Early-life gut microbiome modulation reduces the abundance of antibiotic-resistant bacteria

Abstract

Background: Antibiotic-resistant (AR) bacteria are a global threat. AR bacteria can be acquired in early life and have long-term sequelae. Limiting the spread of antibiotic resistance without triggering the development of additional resistance mechanisms is of immense clinical value. Here, we show how the infant gut microbiome can be modified, resulting in a significant reduction of AR genes (ARGs) and the potentially pathogenic bacteria that harbor them.

Methods: The gut microbiome was characterized using shotgun metagenomics of fecal samples from two groups of healthy, term breastfed infants. One group was fed B. infantis EVC001 in addition to receiving lactation support (n = 29, EVC001-fed), while the other received lactation support alone (n = 31, controls). Coliforms were isolated from fecal samples and genome sequenced, as well as tested for minimal inhibitory concentrations against clinically relevant antibiotics.

Results: Infants fed B. infantis EVC001 exhibited a change to the gut microbiome, resulting in a 90% lower level of ARGs compared to controls. ARGs that differed significantly between groups were predicted to confer resistance to beta lactams, fluoroquinolones, or multiple drug classes, the majority of which belonged to Escherichia, Clostridium, and Staphylococcus. Minimal inhibitory concentration assays confirmed the resistance phenotypes among isolates with these genes. Notably, we found extended-spectrum beta lactamases among healthy, vaginally delivered breastfed infants who had never been exposed to antibiotics.

Conclusions: Colonization of the gut of breastfed infants by a single strain of B. longum subsp. infantis had a profound impact on the fecal metagenome, including a reduction in ARGs. This highlights the importance of developing novel approaches to limit the spread of these genes among clinically relevant bacteria. Future studies are needed to determine whether colonization with B. infantis EVC001 decreases the incidence of AR infections in breastfed infants.

Trial registration: This clinical trial was registered at ClinicalTrials.gov, NCT02457338.

Keywords: Antibiotic resistance gene; Host-microbe interactions; Metagenomics; Microbiology; Probiotics.

Conflict of interest statement

Competing interestsGC, RMD, SAF, LC, RM, and DV are employed by Evolve BioSystems, Inc.

Figures

Fig. 1

Taxonomic classification of metagenomic reads for EVC001-fed infants and controls. a Relative abundance of the top bacterial genera identified between the two groups of infants. b Relative abundance of bacterial species belonging to the Bifidobacterium genus identified among groups. c Hierarchical clustering based on a strain-level analysis of Bifidobacterium longum subspecies. Gene family profiles of a subgroup of reference genomes were selected from a global (n = 38) strain analysis. Each column represents the presence or absence of genes in a sample or a reference genome with respect to the total pangenome. All samples from EVC001-fed infants clustered together with B. longum subsp. infantis ATCC 15697 (B. infantis), whereas the samples from infants in the control group clustered separately with other B. longum subspecies (e.g., B. suis, B. longum DJ01A, and B. longum NCC2705). Functional analysis of gene families confirmed that the EVC001 samples were dominated by B. infantis due to the presence of unique genes (e.g., Blon_2348 in B. infantis), while genes present only in B. longum subsp. longum (e.g., araD; araA), were abundant in the communities from control infants. P-values were computed for each gene via Fisher’s exact test according to group

Fig. 2

Relative abundance of the total resistome profile in each metagenome sample. a Relative abundance of antibiotic resistance genes (ARGs) compared with the overall metagenome for each sample. Each point represents a sample resistome (control, n = 31; EVC001-fed, n = 29). Box plots denote the interquartile range (IQR), with horizontal lines representing the 25th percentile, median, and 75th percentiles. The whiskers represent the lowest and highest values within 1.5 times the IQR from the first and third quartiles, respectively. The asterisks on the top indicate significant P-values (Mann-Whitney test). b Relative abundance of ARGs according to their taxonomic identification. The shade of color represents genera belonging to the same bacterial class. The asterisks on the top indicate significant P-values (Kruskal-Wallis test)

Fig. 3

Comparison of the most significant antibiotic resistance gene types. a Relative abundance of the most significantly different antibiotic resistance genes (ARGs) identified among EVC001-fed infants and controls (P < 0.02; Bonferroni). Percentages are relative to the overall metagenome. These ARGs confer resistance to different drug classes, including beta-lactams, fluoroquinolones, and macrolides. The ARGs are grouped by gene name, followed by CARD identification entry (ARO). The colored bars represent respective drug class to which the ARG is known to confer resistance to. b Heatmap showing a hierarchical cluster analysis of the total ARGs identified (n = 652). Two groups were identified, one characterized by a lower-ARG carriage, containing most of the samples from infants fed EVC001 and one characterized by a higher overall carriage, containing most samples from infants in the control group. Genes clustered based on similar biological mechanisms implicated in drug resistance (see Results). P-values on the bar were computed using a Kruskal-Wallis test normalized with a Bonferroni correction

Fig. 4

Quantification of Enterobacteriaceae family by group-specific qPCR. The data are represented as Log10 CFU per μg of genomic DNA extracted from stool samples. Data in boxplots show the median, first, and third quartiles (P < 0.0001, Mann-Whitney Test)

Fig. 5

Diversity analyses of infant resistomes according to B. infantis EVC001 colonization. a Rarefaction curves showing the number of unique antibiotic resistance genes (ARGs) identified in relation to the increasing number of sequences. Both EVC001 and the control group presented similar curve trends, suggesting that the sequencing depth is not associated with the diversity of antibiotic resistance. P-values were computed with a nonparametric two-sample t-test using Monte Carlo permutations (n = 999). b Global resistome profiles computed via a principal coordinate analysis (PCoA) based on a Bray-Curtis dissimilarity matrix. The EVC001-fed samples clustered closely, indicating that they shared a similar resistome compared to the controls, which had a more dispersed distribution. The effect of B. infantis EVC001 colonization by itself accounted for 31% of the total explained variation (adonis). The P-value was computed using F-tests based on the sequential sums of squares from permutations of the raw data