Circulating microRNAs as biomarkers in patients with allergic rhinitis and asthma

Circulating microRNAs as biomarkers in patients with allergic rhinitis and asthma

Circulating microRNAs as biomarkers in patients with allergic rhinitis and asthma Ronaldo P. Panganiban, BS,a,d Yanli Wang, BS,d,e Judie Howrylak, MD,...

1MB Sizes 0 Downloads 23 Views

Circulating microRNAs as biomarkers in patients with allergic rhinitis and asthma Ronaldo P. Panganiban, BS,a,d Yanli Wang, BS,d,e Judie Howrylak, MD, PhD,a Vernon M. Chinchilli, PhD,b Timothy J. Craig, DO,a Avery August, PhD,c and Faoud T. Ishmael, MD, PhDa,d Hershey and University Park, Pa, and Ithaca, NY Background: MicroRNAs (miRNAs) are emerging as important regulatory molecules that might be involved in the pathogenesis of various diseases. Circulating miRNAs might be noninvasive biomarkers to diagnose and characterize asthma and allergic rhinitis (AR). Objective: We sought to determine whether miRNAs are differentially expressed in the blood of asthmatic patients compared with those in the blood of nonasthmatic patients with AR and nonallergic nonasthmatic subjects. Furthermore, we sought to establish whether miRNAs could be used to characterize or subtype asthmatic patients. Methods: Expression of plasma miRNAs was measured by using real-time quantitative PCR in 35 asthmatic patients, 25 nonasthmatic patients with AR, and 19 nonallergic nonasthmatic subjects. Differentially expressed miRNAs were identified by using Kruskal-Wallis 1-way ANOVA with Bonferroni P value adjustment to correct for multiple comparisons. A random forest classification algorithm combined with a leave-one-out cross-validation approach was implemented to assess the predictive capacities of the profiled miRNAs. Results: We identified 30 miRNAs that were differentially expressed among healthy, allergic, and asthmatic subjects.

From athe Department of Medicine, Division of Pulmonary, Allergy, and Critical Care Medicine, Pennsylvania State University Milton S. Hershey Medical Center, Hershey; b the Department of Public Health Sciences and dthe Department of Biochemistry and Molecular Biology, Pennsylvania State University College of Medicine, Hershey; cthe Department of Microbiology and Immunology, Cornell University, Ithaca; and ethe Bioinformatics and Genomic Program, Pennsylvania State University, University Park. Supported by a Doris Duke Charitable Foundation Clinical Scientist Development Award and National Institutes of Health 1K08HL114100-1. Disclosure of potential conflict of interest: T. J. Craig is an unpaid American Academy of Allergy, Asthma & Immunology Interest Section Leader; is an unpaid board member for the American College of Allergy, Asthma & Immunology, American Lung Association of Pennsylvania, and the Joint Council of Allergy, Asthma & Immunology; has received consultancy fees from CSL Behring, Dyax, Viropharma, Shire, Merck, Biocryst, and Bellrose; has received research support from Viropharma, CSL Behring, Shire, Dyax, Pharming, Merck, Genentech, GlaxoSmithkline, Grifols, Novartis, Sanofi Aventis, and Boehringer Ingelheim; has received lecture fees from CSL Behring, Dyax, Shire, and Grifols; and is coinvestigator for Asthmanet, National Heart, Lung, and Blood Institute. A. August has received research support from the National Institutes of Health (NIH), is employed by Cornell University, and is presenting a research seminar from Biogen. F. T. Ishmael has received research support from Doris Duke Charitable Foundation and the NIH. The rest of the authors declare that they have no relevant conflicts of interest. Received for publication February 8, 2015; revised November 20, 2015; accepted for publication January 8, 2016. Corresponding author: Faoud T. Ishmael, MD, PhD, Pennsylvania State University College of Medicine, 500 University Dr, H171, Hershey, PA 17033. E-mail: [email protected]. 0091-6749/$36.00 Ó 2016 American Academy of Allergy, Asthma & Immunology http://dx.doi.org/10.1016/j.jaci.2016.01.029

These miRNAs fit into 5 different expression pattern groups. Among asthmatic patients, miRNA expression profiles identified 2 subtypes that differed by high or low peripheral eosinophil levels. Circulating miR-125b, miR-16, miR-299-5p, miR-126, miR-206, and miR-133b levels were most predictive of allergic and asthmatic status. Conclusions: Subsets of circulating miRNAs are uniquely expressed in patients with AR and asthmatic patients and have potential for use as noninvasive biomarkers to diagnose and characterize these diseases. (J Allergy Clin Immunol 2016;nnn:nnn-nnn.) Key words: Asthma, allergic rhinitis, microRNA, biomarker, plasma, inflammation, posttranscriptional regulation

Asthma is a heterogeneous disease comprised of numerous phenotypes that are difficult to characterize with current diagnostic tools. It involves a complex interplay of the airway epithelium, innate immune system, and adaptive immunity that is still not completely understood.1 There is a clear need for identification of noninvasive biomarkers to diagnose, characterize, and understand this disease. The aim of this study was to determine whether circulating microRNAs (miRNAs) are differentially expressed in asthmatic patients compared with those in healthy control subjects and patients with allergic rhinitis (AR) and whether their expression could be used as a tool to further characterize asthma. miRNAs are emerging as noninvasive biomarkers that play important roles in cytokine regulation and asthma pathogenesis. miRNAs are short (20-25 nucleotides), single-stranded, noncoding RNAs that posttranscriptionally regulate gene expression through interactions with mRNAs. They associate with members of the Argonaute family of proteins and form the central component of RNA-induced silencing complex.2 Binding of RNA-induced silencing complex (RISC) miRNA to their target mRNA transcripts, usually in the 39 untranslated region, leads to downregulation of gene expression through destabilization of mRNA stability or translational repression.2 In some cases the miRNA-Argonaute complex, possibly in association with a distinct set of regulatory proteins, can enhance gene expression.3-5 miRNAs bind to their targets with partial complementarity, such that any miRNA is capable of binding hundreds or even thousands of targets. Because miRNAs can regulate functionally related genes, it is possible that a few miRNAs or even a single miRNA could regulate entire pathways. miRNAs are produced by a wide variety of cells in different organs and secreted into the blood and other bodily fluids, where they can exert biological effects.6,7 This also allows them to serve as noninvasive biomarkers. It is now evident that miRNAs play significant roles in diverse disease processes.8-10 Their expression 1

2 PANGANIBAN ET AL

J ALLERGY CLIN IMMUNOL nnn 2016

Isolation and characterization of miRNAs Abbreviations used AR: Allergic rhinitis Ct: Cycle threshold KEGG: Kyoto Encyclopedia of Genes and Genomes MAPK: Mitogen-activated protein kinase miRNA: MicroRNA NF-kB: Nuclear factor kB PCA: Principal component analysis qPCR: Real-time quantitative PCR

is dysregulated in a number of diseases, and they can play roles in disease pathogenesis. Because approximately 150 miRNAs are detectable in blood, differential expression patterns can serve as a molecular fingerprint to diagnose and characterize diseases.11 Indeed, multiple studies have already demonstrated their utility as diagnostic or prognostic biomarkers. In diseases such as cancer, circulating miRNAs can diagnose disease, characterize disease biology, predict response to different treatments, and serve as a target for novel therapeutics.12 Circulating miRNAs are poorly studied in patients with AR and asthma but have great potential to diagnose and characterize these diseases. We have previously shown that miRNAs isolated from lungs and blood have utility as biomarkers in asthmatic patients and that differentially expressed miRNAs might be important regulators of TH2 cytokines.5,13 In this study we sought to determine whether plasma miRNAs are differentially expressed in patients with AR and asthma to establish whether they can be used as a tool to characterize asthma subtypes and to identify candidate miRNAs that might play roles in disease pathogenesis. We identified 30 miRNAs that are differentially expressed in the plasma of asthmatic patients, patients with AR, and nonallergic nonasthmatic subjects. These miRNAs can be classified into 5 groups that correlated with different patterns of expression in patients with AR, asthma, or both. Bioinformatic analyses revealed that the differentially expressed miRNAs targeted genes involved in inflammatory pathways. Analysis of miRNA expression in asthmatic patients revealed 2 main clusters that differed in peripheral eosinophil levels. Using random forest classification,14 we were able to implement a prediction model that is 91.1% accurate in predicting AR or asthmatic status. These findings indicate that plasma miRNAs could play roles in AR and asthma and might have potential as biomarkers. The implications of these findings are described herein.

METHODS Patient selection The study was approved by the Penn State College of Medicine Institutional Review Board. All participants provided written informed consent. Patients were classified as asthmatic based on history and lung function, including FEV1 reversible by greater than 12% and greater than 200 mL after bronchodilator or airway hyperresponsiveness caused by methacholine (provocative concentration producing a 20% decrease in FEV1 of less than 8 mg/mL). Patients were considered allergic if they had a clinical history of aeroallergen sensitivity and at least 1 positive skin test response (3 mm larger than that elicited by the negative control) in a standard panel of 19 relevant aeroallergens and were considered nonallergic if the skin test panel result was negative. Asthmatic patients were asked to answer the original 7-item Asthma Control Questionnaire.15

Blood was collected by means of venipuncture in a purple-top tube and then centrifuged at 3000 rpm in a clinical centrifuge to isolate plasma. For isolation of total RNA, 2 mL of 50 nmol/L synthetic cel-miR-39 was added as a ‘‘spike-in’’ normalization control to 500 mL of plasma.16 Afterward, 1.5 mL of TRIzol (Life Technologies, Waltham, Mass) reagent was added, and total RNA was extracted according to the manufacturer’s protocol. RNA concentration was measured based on A260/280 with a NanoDrop Lite Spectrophotometer (Thermo Scientific, Waltham, Mass). Expression of 420 miRNAs in plasma was screened with the Human miRNome v15 PCR array (System Biosciences, Palo Alto, Calif). In short, 400 ng of total RNA was reverse transcribed with the QuantiMir RT Kit (Systems Biosciences), and expression of miRNAs was measured by using the CFX384 Real-Time System (Bio-Rad Laboratories, Hercules, Calif). Each reaction was run in triplicate, and the average cycle threshold (Ct) value was used for analysis. For subsequent analysis by means of quantitative real-time quantitative PCR (qPCR), 250 ng of total RNA was reverse transcribed to cDNA with the qScript microRNA cDNA Synthesis Kit (Quanta BioSciences, Gaithersburg, Md). Quantification of miRNAs with qPCR was performed on the CFX384 Real-Time System with specific primers to miRNAs of interest (250 nmol/L, see Table E1 in this article’s Online Repository at www. jacionline.org), 1 mL of diluted cDNA (diluted 1:10), and the iTaq Universal SYBR Green Supermix (Bio-Rad Laboratories) in a total volume of 10 mL. Primers to each miRNA was synthesized by Integrated DNA Technologies (Coralville, Iowa) and placed in a unique position on 96-well plates at a stock concentration of 1 mmol/L. A multichannel EDP-3 Plus electronic pipette (Bio-Rad Laboratories) was used to transfer the primer to a 384-well plate, where they were mixed with cDNA, SYBR Green mix, and universal primer. Each sample was run in triplicate. A 2-step program was used as follows: 40 cycles of 958C for 10 seconds and 608C for 30 seconds. Sample Ct values were normalized to cel-miR-39 to control for variability in RNA isolation and reverse transcription and then to total RNA expression as a means of normalizing expression data.16 To calculate copy numbers, real-time PCR amplification of multiple dilutions of known concentrations of synthetically synthesized miRNAs (cel-miR-39, miR-155, and Let-7a) was measured. An average of these 3 curves was used to generate a standard curve that could be used to calculate concentrations of unknowns as copy number per microliter.

Statistical analyses Normally distributed data were analyzed by using 1-way ANOVA with the Holm adjustment for pairwise comparisons or the Student t test, where appropriate. Fisher exact tests were used for categorical binary variables.

Analysis of differentially expressed miRNAs Hierarchical cluster analysis was performed in Cluster3.0 by using the average-linkage method.17 Principal component analysis (PCA) and expression analysis were performed in R 3.2.0/Bioconductor.18,19 The Wilcoxon rank sum test or Kruskal-Wallis 1-way ANOVA with the Nemenyi post hoc test was used for nonnormally distributed data.20 To control for multiple testing, we used the Benjamini and Hochberg false discovery rate or the Bonferroni procedure.21 When classifying miRNAs based on expression patterns, miRNAs are considered upregulated or downregulated when their median expression levels show at least 2-fold change between groups and demonstrate adjusted pairwise comparison P values of less than .05.

Bioinformatics analysis of pathways targeted by differentially expressed miRNAs DIANA-miRPath v2.0 Web Service (http://diana.imis.athena-innovation. gr/DianaTools/index.php?r5mirpath) was used to identify the Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways of genes targeted by each group of miRNAs.22 A network consisting of the top 20 pathways and the 30 differentially expressed miRNAs was constructed by using Cytoscape 3.2.1.23 Cytoscape 3.2.1 was also used to create a network that

J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn

was comprised of inflammation-associated genes and the differentially expressed miRNAs that target these genes.

Evaluation of miRNAs as biomarkers for asthma and AR Random forest, as implemented by using the Python machine learning package scikit-learn,24 was used to build a prediction model for asthma and AR based on the 39 profiled miRNAs and 5 available demographic characteristics (see Tables E4 and E5 in this article’s Online Repository at www.jacionline.org). The accuracy and other performance measures of our predictive algorithm was determined with a leave-one-out cross-validation, and the randomness of sample bootstrapping in random forest was controlled by taking the average of 10 runs.

RESULTS Selection of candidate plasma miRNAs for study Our overall goal was to identify miRNA candidates in plasma that might be dysregulated in patients with AR and those with asthma. First, we sought to determine which miRNAs could be readily and reproducibly detected in plasma by using our real-time PCR array methodology. High-throughput profiling of 420 miRNAs was performed on plasma isolated from 5 asthmatic patients and 5 nonasthmatic subjects (see Table E6 in this article’s Online Repository at www.jacionline.org). We found that 135 miRNAs were reproducibly detected in plasma (based on Ct values of between 20 and 33 in all samples, melting curve analysis that showed a single product, and PCR amplification efficiency of 2.0 6 0.1), which is consistent with other reported studies that have confirmed the expression of the majority of these miRNAs in blood.25,26 We then measured the expression of these 135 miRNAs in the plasma of 12 allergic asthmatic patients and 12 nonallergic nonasthmatic subjects as a screen to identify candidate miRNAs that can be differentially expressed in patients with AR and those with asthma (see Table E7 in this article’s Online Repository at www.jacionline.org). There were 30 miRNAs that were found to be differentially expressed, with a greater than 2-fold difference between groups, a Wilcoxon rank sum test significance cutoff of P value of less than .05, and a false discovery rate of 10% (see Fig E1 in this article’s Online Repository at www.jacionline.org). In addition to these miRNAs, we selected 9 additional miRNAs for further study that did not meet the criteria for significance. These included miRNAs that were not significantly different between the 2 groups that could be used as internal controls for normalization, as well as candidates that we previously identified as being differentially expressed in exhaled breath condensates in asthmatic patients.13 We assessed the quality of the qPCR data for these candidates using multiple methodologies. The presence of a single PCR product was confirmed by means of melting curve analysis and gel electrophoresis (see Fig E2 in this article’s Online Repository at www.jacionline.org), the latter of which showed a single product between 50 and 100 nucleotides in length (the expected size of the mature miRNA plus a poly[A] tail and 39 adapter sequence, as described in the Methods section). We also confirmed that our miRNAs of interest had high amplification efficiency (90% to 105%) by using analysis with LinRegPCR.27 To identify miRNAs that might be associated with AR, asthma, or both, the qPCR expression of these 39 miRNAs was subsequently profiled in a larger number of subjects (n 5 79),

PANGANIBAN ET AL 3

which included 35 asthmatic patients (29 of whom had AR), 25 nonasthmatic patients with AR, and 19 nonallergic nonasthmatic (healthy) subjects (see Table E8 in this article’s Online Repository at www.jacionline.org). Subjects’ characteristics are presented in Table I. Unsupervised hierarchical clustering of miRNA expression data revealed the formation of 3 major groups that generally segregated based on the patients’ disease status (Fig 1, A). Expression levels of 30 miRNAs were found to be statistically different among the 3 groups. These differentially expressed miRNAs could be classified into 5 expression pattern groups based on differences in expression between the asthma and AR, AR and healthy, and asthma and healthy groups (Fig 1, B; Table II; and see Fig E3 in this article’s Online Repository at www.jacionline.org; miR-937 is shown as a representative nondifferentially expressed miRNA).20 In group 1 miRNA expression was statistically different between healthy subjects and those with AR, healthy subjects and asthmatic patients, and patients with AR and those with asthma. In these cases the magnitude of change in expression levels of miRNA was greater in asthmatic patients versus healthy control subjects compared with those in patients with AR versus healthy control subjects (Fig 1, B; miR-125b is shown as a representative miRNA). The majority of miRNAs showed increased expression levels in allergic and asthmatic subjects versus healthy control subjects (miR-125b, miR-126, miR-21, let-7b, let-7c, and let-7e), except for miR-1, levels of which were decreased in both groups. miRNAs in group 2 demonstrated expression differences that were unique to asthmatic patients. There was a significant difference in their expression in asthmatic patients versus healthy subjects and asthmatic patients versus patients with AR, but there was no difference in the AR versus healthy groups (Fig 1, B; miR-16 panel). Among asthmatic patients, miR-16, miR-223, miR-148a, and miR-146a were upregulated, and miR-299-5p, miR-570, and miR-150 were downregulated. Group 3 miRNAs exhibited median expression levels that were similar between the AR and asthmatic cohorts but were either downregulated or upregulated compared with those in healthy subjects (Fig 1, B; miR-133b panel). Both miR-145 and miR-422 demonstrated increased plasma levels, whereas miR-133b, miR-133a, miR-26b, miR-1248, miR-330-5p, miR-29, miR-1291, and miR-144 showed decreased plasma levels. miRNAs in the group 4 expression pattern showed similar median expression levels between the healthy and asthmatic cohorts but are either downregulated or upregulated among patients with AR (Fig 1, B; miR-206 panel). In the AR group expression of both circulating miR-206 and miR-328 was upregulated, whereas expression of miR-338-3p and miR-26a were both downregulated. Finally, included in the group 5 expression pattern are miR-106a and miR-155. Both miRNAs were significantly downregulated in patients with AR and asthmatic patients compared with that in healthy subjects; however, their downregulation was more exaggerated among patients with AR compared with asthmatic patients (Fig 1, B; miR-155 panel).

Identification of KEGG pathways regulated by different groups of miRNAs Bioinformatic analyses with DIANA miRPath were used to analyze potential pathways and genes regulated by miRNAs in the

4 PANGANIBAN ET AL

J ALLERGY CLIN IMMUNOL nnn 2016

TABLE I. Subjects’ characteristics Characteristics

Age (y) Sex Male/female, no./no. (%/%) Ethnicity White/nonwhite, no./no. (%/%) BMI (kg/m2) Smoker, no. (%) AR, no. (%) Spirometry FVC (L) FVC (% predicted) FEV1 (L) FEV1 (% predicted) FEV1/FVC ratio

Healthy subjects (n 5 19), no. (%) or mean 6 SEM

Allergic patients (n 5 25), no. (%) or mean 6 SEM

Asthmatic patients (n 5 35), no. (%) or mean 6 SEM

41.5 6 3.25

39.8 6 3.12

43.9 6 2.47

.563*

8/11 (42/58)

11/14 (44/56)

14/21 (40/60)

.957 

16/3 (85/15) 27.7 6 1.64 5 (26) 0 (0)

21/4 (84/16) 28.3 6 1.15 2 (0.08) 25 (100)

34/1 (97/3) 29.2 6 1.09 5 (14) 29 (83)

.129  .701* .268  <.001 

6 6 6 6 6

.126* .081* .036* .008* .041*

3.94 93.3 3.05 87.3 0.76

6 6 6 6 6

0.34 4.55 0.29 4.94 0.02

4.53 102.1 3.64 98.5 0.79

6 6 6 6 6

0.32 3.83 0.29 3.62 0.02

3.70 89.9 2.68 78.4 0.70

0.19 2.84 0.18 3.85 0.02

P value

FVC, Forced vital capacity. *Normally distributed continuous variables were analyzed by using 1-way ANOVA.  Freeman-Halton extension of the Fisher exact test was used to analyze categorical binary data.

5 groups. A network map was generated to identify potential connections between miRNA groups and regulatory pathways and to determine whether there were functional connections among the 5 groups of miRNAs (see Fig E4 in this article’s Online Repository at www.jacionline.org). The complete list of KEGG pathways regulated by each group of miRNAs and the specific gene targets within the pathways for each miRNA group are shown in the supplemental data (see Table E2 in this article’s Online Repository at www.jacionline.org). The top 3 KEGG pathways identified (based on number of genes regulated by miRNAs) were ‘‘PI3K-Akt signaling pathways,’’ ‘‘Pathways in cancers,’’ and ‘‘MAPK signaling pathways,’’ such that miRNAs in each of the 5 groups were predicted to regulate multiple genes in each of these pathways (see Fig E4 in this article’s Online Repository at www.jacionline.org). All 5 groups of miRNAs targeted multiple genes within these pathways (see Fig E4). These genes included important inflammatory mediators, such as nuclear factor kB (NF-kB), IL-8, signal transducer and activator of transcription, activator protein 1, mitogen-activated protein kinase (MAPK) signaling molecules, and the TGF-b receptor (Fig 2). In some cases multiple miRNAs within a group targeted a single gene (eg, TGFBR1 as a target of group 1 miRNAs). There were also cases in which multiple miRNAs within each group targeted multiple genes within pathways, as exemplified by group 2, miRNAs of which were predicted to regulate multiple components of the NF-kB pathway (Fig 2).

Identification of asthma subgroups based on miRNA expression We hypothesized that miRNA expression could be used as a phenotypic tool to identify subsets of asthmatic patients. Using the 39 candidate miRNAs, we performed unsupervised cluster analyses of miRNA expression in the asthmatic group and identified 2 main clusters (cluster 1, n 5 16; cluster 2, n 5 19; Fig 3, A). PCA was applied to reduce the dimensionality of the data set by constructing linear combinations of variables and focusing on the most relevant linear combinations. PCA confirmed the formation of 2 major clusters (see Fig E5, B, in this article’s Online Repository at www.jacionline.org). The first

principal component accounted for 25.1% of the variance in the data, whereas the second principal component accounted for 17.2% of the variance. There were 20 miRNAs differentially expressed between the 2 clusters (see Table E3 and Fig E5, A, in this article’s Online Repository at www.jacionline.org). Cluster 1 contained a higher level of blood eosinophils relative to the second cluster (mean 6 SEM, 285.4 6 44.4 vs 133.3 6 23.4; Fig 3, B). All other characteristics, including demographics, lung function, AR status, aeroallergen sensitivity, Asthma Control Questionnaire scores, and medication use, were similar in the 2 groups (Table III).

Evaluation of miRNAs as biomarkers for asthma and AR We next sought to determine whether miRNA expression could be used diagnostically in patients with AR and asthma with supervised machine learning classification. As a multiclass classification problem, 44 categories of patient information, including levels of 39 miRNAs and 5 demographic characteristics, were used as features (see Table E4) for the 79 subjects to predict their disease status (healthy, AR, or asthma). Using a random forest model that constructed 100 decision trees, we were able to obtain the relevance of all features (see Table E4). Next, we performed manual feature selection to determine the most important miRNA in predicting disease status. The top 6 most relevant miRNAs (miR-125b, miR-16, miR-299-5p, miR-126, miR-206, and miR-133b) were determined to produce the model that was most accurate and resistant to overfitting (see Fig E6, B, in this article’s Online Repository at www.jacionline.org). We also confirmed that the minimum number of decision trees for the random forest required to produce the most robust classification model was 100 (see Fig E6, C). Using our most optimal model of random forest with the 6 most relevant miRNA features and 100 decision trees, we were able to correctly determine whether our subject was healthy, had AR, or was asthmatic 92.4% of the time (73/79 correct predictions; Fig 4, A). Taken as healthy versus diseased (has AR or is asthmatic), our model has high negative predictive value.

PANGANIBAN ET AL 5

J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn

A

B

miR-125b

miR-16

miR-133b

1e+09

Copy Number/μl

1e+06

1e+03 Disease Status Healthy miR-206

miR-155

miR-937

1e+09

AR Asthmatic

1e+06

1e+03

Healthy AR Asthmatic Healthy AR Asthmatic Healthy AR Asthmatic

Disease Status FIG 1. miRNA expression in healthy, allergic, and asthmatic subjects. A, Heat map showing expression of 39 candidate miRNAs in 19 healthy, 25 allergic, and 35 asthmatic subjects after 2-way unsupervised hierarchical clustering. B, Representative box plots showing differential expression patterns of miRNAs among healthy, allergic, and asthmatic cohorts. miR-937 is shown as a nondifferentially expressed miRNA.

Additionally, our model demonstrates high positive predictive value, with low false-positive rates across all disease statuses (Fig 4, B).

DISCUSSION Our study is the first to demonstrate that circulating miRNAs are differentially expressed among subjects who are healthy or have asthma or AR. Specifically, we show that 30 miRNAs that we classified into 5 expression groups are differentially expressed among these cohorts. Differential plasma miRNA expression in patients with asthma and AR is unlikely to be a mere epiphenomenon of these diseases. miRNAs have been shown to directly or indirectly affect the expression of multiple genes involved in the inflammatory response.28,29

AR is an established risk factor for asthma in both children and adults.30,31 Some studies have also shown that longer duration and increased severity of AR correlate with a higher prevalence of asthma.32 These clinical observations posit AR as an intermediate inflammatory phenotype between the healthy and asthmatic states. Group 1 miRNAs exhibit changes in miRNA expression among patients with AR compared with healthy control subjects, and the magnitude of these changes is augmented by asthma. Expression of miRNAs in this group could represent a continuum of airway inflammation with increased dysregulation because the asthmatic inflammatory changes are added to upper airway inflammation. Moreover, group 1 miRNAs could also represent a profile of TH2 airway inflammation because the cluster of asthmatic patients with high eosinophil counts (cluster 1) demonstrated similar expression patterns, with increased expression of

6 PANGANIBAN ET AL

J ALLERGY CLIN IMMUNOL nnn 2016

TABLE II. Differentially expressed miRNAs among the healthy, allergic, and asthmatic cohorts Median miRNA copy number/mL miRNA

Group 1 miR-125b miR-126 miR-21 Let-7b Let-7c miR-1 Let-7e Group 2 miR-16 miR-299-5p miR-223 miR-570 miR-148a miR-146a miR-150 Group 3 miR-133b miR-133a miR-26b miR-1248 miR-330-5p miR-145 miR-29 miR-422 miR-1291 miR-144 Group 4 miR-206 miR-338-3p miR-328 miR-26a Group 5 miR-106a miR-155

Healthy subjects

Allergic patients

Asthmatic patients

P value*

H:ARy

H:Ay

AR:Ay

5.09E103 1.88E104 2.81E105 2.31E105 2.90E105 1.17E106 2.66E105

1.73E105 3.44E105 1.99E106 2.92E106 1.55E106 4.12E104 7.58E105

1.10E106 6.87E106 3.01E107 2.17E107 6.31E106 7.48E102 1.67E106

<.001 <.001 <.001 <.001 <.001 <.001 <.001

.003 .011 .008 <.001 .002 .030 .013

<.001 <.001 <.001 <.001 <.001 <.001 <.001

<.001 <.001 <.001 .003 .003 .001 .021

4.58E105 2.71E105 2.59E106 6.35E104 4.53E105 4.24E105 1.13E105

4.41E105 1.93E106 1.78E107 8.96E104 2.91E105 5.82E105 9.28E105

4.97E107 3.17E104 1.24E108 1.06E104 1.43E106 4.73E106 4.66E106

<.001 <.001 <.001 <.001 <.001 <.001 .001

.993 .229 .117 .185 .894 .954 .999

<.001 <.001 <.001 .009 <.001 <.001 .001

<.001 <.001 <.001 <.001 <.001 <.001 <.001

1.90E106 2.32E106 8.57E106 1.23E106 6.24E106 6.64E105 2.30E107 2.04E105 1.08E106 4.23E105

7.06E104 6.40E104 4.94E105 9.52E104 1.33E105 3.20E106 1.61E106 1.18E106 2.81E105 1.01E105

4.50E104 1.58E105 1.04E106 3.45E104 4.98E104 4.21E106 5.71E106 5.61E105 1.73E105 1.05E104

<.001 <.001 <.001 <.001 <.001 <.001 .002 .006 .011 .018

<.001 <.001 <.001 .001 .007 <.001 <.001 <.001 .039 .036

<.001 <.001 <.001 <.001 <.001 <.001 .025 .043 <.001 <.001

.609 .408 .133 .133 .065 .920 .051 .062 .265 .346

5.64E104 7.16E105 1.58E106 3.59E106

3.06E106 4.35E104 4.45E106 1.64E106

7.60E104 2.71E105 1.66E106 6.35E106

<.001 .013 .017 .034

<.001 <.001 .002 .036

.893 .127 .891 .738

<.001 .035 .002 <.001

2.79E107 1.89E105

3.00E106 1.24E104

8.08E106 4.33E104

<.001 <.001

<.001 <.001

.002 <.001

.002 .036

AR:A, Comparison between the AR and asthmatic groups; H:A, comparison between the healthy and asthmatic groups; H:AR, comparison between the healthy and AR groups. *Listed are Bonferroni-adjusted P values for Kruskal-Wallis 1-way ANOVA.  Indicated are adjusted multiple pairwise comparison P values.20 Shown in italics are P values that do not reach statistical significance (P > .050).

miR-126, miR-21, and let-7b and decreased levels of miR-1, compared with the cluster with low eosinophil counts (see Table E3). It has been shown that miR-21 is upregulated in allergic airways and regulates eosinophil growth33,34 and that the Let7 family regulates IL-13 expression.35 Furthermore, multiple miRNAs in this group were predicted to regulate different isoforms of the TGF-b receptor, raising the possibility that airway remodeling might be a target of these miRNAs. Thus mRNA transcripts targeted by group 1 miRNAs might provide clues about the molecular mechanisms that underlie these clinical observations. Group 2 contained miRNAs that were unique to asthma because these were differentially expressed in the asthma versus AR and healthy groups but were similarly expressed in patients with AR versus healthy subjects. A majority of the group 2 miRNAs have been implicated in asthma in various studies. miR-16 has been found to be upregulated in airways of the house dust mite mouse model of asthma, although its mechanistic role remains unexplored.36,37 More recently, miR-570 has been shown to bind and regulate the expression of HuR, an RNA-binding protein involved in asthmatic inflammation.38-40 These miRNAs

as a group were predicted to regulate prominent inflammatory genes and signal transduction regulators, such as MAPKs, NF-kB components, and signal transducers and activators of transcription, suggesting that they might regulate key components of inflammatory pathways. Because our study only compares asthmatic patients and patients with AR, it is possible that group 2 miRNAs are involved in other allergic diseases as well. miR-223 is a group 2 miRNA that has been shown to regulate the proliferation of eosinophil progenitors,41 and its differential expression has been documented in eosinophilic esophagitis.34 Group 4 miRNAs are particularly interesting because despite the fact that 83% (29/35) of our asthmatic cohort has AR, these miRNAs are only differentially expressed in nonasthmatic patients with AR. Empiric identification and further investigation of genes and signaling pathways that are exclusively targeted by either of these groups might untangle some of the biomolecular complexities of these diseases. These studies can also lead to more personalized approaches in AR and asthma therapies. In our report group 3 miRNAs are those that are concordantly dysregulated by a similar magnitude in patients with both diseases. As such, group 3 miRNAs might be indicative of the

PANGANIBAN ET AL 7

J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn

Group 1

Group 2

hsa-miR-125b-5p

hsa-miR-16-5p

hsa-let-7b-5p

hsa-miR-223-3p

hsa-miR-1

hsa-miR-570-3p

RELA

hsa-let-7e-5p hsa-miR-21-5p

PIK3CB

STAT1

hsa-let-7c

hsa-miR-299-5p

NFKB1

STAT3

MAPK8

TGFBR1

IKBKB

TGFBR2

hsa-miR-1248 hsa-miR-144-3p

IGF1R

TRAF3

FOS CXCL8

hsa-miR-106a-5p

hsa-miR-422a

hsa-miR-155-5p

Group 3

Group 5 hsa-miR-338-3p hsa-miR-206

Group 4 FIG 2. Predicted inflammatory gene targets of miRNAs. Selected inflammatory genes targeted by each group of miRNAs are depicted in a network diagram. Differentially expressed miRNAs that do not target any of the selected genes are not shown.

B

A

Cluster 1 Cluster 2

Eosinophils (cells/ul)

600

**, p=0.007

400

200

0

Cluster 1

Cluster 2

FIG 3. Cluster analysis of miRNA expression in asthmatic patients. A, Heat map showing 2 main clusters of miRNA expression in asthmatic patients. B, Box plot of peripheral eosinophil levels in clusters 1 and 2.

common pathways involved in AR and asthma. AR and asthma are often thought of as a continuum of the same disorder,42,43 and the group 3 miRNA expression pattern is reflective of the shared molecular pathogenesis of AR and asthma. Indeed, the comorbidity between AR and asthma is well recognized.44,45 Similar to group 1 miRNAs, group 5 miRNAs show differential expression among the 3 cohorts in our study. However, group 5

miRNAs demonstrate a greater magnitude of dysregulation among patients with AR compared with that seen in asthmatic patients. Therefore it is likely that targets of group 5 miRNAs are more involved in AR-specific pathways. One of the main limitations of our study is that we had a relatively small sample size given the heterogeneity of AR and asthma. Because of this, the expression patterns of some of the

8 PANGANIBAN ET AL

J ALLERGY CLIN IMMUNOL nnn 2016

TABLE III. Cluster analysis of asthmatic patients Characteristic

Age (y) Sex, male/female (%) BMI (kg/m2) Current smoker FVC (%) FEV1 (%) FEV1/FVC ratio FEV1 DBD ACQ score AR Aeroallergen sensitivity* Trees Grass Weeds Dust mites Animals Cockroach Molds Total ICS use (mean daily dose [mg]) LABA, no. (%) Anti-leukotriene, no. (%) Eosinophils % Eosinophils

Cluster 1 (n 5 16), no. (%) or mean 6 SEM

Cluster 2 (n 5 19), no. (%) or mean 6 SEM

P value

42.8 6 3.13 5 (31)/11 (69) 27.1 6 1.41 3 (19) 90.0 6 3.91 79.1 6 5.20 0.71 6 0.05 13.9 6 3.74 1.04 6 0.14 13 (81)

44.8 6 3.76 9 (47)/10 (53) 30.9 6 1.54 2 (12) 89.7 6 4.16 77.8 6 5.71 0.69 6 0.04 14.1 6 3.04 0.95 6 0.15 16 (84)

.672 .491 .078 .642 .964 .864 .752 .967 .672 1.000

0.94 6 0.35 0.50 6 0.16 0.63 6 0.20 1.00 6 0.26 0.88 6 0.27 0.13 6 0.09 0.44 6 0.18 4.50 6 0.90 503 6 103 7 (43.7) 4 (25.0) 285.4 6 44.4 4.87 6 0.81

0.78 6 0.26 0.21 6 0.10 0.26 6 0.10 0.84 6 0.23 0.53 6 0.23 0.16 6 0.09 0.21 6 0.12 3.00 6 0.74 715 6 117 8 (42.1) 5 (26.3) 133.3 6 23.4 2.02 6 0.39

.741 .130 .124 .653 .339 .787 .310 .208 .191 1.000 1.000 .007 .005

ACQ, Asthma Control Questionnaire; FVC, forced vital capacity; ICS, inhaled corticosteroid; LABA, long-acting b-sgonist. *Mean number of positive skin test responses in a panel of 19 aeroallergens.

A

B

FIG 4. Results of the random forest model with 6 features and 100 decision trees. A, Plot describing the proportion of each predicted phenotype for subjects of each disease status. B, Multiclass receiver operating characteristic curve analysis with pairwise comparisons of one class versus all other classes. The average area under the curve (AUC) for the 3 receiver operating characteristic curves is 0.9736, indicating that the model performs well in discriminating between positive and negative instances.

profiled miRNAs might be miscategorized. It is possible that with much larger cohorts, circulating miRNAs in our study could be classified under different expression patterns. Nevertheless, even with our population size, we made a number of observations that can carry significant importance. Many of the differentially expressed miRNAs targeted pathways specific to inflammation. The role of the MAPK signaling pathway is well known in the pathogenesis of AR and asthma.46,47 The majority of the miRNA groups also target the phosphoinositide 3-kinase–Akt pathway, which is another particularly well-studied signaling cascade in TH2 inflammation.48,49 Furthermore, there is emerging evidence in the literature that implicates neurotrophin50,51 and

insulin signaling in patients with allergic lung diseases.52 Focal adhesion pathways targeted by group 2 and 3 miRNAs in our study are likely involved in changes in airway basement membrane and airway smooth muscle in asthmatic patients.53 Depending on the context of their biomolecular interactions, the differentially expressed miRNAs can play regulatory or pathologic roles. For example, miR-126 indirectly increases GATA-3 expression in T cells, which could promote a TH2 response and is consistent with our observation that expression of this miRNA was increased in patients with AR and asthma.36 In addition, miR-21 has been shown to enhance eosinophilia by promoting eosinophil precursor growth and

PANGANIBAN ET AL 9

J ALLERGY CLIN IMMUNOL VOLUME nnn, NUMBER nn

inhibiting IL-12–mediated TH1 T-cell polarization.33,54 miR-21 is upregulated by at least 8-fold in the eosinophilic asthma cluster (cluster 1). On the other hand, because many miRNAs serve antiinflammatory functions, it is also possible that some of the miRNAs upregulated in patients with AR and asthma are secondarily induced by chronic inflammation and do not promote disease pathogenesis. For instance, miR-146a is known to be stimulated by inflammatory stimuli and acts as a feedback mechanism to limit inflammatory responses by inhibiting NF-kB and cytokine signaling.55 Furthermore, the Let-7 family of miRNAs has also been shown to have important antiinflammatory actions, primarily on IL-13 and TH2 responses.35 Thus the increased plasma levels of these miRNAs in patients with AR and asthmatic patients compared with healthy subjects could be an attempt to attenuate the inflammatory response. It is interesting to note that although expression of these miRNAs was increased in the blood of asthmatic patients, we and others previously demonstrated that Let7 and miR-146a levels were decreased in the lungs of asthmatic patients compared with those in nonasthmatic control subjects (sources of miRNAs included bronchoalveolar lavage fluid, exhaled breath condensates, and airway epithelial cells).13,56,57 miRNAs are produced by most cells in the body and are highly secreted into exosomes, which enter the circulation and traffic to distant sites where they can be taken up by other cells.58,59 It is possible that abnormalities in the transfer of miRNAs between the blood and lungs and/or impairment in the ability of the lungs to express antiinflammatory miRNAs contribute to asthma pathogenesis. The roles of these and other regulatory mechanisms that govern the synthesis and secretion of miRNAs will require further study because these areas have not been well explored in asthmatic patients and patients with AR. In addition to shedding light on the roles of circulating miRNAs in the molecular pathogenesis of asthma and AR, our study has potential significance in the diagnosis and management of these diseases. The finding that levels of many miRNAs clustered with asthmatic patients with high or low peripheral eosinophil counts indicates that expression profiling could be a useful tool to phenotype asthma. This might be important because patients with eosinophilic asthma can have different responses to glucocorticoids compared with those in patients with noneosinophilic asthma.60 Similar to asthma, subphenotyping of patients with AR is of critical importance to the management and treatment of this disease.44 It is likely that circulating miRNAs could also be differentially expressed among patients with different subtypes of AR. Our findings reveal that a random forest prediction model based on 6 circulating miRNAs is sufficient to determine a subject’s AR or asthmatic status. These miRNAs are miR-125b, miR-16, miR-299-5p, miR-126, miR-206, and miR-133b. Notably, our model could accurately differentiate healthy subjects from asthmatic patients, and the few mislabeled cases were misclassified as AR. This goes hand in hand with the fact that many molecular processes implicated in the pathogenesis of AR also contribute to the pathogenesis of asthma. With high negative and positive predictive values, our model possesses excellent diagnostic potential for AR and asthma. One of the key shortcomings of our predictive model is that it cannot distinguish between asthmatic patients with and without AR. Furthermore, 91% (72/79) of our subjects were white. Our findings will need to

be validated through studies with larger cohorts to overcome the phenotypic and demographic limitations of our study. In conclusion, we show that circulating miRNAs have great potential for the diagnosis of AR and asthma, as well the characterization of asthma subtypes. Plasma miRNAs can be easily extracted from peripheral blood with minimal patient risk, and quantification of blood miRNA levels using qPCR is cheap and reproducible and can be multiplexed for high-throughput analyses. Circulating miRNAs are likely involved in different pathologic or regulatory components of AR and asthma. We have identified candidates that might be involved in the regulation of TH2 inflammation and could serve as novel therapeutic targets. We thank Cathy Mende and Alanna Roff for assistance with the collection and processing of blood samples.

Key messages d

miRNAs are differentially expressed in the blood of patients with AR, asthmatic patients, and healthy subjects. Of the 30 differentially expressed miRNAs, miR-125b, miR-16, miR-299-5p, miR-126, miR-206, and miR-133b are most predictive of a subject’s disease status.

d

Expression profiles of circulating miRNAs in asthmatic patients define 2 subgroups that differ by peripheral eosinophil levels. Circulating miRNAs have potential for use as noninvasive biomarkers to characterize asthma and other allergic diseases.

REFERENCES 1. Ishmael FT. The inflammatory response in the pathogenesis of asthma. J Am Osteopath Assoc 2011;111(suppl):S11-7. 2. Jinek M, Doudna JA. A three-dimensional view of the molecular machinery of RNA interference. Nature 2009;457:405-12. 3. Vasudevan S, Steitz JA. AU-rich-element-mediated upregulation of translation by FXR1 and Argonaute 2. Cell 2007;128:1105-18. 4. Vasudevan S, Tong Y, Steitz JA. Switching from repression to activation: microRNAs can up-regulate translation. Science 2007;318:1931-4. 5. Panganiban RP, Pinkerton MH, Maru SY, Jefferson SJ, Roff AN, Ishmael FT. Differential microRNA epression in asthma and the role of miR-1248 in regulation of IL-5. Am J Clin Exp Immunol 2012;1:154-65. 6. Valadi H, Ekstrom K, Bossios A, Sjostrand M, Lee JJ, Lotvall JO. Exosome-mediated transfer of mRNAs and microRNAs is a novel mechanism of genetic exchange between cells. Nat Cell Biol 2007;9:654-9. 7. Chen X, Liang H, Zhang J, Zen K, Zhang CY. Secreted microRNAs: a new form of intercellular communication. Trends Cell Biol 2012;22:125-32. 8. Calin GA, Croce CM. MicroRNA-cancer connection: the beginning of a new tale. Cancer Res 2006;66:7390-4. 9. Nelson PT, Wang WX, Rajeev BW. MicroRNAs (miRNAs) in neurodegenerative diseases. Brain Pathol 2008;18:130-8. 10. Chen RW, Bemis LT, Amato CM, Myint H, Tran H, Birks DK, et al. Truncation in CCND1 mRNA alters miR-16-1 regulation in mantle cell lymphoma. Blood 2008;112:822-9. 11. Cortez MA, Calin GA. MicroRNA identification in plasma and serum: a new tool to diagnose and monitor diseases. Expert Opin Biol Ther 2009;9:703-11. 12. Kosaka N, Iguchi H, Ochiya T. Circulating microRNA in body fluid: a new potential biomarker for cancer diagnosis and prognosis. Cancer Sci 2010;101: 2087-92. 13. Pinkerton M, Chinchilli V, Banta E, Craig T, August A, Bascom R, et al. Differential expression of microRNAs in exhaled breath condensates of patients with asthma, patients with chronic obstructive pulmonary disease, and healthy adults. J Allergy Clin Immunol 2013;132:217-9. 14. Breiman L. Random forests. Machine Learning 2001;45:5-32. 15. Juniper EF, O’Byrne PM, Guyatt GH, Ferrie PJ, King DR. Development and validation of a questionnaire to measure asthma control. Eur Respir J 1999;14:902-7.

10 PANGANIBAN ET AL

16. Zhu W, Qin W, Atasoy U, Sauter ER. Circulating microRNAs in breast cancer and healthy subjects. BMC Res Notes 2009;2:89. 17. de Hoon MJ, Imoto S, Nolan J, Miyano S. Open source clustering software. Bioinformatics 2004;20:1453-4. 18. Team RDC. R: a language and environment for statistical computing. Vienna (Austria): R Foundation for Statistical Computing; 2009. 19. Gentleman RC, Carey VJ, Bates DM, Bolstad B, Dettling M, Dudoit S, et al. Bioconductor: open software development for computational biology and bioinformatics. Genome Biol 2004;5:R80. 20. Pohlert T. The pairwise multiple comparison of mean ranks package (PMCMR). Available at: https://cran.r-project.org/web/packages/PMCMR/index.html. Accessed March 22, 2016. 21. Benjamini Y, Hochberg Y. Controlling the false discovery rate—a practical and powerful approach to multiple testing. J R Stat Soc Series B Stat Methodol 1995;57:289-300. 22. Vlachos IS, Kostoulas N, Vergoulis T, Georgakilas G, Reczko M, Maragkakis M, et al. DIANA miRPath v.2.0: investigating the combinatorial effect of microRNAs in pathways. Nucleic Acids Res 2012;40:W498-504. 23. Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, et al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res 2003;13:2498-504. 24. Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: machine learning in python. J Machine Learn Res 2011;12: 2825-30. 25. Hu Z, Chen X, Zhao Y, Tian T, Jin G, Shu Y, et al. Serum microRNA signatures identified in a genome-wide serum microRNA expression profiling predict survival of non-small-cell lung cancer. J Clin Oncol 2010;28: 1721-6. 26. Arroyo JD, Chevillet JR, Kroh EM, Ruf IK, Pritchard CC, Gibson DF, et al. Argonaute2 complexes carry a population of circulating microRNAs independent of vesicles in human plasma. Proc Natl Acad Sci U S A 2011;108:5003-8. 27. Ruijter JM, Ramakers C, Hoogaars WM, Karlen Y, Bakker O, van den Hoff MJ, et al. Amplification efficiency: linking baseline and bias in the analysis of quantitative PCR data. Nucleic Acids Res 2009;37:e45. 28. O’Connell RM, Rao DS, Baltimore D. microRNA regulation of inflammatory responses. Annu Rev Immunol 2012;30:295-312. 29. Dai R, Ahmed SA. MicroRNA, a new paradigm for understanding immunoregulation, inflammation, and autoimmune diseases. Transl Res 2011; 157:163-79. 30. Rochat MK, Illi S, Ege MJ, Lau S, Keil T, Wahn U, et al. Allergic rhinitis as a predictor for wheezing onset in school-aged children. J Allergy Clin Immunol 2010;126:1170-5.e2. 31. Shaaban R, Zureik M, Soussan D, Anto JM, Heinrich J, Janson C, et al. Allergic rhinitis and onset of bronchial hyperresponsiveness: a population-based study. Am J Respir Crit Care Med 2007;176:659-66. 32. Bousquet J, Annesi-Maesano I, Carat F, Leger D, Rugina M, Pribil C, et al. Characteristics of intermittent and persistent allergic rhinitis: DREAMS study group. Clin Exp Allergy 2005;35:728-32. 33. Lu TX, Munitz A, Rothenberg ME. MicroRNA-21 is up-regulated in allergic airway inflammation and regulates IL-12p35 expression. J Immunol 2009;182: 4994-5002. 34. Lu TX, Sherrill JD, Wen T, Plassard AJ, Besse JA, Abonia JP, et al. MicroRNA signature in patients with eosinophilic esophagitis, reversibility with glucocorticoids, and assessment as disease biomarkers. J Allergy Clin Immunol 2012; 129:1064-75.e9. 35. Kumar M, Ahmad T, Sharma A, Mabalirajan U, Kulshreshtha A, Agrawal A, et al. Let-7 microRNA-mediated regulation of IL-13 and allergic airway inflammation. J Allergy Clin Immunol 2011;128:1077-85.e1-10. 36. Mattes J, Collison A, Plank M, Phipps S, Foster PS. Antagonism of microRNA-126 suppresses the effector function of TH2 cells and the development of allergic airways disease. Proc Natl Acad Sci U S A 2009;106: 18704-9. 37. Collison A, Herbert C, Siegle JS, Mattes J, Foster PS, Kumar RK. Altered expression of microRNA in the airway wall in chronic asthma: miR-126 as a potential therapeutic target. BMC Pulm Med 2011;11:29.

J ALLERGY CLIN IMMUNOL nnn 2016

38. Roff AN, Craig TJ, August A, Stellato C, Ishmael FT. MicroRNA-570-3p regulates HuR and cytokine expression in airway epithelial cells. Am J Clin Exp Immunol 2014;3:68-83. 39. Casolaro V, Fang X, Tancowny B, Fan J, Wu F, Srikantan S, et al. Posttranscriptional regulation of IL-13 in T cells: role of the RNA-binding protein HuR. J Allergy Clin Immunol 2008;121:853-9.e4. 40. Fan J, Ishmael FT, Fang X, Myers A, Cheadle C, Huang S-K, et al. Chemokine transcripts as targets of the RNA-binding protein HuR in human airway epithelium. J Immunol 2011;186:2482-94. 41. Lu TX, Lim EJ, Besse JA, Itskovich S, Plassard AJ, Fulkerson PC, et al. MiR-223 deficiency increases eosinophil progenitor proliferation. J Immunol 2013;190: 1576-82. 42. Bachert C, Vignola AM, Gevaert P, Leynaert B, Van Cauwenberge P, Bousquet J. Allergic rhinitis, rhinosinusitis, and asthma: one airway disease. Immunol Allergy Clin North Am 2004;24:19-43. 43. Feng CH, Miller MD, Simon RA. The united allergic airway: connections between allergic rhinitis, asthma, and chronic sinusitis. Am J Rhinol Allergy 2012;26:187-90. 44. Bousquet J, Schunemann HJ, Samolinski B, Demoly P, Baena-Cagnani CE, Bachert C, et al. Allergic Rhinitis and its Impact on Asthma (ARIA): achievements in 10 years and future needs. J Allergy Clin Immunol 2012;130:1049-62. 45. Cruz AA, Popov T, Pawankar R, Annesi-Maesano I, Fokkens W, Kemp J, et al. Common characteristics of upper and lower airways in rhinitis and asthma: ARIA update, in collaboration with GA(2)LEN. Allergy 2007;62(suppl 84):1-41. 46. Chung KF. p38 mitogen-activated protein kinase pathways in asthma and COPD. Chest 2011;139:1470-9. 47. Galli SJ, Tsai M, Piliponsky AM. The development of allergic inflammation. Nature 2008;454:445-54. 48. Kampe M, Lampinen M, Stolt I, Janson C, Stalenheim G, Carlson M. PI3-kinase regulates eosinophil and neutrophil degranulation in patients with allergic rhinitis and allergic asthma irrespective of allergen challenge model. Inflammation 2012; 35:230-9. 49. Andiappan AK, Wang de Y, Anantharaman R, Parate PN, Suri BK, Low HQ, et al. Genome-wide association study for atopy and allergic rhinitis in a Singapore Chinese population. PLoS One 2011;6:e19719. 50. Prakash Y, Thompson MA, Meuchel L, Pabelick CM, Mantilla CB, Zaidi S, et al. Neurotrophins in lung health and disease. Expert Rev Respir Med 2010;4: 395-411. 51. Scuri M, Samsell L, Piedimonte G. The role of neurotrophins in inflammation and allergy. Inflamm Allergy Drug Targets 2010;9:173-80. 52. Sarlus H, Wang X, Cedazo-Minguez A, Schultzberg M, Oprica M. Chronic airway-induced allergy in mice modifies gene expression in the brain toward insulin resistance and inflammatory responses. J Neuroinflammation 2013;10:99. 53. Hocking DC. Fibronectin matrix deposition and cell contractility: implications for airway remodeling in asthma. Chest 2002;122(suppl):275S-8S. 54. Lu TX, Lim EJ, Itskovich S, Besse JA, Plassard AJ, Mingler MK, et al. Targeted ablation of miR-21 decreases murine eosinophil progenitor cell growth. PLoS One 2013;8:e59397. 55. Taganov KD, Boldin MP, Chang KJ, Baltimore D. NF-kappaB-dependent induction of microRNA miR-146, an inhibitor targeted to signaling proteins of innate immune responses. Proc Natl Acad Sci U S A 2006;103:12481-6. 56. Levanen B, Bhakta NR, Torregrosa Paredes P, Barbeau R, Hiltbrunner S, Pollack JL, et al. Altered microRNA profiles in bronchoalveolar lavage fluid exosomes in asthmatic patients. J Allergy Clin Immunol 2013;131:894-903. 57. Solberg OD, Ostrin EJ, Love MI, Peng JC, Bhakta NR, Hou L, et al. Airway epithelial miRNA expression is altered in asthma. Am J Respir Crit Care Med 2012;186:965-74. 58. Vickers KC, Remaley AT. Lipid-based carriers of microRNAs and intercellular communication. Curr Opin Lipidol 2012;23:91-7. 59. Chen X, Liang HW, Zhang JF, Zen K, Zhang CY. Horizontal transfer of microRNAs: molecular mechanisms and clinical applications. Protein Cell 2012;3:28-37. 60. McGrath KW, Icitovic N, Boushey HA, Lazarus SC, Sutherland ER, Chinchilli VM, et al. A large subgroup of mild-to-moderate asthma is persistently noneosinophilic. Am J Respir Crit Care Med 2012;185:612-9.