Antibiotics and the developing intestinal microbiome, metabolome and inflammatory environment in a randomized trial of preterm infants

Abstract

Antibiotic use in neonates can have detrimental effects on the developing gut microbiome, increasing the risk of morbidity. A majority of preterm neonates receive antibiotics after birth without clear evidence to guide this practice. Here microbiome, metabolomic, and immune marker results from the routine early antibiotic use in symptomatic preterm Neonates (REASON) study are presented. The REASON study is the first trial to randomize symptomatic preterm neonates to receive or not receive antibiotics in the first 48 h after birth. Using 16S rRNA sequencing of stool samples collected longitudinally for 91 neonates, the effect of such antibiotic use on microbiome diversity is assessed. The results illustrate that type of nutrition shapes the early infant gut microbiome. By integrating data for the gut microbiome, stool metabolites, stool immune markers, and inferred metabolic pathways, an association was discovered between Veillonella and the neurotransmitter gamma-aminobutyric acid (GABA). These results suggest early antibiotic use may impact the gut-brain axis with the potential for consequences in early life development, a finding that needs to be validated in a larger cohort.Trial Registration This project is registered at clinicaltrials.gov under the name "Antibiotic 'Dysbiosis' in Preterm Infants" with trial number NCT02784821.

Conflict of interest statement

Dr. Josef Neu is the principal investigator of a study with Infant Bacterial Therapeutics and on the Scientific Advisory Boards of Medela and Astarte. No other authors have competing interest to disclose.

Figures

Figure 1

Antibiotic use 48 h after birth does not significantly affect alpha diversity development. Boxplots displaying (A) the observed ASV richness (B) the Shannon diversity and (C) log10-transformed copies of 16S rRNA by enrollment group across corrected gestational ages between weeks 28 and 39. P values were calculated at each corrected GA timepoint between enrollment groups using the non-parametric Kruskal–Wallis test. Linear mixed-effects modeling of the (D) observed ASV richness and (E) Shannon diversity over time between enrollment groups. Time scale on the x-axis is days of life (DOL) for corrected GA weeks 28–39. Greyed areas around each regression line represent 95% confidence intervals upper and lower around the coefficients.

Figure 2

Antibiotic use explains little effect on beta diversity. PCoA ordination of stool samples using the (A) abundance-based Bray–Curtis and (B) presence/absence-based Jaccard distance metric among enrollment groups of corrected GA between 28–39. Ellipses are calculated based on a 95% confidence interval of a multivariate t-distribution.

Figure 3

Effects of feeding patterns on the gut microbiome are transient over time. Effect of various feeding patterns on the (A) observed ASV richness (B) Shannon diversity and (C) log10-transformed copies of 16S rRNA by enrollment group over corrected GA from weeks 28–39. Only feeding patterns that have at least 2 samples at each timepoint were kept. Statistical comparison of feeding patterns at each corrected GA timepoint was performed using the non-parametric Kruskal–Wallis test.

Figure 4

Linear mixed effects modelling identifies feeding effects by group over time. Linear mixed-effects results plotted as observed bacterial richness over time by group and feeding types (A–F) and the Shannon diversity over time by group and feeding types (G–L). The number of samples used in this analysis by group and by feeding type are listed in Supplementary Table S3.

Figure 5

Integration of clinical and laboratory data gives detailed view of infant stay in NICU. Extensive clinical and laboratory data, when combined, provide a detailed summary of each infant’s stay in the NICU. Data included in each chart from top to bottom include: the infant ID, group assignment, antibiotic change status (bail), gestational age, any adverse clinical events (which are further described in Supplementary Figure S1), the type and duration of antibiotic use (if any), the copy-number corrected absolute composition of each weekly stool sample and it’s log10-scale number of bacterial 16S rRNA copies, the type and duration of each feeding including administration of human milk fortifier, the relative levels of C-reactive protein measured from blood, and relative concentrations of measured stool immune markers (for infants where these measurements were performed). DBM: donor breast milk, MBM: mother’s breast milk, NPO: no enteral nutrition, CRP: C-reactive protein, EGF: epidermal growth factor. Listed below is a key for the color-coded bacterial taxa used in the stool 16S rRNA copy-number corrected composition pie chart for each infant chart. A key for the bacterial color codes, adverse clinical events (including infections by body site) and the administration of human milk fortifier for each infant chart is given below.

Figure 6

Metabolites in stool correlate with abundance of bacterial genera. (A) Heatmap of repeated measures correlation coefficients between peak response heights of identified metabolites in stool and the top 10 bacterial genera from the same samples (n = 90 stool samples). Significant correlations are indicated by a ‘ + ’ with FDR-corrected p values < 0.05. (B) Boxplot comparing the peak response heights for 4-aminobutyric acid (GABA) between enrollment groups. Statistical comparisons were made using the Wilcoxon test. (C) Boxplot comparing the number of rarefied Veillonella counts between the enrollment groups. Statistical comparisons were made using the Wilcoxon test. A summary of the number of infants and samples by group for metabolomics is given in Table 1.

Figure 7

Stool immune marker levels show modest correlation with gut microbiota. (A) Heatmap of repeated measures correlation coefficients between immune markers measured from stool and the most abundant bacterial genera from the same samples (n = 110 stool samples). Only the bacterial genera with at least one significant correlation with an immune marker are displayed (10 genera). Significant correlations are marked with an ‘*’ by the coefficient, with FDR-adjusted p-values < 0.05. (B) Table listing the immune markers used for correlation analysis and their commonly known general functions. (C) Comparison of log10-transformed number of Citrobacter counts by enrollment group and their significance by the Wilcoxon test. A summary of the number of infants and samples by group for immune marker analysis is given in Table 1.