Intestinal microbiome is related to lifetime antibiotic use in Finnish pre-school children

Abstract

Early-life antibiotic use is associated with increased risk for metabolic and immunological diseases, and mouse studies indicate a causal role of the disrupted microbiome. However, little is known about the impacts of antibiotics on the developing microbiome of children. Here we use phylogenetics, metagenomics and individual antibiotic purchase records to show that macrolide use in 2-7 year-old Finnish children (N=142; sampled at two time points) is associated with a long-lasting shift in microbiota composition and metabolism. The shift includes depletion of Actinobacteria, increase in Bacteroidetes and Proteobacteria, decrease in bile-salt hydrolase and increase in macrolide resistance. Furthermore, macrolide use in early life is associated with increased risk of asthma and predisposes to antibiotic-associated weight gain. Overweight and asthmatic children have distinct microbiota compositions. Penicillins leave a weaker mark on the microbiota than macrolides. Our results support the idea that, without compromising clinical practice, the impact on the intestinal microbiota should be considered when prescribing antibiotics.

Figure 1. Microbiota composition in 257 fecal samples as arranged per group.

C denotes control group, no antibiotics for the past 2 years and in total <1 course per year of life on average. E denotes early-life exposure group, no antibiotics for the past 2 years and >1 course per year of life on average. M6 denotes macrolide course within 6 months; M12 denotes macrolide course within 6–12 months; M24 denotes macrolide course within 12–24 months. P6, P12 and P24 denote penicillin courses within 6, 6–12 and 12–24 months, respectively. (a) Phyla composition. (b) Genus-level microbiota composition according to PCoA analysis. The background colour indicates interpolated time since the last macrolide course.

Figure 2. Microbiota richness and maturity relative to age in the different groups.

(a) Number of species-like phylotypes. (b) Score based on 24 age-associated genera (Supplementary Table 2). Asterisks indicate significance of the difference to the control group ‘C'. *P<0.05, **P<0.01 and ***P<0.001, estimated using linear models. Mean values are shown and the error bars show the s.e. values. The number of samples in both panels is 257 (Supplementary Table 1).

Figure 3. Macrolide resistance and bile-salt hydrolase abundance in relation to time since the last macrolide course.

The dashed lines show the model fit (linear or polynomial), R² indicates the variation explained by the model and the P values (estimated using linear models) are indicated. (a) Macrolide resistance potential inferred from metagenomic analysis, N=14. (b) Bile-salt hydrolase abundance in metagenomes, N=14. (c) Macrolide resistance measured as proportion of anaerobic c.f.u.'s growing with erythromycin compared with c.f.u. without erythromycin, N=80. (d) Combined relative abundance of three bile-salt hydrolase genes (bsh) based on qPCR as a function of time since last macrolide course, N=37.