Prenatal dietary supplements influence the infant airway microbiota in a randomized factorial clinical trial

Abstract

Maternal dietary interventions during pregnancy with fish oil and high dose vitamin D have been shown to reduce the incidence of asthma and wheeze in offspring, potentially through microbial effects in pregnancy or early childhood. Here we analyze the bacterial compositions in longitudinal samples from 695 pregnant women and their children according to intervention group in a nested, factorial, double-blind, placebo-controlled, randomized trial of n-3 long-chain fatty acids and vitamin D supplementation. The dietary interventions affect the infant airways, but not the infant fecal or maternal vaginal microbiota. Changes in overall beta diversity are observed, which in turn associates with a change in immune mediator profile. In addition, airway microbial maturation and the relative abundance of specific bacterial genera are altered. Furthermore, mediation analysis reveals the changed airway microbiota to be a minor and non-significant mediator of the protective effect of the dietary interventions on risk of asthma. Our results demonstrate the potential of prenatal dietary supplements as manipulators of the early airway bacterial colonization.

Fig. 1. Bacterial community composition in children’s feces and airway, and maternal vaginal samples over time.

The community composition is represented by the 12 most abundant genera. Each bar is represented by 544 to 665 samples. Source data are provided as a Source Data file.

Fig. 2. Relative bacterial abundances stratified by intervention groups for the 1-month airway samples.

Relative abundances are shown at the phylum-level (a and b), and genus-level (c and d). The three most abundant phyla and nine most abundant genera are shown. Phylum level DA statistics were performed using Wilcoxon rank sum test and genus level DA using the metagenomeseq feature model. Boxplots with first and third quartiles corresponding to the lower and upper hinge, the median represented by a vertical line, the mean by a black dot, upper/lower whiskers extend to the largest/smallest value no further than 1.5 * inter-quartile range (IQR) from the hinge, and outliers are shown as gray circles. In bold are crude p-values reaching FDR corrected significance at level 0.1. n = 653 and 541 independent samples for the n-3 LCPUFA (a + c) and Vitamin D (b + d) plot, respectively. Source data are provided as a Source Data file.

Fig. 3. Ordination of 1-month airway samples stratified by intervention group.

An additive effect is apparent in the right shift observed over PCo1 as samples are subject to either (n-3 LCPUFA/placebo (n = 125), or placebo/vitamin D (n = 141)) or both prenatal dietary interventions (n-3 LCPUFA/vitamin D (n = 131)) compared to double-placebo (placebo/placebo (n = 144)) (p < 0.001, Wilcoxon rank sum test). Boxplots represent the PCo1 values of each intervention group with first and third quartiles corresponding to the lower and upper hinge, the median represented by a vertical line, and the upper/lower whiskers extend to the largest/smallest value no further than 1.5 * inter-quartile range (IQR) from the hinge. Source data are provided as a Source Data file.

Fig. 4. Additive effects of the interventions on the 1-month airway samples.

Relative abundances are shown on (a) Phylum level and (b) Genus level. The correlation was analyzed by Spearman’s rank correlation coefficient and p-values below 0.05 are shown. The p-values in b are not significant when correcting for multiple testing using an FDR corrected significance of 0.1. Boxplots with first and third quartiles corresponding to the lower and upper hinge, the median represented by a vertical line, the mean by a black dot, upper/lower whiskers extend to the largest/smallest value no further than 1.5 * inter-quartile range (IQR) from the hinge, and outliers are shown as gray circles. n = 144, 266, and 131 for the zero, one, and two intervention groups, respectively. Source data are provided as a Source Data file.

Fig. 5. Associations between intervention driven microbiota changes in the 1-month airway and neonatal airway immunology.

PCo1 from Fig. 3 was used as a metric for the intervention driven microbial changes and analyzed with the airway concentration of 20 local immune mediators in the nose, also measured at the 1-month time point. Linear models show that PCo1 is associated with several immune mediators, expressed as relative concentration ratios of immune mediators per standard deviation (SD) increase in PCo1, n = 585 independent samples. Error bars represent 95% confidence intervals. Source data are provided as a Source Data file.