Circulating miRNAs and childhood asthma ICS response: a stratified analysis in the intervention arm of an RCT with vitamin D effect modification

Abstract

Background

miRNAs play a crucial role in the anti-inflammatory effects of inhaled corticosteroids (ICS) in asthma. Vitamin D can modulate the expression of several miRNAs and reduces asthma exacerbations, but its molecular interaction with ICS remains unclear.

Objective

We hypothesised that vitamin D influences long-term ICS response through miRNA regulation.

Methods

Baseline serum miRNAs were sequenced from 462 subjects in the Childhood Asthma Management Program (CAMP), with 187 randomised to ICS treatment included in this study. Linear regression assessed associations between miRNA expression and prebronchodilator forced expiratory volume in 1 s per cent predicted (FEV₁%) change over 4 years, stratified by baseline vitamin D levels and tested in interaction models. Microarray analysis of lymphoblastoid B cells (lymphoblastoid cell lines (LCLs)) from 22 CAMP subjects treated with dexamethasone (DEX), vitamin D or sham identified differentially expressed genes (DEGs). An miRNA target gene network was constructed, clustered and annotated by enrichment analysis. Top miRNAs were evaluated for ICS response prediction.

Results

12 miRNAs were significantly associated with ICS-mediated FEV₁% change in vitamin D insufficient subjects, and 11 miRNAs showed significant interaction with vitamin D (p≤0.05). Three miRNAs were approximately replicated in the Genetics of Asthma in Costa Rica Study. Microarray analysis identified 220 and 240 DEGs in DEX and vitamin D-treated LCLs, respectively. miRNAs hsa-miR-125a-5p, hsa-miR-181a-5p, hsa-miR-101-3p and hsa-miR-107 were enriched in haemopoiesis and leucocyte differentiation pathways (p≤0.05). Two miRNAs, hsa-miR-125a-5p and hsa-miR-181a-5p, predicted ICS response with an area under the receiver operating characteristic curve of 0.86 in the vitamin D insufficient group.

Conclusions

Vitamin D may modulate ICS response through miRNAs involved in immune cell differentiation, which could serve as biomarkers for ICS response, particularly in vitamin D insufficient individuals.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Circulating miRNAs are associated with long-term inhaled corticosteroid (ICS) response and are recognised as reliable biomarkers. Vitamin D plays a critical role in the immune system and affects ICS function.

WHAT THIS STUDY ADDS

• Our study indicates that vitamin D may influence ICS response through the leucocyte differentiation pathway modulated by miRNAs.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our findings highlight that miRNAs are critical modulators in the functional interplay between ICS and vitamin D. Moreover, they serve as promising biomarkers for ICS response in patients with asthma and vitamin D insufficiency.

Introduction

Asthma is a chronic, non-communicable lung disease characterised by airway inflammation, hyper-responsiveness and variable airflow obstruction.¹ According to the Global Burden of Disease Study, asthma affected approximately 262 million people worldwide in 2019, placing a significant burden on both patients’ families and healthcare systems.²

Inhaled corticosteroids (ICS) are recommended as the first-line treatment for long-term asthma control and prevention of exacerbations. Corticosteroids can exert rapid effects through non-genomic mechanisms via cell surface receptors or intracellular signalling pathways, as well as through genomic effects.³ In the latter, corticosteroids bind to glucocorticoid receptors (GRs) in the cytoplasm, allowing the complex to enter the nucleus, where it exerts anti-inflammatory effects. These include activating anti-inflammatory genes as transcription factors, interacting with coactivator molecules to suppress inflammatory genes and recruiting histone deacetylase-2 to inhibit inflammatory genes.⁴

Vitamin D and its metabolites, a group of fat-soluble secosteroids, help regulate calcium and phosphorus in the body but also influence immune function.⁵ Similar to corticosteroids, calcitriol (1,25-dihydroxyvitamin D₃ (1,25(OH)₂D₃), the active form of vitamin D, binds to the vitamin D receptor in the cytoplasm, forming a heterodimer that functions as a transcription factor in the nucleus to activate downstream genes.⁶ Vitamin D plays a critical role in bone development and immune regulation and inflammation. Deficiency in vitamin D has been associated with increased airway hyper-responsiveness, as well as poor lung function in childhood asthma in our previous study.⁷

Vitamin D may enhance or interact with the efficacy of corticosteroids.⁸ Reduced vitamin D levels have been associated with a diminished glucocorticoid response in asthma.⁹ Some studies have shown that vitamin D supplementation improves corticosteroid responsiveness.⁷ Xystrakis et al demonstrated that vitamin D could reverse the impaired interleukin 10 (IL-10) induction by corticosteroids in CD4+ T cells from steroid-resistant patients, suggesting a potential immune mechanism for this interaction.¹⁰ Nanzer et al revealed that vitamin D can enhance the corticosteroid response by inhibiting the capability of helper T 17 (Th17) cells and reducing IL-17A secretion from Th17 cells.¹¹ However, the genomic basis of these effects is still unknown.

MicroRNAs (miRNAs) are a class of small, non-coding, single-stranded RNAs about 22 nucleotides long that function post-transcriptionally by binding to mRNAs, downregulating gene expression.¹² miRNAs are considered as excellent non-invasive biomarkers of many diseases.¹³ Previous studies have reported associations between extracellular circulating miRNAs and asthma severity across different vitamin D levels, as well as treatment response to ICS.14,16 In this study, we hypothesised that miRNAs are potential modulators and biomarkers of the effect of vitamin D on long-term ICS response.

Methods

Study populations

The Childhood Asthma Management Program (CAMP) (ClinicalTrials.gov: NCT00000575) was a double-blind randomised controlled clinical trial designed to evaluate the efficacy and safety of long-term ICS treatment in children with mild to moderate asthma.¹⁷ A total of 1041 children with asthma were enrolled and randomly assigned to one of three groups: budesonide (311 children), nedocromil (312 children) or placebo (418 children). This trial demonstrated that inhaled budesonide improves airway responsiveness and provides better asthma control.¹⁷ A more detailed description can be found in the online supplemental material.

We sequenced baseline serum miRNAs from 187 subjects randomised to the budesonide treatment group. Blood samples were collected at baseline, prior to randomisation, and participants were followed for 4–6 years. Serum 25-hydroxyvitamin D₃ (25(OH)D₃) concentrations were measured using a radioimmunoassay in Dr Bruce Hollis’ laboratory at the Medical University of South Carolina. Vitamin D levels were categorised as insufficient (≤30 ng/mL) or sufficient (>30 ng/mL) and were used in subsequent stratified and interaction analyses. Lung function was monitored throughout the follow-up period.

Replication analyses were performed using the Genetic Epidemiology of Asthma in Costa Rica Study (GACRS) (ClinicalTrials.gov: NCT00021840), a cross-sectional study involving 1165 children with asthma, aged 6–14 years. Detailed information about the GACRS cohort has been previously published.¹⁸ ICS usage history was collected through questionnaires, and serum vitamin D levels were measured from blood samples taken at enrolment. A total of 568 subjects had a history of ICS treatment at enrolment, of which 104 were in the vitamin D insufficient group, 285 were in the vitamin D sufficient group and 179 had missing vitamin D data.

Small RNA-Seq and profiling

Total RNA was isolated using the Qiagen miRNeasy Serum/Plasma Extraction Kit with QIAcube automation. RNA concentration was measured using a Nanodrop spectrophotometer, and RNA-Seq libraries were prepared with the Norgen Biotek Small RNA Library Prep Kit. Small RNA sequencing was performed on the Illumina NextSeq 500 platform, generating 51 bp single-end reads. COMPSRA was used to assess read quality and annotate miRNAs.¹⁹ miRNAs with fewer than five mapped reads or present in less than 90% of subjects were excluded. Read counts were log transformed and quantile normalised. Further details on the quality control of the sequencing data are provided in online supplemental figures S1–S3. The sequencing data have been deposited in the Gene Expression Omnibus under accession number GSE134897.¹⁴

Statistical analysis

The change in ICS response was defined as the difference in prebronchodilator forced expiratory volume in 1 s per cent predicted (FEV₁%) between baseline and the 4-year clinical trial. Univariate and multivariate linear regression models were used to assess the association between miRNA and changes in FEV₁% in both vitamin D sufficient and insufficient groups, and the interaction between miRNA and vitamin D levels was also evaluated. Age, sex, race and body mass index (BMI) were included a priori as confounders in the multivariate regression model. P values were adjusted for multiple testing using false discovery rate (FDR) with the R package ‘fdrtool’.

CAMP lymphoblastoid cell line gene expression arrays

22 participants randomised to ICS treatment provided additional blood samples roughly 6 years after the study’s conclusion. Lymphoblastoid B cells were isolated and immortalised using Epstein-Barr virus to create lymphoblastoid cell lines (LCLs). LCLs were cultured in RPMI (Roswell Park Memorial Institute)-1640 medium supplemented with 5% FBS (Fetal Bovine Serum) and 1× penicillin/streptomycin/L-glutamine, maintaining an average density of 200 000 cells/mL. These LCLs were then cultured separately with dexamethasone (DEX) (10⁻⁶ _M) for 6 hours and 1,25-OH vitamin D (10⁻⁶ _M) for 72 hours, as well as with a sham treatment. The transcriptome expression levels were measured using the Illumina HumanRef8 v2 BeadChip (Illumina, San Diego, California). The R package ‘limma’ was employed for quality control, batch effect adjustment, normalisation and differential gene expression analysis. P values were adjusted for multiple comparisons using the FDR.

miRNA and target gene mixed network

We constructed a mixed network of miRNAs and their target genes by integrating miRNA target gene regulation data with protein-protein interaction information. Target gene data for miRNAs were obtained from miRTarBase v9.0, which includes miRNA target gene relationships identified through various experimental methods (eg, luciferase reporter assay, HITS-CLIP (High-Throughput Sequencing of RNA isolated by CrossLinking and ImmunoPrecipitation), CLASH (Cross-linking Ligation and Sequencing of Hybrids), quantitative reverse transcription PCR, western blot and microarray). We selected miRNA target genes based on strong experimental evidence (eg, reporter assay or western blot). Protein-protein interaction data for target genes were downloaded from STRING, which integrates both physical interactions and functional associations. The mixed network was clustered using the RandomWalkRestartMH method, starting with each miRNA as the initial node. These clusters were subsequently enriched with Gene Ontology (GO) biological functions using the R package ‘clusterProfiler’. The entire network and its clusters were visualised using Cytoscape V.4.1. The regulatory effects of miRNAs were assessed using the Kolmogorov-Smirnov test and further validated with a rank-based enrichment score method, similar to GSEA (Gene Set Enrichment Analysis).²⁰

Prediction

We employed a multivariate logistic regression model to predict ICS response in the vitamin D insufficient group. To account for the intrinsic differences in FEV₁%, we focused on predicting the highest (Y=1) versus lowest (Y=0) tertiles of the change in FEV₁%. The normalised miRNA counts, along with potential confounders, such as age, sex, race and BMI, were used as input variables. Model performance was evaluated using the area under the receiver operating characteristic (AUROC) curve. The 95% CI of the AUROC was calculated using bootstrap resampling with 1000 iterations. Hedges’ g was calculated through the R package ‘effectsize’.

Results

Baseline characteristics

A total of 187 subjects were treated with budesonide, with 132 in the vitamin D sufficient group and 55 in the vitamin D insufficient group. Baseline characteristics are summarised in table 1. Race (p=6.73×10⁻⁶) showed a significant difference between the vitamin D sufficient and insufficient groups, consistent with our previous findings.¹⁵ The average change in FEV₁% predicted from baseline to follow-up was higher in the vitamin D insufficient group compared with the sufficient group (5.70±12.82 vs 3.53±12.41), although the difference was not statistically significant due to the large SD.

Table 1. Baseline characteristics of study participants from CAMP.

Characteristic	Vitamin D insufficient* (n=55)	Vitamin D sufficient* (n=132)	P value
Age, year	9.41±2.11	8.81±2.14	0.08
Sex			0.09†
Male	26 (47.27)	82 (62.12)
Female	29 (52.73)	50 (37.88)
Race			6.73×10−6†
White	25 (45.45)	102 (77.27)
Black	24 (43.63)	16 (12.12)
Hispanic	6 (10.92)	14 (10.61)
BMI, kg/m2	19.24±4.37	18.28±3.31	0.15
Bone age, months	168.44±32.98	161.97±30.84	0.23
Household income			0.87†
≥$30 000	40 (73.73)	94 (71.21)
<$30 000	15 (27.27)	31 (23.48)
Don’t know	0 (0)	7 (5.3)
Asthma severity			0.68†
Mild	25 (45.45)	54 (40.91)
Moderate	30 (54.55)	78 (59.09)
Parental asthma history
Yes	29 (52.73)	56 (42.42)	0.19†
No	23 (41.82)	72 (54.55)
Don’t know	3 (5.45)	4 (3.03)
Child has ever been hospitalised for asthma			1†
Yes	19 (34.55)	44 (33.33)
No	36 (65.45)	88 (66.67)
Parental smoking status			0.87†
Yes	22 (40)	50 (37.88)
No	32 (58.18)	81 (61.36)
Missing	1 (1.82)	1 (0.76)
FEV1% predicted at baseline	92.09±13.02	93.42±13.27	0.53
Changes in FEV1% predicted between baseline and follow-up	5.70±12.82	3.53±12.41	0.33

Data presented as n (%) or mean±SD.

^*Vitamin D sufficient was defined as serum 25(OH)D₃>30 ng/mL and vitamin D insufficient was defined as serum 25(OH)D₃≤30 ng/mL.

^†P value from χ2 test.

BMI, body mass index; CAMP, Childhood Asthma Management Program; FEV₁% predicted, forced expiratory volume in 1 s, expressed as a percentage of the predicted value.

Significant miRNAs in the vitamin D insufficient group

We identified 12 miRNAs significantly associated with changes in FEV₁% predicted between baseline and follow-up in the vitamin D insufficient group, based on both univariate and multivariate linear regression models (table 2). Four miRNAs were positively associated with changes in FEV₁% predicted, indicating a positive correlation between increased miRNA count and improved lung function. Conversely, eight miRNAs were negatively associated, meaning that higher miRNA counts were linked to decreased FEV₁% predicted. Among these, hsa-miR-125a-5p showed the strongest positive effect (β=6.52, p=0.005), while hsa-miR-181a-5p exhibited the strongest negative effect (β=−8.32, p=0.02). All the 12 miRNAs were significant at an FDR≤0.1. Notably, no significant miRNAs were identified in the vitamin D sufficient group in relation to changes in FEV₁% predicted.

Table 2. Significant miRNAs in 55 participants from the vitamin D insufficient group.

miRNA	Univariate				Multivariate*
	β	2.50%	97.50%	P value	β	2.50%	97.50%	P value
hsa-miR-186-5p	−6.69	−10.46	−2.92	0.001	−6.07	−9.92	−2.22	0.004
hsa-miR-125a-5p	6.48	2.32	10.64	0.004	6.52	2.23	10.80	0.005
hsa-miR-107	−6.18	−10.50	−1.86	0.007	−6.45	−10.71	−2.19	0.005
hsa-miR-92b-3p	3.87	0.06	7.68	0.05	5.13	1.21	9.04	0.014
hsa-miR-181a-5p	−8.68	−15.01	−2.34	0.010	−8.32	−15.05	−1.59	0.020
hsa-miR-125b-5p	5.03	0.35	9.71	0.041	5.55	0.94	10.15	0.023
hsa-miR-502-3p	−5.46	−9.53	−1.38	0.012	−5.11	−9.35	−0.87	0.023
hsa-miR-93-5p	−3.78	−6.73	−0.82	0.016	−3.86	−7.09	−0.62	0.025
hsa-miR-320a	−5.83	−10.72	−0.94	0.024	−5.78	−11.10	−0.45	0.040
hsa-miR-744-5p	4.21	0.38	8.03	0.036	4.30	0.32	8.27	0.040
hsa-miR-320d	−3.75	−7.16	−0.35	0.036	−3.97	−7.67	−0.28	0.041
hsa-miR-363-3p	−5.61	−10.31	−0.90	0.024	−5.27	−10.27	−0.26	0.046

^*Adjusted for age, sex, race and body mass index (BMI).

Significant miRNAs by vitamin D interaction

11 miRNAs were significantly associated with changes in FEV₁% predicted when considering their interaction with vitamin D. Of these, three miRNAs showed positive associations, while eight were negatively associated (table 3). For instance, hsa-miR-125a-5p exhibited the strongest positive effect (β=6.92, p=0.008), where increased expression of this miRNA was linked to a relative improvement in FEV₁% predicted between vitamin D insufficient and sufficient groups (figure 1A). hsa-miR-181a-5p had the strongest negative effect (β=−7.48, p=0.027), with higher expression correlating to a relative decline in FEV₁% predicted (figure 1B). Six miRNAs, including hsa-miR-186-5p, hsa-miR-125a-5p, hsa-miR-107, hsa-miR-181a-5p, hsa-miR-320a and hsa-miR-320d, remained significant in the vitamin D insufficient group, as noted in table 2.

Table 3. Significant miRNAs in 187 participants based on vitamin D interaction analysis.

miRNA	Univariate interaction				Multivariate interaction*
	β	2.50%	97.50%	P value	β	2.50%	97.50%	P value
hsa-miR-186-5p	−7.29	−11.72	−2.87	0.002	−6.25	−10.51	−1.99	0.005
hsa-miR-125a-5p	7.67	2.50	12.84	0.004	6.92	1.87	11.96	0.008
hsa-miR-320a	−6.71	−11.83	−1.59	0.011	−6.69	−11.70	−1.68	0.010
hsa-miR-101-3p	−5.45	−10.04	−0.86	0.021	−5.75	−10.19	−1.30	0.012
hsa-miR-320d	−4.75	−8.46	−1.05	0.013	−4.55	−8.18	−0.92	0.015
hsa-miR-320b	−4.48	−8.33	−0.63	0.024	−4.68	−8.47	−0.89	0.017
hsa-miR-181a-5p	−8.86	−15.58	−2.14	0.011	−7.48	−14.06	−0.90	0.027
hsa-miR-28-3p	5.24	0.23	10.25	0.042	5.17	0.45	9.88	0.033
hsa-miR-107	−6.77	−11.75	−1.79	0.008	−5.14	−9.95	−0.33	0.038
hsa-miR-502-3p	−5.07	−9.50	−0.63	0.027	−4.40	−8.70	−0.11	0.046
hsa-miR-375	4.74	0.40	9.07	0.034	4.30	0.07	8.53	0.048

^*Adjusted for age, sex, race and body mass index (BMI).

Figure 1. Scatter plots of (A) hsa-miR-125a-5p and (B) hsa-miR-181a-5p in the interaction analysis. Red points denote subjects in the vitamin D insufficient group, and green points denote those in the vitamin D sufficient group. The red line indicates the regression line for the vitamin D insufficient group, while the green line represents the regression line for the vitamin D sufficient group. FEV₁, forced expiratory volume in 1 s.

Replication of miRNAs in the ICS usage and vitamin D insufficient group in GACRS

Since GACRS is a cross-sectional study without follow-up data, we validated the association between the significant miRNAs in table 2 and FEV₁% predicted in GACRS. We identified six miRNAs with p values <0.1 (online supplemental table S1), including hsa-miR-181a-5p, hsa-miR-320a and hsa-miR-363-3p, which had p values <0.05. This replication suggests that these miRNAs are associated with ICS response in the context of ICS usage and vitamin D insufficiency.

Differentially expressed genes in LCL gene expression arrays

We established a cell model using immortalised LCLs from 22 participants in the ICS treatment group of the CAMP study (figure 2A). A total of 220 genes were differentially expressed in DEX-treated LCLs in vitro, with |logFC|≥0.5 and an adjusted p value <0.05 (figure 2B). Additionally, 212 genes were differentially expressed in vitamin D-treated LCLs in vitro, with |logFC|≥0.25 and adjusted p value <0.05 (figure 2C). Of these, 20 genes were significantly differentially expressed in both DEX and vitamin D-treated LCLs (figure 2D). These genes were enriched in immune and inflammation-related functions, such as leucocyte migration, positive regulation of cytokine production and nuclear factor kappa-B signal transduction (online supplemental figure S4). We also identified 2900 target genes with strong evidence from miRTarBase, 79 of which overlapped with the differentially expressed genes (DEG) in either the DEX or vitamin D groups (figure 2D).

Figure 2. (A) Workflow of the microarray functional validation. (B) Volcano plot displaying differentially expressed genes (DEGs) in the dexamethasone (DEX)-treated lymphoblastoid cell line (LCL) group. (C) Volcano plot of DEGs in the vitamin D-treated LCL group. The logarithmic fold change is capped at 2 to provide a clearer view of the x-axis. (D) Venn diagram (VD) showing overlap of DEGs in DEX and vitamin D-treated LCLs with the target genes of miRNAs. CAMP, Childhood Asthma Management Program.

Analysis of miRNA and target gene mixed network

The total mixed network is illustrated in figure 3, which encompassed both miRNA target gene regulation and functional associations between target genes. Four clusters (online supplemental figure S5), including cluster hsa-miR-125a-5p, cluster hsa-miR-181a-5p, cluster hsa-miR-101-3p and cluster hsa-miR-107, were all significantly enriched in GO:1903706 (regulation of haemopoiesis) and GO:1902105 (regulation of leucocyte differentiation) (online supplemental figure S6). Transforming growth factor beta receptor 3 (TGFBR3), also known as betaglycan, and B and T lymphocyte associated (BTLA), also known as CD272, were differentially expressed in both DEX and vitamin D-treated LCLs and were present in all four clusters (online supplemental figure S5). hsa-miR-181a-5p exhibited a significant upregulatory effect on DEGs in the vitamin D group, as identified by both the KS (Kolmogorov–Smirnov) test and the GSEA-like approach (online supplemental table S2).

Figure 3. Comprehensive miRNA and target gene interaction network. Diamonds represent miRNAs, while circles represent target genes. Pink indicates differentially expressed genes (DEGs) in dexamethasone (DEX)-treated lymphoblastoid cell lines (LCLs), green represents DEGs in vitamin D-treated LCLs and purple denotes DEGs present under both conditions. Red borders indicate upregulation in both DEX and vitamin D conditions, while green borders show downregulation in both. Blue borders indicate downregulation in the DEX condition and upregulation in the vitamin D condition, and yellow borders show upregulation in the DEX condition and downregulation in the vitamin D condition.

Prediction of ICS response based on miRNAs

We assessed the prognostic value of two miRNAs, hsa-miR-125a-5p and hsa-miR-181a-5p, which exhibited the strongest effects in opposing directions within the vitamin D insufficient group. These miRNAs were used to differentiate between the highest tertile (mean±SD FEV₁% change, 18.44±9.89) and the lowest tertile (mean±SD FEV₁% change, −8.31±8.15). The AUROC curve was used to evaluate the performance of the multivariate logistic regression model. The AUROC of hsa-miR-125a-5p was 0.83 (95% CI 0.74 to 1, Hedges’ g=1.43), while for hsa-miR-181a-5p was 0.76 (95% CI 0.7 to 0.98, Hedges’ g=0.98). Combined, the AUROC for both miRNAs reached 0.86 (95% CI 0.79 to 1, Hedges’ g=1.59), indicating good to excellent prognostic ability (figure 4). If confounding factors are not taken into account and the ICS response is predicted using the two miRNAs, the AUROC still reaches 0.83 (95% CI 0.65 to 0.99, Hedges’ g=1.37) (online supplemental figure S7). The model discrimination and calibration metrics were reported in online supplemental table S3. In contrast, hsa-miR-125a-5p and hsa-miR-181a-5p demonstrated low AUROC values of 0.54 (95% CI 0.47 to 0.66) in the vitamin D sufficient group (online supplemental figure S8).

Figure 4. Prediction of hsa-miR-125a-5p and hsa-miR-181a-5p using a logistic regression model. The red curve shows the performance of hsa-miR-125a-5p, the green curve denotes hsa-miR-181a-5p and the blue curve illustrates the combined effect of hsa-miR-125a-5p and hsa-miR-181a-5p. ROC, receiver operating characteristic.

Discussion

In this study, we investigated the relationship between miRNA expression and ICS response at varying serum vitamin D levels in children who had asthma from the intervention arm of CAMP study. We identified 12 miRNAs with significant associations in the vitamin D insufficient group, whereas no miRNAs were significant in the vitamin D sufficient group. Three miRNAs were successfully replicated for their significant association with ICS response in an independent cohort, GACRS. 11 miRNAs showed notable interactions with vitamin D levels. We used microarray analysis to identify DEGs in LCLs treated with either DEX or vitamin D, isolated and immortalised from 22 CAMP participants. A mixed network, encompassing both miRNA target gene regulation and gene-gene functional associations, was constructed to connect miRNAs to DEGs for DEX or vitamin D. We identified four functional gene clusters, all of which were associated with the regulation of haemopoiesis and leucocyte differentiation.

Hsa-miR-125a-5p exhibits the strongest positive effect in both the main analysis (β=6.52; p=0.005) and the interaction analysis (β=6.92; p=0.008). This suggests that increased expression of hsa-miR-125a-5p could enhance lung function in children with asthma who have insufficient serum vitamin D. Cay et al found that miR-125a-5p was underexpressed in lung samples from a house dust mite (HDM) mouse model and demonstrated strong predictive power for asthma.²¹ Banerjee et al reported that miR-125a-5p promotes M2 macrophage polarisation and exhibits anti-inflammatory effects.²² Chen et al showed that miR-125a-5p is highly expressed in oxidised low-density lipoprotein-stimulated monocytes/macrophages and reduces the secretion of inflammatory cytokines such as IL-2, IL-6, tumour necrosis factor-alpha and TGF-β.²³ Naguib et al observed that miR-125a-5p was reduced in patients with immune thrombocytopenia, but was significantly higher in patients who responded to steroid treatment.²⁴

Hsa-miR-181a-5p exhibits the strongest negative effect in both the main analysis (β=−8.32; p=0.02) and the interaction analysis (β=−7.48; p=0.027), indicating that increased expression of hsa-miR-181a-5p is associated with a decreased ICS response. hsa-miR-181a-5p also exhibits a significant upregulatory effect on DEGs in the vitamin D group, suggesting a complex indirect regulatory mechanism. Hong et al confirmed the interaction between hsa-miR-181a-5p and vitamin D in relation to a history of childhood asthma hospitalisation.¹⁵ Mirra et al observed significantly lower expression of miR-181a-5p in serum samples from patients with asthma.²⁵ Similarly, Liu et al found reduced miR-181a-5p expression in regulatory T cells (Tregs) of children with allergic rhinitis, noting that miR-181a-5p levels were positively correlated with IL-10 and TGF-β and negatively correlated with total nasal severity scores.²⁶ Additionally, lower miR-181a-5p expression was observed in mice with allergic rhinitis, and overexpression of miR-181a-5p was shown to alleviate allergic behaviours and delay asthma progression.²⁷

Kho et al found that hsa-miR-186-5p was associated with the FEV₁/forced vital capacity ratio,²⁸ and Sharma et al identified hsa-miR-186-5p as a central hub in association with metabolites in childhood asthma.²⁹ Hu et al reported a positive correlation between miR-107 and the absolute count of peripheral eosinophils,³⁰ while Kim et al suggested that this effect might be mediated by HDM sensitisation in childhood asthma.³¹ Additionally, Boudewijn et al observed that the miR-320 family, including miR-320a, miR-320b, miR-320c and miR-320d, was upregulated among subjects with complete remission compared with those with persistent asthma.³²

Vitamin D plays a crucial role in the human immune system, and deficiencies in vitamin D are linked to an increased susceptibility to diseases such as asthma.³³ Our study identified 12 miRNAs significantly associated with improved ICS response in the vitamin D insufficient group, while no miRNAs were significant in the vitamin D sufficient group. This suggests that vitamin D may act as a modulator of immune system or inflammatory responses. Chauss et al demonstrated that autocrine vitamin D can facilitate the transition from proinflammatory interferon-γ+ TH1 cells to suppressive IL-10+ cells.³⁴ Grund et al observed that vitamin D supplementation could shift proinflammatory immune responses to anti-inflammatory ones in lung innate lymphoid cells and tissue-resident memory cells via PRDM1.³⁵ Moreover, PRDM1 was differentially expressed in DEX-treated LCLs (figure 2B) and regulated by hsa-miR-125a-5p, hsa-miR-101-3p, BTLA and syndecan-1 (SDC1) (figure 3). BTLA is a critical costimulatory molecule that induces immunosuppression by inhibiting B and T cell activation and proliferation.³⁶ SDC1 is a transmembrane heparan sulfate proteoglycan that can reduce allergic lung inflammation by suppressing CC chemokine-mediated Th2 cell recruitment to the lungs.³⁷ Additionally, the four clusters identified through the random walk approach (online supplemental figure S2) were enriched in pathways related to the regulation of haemopoiesis and leucocyte differentiation, suggesting that vitamin D may influence ICS response through the leucocyte differentiation pathway modulated by miRNAs.

TGFBR3 (betaglycan) is differentially expressed in both DEX and vitamin D-treated LCLs and regulated by hsa-miR-125a-5p, hsa-miR-181a-5p, hsa-miR-101-3p and hsa-miR-107 (figure 3). Xystrakis et al confirmed that TGFBR3 was significantly downregulated in CD4+ Tregs stimulated by DEX and calcitriol.¹⁰ Kim et al found that genetic variations in TGFBR3 were associated with a predisposition to asthma in the Korea population.³⁸ TGF-β plays a significant role in immune cell differentiation and has both proinflammatory and anti-inflammatory effects in asthma, such as activating Th17 lymphocytes, promoting Treg generation and inhibiting the differentiation of Th1 and Th2 lymphocytes.³⁹ TGFBR3 is the most widely expressed TGF-β receptor without kinase activity and can modulate TGF-β signalling as an antagonist of other TGF-β receptors.⁴⁰ Our analysis suggests that miRNAs influencing the effect of vitamin D on ICS response may play a role in leucocyte differentiation via the TGF-β signalling pathway.

Our study has several limitations. First, the sample size is relatively small, with only 187 subjects included; however, the microarray data from LCLs of 22 subjects in the same cohort help strengthen the functional validation of the significant miRNAs. Second, the replication of the association between target miRNAs and ICS response in the GACRS cohort is approximate, as GACRS is a cross-sectional study. Therefore, further validation through large-scale studies with follow-up data or molecular experiments is needed to confirm these findings.

In summary, our study successfully associates miRNA expression with ICS response when stratified by serum vitamin D levels, as well as in the interaction analysis. Specifically, we identified 11 miRNAs with significant vitamin D effect modification, and our in vitro microarray experiment demonstrated that the influence of vitamin D on ICS response may be modulated by miRNAs regulating leucocyte differentiation. Our findings suggest that miRNAs have potential as biomarkers for predicting ICS response in case of vitamin D insufficiency, with a prognostic accuracy reaching 0.86 in our study. However, these results are preliminary and require further validation in larger independent cohorts to confirm their clinical utility.

Data availability statement

Data are available upon reasonable request.