Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2024 Feb 1.
Published in final edited form as: Allergy. 2022 Sep 24;78(2):418–428. doi: 10.1111/all.15516

The Maternal Prenatal and Offspring Early-Life Gut Microbiome of Childhood Asthma Phenotypes

Kathleen A Lee-Sarwar 1,2,*, Yih-Chieh Chen 1,2, Yuan Yao Chen 3, Anita L Kozyrskyj 3, Piush J Mandhane 3, Stuart E Turvey 4, Padmaja Subbarao 5, Hans Bisgaard 6, Jakob Stokholm 6, Bo Chawes 6, Søren J Sørensen 6, Rachel S Kelly 1, Jessica Lasky-Su 1, Robert S Zeiger 7, George T O’Connor 8, Megan T Sandel 9, Leonard B Bacharier 11, Avraham Beigelman 10,12, Vincent J Carey 1, Benjamin J Harshfield 1, Nancy Laranjo 1, Diane R Gold 1,13, Scott T Weiss 1, Augusto A Litonjua 14
PMCID: PMC9892205  NIHMSID: NIHMS1838692  PMID: 36107703

Abstract

Background

The infant fecal microbiome is known to impact subsequent asthma risk, but the environmental exposures impacting this association, the role of the maternal microbiome, and how the microbiome impacts different childhood asthma phenotypes are unknown.

Methods

Our objective was to identify associations between features of the prenatal and early-life fecal microbiomes and child asthma phenotypes. We analyzed fecal 16s rRNA microbiome profiling and fecal metabolomic profiling from stool samples collected from mothers during the third trimester of pregnancy (n=120) and offspring at ages 3–6 months (n=265), 1 (n=436) and 3 years (n=506) in a total of 657 mother-child pairs participating in the Vitamin D Antenatal Asthma Reduction Trial. We used clinical data from birth to age 6 years to characterize subjects with asthma as having early, transient or active asthma phenotypes. In addition to identifying specific genera that were robustly associated with asthma phenotypes in multiple covariate-adjusted models, we clustered subjects by their longitudinal microbiome composition and sought associations between fecal metabolites and relevant microbiome and clinical features.

Results

Seven maternal and two infant fecal microbial taxa were robustly associated with at least one asthma phenotype, and a longitudinal gut microenvironment profile was associated with early asthma (Fisher exact test p=0.03). Though mode of delivery was not directly associated with asthma, we found substantial evidence for a pathway whereby cesarean section reduces fecal Bacteroides and microbial sphingolipids, increasing susceptibility to early asthma.

Conclusion

Overall, our results suggest that the early-life, including prenatal, fecal microbiome modifies risk of asthma, especially asthma with onset by age 3 years.

Keywords: Bacteroides, cesarean section, metabolomics, sphingolipids, wheeze

Graphical Abstract

graphic file with name nihms-1838692-f0001.jpg

INTRODUCTION

It is now well-established that the trillions of microbes that reside in the human gut influence health throughout the body1. Colonization early in life impacts immune development with relevance to subsequent inflammatory and allergic disease risk2 and emerging literature suggests that the intestinal microbiome-asthma connection may begin before birth, during pregnancy3–5. Prior studies have identified microbiome features present during the first months of life that are associated with later asthma development, but few have differentiated between asthma phenotypes or included maternal prenatal microbiome profiling3,6. In this ancillary analysis of the Vitamin D Antenatal Asthma Reduction Trial (VDAART), we hypothesized that the prenatal and early-life bacterial microbiomes, evaluated by 16S rRNA sequencing during the third trimester of pregnancy in maternal stool samples and at ages 3–6 months, 1 year and 3 years in offspring stool samples, are associated with childhood asthma. In recognition of the heterogeneity in childhood asthma phenotypes and risk factors7, we characterized three phenotypes: early asthma or recurrent wheeze by age 3 years (hereafter referred to as early asthma), active asthma at age 6 years and transient asthma (early asthma without active asthma at age 6 years). We found evidence for longitudinal fecal microbiome and metabolome patterns that portended increased risk of asthma, including a pathway whereby birth by cesarean section delivery leads to asthma-associated microbiome perturbations.

METHODS

Detailed methods are available in the Supporting Information. Subjects were participants in VDAART, a multi-site randomized controlled trial of Vitamin D supplementation during pregnancy for prevention of asthma in offspring (NCT00920621)8. We analyzed three childhood asthma phenotypes: early asthma (asthma diagnosis and/or recurrent wheeze by age 3 years), active asthma (asthma diagnosis and report of wheeze and/or asthma medication use between ages 5 and 6 years) and transient asthma (early asthma without active asthma at age 6 years). Stool samples were collected from mothers during the third trimester of pregnancy (n=120) and in offspring at ages 3 to 6 months (n=265), 1 year (n=436) and 3 years (n=506) only if the subject had not been exposed to antibiotics in the prior 7 days. Microbiome profiling was performed by 16S rRNA sequencing and metabolomic profiling was performed using ultra-high performance liquid chromatography coupled with tandem mass spectrometry.

Analyses were adjusted for sex, race/ethnicity and VDAART study site (Boston, St. Louis and San Diego) unless otherwise specified. Analyses of stool samples collected at age 3–6 months were additionally adjusted for exact age at stool sample collection, and analyses of stool samples collected during pregnancy were additionally adjusted for maternal age.

To ensure robust associations between fecal bacterial genera and asthma phenotypes, we used three methods: 1) Analysis of Compositions of Microbiomes with Bias Correction (ANCOM-BC)9, which utilizes a regression framework; 2) Songbird, a reference frame-based approach10 that produces log-fold changes for each taxon and a differential ranking of taxa by the strength of their association with the outcome, in this case, asthma phenotype; and 3) a comparison of frequency of presence vs absence of each taxa by phenotype using logistic regression.

We performed similarity network fusion (SNF) with spectral clustering to cluster subjects by their longitudinal fecal microbiome and microbiome/metabolome trajectories. Associations of SNF clusters with asthma phenotypes were analyzed using Fisher’s exact test and covariate-adjusted logistic regression. ANCOM-BC was used to test associations of fecal microbes with SNF clusters. We performed logistic regression analyses of associations of fecal metabolites with asthma phenotypes and with SNF clusters.

We tested the hypothesis that exposures, including breastfeeding, mode of delivery, antibiotics exposure and dog ownership, may influence the microbiome and thereby influence asthma risk. First, we tested associations of these exposures with Bray-Curtis microbiome beta diversity using PERMANOVA. Mediation analyses were performed to further evaluate potential causal pathways. To identify associations of the microbial dysbiosis of Cesarean section with asthma, we trained a random forest algorithm to predict mode of delivery based solely on fecal microbiome data from age 3–6 months and used Fisher exact test and adjusted logistic regression to determine associations with asthma.

We performed sensitivity analyses of key models to test biologically plausible roles of VDAART antenatal vitamin D treatment assignment and maternal asthma in the hypothesized causal relationship between the microbiome and asthma, testing whether antenatal vitamin D or maternal asthma could be confounders (tested by addition of these variables as model covariates), upstream causal factors (tested by PERMANOVA and logistic regression to determine associations between these variables and beta-diversity and SNF cluster membership, respectively), or effect modifiers (tested by addition of interaction terms to models followed by stratification where the interaction term p<0.05).

RESULTS

Subject Characteristics

We analyzed data from 657 mother-child pairs who provided at least one stool sample, a subgroup comparable to the overall VDAART cohort (Figures E1 & E2, Table E1). Of the offspring, 115 (19%) had active asthma at age 6 years, 90 (15%) had transient asthma and 405 (66%) had no asthma, and these categories were mutually exclusive. Early asthma (onset before age 3 years) was present in 185 (28%) offspring (Figure E3).

Children with asthma of any phenotype were more likely to be male (Chi square p=0.03) and of Black non-Hispanic race/ethnicity (Chi square p<0.01) (Table E1). Mothers of children with asthma of any phenotype were younger (ANOVA p<0.01), less likely to have a college degree (Chi square p<0.01) and more likely to have asthma (Chi square p<0.01). As reported previously11–14, high-dose antenatal vitamin D supplementation (the VDAART intervention) was associated with reduced early, but not active, asthma, though this did not reach statistical significance in this subgroup (Chi square p=0.06). Allergic sensitization, based on serum specific IgE to common inhalant allergens, was most frequent among subjects with active asthma (Chi square p<0.01). Asthma status also differed by study center and was more frequent among subjects born preterm (Chi square p<0.01). Those with asthma of any phenotype were more likely to have received perinatal antibiotics (Chi square p<0.01) and less likely to have been exclusively breastfed for the first four months of life (Chi square p<0.01) (Table E1). Additionally, there was a lower likelihood of having a dog in the home among those with early asthma (Chi square p=0.03). Asthma was not associated with mode of delivery or number of older siblings.

Fecal Microbial Diversity and Taxa Associations with Asthma

Alpha diversity was highest in maternal samples and increased in offspring from a minimum in infancy (samples from age 3–6 months, mean 4.5 months, standard deviation 1.0 months, Figure E2B) to levels approaching maternal diversity at age 3 years (Figure 1). In analyses of microbial diversity associations with asthma phenotypes, all of which were adjusted for sex, race/ethnicity and study site (analyses of maternal samples were additionally adjusted for maternal age and analyses of infant samples were additionally adjusted for exact age at stool sample collection), alpha diversity was not associated with asthma phenotype (adjusted linear regression p>0.05 across all time points and asthma phenotypes, Table E2). Beta diversity (Bray-Curtis dissimilarity) exhibited profound shifts from infancy to age 3 years, at which point offspring microbiome composition came to resemble maternal composition (Figure 1). Beta diversity at age 3 years was associated with active asthma (PERMANOVA F=2.50, p=0.004), while maternal fecal beta diversity was associated with early asthma (PERMANOVA F=1.65, p=0.05) (Table E3).

Figure 1.

Figure 1.

Open in a new tab

Fecal microbiome diversity trajectories over the first years of life. A. Line plots depict average alpha diversity (Shannon Index and Faith’s phylogenetic diversity) with standard errors at each time point by asthma phenotype. B. Principal coordinates plots depict fecal microbiome composition (Bray-Curtis Dissimilarity) color-coded by age at stool sample collection in all subjects and by asthma phenotype.

We sought genera that differed by asthma phenotype at each stool sample collection time point. Recognizing that bioinformatic tools may yield discordant results15,16, we used an ensemble approach of three methods: 1) ANCOM-BC9, which employs a linear regression framework; 2) Songbird10, a reference frame-based approach that produces rankings of taxa with log-fold changes with respect to asthma outcomes; and 3) adjusted logistic regression models of presence vs absence of taxa. All analyses were adjusted for sex, race/ethnicity, study site; analyses of maternal samples were additionally adjusted for maternal age and analyses of infant samples were additionally adjusted for exact age at stool sample collection. We defined a robust association as one observed using at least two of these methods (see Supporting Information). Nine identified robust associations included lower Staphylococcus abundance at age 3–6 months in association with transient asthma, lower Bacteroides at age 3–6 months in association with transient and early asthma, and associations of 5 maternal taxa with transient asthma (Erysipelatoclostridium, Terrisporobacter, Clostridia UCG-014, UBA1819 and Lactobacillus) and 2 maternal taxa with active asthma compared to no asthma (Faecalitalea and Prevotella, Table E4, Figure E4, Figure 2). We found modest evidence that maternal fecal asthma-associated taxa are associated with increased preterm birth (results available in the Supporting Information). No fecal genera at age 1 or 3 years were associated with asthma of any phenotype (all FDR>0.05).

Figure 2.

Figure 2.

Open in a new tab

Violin plots display distributions of read counts (y axis) by asthma phenotype for genera that demonstrated robust associations with at least one asthma phenotype with asterisks denoting significant (FDR<0.05) associations in ANCOM-BC analysis or present vs absent analyses. Genera from stool samples collected from mothers during pregnancy are displayed in panel A and genera from stool samples collected from offspring at age 3–6 months are displayed in panel B.

Maternal Microbiome Features are Associated with both Preterm Birth and Offspring Asthma

Given a potential link between the maternal microbiome and preterm birth17 and our observation of increased asthma occurrence among VDAART offspring born premature (Table E1), we hypothesized that asthma-associated maternal fecal genera may predispose to premature birth. Overall maternal fecal microbiome composition was associated with gestational age (PERMANOVA for maternal Bray-Curtis dissimilarity R2=0.014, F=1.68, p=0.04), and of the 7 maternal taxa associated with offspring asthma, two were associated with offspring gestational age at delivery: Clostridia UCG-014 (adjusted linear regression beta=12.3, 95% CI 0.1, 24.5, p=0.049) and the genus UBA1819 of the family Ruminococcaceae (beta=−7.5, 95% CI −13.6, −1.42, p=0.02). These patterns of associations are suggestive of a pathway whereby reduced maternal fecal Clostridia UCG-014 and increased UBA1819 may mark increased risk of early delivery and thereby increase risk of offspring asthma. However, these data are limited by the small number of premature births (gestational age < 37 weeks) among mothers with prenatal microbiome profiling (7 premature births out of 120 mothers).

Longitudinal Microbiome Profiling

We used similarity network fusion (SNF) followed by spectral clustering to group subjects based on longitudinal genus-level microbiome data. As SNF requires complete data and there were limited maternal stool samples available (Figure E2A), we included microbiome data from offspring only at ages 3–6 months, 1 year and 3 years (n=156). SNF yielded 5 clusters based on longitudinal microbiome composition, which we refer to as SNF-M clusters. SNF-M cluster 3 (n=27) was associated with reduced early asthma (Fisher exact test p=0.03) and a nearly statistically significant reduction in any asthma (compared to no asthma, p=0.06) (Figure 3A). Regarding potential confounders, SNF-M was not associated with child sex or race/ethnicity (Fisher exact test p>0.05) but was associated with study center (p=0.03). In logistic regression models adjusted for study center, associations of SNF-M with early asthma and any asthma were preserved, though with an attenuated level of significance (OR=0.30, 95% CI 0.07, 0.96, p=0.07 for early asthma; OR=0.36, 95% CI 0.10, 1.04, p=0.08 for any asthma). Subjects in the low-asthma-risk cluster exhibited notable enrichments in fecal Bacteroides at both age 3–6 months and 1 year (Table E5, Figure E5B). Though SNF-M cluster generation did not utilize maternal microbiome data, 21 maternal genera were associated (FDR<0.05) with SNF-M, including 5 of the 7 genera with robust asthma associations, reinforcing the potential import of the maternal microbiome in determining offspring microbiome trajectories and asthma risk.

Figure 3.

Figure 3.

Open in a new tab

Percentage of subjects with asthma phenotypes by similarity network fusion (SNF) cluster membership. SNF-M clusters based on microbiome composition at ages 3–6 months, 1 year and 3 years are displayed in panel A. There were 156 subjects with complete data for SNF-M cluster determination, including 44 with early asthma, 19 with transient asthma, and 32 with active asthma. SNF-M2 clusters based on microbiome composition at ages 3–6 months, 1 year and 3 years and metabolome composition at ages 1 year and 3 years are displayed in panel B. There were 130 subjects with complete data for SNF-M2 cluster determination, including 43 with early asthma, 18 with transient asthma, and 28 with active asthma. p values are for Fisher’s exact test for the comparison of the asthma-associated cluster with other clusters.

Fecal Metabolomics Associations with Asthma

To characterize the functional significance of microbiome associations with asthma, we analyzed untargeted stool metabolomic data from the same samples used for microbiome profiling. Models were adjusted for sex, race/ethnicity and study site; analyses of maternal samples were additionally adjusted for maternal age and analyses of infant samples were additionally adjusted for exact age at stool sample collection. Only one fecal metabolite was associated (FDR<0.05) with an asthma phenotype: the sialic acid N-acetylneuraminate at age 3–6 months was reduced in offspring who developed early asthma (logistic regression combined OR 0.31, 95% CI 0.19, 0.53, FDR=0.01).

We next sought fecal metabolites associated with SNF-M. At age 3–6 months, a single association was identified: 3-ketosphinganine was enriched in the low asthma risk SNF-M cluster (combined OR 4.5, 95% CI 2.1, 9.3, FDR=0.047) and was also non-significantly but consistently associated with reduced asthma (early asthma OR 0.75 (95% CI 0.54, 1.04), p=0.08; transient asthma OR 0.71 (0.44, 1.16), p=0.18; active asthma OR 0.76 (0.53, 1.10), p=0.15). 3-ketosphinganine is a metabolite of the de novo sphingolipid synthesis pathway, which is utilized by limited gut bacteria, including Bacteroides18. As both Bacteroides and 3-ketosphinganine were reduced in infant stool samples among offspring with early asthma, we hypothesized that Bacteroides could be responsible for 3-ketosphinganine production. Consistent with this, we found a strong correlation between Bacteroides and 3-ketosphinganine at age 3–6 months (Spearman rho=0.63 and 0.68, p<0.0001 for metabolomics batches 1 (n=63) and 2 (n=153), respectively, Figure E5A).

In stool samples collected at age 1 year, 11 metabolites were associated with SNF-M (FDR<0.05, Table E6). Only one – the polyunsaturated fatty acid linoleic acid – was associated with asthma in a directionally consistent manner; specifically, fecal linoleic acid at age 1 year was reduced among subjects with active asthma (OR 0.67 (95% CI 0.45, 1.00), p=0.049). No maternal fecal metabolites or metabolites from stool samples collected at age 3 years were associated with SNF-M.

We performed SNF a second time, clustering subjects based on fecal metabolome at ages 1 and 3 years in addition to the longitudinal microbiome data used in our first SNF analysis. Batch effects prohibited inclusion of metabolomics from age 3–6 months. We refer to the resulting two clusters as SNF-M2 clusters because they are based on both microbiome and metabolome data. Similar to the low-asthma-risk SNF-M cluster generated using microbiome data alone, SNF-M2 cluster 1 (n=67) was associated with reduced early asthma (Fisher exact test p=0.03) and any asthma (compared to no asthma, p=0.04) (Figure 3B). Regarding potential confounders, SNF-M2 cluster membership was not associated with child sex (Fisher exact test p=0.73) but was associated with race/ethnicity and study center (p<0.01 for both). In logistic regression models adjusted for race/ethnicity and study center, associations of SNF-M2 with early asthma and any asthma were preserved, though with an attenuated level of significance (early asthma OR=0.44, 95% CI 0.17, 1.07, p=0.07; any asthma OR=0.43, 95% CI 0.17, 1.05, p=0.07). Of 24 subjects with complete microbiome and metabolome data in the original low-risk SNF-M cluster, 22 (92%) fell into the low-risk SNF-M2 cluster (Fisher exact test p<0.01). In contrast to the few fecal metabolites associated with SNF-M, numerous metabolites were associated with SNF-M2 (FDR<0.05 for 215 (33% of) metabolites from age 3–6 months; 241 (32%) from age 1 year; 349 (47%) from age 3 years, Table E7). However, pathway analysis did not identify specific pathways driving these associations (FDR>0.05 for all pathways at all time points).

Fecal Dysbiosis of Cesarean Section Associations with Asthma

We hypothesized that exposures associated with asthma, including breastfeeding, perinatal antibiotics and dog ownership (Table E1), impact asthma risk by modifying the microbiome. We also analyzed mode of delivery given our observed association of Bacteroides at 3–6 months with asthma and previously reported associations between mode of delivery and infant fecal Bacteroides19; we too observed a pronounced reduction in Bacteroides at age 3–6 months in those born by cesarean section (Wilcoxon rank sum test p<0.001, Figure E5B). All examined exposures were associated with microbiome composition at one or more time points (PERMANOVA p<0.05, Figure 4).

Figure 4.

Figure 4.

Open in a new tab

F statistics for associations of upstream exposures with fecal microbiome beta-diversity (Bray-Curtis) in PERMANOVA analyses including all analyzed upstream exposures (breastfeeding, mode of delivery, perinatal antibiotics and dog ownership), child sex, race/ethnicity and study center. Analyses of the microbiome at age 3–6 months were additionally adjusted for age at stool sample collection. Asterisks denote significant (p<0.05) results.

Adding to the evidence that mode of delivery may impact dysbiosis in early asthma, birth by cesarean section was less frequent in the low-asthma-risk SNF-M cluster (3 of 27 (11%) in low-risk vs 45 of 129 (35%) in high-risk cluster, Fisher exact test p=0.02). This association was preserved in models adjusted individually for potential confounders and in a fully adjusted model (logistic regression p range 0.01–0.04 with covariates: breastfeeding, antibiotics, dog ownership, study center and race/ethnicity; fully adjusted p=0.02). Accordingly, SNF-M was a significant mediator of the association between mode of delivery and early asthma (33% of association mediated; p for indirect effect mediated through SNF-M=0.03; p for direct effect not mediated through SNF-M=0.92). Similarly, fecal Bacteroides at age 3–6 months was a significant mediator of the association between mode of delivery and asthma (52% and 55% of associations with early and any asthma, respectively, mediated; p for indirect effects < 0.05 in analyses adjusted for antibiotics, sex, race/ethnicity and study center). No other analyzed exposures were associated with SNF-M. In summary, despite no overall association of mode of delivery with asthma, we found substantial evidence for a pathway whereby birth by cesarean section results in a high-risk microbiome trajectory with reduced fecal Bacteroides and de novo intestinal sphingolipid synthesis, increasing risk of early asthma (Figure 5).

Figure 5.

Figure 5.

Open in a new tab

Schematic of associations between mode of delivery and early asthma. Significant (p<0.05) associations are denoted with red lines.

To disentangle the effects of mode of delivery on the microbiome from its non-microbial effects, we trained a random forest model to predict mode of delivery using only fecal microbiome data from age 3–6 months. Bacteroides was the most informative genus in the resulting model (Figure E6), consistent with its strong association with mode of delivery. In a test dataset that did not include subjects used to train the model, the model correctly classified 8 out of 31 subjects born by cesarean section and 55 out of 63 subjects born by vaginal delivery, with the relatively high error rate possibly due to small training dataset and likely reflecting that gut microbiome differences by birth mode are gradually attenuated over the first year of life20,21. We posited that test dataset subjects predicted to have been born by cesarean section (correctly or incorrectly) had microbiomes most consistent with cesarean section delivery and refer to these subjects as exhibiting a cesarean-section-associated microbiome. While actual mode of delivery was not associated with early asthma or any asthma (test dataset Fisher test p=0.47 and 0.82, respectively; logistic regression adjusted for age at sample collection and perinatal antibiotics p=0.58 and 0.89, respectively), having a cesarean-section-associated microbiome was associated with increased asthma (Fisher exact test p=0.03 and 0.04 for early and any asthma, respectively), and this association was largely preserved after adjustment for age at stool sample collection and perinatal antibiotics (logistic regression OR=3.5 (95% CI 1.1, 11.1) and 2.8 (95% CI 0.9, 9.0); p=0.03 and 0.07 for early asthma and any asthma, respectively). These findings suggest that mode of delivery has effects both dependent on and independent of the microbiome, and that its effects specifically on the microbiome months after birth may increase asthma risk22,23.

In contrast, mode of delivery was not associated with the SNF-M2 clusters generated using both microbiome and metabolomic data (Fisher exact test p=0.45), nor were dog ownership or antibiotics. Instead, breastfeeding was markedly more common in the low-asthma-risk SNF-M2 cluster (40 out of 66 subjects (61%)) compared to the high-risk cluster (10 out of 62 subjects (16%), Fisher exact test p<0.001). This association was robust in sensitivity logistic regression analyses (p<0.001 including individual covariates: breastfeeding, antibiotics, dog ownership, study center and race/ethnicity; p=0.001 in a fully adjusted model). SNF-M2 mediated a substantial estimated proportion of the association between breastfeeding and asthma (23% for early asthma, 38% for active asthma, 29% for any asthma), but this was not statistically significant (p indirect effects > 0.05). These results suggest that while mode of delivery has a prominent impact on the early intestinal dysbiosis of asthma, breastfeeding impacts the gut environment beyond the microbiome, including its biochemical features, with independent relevance to asthma risk.

Antenatal Vitamin D and Maternal Asthma May Modify Associations of the Microbiome with Asthma

This study is an ancillary analysis of a clinical trial of prenatal vitamin D supplementation to prevent asthma in offspring at elevated genetic risk of allergic diseases, and maternal asthma was present in 40% of our study sample. There are multiple biologically plausible ways in which prenatal vitamin D or maternal asthma could participate in hypothesized causal relationships of the microbiome with offspring asthma, including as upstream causal factors, confounders, or effect modifiers (Figure E7). We tested these possibilities in sensitivity analyses, and to limit multiple testing, focused on asthma associations with microbiome composition (Bray-Curtis beta diversity) and SNF clusters.

Neither vitamin D nor maternal asthma was associated with fecal microbiome composition at any time point (adjusted PERMANOVA p>0.05), or with SNF-M or SNF-M2 (adjusted logistic regression p>0.05). There were minimal changes to microbiome composition or SNF associations with asthma after adjustment for vitamin D treatment assignment or maternal asthma (Figure E8). We concluded that neither antenatal vitamin D nor maternal asthma is an upstream causal factor or meaningful confounder of observed microbiome-asthma associations.

In contrast, we did find modest evidence of effect modification by both vitamin D treatment assignment and maternal asthma. In analyses of beta diversity, there was one significant (p=0.03) interaction identified between maternal asthma and microbiome composition at age 3–6 months on the outcome of active asthma. Microbiome composition at age 3–6 months was associated with active asthma among offspring of mothers with asthma (n=88, F=2.5, p=0.01) but not among offspring of mothers without asthma (n=146, F=1.1, p=0.36). In analyses of SNF clusters, there was one significant (p=0.04) interaction identified between SNF-M2 and vitamin D treatment assignment on the outcome of active asthma. Specifically, there was no association between SNF-M2 and active asthma among offspring of mothers randomized to high-dose vitamin D (n=74, OR=0.93 (95% CI 0.30, 2.86), p=0.90) but a significant inverse association between membership in the low-asthma-risk SNF-M2 cluster and active asthma among offspring of mothers randomized to low-dose vitamin D (n=51, OR=0.15 (95% CI 0.02, 0.68), p=0.03). A similar pattern was seen for the early asthma phenotype (Figure E9). We concluded that associations of the microbiome with asthma may differ by antenatal vitamin D or maternal asthma status.

DISCUSSION

In 657 mother-child pairs, early-life longitudinal patterns of fecal microbiome and metabolome development were associated with asthma, especially with onset by age 3 years. Despite no direct association between mode of delivery and asthma, we found suggestive evidence for a pathway whereby birth by cesarean section results in dysbiosis with reduced infant Bacteroides and microbial sphingolipid production and increases risk of early asthma. A rich prior literature has linked mode of delivery to infant fecal Bacteroides24–26 and Bacteroides with sphingolipid production27. Cesarean section delivery may have both microbiome-modifying and non-microbial consequences, and microbial (but not necessarily non-microbial) effects may have greatest relevance to asthma risk. We predicted mode of delivery using only infant microbiome data to isolate the microbiome-modifying effects of cesarean section, which were associated with early asthma. Our results are in accordance with a recent report that delivery by cesarean section is associated with asthma only when gut microbiome composition at age 1 year retains a cesarean section microbial signature23, and support efforts to restore the microbiome after cesarean section delivery28.

Reduced fecal Bacteroides in infancy has previously been associated with atopy and wheeze at age 5 years29, childhood allergic sensitization30–32 and atopic eczema development33. However, other studies have not detected an inverse association of Bacteroides with asthma and instead identified other fecal taxa associations with asthma, most of which were not replicated in our analysis6. We used a stringent approach to reduce false positive associations, and the downside of this is the possibility of false negatives where we otherwise may have replicated prior associations. We did detect novel microbe-asthma associations, especially with maternal genera. In fact, seven of our nine identified robust microbe-asthma associations were with maternal fecal microbes, and we found evidence for associations between the microbiome during pregnancy and all asthma phenotypes: early asthma was associated with maternal beta diversity and both transient and active asthma were associated with specific maternal taxa. Very few studies have examined associations of the maternal microbiome with asthma in offspring, though associations of maternal taxa have been identified with offspring food allergy4 and allergy-associated dermatitis in early infancy5 and a rich literature supports the potential importance of the maternal microbiome in asthma inception3.

N-acetylneuraminate, the dominant sialic acid in humans, was the only fecal metabolite directly associated with asthma and was reduced at age 3–6 months in those with early asthma. Sialic acid is ubiquitously expressed and could impact asthma risk via immune-modulatory interactions with Siglecs or selectins34, differential binding to IgE35, or modulation of microbiome or gut barrier function36. Fecal linoleic acid, the essential omega-6 fatty acid, was reduced at age 1 year in children with early asthma and was associated with a protective early-life microbiome profile. While omega-6 fatty acids are generally considered pro-inflammatory, complex relationships of omega-3 and omega-6 fatty acids have been found with asthma37,38 and we have observed that omega-3 and omega-6 fatty acids are highly correlated in both stool and plasma in VDAART39,40. The relationships of fecal omega-3 and omega-6 fatty acids to one another and to dietary fatty acid intake are worthy of future study. Finally, breastfeeding was associated with a low asthma risk profile based on both fecal microbiome and metabolome data, but not one based on microbiome data only, suggesting that benefits of breastfeeding may be due to effects on the intestinal biochemical environment, beyond effects on microbiome composition.

Our analysis had strengths: we followed diverse mother-child pairs longitudinally and differentiated between childhood asthma phenotypes. However, future studies could utilize more numerous maternal stool samples than the 120 analyzed here, and could employ a more granular assessment of phenotypes; for example, some children with active asthma at age 6 years did not have early asthma but due to sample size limitations were not analyzed separately as a “late-onset asthma” phenotype41. Additionally, subjects with persistent vs transient asthma or with early vs no early asthma were more likely to exhibit allergic sensitization (Table E1), and future analyses could determine phenotypes based on atopy rather than timing of asthma symptoms, though these phenotype groupings are likely to be closely correlated41. Given that our study population was at elevated genetic risk of allergic diseases, the distinction between atopic and non-atopic status may be more meaningful in an unselected study sample. We sequenced microbial 16S rRNA in stool samples, which results in excellent sensitivity but limits taxonomic resolution, so sequences could reliably be identified only at the genus or higher level. In this observational study, we cannot establish a causal relationship between the microbiome and asthma; for example, though it is possible that microbiome perturbations cause increased asthma risk, it is also possible that early signs of asthma lead to dietary or other environmental changes that account for microbiome differences. Finally, generalizability may be limited in this ancillary analysis of a trial of antenatal vitamin D to prevent asthma in offspring at elevated genetic risk of atopy. We did find modest evidence of effect modification by vitamin D treatment and maternal asthma on microbiome-asthma associations, and such interactions are worthy of further investigation to determine whether a personalized approach to microbiome-directed therapeutics in asthma prevention is warranted based on these or other factors.

Supplementary Material

SUPINFO2
SUPINFO1

  • This analysis identified associations between the fecal microbiome of mothers during pregnancy and offspring at ages 3–6 months, 1 and 3 years with childhood asthma phenotypes.

  • Associations were identified between maternal microbiome diversity and taxa with offspring asthma.

  • Results suggest that dysbiosis associated with cesarean section birth increases early asthma risk.

Acknowledgments:

VDAART was funded by U01HL091528 from the National Heart, Lung, and Blood Institute. Additional funding came from NIH grants 1K08HL148178, R01HL108818, R01HL123915, R01HL141826, 5T32HL007427 and ECHO grant OD023268.

Footnotes

Conflict of Interest Statement: AAL has received author royalties from UpToDate, Inc. STW has received royalties from UpToDate, Inc. LBB reports grants from NIH/NIAID and NHLBI, personal fees from GlaxoSmithKline Genentech/Novartis, DBV Technologies, Teva, Boehringer Ingelheim, AstraZeneca, WebMD/Medscape, Sanofi, Regeneron, Vectura, Circassia, Kinaset, Vertex, OM Pharma; and royalties from Elsevier outside the submitted work. AB holds stock from DBV Technologies, is a consultant for AstraZeneca and Raffa, and received speaking honoraria from AstraZenica, Novartis and Sanofi. RSZ is a consultant for AstraZeneca, DBV Technologies, Genentech, Inc., GlaxoSmithKline, Merck & Co., Novartis, Quest Diagnostics, Regeneron/Sanofi, TEVA Pharmaceuticals, and has received research support from ALK Pharmaceuticals, AstraZeneca, Genentech, Inc., GlaxoSmithKline, NHLBI, MedImmune, Merck, and Teva Pharmaceuticals. GTO has been compensated for speaking at a conference supported by Menarini, Inc. and for serving on a Data and Safety Monitoring Committee for Dicerna, Inc. VJC has received research support from Bayer and has a financial interest in Gilead Sciences. KL-S, YSC, YYC, ALK, PJM, SET, PS, HS, JS, BC, SJS, RSK, JL-S, MS, BJH, NL and DRG have nothing to disclose.

Trial Registration: ClinicalTrials.gov NCT00920621

References

  • 1.Lynch SV, Pedersen O. The Human Intestinal Microbiome in Health and Disease. N Engl J Med. 2016. Dec 15;375(24):2369–2379. [DOI] [PubMed] [Google Scholar]
  • 2.Gensollen T, Iyer SS, Kasper DL, Blumberg RS. How colonization by microbiota in early life shapes the immune system. Science. 2016. Apr 29;352(6285):539–44. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Gao Y, Nanan R, Macia L, Tan J, Sominsky L, Quinn TP, et al. The maternal gut microbiome during pregnancy and offspring allergy and asthma. J Allergy Clin Immunol. 2021. Sep;148(3):669–678. [DOI] [PubMed] [Google Scholar]
  • 4.Vuillermin PJ, O’Hely M, Collier F, Allen KJ, Tang MLK, Harrison LC, et al. ; J. Craig Venter Institute; BIS Investigator Group. Maternal carriage of Prevotella during pregnancy associates with protection against food allergy in the offspring. Nat Commun. 2020. Mar 24;11(1):1452. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Tanabe H, Sakurai K, Kato T, Kawasaki Y, Nakano T, Yamaide F, et al. Association of the maternal microbiome in Japanese pregnant women with the cumulative prevalence of dermatitis in early infancy: A pilot study from the Chiba study of Mother and Child Health birth cohort. World Allergy Organ J. 2019. Nov 1;12(10):100065. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Zimmermann P, Messina N, Mohn WW, Finlay BB, Curtis N. Association between the intestinal microbiota and allergic sensitization, eczema, and asthma: A systematic review. J Allergy Clin Immunol. 2019. Feb;143(2):467–485. [DOI] [PubMed] [Google Scholar]
  • 7.Stein RT, Martinez FD. Asthma phenotypes in childhood: lessons from an epidemiological approach. Paediatr Respir Rev. 2004. Jun;5(2):155–61. [DOI] [PubMed] [Google Scholar]
  • 8.Litonjua AA, Lange NE, Carey VJ, Brown S, Laranjo N, Harshfield BJ, et al. The Vitamin D Antenatal Asthma Reduction Trial (VDAART): rationale, design, and methods of a randomized, controlled trial of vitamin D supplementation in pregnancy for the primary prevention of asthma and allergies in children. Contemp Clin Trials. 2014. May;38(1):37–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Lin H, Peddada SD. Analysis of compositions of microbiomes with bias correction. Nat Commun. 2020. Jul 14;11(1):3514. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Morton JT, Marotz C, Washburne A, Silverman J, Zaramela LS, Edlund A, et al. Establishing microbial composition measurement standards with reference frames. Nat Commun. 2019. Jun 20;10(1):2719. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Wolsk HM, Chawes BL, Litonjua AA, Hollis BW, Waage J, Stokholm J, et al. Prenatal vitamin D supplementation reduces risk of asthma/recurrent wheeze in early childhood: A combined analysis of two randomized controlled trials. PLoS One. 2017. Oct 27;12(10):e0186657. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 12.Litonjua AA, Carey VJ, Laranjo N, Harshfield BJ, McElrath TF, O’Connor GT, et al. Effect of Prenatal Supplementation With Vitamin D on Asthma or Recurrent Wheezing in Offspring by Age 3 Years: The VDAART Randomized Clinical Trial. JAMA. 2016. Jan 26;315(4):362–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Chawes BL, Bønnelykke K, Stokholm J, Vissing NH, Bjarnadóttir E, Schoos AM, et al. Effect of Vitamin D3 Supplementation During Pregnancy on Risk of Persistent Wheeze in the Offspring: A Randomized Clinical Trial. JAMA. 2016. Jan 26;315(4):353–61. [DOI] [PubMed] [Google Scholar]
  • 14.Litonjua AA, Carey VJ, Laranjo N, Stubbs BJ, Mirzakhani H, O’Connor GT, et al. Six-Year Follow-up of a Trial of Antenatal Vitamin D for Asthma Reduction. N Engl J Med. 2020. Feb 6;382(6):525–533. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Lin H, Peddada SD. Analysis of microbial compositions: a review of normalization and differential abundance analysis. NPJ Biofilms Microbiomes. 2020. Dec 2;6(1):60. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Wallen ZD. Comparison study of differential abundance testing methods using two large Parkinson disease gut microbiome datasets derived from 16S amplicon sequencing. BMC Bioinformatics. 2021. May 25;22(1):265. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Vinturache AE, Gyamfi-Bannerman C, Hwang J, Mysorekar IU, Jacobsson B; Preterm Birth International Collaborative (PREBIC). Maternal microbiome - A pathway to preterm birth. Semin Fetal Neonatal Med. 2016. Apr;21(2):94–9. [DOI] [PubMed] [Google Scholar]
  • 18.Olsen I, Jantzen E. Sphingolipids in bacteria and fungi. Anaerobe. 2001;7(2):103–112. [Google Scholar]
  • 19.Shao Y, Forster SC, Tsaliki E, Vervier K, Strang A, Simpson N, et al. Stunted microbiota and opportunistic pathogen colonization in caesarean-section birth. Nature. 2019. Oct;574(7776):117–121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Stokholm J, Thorsen J, Chawes BL, Schjørring S, Krogfelt KA, Bønnelykke K, et al. Cesarean section changes neonatal gut colonization. J Allergy Clin Immunol. 2016. Sep;138(3):881–889.e2. [DOI] [PubMed] [Google Scholar]
  • 21.Korpela K Impact of Delivery Mode on Infant Gut Microbiota. Ann Nutr Metab. 2021. Aug 30:1–9. [DOI] [PubMed] [Google Scholar]
  • 22.Gürdeniz G, Ernst M, Rago D, Kim M, Courraud J, Stokholm J, et al. Neonatal metabolome of cesarean section and risk of childhood asthma. Eur Respir J. 2021. Dec 9:2102406. [DOI] [PubMed] [Google Scholar]
  • 23.Stokholm J, Thorsen J, Blaser MJ, Rasmussen MA, Hjelmsø M, Shah S, et al. Delivery mode and gut microbial changes correlate with an increased risk of childhood asthma. Sci Transl Med. 2020. Nov 11;12(569):eaax9929. [DOI] [PubMed] [Google Scholar]
  • 24.Mitchell CM, Mazzoni C, Hogstrom L, Bryant A, Bergerat A, Cher A, et al. Delivery Mode Affects Stability of Early Infant Gut Microbiota. Cell Rep Med. 2020. Dec 22;1(9):100156. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Princisval L, Rebelo F, Williams BL, Coimbra AC, Crovesy L, Ferreira AL, et al. Association Between the Mode of Delivery and Infant Gut Microbiota Composition Up to 6 Months of Age: A Systematic Literature Review Considering the Role of Breastfeeding. Nutr Rev. 2021. Dec 8;80(1):113–127. [DOI] [PubMed] [Google Scholar]
  • 26.Shaterian N, Abdi F, Ghavidel N, Alidost F. Role of cesarean section in the development of neonatal gut microbiota: A systematic review. Open Med (Wars). 2021. Apr 9;16(1):624–639. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Heaver SL, Johnson EL, Ley RE. Sphingolipids in host-microbial interactions. Curr Opin Microbiol. 2018. Jun;43:92–99. [DOI] [PubMed] [Google Scholar]
  • 28.Korpela K, Helve O, Kolho KL, Saisto T, Skogberg K, Dikareva E, et al. Maternal Fecal Microbiota Transplantation in Cesarean-Born Infants Rapidly Restores Normal Gut Microbial Development: A Proof-of-Concept Study. Cell. 2020. Oct 15;183(2):324–334.e5. [DOI] [PubMed] [Google Scholar]
  • 29.Arrieta MC, Arévalo A, Stiemsma L, Dimitriu P, Chico ME, Loor S, et al. Associations between infant fungal and bacterial dysbiosis and childhood atopic wheeze in a nonindustrialized setting. J Allergy Clin Immunol. 2018. Aug;142(2):424–434.e10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Azad MB, Konya T, Guttman DS, Field CJ, Sears MR, HayGlass KT, et al. ; CHILD Study Investigators. Infant gut microbiota and food sensitization: associations in the first year of life. Clin Exp Allergy. 2015. Mar;45(3):632–43. [DOI] [PubMed] [Google Scholar]
  • 31.Björkstén B, Sepp E, Julge K, Voor T, Mikelsaar M. Allergy development and the intestinal microflora during the first year of life. J Allergy Clin Immunol. 2001. Oct;108(4):516–20. [DOI] [PubMed] [Google Scholar]
  • 32.Tun HM, Peng Y, Chen B, Konya TB, Morales-Lizcano NP, Chari R, et al. Ethnicity Associations With Food Sensitization Are Mediated by Gut Microbiota Development in the First Year of Life. Gastroenterology. 2021. Jul;161(1):94–106. [DOI] [PubMed] [Google Scholar]
  • 33.Abrahamsson TR, Jakobsson HE, Andersson AF, Björkstén B, Engstrand L, Jenmalm MC. Low diversity of the gut microbiota in infants with atopic eczema. J Allergy Clin Immunol. 2012. Feb;129(2):434–40, 440.e1–2. [DOI] [PubMed] [Google Scholar]
  • 34.Varki A Sialic acids in human health and disease. Trends Mol Med. 2008. Aug;14(8):351–60. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Shade KC, Conroy ME, Washburn N, Kitaoka M, Huynh DJ, Laprise E, et al. Sialylation of immunoglobulin E is a determinant of allergic pathogenicity. Nature. 2020. Jun;582(7811):265–270. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Han Z, Thuy-Boun PS, Pfeiffer W, Vartabedian VF, Torkamani A, Teijaro JR, et al. Identification of an N-acetylneuraminic acid-presenting bacteria isolated from a human microbiome. Sci Rep. 2021. Feb 26;11(1):4763. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Birch EE, Khoury JC, Berseth CL, Castañeda YS, Couch JM, Bean J, et al. The impact of early nutrition on incidence of allergic manifestations and common respiratory illnesses in children. J Pediatr. 2010. Jun;156(6):902–906.e1. [DOI] [PubMed] [Google Scholar]
  • 38.Ekström S, Sdona E, Klevebro S, et al. Dietary and plasma levels of polyunsaturated fatty acids in childhood and adolescence in relation to asthma and lung function up to adulthood. Am J Clin Nutr. December 2021. doi: 10.1093/ajcn/nqab427 [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Lee-Sarwar KA, Kelly RS, Lasky-Su J, Zeiger RS, O’Connor GT, Sandel MT, et al. Integrative analysis of the intestinal metabolome of childhood asthma. J Allergy Clin Immunol. 2019. Aug;144(2):442–454. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Lee-Sarwar K, Kelly RS, Lasky-Su J, Kachroo P, Zeiger RS, O’Connor GT, et al. Dietary and Plasma Polyunsaturated Fatty Acids Are Inversely Associated with Asthma and Atopy in Early Childhood. J Allergy Clin Immunol Pract. 2019. Feb;7(2):529–538.e8. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Guilbert TW, Mauger DT, Lemanske RF Jr. Childhood asthma-predictive phenotype. J Allergy Clin Immunol Pract. 2014. Nov-Dec;2(6):664–70. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

SUPINFO2
NIHMS1838692-supplement-SUPINFO2.xlsx (95.5KB, xlsx)
SUPINFO1
NIHMS1838692-supplement-SUPINFO1.docx (676.8KB, docx)

RESOURCES