State Children’s Health Insurance Program; and 1% to 4% T2LOW/T17LOW) and the 5-gene mean was higher in the T17HIGH
were uninsured. profile than in the T2HIGH and T2LOW/T17LOW profiles (P < .001
After conducting k-means clustering based on the 8 origi- in all instances) (eFigure 1 and eTable 1 in Supplement 1). In
nal signature genes,5,6 3 profiles were identified in each study: all studies, there was high concordance for identification of the
(1) upregulated expression of T2 signature genes (3-gene mean) T2 HIGH and T17 HIGH profiles across the 5 clustering ap-
or T2HIGH profile (STAR, n = 40 [25.6%]; EVA-PR, n = 69 [29.1%]; proaches (91.8%, 78.3%, and 80.4% in STAR, EVA-PR, and
and VDKA, n = 15 [22.7%]); (2) upregulated expression of T17 VDKA, respectively), with lower but substantial concordance
signature genes (5-gene mean) or T17HIGH profile (STAR, n = 57 (≥80% for ≥3 approaches) for the T2 LOW /T17 LOW profile
[36.5%]; EVA-PR, n = 83 [35.0%]; and VDKA, n = 31 [47.0%]), (eTables 1-3, eFigures 2-4, and eResults in Supplement 1).
and (3) no upregulated expression of T2 or T17 signature genes In STAR, participants with a T2HIGH profile were more likely
or T2LOW/T17LOW profile (STAR, n = 59 [37.8%]; EVA-PR, n = 85 to be Puerto Rican and have physician-diagnosed allergic rhi-
[35.9%]; and VDKA, n = 20 [30.3%]) (Table; Figure 2; eFig- nitis, a higher annual household income, and 1 or more asthma-
ure 1 in Supplement 1). Results were similar in a sensitivity related visits to the emergency department or urgent care in
analysis including only individuals assigned to the same en- the previous year than participants with T2 LOW profiles
dotype profile using all 5 clustering methods deemed to have (eTable 4 in Supplement 1). A total of 81.9% of STAR partici-
a consensus profile (STAR, n = 129; EVA-PR, n = 177; and VDKA, pants with T2LOW profiles were African American. In STAR and
n = 52) (Figure 2). In all studies, the 3-gene mean was higher VDKA, participants with a T17HIGH profile were younger than
in the T2HIGH profile than in the T2LOW profiles (T17HIGH and those with a T2LOW/T17LOW profile. Overweight or obesity was

E4 JAMA Published online January 2, 2025 (Reprinted) jama.com

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes Original Investigation Research

Table. Demographic and Clinical Characteristics, Biomarkers, and Lung Function of Participants in Each Study
STAR, No./total No. EVA-PR, No./total No. VDKA, No./total No.
(%)a (N = 156) (%)a (N = 237) (%)a (N = 66)
Age, mean (SD), y 14.2 (3.0) 15.4 (2.8) 10.3 (2.6)
Sex
Male 73 (46.8) 135 (57.0) 39 (59.1)
Abbreviations: BMI, body mass index
Female 83 (53.2) 102 (43.0) 27 (41.0) (calculated as weight in kilograms
Race and ethnicityb divided by height in meters squared);
Black or African American 112 (71.8) 0 38 (57.6) EVA-PR, Epigenetic Variation and
Childhood Asthma in Puerto Ricans
Puerto Rican 44 (28.2) 237 (100) 0 study; FeNO, fractional exhaled nitric
White 0 0 18 (27.3) oxide; FEV1, forced expiratory volume
Other 0 0 10 (15.2) in one second; FVC, forced vital
capacity; ICS, inhaled corticosteroids;
Annual household income ≥$15 000c 57/128 (44.5) 98/236 (41.5) 37/62 (59.7) IgE, immunoglobulin E; STAR, Stress
≥1 Parent completed some college 101 (64.7) 131 (55.3) 44 (66.7) and Treatment Response in Puerto
education Rican and African American Children
Private or employer-based health insurance 25/146 (17.1) 89/237 (37.6) 22/65 (33.8) with Asthma study; VDKA, Vitamin D
Season at nasal and serum sampling visit Kids Asthma study.
a
September-November 52/156 (33.3) 59/236 (25.0) 16/66 (24.2) Unless otherwise indicated.
b
In STAR, all participants had to have
December-February 35/156 (22.4) 54/236 (22.9) 19/66 (28.8)
at least 3 Puerto Rican grandparents
March-May 38/156 (24.4) 64/236 (27.1) 23/66 (34.8) (at the Puerto Rico site) or at least 3
June-August 31/156 (19.9) 59/236 (25.0) 8/66 (12.1) African American grandparents
(at the Pittsburgh site). All EVA-PR
Naso-ocular symptoms apart from colds, 100/144 (69.4) 203/237 (85.7) 45/63 (71.4)
past yeard participants had to have 4 Puerto
Rican grandparents. In VDKA,
Physician-diagnosed allergic rhinitise 39/142 (27.5) 40/232 (17.2) 17/45 (37.8)
parents selected 1 option from the
≥1 Asthma-related emergency 65/142 (45.8) 89/237 (37.6) 58/66 (87.9) following categories: American
department/urgent care visits Indian or Alaska Native, Asian, Black
in the prior year
or African American, Multiracial
≥1 Asthma-related hospitalizations 19/143 (13.3) 25/236 (10.6) 25/38 (65.8) (checked more than 1 category),
in the prior year
White, or Other. Small numbers
BMI z score, mean (SD)f 1.20 (1.17) 0.84 (1.12) 0.98 (1.17) were reported as the composite
Overweight or obesityg 102 (65.4) 108 (45.6) 37 (56.1) Other (which includes American
Use of nasal corticosteroids <4 wk 0 9 (3.8) 2 (3.0) Indian or Alaska Native, Asian,
prior to nasal specimen collection Multiracial, and Other) to protect
Use of ICS <4 wk prior to nasal specimen 0 62 (26.2) 66 (100) participant anonymity.
collection c
Annual household income $30 000
ICS response h
56/146 (38.4) or greater shown for VDKA and
Pittsburgh participants in STAR.
≥1 IgE to allergens, ≥0.35 IU/mL, No. 89/145 (61.4) 165/229 (72.1) 43/65 (66.2)
d
Defined as having had a runny or
Serum total IgE, median (IQR), IU/mL 244 (70-605) 272 (110-692) 392 (190-724)
stuffy nose accompanied by sneezing
No. 145 229 65 and itching without a cold or flu in the
Blood eosinophil count, median (IQR), 230 (120-389) 244 (132-387) 451 (198-606) last 12 months as a result of exposure
cells/mL to pollens (eg, grass, trees, weeds),
No. 153 236 65 house dust, cats, or dogs.
e
FeNO, median (IQR), parts per billion 23.5 (12-50) Physician-diagnosed hay fever or
allergic rhinitis in participants with
No. 154
naso-ocular symptoms.
FEV1, mean (SD), % predicted 91.3 (15.5) 96.6 (15.0) 98.1 (16.0) f
z scores were calculated using
No. 154 231 growth charts from the US Centers
FVC, mean (SD), % predicted 100.5 (13.1) 102.4 (14.0) 103.9 (13.9) for Disease Control and Prevention.
g
No. 154 231 A BMI z score in the 85th percentile
or above.
FEV1/FVC, mean (SD) 0.80 (0.09) 0.83 (0.08) 94.0 (10.2) h
ICS response was defined as an
No. 154 231 increment in FEV1 of 5% or more
T2HIGH cluster assignmenti 40 (25.6) 69 (29.1) 15 (22.7) from baseline after 6 weeks of ICS
treatment.
T17HIGH cluster assignmenti 57 (36.5) 83 (35.0) 31 (47.0)
i
Cluster assignment determined by
T2LOW/T17LOW cluster assignmenti 59 (37.8) 85 (35.9) 20 (30.3)
k-means clustering.

more common in STAR and EVA-PR participants with a T2LOW P = .08 for eosinophils in VDKA). The proportion of partici-
profile and in VDKA participants with a T2HIGH profile. pants with 1 or more positive (≥0.35 IU/mL) allergen-specific
In all studies, total IgE and eosinophils were higher in par- IgEs was high in those with the T2HIGH profile (91.9% in STAR,
ticipants with a T2HIGH profile than in those with T2LOW pro- 95.5% in EVA-PR, and 66.7% in VDKA). However, 50% to 73.3%
files (P < .001 in STAR and EVA-PR; P = .12 for total IgE and of the T2LOW profile participants in STAR, EVA-PR, and VDKA

jama.com (Reprinted) JAMA Published online January 2, 2025 E5

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Research Original Investigation Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes

Figure 2. Heat Map of Transcriptomic Clusters in the STAR, EVA-PR, and VDKA Studies Using K-Means and Consensus Clustering Profiles

K-means cluster Consensus cluster

A STAR (n = 156) D STAR (n = 129)

–2 0 2 –2 0 2
Normalized expression value of gene Normalized expression value of gene

T2HIGH/T17LOW T2LOW/T17HIGH T2LOW/T17LOW T2HIGH/T17LOW T2LOW/T17HIGH T2LOW/T17LOW

POSTN POSTN
T2 marker

T2 marker
SERPINB2 SERPINB2
CLCA1 CLCA1
CSF3 CSF3
T17 marker

CXCL1

T17 marker
CXCL1
CXCL2 CXCL2
CXCL3 CXCL3
IL-8 IL-8

B EVA-PR (n = 237) E EVA-PR (n = 177)

–2 0 2 –3 –2 –1 0 1 2 3
Normalized expression value of gene Normalized expression value of gene

T2HIGH/T17LOW T2LOW/T17HIGH T2LOW/T17LOW T2HIGH/T17LOW T2LOW/T17HIGH T2LOW/T17LOW

POSTN POSTN
T2 marker

T2 marker
SERPINB2 SERPINB2
CLCA1 CLCA1
CSF3 CSF3
T17 marker

CXCL1
T17 marker

CXCL1
CXCL2 CXCL2
CXCL3 CXCL3
IL-8 IL-8

C VDKA (n = 66) F VDKA (n = 52)

–4 –2 0 2 4 –2 0 2
Normalized expression value of gene Normalized expression value of gene

T2HIGH/T17LOW T2LOW/T17HIGH T2LOW/T17LOW T2HIGH/T17LOW T2LOW/T17HIGH T2LOW/T17LOW

POSTN POSTN
T2 marker

T2 marker

SERPINB2 SERPINB2
CLCA1 CLCA1
CSF3 CSF3
T17 marker

CXCL1
T17 marker

CXCL1
CXCL2 CXCL2
CXCL3 CXCL3
IL-8 IL-8

Study participants who had the same transcriptomic profile assignment EVA-PR indicates Epigenetic Variation and Childhood Asthma in Puerto Ricans
(eg, T2HIGH, T17HIGH, or T2LOW/T17LOW) across k-means clustering, k-medoid study; STAR, Stress and Treatment Response in Puerto Rican and African
clustering, hierarchical clustering, Gaussian mixture model, and spectral American Children with Asthma study; and VDKA, Vitamin D Kids Asthma study.
clustering approaches were deemed to have a consensus profile.

had 1 or more positive allergen-specific IgEs (eTable 4 in Supple- Supplement 1 shows the characteristics associated with ICS re-
ment 1). In STAR, FeNO was higher in the T2HIGH than in the sponsiveness across transcriptomic profiles in STAR.
T2LOW profiles (P < .001). Total IgE and eosinophils (in all stud- Next, we assessed how well common biomarkers identi-
ies) and FeNO (in STAR) were moderately correlated with the fied the T2HIGH profile in STAR. Optimal cutoffs for indi-
T2 3-gene mean (eFigure 5 in Supplement 1). There were no vidual biomarkers, defined using the Youden Index,24 were
significant differences in lung function measures across pro- 417.5 IU/mL for total IgE, 210.4 cells/μL for eosinophils, and
files in any study. Results were similar for participants with a 32.5 parts per billion (ppb) for FeNO, with similar results for
consensus profile (eTable 5 in Supplement 1). eTable 6 in participants with a consensus profile (Figure 3). Applying these

E6 JAMA Published online January 2, 2025 (Reprinted) jama.com

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes Original Investigation Research

Figure 3. Predictive Statistics for Fractional Exhaled Nitric Oxide (FeNO), Total Immunoglobulin E (IgE),
and Eosinophils in Determining the T2HIGH Profile in the STAR Cohort

A Predictive statistics for all individuals in the cohort (N = 156)

1.0
Eosinophils (AUC = 0.820)
FeNO (AUC = 0.800)
Total IgE (AUC = 0.849)
0.8

0.6
Sensitivity

0.4

0.2

0
0 0.2 0.4 0.6 0.8 1.0
1 – Specificity
Eosinophils FeNO Total IgE
T2LOW T2HIGH T2LOW T2HIGH T2LOW T2HIGH
Eosinophils <210.4 68 3 FeNO <32.5 85 8 Total IgE <417.5 93 9
Eosinophils ≥210.4 45 37 FeNO ≥32.5 31 30 Total IgE ≥417.5 15 28
Accuracy 0.686 0.747 0.835
Sensitivity 0.602 0.733 0.861
Specificity 0.925 0.790 0.757
PPV 0.958 0.914 0.912
NPV 0.451 0.492 0.651

B Predictive statistics among individuals with a consensus profile (N = 129)

1.0
Eosinophils (AUC = 0.856)
FeNO (AUC = 0.819)
Total IgE (AUC = 0.882)
0.8

0.6
Sensitivity

0.4

0.2

A, Cutoffs were 417.5 IU/mL for total

IgE, 210.4 cells/μL for eosinophils,
0
0 0.2 0.4 0.6 0.8 1.0 and 32.5 ppb for FeNO. B, Cutoffs
were 417.5 IU/mL for total IgE,
1 – Specificity
231.9 cells/μL for eosinophils, and
Eosinophils FeNO Total IgE 31.5 ppb for FeNO. All biomarker
T2LOW T2HIGH T2LOW T2HIGH T2LOW T2HIGH cutoffs were defined using the
Eosinophils <231.9 61 3 FeNO <31.5 69 7 Total IgE <417.5 76 7 Youden Index.
Eosinophils ≥231.9 29 33 FeNO ≥31.5 24 27 Total IgE ≥417.5 11 26 AUC indicates area under the curve;
Accuracy 0.746 0.756 0.850 NPV, negative predictive value; ppb,
Sensitivity 0.678 0.742 0.874 parts per billion; PPV, positive
Specificity 0.917 0.794 0.788 predictive value; and STAR, Stress
PPV 0.953 0.908 0.916 and Treatment Response in Puerto
NPV 0.532 0.529 0.703 Rican and African American Children
with Asthma study.

cutoff values (using predicted FeNO for EVA-PR and VDKA accuracies (correctly classified proportions) for the T2HIGH pro-
[eFigure 6 and eTable 7 in Supplement 1]) led to the following file in (1) STAR: 84% for total IgE, 69% for eosinophils, and 75%

jama.com (Reprinted) JAMA Published online January 2, 2025 E7

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Research Original Investigation Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes

for FeNO; (2) EVA-PR: 70% for IgE, 61% for eosinophils, and T2 inflammation. 6 This study accounted for seasonality,
77% for predicted FeNO; and (3) VDKA: 58% for total IgE, 48% participants did not currently smoke (eliminating the con-
for eosinophils, and 79% for predicted FeNO. Combining these cern that smoking may have attenuated similarities in gene
biomarkers, we also predicted the T2HIGH profile using a clas- expression between nasal and bronchial epithelium), and
sification rule based on a decision tree machine learning participants did not have a recent self-reported URI. Inclu-
method (eTable 8 and eFigure 7 in Supplement 1) and a ran- sion of youths with mostly mild to moderate persistent
dom forest model (eTable 8 in Supplement 1). asthma enhanced the generalizability of results. Unlike a
A differential expression meta-analysis was conducted for prior study of nasal epithelial gene expression in children
the T2HIGH and T17HIGH transcriptomic profiles among 358 par- with asthma that only included T2 gene signatures,16 this
ticipants with consensus profiles in STAR (n = 129), EVA-PR study measured both T2 and T17 gene signatures to compre-
(n = 177), and VDKA (n = 52), identifying 3516 differentially ex- hensively characterize transcriptomic profiles associated
pressed genes for the T2HIGH profile and 2494 differentially with asthma and identified a lower frequency of T2-high
expressed genes for the T17HIGH profile (Figure 4). Based on asthma in youth.
these results, a canonical pathway enrichment analysis was The finding that sensitization to 1 or more allergens
performed on upregulated genes, identifying enriched path- (atopy) is common among participants with a T2LOW profile is
ways for the T2HIGH and T17HIGH profiles at false discovery rate– supported by prior epidemiologic studies.8 In a worldwide
adjusted P < .05 (Figure 5A-B). The highest enriched path- study, atopy was present in only 2% to 40% of cases of
ways for the T2HIGH and T17HIGH profiles were related to cellular asthma in children aged 8 to 12 years29 in Mumbai, India, and
signaling and neutrophils or T17 signaling, respectively. Munich, Germany. Another study showed that high intensity
A consensus weighted gene coexpression network analy- and earlier development of atopy (more positive test results
sis identified 41 gene modules (clusters of highly intercon- to allergens at a younger age) were associated with a higher
nected genes) (eFigure 8A in Supplement 1). Subsequently, we risk of asthma than any or later atopy during childhood.30
correlated the eigengene (the principal component of the gene Moreover, a study of German children identified 4 asthma
module and a single representative expression profile) for each clusters based on atopy and blood eosinophilia (≥90th per-
module with the 3 transcriptomic profiles (eFigure 8B in centile or ≥470 cells/μL): (1) atopy only (41.3%), (2) eosino-
Supplement 1). In eFigure 8B in Supplement 1, the light yel- philia only (2.4%), (3) no atopy or eosinophilia (16.1%, labeled
low, white, and dark green modules were the top 3 modules T2-low), and (4) atopy and eosinophilia (40.2%).31 The mean
positively correlated with T2HIGH profiles, with correlations of FeNO for groups 1 (22.5 ppb) and 3 (13.5 ppb) were not signifi-
0.73, 0.53, and 0.34, respectively; for T17HIGH profiles, the green cantly different, but both were lower than that for groups 2
and dark red gene modules showed high positive correla- and 4 (39.5-42.3 ppb), suggesting that the atopy-only group
tions. We conducted gene ontology analyses on select T2 and included children with T2-low asthma.
T17 gene modules to identify enriched pathways in the respec- This study confirmed that atopy only or any single bio-
tive profiles. Glycoprotein-related pathways and IL-13 signal- marker has poor accuracy to define T2-high asthma and may
ing were upregulated in the T2HIGH profile, while pathways re- explain why some individuals considered to express a T2-
lated to responses to viruses, cytokines, and interferon-γ were high endotype based on a single biomarker (usually blood
upregulated in the T17HIGH profile (Figure 5C-D). eFigure 9 in eosinophils) do not respond to monoclonal antibodies.32-34 For
Supplement 1 shows an expanded signature gene panel de- this reason, the study used a decision tree learning method
rived from these analyses. based on a T2HIGH transcriptomic standard to develop a tool
that could be clinically applied to identify T2-high asthma with
high accuracy (77%-88%) to predict a T2HIGH profile, based on
total IgE of 418 IU/mL or above and FeNO 28 ppb or above.
Discussion Of note, this tool utilizes higher cutoffs for IgE than typically
T2-high has been considered the most common asthma en- used in clinical practice.32-34
dotype in school-aged youths aged 6 to 20 years because atopy In the absence of T17-high biomarkers, trials of T17 path-
frequently coexists with asthma in this group. However, this way blocking agents have been conducted in patients with any
study showed that T2-low asthma endotypes (including asthma endotype, yielding disappointing results.35,36 Nasal
T17HIGH and T2LOW/T17LOW transcriptomic profiles) were more transcriptomic profiling may enable endotype-targeted trials
prevalent than the T2HIGH profile in 3 studies of predomi- of much-needed therapies for T17-high asthma, particularly
nantly racially and ethnically minoritized youths with asthma. in racially and ethnically minoritized youths.37,38 Further, some
Prior studies of airway epithelial gene expression and of the identified genes, including FPR1 (a regulator of neutro-
asthma have consisted primarily of adults or patients taking phil recruitment), TREM1 (which amplifies responses against
inhaled and intranasal corticosteroids. In contrast, none of infection), and IL23A (which enhances expansion of T17 cells)
the participants in this study received systemic steroids for could be novel targets for developing therapies for youths with
at least 4 weeks prior to nasal sampling, and 96% or more of T17-high asthma. This study developed expanded T2 and T17
youths in this study were not using intranasal corticoste- transcriptomic signatures (eFigure 9 in Supplement 1) as a re-
roids, although in the VDKA cohort, all 66 participants were source for the identification of additional pathway-related
treated with ICS. Thus, study results were unlikely to be genes to inform the development of precision biomarker pro-
confounded by corticosteroid exposure, which suppresses files and therapeutics.

E8 JAMA Published online January 2, 2025 (Reprinted) jama.com

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Figure 4. Differential Expression Meta-Analysis Results

jama.com
A Volcano plot of DEGs with T2HIGH vs T2LOW/T17LOW in meta-analysis B Volcano plot of DEGs with T17HIGH vs T2LOW/T17LOW in meta-analysis
150 120
CDH26

CXCL2
100

NTRK2 POSTN CLCA1

CSF3

100 CISH 80
CST1
CXCL1
ALOX15

CCL26 CXCL8
FETUB
GCNT4
CD274 60
Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes

PRB2

–Log10 P
–Log10 P
PP7080 GATA2 CST2
ELOVL5 TNIP3
KCNJ16 ITLN1
SLC7A1 MS4A2 NCOA7
DSE ARHGAP40 TPSAB1
50 CACNB4 SLC9A3 40 CXCL6
TPSB2

Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025

LOC101927023 NEK6 NFE2 GCSAML BCL2A1
CDC42EP5 MEFV IL1B
SLC6A20
SERPINB2 TAL1
EDIL3 HS3ST4 TPSD1
ENTPD2 IL23A CR1 TREM1
FAM83D SLC5A5 LCN2
SCGB3A1 SYT7 G0S2 LILRA5
ERMN CST4 ITGAM
SSPN SAMD3 CEBPE 20 CFB GLT1D1
TMC1 SAA1 TNFAIP6
DDIT4L EDN2 SH3TC2 LILRA2
BPI SERPINF1 PROKR1 FAM74A3 SLC2A6
ACSS3 CCK C4BPA FCN1
CXCL12 GSTA1 MAMDC4 COL13A1 P2RY6
IDO2 LRRC55
AFAP1L1 BTNL9 FFAR3 KDR CCL2
SLC38A4 GMNC IL20 FDCSP
CCL18 EDAR
MMP3 KRT1 CES1P1 SPRR2B
DEFB4A LINC00707 C8B
FZD10
0 0

–3 –2 –1 0 1 2 3 4 5 6 7 8 9 10 –2 –1 0 1 2 3 4
Log2 fold change Log2 fold change

The magnitude, direction, and significance of changes in gene expression for the T2HIGH and T17HIGH profiles, expression levels between 2 groups, where fold change is the ratio of the expression level of a gene in 1 group
respectively. A, Includes DEGs among individuals in the consensus T2HIGH profile compared with individuals in the (eg, T2HIGH/T17LOW) to the expression level in another group (eg, T2LOW/T17LOW). Red = |log2 fold change|>1 and
consensus T2LOW/T17LOW profile in the meta-analysis of all 3 studies (STAR, n = 76; EVA-PR, n = 114; and VDKA, FDR-adjusted P value <0.01. Blue = FDR-adjusted P value <0.01 but fold change below threshold. Green = |log2
n = 27). B, Includes DEGs among individuals in the consensus T17HIGH profile compared with individuals in the fold change|>1 but FDR-adjusted P value below significance threshold.
consensus T2LOW/T17LOW profile in the meta-analysis of all 3 studies (STAR, n = 93; EVA-PR, n = 121; and VDKA, DEGs indicates differentially expressed genes; EVA-PR, Epigenetic Variation and Childhood Asthma in Puerto
n = 40). Volcano plots show –log10 (raw P values) vs log2 fold change. −Log10 P refers to the logarithm (base 10) of Ricans study; STAR, Stress and Treatment Response in Puerto Rican and African American Children with Asthma
the raw P value in the differential expression analysis. Genes with small P values, corresponding to strong study; and VDKA, Vitamin D Kids Asthma study.
differential effects, appear at the top. Log2 fold change refers to the logarithm (base 2) of the fold change in

(Reprinted) JAMA Published online January 2, 2025

Original Investigation Research

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
E9
 Research Original Investigation Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes

Figure 5. Pathway Enrichment Analysis Results

A Top enriched pathways of upregulated DEGs from T2HIGH B Top enriched pathways of upregulated DEGs from T17HIGH
vs T2LOW/T17LOW in meta-analysis vs T2LOW/T17LOW in meta-analysis

Molecular mechanisms of cancer Granulocyte adhesion and diapedesis

cAMP-response element binding Neutrophil degranulation
protein signaling in neurons IL-10 signaling
Integrin cell surface interactions Pathogen-induced cytokine storm
Class A/1 (rhodopsin-like receptors) signaling pathway
Phagosome formation IL-4 and IL-13 signaling
Breast cancer regulation by stathmin1 IL-17 signaling
Signal transducer and activator of Airway pathology in chronic obstructive
transcription 3 pathway pulmonary disease
Gustation pathway Agranulocyte adhesion and diapedesis

G protein–coupled receptor signaling TREM1 signaling

Differential regulation of cytokine production in
S100 family signaling pathway
intestinal epithelial cells by IL-17A and IL-17F
BBSome signaling pathway
Class A/1 (rhodopsin-like receptors)
Serotonin receptor signaling S100 family signaling pathway
O-linked glycosylation Liver X receptor/retinoid X receptor activation
Hepatic fibrosis/hepatic stellate Tumor microenvironment pathway
cell activation
Role of IL-17A in psoriasis
Granulocyte adhesion and diapedesis 6
6 IL-10 signaling
Cardiac hypertrophy signaling 4
5 Acute phase response signaling
z score

z score
(enhanced)
Macrophage classical activation 2
IL-13 signaling pathway 4
signaling pathway
Focal adhesion kinase signaling 3 0
Hepatic fibrosis/hepatic stellate cell activation
G alpha (q) signaling events –2
Cyclic GMP-AMP synthase-stimulator of
NA NA
Ion channel transport interferon genes signaling pathway

0 2 4 6 8 0 5 10 15 20 25
–log (FDR P value) –log (FDR P value)

C Enriched gene ontology pathways of T2 gene module identified D Enriched gene ontology pathways of T17 gene module identified
in weighted gene coexpression network analysis in weighted gene coexpression network analysis

Biological process Biological process

Glycoprotein metabolic process Response to virus
Glycoprotein biosynthetic process Defense response to virus
Protein glycosylation Defense response to symbiont
Macromolecule glycosylation Cytokine-mediated signaling pathway
Regulation of response to biotic stimulus
Glycosylation
Regulation of innate immune response
Regulation of supramolecular Regulation of viral process
fiber organization Response to interferon-γ
Regulation of protein polymerization Regulation of viral life cycle
Protein O-linked glycosylation Negative regulation of viral process
O-glycan processing Cellular component
Protein exit from endoplasmic Vacuolar membrane
reticulum Lysosomal membrane
Cellular component Lytic vacuole membrane
MHC protein complex
Golgi apparatus subcompartment
Coat protein complex II-coated ER
Golgi cisterna
to Golgi transport vesicle
Golgi stack Integral component of lumenal side
Golgi cisterna membrane Count
of endoplasmic reticulum membrane 10
Molecular function Lumenal side of endoplasmic reticulum membrane 20
Glycosyltransferase activity Lumenal side of membrane 30
Hexosyltransferase activity Count ER to Golgi transport vesicle membrane 40
Uridine diphosphate- 10 MHC class II protein complex 50
glycosyltransferase activity 20 Molecular function 60
Isomerase activity Cytokine receptor binding
Acetylgalactosaminyltransferase activity 1.0e-04 Cytokine activity
Intramolecular oxidoreductase activity Chemokine receptor binding 2.0e-06
FDR P value

FDR P value

7.5e-05 MHC protein complex binding

Nucleoside-diphosphatase activity 1.5e-06
Peptide antigen binding
Protein disulfide isomerase activity 5.0e-05 Double-stranded RNA binding 1.0e-06
Intramolecular oxidoreductase activity, MHC class II protein complex binding
transposing disulfide bonds 2.5e-05 Chemokine activity 5.0e-07
Polypeptide N- CXC chemokine receptor binding
acetylgalactosaminyltransferase activity MHC class II receptor activity
0 0.025 0.050 0.075 0.100 0.125 0 0.05 0.10 0.15 0.20
Gene ratio Gene ratio

Top 20 enriched (statistically overrepresented) canonical pathways of upregulated ontology pathway enrichment for T2-related (C) and T17-related (D) modules
genes from T2 (A) or T17 (B) models, ranked by −log10 (FDR-adjusted P value). Red identified in WGCNA, plotting enrichment significance (−log10) against ratio of
(positive enrichment scores), pathway activation; purple (negative enrichment genes involved in biological process, cellular component, or molecular function.
scores), pathway inhibition; and gray, no activity prediction can be made. Gene Dot size corresponds to number of genes enriched.

E10 JAMA Published online January 2, 2025 (Reprinted) jama.com

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes Original Investigation Research

Limitations
This study has several limitations. First, there is no criterion Conclusions
standard to define asthma endotypes.37,38 Second, assign-
ment to a transcriptomic profile was based on a single mea- In 3 studies that included predominantly Puerto Rican and
surement, but individuals may have had 1 or more asthma en- Black youths with asthma, a nasal transcriptomic profile con-
dotypes and endotypes may have changed over time. Third, sistent with the T2HIGH profile was least common across all
nasal epithelial transcriptomics, although well-correlated, may studies. The distribution of T2-high, T17-high, and T2-low/
not fully reflect bronchial epithelial transcriptomics. Fourth, T17-low asthma endotypes was similar across these studies.
this study included mostly Puerto Rican and Black youths, Most participants had T2 LOW profiles (T17 HIGH or T2 LOW /
so results may not be generalizable to youths of other races T17LOW) and sensitization to 1 or more allergens was common
or ethnicities. among these endotypes.

ARTICLE INFORMATION manuscript; and decision to submit the manuscript 12. Conrad LA, Cabana MD, Rastogi D. Defining
Accepted for Publication: October 7, 2024. for publication. pediatric asthma: phenotypes to endotypes and
Data Sharing Statement: See Supplement 2. beyond. Pediatr Res. 2021;90(1):45-51. doi:10.1038/
Published Online: January 2, 2025. s41390-020-01231-6
doi:10.1001/jama.2024.22684
REFERENCES 13. Kuruvilla ME, Lee FE, Lee GB. Understanding
Author Contributions: Drs Chen and Celedón had asthma phenotypes, endotypes, and mechanisms
full access to all of the data in the study and take 1. Porsbjerg C, Melén E, Lehtimäki L, Shaw D.
Asthma. Lancet. 2023;401(10379):858-873. doi:10. of disease. Clin Rev Allergy Immunol. 2019;56(2):
responsibility for the integrity of the data and the 219-233. doi:10.1007/s12016-018-8712-1
accuracy of the data analysis. Mr Yue and Dr Gaietto 1016/S0140-6736(22)02125-0
are co–first authors. Drs Chen and Celedón are 2. US Centers for Disease Control and Prevention. 14. Woodruff PG, Modrek B, Choy DF, et al.
co–senior authors. Data, statistics, and surveillance—asthma T-helper type 2-driven inflammation defines major
Concept and design: Yue, Gaietto, Canino, Forno, surveillance data. Reviewed July 24, 2024. subphenotypes of asthma. Am J Respir Crit Care Med.
Chen, Celedón. Accessed June 25, 2024. https://www.cdc.gov/ 2009;180(5):388-395. doi:10.1164/rccm.200903-
Acquisition, analysis, or interpretation of data: Yue, asthma/asthmadata.htm 0392OC
Gaietto, Han, Rosser, Xu, Qoyawayma, 3. Pate CA, Zahran HS, Qin X, Johnson C, 15. Fahy JV. Type 2 inflammation in
Acosta-Perez, Forno, Chen, Celedón. Hummelman E, Malilay J. Asthma asthma—present in most, absent in many. Nat Rev
Drafting of the manuscript: Yue, Gaietto, Rosser, surveillance—United States, 2006-2018. MMWR Immunol. 2015;15(1):57-65. doi:10.1038/nri3786
Acosta-Perez, Chen, Celedón. Surveill Summ. 2021;70(5):1-32. doi:10.15585/ 16. Poole A, Urbanek C, Eng C, et al. Dissecting
Critical review of the manuscript for important mmwr.ss7005a1 childhood asthma with nasal transcriptomics
intellectual content: Yue, Han, Rosser, Xu, distinguishes subphenotypes of disease. J Allergy
Qoyawayma, Canino, Forno, Chen, Celedón. 4. Woodruff PG, Boushey HA, Dolganov GM, et al.
Genome-wide profiling identifies epithelial cell Clin Immunol. 2014;133(3):670-8.e12. doi:10.1016/j.
Statistical analysis: Yue, Gaietto, Han, Rosser, Xu, jaci.2013.11.025
Chen, Celedón. genes associated with asthma and with treatment
Obtained funding: Celedón. response to corticosteroids. Proc Natl Acad Sci U S A. 17. Forno E, Wang T, Qi C, et al. DNA methylation in
Administrative, technical, or material support: Han, 2007;104(40):15858-15863. doi:10.1073/pnas. nasal epithelium, atopy, and atopic asthma in
Rosser, Qoyawayma, Acosta-Perez, Celedón. 0707413104 children: a genome-wide study. Lancet Respir Med.
Supervision: Canino, Forno, Chen, Celedón. 5. Choy DF, Hart KM, Borthwick LA, et al. TH2 and 2019;7(4):336-346. doi:10.1016/S2213-2600(18)
Other - oversaw direction of EVA-PR: Acosta-Perez. TH17 inflammatory pathways are reciprocally 30466-1

Conflict of Interest Disclosures: Dr Gaietto regulated in asthma. Sci Transl Med. 2015;7(301): 18. Forno E, Bacharier LB, Phipatanakul W, et al.
reported receiving a T32 training grant (HL129949) 301ra129. doi:10.1126/scitranslmed.aab3142 Effect of vitamin D3 supplementation on severe
from the National Institutes of Health (NIH). 6. Diver S, Sridhar S, Khalfaoui LC, et al. Feno asthma exacerbations in children with asthma and
Dr Rosser reported receiving grants from NIH (K08 differentiates epithelial gene expression clusters: low vitamin D levels: the VDKA randomized clinical
HL159333). Dr Forno reported receiving grants exploratory analysis from the MESOS randomized trial. JAMA. 2020;324(8):752-760. doi:10.1001/
from NIH (HL149693). Dr Celedón reported controlled trial. J Allergy Clin Immunol. 2022;150(4): jama.2020.12384
receiving nonfinancial support from Merck during 830-840. doi:10.1016/j.jaci.2022.04.024 19. Rubin J. Optimal classification into groups: an
the conduct of the Stress and Treatment Response 7. Brusselle GG, Koppelman GH. Biologic therapies approach for solving the taxonomy problem.
in Puerto Rican and African American Children with for severe asthma. N Engl J Med. 2022;386(2):157-171. J Theor Biol. 1967;15(1):103-144. doi:10.1016/0022-
Asthma (STAR) study and Pharmavite and GSK doi:10.1056/NEJMra2032506 5193(67)90046-X
during the conduct of the Vitamin D Kids Asthma 20. Charrad M, Ghazzali N, Boiteau V, Niknafs A.
(VKDA) study. No other disclosures were reported. 8. Mishra PE, Melén E, Koppelman GH, Celedón JC.
T2-low asthma in school-aged children: NbClust: an R package for determining the relevant
Funding/Support: STAR and the Epigenetic unacknowledged and understudied. Lancet Respir number of clusters in a data set. J Stat Softw.
Variation and Childhood Asthma in Puerto Ricans Med. 2023;11(12):1044-1045. doi:10.1016/S2213- 2014;61(6):1-36. doi:10.18637/jss.v061.i06
(EVA-PR) study were funded by grants HL079966, 2600(23)00369-7 21. Martin RJ, Szefler SJ, King TS, et al; National
HL117191, and HL150431 (Dr Celedón). Merck Heart, Lung, and Blood Institute’s Asthma Clinical
provided the Asmanex given to participants in 9. Hudey SN, Ledford DK, Cardet JC. Mechanisms
of non-type 2 asthma. Curr Opin Immunol. 2020; Research Center. The Predicting Response to
STAR. VDKA was funded by NIH grant HL119952 Inhaled Corticosteroid Efficacy (PRICE) trial.
and by the Pediatric Clinical and Translational 66:123-128. doi:10.1016/j.coi.2020.10.002
J Allergy Clin Immunol. 2007;119(1):73-80. doi:10.
Research Center at the UPMC Children’s Hospital of 10. Gans MD, Gavrilova T. Understanding the 1016/j.jaci.2006.10.035
Pittsburgh through NIH grant UL1TR001857. immunology of asthma: pathophysiology,
Pharmavite provided the vitamin D and placebo biomarkers, and treatments for asthma endotypes. 22. Sachs MC. plotROC: a tool for plotting ROC
capsules and GSK the Flovent given to VDKA Paediatr Respir Rev. 2020;36:118-127. doi:10.1016/ curves. J Stat Softw. 2017;79:79. doi:10.18637/jss.
participants; and Merck provided the Asmanex j.prrv.2019.08.002 v079.c02
given to STAR participants. 11. McKinley L, Alcorn JF, Peterson A, et al. TH17 23. Youden WJ. Index for rating diagnostic tests.
Role of the Funder/Sponsor: The NIH had no role cells mediate steroid-resistant airway inflammation Cancer. 1950;3(1):32-35. doi:10.1002/1097-0142
in the design and conduct of the study; collection, and airway hyperresponsiveness in mice. J Immunol. (1950)3:1<32::AID-CNCR2820030106>3.0.CO;2-3
management, analysis, and interpretation of 2008;181(6):4089-4097. doi:10.4049/jimmunol.181. 24. Hothorn T, Hornik K, Zeileis A. Unbiased
the data; preparation, review, or approval of the 6.4089 recursive partitioning: a conditional inference

jama.com (Reprinted) JAMA Published online January 2, 2025 E11

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025
 Research Original Investigation Transcriptomic Profiles in Nasal Epithelium and Asthma Endotypes

framework. J Comput Graph Stat. 2006;15(3):651- 2007;176(6):565-574. doi:10.1164/rccm.200607- 34. Pearce N, Pekkanen J, Beasley R. How much
674. doi:10.1198/106186006X133933 994OC asthma is really attributable to atopy? Thorax. 1999;
25. Love MI, Huber W, Anders S. Moderated 30. Filiou A, Holmdahl I, Asarnoj A, et al. 54(3):268-272. doi:10.1136/thx.54.3.268
estimation of fold change and dispersion for Development of sensitization to multiple allergen 35. Busse WW, Holgate S, Kerwin E, et al.
RNA-seq data with DESeq2. Genome Biol. 2014;15 molecules from preschool to school age is related to Randomized, double-blind, placebo-controlled
(12):550. doi:10.1186/s13059-014-0550-8 asthma. Int Arch Allergy Immunol. 2022;183(6): study of brodalumab, a human anti-IL-17 receptor
26. Krämer A, Green J, Pollard J Jr, Tugendreich S. 628-639. doi:10.1159/000521324 monoclonal antibody, in moderate to severe
Causal analysis approaches in Ingenuity Pathway 31. Maison N, Omony J, Illi S, et al; ALLIANCE Study asthma. Am J Respir Crit Care Med. 2013;188(11):
Analysis. Bioinformatics. 2014;30(4):523-530. doi: Group. T2-high asthma phenotypes across lifespan. 1294-1302. doi:10.1164/rccm.201212-2318OC
10.1093/bioinformatics/btt703 Eur Respir J. 2022;60(3):2102288. doi:10.1183/ 36. Brightling CE, Nair P, Cousins DJ, Louis R,
27. Langfelder P, Horvath S. WGCNA: an R package 13993003.02288-2021 Singh D. Risankizumab in severe asthma—a phase
for weighted correlation network analysis. BMC 32. Fitzpatrick AM, Gaston BM, Erzurum SC, 2a, placebo-controlled trial. N Engl J Med. 2021;385
Bioinformatics. 2008;9(1):559. doi:10.1186/1471-2105- Teague WG; National Institutes of Health/National (18):1669-1679. doi:10.1056/NEJMoa2030880
9-559 Heart, Lung, and Blood Institute Severe Asthma 37. Fitzpatrick AM, Chipps BE, Holguin F,
28. Ashburner M, Ball CA, Blake JA, et al; The Gene Research Program. Features of severe asthma in Woodruff PG. T2-“low” asthma: overview and
Ontology Consortium. Gene ontology: tool for the school-age children: atopy and increased exhaled management strategies. J Allergy Clin Immunol Pract.
unification of biology. Nat Genet. 2000;25(1):25-29. nitric oxide. J Allergy Clin Immunol. 2006;118(6): 2020;8(2):452-463. doi:10.1016/j.jaip.2019.11.006
doi:10.1038/75556 1218-1225. doi:10.1016/j.jaci.2006.08.019 38. Guida G, Bagnasco D, Carriero V, et al. Critical
29. Weinmayr G, Weiland SK, Björkstén B, et al; 33. Herr M, Just J, Nikasinovic L, et al. Risk factors evaluation of asthma biomarkers in clinical practice.
ISAAC Phase Two Study Group. Atopic sensitization and characteristics of respiratory and allergic Front Med (Lausanne). 2022;9:969243. doi:10.
and the international variation of asthma symptom phenotypes in early childhood. J Allergy Clin Immunol. 3389/fmed.2022.969243
prevalence in children. Am J Respir Crit Care Med. 2012;130(2):389-96.e4. doi:10.1016/j.jaci.2012.05.
054

E12 JAMA Published online January 2, 2025 (Reprinted) jama.com

© 2025 American Medical Association. All rights reserved, including those for text and data mining, AI training, and similar technologies.
Downloaded from jamanetwork.com by Universidad de Guadalajara user on 01/08/2025