Breast milk-derived extracellular vesicle miRNAs are associated with maternal asthma and atopy

Abstract

Background:

Breast milk-derived extracellular vesicle (EV) miRNAs may program child health outcomes associated with maternal asthma and atopy. The authors investigated associations between maternal asthma/atopy and EV miRNAs in the Programming of Intergenerational Stress Mechanisms cohort.

Methods:

Breast milk-derived EV miRNAs collected 6.1 ± 5.9 weeks postnatally (n = 80 mothers) were profiled using the TaqMan OpenArray Human MicroRNA Panel. The authors assessed associations using adjusted robust regression.

Results:

Nine EV miRNAs were associated with asthma during pregnancy (a priori criteria: nominal p < 0.05; |B_regression| >0.2). miR-1290 was associated with asthma and atopy during pregnancy (p < 0.05; |B_regression| >0.2). Enriched Kyoto Encyclopedia of Genes and Genomes pathways included TGF-β signaling and extracellular matrix–receptor interaction (false discovery rate <0.05).

Conclusion:

In this study, maternal asthma and atopy were associated with breast milk-derived EV miRNAs. Additional studies are needed to understand whether EV miRNAs have direct effects on infant and child health.

Plain language summary

Maternal asthma is associated with child health outcomes, although the biological mechanisms involved are not fully understood. miRNAs are small molecules involved in regulating gene expression. miRNAs packaged into membrane-bound particles called extracellular vesicles (EVs) are present in human breast milk and may pass from mother to infant to signal which genes to translate into proteins. This study investigated the extent to which maternal asthma and atopy influenced levels of 130 EV miRNAs measured in breast milk. Nine EV miRNAs were associated with maternal asthma during pregnancy, and one EV miRNA was associated with maternal atopy. miRNAs associated with asthma target genes in pathways related to asthma; however, future research is needed to determine whether changes in breast milk-derived EV miRNAs impact child health.

Maternal asthma is associated with adverse infant and child health outcomes, including preterm and small-for-gestational-age births and perinatal mortality [1]. Maternal asthma and atopy have also been associated with respiratory health issues in offspring, including bronchitis [2] and acute respiratory tract infections in infants [3] and recurrent wheeze [4,5], atopy [6] and asthma [4,5] in childhood. The biological mechanisms through which maternal asthma impacts child health are not fully understood, although there is increasing evidence that mother–infant biochemical communication via breast milk may represent one pathway by which maternal health impacts child development and health [7–9]. Cohort studies have found that maternal asthma may modify associations between breastfeeding and offspring health, with greater protective effects of breastfeeding on wheeze and lung function among mothers with asthma [10,11].

Breast milk contains regulatory biomolecules, including hormones, growth factors, cytokines and miRNAs, which may impact early life programming [12–14]. miRNAs are ncRNA molecules measuring approximately 22 nucleotides in length [15]. Mature miRNAs can influence post-transcriptional gene repression by inhibiting translation or guiding mRNA degradation [16]. An individual miRNA may have hundreds of target mRNA molecules [17]. miRNAs in breast milk may be encapsulated in breast milk-derived extracellular vesicles (EVs). EVs protect miRNAs from degradation in the GI tract, enabling miRNAs to be taken up by intestinal cells [18,19]. Infants also have low levels of gastric acid and digestive enzymes [20] as well as immature intestinal cells with high endocytic activity, facilitating intestinal macromolecule uptake [21]. In an in vitro study of simulated digestion, human exosomal miRNAs were shown to enter human intestinal cells [22]. In animal models, labeled EV miRNAs have been shown to accumulate in the intestinal mucosa, spleen, liver, heart and brain of mice [23]. In addition, ingestion by rodents of bovine EVs containing miRNAs has been shown to elicit changes in the gut microbiome [24], muscle growth [25], spatial learning and memory [26] and modify the inflammatory response to dust [27]. The packaging of miRNAs into EVs may also facilitate transport to particular cells; antigen-specific surface antibody light chains may allow for targeting of immune cells [28] to elicit a biological response even at low miRNA concentrations [29].

Intervention studies of the effects of breast milk-derived EV miRNAs on infant health in humans have been limited in part because of ethical concerns, and prior research has largely focused on linking miRNAs identified in breast milk to mRNA targets using computational approaches [30]. miRNAs with medium to high abundance in human breast milk have been linked to diverse functions, including lipid metabolism, glucose metabolism, gut maturation and neurogenesis [31]. A large proportion of breast milk miRNAs have been associated with immune function and the inflammatory response. In humans, 68% of pre-miRNAs associated with immune function have been found to be enriched in breast milk exosomes [32], and miRNAs related to T- and B-cell maturation and regulation and the innate immune response have been found to be highly expressed [32–34]. Breast milk-derived miRNAs may also influence cytokine expression. For example, miR-181a, which has been identified as highly expressed in human breast milk exosomes [31], may have anti-inflammatory effects by targeting IL1A mRNA [35].

miRNAs in breast milk are influenced by maternal factors. The authors previously identified associations between stressful events experienced over mothers' lifetimes and during pregnancy and breast milk-derived EV miRNA profiles in the Programming of Intergenerational Stress Mechanisms pregnancy cohort (n = 74) [36]. Differential expression of breast milk exosomal miRNAs has been associated with maternal Type 1 diabetes (n = 52) [37]. In addition, maternal obesity has been found to be associated with breast milk miRNAs involved in adipogenesis and glucose metabolism (n = 60) [38]. In a cohort of 59 mothers, maternal obesity modified the association between breast milk leptin and adiponectin concentration and miRNA expression [39].

Considering the epidemiological associations between maternal asthma, breastfeeding and child health, the authors hypothesized that maternal asthma and atopy would be associated with breast milk-derived EV miRNA profiles. Leveraging data from the Programming of Intergenerational Stress Mechanisms cohort, the authors analyzed associations between active maternal asthma and atopy during pregnancy and EV miRNA expression levels. Associations between inactive maternal asthma and atopy during pregnancy and EV miRNAs were included as secondary outcomes.

Methods

Study design

The Programming of Intergenerational Stress Mechanisms cohort is a Boston- and New York City-based prospective cohort of mother–child pairs recruited in pregnancy designed to examine associations among perinatal stress, other environmental exposures (e.g., air pollutants, chemicals, nutrition) and child development [40]. Analyses presented here include mothers recruited in New York City, as this site was still actively recruiting during pregnancy and supplemental funding was obtained to add breast milk to sample collection protocols. Mothers were recruited by the Icahn School of Medicine at Mount Sinai from November 2012 to August 2014 and were eligible if they were ≥18 years old, in the first or second trimester of pregnancy, pregnant with a singleton fetus and English- or Spanish-speaking. Exclusion criteria were reporting seven or more alcoholic drinks per week before recognition of pregnancy, alcohol or drug use during pregnancy and HIV-positive status.

Ethics

The institutional review board at the Icahn School of Medicine at Mount Sinai approved the study protocol. Mothers provided written consent in their primary language.

Sociodemographics & maternal characteristics

Sociodemographic information was collected during a structured survey at enrollment. Maternal race/ethnicity was categorized as Black, Hispanic and non-Hispanic White/other. Education was categorized as ≤completing high school/General Educational Development Test or >high school/General Educational Development Test. Self-reported average weight prior to the current pregnancy was used to calculate pre-pregnancy BMI, and women were classified as obese (≥30 kg/m²) or not obese (<30 kg/m²). Prenatal tobacco smoke exposure was defined as self-reported maternal smoking or environmental tobacco smoke exposure ≥1 h per week.

Maternal asthma & atopy

Briefly, mothers were classified as having asthma based on the question “Have you ever had asthma?” and self-reported history of treatment for asthma (e.g., emergency visit or hospitalization), use of asthma medications and asthma symptoms (e.g., wheezing/whistling or nighttime awakening with wheeze or cough). Mothers with a history of asthma were classified as having active asthma during pregnancy based on the question “Do you still have asthma?” and asthma treatment, medication use or symptoms reported within the past year. Atopy (including asthma, allergic rhinitis and atopic dermatitis) and symptoms during pregnancy were similarly defined based on self-reported diagnosis, treatment and medication use ever and within the past year. Specific survey questions used to determine maternal asthma and atopy status are included in the Supplementary Methods.

Sample collection, EV isolation & RNA extraction

Breast milk miRNA analysis has been described in detail [36]. Briefly, breast milk samples were collected in mothers' homes (n = 80) 6.1 ± 5.9 weeks postpartum (Supplementary Figure 1). Before the first morning feeding, mothers washed their breasts, and 3–10 ml of breast milk was collected with a manual pump. Samples were stored at 4°C and transported to the laboratory, where they were frozen at -80°C until EV isolation.

Stored breast milk samples were thawed on ice, and 1400–1500 μl was centrifuged to separate a top layer of fat. Although miRNAs contained in lipid fraction might have been lost [41], the lipid layer was removed to limit variation in miRNA profiles due to differences in the ratio of skim milk to lipids. Samples were again centrifuged at a faster speed to remove cellular debris and apoptotic bodies. The exoEasy Maxi Kit (QIAGEN, MD, USA) was used to isolate exosomes and other EVs using membrane affinity spin columns [42]. The supernatant (approximately 50–60% of the starting volume) was aspirated and mixed with Buffer XBP (QIAGEN) in a 1:1 ratio and centrifuged on an exoEasy spin column. The residual liquid was removed, bound EVs were washed with Buffer XWP (QIAGEN), samples were centrifuged again, flow-through was discarded and EVs were eluted with 400 μl Buffer XE (QIAGEN). Although membrane affinity spin columns do not distinguish EVs by size or cellular origin, the authors have previously demonstrated common CD63 and CD9 surface markers and size distribution (mean diameter approximately 100 nm) of isolated EVs [13].

Total RNA was extracted from EVs using the miRNeasy Serum/Plasma Kit (QIAGEN). For each sample, 200 μl EVs was mixed with 1000 μl QIAzol (QIAGEN). After a brief incubation, 200 μl chloroform was added to each sample, vortexed and then centrifuged. The aqueous phase was recovered, mixed with 2100 μl 100% ethanol and centrifuged on an RNeasy MinElute column (QIAGEN). The flow-through was discarded, and samples were washed once with 500 μl Buffer RWT (QIAGEN) and twice with 500 μl Buffer RPE (QIAGEN). Samples were centrifuged with 500 μl 80% ethanol, flow-through was discarded and RNA was eluted with 14 μl RNase-free water.

EV miRNA profiling

RNA samples labeled with deidentified identifiers were shipped to the Genomics Core Facility at the University of Utah (UT, USA) for measurement of miRNAs. Samples were processed with RNA Clean & Concentrator kits (Zymo Research Corporation, CA, USA), and RNA concentration and quality were determined with a Bioanalyzer (Agilent Technologies, Inc., CA, USA). Sufficient RNA was available for 75 mothers (Supplementary Figure 1). Samples were diluted to 34 ng RNA/ul, with the exception of 11 samples with low RNA concentrations (4.76–33 ng RNA/μl). RNA was analyzed with the TaqMan OpenArray Human MicroRNA Panel (Thermo Fisher Scientific, MA, USA). Two primer pools of 170 ng were used, and Applied Biosystems Megaplex RT and PreAmp (Thermo Fisher Scientific) reactions were performed in a 96-well plate on an Applied Biosystems 9700 PCR thermal cycler (Thermo Fisher Scientific). The product was diluted, mixed with TaqMan OpenArray Real-Time PCR Master Mix (Thermo Fisher Scientific), and 5 μl of each sample was transferred to eight wells in a 384-well plate. Batches of four plates were run on the QuantStudio 12K Flex Real-Time PCR System (Thermo Fisher Scientific). Samples were run in a single day; nine samples were rerun at a later date because of a defective array.

miRNA data processing & quality control

Control and incorrectly annotated probes were removed, resulting in 752 probes measuring known human miRNAs. Relative cycle threshold (C_q) values with amplification scores <1.1, C_q confidence <0.8, and values <10.1 (i.e., value of the control probe U6 rRNA_001973) or >35 (i.e., non-detect value) were replaced with missing. A total of 550 miRNAs were detected in at least one sample (Supplementary Spreadsheet 1).

Missing C_q values were treated as non-detectable and set to 35. Reference genes are not available for breast milk-derived miRNAs, and data were therefore normalized using the global mean method [43]. The ΔC_q values were calculated as follows: for each miRNA i in sample j with n miRNAs measured, ΔC_qi,j = C_qi,j - geometric mean of C_q1:n,j. Negative ΔC_q values (-ΔC_q) were used as the primary measure of analysis so that negative values indicated lower expression relative to the geometric mean within a sample and positive values indicated higher expression. Median (interquartile range [IQR]) and mean (standard deviation) of -ΔC_q values are included in Supplementary Spreadsheet 1.

Data analysis

Maternal asthma and atopy data were missing for one mother; therefore, analyses included 74 mother–infant pairs. The authors calculated descriptive statistics using median and IQR for continuous variables and frequency and proportion for categorical variables.

Analyses included 130 EV miRNAs detected in ≥80% of samples. Maternal asthma and atopy were classified as never (reference), inactive during pregnancy and active during pregnancy. Associations with EV miRNA expression levels (i.e., -ΔC_q values among 130 EV miRNAs detected in ≥80% of samples) were analyzed using robust linear regression implemented with the rlm function in the R MASS package [44] using the M estimator. The p-values were estimated using the coeftest function in the R lmtest package [45] with the vcovHC covariance matrix estimation function with White's estimator in the R sandwich package [44,46]. To address limitations in power due to small sample size, the authors used the a priori criteria, incorporating a threshold for effect size [47] of nominal p < 0.05 and |B_regression| >0.2 (i.e., approximately 1% difference in the range of -ΔC_q values) to determine significance of associations [36]. Fully adjusted models included infant sex, maternal race and education and postpartum week of breast milk collection.

Principal component analysis was performed on -ΔC_q values to characterize expression profiles of the 130 EV miRNAs detected in ≥80% of samples. Associations between the top 13 principal components (PCs) explaining 80% of variance and maternal asthma and atopy variables and covariates were evaluated using Wilcoxon rank-sum or Kruskal–Wallis tests for categorical variables and Spearman's correlation for continuous variables.

The authors conducted sensitivity analyses, adjusting for pre-pregnancy obesity. Because of observed associations between postpartum week of breast milk collection and EV miRNA expression levels [36], the authors also performed sensitivity analyses, excluding five samples collected >12 weeks postpartum. All described analyses were performed in R 4.0.2 [48].

To determine predicted mRNA targets and biological roles of differentially expressed EV miRNAs (p < 0.05 and |B_regression| >0.2 for maternal asthma during pregnancy vs never diagnosed), the authors performed Kyoto Encyclopedia of Genes and Genomes (KEGG) [49] enrichment analyses. Analyses were implemented with DIANA-miRPath v.3 [50] using the TarBase v.7.0 database and Fisher's exact test with a conservative adjustment, and results were combined using an a posteriori union of pathways. TarBase v.7.0 indexes miRNA–mRNA interactions determined from experimental data [51]. Pathways with p < 0.05 after a false discovery rate correction were considered significant.

Results

Participant characteristics

A total of 74 mother–infant pairs were included in the analyses (Supplementary Figure 1). Characteristics of mother–infant pairs are shown in Table 1. Mothers had a median (IQR) age of 27.3 (23.1–32.1) years. Over half self-identified as Black (56.8%), and 36.5% self-identified as Hispanic. The majority (70.2%; n = 52) of mothers were never diagnosed with asthma, whereas 24.3% (n = 18) and 5.4% (n = 4) reported, respectively, having active and inactive asthma during pregnancy. Approximately half (48.6%; n = 36) of mothers were never diagnosed with atopy, whereas 44.6% (n = 33) and 6.8% (n = 5) reported, respectively, active and inactive atopy during pregnancy. Breast milk was collected at a median (IQR) of postpartum week 4.2 (2.6–7.9), and the median (IQR) number of miRNAs detected per sample was 182.0 (154.5–193.0).

Table 1. . Characteristics of mother–infant pairs.

	n (%)
Maternal
Age at delivery, years, median (IQR)	27.3 (23.1–32.1)
Education
≤High school diploma or GED	26 (35.1)
>High school diploma or GED	48 (64.9)
Prenatal tobacco smoke exposure
Environmental exposure†	19 (27.9)
Maternal smoking	10 (13.5)
Race/ethnicity
White/other	5 (6.8)
Black	42 (56.8)
Hispanic	27 (36.5)
Asthma
Active during pregnancy	18 (24.3)
Inactive during pregnancy	4 (5.4)
Never diagnosed	52 (70.3)
Atopy
Active during pregnancy	33 (44.6)
Inactive during pregnancy	5 (6.8)
Never diagnosed	36 (48.6)
Infant
Sex
Male	41 (55.4)
Female	33 (44.6)
Gestational age, weeks, median (IQR)	39.0 (38.0–40.0)
Preterm, <37 weeks' gestation	11 (14.9)
Birth weight, kg, median (IQR)	3030 (2795–3425)
Low birth weight, <2500 g	9 (12.2)
Postpartum week of breast milk collection, median (IQR)	4.2 (2.6–7.9)
Number of miRNAs detected, median (IQR)	182.0 (154.5–193.0)

^†Defined as being exposed to environmental tobacco smoke ≥1 h per week during pregnancy.

GED: General Educational Development Test; IQR: Interquartile range.

EV miRNA expression levels

Median (IQR) and mean (standard deviation) of -ΔC_q values are listed in Supplementary Spreadsheet 1. Analyses of EV miRNA expression levels were restricted to miRNAs detected in ≥80% of samples. In principal component analysis, 80% of the variance in EV miRNA expression levels was explained by the top 13 PCs. PCs 7 and 12 were associated with maternal atopy (Kruskal–Wallis test p < 0.05), and PC 12 was associated with maternal asthma (Kruskal–Wallis test p < 0.05) (Figure 1). Moreover, at p < 0.10, PC 11 was associated with maternal asthma. The authors also identified associations between covariates and PCs (infant sex: PC 1; maternal education: PC 4; postpartum week of breast milk collection: PCs 1, 3, 5, 6 and 9; p < 0.05).

Figure 1. . Associations between extracellular vesicle miRNA principal components, maternal asthma and atopy and characteristics of mother–infant pairs.

PCA was performed on 130 miRNAs detected in ≥80% of samples. Associations were assessed using Wilcoxon rank-sum or Kruskal–Wallis test for categorical variables and Spearman's correlation for continuous variables.

*p < 0.10; **p < 0.05.

PC: Principal component; PCA: Principal component analysis.

Associations between individual EV miRNA expression levels and maternal asthma and atopy (active during pregnancy and inactive during pregnancy versus never diagnosed) were assessed using robust linear models and the a priori criteria for significance of nominal p < 0.05 and |B_regression| >0.2. Final models were adjusted for infant sex, maternal race and education and postpartum week of breast milk collection. Results from all robust linear regression models are included in Supplementary Spreadsheets 2 & 3.

In adjusted models, expression of nine EV miRNAs was associated with asthma that was active during pregnancy versus never diagnosed (p < 0.05; |B_regression| >0.2). miRNAs associated with active asthma during pregnancy are shown in Table 2, a volcano plot is shown in Figure 2 and raincloud plots are included in Supplementary Figure 2. Eight of the significant EV miRNAs had lower mean expression levels among mothers with asthma that was active during pregnancy versus never diagnosed, and overall, these miRNAs were positively associated with each other (Supplementary Figure 3). miR-1290 was positively associated with active asthma and was negatively correlated with five of the significant miRNAs (miR-331-3p, mir-200a-3p, miR-106b-5p, miR-324-5p, miR-148b-3p). A total of 31 miRNAs were associated with asthma that was inactive during pregnancy (p < 0.05; |B_regression| >0.2) (Supplementary Spreadsheet 2); however, effect size estimates may have been influenced by the small number of mothers in this category (n = 4). Three EV miRNAs were associated with asthma that was both active and inactive during pregnancy: miR-324-5p (B [95% CI]_active = -0.61 [-1.14 to -0.07]; B [95% CI]_inactive = 0.66 [0.13–1.19]), miR-30a-3p (B [95% CI]_active = -0.55 [-1.04 to -0.05]; B [95% CI]_inactive = 0.45 [0.03–0.87]) and miR-148b-3p (B [95% CI]_active = -0.76 [-1.52 to -0.001]; B [95% CI]_inactive = 0.88 [0.13–1.64]).

Table 2. . Extracellular vesicle miRNAs associated with maternal asthma and atopy that were active during pregnancy versus never diagnosed.

EV miRNA	B	95% CI	p-value
Maternal asthma that was active during pregnancy versus never diagnosed
miR-191-5p	-0.65	-1.17 to -0.13	0.015
miR-200a-3p	-0.61	-1.11 to -0.12	0.015
miR-324-5p	-0.61	-1.14 to -0.07	0.026
miR-29a-5p	-1.06	-1.99 to -0.12	0.027
miR-331-3p	-0.66	-1.24 to -0.07	0.028
miR-30a-3p	-0.55	-1.04 to -0.05	0.030
miR-1290	2.14	0.19–4.08	0.032
miR-106b-5p	-0.68	-1.33 to -0.04	0.039
miR-148b-3p	-0.76	-1.52 to -0.001	0.050
Maternal atopy that was active during pregnancy versus never diagnosed
miR-1290	2.38	0.15 to 4.62	0.037

Associations of 130 miRNAs detected in ≥80% of samples were evaluated using robust linear regression adjusted for infant sex, maternal race and education and postpartum week of breast milk collection. EV miRNAs that met a priori criteria for significance (p < 0.05; |B_regression| >0.2) are listed.

EV: Extracellular vesicle.

Figure 2. . Volcano plots of associations between EV miRNA expression levels and (A) maternal asthma that was active during pregnancy versus never diagnosed and (B) maternal atopy that was active during pregnancy versus never diagnosed.

Associations of 130 miRNAs identified in ≥80% of samples were evaluated using robust linear regression adjusted for infant sex, maternal race and education and postpartum week of breast milk collection. A priori criteria for significance (p < 0.05; |B_regression| >0.2) are indicated by dashed red lines.

EV: Extracellular vesicle.

In adjusted models, atopy that was active during pregnancy versus never diagnosed was associated with expression levels of one EV miRNA – miR-1290 (p < 0.05; |B_regression| >0.2) (Table 2, Figure 2 & Supplementary Figure 4). miR-1290 was also associated with asthma that was active during pregnancy with a similar effect size (B [95% CI]_asthma = 2.14 [0.19–4.08]; B [95% CI]_atopy = 2.14 [0.15–4.62]). Seven miRNAs were associated with maternal atopy that was inactive during pregnancy versus never diagnosed (p < 0.05; |B_regression| >0.2) (Supplementary Spreadsheet 3), although the small number of mothers with inactive atopy should be noted (n = 5). Five miRNAs were positively associated with both asthma and atopy that were inactive during pregnancy versus never diagnosed: miR-193a-5p (B [95% CI]_asthma = 1.22 [0.60–1.84]; B [95% CI]_atopy = 1.09 [0.36–1.82]), miR-224-5 (B [95% CI]_asthma = 1.65 [0.68–2.63]; B [95% CI]_atopy = 0.91 [0.27–1.55]), miR-24-3p (B [95% CI]_asthma = 0.72 [0.35–1.10]; B [95% CI]_atopy = 0.68 [0.19–1.17]), miR-218-5p (B [95% CI]_asthma = 0.60 [0.10–1.10]; B [95% CI]_atopy = 0.75 [0.13–1.36]) and miR-19b-3p (B [95% CI]_asthma = 0.84 [0.24–1.45]; B [95% CI]_atopy = 0.63 [0.06–1.19]).

KEGG analysis

The authors performed pathway enrichment analysis using the KEGG database. Sixteen pathways were enriched among EV miRNAs associated with asthma during pregnancy versus never diagnosed (false discovery rate <0.05), including the TGF-β signaling pathway, extracellular matrix–receptor interaction, adherens junction and FoxO signaling (Figure 3). miR-30a-3p and miR-106b-5p were involved in the most significant pathways (12 and 11, respectively).

Figure 3. . Kyoto Encyclopedia of Genes and Genomes pathways were enriched among extracellular vesicle miRNAs associated with asthma during pregnancy versus never diagnosed.

The number of genes targeted by differentially expressed miRNAs (p < 0.05; |B_regression| >0.2) and FDR-adjusted p-values are shown in the left panel, and differentially expressed miRNAs involved in each pathway are shown in the right panel.

FDR: False discovery rate; KEGG: Kyoto Encyclopedia of Genes and Genomes.

Sensitivity analyses

In analyses adjusting for pre-pregnancy obesity, seven EV miRNAs were associated with maternal asthma during pregnancy versus never diagnosed (p < 0.05; |B_regression| >0.2), all of which were also identified in the primary analyses; miR-324-5p (B [95% CI] = -0.55 [-1.11 to 0.02]) and miR-106b-5p (B [95% CI] = -0.64 [-1.29 to 0.02]) did not achieve statistical significance. miR-9a-5p was associated with maternal atopy during pregnancy (B [95% CI] = -0.72 [-1.41 to -0.03]), but miR-1290 was not statistically significant (B [95% CI] = 2.45 [-0.12 to 5.03]). Because of the observed association between postpartum week of breast milk collection and EV miRNA profile, sensitivity analyses were conducted, excluding five samples collected >12 weeks postpartum (one with asthma that was active during pregnancy and four with asthma that was never diagnosed; two with atopy that was active during pregnancy and three with atopy that was never diagnosed). Fourteen miRNAs were associated with maternal asthma during pregnancy versus never diagnosed (p < 0.05; |B_regression| >0.2), nine of which were identified in the primary analyses. The five additional EV miRNAs had negative associations: miR-324-3p (B [95% CI] = -0.68 [-1.21 to -0.15]), miR-511-5p (B [95% CI] = -1.31 [-2.50 to -0.12]), miR-25-3p (B [95% CI] = -0.68 [-1.33 to -0.05]), miR-29c-5p (B [95% CI] = -0.73 [-1.41 to -0.04]) and miR-19b-3p (B [95% CI] = -0.61 [-1.19 to -0.02]). Two miRNAs were negatively associated with atopy during pregnancy: hsa-miR-483-5 (B [95% CI] = -1.33 [-2.42 to -0.24]) and hsa-miR-29a-5p (B [95% CI] = -0.65 [-1.29 to -0.01]).

Discussion

To the authors' knowledge, this is the first study to investigate associations between maternal asthma or atopic status and EV miRNAs in breast milk. Nine miRNAs were associated with maternal asthma that was active during pregnancy versus never diagnosed (p < 0.05; |B_regression| >0.2), eight of which had decreased expression with asthma during pregnancy. One miRNA was positively associated with both maternal asthma and atopy that were active during pregnancy versus never diagnosed.

Previous studies have identified miRNAs involved in lung development and disease, immune response and asthma [52]. Differentially expressed miRNAs have been observed in epithelial and peripheral blood mononuclear cells of asthmatic compared with healthy patients, with roles in inflammatory response, immune cell activation and ciliated cell differentiation [53]. Several previously identified asthma-associated miRNAs were also differentially expressed in the current analyses. In a cross-species study, miR-30a-3p was found to be downregulated in the peripheral blood of asthmatic patients compared with controls (n = 30) [54]. This direction of association between miR-30a-3p and asthma is consistent with the findings presented here. In a separate study of plasma EV miRNAs in patients with severe asthma compared with controls (n = 30), a suggestive positive association between miR-191-5p expression and asthma was observed (false discovery rate = 0.12), as was an association between miR-191-5p and percent forced expiratory volume in the first second of expiration predicted [55]. Although miR-191-5p was identified as differentially expressed in the breast milk of mothers with asthma during pregnancy in the current analyses, a direction of association was observed.

As demonstrated in asthma mouse models, changes in miRNA expression may be involved in regulating asthma-related pathways. For example, miR-30a-3p has been found to be inversely associated with expression of CCR3 [54], a chemokine receptor involved in eosinophilic inflammation [56]. In addition, miR-106b-5p, which was found to be negatively associated with asthma in this study, was downregulated in an asthma mouse model with upregulation of SIX1, a transcription factor involved in pulmonary fibrosis [57].

One miRNA, miR-1290, was found to be positively associated with both active asthma and atopy during pregnancy. Similarly, in a small study of patients with asthma compared with healthy controls (N = 22), miR-1290 had greater expression in naive T cells among individuals with mild asthma [58]. miR-1290 is a miRNA involved in carcinogenesis and inflammatory diseases [59] and has been associated with sex-specific lung function among children with asthma [60].

The authors also identified KEGG pathways enriched for miRNAs associated with active maternal asthma during pregnancy with potential roles in asthma-related inflammation, including the TGF-β signaling pathway, extracellular matrix–receptor interaction, adherens junction and FoxO signaling. TGF-β is a cytokine involved in asthma-related inflammation and airway remodeling [61]. Inflammation and physical strain from bronchoconstriction associated with asthma may result in extracellular matrix remodeling, and remodeling markers, including collagen, actin and elastic fibers, in the lungs and airways have been shown to be increased in cases of asthma [62]. In addition, inflammation may contribute to dysfunction of apical junctional complexes – including adherens junctions – between epithelial cells [63]. FoxO1 signaling may also be involved in allergic inflammation as a regulator of IRF4 in macrophages [64]. Mothers with asthma that was active during pregnancy had lower expression of miR-30a-3p, miR-106b-5p, miR-331-3p, miR-200a-3p and miR-324-5p, miRNAs involved in these enriched KEGG pathways, suggesting upregulation of genes implicated in asthma-related inflammation. The authors also identified the Hippo signaling pathway as enriched for maternal asthma-associated miRNAs. Genetic variants of YAP1, a gene involved in the Hippo pathway, have been associated with exercise-induced asthma and asthma severity [65].

Strengths of this study include a well-characterized prospective cohort of pregnant women recruited in early to mid-pregnancy with asthma and atopy defined by standardized questionnaires. The study protocol obtained breast milk following standardized procedures. MiRNAs were measured from isolated EVs, which are more likely to be taken up by the intestinal cells of infants and elicit downstream biological effects. In addition, the TaqMan OpenArray Human MicroRNA Panel provided an efficient method for measuring expression of known miRNAs in a larger sample size than previous studies. The TaqMan OpenArray uses quantitative real-time PCR, which is the gold standard for RNA measurement and has higher reproducibility than other platforms [66]. Furthermore, results of sensitivity analyses to address potential effect modification by postpartum week of breast milk collection were largely consistent with the primary analyses.

This study focused on expression levels of EV miRNAs detected in ≥80% of samples. Although an alternative approach would be to analyze miRNAs as detected versus not detected, the authors chose to analyze miRNAs with higher expression levels, as these are thought to be more likely to be taken up by intestinal cells and have greater bioavailability. It is not known whether there is a threshold level of expression for breast milk-derived EV miRNAs to affect biological processes in infants. Several additional limitations of this study should be noted. Power was limited by a small sample size, and the authors therefore used the a priori criteria for determining significant associations in primary analyses of expression levels (nominal p < 0.05; |B_regression| >0.2). All miRNAs with nominal p < 0.05 in linear models also met the criteria (|B_regression| >0.2). This approach may increase type 1 errors, and future studies may therefore incorporate a larger effect size and corrections to adjust for multiple testing.

The number of mothers classified as having asthma or atopy that was inactive during pregnancy was particularly small (n = 4 and 5, respectively). Although previous studies have identified associations between miRNA expression and asthma severity [53], associations with inactive asthma and atopy were secondary outcomes in this study, and results from these analyses should be interpreted with caution. A larger sample size would also allow for testing effect modification by factors such as preterm delivery and birth weight – outcomes associated with maternal asthma [1]. In addition, the authors did not have data on asthma or atopy symptoms at the time of breast milk collection and therefore could not determine if symptoms during breastfeeding were associated with EV miRNA expression. This study was also limited by lack of mRNA expression data. Further research is necessary to determine downstream biological effects of breast milk-derived miRNA profiles on child health outcomes.

Conclusion

This study demonstrates that maternal asthma and atopic status is associated with breast milk-derived EV miRNA profiles, providing evidence that miRNAs may be involved in biochemical communication between mothers and infants. EV miRNAs and enriched pathways associated with maternal asthma and atopy in our study have established roles in immune and inflammatory responses associated with asthma and allergic diseases. Additional research, including larger studies, is needed to fully understand the relationship between maternal asthma and atopy and breast milk-derived EV miRNA expression and to determine the effect of differential miRNA expression on biological pathways and health outcomes in offspring.

Future perspective

Understanding of the biological mechanisms involved in early life programming is rapidly evolving. Future epidemiological studies that include larger sample sizes and more diverse populations may provide insights into the maternal and environmental factors influencing breast milk-derived EV miRNA profiles and downstream effects on child health. Results may inform interventions for improving infant and child health, such as the addition of miRNAs to infant formula.

Summary points.

• Extracellular vesicle (EV)-encapsulated miRNAs in breast milk are a form of mother–infant biochemical communication involved in early life programming.

• EV miRNAs in breast milk may be influenced by maternal asthma and atopy.

• The authors identified nine breast milk-derived EV miRNAs with expression levels (a priori criteria: p < 0.05; |B_regression| >0.2) associated with maternal asthma.

• Eight of the nine differentially expressed EV miRNAs had lower expression among mothers with active asthma during pregnancy.

• One EV miRNA, miR-1290, was positively associated with both asthma and atopy during pregnancy.

• Kyoto Encyclopedia of Genes and Genomes pathways enriched for differentially expressed miRNAs (false discovery rate <0.05) were related to asthma, including the TGF-β signaling pathway, extracellular matrix–receptor interaction, adherens junction and FoxO signaling.

• Further research may provide insights into the effects of changes in miRNA profiles on biological pathways and child health outcomes.