Abstract
Early allergic sensitisation (atopy) is the first step in the development of allergic diseases such as atopic asthma later in life. Genes and pathways associated with atopy and atopic asthma in children and adolescents have not been well characterised.
A transcriptome-wide association study (TWAS) of atopy and atopic asthma in white blood cells (WBCs) or whole blood was conducted in a cohort of 460 Puerto Ricans aged 9–20 years (EVA-PR study) and in a cohort of 250 Swedish adolescents (BAMSE study). Pathway enrichment and network analyses were conducted to further assess top findings, and classification models of atopy and atopic asthma were built using expression levels for the top differentially expressed genes (DEGs).
In a meta-analysis of the study cohorts, both previously implicated genes (e.g. IL5RA and IL1RL1) and genes not previously reported in TWASs (novel) were significantly associated with atopy and/or atopic asthma. Top novel genes for atopy included SIGLEC8 (p=8.07×10−13), SLC29A1 (p=7.07×10−12) and SMPD3 (p=1.48×10−11). Expression quantitative trait locus analyses identified multiple asthma-relevant genotype–expression pairs, such as rs2255888/ALOX15. Pathway enrichment analysis uncovered 16 significantly enriched pathways at adjusted p<0.01, including those relevant to T-helper cell type 1 (Th1) and Th2 immune responses. Classification models built using the top DEGs and a few demographic/parental history variables accurately differentiated subjects with atopic asthma from nonatopic control subjects (area under the curve 0.84).
We have identified genes and pathways for atopy and atopic asthma in children and adolescents, using transcriptome-wide data from WBCs and whole blood samples.
Abstract
In Puerto Rican and Swedish children, a TWAS identified 59 DEGs by atopy and 40 DEGs by atopic asthma, including SIGLEC8 and IL17RB. Transcriptomic data from WBCs could be used to develop predictive models of atopy/atopic asthma in prospective studies. http://ow.ly/kF6v30nYnYC
Introduction
Over the last few decades, the prevalence of allergic diseases has increased in the USA and worldwide. Sensitisation to allergens (atopy) in early life precedes the development of allergic diseases later in childhood, including food allergies, allergic rhinitis and atopic asthma [1]. Puerto Rican and Swedish children bear a high burden of atopy and asthma [1–3]. In studies of school-aged children, estimates of the prevalence of asthma and allergic rhinitis in Puerto Ricans were 16.1% and 48.4%, respectively [2, 4]. Among Swedish children aged 7–8 years, estimates of the prevalence of current wheeze and allergic rhinitis were 13% and 13.8%, respectively [5].
Recent meta-analyses of genome-wide association studies (GWASs) identified 18 susceptibility loci for asthma [6] and 41 susceptibility loci for allergic rhinitis [7]. However, such loci accounted for a small proportion of the heritability of asthma (∼3.5%) and allergic rhinitis (∼7.8%), and alternative approaches such as transcriptomics could help identify the “missing heritability” of atopic diseases. Yet another complementary approach is to study intermediate phenotypes such as atopy, as susceptibility genes for atopy often overlap with those for common allergic diseases [7].
In contrast to the many GWASs conducted to date, there have been few transcriptome-wide association studies (TWASs) of atopy and asthma, all limited by modest sample size and no or insufficient replication [8–12]. For example, a study of targeted RNA sequencing of 105 genes in nasal brushings from subjects with (n=50) and without (n=50) asthma showed that nasal transcriptomic profiles could identify subjects with IL13-driven asthma and T-helper cell type 2 (Th2)-skewed immune responses [9]. In another study (including mostly adults), RNA sequencing profiles from nasal brushings were used to first develop a panel of 90 genes to differentiate subjects with mild-to-moderate asthma (n=53) from control subjects (n=97), and then test the “classifier panel” in 40 subjects with (n=13) and without (n=27) asthma [10]. Although the panel performed well in training cohorts, external replication was limited and thus no individual genes could confidently be identified as conferring susceptibility to asthma.
While airway epithelial studies have shown promising results, blood samples are easy to collect and often available in epidemiological studies. A genome-wide analysis of transcriptomics in circulating CD19+ B-lymphocytes from 41 adults provided suggestive evidence of increased IL4R expression in subjects with dust mite allergy and asthma compared with control subjects [11]. In another study in young adolescents with asthma, microarray expression profiles from CD4+ T-lymphocytes were differentially expressed by atopic status [12] and five “atopic signature” genes were then nominally replicated in a cohort of 30 subjects. A later integrative study in asthma controls focused on the transcriptomic components in blood that vary with degree of asthma control [13].
Here, we report the findings from the first TWAS of atopy and atopic asthma in white blood cells (WBCs) from children and adolescents in Puerto Rico, with validation studies in a cohort of Swedish children.
Methods
Further details are given in the supplementary material.
Study population
The EVA-PR (Epigenetic Variation and Childhood Asthma in Puerto Ricans) study is a case–control study of asthma in Puerto Ricans aged 9–20 years, recruited using an approach similar to that used in previous studies in the metropolitan area of San Juan (Puerto Rico) [14, 15]. Of the 543 participating children, 460 had measurements of IgE to five common allergens in Puerto Rico (dust mite (Der p 1), cockroach (Bla g 2), cat dander (Fel d 1), dog dander (Can f 1) and mouse urinary protein (Mus m 1)) [16], WBC count and differential, and RNA from WBCs, and were thus included in the current analysis. The study was approved by the institutional review boards of the University of Puerto Rico (San Juan) and the University of Pittsburgh (Pittsburgh, PA, USA). Written parental consent and assent were obtained from all participants.
RNA sequencing and data pre-processing
Sequencing was performed using 350 ng of high-quality RNA extracted from WBCs after removing haemoglobin (RNA integrity number >7). Library preparation was done using the TruSeq Stranded Total RNA Library Prep Kit with the Ribo-Zero Gold High Throughput Kit (Illumina, San Diego, CA, USA), according to the manufacturer's protocol. Libraries were run on the NextSeq 500 using the NextSeq 500/550 High Output Kit version 2 (Illumina).
Quality control for raw RNA sequencing FASTQ files was performed using FastQC [17]. Low-quality reads and 3′ adapters were trimmed with Trim Galore! [18] and Cutadapt [19]. Trimmed reads were then aligned to reference human genome hg19 with STAR [20] and subsequently annotated in the Illumina iGenomes database using RSEM [21]. Samples with low alignment percentage and genes with low expression were removed from downstream analyses. After pre-processing, a total of 16 880 genes and 460 samples were retained in the final analysis.
Principal components were calculated from genome-wide genotypic data, which were obtained using the HumanOmni2.5 BeadChip platform (Illumina), as previously described [22]. Imputation of nongenotyped single nucleotide polymorphisms (SNPs) was performed with the Imputation Server [23], using the Haplotype Reference Consortium r1.1 2016 [24] as the reference panel.
The BAMSE study
We also analysed whole blood samples using Affymetrix Human Transcriptome Array 2.0 (Affymetrix, Santa Clara, CA, USA) in 269 16-year-old children selected from BAMSE (Children [Barn], Allergy, Milieu, Stockholm, Epidemiology), a Swedish population-based cohort (with oversampling of asthma and rhinitis cases) [25]. After quality control (RNA integrity number >6) and outlier detection, gene expression data from 250 blood samples were included in the current analysis of atopy and atopic asthma with available WBC count and differential. Allergen-specific IgE levels were measured using the ImmunoCAP System (Thermo Fisher/Phadia, Uppsala, Sweden).
Statistical analyses
In EVA-PR and BAMSE, atopy was defined as at least one positive IgE (≥0.35 IU·mL−1) to the allergens tested. Atopic asthma was defined as current asthma (physician-diagnosed asthma and current wheeze) and atopy in EVA-PR, and as current asthma (assessed using a modified version of the GA2LEN questionnaire2) and atopy in BAMSE [26].
Differentially expressed genes (DEGs) were analysed based on the raw count table using DESeq2 [27] for RNA sequencing data in EVA-PR and LIMMA for microarray data in BAMSE. In both EVA-PR and BAMSE, multivariable models of atopic asthma were adjusted for age, sex, batch label and the proportions (percentages) of WBC subtypes (eosinophils, lymphocytes, monocytes and neutrophils). All multivariable models of atopy were additionally adjusted for asthma status. In EVA-PR, all models were additionally adjusted for the first five principal components derived from genotypes, to account for population stratification. A zero-mean normal prior was put on the nonintercept coefficients to ensure that fold changes were independent of the choice of reference level. Next, we conducted a meta-analysis of the cohort-specific results using a fixed effect model with in-house code using the R package meta [28, 29]. The false discovery rate (FDR) approach was applied to adjust for multiple testing.
In EVA-PR, we performed an expression quantitative trait loci (eQTL) analysis (at a distance <1 Mb from gene start and end sites) for the top findings in the meta-analysis of atopy using the R package Matrix eQTL [30].
Pathway and network analysis
Ingenuity Pathway Analysis was performed for the top DEGs (FDR-adjusted p<0.1) in the meta-analysis of atopy and atopic asthma to identify enriched canonical pathways and gene networks [31]. We then calculated the partial correlation between gene expression levels in our discovery cohort, first in all subjects and then separately by asthma status, using the R package FastGGM [32], a powerful tool to detect conditional gene networks under the framework of the Gaussian graphical model.
Classification models of atopy and atopic asthma
Logistic regression with penalised maximum likelihood was applied to build classification models of atopy and atopic asthma in EVA-PR. We first built two models for comparison: one including only demographic and parental history variables (age, sex and parental history of asthma) as predictors, and another including demographic variables, parental history and gene expression levels of randomly chosen genes. Next, we built a model using the top DEGs in our meta-analysis, and another using the top DEGs plus demographic variables and parental history. Models were built separately for atopy and atopic asthma.
Results
The main characteristics of study participants are shown in table 1. In EVA-PR, subjects with atopy were significantly more likely to be male and to have asthma than control subjects. Compared with participants in EVA-PR, those in BAMSE had, as expected for a population-based cohort study, a lower proportion of asthmatic subjects and a narrower age range.
Characteristics of participants in the EVA-PR and BAMSE studies
TWAS of atopy
In the multivariable analysis of atopy, which was adjusted for WBC subtype and other covariates, we identified 69 DEGs in EVA-PR at FDR-adjusted p<0.05. In BAMSE, there were no DEGs at FDR-adjusted p<0.05, but 17 of the 69 DEGs in EVA-PR were differentially expressed at p<0.05, with similar estimates of effect size across the two studies (supplementary table S1). Next, to increase the statistical power of DEG analysis, we combined the results from the two cohorts in a transcriptome-wide meta-analysis, which identified 59 DEGs by atopy, at an FDR-adjusted p<0.05 (figure 1a and supplementary figure S1). Table 2 lists the top 20 genes (by p-value) in this meta-analysis, for which the direction of the association between gene expression and atopy was consistent across the two studies.
Meta-analysis of gene expression demonstrates substantial differences in gene expression between atopic subjects and control subjects. DEG: differentially expressed gene. a) Volcano plot showing the log2(fold change) and −log10(adjusted p-value) of all the genes in the meta-analysis. The statistically significant DEGs between atopic subjects and control subjects are represented by red (upregulated) and blue (downregulated) symbols. The statistical criterion for a gene to be considered differentially expressed was false discovery rate-adjusted p<0.05. b). Venn diagram depicting the distribution of the DEGs in the analysis between atopic subjects and control subjects, and between atopic asthmatic subjects and nonatopic nonasthmatic control subjects.
Top differentially expressed genes from the meta-analysis of atopy
For the top DEGs by atopy, we found 43 enriched pathways at FDR-adjusted p<0.05. Table 3 lists the top 20 enriched pathways, of which the two most significant ones were the Th2 pathway and the Th1/Th2 activation pathway. Other top pathways included those for nerve growth factor (NGF) and adenine/adenosine salvage I.
Top functional Ingenuity pathways enriched by the top differentially expressed genes in the meta-analysis of atopy
TWAS of atopic asthma
We then conducted a transcriptomic analysis of atopic asthma, finding substantial overlap between the results from this analysis and those for atopy (figure 1b and supplementary table S2). We identified 49 DEGs by atopic asthma in EVA-PR at FDR-adjusted p<0.05. In BAMSE, there were no DEGs at FDR-adjusted p<0.05, but five of the 49 DEGs by atopic asthma in EVA-PR were differentially expressed at p<0.05 in BAMSE, with similar estimates of effect size. 40 DEGs were then identified in the meta-analysis at FDR-adjusted p<0.05. For atopic asthma, a pathway enrichment analysis revealed pathways related to lipid metabolism (e.g. docosahexaenoic acid signalling) and arachidonic acid signalling (e.g. eicosanoids signalling) (supplementary table S3).
eQTL analysis
Next, we conducted a cis-eQTL analysis in EVA-PR, which identified 3772 SNPs that were significantly associated with our top genes from the expression analysis for atopy at FDR-adjusted p<0.05 (supplementary table S4). Such SNPs included multiple eQTLs in the GTEx database (e.g. rs2255888 in ALOX15 and rs35092096 in CCR3) [33]. Of interest, multiple SNPs were associated with expression of IL1RL1 (a known asthma-susceptibility gene) [6].
Network analysis
A network analysis was performed to illustrate interactions among our top DEGs by atopy (supplementary figure S2), first in all subjects and then separately in atopic subjects and in control subjects. Partial correlation among these genes remained strong, even after conditioning on global expression profile. This was particularly true for the gene most differentially expressed, SIGLEC8. We also found that some strong correlations in control subjects weakened significantly in atopic children, such as those for the gene pair IL34 and CYSLTR2.
We next performed hierarchical clustering with log-transformed transcripts per kilobase million values in EVA-PR, based on the 59 DEGs (FDR-adjusted p<0.05) in the meta-analysis of atopy, identifying four individual clusters (figure 2). In this analysis, the prevalence of atopy differs between the first two clusters, but not between the last two clusters (supplementary figure S3). However, when we compared eosinophil proportions in samples across these four clusters, we found that the samples with higher expression of the top DEGs for atopy had more eosinophils in peripheral blood (figure 3). We then tested for an association between expression levels of the top DEGs for atopy or atopic asthma and the measured proportions of each cell subset in peripheral blood (supplementary table S5). Of the top 59 DEGs for atopy, 51 genes were significantly associated (FDR-adjusted p<0.05) with the proportion of eosinophils in peripheral blood. However, a few of the top DEGs were associated with the proportions of other cell types (e.g. APRT and TMEM160 with lymphocyte proportions, and PEPD with neutrophil proportions). All top 40 DEGs for atopic asthma were significantly associated with eosinophils in peripheral blood. Next, we deconvolved the expression profile of WBCs based on known cell proportions and permutations in order to investigate cell type-specific differential expression using CIBERSORT [34]. 38 genes were significantly enriched in eosinophils (FDR-adjusted p<0.05). Of these 38 genes, 24 were among the top DEGs for atopy and 20 were among the top DEGs for atopic asthma.
Hierarchical clustering based on the top differentially expressed genes in the meta-analysis. Four clusters are identified, with most of the genes upregulated from cluster 1 to cluster 4.
Eosinophil proportion comparison for each cluster indentified in figure 2, showing significant difference both globally and between clusters individually in the Kruskal–Wallis test. Box-and-whisker plots show median with interquartile range (IQR) (boxes) and upper/lower limit within 1.5 IQR from the box range (whiskers); outliers are indicated by individual data points.
Classification models for atopy and atopic asthma
Finally, we built classification models in our discovery cohort using only demographic variables and parental history, achieving area under the curve (AUC) values of 0.56 for atopy and 0.74 for atopic asthma. These results were similar after adding expression levels for a random set of genes (data not shown). When using top DEGs alone as predictors, the models reached AUC values of 0.77 and 0.81 for atopy and atopic asthma, respectively. However, the performance of the classification models improved substantially after expression levels of the top DEGs, demographic variables and parental history were included as predictors, achieving AUC values of 0.77 for atopy and 0.84 for atopic asthma (figure 4). For atopy, we used 51 genes (FDR-adjusted p<0.05 in the meta-analysis for atopy and available in both cohorts), while for atopic asthma we included the top 33 genes (FDR-adjusted p<0.05 in the meta-analysis for atopic asthma and available in both cohorts).
Classification models of a) atopy and b) atopic asthma. H: age, sex and asthma history of parents as predictors; R: expression profile of randomly chosen genes (the same number as G) as predictors; G: expression profile of top differentially expressed gene predictors (false discovery rate-adjusted p<0.05 in the meta-analysis); AUC: area under the receiver operating characteristic curve.
Discussion
In a TWAS of WBCs, we identified DEGs by atopy and atopic asthma in 460 Puerto Rican children. In parallel, we conducted a TWAS of atopy and atopic asthma in 250 Swedish children, which yielded nonsignificant results due to the small sample size, but yielded replication of some of our top findings in Puerto Rican children at p<0.05, in the same direction of association and with similar estimates of effect size. Next, we conducted a meta-analysis of TWASs from both study cohorts, identifying 59 DEGs by atopy and 40 DEGs by atopic asthma.
In our meta-analysis, the top DEG by atopy was SIGLEC8, which has an extracellular binding domain that induces apoptosis in eosinophils and inhibits FcεRI-dependent mediator release in mast cells [35]. SNPs in SIGLEC8 were associated with asthma and eosinophilic oesophagitis in candidate gene studies [36, 37], but this gene was not associated with asthma in a recent meta-analysis of GWASs [6]. Of interest, expression of one of the top 20 genes in our meta-analysis of atopy (OLIG2, encoding oligodendrocyte transcription factor 2) was previously reported to downregulate RNA and protein levels of SIGLEC8 in peripheral blood eosinophils [38]. Moreover, expression levels of OLIG2 and CCL23 (one of our top DEGs for atopic asthma) in whole blood were significantly associated with suboptimal disease control in a meta-analysis of data from 693 ethnically diverse children and adults [13]. Both OLIG2 and CCL23 are TREM-1 (triggering receptor expressed in myeloid cells-1)/lipopolysaccharide signalling genes, and TREM-1 is a key amplifier of inflammatory responses induced by pattern recognition receptors.
Other top study results are supported by experimental evidence or findings from other human studies. SLC29A1 is involved in adenosine-regulated inhibition of IgE-dependent degranulation of mast cells [39]. IL5RA, a gene implicated in asthma in our prior study [40], regulates Th2 cell differentiation and Th2 cell effector functions [41]. Moreover, IL5RA and ALOX15 (also among our top genes) were both significantly associated with eosinophil count in a transcriptomic analysis of blood samples from subjects with asthma [36]. This is consistent with our prior findings for total IgE [42] and, together with the results from our hierarchical cluster analysis and other complementary analyses, further emphasises the key role of eosinophil-dependent mechanisms in the pathogenesis of atopy and atopic asthma.
Our findings for IL17RB are particularly interesting. In mouse models, allergen challenge triggers the interleukin (IL)-4 pathway, which promotes allergic airway inflammation and airway responsiveness. In turn, activation of IL-4 pathways induces expression of IL-25 and IL-17RB in serum and the lungs [43, 44]. In a prior study of human subjects with asthma, IL17RB expression in peripheral blood CD4+ T-cells was significantly associated with serum total IgE [45], albeit only in males. Consistent with a key role for IL17RB in atopy in humans, genetic deletion of IL17RB significantly lessens allergic sensitisation in mice [44].
We identified other plausible candidate genes for atopy and atopic asthma. SMPD3, encoding sphingomyelin phosphodiesterase 3 [46], is hypermethylated and downregulated in the small airways of subjects with chronic obstructive pulmonary disease [47]. PIK3R6, an IL-17A gene target in airway smooth muscle cells [48], has been shown to interact with second-hand smoke exposure on forced expiratory volume in 1 s. Other genes, such as ADAMTS7P1, EPN2 and RAB44, have not been previously reported in TWASs and merit further investigation.
DEGs by atopy often overlapped with those for atopic asthma. Most of our top findings were correlated with genotypic data, emphasising a genetic predisposition to atopy. Moreover, a network analysis based on partial correlation revealed close interactions between significant DEGs.
Consistent with known mechanisms of atopy, we found that the top two most enriched pathways were the Th2 and Th1/Th2 activation pathways. Other top pathways in our analysis were that for NGF (a critical factor in the pathogenesis of allergic airway inflammation and proliferation of airway smooth muscle cells) [49] and that for adenine/adenosine salvage I (adenosine has been contraindicated in allergy or severe asthma) [50].
We compared our findings for gene expression and atopic asthma in WBCs with those of a prior meta-analysis of gene expression in airway epithelium and asthma [51]. Of the top 40 DEGs by atopic asthma in our study, seven were differentially expressed by asthma in airway epithelium (supplementary table S6). Some of these differences may be due to inclusion of cases of both atopic and nonatopic asthma in the prior analysis in airway epithelium, while others may indicate true tissue-specific differences in gene expression in atopic asthma. We also compared our results with those of a prior European study of sputum transcriptomics in 104 subjects with moderate-to-severe asthma and 16 healthy volunteers [52]. In that study, three transcriptome-associated clusters were identified, with one of them enriched for the IL-13/Th2 pathway. Of the 20 genes comprising this cluster, seven were also associated with atopy or atopic asthma (FDR-adjusted p<0.05) in our meta-analysis (PRSS33, CLC, LGALS12, ALOX15, OLIG2, HRH4 and CCR3).
We mapped the top DEGs in our meta-analysis to the top CpG sites discovered in epigenome-wide association studies (EWASs) of atopy and atopic asthma in blood and airway epithelial samples. None of the genes containing or near the top 30 CpG sites recently associated with atopy in an EWAS of nasal epithelium [53] were DEGs by atopy in our study, but two of the top DEGs by atopic asthma in our study (IL5RA and SIGLEC8) were part of differentially methylated regions by asthma in an EWAS of whole blood [54, 55].
Results from our classification models further support the validity of our top findings, as well as the plausibility of developing predictive transcriptomic panels of atopy and atopic asthma, using WBCs, in future longitudinal studies. Although classification models of atopic asthma using transcriptomic data from airway epithelium in adults have shown greater accuracy (AUC 0.99) [10] than that shown in the current study (AUC 0.84), WBCs are easier to collect than nasal or bronchial epithelium [10].
There were differences in both the type of sample (WBCs versus whole blood) and the platform used to assess gene expression (RNA sequencing versus microarray) between EVA-PR and BAMSE. This, together with differences in ethnicity and sample size between the two studies, likely explains lack of significant replication of our findings in EVA-PR in BAMSE. However, a previous study of the transcriptomics of atopic dermatitis found generally good correlation between the results of RNA sequencing and microarrays, except for genes with low expression levels [56]. Moreover, we found nominal replication at p<0.05, with consistent direction and estimated effect size, for some of our top findings in EVA-PR in BAMSE. Although there were no significant DEGs for atopy or atopic asthma in BAMSE after FDR correction for multiple testing, the top genes associated with atopy or atopic asthma in BAMSE (shown in supplementary table S7) were enriched in pathways for energy-coupled proton and ion transmembrane transport.
In summary, our combined transcriptomic analysis in Puerto Rican and Swedish children identified 59 DEGs by atopy and 40 DEGs by atopic asthma, including SIGLEC8 and IL17RB. Transcriptomic data from WBCs could be used to develop predictive models of atopy and atopic asthma in future prospective studies.
Supplementary material
Supplementary Material
Please note: supplementary material is not edited by the Editorial Office, and is uploaded as it has been supplied by the author.
Supplementary material ERJ-00102-2019.Supplement
Acknowledgements
We thank all the participating children and families. The BAMSE team would like to thank Charles Auffray (European Institute for Systems Biology and Medicine, Lyon, France) and his team for generating the transcriptomics data.
Footnotes
This article has supplementary material available from erj.ersjournals.com
Author contributions: W. Chen, E. Melén and J.C. Celedón conceived and designed the study. Y. Jiang, O. Gruzieva and S.K. Merid conducted the primary analysis and interpreted data. E. Forno, T. Wang, E. Acosta-Pérez, G. Canino, I. Kull, J.M. Antó and J. Bousquet participated in data collection and data analysis. Y. Jiang, W. Chen and J.C. Celedón prepared the first draft of the manuscript. All authors reviewed the draft for intellectual content and approved submission of the final version of the manuscript.
Conflict of interest: Y. Jiang has nothing to disclose.
Conflict of interest: O. Gruzieva has nothing to disclose.
Conflict of interest: T. Wang has nothing to disclose.
Conflict of interest: E. Forno has nothing to disclose.
Conflict of interest: N. Boutaoui has nothing to disclose.
Conflict of interest: T. Sun has nothing to disclose.
Conflict of interest: S.K. Merid has nothing to disclose.
Conflict of interest: E. Acosta-Pérez has nothing to disclose.
Conflict of interest: I. Kull has nothing to disclose.
Conflict of interest: G. Canino has nothing to disclose.
Conflict of interest: J.M. Antó has nothing to disclose.
Conflict of interest: J. Bousquet reports personal fees for advisory board work, consultancy and lectures from Chiesi, Cipla, Hikma, Menarini, Mundipharma, Mylan, Novartis, Sanofi-Aventis, Takeda, Teva and Uriach, and owns shares in Kyomed, outside the submitted work.
Conflict of interest: E. Melén has nothing to disclose.
Conflict of interest: W. Chen has nothing to disclose.
Conflict of interest: J.C. Celedón has received research materials from GSK, Merck and Pharmavite, outside the submitted work.
Support statement: This study was supported by grants HL079966, HL117191 and MD011764 (Principal Investigator: J.C. Celedón) from the US National Institutes of Health (NIH). E. Forno's contribution was supported by NIH grant HL125666. Y. Jiang was supported by the China Scholarship Council. The BAMSE cohort was supported by grants from the Swedish Heart-Lung Foundation, the Swedish Research Council, Stockholm County Council (ALF), the Strategic Research Program (SFO) in Epidemiology at Karolinska Institutet and MeDALL (Mechanisms of the Development of ALLergy; European Union grant agreement number 261357). E. Melén is supported by a grant from the European Research Council (grant agreement number 757919, TRIBAL). Funding information for this article has been deposited with the Crossref Funder Registry.
- Received January 15, 2019.
- Accepted March 2, 2019.
- Copyright ©ERS 2019
References
