LAUSR

Significance In many clinically important bacteria, antibiotic resistance genes are primarily carried on plasmids. These can spread horizontally between different strains and species. However, current surveillance systems track chromosomal lineages of bacteria only, leading to an incomplete understanding of how resistance spreads, from within an individual hospital to across country borders. We present an integrated, high-resolution analysis of both chromosome and plasmid sequences using Klebsiella pneumoniae isolates sampled during a European survey. We show that carbapenemase genes, which confer resistance to last-line antibiotics, have spread in diverse ways including via one plasmid/multiple lineages (blaOXA-48-like), multiple plasmids/multiple lineages (blaVIM, blaNDM), and multiple plasmids/one lineage (blaKPC). These different trajectories must be considered in genomic surveillance systems and the design of new interventions. Molecular and genomic surveillance systems for bacterial pathogens currently rely on tracking clonally evolving lineages. By contrast, plasmids are usually excluded or analyzed with low-resolution techniques, despite being the primary vectors of antibiotic resistance genes across many key pathogens. Here, we used a combination of long- and short-read sequence data of Klebsiella pneumoniae isolates (n = 1,717) from a European survey to perform an integrated, continent-wide study of chromosomal and plasmid diversity. This revealed three contrasting modes of dissemination used by carbapenemase genes, which confer resistance to last-line carbapenems. First, blaOXA-48-like genes have spread primarily via the single epidemic pOXA-48–like plasmid, which emerged recently in clinical settings and spread rapidly to numerous lineages. Second, blaVIM and blaNDM genes have spread via transient associations of many diverse plasmids with numerous lineages. Third, blaKPC genes have transmitted predominantly by stable association with one successful clonal lineage (ST258/512) yet have been mobilized among diverse plasmids within this lineage. We show that these plasmids, which include pKpQIL-like and IncX3 plasmids, have a long association (and are coevolving) with the lineage, although frequent recombination and rearrangement events between them have led to a complex array of mosaic plasmids carrying blaKPC. Taken altogether, these results reveal the diverse trajectories of antibiotic resistance genes in clinical settings, summarized as using one plasmid/multiple lineages, multiple plasmids/multiple lineages, and multiple plasmids/one lineage. Our study provides a framework for the much needed incorporation of plasmid data into genomic surveillance systems, an essential step toward a more comprehensive understanding of resistance spread.