Gene network analysis for identification of microRNA biomarkers for asthma

Background To date, reliable biomarkers for asthma have not been identified. MicroRNAs (miRNAs) are small, non-coding RNAs that negatively regulate post-transcriptional gene expression, and they are involved in various diseases, including asthma. MiRNAs may serve as ideal biomarkers due to their ability to regulate multiple pathways. This study aims to identify miRNA biomarker signatures for asthma. Methods We used the house dust mite (HDM) mouse model of allergic inflammation. Mice were phenotyped by assessing lung function, allergic response, airway inflammation, and remodeling. The miRNA signature profiles in serum and lung tissue were determined by small RNA sequencing, and data were analyzed using Qiagen CLC Genomics Workbench. To identify relevant gene targets, we performed mRNA sequencing, followed by miRNA-targets analysis. These miRNAs and targets were subject to subsequent pathway and functional analyses. Results Mice exposed to HDM developed phenotypic features of allergic asthma. miRNA sequencing analysis showed that 213 miRNAs were substantially dysregulated (FDR p-value < 0.05 and fold change expression >  + 1.5 and < − 1.5) in the lung of HDM mice relative to the control mice. In contrast, only one miRNA (miR-146b-5p) was significantly increased in serum. Target analysis of lung dysregulated miRNAs revealed a total of 131 miRNAs targeting 211 mRNAs. Pathway analysis showed T helper 2/1 (Th2/Th1) as the top significantly activated signaling pathway associated with the dysregulated miRNAs. The top enriched diseases were inflammatory response and disease, which included asthma. Asthma network analysis indicated that 113 of 131 miRNAs were directly associated with asthma pathogenesis. Conclusions These findings suggest that most dysregulated miRNAs in the HDM model were associated with asthma pathogenesis via Th2 signaling. We identified a panel of 30 miRNAs as potential biomarker candidates for asthma. Supplementary Information The online version contains supplementary material available at 10.1186/s12931-022-02304-2.


Introduction
Asthma is a prevalent chronic inflammatory disease of the airways that affects over 300 million people worldwide and 25 million people in the United States [1]. It is characterized by airway hyper-responsiveness (AHR), remodeling, and smooth muscle cell hyperplasia and hypertrophy [2]. Asthma is classified into multiple phenotypes and endotypes, including allergic, eosinophilic, and neutrophilic with allergic asthma the most prevalent subtype [3]. Heterogeneity in the origin of inflammation contributes to challenges in finding effective treatment and there is no cure or specific biomarkers for asthma. Thus, identifying reliable biomarkers to predict disease pathogenesis would fill a gap in our knowledge that could be useful in therapeutic development.
While the use of circulating miRNAs as potential biomarkers for asthma has been investigated, conflicting results that associate circulating miRNAs and asthma have impeded progress toward finding reliable diagnostic and prognostic biosignatures. For instance, Rodrigo-Muñoz et al., identified a panel of three circulating miRNAs (miR-185-5p, miR-320a, and miR-320b) as biomarkers for asthma [11]. In contrast, another group showed a distinct panel of four miRNAs (miRNA-16-5p, miR-223-3p, miR-570-3p, and miR-299-5p) differentially expressed in asthmatic patients relative to the healthy group and individuals with allergic rhinitis [12]. Subsequent study found a subset of seven miRNAs comprising five miRNA ratios as biomarker signatures for asthma [16].
Another confounding problem is that the expression profiles of miRNAs in response to physiological or pathological stimuli are tissue-specific. Therefore, lung tissue might be a more promising source of biomarker discovery for asthma than circulating miRNAs. In this study, we compared lung miRNA signatures to circulating miRNA profiles using a HDM preclinical model of asthma to address the hypothesis that the lung contains distinct miRNA signature patterns that can serve as unique biomarkers for asthma. We have identified a panel of 30 miRNAs as biomarker candidates for asthma. While 18 of the 30 miRNAs had been previously associated with asthma pathogenesis, we have found 12 novel miRNAs (miR-3473b, 7061-5p, 217-5p, 369-3p, 411-3p, 381-3p, miR-7224-3p, 491-5p, 3097-5p, miR-6540-5p, 33-3p, and 1943-5p) correlated with asthma features.

Mice
Male and female BALB/c mice, aged 6-8 weeks old, were purchased from the Jackson Laboratory and housed in a pathogen-free environment in the laboratory animal medicine facility located in the Center for Molecular Medicine at the University of Nevada, Reno (UNR), School of Medicine. All experimental procedures were approved by the institutional Animal Care and Use Committee (IACUC).

HDM sensitization and challenge protocol
The HDM protocol was carried out as previously described [17]. Briefly, mice were sensitized with an intranasal administration of 50 μl of a 25 μg HDM solution or saline once daily for five consecutive days. This was followed by a daily challenge of HDM or saline for five consecutive days per week for 3 weeks.

Tracheostomy and lung function assessment
Mice were anesthetized by an intraperitoneal (i.p.) injection of a ketamine (90 mg/kg)/xylazine (10 mg/kg) cocktail. The anesthesia depth was monitored with a toe pinch and heart rate assessment. Mice were laid supine and hair removed from the chest and neck area using commercial hair removal product followed by wet cotton swabs to reduce irritation. The skin, muscles, and fat in the neck were removed, with stainless-steel surgical scissors, to expose the trachea. A small lateral incision was made to the upper trachea, and a surgical tube was inserted 1/3 deep into the trachea above the primary bronchi. The surgical tube was secured into the trachea using a sterile suture. Additional ketamine was often administered after the tracheostomy to ensure proper sedation before lung function assessment. Mice were mechanically ventilated and lung airway resistance and compliance were measured in response to increasing doses (0-12.5 mg/ ml in PBS) of aerosolized methacholine using Buxco equipment and FinePoint software from Data Sciences International (St. Paul, MN).

Blood collection and serum preparation
Blood was collected after lung function analysis via cardiac puncture as previously described [18] using a 23G needle attached to a 1 ml syringe. After collection, the needle was removed from the syringe, and blood was transferred into serum gel with clotting activator tubes (Sarstedt, Newton, NC). The blood was clotted at room temperature (RT) for 30 min, followed by centrifugation at 2000×g for 10 min at 4 °C, and the serum was transferred to a new 1.5 ml tube. The serum was then centrifuged at 16,000×g at 4 °C for 10 min to remove additional debris and minimize gDNA contamination. The clarified serum was aliquoted into 200 μl aliquots.

Bronchoalveolar lavage and Luminex assays
After the cardiac puncture procedure, bronchoalveolar lavage (BAL) was performed on tracheostomized mice. A total of five lavages were performed per mouse using a 1 ml syringe with a 0.8 ml of sterile 2.6 mM EDTA-saline solution (BAL solution) per lavage. Cell pellets from all five lavages were collected by centrifugation (300×g at 4 °C for 5 min), combined into a single tube, resuspended in 500 μ of BAL solution, and subject to cell counting using a Coulter counter (Beckman Coulter Z series, Brea, CA). The supernatants from the first two lavages were pooled and used for Luminex assays. The Luminex multiplex assays were performed at Eve Technologies Corp (Calgary, AB, Canada) using 50 µl BAL fluid per sample with the Mouse Cytokine Array/Chemokine Array 32-Plex Discovery Assay ® (MD31) according to the manufacturer's instructions.

Lung tissue collection and histology
The right lung lobes and post-caval lobe were collected and placed, at one lobe per tube, into 1 ml tubes containing RNAlater solution (Invitrogen, ThermoFisher Scientific, cat# AM7021, Carlsbad, CA) and stored at − 80 °C for downstream analysis. The left lung lobe was inflated and fixed with 10% neutral buffered formalin (Fisher Scientific, Kalamazoo, MI) at 25 cm above the mouse level. Fixed lungs were embedded in paraffin (FFPE), cut into five μM sections onto positively charged slides. Slides were stained with hematoxylin-eosin (H&E) and Masson's trichrome according to the Reveal Biosciences' standard protocol. The whole slide images were obtained using a Panoramic SCAN (3D Histech, Hungary). The quantitative analysis of whole slide images was performed using imageDX software (Reveal Biosciences, San Diego, CA). The inflammatory cells in the H&E-stained lung were determined and quantitated as the number of immune cells within the total image analysis area (mm 2 ). The collagen and extracellular matrix (ECM) depositions in the Masson's Trichrome-stained sections were evaluated as a percent of the entire image analysis area (mm 2 ). The representative 20X photos were acquired from whole slide images using the CaseViewer software (3D Histech, Hungary).

Detection of serum HDM-specific IgE
HDM-specific IgE in the mouse serum was quantified using the mouse serum anti-HDM IgE antibody assay kit (cat# 3037, Chondrex, Redmond, WA) according to the manufacturer's recommendations. Briefly, anti-HDM Mouse IgE antibody standards (diluted at a range of 0-50 ng/ml) and mouse serum samples (diluted at 1:10) were added in duplicate wells onto a 96-well ELISA plate precoated with anti-mouse IgE antibody. The plate was incubated for 2 h at RT and washed 3× with wash buffer. The plate was incubated with the biotinylated HDM detection antibody for 2 h at RT and washed 3×. The streptavidin peroxidase was added, incubated at RT for 30 min, and washed 3×. The plate was incubated with TMB (3,3′,5,5′-tetramethylbenzidine) substrate for 25 min at RT, the reaction was stopped with 2N sulfuric acid, and the plate was read at 450 mm.

microRNAs sequencing and bioinformatic analysis
microRNA sequencing (miRNA-seq) from serum and lung tissue was performed at the Qiagen Genomic Center (Hilden, Germany) according to the facility's established protocols. For serum samples, RNA was extracted from 200 μl of serum using the miRNeasy serum/plasma advanced kit, and the RNA quality control (QC) was determined by QPCR of serum endogenous miRNAs (such as miRNA-103a-3p and miR-191-5p), hemolysis markers (miR-23a and miR451a), and Spike-Ins control (UniSp6). For lung tissue samples, RNA was extracted with the RNeasy mini kit, quantified using a Nanodrop, and the RNA integrity was assessed using the Agilent TapeStation. miRNA libraries for serum and lung were prepared using Qiagen's miRNA Library Kit following the manufacturer's protocol. cDNA libraries with Unique Molecular Index (UMI) assignment were amplified, cleaned up, QC done by Bioanalyzer, and sequenced using single-end reads, 1 × 75 bp, to a depth of 12-14 million reads (NextSeq 550, Illumina). FASTQ files with quality scores (QS) > 30 were analyzed by CLC Genomics Workbench (version 12.0.2) using the QIAseq miRNA quantification workflow. The de-duplicated Unique Molecular Index (UMI) reads were grouped into unique search sequences and mapped to the miRNAs annotated in the miRbase v22. miRNA expression was calculated via counting the number of distinctive UMI reads corresponding to each miRbase-annotated miRNA and normalized to the sequencing depth (counts per million). Differential miRNA expression between control and HDM-treated mice was assessed using the "Exact Test" for two-group comparisons available as part of the EdgeR Bioconductor package in the CLC Genomic Workbench. The microRNA-Seq data have been deposited at the NCBI Gene Expression Omnibus (https:// www. ncbi. nlm. nih. gov/ geo) under series accession number GSE214937.

mRNA sequencing and bioinformatic analysis
Libraries were prepared from the same lung RNA samples used for miRNA sequencing using QIAseq Stranded mRNA Select Kit with poly-A enrichment. cDNA libraries were amplified, cleaned up, QC performed with a Bioanalyzer, and sequenced using paired-end 2 × 75 bp read, to a depth of 60 million reads (NextSeq 550, Illumina) FASTQ files with quality scores (QS) > 30 were analyzed, mapped, and quantified using the RNA-seq analysis tool from CLC Genomics Workbench (version 21.0.4). Reads were trimmed, mapped to the mouse genome version GRCm38, and annotated using the ENSEMBL Mus_ musculus. GRCm38 version 98. Gene expression was calculated by counting number of reads mapping to the annotated gene loci, and the resulting expression values were normalized to transcripts per million (TPM). Differential gene expression between control and HDMexposed mice was assessed using the "Exact Test" for two-group comparisons available as part of the EdgeR Bioconductor package in the CLC Genomic Workbench. The RNA-Seq data have been deposited at the NCBI Gene Expression Omnibus (https:// www. ncbi. nlm. nih. gov/ geo) under series accession number GSE214937.

miRNA targets identification and pathway analysis
miRNA target identification and pathway analysis were performed using Ingenuity Pathway Analysis (IPA, Qiagen). To determine the most significantly dysregulated molecules, miRNAs and mRNA datasets were filtered using the filter dataset tab on IPA. The cut-off values were log2 fold-change expression (log2 FC) > 0.58 or < − 0.58 and FDR p < 0.05. Next, the mRNA targets were identified using the miRNA target filter. It is known that one miRNA can target up to 200 mRNAs. Therefore, to reduce the list of mRNA targets, we applied filters to include diseases relevant to asthma pathogenesis, such as inflammatory disease, inflammatory response, and respiratory disease. Further, the pathways list was filtered to include those related to a disease-specific pathway, such as airway inflammation in asthma and cellular growth, proliferation, and development pathways (including Th1, Th2, and Th17 activation pathways). The resulting miRNA and mRNA targets were subject to the IPA core analysis [19].

cDNA synthesis and QPCR
The lung RNA samples from the same mice used for all the experiments listed above, including pulmonary function, lung histology, and miRNA/RNA sequencing, were utilized to validate the miRNA sequencing data. miRNA cDNA was synthesized from five ng of total RNA using TaqMan Advanced miRNA cDNA Kit (Cat# A28007, Thermofisher Scientific, Waltham, MA) according to the manufacturer's recommendations. QPCR was performed on the QuantStudio3 Fast Real-time PCR system, 96-well, 0.1 ml plate with 2X Advanced TaqMan Fast Master Mix and specific miRNA probe-primer mix. miRNA expression was normalized to mmu-miR5121. The miRNA mmu-miR-5121 was chosen as reference miRNA because its expression in the dataset remains unchanged in response to HDM exposure.

Statistics
Mice characterization data were graphed and analyzed using Graph Pad Prism software. Data represent mean ± standard error mean (SEM). An unpaired t-test with Welch's correction was used for two-group comparison and p ≤ 0.05 is considered statistically significant. For the miRNA and RNA sequencing data, the false discovery rate (FDR) p-values were calculated using the Benjamini-Hochberg method to account for multiple testing issues, and FDR < 0.05 was considered significant. The networks, pathways, and functional analyses were done through IPA. Right-tailed Fisher's Exact Test was used to calculate p-values, and p < 0.05 (− log 10 p-value > 1.3) was considered significant.

HDM exposure increases airway resistance and induces allergic inflammatory response
To confirm the efficacy of HDM exposure on lung function, mice were anesthetized with ketamine/xylazine, tracheostomized, and airway resistance was measured in response to increasing doses of methacholine (0-12.5 mg/ml). HDM sensitized and challenged mice displayed a significantly increased lung resistance compared to the controls (Fig. 1A). These mice also showed a substantial elevation in the concentration of serum HDM-specific IgE relative to the saline controls (Fig. 1B).
To analyze the effect of HDM sensitization and challenge on inflammation and remodeling, we assessed the total inflammatory cell count in the BAL fluid and performed histology in FFPE lung sections. The total BAL inflammatory cell counts were significantly higher in mice exposed to HDM compared to controls ( Fig. 2A).
Hematoxylin-eosin (H&E) staining revealed that HDM mice had a substantial increase in lung inflammation and thicker airways than the saline controls ( Fig. 2B-D). However, the collagen deposition levels were unchanged between the two groups ( Fig. 2E, F). These results suggest that HDM sensitization and challenge trigger airway inflammation and remodeling but do not affect ECM deposition.
The effect of HDM sensitization and challenge on various inflammatory cytokines and chemokines was assessed via Luminex assays. The secretion of major Th2 cytokines, IL-4, 5, 6, 10, and 13, were significantly enhanced in HDM mice compared to the controls ( Fig. 3A-E). In contrast, IL-1α levels were reduced in HDM mice relative to controls (Fig. 3F). Additionally, the production levels of chemokines, including eotaxin/ CCL11, IP-10/CXCL10, KC/CXCL3, MCP-1/CCL2, MIG/CXCL9 and MIP-1β/CCL4, and were significantly Fig. 1 Assessment of airway hyper-responsiveness and IgE production. A Airway resistance was measured in response to increasing doses of methacholine, and B serum HDM-specific IgE was measured using ELISA. The graphs represent mean ± SEM, n = 16-18, *p < 0.05 and ****p < 0.0001 HDM-challenged mice versus alum control, two-way ANOVA followed by Dunnett's multiple comparisons (A) or unpaired t-test with Welch's correction (B) Data represent mean ± SEM, n = 12-18 per group, ***p < 0.001 and ****p < 0.0001 HDM-challenged mice relative alum control using unpaired t-test with Welch's correction Data are expressed as mean ± SEM, n = 16-18, **p < 0.01 ***p < 0.001 and ****p < 0.0001 versus alum control, unpaired t-test with Welch's correction higher in BAL fluid of HDM mice relative to saline controls (Fig. 4). As expected, the HDM model exhibited increased production of key inflammatory mediators. Taken together, these results validate the use of the HDM model for these studies. We also analyzed the phenotypic features of the HDM mouse model based on sex and found that female HDM-exposed mice exhibited a similar allergic inflammatory profile as the male HDM counterparts (data not shown). Therefore, both sexes were combined for all subsequent analyses.

Dysregulation of miRNA expression in the HDM mouse model of allergic inflammation
Principal component analysis (PCA) of miRNA expression in lung samples showed two nicely separated clusters between mice treated with saline and HDM (Fig. 5A). In contrast, there was no clear clustering observed between saline and HDM mice in serum samples (Fig. 5B). In the lung tissue, a total of 345 miRNAs were significantly differentially expressed (FDR < 0.05 or − log 10 (FDR) > 1. 3) in HDM mice relative to control (Fig. 5C), but only one miRNA (miR-146b-5p) was substantially upregulated (FDR < 0.05 or − log 10 (FDR) > 1.3) in serum (Fig. 5D, E). Thus, the rest of this study focuses on lung miRNA signatures. Further analysis of the lung miRNA profile indicated that 125 miRNAs were up-regulated and 88 miRNAs were downregulated by at least 1.5-fold change in HDM mice compared to saline control (Fig. 5E, F). This lung miRNA signature was the focus of the rest of this study ( Fig. 5F and Table 1).

Target identification of dysregulated lung miRNAs and pathway analysis of target genes
miRNA target analysis releveled a total of 131 microR-NAs, with conserved seed sequence homology between mouse and human, targeting 211 mRNAs. There were 78 miRNAs upregulated and 53 miRNAs downregulated in HDM mice relative to saline controls. The expression pairing between miRNAs and corresponding mRNA targets is displayed in Additional file 2: Table S1. These molecules (Additional file 2: Table S1) were subject to pathway analysis to identify relevant signaling pathways and biological functions. We found that 38 canonical pathways were substantially impacted (− log 10 (p-value) > 1.3) with significant predicted activity (Z-score ≥ + 2) or inhibition (Z-score ≤ − 2) (Additional file 1: Fig. SA). The top 15 significantly activated pathways associated with the target genes were identified using IPA and included T helper 2 and 1 (Th2 and Th1), high mobility group box protein1 (HMGB1), interleukin-17 (IL-17), T and B cell signaling, and RhoA signaling (Fig. 7).

Functional analysis of dysregulated miRNAs and their targets
Using pathway analysis, we identified several pathologies significantly associated with miRNAs and corresponding  3). A − log 10 (FDR) > 1.3 is considered as statistically significant. E Venn diagram of significantly up-regulated miRNAs in lung (green circle), serum (orange circle) samples, and down-regulated miRNAs in lung (blue circle). The miRNAs included in this analysis have a log2 fold change expression cut off > + 0.58 and cut off < − 0.58 in HDM vs saline mice. n = 14-16 per group, and FDR < 0.05 is considered statistically significant. The Venn diagram was designed using the online Venny 2.1.0, https:// bioin fogp. cnb. csic. es/ tools/ venny/ index. html, with slight modifications. F Heat map of the 213 miRNAs that were substantially dysregulated in the HDM mice vs. saline controls. Red squares show upregulated miRNAs and green represent downregulated miRNAs. The heat map was generated by scaling the data using Z-scores of each miRNA (normalized CPM were converted into z-scores)     Fig. SB). The five top impacted pathologies were inflammation, organismal injury and abnormalities, connective tissue disorders,   and skeletal and muscular disorders (Fig. 8A). The anomalies with significantly increased (z-score > 2) biological functions were inflammatory response, organismal injury and abnormalities, and inflammatory disease ( Fig. 8B-D). These biological functions were mainly related to the activation, accumulation, and chemotaxis of immune cells (Fig. 8B), lung damage, injury, fibrosis (Fig. 8C), and airway hyper-responsiveness (Fig. 8D). All these functions are associated with phenotypic features of asthma. Additionally, we noted that asthma was substantially enhanced in the top-enriched disease categories ( Fig. 8B-D), indicating the correlation of the lung miRNA signatures with asthma pathogenesis.

Identification of miRNA and mRNA targets associated with an asthma network
We performed further downstream analysis, based on published data found in the IPA knowledge base, to determine genes and targeting miRNAs associated with asthma (z-score = 2.41) in mice exposed to HDM. This analysis identified 45 genes and three miRNAs (miR-135-b-5p, miR-148a-3p, and miR-215-5p). These genes included cytokines (IL-4, IL-5, and IL-13), chemokines (CCL2), growth factors (TGFβ1), peptidases (MMP9), transcription regulators (FOXP3 and GATA3), transmembrane receptors (IL13RA2 and IL17RB), enzymes and Kinases (Fig. 9A, Table 2). To connect the genes in Fig. 9A to the targeting miRNAs, we utilized the build and grow tabs in the path designer tool. These genes were targeted by 113 distinct miRNAs (Fig. 9B, Table 3). These results may indicate that most miRNAs that were differentially dysregulated in HDM mice are involved in asthma pathogenesis.
These data indicate that 22 of the 30 miRNAs tested were associated with three features of asthma, and eight of the 30 miRNAs were correlated with at least one asthma parameter.

Discussion
The primary goal of this study was to identify miRNA biomarkers for asthma using a house dust mite (HDM) mouse model of allergic inflammation. Although several common allergens are used to induce allergic inflammation, HDM remains the most clinically-relevant antigen, as it affects nearly 85% of asthma patients worldwide [25]. Even though HDM is a well-established model, the outcomes of the allergic response and associated molecular signatures, such as miRNA profiles, can vary significantly depending on the dose of the allergen, sensitization route, mouse strain, and protocol used. Our results show that HDM-exposed mice exhibited an increased airway resistance, prominent airway inflammation and thickening, elevated serum HDM-specific IgE, and increased production of key inflammatory mediators. These results are in line with previous studies and indicate that the HDM protocol used in this study elicited a potent allergic inflammatory response with airway remodeling that recapitulates key features of human asthma [17,26]. We utilized miRNA sequencing followed by bioinformatic analysis to identify miRNA signatures in lung tissue and serum of HDM-exposed mice relative to the saline control group. The comparison between lung tissue and serum miRNA profiles revealed 213 miR-NAs significantly dysregulated in lung tissue, with the Fig. 10 The miRNA biomarker candidates for asthma. Biomarker analysis was performed using prior knowledge from IPA. There are four miRNA biomarkers categories (efficacy, safety, diagnosis, and disease progression), and each category is connected to the molecules in the network by solid black lines. Nodes represent the molecules and lines indicate the relationship between two nodes. Green nodes represent downregulated miRNAs, and the upregulated ones are in pink or red. All mRNA targets, except MMP9, are represented in V-shaped dark pink. Each group of miRNAs is pointing to its specific mRNA target. BM: Biomarker log2 fold change expression cut off (> + 0.58 and cut off < − 0.58) and FDR p-values < 0.05. Unexpectedly, only miR-146b-5p was substantially upregulated in the serum. These results suggest that lung tissue is a more reliable source of miRNA-based biomarker discovery for asthma than circulating miRNAs. To the best of our knowledge, this is the second study to analyze the free circulating miRNA profiles in a murine model of allergic inflammation. The first group (Milger K. et al.) to profile plasma in a mouse model of inflammation identified 11 significantly dysregulated miRNAs with the fold change (FC) from 0.59 to 1.75 and p < 0.01 [16]. While we only found one miRNA significantly upregulated (FC = 2.5 and FDR p-value = 5.13E−13). In Milger K. et al. 's study, miRNA signatures analysis was done using a focus panel containing 179 known miRNAs in plasma samples from female mice exposed to HDM or PBS. However, our study used serum from male and female mice, and miRNA profiles were determined via deep sequencing. The discrepancy between Miller et al. 's results and ours may be due to the gender and experimental methodology differences between the two studies.
The lung is a complex organ comprised of various cell types, including epithelial, smooth muscle, fibroblast, and endothelial cells [27] that undergoes structural changes and inflammatory cell recruitment when exposed to allergens such as HDM. Thus, analysis of miRNA signatures from whole lungs likely reflects a pool of various miRNAs from specific lung cell types and infiltrating immune cells, such as eosinophils. This would probably be more beneficial in unveiling the accurate landscape of the miRNA signature profile for asthma rather than using a specific lung cell type. Based on this rationale, we decided to focus solely on lung miRNA signatures since miR-146b-5p was the only miRNA dysregulated in serum, and miR146b-5p was also included in the whole lung analysis.
Target analysis of lung miRNA signatures showed 131 microRNAs targeting 211 mRNAs. These miRNAs had conserved seed sequence homology with human miR-NAs, indicating their relevance to human asthma. Pathway analysis demonstrated that the mRNA targets were implicated in various immune and inflammatory signaling pathways, and Th2 signaling was the most significantly enriched pathway. Functional analysis indicated that the dysregulated miRNAs and corresponding mRNA targets were involved in several inflammatory responses and diseases, specifically in immune cell activation, accumulation, and chemotaxis, lung damage, lung injury, lung fibrosis, airway hyper-responsiveness, and asthma. Interestingly, all these biological functions are involved in the phenotypic features of asthma. Taken together, these findings suggest that lung miRNA signatures identified herein are associated with Th2-mediated asthma pathogenesis, the most predominant feature observed in mice models of allergic inflammation, including the BALB/c mouse strain [28]. Therefore, these miRNA signatures would be specifically relevant for the diagnosis of patients with the Th2-high asthma phenotype but not those with other asthmatic phenotypes, such as Th2-low asthma. The asthma network analysis unveiled 113 miRNAs targeting 45 genes. These genes included cytokines (IL-4, IL-5, and IL-13), chemokines (CCL2), growth factors (TGFβ1), peptidases (MMP9), transcription regulators (FOXP3 and GATA3), transmembrane receptors (IL13RA2 and IL17RB), enzymes and kinases that were significantly dysregulated with 41 genes were upregulated and four were downregulated in our dataset. These results confirm the implication of these genes in asthma and agree with previous human asthma studies [3,29].
Similar studies have been done in ovalbumin models using either male or female mice [8,37,45]. However, our study has several strengths, including using the HDM mouse model, a most clinically-relevant model of allergic inflammation. Additionally, we surveyed an extensive number of miRNAs using male and female mice. Further, each of the 30 miRNA candidates had good predictive power (AUC ≥ 80%), making these miRNAs stand-alone biomarkers for asthma diagnosis. The weakness of the study is that we did not confirm our findings in human samples. Despite this, several miRNAs discovered herein were also implicated in human asthma, indicating the validity of HDM models in profiling lung miRNAs. Subsequent studies will be performed to analyze how these miRNAs regulate asthma. Further, we will test if these miRNAs can serve as biomarkers in clinical settings using lung biopsy samples from asthmatic and healthy individuals.

Conclusions/future directions
These results suggest that lung tissue is a more reliable source of miRNA-based biomarker discovery for asthma than circulating miRNAs. Overall, these findings indicate that mouse models of inflammation might not be ideal models for circulating miRNA studies. Regardless of this plausible specific pitfall, the HDM model used herein was very useful in recapacitating all key features of human asthma, leading to the discovery of unique lung miRNA signatures for allergic asthma. Although most of these miRNAs were previously implicated with asthma, we have identified 12 novel biomarkers. Subsequent studies