Rationale: Children with preschool wheezing or school-age asthma are reported to have airway microbial imbalances.
Objectives: To identify clusters in children with asthma or wheezing using oropharyngeal microbiota profiles.
Methods: Oropharyngeal swabs from the U-BIOPRED (Unbiased Biomarkers for the Prediction of Respiratory Disease Outcomes) pediatric asthma or wheezing cohort were characterized using 16S ribosomal RNA gene sequencing, and unsupervised hierarchical clustering was performed on the Bray-Curtis β-diversity. Enrichment scores of the Molecular Signatures Database hallmark gene sets were computed from the blood transcriptome using gene set variation analysis. Children with severe asthma or severe wheezing were followed up for 12–18 months, with assessment of the frequency of exacerbations.
Measurements and Main Results: Oropharyngeal samples from 241 children (age range, 1–17 years; 40% female) revealed four taxa-driven clusters dominated by Streptococcus, Veillonella, Rothia, and Haemophilus. The clusters showed significant differences in atopic dermatitis, grass pollen sensitization, FEV1% predicted after salbutamol, and annual asthma exacerbation frequency during follow-up. The Veillonella cluster was the most allergic and included the highest percentage of children with two or more exacerbations per year during follow-up. The oropharyngeal clusters were different in the enrichment scores of TGF-β (transforming growth factor-β) (highest in the Veillonella cluster) and Wnt/β-catenin signaling (highest in the Haemophilus cluster) transcriptomic pathways in blood (all q values <0.05).
Conclusions: Analysis of the oropharyngeal microbiota of children with asthma or wheezing identified four clusters with distinct clinical characteristics (phenotypes) that associate with risk for exacerbation and transcriptomic pathways involved in airway remodeling. This suggests that further exploration of the oropharyngeal microbiota may lead to novel pathophysiologic insights and potentially new treatment approaches.
Children with asthma or wheezing were reported to have airway microbial imbalances. However, it is not yet clear whether characterizing the microbiota in oropharyngeal swabs, as a noninvasive sampling compartment, could help identify clinically relevant phenotypes in children with asthma or wheezing.
In this study, we show that the oropharyngeal microbiota could identify four distinct clusters that were different in allergic status and spirometry and that exhibited differential expression of TGF-β (transforming growth factor-β) and Wnt/β-catenin signaling pathways in blood. These clusters were independent predictors of future exacerbation events in children with severe disease within 12–18 months of follow-up.
Asthma is the most common chronic disease affecting children. Childhood asthma is a multifaceted disease that presents with varied clinical manifestations and treatment responses. In addition, childhood asthma–like symptoms were reported to increase risk of chronic obstructive pulmonary disease in adulthood (1). The underlying mechanisms of childhood asthma or preschool wheezing are not completely understood and are believed to originate from multiple interacting factors, including genetic susceptibility, environmental exposures, and microbiota (2). The host-residing microbiota regulates and works in alliance with the immune system and is believed to partly mediate the link between environmental exposures and asthma pathophysiology (3, 4). Therefore, assessing microbiota profiles in children with wheezing or asthma may reveal distinct host, microbiome, and disease links that could play a role in disease development.
Omics-guided asthma characterization approaches have been used recently to delineate groups (i.e., clusters) of patients with asthma who share similar molecular profiles and exhibit distinct phenotypic asthma characteristics (5). Underpinning these asthma phenotypes with specific molecular fingerprints may lead to tailoring diagnostic and/or therapeutic decisions. This may result in better patient care, reduce the risk of severe complications, and prevent inappropriate medication use. In adults, the U-BIOPRED (Unbiased Biomarkers for the Prediction of Respiratory Disease Outcomes) study revealed sputum taxa-driven clusters linked to asthma severity and underlying inflammatory status, particularly sputum neutrophilia (6). Similar findings in adults with asthma have been reported elsewhere (7), suggesting that the sputum microbiota may help refine the neutrophilic asthma phenotype in adult patients with severe asthma. Combining other omics, including transcriptomics, delineates associated molecular mechanisms characterizing the taxa-driven clusters and may improve our understanding of the pathophysiology (8). In children, obtaining samples representing the lower airways (e.g., induced sputum, bronchoalveolar lavage fluid, endobronchial biopsy samples) is challenging. Evidence suggests the ecological continuity of the microbiota structure within the respiratory tract (9–11), highlighting the need to investigate the value of noninvasive microbiota sampling in the assessment of children with wheeze or asthma. Studies have shown that changes in pharyngeal microbiota are associated with the risk of later developing asthma in young children (12–15) or associated with different asthma characteristics (16, 17). However, attempts to assess the value of airway microbiota in stratifying children with wheezing or asthma into clinically meaningful groups are scarce.
To clarify the role of the oropharyngeal microbiota in pediatric preschool-age wheeze and school-age asthma, we examined oropharyngeal swabs using 16S ribosomal RNA gene sequencing in the European U-BIOPRED pediatric cohort (18). In a previous report from this data set, we found limited differences in diversity and differential abundances between children with severe and mild to moderate asthma or wheeze but a significant association with age (19). Building on these results, we hypothesized that oropharyngeal microbiota profiles may reveal distinct clusters in children with preschool-age wheezing or asthma during school age. The specific aims of this study are to 1) perform unsupervised unbiased clustering of oropharyngeal microbiota profiles of children and assess whether these clusters are linked with district asthma characteristics (phenotypes); 2) elucidate the possible underlying biological pathways of the revealed phenotypic clusters using the peripheral blood transcriptome; and 3) reveal whether the baseline clusters are associated with future exacerbations during 12–18 months of follow-up in children with severe school-age asthma and severe preschool-age wheezing.
The U-BIOPRED study is a multicenter European observational cohort study including children with physician-diagnosed mild to moderate and severe school-age asthma and preschool wheezing aged 1–17 years, as previously described (18). Children were recruited from seven study centers in five European countries (the United Kingdom, Denmark, Sweden, the Netherlands, and Switzerland) and were mainly of White ethnic origin. Children were allocated into four cohorts comprising severe school-age asthma (cohort A), mild to moderate school-age asthma (cohort B), severe preschool-age wheeze (cohort C), and mild to moderate preschool-age wheeze (cohort D). All study centers obtained ethics committee approval, and parents/caregivers gave written informed consent. Children gave assent where age appropriate. The study was registered at ClinicalTrials.gov (NCT 01982162). All study centers followed uniform standard operating procedures in all study procedures as described (18).
A total of 131 preschool-age children with wheeze (n = 77 severe, n = 54 mild to moderate) and 140 school-age children with asthma (n = 97 severe, n = 43 mild to moderate) were recruited to the U-BIOPRED pediatric cohorts. The flow diagram of patient inclusion is shown in Figure E1 in the online supplement. Asthma or wheeze severity was defined according to the Global Initiative for Asthma and the Innovative Medicines Initiative guidelines (20, 21) and has been fully described elsewhere (18). Body mass index was converted into z-scores according to the World Health Organization growth charts (22). Participants were assessed for atopy (23, 24) (see the online supplement for details) and environmental tobacco smoke exposure by measuring urinary cotinine concentrations (18). In addition, they underwent spirometry and fractional exhaled nitric oxide testing (for older capable children) as described (18). Quality of life was evaluated using the standard Pediatric Asthma (Caregiver’s) Quality of Life Questionnaire (25, 26), and asthma control was evaluated using the Childhood Asthma Control Test (27, 28). A Childhood Asthma Control Test score of >19 indicates well-controlled asthma. Other family and medical histories were collected from children or caregivers and physicians as appropriate.
Children with severe asthma or severe wheezing (n = 125 of 151 [82.8%]) were followed up prospectively in time (12–18 mo after the baseline visit), during which the frequency of asthma exacerbation (as defined [29]) per year was documented (i.e., future exacerbations).
Full details regarding sample collection and omics analysis were described previously (19) and in the online supplement.
Briefly, oropharyngeal (throat) swabs for microbiota processing were taken from the posterior pharyngeal wall, behind the uvula. The 16S ribosomal RNA gene in the collected samples was amplified for the hypervariable V4 region, and sequencing was performed using the Illumina MiSeq desktop sequencer. The microbiome sequencing data can be found in the European Nucleotide Archive database (accession number PRJEB47973). Paired-end reads (250 bp × 2) from amplicon sequencing were quality checked using FastQC (30) and MultiQC (31) and then cleaned by removing V4 region primers using Cutadapt (32). Further processing was performed using the DADA2 pipeline as described (33). Briefly, the pipeline works by performing quality filtering and trimming, dereplicating sequences, learning data set–specific error rates, denoising by removing potentially containing errors sequences, merging paired-end reads while removing mismatches to reduce errors, constructing amplicon sequence variants, and removing chimera by implementing the “bimera” method. Finally, taxonomic classification of amplicon sequence variants was done using the SILVA database version 132 (34).
Peripheral blood was isolated and gene expression was assessed using Affymetrix HT HG-U133+ PM arrays. The microarray data are deposited in the Gene Expression Omnibus database (accession number GSE123750). Raw microarray data were quality assessed and preprocessed using robust multiarray average normalization method (35) as implemented in the affy (https://bioconductor.org/packages/release/bioc/html/affy.html) package for R (https://www.r-project.org/). Transcriptomic pathway enrichment analysis was performed using an unsupervised approach using the GSVA (https://www.bioconductor.org/packages/release/bioc/html/GSVA.html) package for R against gene sets from the Molecular Signatures Database version 7.4 (36).
The general data analysis workflow is shown in Figure E2 and was further described previously (6) and in detail in the online supplement.
Briefly, a clustering benchmarking strategy using hierarchical Ward2, partition around the medoids (PAM), and topological data analysis (TDA) was implemented to subtype children with asthma or wheezing according to their microbiota profiles. The clusters were checked for differences in microbial profiles using Shannon α-diversity (using a Kruskal-Wallis H test followed by Dunn’s post hoc test) and by linear discriminant analysis effect size (LEfSe) for differentially abundant bacteria at the phylum and species levels (using a linear discriminant analysis score >2 and a false discovery rate [FDR] q value <0.01). Furthermore, the taxa-driven clusters were tested for differences in certain demographic and clinical characteristics (such as allergic status, atopy, and spirometry) and the frequency of future exacerbations per year using Kruskal-Wallis H or Pearson chi-square tests as appropriate. Clinical characteristics that were statistically significant after univariate analysis were adjusted for several potential confounders using logistic or linear multiple regression models as appropriate. Potential confounders (covariates) to adjust for in the regression models were determined using a directed acyclic graph (37) (see Figure E3). A simple mediation analysis was conducted to investigate whether microbiota cluster 2 was an independent predictor of future asthma exacerbations (two or more) or if this was mediated by atopic dermatitis status.
The gene set variation analysis enrichment scores (ESs) of the tested gene sets (i.e., transcriptomic pathways) were compared between the clusters using the Kruskal-Wallis H test, and multiple testing was corrected by the FDR, followed by permutation analysis (10,000 times). Linear regression models were used to investigate whether the ESs of the significant gene sets were dependent on the participants’ cluster assignments after adjusting for multiple covariates (as defined in the directed acyclic graph).
All analyses were performed using RStudio (https://posit.co/download/rstudio-desktop/) (version 1.1.453) with R software (version 3.6.1).
A total of 241 participants provided baseline oropharyngeal swabs that were sequenced to a read depth of at least 1,000 and passed quality control. The baseline characteristics of the subjects are shown in Table 1. The median age of the included participants was 6 years, 40% were female, 80% were White, and 70% were atopic.
Characteristic | All Children with Microbiota Profiles (n = 241) |
---|---|
Age, yr, median (IQR) | 6.0 (4.0–13.0) |
Female, n/N (%) | 96/241 (40) |
Ethnicity, n/N (%) | |
White | 192/241 (80) |
Black African | 11/241 (5) |
Central Asian | 6/241 (2) |
East Asian | 2/241 (1) |
South Asian | 2/241 (1) |
Arabic north heritage | 6/241 (2) |
Multiple races | 18/241 (7) |
Other | 4/241 (2) |
Body mass index z-score, median (IQR) | 0.61 (0.01–1.42) (n = 240) |
Asthma or wheezing cohort, n/N (%) | |
Cohort A (severe school-age asthma) | 86/241 (36) |
Cohort B (mild to moderate school-age asthma) | 39/241 (16) |
Cohort C (severe preschool wheezing) | 65/241 (27) |
Cohort D (mild to moderate preschool wheezing) | 51/241 (21) |
Overall atopic sensitization*, n (%) | 159/226 (70) |
Asthma control (well controlled)†, n (%) | 72/181 (40) |
Average quality-of-life score, median (IQR) | 5.30 (3.70–6.47) (n = 238) |
Lung function test, median (IQR) | |
FEV1% predicted before salbutamol | 92.9 (81.6–106.7) (n = 135) |
FEV1% predicted after salbutamol | 103.5 (92.6–113.0) (n = 143) |
FeNO, ppb | 30.0 (15.6–57.0) (n = 117) |
Asthma medications, n/N (%) | |
ICS | 212/241 (88) |
SABA | 227/241 (94) |
LABA | 137/241 (57) |
OCS (maintenance) | 40/241 (17) |
SAMA | 15/241 (6) |
LAMA | 3/241 (1) |
LTRA | 143/241 (59) |
Nasal steroids | 18/241 (7) |
Antibiotic intake, n/N (%) | |
Antibiotics (recent) | 30/241 (12) |
Antibiotics (recent and previous) | 36/241 (15) |
The median number of sequencing reads passing quality control was 11,975 (interquartile range, 7,613–21,098) per sample. Negative extraction controls showed different microbial composition and significantly lower Shannon α-diversity compared with oropharyngeal swabs, while the correct classification of all bacterial genera was achieved in the mock community (see Figure E4). The main bacterial genera detected in the oropharyngeal swab samples were Streptococcus, Veillonella, Haemophilus, Prevotella, and Rothia (see Figure E5).
Clustering performed on Bray-Curtis β-diversity suggested four optimum clusters, as evaluated by the majority vote of multiple indices (see Figure E6), resulting in four groups of children by hierarchical Ward2 (Figure 1A) and PAM (Figure 1B) clustering. Quantitative assessment of similarity in participants’ assignment between hierarchical Ward2 and PAM clustering was performed by means of the Rand index (0.85) and the Pearson chi-square test (χ2 = 498.1; P < 1 × 10−4), suggesting high similarity in the two clustering assignments. Visual representation by TDA shows that the four microbiota clusters are enriched (according to spatial analysis of functional enrichment scores) into four distinct regions of the TDA network (see Figures E7A and E7B). The four clusters were also visualized by the principal coordinates analysis plot on the Bray-Curtis distance with depiction of the 95% ellipses (see Figure E8).
![figure](/na101/home/literatum/publisher/thoracic/journals/content/ajrccm/2023/ajrccm.2023.208.issue-2/rccm.202211-2107oc/20230703/images/medium/rccm.202211-2107ocf1.gif)
Figure 1. (A) Hierarchical agglomerative cluster dendrogram of the Bray-Curtis dissimilarities showing the four main groups of children. (B) Principal-component plot of the clustering by partition around the medoids (PAM) shows four relatively detached 95% reference interval ellipses of four groups of children. The similarity in participants’ assignments between the two clustering algorithms was assessed using the Rand index (0.85) and the Pearson chi-square test (χ2 = 498.1; P < 1 × 10−4) showing high similarity in the clustering assignments. For assessing cluster stability, replacement of points by noise schemes (1,000 iterations) resulted in mean Jaccard similarity indices ranging from 0.74 to 1 for the four clusters by either hierarchical clustering or PAM, suggesting relatively stable clusters. Dim = dimension.
[More] [Minimize]The hierarchical clusters had significantly different microbial Shannon α-diversity (see Figure E9), showing a decreasing order of median diversity as follows: cluster 4 > cluster 2 > cluster 3 = cluster 1. The mean relative abundances of the most abundant bacterial genera and the individual samples’ compositions were stratified according to the clusters’ assignments (see Figures E10 and E11, respectively). Analysis using the LEfSe method revealed the differential abundance of the most abundant bacterial phyla and species among the clusters (Figure 2). Cluster 1 showed main enrichment of bacterial genus Streptococcus and phylum Firmicutes, cluster 2 of genus Veillonella and phylum Patescibacteria, cluster 3 of genus Rothia and phylum Actinobacteria, and cluster 4 of genera Haemophilus and Neisseria, as well as phylum Proteobacteria. The TDA was consistent with the LEfSe method, in which the same regions of the TDA network enriched with the cluster’s assignment showed increased relative abundances of the aforementioned bacterial genera (see Figures E12 and E13). Of note, Moraxella was detected in a small percentage of samples and in low abundances (5.0% of samples, 0.01% mean relative abundance). It was relatively higher in abundance in cluster 1 compared with the other clusters, although the results were not statistically significant after FDR correction. Decision tree analysis showed that the relative abundances of only three bacterial genera (Streptococcus, Veillonella, and Haemophilus) (see Figure E14) could be used to classify the children into the corresponding clusters with an accuracy of 0.86 (95% confidence interval [CI], 0.75–0.94; P = 8.24 × 10−15) using a holdout validation set.
![figure](/na101/home/literatum/publisher/thoracic/journals/content/ajrccm/2023/ajrccm.2023.208.issue-2/rccm.202211-2107oc/20230703/images/medium/rccm.202211-2107ocf2.gif)
Figure 2. Differentially abundant bacterial taxa among the taxa-driven clusters using the linear discriminant analysis (LDA) effect size (LEfSe) statistical method. (A and B) The highest enriched bacterial taxa related to each cluster are shown at the phylum (A) and species (B) levels. Only differentially abundant bacterial species with false discovery rate q values <0.01 and absolute LDA scores >2 are depicted.
[More] [Minimize]There were no statistically significant associations between the participants’ cluster assignments and the seven study centers from which samples were collected or the season of sample collection (P = 0.390 and P = 0.526, respectively).
Table 2 shows the demographic and clinical characteristics of the four taxa-driven clusters. The clusters showed statistically significant differences in atopic dermatitis diagnosis and FEV1% predicted values after salbutamol (the latter are also shown in Figure E15). These differences remained significant after adjusting for several covariates (see Tables E2 and E3; q < 0.05 for all). Table 3 shows characteristics of the atopic sensitization of the clusters. The clusters showed statistically significant differences in (aeroallergen) atopic sensitization specifically to grass pollens and the number of aeroallergen sensitizations (the latter is also shown in Figure E16). After adjusting for covariates, grass aeroallergen sensitization (but not overall sensitization) was statistically significantly associated with the clusters (see Table E4).
Characteristic | Cluster 1 (n = 50) | Cluster 2 (n = 88) | Cluster 3 (n = 21) | Cluster 4 (n = 82) | P Value |
---|---|---|---|---|---|
Demographics | |||||
Age, yr, median (IQR) | 5.0 (3.3 to 13.0) | 9.5 (4.8 to 13.0) | 5.0 (3.0 to 12.0) | 5.0 (3.3 to 12.0) | 0.144 |
Female, n/N (%) | 21/50 (42) | 38/88 (43) | 5/21 (24) | 32/82 (39) | 0.430 |
Ethnicity, n/N (%) | 0.427 | ||||
White | 44/50 (88) | 64/88 (73) | 19/21 (90) | 65/82 (79) | |
Black African | 0/50 (0) | 4/88 (5) | 1/21 (5) | 6/82 (7) | |
Central Asian | 2/50 (4) | 3/88 (3) | 0/21 (0) | 1/82 (1) | |
East Asian | 1/50 (2) | 0/88 (0) | 0/21 (0) | 1/82 (1) | |
South Asian | 0/50 (0) | 0/88 (0) | 0/21 (0) | 2/82 (2) | |
Arabic north heritage | 1/50 (2) | 4/88 (5) | 0/21 (0) | 1/82 (1) | |
Multiple races | 2/50 (4) | 11/88 (12) | 1/21 (5) | 4/82 (5) | |
Other | 0/50 (0) | 2/88 (2) | 0/21 (0) | 2/82 (2) | |
Body mass index z-score, median (IQR) | 0.59 (−0.12 to 1.01) (n = 49) | 0.77 (0.05 to 1.77) (n = 88) | 0.27 (0.10 to 1.08) (n = 21) | 0.61 (0.02 to 1.15) (n = 82) | 0.544 |
Delivery mode (cesarean section), n/N (%) | 12/50 (24) | 25/88 (28) | 3/21 (14) | 18/82 (22) | 0.335 |
Smoking history, n/N (%) | |||||
Maternal smoking during pregnancy | 5/50 (10) | 13/86 (15) | 2/20 (10) | 6/82 (7) | 0.453 |
Secondhand smoking | 12/50 (24) | 18/85 (21) | 4/21 (19) | 11/79 (14) | 0.494 |
Cotinine present (in urine) | 4/43 (9) | 10/74 (14) | 2/20 (10) | 8/74 (11) | 0.911 |
Clinical characteristics | |||||
Asthma or wheezing cohort, n/N (%) | 0.057 | ||||
Cohort A (severe school-age asthma) | 13/50 (26) | 39/88 (44) | 5/21 (24) | 29/82 (35) | |
Cohort B (mild to moderate school-age asthma) | 10/50 (20) | 16/88 (18) | 4/21 (19) | 9/82 (11) | |
Cohort C (severe preschool wheezing) | 15/50 (30) | 20/88 (23) | 10/21 (48) | 20/82 (24) | |
Cohort D (mild to moderate preschool wheezing) | 12/50 (24) | 13/88 (15) | 2/21 (10) | 24/82 (29) | |
History of diagnosed allergic disorders (ever), n/N (%) | |||||
Atopic dermatitis (ever) | 24/50 (48) | 68/87 (78) | 13/21 (62) | 55/81 (68) | 0.003 |
Allergic rhinitis (ever) | 23/46 (50) | 42/82 (51) | 12/21 (57) | 34/78 (44) | 0.633 |
Asthma control (well controlled)*, n/N (%) | 18/38 (47) | 24/72 (33) | 8/13 (62) | 22/58 (38) | 0.188 |
Average quality-of-life score, median (IQR) | 5.4 (4.1 to 6.7) (n = 50) | 5.3 (3.3 to 6.4) (n = 86) | 5.3 (4.6 to 6.4) (n = 21) | 5.2 (3.7 to 6.5) (n = 81) | 0.853 |
Lung function test, median (IQR) | |||||
FEV1% predicted before salbutamol | 97.5 (86.6 to 108.1) (n = 28) | 88.9 (74.3 to 99.31) (n = 55) | 96.3 (87.7 to 108.8) (n = 11) | 95.5 (87.0 to 106.9) (n = 41) | 0.108 |
FEV1% predicted after salbutamol | 108.7 (95.9 to 117.5) (n = 29) | 98.3 (88.8 to 107.7) (n = 59) | 107.2 (104.2 to 118.3) (n = 10) | 102.9 (94.1 to 109.9) (n = 45) | 0.047 |
FeNO, ppb | 43.0 (13.5 to 58.3) (n = 20) | 29.0 (16.0 to 50.0) (n = 53) | 35.0 (14.0 to 97.0) (n = 9) | 27.0 (14.5 to 58.0) (n = 35) | 0.954 |
Asthma/other medications, n/N (%) | |||||
ICS | 43/50 (86) | 80/88 (91) | 20/21 (95) | 69/82 (84) | 0.390 |
SABA | 48/50 (96) | 85/88 (97) | 19/21 (90) | 75/82 (91) | 0.422 |
LABA | 25/50 (50) | 56/88 (64) | 11/21 (52) | 45/82 (55) | 0.406 |
OCS (maintenance) | 6/50 (12) | 14/88 (16) | 5/21 (24) | 15/82 (18) | 0.623 |
SAMA | 4/50 (8) | 6/88 (7) | 0/21 (0) | 5/82 (6) | 0.670 |
LAMA | 1/50 (2) | 0/88 (0) | 0/21 (0) | 2/82 (2) | 0.503 |
LTRA | 24/50 (48) | 58/88 (66) | 13/21 (62) | 48/82 (59) | 0.235 |
Nasal steroids | 3/50 (6) | 9/88 (10) | 0/21 (0) | 6/82 (7) | 0.415 |
Antibiotic intake, n/N (%) | |||||
Antibiotics (recent) | 7/50 (14) | 9/88 (10) | 2/21 (10) | 12/82 (15) | 0.803 |
Antibiotics (recent and previous) | 7/50 (14) | 10/88 (11) | 4/21 (19) | 15/82 (18) | 0.600 |
Characteristic | Cluster 1 (n = 50) | Cluster 2 (n = 88) | Cluster 3 (n = 21) | Cluster 4 (n = 82) | P Value |
---|---|---|---|---|---|
Overall atopic sensitization*, n/N (%) | 25/47 (53) | 65/83 (78) | 13/18 (72) | 56/78 (72) | 0.022 |
Common aeroallergen sensitization, n/N (%) | 25/47 (53) | 62/83 (75) | 13/18 (72) | 46/78 (59) | 0.045 |
Common aeroallergens, n (%) | |||||
House dust mites | 19/46 (41) | 46/81 (57) | 11/17 (65) | 31/77 (40) | 0.068 |
Tree | 17/46 (37) | 33/79 (42) | 6/16 (38) | 21/70 (30) | 0.522 |
Mold | 6/43 (14) | 27/78 (35) | 4/16 (25) | 19/71 (27) | 0.111 |
Grass | 13/46 (28) | 45/81 (56) | 8/18 (44) | 29/74 (39) | 0.021 |
Pet animal (cat/dog) | 22/46 (48) | 52/81 (64) | 7/18 (39) | 36/75 (48) | 0.079 |
Number of aeroallergen classes, median (IQR) | 0.0 (0.0–3.0) (n = 47) | 3.0 (1.0–4.00) (n = 83) | 1.5 (0.0–3.8) (n = 18) | 1.0 (0.0–3.0) (n = 78) | 0.021 |
Total IgE, kU/L, median (IQR) | 292.6 (54.7–924.7) (n = 38) | 245.1 (95.8–1,002.5) (n = 76) | 105.0 (52.8–310.0) (n = 17) | 150.0 (38.9–682.3) (n = 68) | 0.240 |
A total of 188 of 241 (78%) participants provided peripheral blood samples for transcriptomic analysis. Figure 3 shows the statistically significant transcriptomic pathways among the taxa-driven clusters. A total of three hallmark pathways were statistically different among the taxa-driven clusters, of which the Wnt/β-catenin and the TGF-β (transforming growth factor-β) signaling transcriptomic pathways remained significant after multiple testing corrections. Cluster 4 had the highest median ESs for the Wnt/β-catenin signaling pathway, while cluster 2 had the highest median ESs for the TGF-β signaling pathway. The permutation analysis P values were 9.9 × 10−5 and 0.001 for the Wnt/β-catenin and TGF-β signaling pathways, respectively. These pathways showed statistically significant associations with the clusters after adjusting for different covariates (see Table E5).
![figure](/na101/home/literatum/publisher/thoracic/journals/content/ajrccm/2023/ajrccm.2023.208.issue-2/rccm.202211-2107oc/20230703/images/medium/rccm.202211-2107ocf3.gif)
Figure 3. Box-and-whisker plots showing the enrichment scores of the Molecular Signatures Database gene sets among the taxa-driven clusters. Only the statistically significant gene sets are shown, and those passing the multiple testing threshold (q < 0.05) are shown in boldface type. Overall P values were generated using the Kruskal-Wallis H test, and pairwise P values were generated using Dunn’s post hoc test. TGF-β = transforming growth factor-β.
[More] [Minimize]The baseline microbiota clusters were significantly associated with exacerbation risk during a follow-up period of 12–18 months (Table 4). Logistic regression analysis showed that microbiota clusters are a significant predictor of having two or more future exacerbations per year after adjusting for several covariates (see Table E5). In particular, cluster 3 and cluster 4 had statistically significantly lower odds ratio (ORs) of having two or more future exacerbations per year relative to cluster 2 (ORs, 0.06 [95% CI, 0.01–0.43] and 0.11 [95% CI, 0.03–0.34], respectively; q < 0.05 for both; see Table E6). Other covariates that were associated with the frequency of two or more exacerbations per year included atopic dermatitis diagnosis, center of inclusion, and recent prescription of antibiotics at baseline visit (q < 0.05 for all; see Table E6). Notably, the transcriptomic (Wnt/β-catenin and TGF-β signaling) pathways’ ESs were not selected by the stepwise regression model as significant predictors of having two or more future exacerbations. Mediation analysis (see Table E7) showed that microbiota cluster 2 was an independent predictor of having two or more future exacerbations, and this association was not mediated by atopic dermatitis status, whether this was not adjusted or adjusted for the different covariates.
![figure](/na101/home/literatum/publisher/thoracic/journals/content/ajrccm/2023/ajrccm.2023.208.issue-2/rccm.202211-2107oc/20230703/images/medium/rccm.202211-2107octf1.gif)
Table 4. Baseline Microbiota Cluster Assignments and the Frequency of Exacerbations per Year during Follow-Up (12–18 mo after Baseline Visit)
The bottom panel shows a Sankey diagram for baseline microbiota cluster assignments and the frequency of exacerbations per year during follow-up.
[More] [Minimize]A summary overview of clinical features associated with the taxa-driven clusters is provided in Box 1 and Figure E17.
Cluster 1: Cluster 1 included 50 (20.7%) patients. Cluster 1 had the fewest children with diagnoses of atopic dermatitis (n = 24 of 50 [48%]; post hoc P = 0.001); the fewest children with atopic sensitization (n = 25 of 47 [53%]; post hoc P = 0.003), particularly to grass pollens (n = 13 of 46 [28%]; post hoc P = 0.020); and the lowest median number of aeroallergen sensitizations (median, 0 [interquartile range (IQR), 0–3]) compared with the other clusters (P < 0.05 for all). Generally, patients within cluster 1 showed the lowest median enrichment scores (ESs) of blood transcriptomic signatures related to the Wnt/β-catenin (P < 0.001, q < 0.01), TGF-β (transforming growth factor-β) signaling (P < 0.01, q < 0.05), and myogenesis (P < 0.05, q > 0.05) Molecular Signatures Database (MSigDB) hallmark pathways.
Cluster 2: Cluster 2 was the largest cluster and included 88 patients (36.5%). Cluster 2 included older school-age patients with asthma (median age, 9.5 yr [IQR, 4.8–13.0 yr]) and exhibited the highest percentage of school-age patients with severe asthma (n = 39 of 88 [44%]; post hoc P = 0.034) compared with the other clusters (median age, 5 yr), but neither difference was statistically significant at the overall cluster level (P = 0.144 and P = 0.057, respectively). Cluster 2 showed the highest percentage of atopic dermatitis diagnosis (n = 68 of 87 [78%]; post hoc P = 0.005); the highest percentage of atopic sensitization to aeroallergens, particularly grass pollens (n = 45 of 81 [56%]; post hoc P = 0.005); and the highest median number of aeroallergen sensitizations (median, 3 [IQR, 1–4]) compared with the other clusters. In addition, cluster 2 had the lowest median value of FEV1% predicted after bronchodilator therapy (median, 98.3% [IQR, 88.8–107.7%]) and the highest percentage of children with severe symptoms who experienced future exacerbations (n = 37 of 48 [77%]), particularly two or more future exacerbations per year (n = 33 of 48 [69%]; post hoc P = 0.002) during the follow-up period (P < 0.05 for all). This cluster showed the relatively highest median ESs of the TGF-β signaling MSigDB hallmark pathway (P < 0.01, q < 0.05).
Cluster 3: Cluster 3 was the smallest cluster and included 21 patients (8.7%). This cluster had the highest percentage of preschoolers with severe wheezing (n = 10 of 21 [48%]; post hoc P = 0.026) compared with the other clusters, but this difference was not statically significant at the overall cluster level (P = 0.057). This cluster had significantly higher median values of FEV1% predicted after bronchodilator therapy compared with cluster 2 (medians, 107.2% [IQR, 104.2–118.3%] and 98.3% [IQR, 88.8–107.7%], respectively; P < .05) and had the highest percentage of children with severe symptoms (n = 8 of 11 [73%]; post hoc P = 0.006) who did not experience any future exacerbations during the follow-up period.
Cluster 4: Cluster 4 was the second largest cluster and included 82 patients (34%). This cluster included the second-highest percentage of children with severe symptoms (n = 28 of 42 [67%]) who experienced future exacerbations and particularly included the highest percentage of children with severe symptoms who experienced one exacerbation per year (n = 18 of 42 [24%]; post hoc P = 0.018) during the follow-up period compared with the other clusters (P < 0.05). This cluster showed the relatively highest median ESs of the Wnt/β-catenin signaling pathway and the second highest median ESs of the TGF-β signaling MSigDB hallmark pathway (P < 0.01 for all, q < 0.05).
Using unsupervised clustering of oropharyngeal microbiota profiles, we found that children with school-age asthma and preschool-age wheezing could be stratified into four clusters that differed significantly by 1) atopic dermatitis diagnosis, grass pollen sensitization, and postbronchodilator FEV1% predicted; 2) the Wnt/β-catenin and TGF-β signaling peripheral blood transcriptomic pathways; and 3) the frequency of future exacerbations among children with severe disease.
Age has been reported to be a significant contributor to the variability in microbiome compositions, while the largest influence of age on microbial profiles was observed very early in life when microbial maturation is achieved (38, 39). In this study, the clustering strategy performed only showed a trend of statistically nonsignificant differences among the clusters concerning age, with cluster 2 being more likely to include older school-age children compared with the other clusters. In addition, in this study, no clear separation of the oropharyngeal microbiota clusters between children with school-age asthma and children with preschool wheezing stratified by disease severity was found, showing an overlapping oropharyngeal microbiome composition in this studied cohort (1–17 years of age). These findings are in line with those of the previous U-BIOPRED study (19), suggesting that classical clinical labeling of asthma or wheezing severity cannot be fully explained by the underlying microbiota composition, and attempts to reclassify/rephenotype children with asthma or wheezing should be sought. In this regard, clustering children by underlying microbiota composition may offer new insight in disease pathophysiology, particularly if this could be linked with objective measures of disease burden, such as future exacerbations.
The taxa-driven clusters were significantly different in atopic dermatitis diagnosis and grass pollen sensitization. Particularly, cluster 2 included the highest percentage of children with atopic dermatitis, grass pollen sensitization, and school-age severe asthma. In addition, children with severe asthma or wheezing within this cluster included the highest percentage of those who experienced two or more exacerbations in the following year. Atopic sensitization and comorbid allergy are among the main characteristics that shape childhood asthma phenotypes (40, 41). Our findings are partly in line with those of other studies; for example, in one study it was found that concomitant atopic dermatitis might be one of the factors associated with severe asthma (42). Moreover, reports have shown that atopic sensitization to inhalant allergens increases the risk of acute exacerbations and asthma-related hospital admissions in childhood asthma (43). In a study conducted in children with asthma 3–17 years of age, it was found that viral infections and allergen sensitization synergistically increased the risk of asthma-related hospital admissions (44). In our study, both the taxa-driven clusters and atopic dermatitis diagnosis were found to be significantly associated with the frequency of future exacerbations, confirming previous observations. This suggests that the microbiota–allergen interaction may play a role in the pathogenesis of childhood asthma (45), and this interaction needs further investigations in future childhood asthma studies.
Cluster 2 (Veillonella dominant and the second highest cluster in Shannon α-diversity) and cluster 4 (Haemophilus and Neisseria dominant and the highest cluster in Shannon α-diversity) included the highest percentage of children with severe asthma or wheezing who experienced episodes of future exacerbations (two or more and one exacerbation, respectively) compared with the other clusters. In addition, children in cluster 2 had lower FEV1% predicted after salbutamol compared with those in the other clusters. These results are partly in line with findings from another study showing that increased bacterial Shannon α-diversity and increased Veillonella and Neisseria abundances in hypopharyngeal aspirates were associated with increased duration of asthma-like episodes in preschoolers (46). In the same study, an augmented effect of azithromycin in decreasing asthma-like episodes was particularly seen in children with elevated bacterial relative abundances of Veillonella and Neisseria genera (46). In other microbiome studies in which the oropharynx was sampled, preschoolers with wheezing had partly increased concentrations of Neisseria genus compared with those without wheezing (47), while children with school-age asthma (6–12 years of age) had increased abundances of Haemophilus genus compared with healthy control subjects (48). These findings suggest that the pharyngeal relative compositions of the aforementioned bacterial genera may play a role in childhood asthma pathogenesis and may be useful in the phenotypic assessment of children with asthma or wheezing.
Interestingly, cluster 3 (Rothia dominant) included the highest percentage of children who did not experience any future exacerbations during the follow-up period. This is in line with findings from the U-BIOPRED adult cohort, in which high-abundance R. mucilaginosa in sputum was associated with a less severe asthma phenotype (6). Moreover, R. mucilaginosa was found to be negatively correlated with MMP-1 (matrix metalloproteinase-1), MMP-8, and MMP-9, and proinflammatory markers (IL-8 and IL-1β) in sputum from adults with bronchiectasis (49). In vivo/in vitro mechanistic investigation in the same study revealed that R. mucilaginosa potentially mitigates inflammation by inhibiting NF-κB (nuclear factor-κB) pathway activation, which may influence the severity and development of chronic airway diseases (49). These findings suggest that Rothia species may have a protective effect in adults and children with severe asthma or wheezing.
The peripheral blood transcriptome showed statistically significant differences among the clusters in two main pathways: the TGF-β and Wnt/β-catenin signaling pathways. The TGF-β pathway was relatively upregulated in cluster 2 (the most atopic) and, to a lesser extent, cluster 4 compared with cluster 1. The TGF-β signaling pathway has been reported to be involved in airway remodeling in atopic asthma after exposure to an allergen (50) and may play a role in asthma severity (51). The TGF-β pathway has been reported to regulate the immune system in response to bacterial infections (52). On the other hand, the Wnt/β-catenin signaling pathway was relatively upregulated in the Haemophilus/Neisseria-predominant cluster 4 compared with the other clusters. Similarly, the Wnt/β-catenin pathway has been reported to regulate remodeling in asthmatic airways and may play a role in asthma pathogenesis (53–55). In addition, Wnt signaling is believed to be one of the key pathways involved in host–bacterial pathogen interactions (56). The canonical Wnt/β-catenin signaling pathway was found to be activated after infection with virulent strains of Haemophilus species, leading to disruption of the epithelial barrier in pigs (57). In addition, it was recently found that WNT/β-catenin signaling is a key regulator of macrophage phagocytosis after exposing chronic obstructive pulmonary disease airway epithelium to nontypeable H. influenzae (58). In light of the findings of this study and evidence of involvement of these pathways in airway remodeling and pathogenesis, further research is needed to explore the interaction of the oropharyngeal microbiota with these pathways in childhood asthma.
In this study, we showed that clustering on the oropharyngeal microbiota revealed subtypes of children with asthma or wheezing with distinct clinical characteristics (phenotypes). A decision tree based on the relative abundances of only three main bacterial genera showed that we can classify the children with 86% accuracy, as estimated from a hold-out validation set. These bacterial genera were among the most abundant and prevalent bacterial taxa found in other pediatric cohorts in which the pharynx was sampled for microbiota analysis (46–48). This suggests that the taxa-driven clustering strategy in this study may be generalizable to other asthma or wheezing cohorts and settings. Phenotyping in children with asthma or wheezing remains challenging, particularly because of the difficulty in obtaining invasive samples representing the lower airways and their inflammatory status. Oropharyngeal (throat) swabs are relatively easy to obtain, and 16S sequencing is likely to be an affordable option to sequence the bacterial profiles within these samples. Hence, using oropharyngeal taxa-driven clustering may help clinicians noninvasively define subgroups among children with asthma or wheezing who may benefit from personalized monitoring and/or treatment options. Yet validation of these findings in external cohorts and different populations remains necessary before they can be rendered translatable to clinical practice, as well as to assess whether the airway microbiome can be therapeutically targeted in children with asthma or wheezing (59).
Our study has many strengths. First, the adopted analysis strategy is unsupervised and is not driven by a priori hypothesis. Second, the prospective international multicenter design of this study, the relatively large sample size, and the wide age range of the included cohort make the results likely more generalizable and valid than previously reported single-center studies with smaller sample sizes. Third, oropharyngeal samples are easy to collect and convenient for patients and healthcare professionals, making phenotypic assessment on the basis of oropharyngeal microbiota feasible. Finally, internally validating the findings by using different clustering algorithms makes the statistical analysis more robust.
However, there are also limitations. First, the 16S amplicon sequencing technique used is limited in identifying bacterial taxa up to the species level, and this imposes some hindrance in identifying the pathogenicity of the differentially abundant bacterial genera among the clusters. More sophisticated techniques, such as shotgun metagenomics that enable species or strain-level identification of bacteria, may therefore provide better insight in this regard. However, in another U-BIOPRED study conducted in adult patients with severe asthma, clustering on the basis of 16S amplicon sequencing revealed similar patient clusters to shotgun metagenomics, and this suggests that the affordable 16S amplicon sequencing is potentially applicable in clinical settings in which metagenomics cannot be performed (6). Second, we sampled only one compartment for bacterial sequencing (oropharyngeal swabs). Whether other sampling compartments and detecting other microorganisms (e.g., viruses/fungi) will provide additional information needs to be determined. Third, although patients were followed up clinically for 12–18 months, no additional biological samples were collected to assess the longitudinal shifts in microbial profiles with time. Although the taxa-driven clusters were found to be relatively stable after 12–18 months in adults with severe asthma (6), further longitudinal investigations are still needed in children with asthma. Fourth, we did not collect some parameters in the U-BIOPRED pediatric cohort, such as birth weight and antenatal (in utero) and perinatal antibiotic treatment history; therefore, we were not able to investigate whether they are associated with oropharyngeal microbiota profiles. Further research is needed to investigate whether these characteristics could influence the oropharyngeal microbiota composition. Fifth, assessing the peripheral blood gene expression may not reflect airway-centric asthma pathobiology, and therefore further research is needed to investigate whether the microbiota composition is directly linked to pathophysiologic biological pathways underlying the diseased airways. Sixth, corticosteroids have been reported to influence the composition of the airway microbiome (60). Although we have adjusted for corticosteroid intake in all regression analyses performed, this may not reflect the systemic concentrations of corticosteroids, and this could influence the oropharyngeal microbiota composition. Finally, using an external cohort for validation is still needed to confirm the present findings.
We have shown that oropharyngeal taxa-driven clustering can be used as an unsupervised method of detecting subtypes in children with preschool wheezing and school-age asthma. These subtypes exhibited differences in allergic status, lung function, blood transcriptomic pathways involved in airway remodeling, and the frequency of future exacerbations. These findings suggest that the oropharyngeal microbiota can be used as a noninvasive approach to phenotype children with asthma or wheezing.
The authors thank all the participating children and their families. In addition, U-BIOPRED is a consortium effort, and the authors acknowledge the help and expertise of the individuals and groups whose names are mentioned in the U-BIOPRED study group list available in the online supplement.
U-BIOPRED Collaborators: Johan Kolmert, Anna James, Ana R. Sousa, John H. Riley, Stewart Bates, Per S. Bakke, Massimo Caruso, Pascal Chanez, Stephen J. Fowler, Thomas Geiser, Peter Howarth, Ildiko Horvath, Norbert Krug, Paolo Montuschi, Marek Sanak, Annelie Behndig, Dominick E. Shaw, Richard G. Knowles, Cecile T. J. Holweg, Asa M. Wheelock, Barbro Dahlen, Kjell Alving, Kian Fan Chung, Ian M. Adcock, Peter J. Sterk, Ratko Djukanovic, Sven-Erik Dahlen, Craig E. Wheelock, Lars Andersson, Charles Auffray, Hans Bisgaard, Bertrand De Meulder, Louise J. Fleming, Urs Frey, Rene Lutter, Sharon Mumby, Graham Roberts, M. Uddin, S. S. Wagers, and N. Zounemat Kermani.
1. | Bisgaard H, Nørgaard S, Sevelsted A, Chawes BL, Stokholm J, Mortensen EL, et al. Asthma-like symptoms in young children increase the risk of COPD. J Allergy Clin Immunol 2021;147:569–576.e9. |
2. | Pijnenburg MW, Frey U, De Jongste JC, Saglani S. Childhood asthma: pathogenesis and phenotypes. Eur Respir J 2022;59:2100731. |
3. | Belkaid Y, Hand TW. Role of the microbiota in immunity and inflammation. Cell 2014;157:121–141. |
4. | Abdel-Aziz MI, Vijverberg SJH, Neerincx AH, Kraneveld AD, Maitland-van der Zee AH. The crosstalk between microbiome and asthma: exploring associations and challenges. Clin Exp Allergy 2019;49:1067–1086. |
5. | Abdel-Aziz MI, Neerincx AH, Vijverberg SJ, Kraneveld AD, Maitland-van der Zee AH. Omics for the future in asthma. Semin Immunopathol 2020;42:111–126. |
6. | Abdel-Aziz MI, Brinkman P, Vijverberg SJH, Neerincx AH, Riley JH, Bates S, et al.; U-BIOPRED Study Group. Sputum microbiome profiles identify severe asthma phenotypes of relative stability at 12 to 18 months. J Allergy Clin Immunol 2021;147:123–134. |
7. | Diver S, Richardson M, Haldar K, Ghebre MA, Ramsheh MY, Bafadhel M, et al. Sputum microbiomic clustering in asthma and chronic obstructive pulmonary disease reveals a Haemophilus-predominant subgroup. Allergy 2020;75:808–817. |
8. | Abdel-Aziz MI, Vijverberg SJH, Neerincx AH, Brinkman P, Wagener AH, Riley JH, et al.; U-BIOPRED Study Group. A multi-omics approach to delineate sputum microbiome-associated asthma inflammatory phenotypes. Eur Respir J 2022;59:2102603. |
9. | Charlson ES, Bittinger K, Haas AR, Fitzgerald AS, Frank I, Yadav A, et al. Topographical continuity of bacterial populations in the healthy human respiratory tract. Am J Respir Crit Care Med 2011;184: 957–963. |
10. | Dickson RP, Erb-Downward JR, Freeman CM, McCloskey L, Beck JM, Huffnagle GB, et al. Spatial variation in the healthy human lung microbiome and the adapted island model of lung biogeography. Ann Am Thorac Soc 2015;12:821–830. |
11. | Ahmed B, Cox MJ, Cuthbertson L, James PL, Cookson WOC, Davies JC, et al. Comparison of the upper and lower airway microbiota in children with chronic lung diseases. PLoS One 2018;13:e0201156. |
12. | Bisgaard H, Hermansen MN, Buchvald F, Loland L, Halkjaer LB, Bønnelykke K, et al. Childhood asthma after bacterial colonization of the airway in neonates. N Engl J Med 2007;357:1487–1495. |
13. | Thorsen J, Rasmussen MA, Waage J, Mortensen M, Brejnrod A, Bønnelykke K, et al. Infant airway microbiota and topical immune perturbations in the origins of childhood asthma. Nat Commun 2019;10:5001. |
14. | Raita Y, Camargo CA Jr, Bochkov YA, Celedón JC, Gern JE, Mansbach JM, et al. Integrated-omics endotyping of infants with rhinovirus bronchiolitis and risk of childhood asthma. J Allergy Clin Immunol 2021;147:2108–2117. |
15. | Raita Y, Pérez-Losada M, Freishtat RJ, Hahn A, Castro-Nallar E, Ramos-Tapia I, et al. Nasopharyngeal metatranscriptome profiles of infants with bronchiolitis and risk of childhood asthma: a multicentre prospective study. Eur Respir J 2022;60:2102293. |
16. | Chiu CY, Chan YL, Tsai MH, Wang CJ, Chiang MH, Chiu CC, et al. Cross-talk between airway and gut microbiome links to IgE responses to house dust mites in childhood airway allergies. Sci Rep 2020;10:13449. |
17. | Pérez-Losada M, Authelet KJ, Hoptay CE, Kwak C, Crandall KA, Freishtat RJ. Pediatric asthma comprises different phenotypic clusters with unique nasal microbiotas. Microbiome 2018;6:179. |
18. | Fleming L, Murray C, Bansal AT, Hashimoto S, Bisgaard H, Bush A, et al.; U-BIOPRED Study Group. The burden of severe asthma in childhood and adolescence: results from the paediatric U-BIOPRED cohorts. Eur Respir J 2015;46:1322–1333. |
19. | Thorsen J, Stokholm J, Rasmussen MA, Roggenbuck-Wedemeyer M, Vissing NH, Mortensen MS, et al. Asthma and wheeze severity and the oropharyngeal microbiota in children and adolescents. Ann Am Thorac Soc 2022;19:2031–2043. |
20. | Bel EH, Sousa A, Fleming L, Bush A, Chung KF, Versnel J, et al.; Unbiased Biomarkers for the Prediction of Respiratory Disease Outcome (U-BIOPRED) Consortium, Consensus Generation. Diagnosis and definition of severe refractory asthma: an international consensus statement from the Innovative Medicine Initiative (IMI). Thorax 2011;66:910–917. |
21. | Bousquet J. Global Initiative for Asthma (GINA) and its objectives. Clin Exp Allergy 2000;30:2–5. |
22. | de Onis M, Onyango AW, Borghi E, Siyam A, Nishida C, Siekmann J. Development of a WHO growth reference for school-aged children and adolescents. Bull World Health Organ 2007;85:660–667. |
23. | Siles RI, Hsieh FH. Allergy blood testing: a practical guide for clinicians. Cleve Clin J Med 2011;78:585–592. |
24. | Abdel-Aziz MI, Brinkman P, Vijverberg SJH, Neerincx AH, de Vries R, Dagelet YWF, et al.; U-BIOPRED Study Group; Amsterdam UMC Breath Research Group. eNose breath prints as a surrogate biomarker for classifying patients with asthma by atopy. J Allergy Clin Immunol 2020;146:1045–1055. |
25. | Juniper EF, Guyatt GH, Feeny DH, Ferrie PJ, Griffith LE, Townsend M. Measuring quality of life in children with asthma. Qual Life Res 1996;5:35–46. |
26. | Juniper EF, Guyatt GH, Feeny DH, Ferrie PJ, Griffith LE, Townsend M. Measuring quality of life in the parents of children with asthma. Qual Life Res 1996;5:27–34. |
27. | Nathan RA, Sorkness CA, Kosinski M, Schatz M, Li JT, Marcus P, et al. Development of the asthma control test: a survey for assessing asthma control. J Allergy Clin Immunol 2004;113:59–65. |
28. | Liu AH, Zeiger R, Sorkness C, Mahr T, Ostrom N, Burgess S, et al. Development and cross-sectional validation of the Childhood Asthma Control Test. J Allergy Clin Immunol 2007;119:817–825. |
29. | Reddel HK, Taylor DR, Bateman ED, Boulet LP, Boushey HA, Busse WW, et al.; American Thoracic Society/European Respiratory Society Task Force on Asthma Control and Exacerbations. An official American Thoracic Society/European Respiratory Society statement: asthma control and exacerbations: standardizing endpoints for clinical asthma trials and clinical practice. Am J Respir Crit Care Med 2009;180:59–99. |
30. | Andrews S. FastQC: a quality control tool for high throughput sequence data. Cambridge, United Kingdom: Babraham Institute; 2010 [updated 2023 Mar 1; accessed 2019 Oct 20]. Available from: https://www.bioinformatics.babraham.ac.uk/projects/fastqc/. |
31. | Ewels P, Magnusson M, Lundin S, Käller M. MultiQC: summarize analysis results for multiple tools and samples in a single report. Bioinformatics 2016;32:3047–3048. |
32. | Martin M. Cutadapt removes adapter sequences from high-throughput sequencing reads. EMBnet J 2011;17:3. |
33. | Callahan BJ, McMurdie PJ, Rosen MJ, Han AW, Johnson AJ, Holmes SP. DADA2: high-resolution sample inference from Illumina amplicon data. Nat Methods 2016;13:581–583. |
34. | Quast C, Pruesse E, Yilmaz P, Gerken J, Schweer T, Yarza P, et al. The SILVA ribosomal RNA gene database project: improved data processing and web-based tools. Nucleic Acids Res 2013;41:D590–D596. |
35. | Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, Speed TP. Summaries of Affymetrix GeneChip probe level data. Nucleic Acids Res 2003;31:e15. |
36. | Liberzon A, Birger C, Thorvaldsdóttir H, Ghandi M, Mesirov JP, Tamayo P. The Molecular Signatures Database (MSigDB) hallmark gene set collection. Cell Syst 2015;1:417–425. |
37. | Textor J, van der Zander B, Gilthorpe MS, Liskiewicz M, Ellison GT. Robust causal inference using directed acyclic graphs: the R package ‘dagitty’. Int J Epidemiol 2016;45:1887–1894. |
38. | Mortensen MS, Brejnrod AD, Roggenbuck M, Abu Al-Soud W, Balle C, Krogfelt KA, et al. The developing hypopharyngeal microbiota in early life. Microbiome 2016;4:70. |
39. | Stokholm J, Blaser MJ, Thorsen J, Rasmussen MA, Waage J, Vinding RK, et al. Maturation of the gut microbiome and risk of asthma in childhood. Nat Commun 2018;9:141. |
40. | Bel EH. Clinical phenotypes of asthma. Curr Opin Pulm Med 2004;10:44–50. |
41. | Howrylak JA, Fuhlbrigge AL, Strunk RC, Zeiger RS, Weiss ST, Raby BA; Childhood Asthma Management Program Research Group. Classification of childhood asthma phenotypes and long-term clinical responses to inhaled anti-inflammatory medications. J Allergy Clin Immunol 2014;133:1289–1300.e1–e12. |
42. | Lee JK, Han D. Atopic dermatitis is an important comorbidity in severe asthma. Ann Allergy Asthma Immunol 2018;120:661–662. |
43. | Jackson DJ, Sykes A, Mallia P, Johnston SL. Asthma exacerbations: origin, effect, and prevention. J Allergy Clin Immunol 2011;128:1165–1174. |
44. | Murray CS, Poletti G, Kebadze T, Morris J, Woodcock A, Johnston SL, et al. Study of modifiable risk factors for asthma exacerbations: virus infection and allergen exposure increase the risk of asthma hospital admissions in children. Thorax 2006;61:376–382. |
45. | Jackson DJ, Busse WW, Bacharier LB, Kattan M, O’Connor GT, Wood RA, et al. Association of respiratory allergy, asthma, and expression of the SARS-CoV-2 receptor ACE2. J Allergy Clin Immunol 2020;146:203–206.e3. |
46. | Thorsen J, Stokholm J, Rasmussen MA, Mortensen MS, Brejnrod AD, Hjelmsø M, et al. The airway microbiota modulates effect of azithromycin treatment for episodes of recurrent asthma-like symptoms in preschool children: a randomized clinical trial. Am J Respir Crit Care Med 2021;204:149–158. |
47. | Powell EA, Fontanella S, Boakes E, Belgrave D, Shaw AG, Cornwell E, et al. Temporal association of the development of oropharyngeal microbiota with early life wheeze in a population-based birth cohort. EBioMedicine 2019;46:486–498. |
48. | Boutin S, Depner M, Stahl M, Graeber SY, Dittrich SA, Legatzki A, et al. Comparison of oropharyngeal microbiota from children with asthma and cystic fibrosis. Mediators Inflamm 2017;2017:5047403. |
49. | Rigauts C, Aizawa J, Taylor SL, Rogers GB, Govaerts M, Cos P, et al. Rothia mucilaginosa is an anti-inflammatory bacterium in the respiratory tract of patients with chronic lung disease. Eur Respir J 2022;59:2101293. |
50. | Torrego A, Hew M, Oates T, Sukkar M, Fan Chung K. Expression and activation of TGF-beta isoforms in acute allergen-induced remodelling in asthma. Thorax 2007;62:307–313. |
51. | Al-Alawi M, Hassan T, Chotirmall SH. Transforming growth factor β and severe asthma: a perfect storm. Respir Med 2014;108:1409–1423. |
52. | Sanjabi S, Oh SA, Li MO. Regulation of the immune response by TGF-β: from conception to autoimmunity and infection. Cold Spring Harb Perspect Biol 2017;9:a022236. |
53. | Kwak HJ, Park DW, Seo JY, Moon JY, Kim TH, Sohn JW, et al. The Wnt/β-catenin signaling pathway regulates the development of airway remodeling in patients with asthma. Exp Mol Med 2015;47:e198. |
54. | Jia XX, Zhu TT, Huang Y, Zeng XX, Zhang H, Zhang WX. Wnt/β-catenin signaling pathway regulates asthma airway remodeling by influencing the expression of c-Myc and cyclin D1 via the p38 MAPK-dependent pathway. Exp Ther Med 2019;18:3431–3438. |
55. | Sharma S, Tantisira K, Carey V, Murphy AJ, Lasky-Su J, Celedón JC, et al. A role for Wnt signaling genes in the pathogenesis of impaired lung function in asthma. Am J Respir Crit Care Med 2010;181:328–336. |
56. | Rogan MR, Patterson LL, Wang JY, McBride JW. Bacterial manipulation of wnt signaling: a host-pathogen tug-of-Wnt. Front Immunol 2019;10:2390. |
57. | Hua K, Li Y, Zhou H, Hu X, Chen Y, He R, et al. Haemophilus parasuis infection disrupts adherens junctions and initializes EMT dependent on canonical Wnt/β-catenin signaling pathway. Front Cell Infect Microbiol 2018;8:324. |
58. | Skronska-Wasek W, Durlanik S, Le HQ, Schroeder V, Kitt K, Garnett JP, et al. The antimicrobial peptide S100A8/A9 produced by airway epithelium functions as a potent and direct regulator of macrophage phenotype and function. Eur Respir J 2022;59:2002732. |
59. | Chotirmall SH, Bogaert D, Chalmers JD, Cox MJ, Hansbro PM, Huang YJ, et al. Therapeutic targeting of the respiratory microbiome. Am J Respir Crit Care Med 2022;206:535–544. |
60. | Leitao Filho FS, Takiguchi H, Akata K, Ra SW, Moon JY, Kim HK, et al. Effects of inhaled corticosteroid/long-acting β2-agonist combination on the airway microbiome of patients with chronic obstructive pulmonary disease: a randomized controlled clinical trial (DISARM). Am J Respir Crit Care Med 2021;204:1143–1152. |
U-BIOPRED has received funding from the Innovative Medicines Initiative Joint Undertaking under FP7 Health grant agreement 115010, resources of which are composed of financial contributions from the European Union’s Seventh Framework Programme (FP7/2007-2013) and European Federation of Pharmaceutical Industries and Associations companies’ in-kind contributions (www.imi.europa.eu). M.I.A.-A. has received a full scholarship from the Ministry of Higher Education of the Arab Republic of Egypt (2015/2016).
Author Contributions: M.I.A.-A. performed the analysis and drafted the initial version of the manuscript. M.I.A.-A., A.H.M.-v.d.Z., J.T., G.R., and P.J.S. contributed to the design of the analysis plan. All authors contributed to the acquisition of data, interpretation of the analysis, revision, drafting, critical appraisal, and ensuring the accuracy and integrity of the analysis. All authors provided final approval of the version to be published.
This article has an online supplement, which is accessible from this issue’s table of contents at www.atsjournals.org.
Originally Published in Press as DOI: 10.1164/rccm.202211-2107OC on May 10, 2023
Author disclosures are available with the text of this article at www.atsjournals.org.