Gut resistome development in healthy twin pairs in the first year of life

Abstract

Background: The early life of the human host marks a critically important time for establishment of the gut microbial community, yet the developmental trajectory of gut community-encoded resistance genes (resistome) is unknown. We present a longitudinal study of the fecal antibiotic resistome of healthy amoxicillin-exposed and antibiotic-naive twins and their mothers during the first year of life.

Results: We extracted metagenomic DNA (mgDNA) from fecal samples collected from three healthy twin pairs at three timepoints (1 or 2 months, 6 or 7 months, and 11 months) and from their mothers (collected at delivery). The mgDNA was used to construct metagenomic expression libraries in an Escherichia coli host. These libraries were screened for antibiotic resistance, and functionally selected resistance genes were sequenced and annotated. A diverse fecal resistome distinct from the maternal resistome was apparent by 2 months of age, and infants' fecal resistomes included resistance to clinically important broad-spectrum beta-lactam antibiotics (e.g., piperacillin-tazobactam, aztreonam, cefepime) not found in their mothers. Dissemination of resistance genes among members of a given family was positively correlated with sharing of those same resistance genes between unrelated families, potentially identifying within-family sharing as a marker of resistance genes emerging in the human community at large. Finally, we found a distinct developmental trajectory for a community-encoded function: chloramphenicol resistance. All study subjects at all timepoints harbored chloramphenicol resistance determinants, but multidrug efflux pumps (rarely found in mothers) were the primary effectors of chloramphenicol resistance in young infants. Chloramphenicol acetyltransferases were more common in mothers than in infants and were found in nearly all the infants at later timepoints.

Conclusions: Our results suggest that healthy 1-2-month-old infants' gut microbes harbor clinically relevant resistance genes distinct from those of their mothers, and that family-specific shared environmental factors early in life shape resistome development.

Keywords: Antibiotic resistance; Beta lactamase; Chloramphenicol resistance; Fecal microbiome; Gut microbiome; Pediatrics.

Fig. 1

Twin infant fecal resistomes resemble those of their siblings. Predicted resistance proteins were collapsed into 97 % identity clusters. Binary Jaccard resistance protein cluster composition similarity was determined for (1) the same infant at different timepoints (self-sharing), (2) twin siblings, (3) unrelated infants, and (4) mothers and infants from the same family. a All resistance proteins at left; b the subset of β-lactamases and penicillin-binding proteins at right. Significance was calculated using the Student’s t-test with 1000 Monte Carlo simulations (**p < 0.01). Infant resistomes overall (a) were significantly more similar to a twin sibling or to the same subject at different timepoints than to their mothers or unrelated infants. Infant resistomes were no more similar to those of their mothers than to unrelated infants. There was also no significant difference between the similarity between infants and their twin sibling and the persistence of resistance proteins within a given individual at different timepoints (self-sharing)

Fig. 2

Sharing of resistance-associated proteins within and between families. The top graph shows absolute counts of resistance-associated protein clusters, grouped by the number of members within a given family they were identified in. Most protein clusters were only identified in one member of one family; much smaller numbers were identified in multiple family members or multiple families. The lower graph shows the proportions of resistance-associated protein clusters identified in multiple families, grouped by the number of members of a single family they were identified in. Larger proportions of resistance-associated proteins that are shared are multiple members of a single family are also identified in multiple unrelated families

Fig. 3

β-lactamase phylogenetic tree. Predicted β-lactamase protein sequences were collapsed into 97 % ID clusters. All β-lactamase protein sequences with at least 90 % coverage of the nearest hit in the NCBI nr database were included in the tree. Multiple alignment was done with Muscle and the tree was made using FastTree. Nodes with an S-H value >0.7 are marked with a square. All classes of β-lactamases are present. Class A β-lactamases separated into two groups: one with high identity to TEM extended-spectrum β-lactamases and one without. β-lactamases co-localized with mobile genetic elements are marked with a gray dot. Novel β-lactamases with less than 70 % identity to any known β-lactamase are marked with a star. β-lactamases found in cefepime selections are marked with a triangle

Fig. 4

Populations of chloramphenicol resistance proteins change over time. Predicted proteins found when fecal metagenomic libraries were screened on chloramphenicol-containing media were collapsed into 97 % ID clusters. Black boxes signify genes encoding a resistance protein that were identified in the fecal metagenome of a study subject at a given timepoint, while white or light gray squares indicate that the protein was not present. Proteins that were co-localized with a mobile genetic element are marked with an asterisk. Chloramphenicol acetyltransferases were found in all mothers and in five of the six infants at the final timepoint, but were qualitatively less common in infants at earlier timepoints. By contrast, multidrug efflux pumps were rare in mothers and in 11-month-old infants, but were commonly found in earlier samples

Fig. 5

β-lactam phylogenetic tree, annotated by study subject. Maternal subjects are marked with an “M”. Infant fecal samples are marked with a number; “1” indicating the first (baseline) sample collected at 1–2 months of age, “2” indicating the second sample collected at 6–7 months of age, and “3” indicating the third sample collected at 11 months of age. The antibiotic-naïve control family is colored green, the family with infants discordant for amoxicillin exposure at 8 months of age is colored purple, and the family with infants concordant for amoxicillin exposure at 8 months of age is colored brown. Infant twin A subjects are shaded darker; twin B subjects are shaded lighter. β-lactamases were commonly present in both members of a twin pair, and frequently persisted at more than one timepoint within a given subject. Many β-lactamases identified in the infant fecal microbiomes were not present in the maternal microbiome