Dynamics of gut resistome and mobilome in early life: a meta-analysis

Abstract

Background: The gut microbiota of infants harbours a higher proportion of antibiotic resistance genes (ARGs) compared to adults, even in infants never exposed to antibiotics. Our study aims to elucidate this phenomenon by analysing how different perinatal factors influence the presence of ARGs, mobile genetic elements (MGEs), and their bacterial hosts in the infant gut.

Methods: We searched MEDLINE and Embase up to April 3rd, 2023, for studies reporting infant cohorts with shotgun metagenomic sequencing of stool samples. The systematic search identified 14 longitudinal infant cohorts from 10 countries across three continents, featuring publicly available sequencing data with corresponding metadata. For subsequent integrative bioinformatic analyses, we used 3981 high-quality metagenomic samples from 1270 infants and 415 mothers.

Findings: We identified distinct trajectories of the resistome and mobilome associated with birth mode, gestational age, antibiotic use, and geographical location. Geographical variation was exemplified by differences between cohorts from Europe, Southern Africa, and Northern America, which showed variation in both diversity and abundance of ARGs. On the other hand, we did not detect a significant impact of breastfeeding on the infants' gut resistome. More than half of detected ARGs co-localised with plasmids in key bacterial hosts, such as Escherichia coli and Enterococcus faecalis. These ARG-associated plasmids were gradually lost during infancy. We also demonstrate that E. coli role as a primary modulator of the infant gut resistome and mobilome is facilitated by its increased abundance and strain diversity compared to adults.

Interpretation: Birth mode, gestational age, antibiotic exposure, and geographical location significantly influence the development of the infant gut resistome and mobilome. A reduction in E. coli relative abundance over time appears as a key factor driving the decrease in both resistome and plasmid relative abundance as infants grow.

Funding: Centre for Advanced Study in Oslo, Norway. Centre for New Antibacterial Strategies through the Tromsø Research Foundation, Norway.

Keywords: Bacterial drug resistance; Cohort studies; Escherichia coli; Gut microbiota; Infant; Meta-analysis; Metagenomics; Mobile genetic elements; Plasmids.

Fig. 1

Infant gut resistome dynamics. (a) Map indicating the geographical locations of included cohorts. (b) Overview of the cohorts' characteristics. A grey bar in panel B denotes missing data. The impact of selected variables on the infant gut resistome α-diversity (c) and abundance (d), as determined by a linear mixed effect modelling (LMM) (n = 3497 samples from 1270 infants). The fixed effects used in LMM included infant age, the presence/absence of bead-beating, and the use of various sequencing platforms. The estimates are in Confidence Interval (CI) of 95% as determined by LMM. The green error bars indicate statistically significant results (p < 0.05), while the orange error bars represent non-significant results (p > 0.05). (e) The relative abundance of the gut resistome quantified in reads per kilobase per million mapped reads (RPKM). The main antibiotic resistant genes families in three selected infant groups as compared to maternal samples. The ‘C-section’ (n = 330 samples from 173 infants) and ‘Vaginal’ groups (n = 750 samples from 513 infants) are full-term infants that were exclusively breastfed the first three months of life and not exposed to antibiotics during the displayed period. The ‘Premature’ group (n = 372 samples from 135 infants) is preterm babies without discrimination of birth mode, feeding, or exposure to antibiotics. In addition, 484 maternal samples were used in the analysis. Multi-drug is ARG conferring resistance to two or more antibiotic classes.

Fig. 2

Infant gut mobilome dynamics. The impact of selected variables on the infant gut mobilome α-diversity (a) and abundance (b) as determined by linear mixed effect modelling (LMM) (n = 3497 samples from 1270 infants). The fixed effects used in LMM included infant age, the presence/absence of bead-beating, and the use of various sequencing platforms. The green error bars indicate statistically significant results (p < 0.05), while the orange error bars represent non-significant results (p > 0.05). (c) The relative abundance in reads per kilobase per million mapped reads (RPKM) of three mobile genetic element classes detected in the metagenomic data of infant and maternal samples. The ‘C-section’ (n = 330 samples from 173 infants) and ‘Vaginal’ groups (n = 750 samples from 513 infants) cover samples from full-term infants that were exclusively breastfed the first three months of life and not exposed to antibiotics during the displayed period. The ‘Premature’ group (n = 372 samples from 135 infants) covers samples from preterm infants without discrimination of birth mode, feeding, or exposure to antibiotics. In addition, 484 maternal samples were used in the analysis.

Fig. 3

Major bacterial hosts of infant gut resistome. The heatmap shows the prevalence of antibiotic resistant gene (ARG) families in (a) ‘Vaginal’, (b) ‘C-section’, and (c) ‘Premature’ infant groups for the top 20 bacterial species associated with ARGs. The scale represents the log₂ of the ARG count. Red stars in the heatmaps denote five bacterial species, which are significant contributors to global mortality, and these are also shown in Sankey diagrams of panels d–f. Each Sankey diagram connects indicated bacterial species to ARG families and mobile genetic element categories for samples from (d) ‘Vaginal’, (e) ‘C-section’, and (f) ‘Premature’ infant groups. The length of each species node signifies the total count of resistance genes, mobile genetic elements, or contigs assigned to a specific bacterium within each group, enabling intra-group analysis but not direct comparison between the infant groups (i.e., the node lengths are not comparable across d, e, and f panels). The ‘C-section’ (n = 330 samples from 173 infants) and ‘Vaginal’ groups (n = 750 samples from 513 infants) are full-term infants that were exclusively breastfed the first three months of life and not exposed to antibiotics. The ‘Premature’ group (n = 372 samples from 135 infants) is preterm babies without discrimination of birth mode, feeding, or exposure to antibiotics. NA denotes contigs that have been assigned to both species and resistance genes but not to mobile genetic elements. Multi-drug is ARG conferring resistance to two or more antibiotic classes.

Fig. 4

E. coli as a major host of antibiotic resistome. (a) The relative abundance of each antibiotic resistant gene (ARG) family, measured in reads per kilobase per million mapped reads (RPKM), and predicted associations with six microbiota community (MC) types identified through Non-negative Matrix Factorisation (n = 3497 samples from 1270 infants). (b) A comparison of E. coli relative abundance between infants (n = 3497 samples from 1270 infants) and mothers (n = 484). The trajectory of E. coli abundance over time was analysed using linear mixed-effects modelling (LMM). The association between the α-diversity of E. coli strains and ARGs (c) and the relative abundance of E. coli strains and ARGs (d) in three selected infant groups was determined by Spearman's correlation. (e) The impact of selected variables on E. coli strains α-diversity employing LMM (n = 3497 samples from 1270 infants). The fixed effects used in LMM included infant age, presence/absence of bead-beating, and the use of various sequencing platforms. The green colour indicates a significant association, and the estimates are in a Confidence Interval (CI) of 95% as determined by LMM. The ‘C-section’ (n = 330 samples from 173 infants) and ‘Vaginal’ groups (n = 750 samples from 513 infants) are full-term infants that were exclusively breastfed the first three months of life and not exposed to antibiotics during the displayed period. The ‘Premature’ group (n = 372 samples from 135 infants) is preterm babies without discrimination of birth mode, feeding, or exposure to antibiotics. The green error bars indicate statistically significant results (p < 0.05), while the orange error bars represent non-significant results (p > 0.05).