Preterm infants harbour diverse Klebsiella populations, including atypical species that encode and produce an array of antimicrobial resistance- and virulence-associated factors

Klebsiella spp. are frequently enriched in the gut microbiota of preterm neonates, and overgrowth is associated with necrotizing enterocolitis, nosocomial infections and late-onset sepsis. Little is known about the genomic and phenotypic characteristics of preterm-associated Klebsiella as previous studies have focussed on recovery of antimicrobial-resistant isolates or culture-independent molecular analyses. Faecal samples from a UK cohort of healthy and sick preterm neonates (n=109) were screened on MacConkey agar to isolate lactose-positive Enterobacteriaceae. Whole-genome sequences were generated for isolates. Approximately one-tenth of faecal samples harboured Klebsiella spp. (Klebsiella pneumoniae, 7.3 %; Klebsiella quasipneumoniae, 0.9 %; Klebsiella grimontii, 2.8 %; Klebsiella michiganensis, 1.8 %). Isolates recovered from NEC- and sepsis-affected infants and those showing no signs of clinical infection (i.e. healthy) encoded multiple {beta}-lactamases, which may prove problematic when defining treatment regimens for NEC or sepsis, and suggest healthy preterm infants contribute to the resistome. No difference was observed between isolates recovered from healthy and sick infants with respect to in vitro siderophore production (all encoded enterobactin in their genomes). All K. pneumoniae, K. quasipneumoniae, K. grimontii and K. michiganensis faecal isolates tested were able to reside and persist in macrophages, indicating their immune evasion abilities. Using a curated dataset of Klebsiella oxytoca, K. grimontii and K. michiganensis whole-genome sequences, metapangenome analyses of published metagenomic data confirmed our findings regarding the presence of K. michiganensis in the preterm gut, and highlight the importance of refined analyses with curated sequence databases when studying closely related species present in metagenomic data.