The airway fungal microbiome in asthma

Fungal involvement in asthma is associated with severe disease. The full spectrum of fungal species in asthma is not well described and is derived largely from insensitive culture techniques.


| INTRODUC TI ON
Asthma is a common, heterogeneous, chronic inflammatory disorder of the airways. It causes considerable morbidity and avoidable mortality. Up to 70% of severe asthmatics are IgE sensitized to fungi compared with ~10% of mild-to-moderate asthmatics and 5% of the general population. 1,2 Thermotolerant fungi such as Aspergillus fumigatus, unlike mesophilic species such as Cladosporium and Alternaria spp. which act predominantly as allergens, are able to non-invasively colonize the bronchial tree causing a variety of clinical presentations grouped under the term allergic fungal airway disease (AFAD). [3][4][5] AFAD includes all patients where IgE sensitization to thermotolerant filamentous fungi is complicating their asthma in contrast to allergic bronchopulmonary aspergillosis (ABPA) and severe asthma with fungal sensitization (SAFS) which exclude many patients with fungal-related asthma based on relatively arbitrary criteria, particularly total IgE which has recently been shown to be less suitable than A. fumigatus-specific IgE as a biomarker of fungal allergy and lung damage. 6 Over a prolonged period, fungal colonization in association with IgE sensitization appears to cause lung damage with bronchiectasis, lung fibrosis and fixed airflow obstruction. 6,7 Methods for culturing fungi from sputum are often insensitive, favouring faster growing species that can be cultured on generic media. 8,9 Aspergillus fumigatus is the fungus most readily isolated from respiratory samples of patients with asthma. However, using a more sensitive culture method twenty-seven different filamentous fungal taxa were detected in sputum from asthmatics. 10 High-throughput sequencing (HTS) has been used successfully to comprehensively define the microbiological component of an ecological niche. Although most human studies have focused on the community of microbes (microbiota) in the gastrointestinal tract, interest has expanded to include studies of the airways including asthmatics. 11 Bacterial microbiota studies in asthma have suggested a discordance between health and asthma. 12,13 Studies of fungi (mycobiota) lag behind those of bacteria and have focused on the gut and other body sites more than the lungs. 14 Nevertheless, specific mycobiota have been identified in the lungs 15,16 and there is growing recognition that the lung mycobiota has a significant impact on the clinical outcome of chronic respiratory diseases 17 with particular progress having been made in cystic fibrosis 18,19 and bronchiectasis. 20,21 Relatively, little is known about the lung mycobiota in asthma. 22,23 The aim of this study was to characterize the airway mycobiota in asthmatics with and without fungal sensitization, to correlate findings with clinical outcomes and to compare it with the mycobiota of healthy controls and concurrent regional outdoor air samples.

| Subjects
Patients with asthma were recruited from respiratory and allergy clinics at Glenfield hospital, Leicester, UK between 2013 and 2017.
The inclusion criteria included age ≥18 years, with a physician diagnosis of asthma and objective evidence of variable airflow obstruction. Full details on inclusion and exclusion criteria and clinical characterization are provided in Appendix S1. Healthy volunteers served as controls and included members of the public and staff at Glenfield hospital. The study was approved by an appropriately constituted National Health Service ethics committee (REACT IRAS 159074:CRN 159074 and ABPA UHL 10111). All subjects gave their written informed consent.

| Allergy testing
Atopy was assessed using skin prick tests to common allergens, and IgE fungal sensitization assessed using skin prick tests and blood immunology to a panel of eight fungi (Alternaria alternata, A. fumigatus, Candida albicans, Cladosporium herbarum, Malassezia spp., Mucor racemosus, Penicillium chrysogenum and Trichophyton rubrum). A positive test was defined as a weal of ≥3 mm above negative control and specific IgE ≥ 0.35 kU/L. Details of the allergens and fungi tested are provided in Appendix S1.

| Sputum and bronchoscopic samples
Briefly, sputum was obtained either spontaneously or induced, 24,25 with sputum plugs separated from saliva and homogenized. A subset of subjects who were willing to consent to undertake the procedure underwent bronchoscopy in accordance with recent British Thoracic Society guidelines, 26 providing up to three samples for analysis; bronchial wash, bronchoalveolar lavage (BAL) and protected brush samples. Towards the end of the study, we obtained ethical approval to obtain a mouthwash sample prior to producing sputum from a subset of subjects. Further details are available in Appendix S1.
Aliquots of 100 µL of sputum homogenate, concentrated mouthwash, BAL, bronchial wash and brush samples were used for culture and DNA extraction. Fungal culture was performed as previously described. 7,9,10 Total genomic DNA was extracted using a DNeasy plant mini kit (Qiagen) with a modified protocol including bead-beating. 27

| Air samples
Outdoor air samples collected 12 m above ground and representative of a wide geographic area, 29 corresponding to patient sampling dates were obtained for a subset of patients. More than 23K Litres of air was sampled into a microcentrifuge tube over a 24-hour period from midnight and DNA extracted as before. 27 Further details are available in Appendix S1.

| Sequencing
The internal transcribed spacer region 2 (ITS2) of the nuclear ribosomal operon was amplified using barcoded ITS3 and ITS4 primers 30 using a dual index nested PCR approach. Full details including PCR cycling conditions, amplicon cleaning and quantification, and control samples (DNA extraction kit, PCR reagents, sample processing reagents and two fungal mock communities) are provided in Appendix S1. The products were pooled and submitted to the Centre for Genomic Research, University of Liverpool, for paired-end sequencing (2 × 250 bp) on the Illumina MiSeq platform. Sequence data, including those from controls and mock communities, were deposited in the European Nucleotide Archive; study accession number PRJEB31078.

| Data handling, statistical analysis, bioinformatic processing and data analysis
Study data were collected and managed using a research electronic data capture tool, 31 and statistical analysis performed using PRISM version 7.02 (GraphPad Software). Sequence processing incorporated recommended quality-filtering guidelines. 32 Further processing was implemented using the Quantitative Insights Into Microbial Ecology (QIIME) open-source bioinformatics pipeline version 1.9.1 33 and QIIME2.2019.4. 34 Taxonomy was assigned using the UNITE (Userfriendly Nordic ITS Ectomycorrhiza) fungal ITS database. Three alpha diversity indices were used: Chao1 index, a measure of species richness useful for low abundance data sets; observed OTUs, a measure of the number of distinct OTUs; and Shannon index, a measure of richness and evenness. Beta diversity was calculated using the nonphylogenetic Bray-Curtis dissimilarity. To assess the statistical significance of compositional differences, two complementary approaches for differential abundance testing were used: analysis of composition of microbes (ANCOM) 35 and balance trees analysis, 36 sometimes referred to as gneiss analysis. Full details are provided in Appendix S1.

| Subjects
Sputum samples were obtained from 83 people with asthma (54 IgE sensitized to fungi), and 14 healthy controls ( Table 1). The three groups were well matched for sex and age. For asthma, the fungalsensitized group had earlier onset asthma (P = .003), longer asthma duration (P < .001), received lower levels of asthma treatment (P = .041) and required less oral steroids (P = .025), trended towards impaired post-bronchodilator FEV 1 (P = .077) and had increased immunological markers of fungal sensitization. Sputum fungal culture rates were low, with no significant difference between the three groups. Approximately three-quarters of the fungal-sensitized group were sensitized to A. fumigatus (Table 1).

| Fungal ITS2 amplicons
Fungal amplicons were generated from the majority of samples, although success rates were notably lower for bronchial brush samples, particularly from healthy subjects ( Table 1). Forty-five air samples were obtained and sequenced; 28 matched dates when sputum from 31 subjects was obtained and 18 matched days on which 19 bronchoscopies were performed. The average number of high-quality fungal reads remaining after quality control was 123106 per sample (range 5085-543347, Table S1) with the rarefaction curves reaching a plateau, indicating that almost all the mycobiota present were detected ( Figure S1).

| Sputum airway mycobiota
Fungal sequences clustered into 830 operational taxonomic units (OTUs), which after removal of very rare (<0.005% relative abundance) OTUs represented 206 taxa at the species level, of which 154 were named species (Table S1). The sputum mycobiota was dominated by three species: A. fumigatus, C. albicans and Mycosphaerella tassiana ( Figure 1A), although, the relative abundance per individual was highly variable. C. albicans was the most prevalent fungus detected, being present in all samples, whilst A fumigatus had the highest average relative abundance (Table S2). There were no significant differences in alpha diversity between asthma and health, although there was a trend towards increased alpha diversity in sputum obtained from people with asthma. Grouping samples from asthmatics according to whether the individual was sensitized to fungi also showed no difference in alpha diversity (Table S3, Figure 1B). Neither beta diversity nor fungal taxa detected were significantly different grouping samples according to asthma status or sensitization (Tables S4 and 2, Figure 1C).
Culture positivity likely indicates a higher burden of fungus; therefore, we investigated whether culture of yeasts or A. fumigatus from sputum influenced the mycobiota. Whilst there was no significant effect on alpha diversity, PERMANOVA analysis indicated significant differences in beta-diversity grouping samples as yeast culture positive or negative, or A. fumigatus culture positive or negative, with or without healthy controls (Table S4)  ANCOM and balance trees analysis confirmed that yeast culture was highly associated with increased relative abundance of C. albicans sequences (Table 2, Figure S3A). Similarly, balance trees analysis found a cluster of four species comprising A. fumigatus,

Aspergillus tubingensis, Penicillium roqueforti and an unidentified
Ascomycota associated with the culture of A. fumigatus (Table 2, Figure S3B).

| Comparing lung compartments
Twenty-nine subjects underwent a bronchoscopy, resulting in ITS2 sequence data from 23 bronchial washes, 21 BAL and 14 brush samples, as detailed in Table S1. Consistent with sputum, C. albicans, M. tassiana and A. fumigatus were amongst the most prevalent and relatively abundant fungi detected from all sample types; however,

TA B L E 1 (Continued)
F I G U R E 1 Sputum samples grouped into healthy controls (n = 13), fungal-sensitized asthmatics (n = 53) and non-fungal-sensitized asthmatics (n = 26). A, Heatmap showing the relative abundance of fungal ITS2 sequence data between samples. The data were rarefied then transformed using ln(x + 1) due to zeros in the data set. Only taxa with a relative abundance ≥1% are shown. other fungi were also prominent from the bronchoscopy-derived samples, such as A. tubingensis. The fungus with the highest relative abundance remained A. fumigatus (Figure 2A, Table S5), and however, C. albicans was no longer the most prevalent.
There was no significant difference in alpha diversity between the sample types ( Figure 2B, Table S6). In contrast, there were significant differences in beta diversity between the brush and wash and brush and sputum ( Figure 2C, Table S7), although significance was lost when correcting for multiple samples and only brush samples showed clustering on a PCoA plot ( Figure 2D).
As with sputum, grouping by culture of yeast resulted in significant differences in beta diversity (pseudo-F = 4.52, P = .003, Figure S2C). There was no significant effect on beta-diversity grouping samples according to whether they came from healthy controls or asthmatics; however, there were significant differences when samples were grouped according to the patient they were obtained from (PERMANOVA pseudo-F = 3.02, P = .001, Figure S2D).
Balance tree analysis, whilst not statistically significant, grouped  Figure S5).
The unidentified fungus was found to have a higher mean relative abundance in healthy control samples, whereas the other three were higher from asthmatics. Balance trees analysis did not find any significant differences; however, the box plot suggested a trend towards differences and the heatmap revealed a cluster of six fungi with sputum and seven without that may explain some of the compositional variability (Table S9, Figure S5). H. radula and C. pseudolongus were present in both clusters.
Given the variability of the mycobiota and the significant effect of the individual, we used a paired analysis matching samples by subject. Significant differences in Shannon diversity index were observed between health and asthma pairing sputum samples with BAL or wash, Figure S4.  Figure S6A, Table S10). No filamentous fungi were detected with a mean relative abundance >1%.

| Influence of the oral cavity
Alpha diversity was lower from mouthwash than sputum samples; particularly using the Shannon index, although did not reach statistical significance (Table S6, Figure S6B, P = .061). There was no significant difference in beta-diversity measures (Table S7) between mouthwash and sputum and no differentially abundant taxa detected (Table S8). PCoA plots showed no clustering by sample type or patient ID ( Figure S6C,D).

| Impact of outdoor air
Mycosphaerella tassiana dominated air samples, being the most prevalent and abundant fungus detected (Table S11, Figure 3A). In  (Tables S2 and S5), whereas were not abundant in outdoor air samples (Table S10). Further details can be found in Appendix S1.

| Relationship between mycobiota and clinical outcomes
To determine whether any clinical characteristics were associated with the lung mycobiota, we performed analyses with and without healthy controls, unless asthma specific. Asthma duration had no association with measures of alpha or beta diversity (Tables S3   and S4); however, the balance trees analysis suggested the fungal community composition changes significantly with time since diagnosis ( Figure 5G). GINA class, bronchiectasis, post-bronchodilator FEV 1 , smoking status, levels of fractional exhaled nitric oxide (FeNO), use of oral steroids and dose of inhaled steroid had no significant associations with diversity metrics, or associations with differentially abundant taxa (Tables 2, S3 and S4). However, there was a trend towards increased beta diversity between people with asthma not taking or on low doses of inhaled steroids compared to those taking high doses ( Figure S7E). Balance trees analysis suggested groups of fungi changed with increasing levels of FeNO ( Figure 5H).  Figure S7). The relationship was also observed when excluding healthy controls; however, significance was lower and only significant with the Shannon index. Beta diversity was also significantly different according to A. fumigatus IgG status, with and without healthy controls (Table S4), but no taxa were found to be differentially abundant ( Table 2). For samples from asthmatics, there was a statistically significant shift in the mycobiota as levels of A. fumigatus IgG increased (Table S12, Figure 5E), with A. fumigatus being amongst the associated group of fungi.
To investigate whether the mycobiota was influenced by biomarkers of inflammation, patients were grouped according to sputum inflammatory phenotypes. Peripheral blood eosinophilia was not associated with differences in alpha diversity (Table S3), and however, there were significant beta-diversity differences between samples from people with low levels (<0.2 × 10 −9 /L) compared to high (>0.4 × 10 −9 /L, Table S4, Figure S7). Balance trees analysis suggested significant shifts in balances between groups of fungi for people with asthma associated with increasing blood eosinophil counts (Table S12, Figure 5F). Using sputum cell differential data, we first compared individuals who were eosinophilic (sputum eosinophils ≥3%) against non-eosinophilic, and individuals who were neutrophilic (sputum neutrophils ≥61%) against non-neutrophilic. We then grouped patients by their inflammatory phenotype 37 ; classifying patients as neutrophilic, F I G U R E 2 Samples from patients who underwent a bronchoscopy representing sputum samples (n = 25), bronchial wash (n = 23), bronchoalveolar lavage (BAL) (n = 21) and protected brush samples (n = 14). A, Heatmap showing the relative abundance of fungal ITS2 sequence data between samples. The data were rarefied then transformed using ln(x + 1) due to zeros in the data set. Only taxa with a relative abundance ≥1% are shown. The relative abundance is indicated by colour, with the lighter shades representing the more abundant sequences. B, Box plot of Shannon alpha diversity indices, showing no significant differences in alpha diversity between groups. C, Box plots of Bray-Curtis dissimilarity showing distance from brush (brush to sputum, P = .013, q = 0.078, brush to wash P = .035, q = 0.105). D, Principal coordinate analysis plots based on the Bray-Curtis distance values suggesting weak clustering of protected brush samples eosinophilic, mixed granulocytic (≥61% neutrophils and ≥3% eosinophils) or paucigranulocytic (<61% neutrophils and <3% eosinophils). Sputum inflammatory data were only available for a subset of subjects, comprising 18, 14, 17 and 5 people with asthma, or 24, 14, 17 and 7 combined with healthy controls, for neutrophilic, eosinophilic, mixed granulocytic and paucigranulocytic phenotypes respectively. No association was seen with diversity metrics (Tables S3 and S4). Sputum eosinophils and neutrophils as continuous variables were associated with changes in the balance of groups of fungi in the mycobiota (Table S12, Figure 5C,D). Both ANCOM and balance trees analysis found significantly different abundant species between samples from patients with high versus low neutrophil counts and amongst individuals grouped into the four inflammatory phenotypes. The strongest signal was from A. tubingensis, with and without the inclusion of healthy controls, which were negatively associated with neutrophils (Table S12, Figure 6). Penicillium roqueforti was often grouped with A. tubingensis.

| D ISCUSS I ON
This study provides a comprehensive assessment of the airway mycobiota in asthma compared to healthy controls. We also compared the oral mycobiota and aeromycobiota allowing an assessment of those species that were present as a result of inhalation rather than colonization. The analysis of the airway mycobiota revealed a complex community with high diversity between individuals, but distinct shifts in the mycobiota detected dependent on the lung compartment studied and the health status of the individual. This was particularly true of the epithelial lining of the lung, sampled using protected brushes which minimizes risk of oropharyngeal contamination. 38  the mouthwash that were not found in the lower airways. It is clear from this study that C. albicans and other yeasts can also be found in substantial amounts in the lower airway. It is possible that there was contamination from oral flora in sputum, BAL and wash samples, but C. albicans was also present, albeit at lower levels, in the protected brush specimen where minimal contamination would be anticipated.
The increased beta diversity between brush and wash, and between brush and sputum supports this. A. fumigatus is a dominant pathogen in fungal lung disease, but it was interesting to discover that it was present in nearly all subjects, including the non-asthmatics. Other studies of the normal airways have also demonstrated the presence of Aspergillus as part of the normal airway mycobiota. 43 We have previously shown increased rates of sputum culture for A. fumigatus in asthmatics (~50%) compared to healthy subjects (~10%), especially in people with fungal IgE sensitization. 7,10 In this study, we only culturedA. fumigatus from asthmatics, although at lower rates than our earlier studies because most of the sample was used for DNA extraction.
Fraczek et al 22 in one of few studies of the mycobiota in asthma focussed on overall fungal load in BAL fluid suggesting that corticosteroids could encourage fungal growth, although that observation was confounded by asthma severity and treatment with azoles. In our study, all the asthmatics were on IHC, and although the dose of IHC did not appear to influence the prevalence and relative abundance of the airway mycobiota, there was a trend towards increased beta diversity between the high and low dose groups. The correlation between A. fumigatus and C. albicans relative abundances with culture outcomes validated our HTS findings.
More than 200 fungal taxa were detected in the airway; however, the majority were only present at low concentrations in a minority of individuals. Due to the compositional nature of HTS data, conventional statistical tools such as the t test and ANOVA F I G U R E 3 Outdoor air samples were analysed from 45 d, matching clinical samples from sputum (n = 31), bronchial wash (n = 18), BAL (n = 18) and protected brush samples (n = 13). A, Heatmap showing the relative abundance of fungal ITS2 sequence data between samples. The data were rarefied then transformed using ln(x + 1) due to zeros in the data set. Only taxa with a relative abundance ≥1% are shown. The relative abundance is indicated by colour, with the lighter shades representing the more abundant sequences. B, Box plot of Shannon alpha diversity indices, showing greater alpha diversity in outdoor air samples than clinical samples. C, Box plots of Bray-Curtis dissimilarity showing distance from outdoor air (outdoor air to brush, sputum, wash and BAL were each P = .001, q = 0.003, brush to sputum P = .025, q = 0.05). D, E, Principal coordinate analysis plots based on the Bray-Curtis distance values, the purple ellipse surrounding the air samples; in (D) clinical samples are coloured according to whether they were from healthy controls or people with asthma and in (E) according to sample type. Grey spheres in (C) indicate the coordinates where the main fungal species are centred, with size correlating to relative abundance of the labelled organism struggle to accurately detect differentially abundant taxa between groups. 44 To account for this compositionality, analysis of composition of microbiomes (ANCOM) 35 and balance trees analysis 36 were used in this study. ANCOM is a powerful method for determining microbial differential abundance, with good control of false-discovery rates. 45 ANCOM does, however, have issues with large numbers of zeros and assumes that less than 25% of taxa are changing in abundance. 36 To limit the possible increase in error, balance tree analysis was used in addition to ANCOM that attempts to calculate changes in the balance (log ratio of abundance) between subcommunities of taxa rather than individuals.
These subsets of taxa were constructed either by clustering on how often they co-occur with one another (categorical data) or clustering on a continuous variable if continuous data were investigated. Gradient clustering is more prone to false positives, and hence, caution was used in interpreting implicated fungi. The use of balances enables the identification of taxa whose relative abundances were most significant in explaining compositional variations.
Few bacterial studies have compared diversity scores between asthma and health, and those that have had conflicting findings ranging from decreased 46 to increased alpha diversity in asthmatics, 12 with other studies finding no differences in diversity metrics. 47 Fewer studies have compared the mycobiota, but adult, 22 and paediatric 23 studies have found no significant differences in standard diversity metrics between asthma and health. Using paired samples between sputum and bronchoscopy-derived samples from individuals, we found significant differences in alpha diversity between asthma and health. We found differentially abundant species, including C. pseudolongus, between asthma and health when analysing bronchoscopy-derived but not sputum samples, and noted significant shifts in the mycobiota with asthma duration.  Table  S12 There has been increasing momentum towards using endotyping as a means of grouping individuals by a distinct pathophysiological mechanism. 48 Stratifying patients according to their underlying inflammation is one such approach, and studies have compared the composition of the bacterial microbiota of the lower airways with inflammatory features such as neutrophilia or eosinophilia with conflicting findings. Sputum neutrophilia but not eosinophilia was correlated with decreased diversity measures in severe asthma 37 and moderate-to-severe asthma, 49 whereas relationships seen with eosinophils in mild asthma. 50 To our knowledge, this is the first study to compare inflammatory phenotypes with the lung mycobiota, finding significant differences with blood eosinophils and sputum differential counts. Beta diversity was significantly different between individuals with low and high blood eosinophil counts, and groups of fungi within the mycobiota were seen to shift with increasing blood eosinophil count. Whilst increases in both sputum eosinophils and neutrophils were seen to cause shifts in the mycobiota, it was associations with neutrophils that were most interesting, with rela-   Although this study includes more subjects than many comparable respiratory bacterial microbiota studies [55][56][57] or asthma mycobiota studies, 22,23 we lacked power to measure significant differences in community structure and were unable to effectively adjust for factors such as age or medication usage. There are inherent challenges studying the mycobiota using a PCR-based approach. Non-homogenous lysis of fungal cells during DNA extraction may introduce biases, and however, we utilized a recommended bead-beating approach. 58 We have previously shown our DNA extraction method and primer choice to be highly suitable for capturing a diverse range of fungal diversity 27 and included in this study two mock fungal communities 28 to test the DNA extraction and HTS-PCR protocols and to optimize our bioinformatic pipeline. Recently, our region of choice (ITS2) has been shown to introduce less bias than the more commonly used ITS1 region in both mock communities and sputum samples, 58 supporting our choice of regions. One caveat with either ITS region is that many clinically relevant species are indistinguishable from close relatives using ITS alone 59,60 and therefore are more reliably referred to by the species complex within which they reside. Primers that are universal for the organisms of interest are a requirement of microbiota studies, but will preferentially amplify the most prevalent flora.
Our study did not identify Pneumocystis which is sometimes identified in respiratory samples. This was not unexpected as there is only one copy of the ITS2 locus in the Pneumocystis genome. 61 A similar study utilizing ITS2 for high-throughput-sequencing cystic fibrosis sputum samples had negative HTS results from samples that had been positive using a highly sensitive species-specific nested PCR approach. 17 The sample preparation for HTS is prone to contamination due to multiple processing steps 62 and reagents like those used for DNA extraction. 63 We included PCR, DNA extraction and sample processing reagent controls which were examined alongside the samples they controlled for prior to commencing the main analysis.
Whilst choice of primers may influence the findings, 64 as could index hopping whereby incorrect barcode identifiers are incorporated into growing clusters, 65 we have mitigated some of these risks by utilizing a conservative quality-filtering approach 32 and focusing on the more abundant members of the mycobiota.

IgE sensitization to filamentous thermotolerant fungi such as
A. fumigatus is associated with poor asthma control, fixed airflow obstruction, bronchiectasis and various radiological abnormalities resulting in a pattern of asthma we have termed AFAD. 4,6 Whilst a number of clinical correlations were observed, a larger study utilizing a quantitative approach may be needed to fully address the clinical significance of the fungi detected. This was a cross-sectional study, and one of the next steps should be a longitudinal investigation to establish the stability of the lung mycobiota during stable state and exacerbations in order to interpret structural abnormalities that develop over decades.
In summary, we have shown that the airways are colonized by a community of fungi dominated by A. fumigatus and C. albicans, which are the main components of the normal airway mycobiota. This community is highly variable; however, notable shifts in the balance of fungi within communities and between compartments of the lung were associated with asthma status, asthma duration and biomarkers of inflammation. In particular, members of the A. niger and C. humicola species complexes were revealed by this molecular approach to be playing an unexpected potential role in asthma pathogenesis.