Abstract
Early in life, infants are colonized with multiple bacterial strains whose differences in gene content can have important health consequences. Metagenomics-based approaches have revealed gene content differences between different strains co-colonizing newborns, but less is known about the rate, mechanism, and phenotypic consequences of gene content diversification within strains. Here, focusing on Staphylococcus epidermidis, we whole-genome sequence and phenotype more than 600 isolates from newborns. Within days of birth, infants are co-colonized with a highly personalized repertoire of S. epidermidis strains, which are spread across the newborn body. Comparing the genomes of multiple isolates of each strain, we find very little evidence of adaptive evolution via single-nucleotide polymorphisms. By contrast, we observe gene content differences even between otherwise genetically identical cells, including variation of the clinically important methicillin resistance gene, mecA, suggesting rapid gene gain and loss events at rates higher than point mutations. Mapping the genomic architecture of structural variants by long-read Nanopore sequencing, we find that deleted regions were always flanked by direct repeats, consistent with site-specific recombination. However, we find that even within a single genetic background, recombination occurs at multiple, often non-canonical repeats, leading to the rapid evolution of patient-specific diverse structural variants in the SCCmec island and to differences in antibiotic resistance.