Genomic epidemiology and transmission dynamics of plasmids carrying New Delhi metallo-β-lactamase (blaNDM) at a single hospital system over five years

Background: Conjugative plasmids encoding New Delhi metallo-beta-lactamase (blaNDM) pose a threat for the spread of carbapenem resistance among healthcare acquired pathogens. Plasmid-associated outbreaks of blaNDM-producing bacteria can involve multiple bacterial species and persist over long time periods, making their detection and control difficult. We systematically studied the genomic epidemiology of blaNDM-encoding plasmids detected within a single hospital system over a five-year period. Methods: blaNDM-producing isolates were collected from clinical cultures as part of the Enhanced Detection System for Healthcare-Associated Transmission (EDS-HAT) genomic sequencing active surveillance program, or during infection prevention and control (IP&C) investigations. Isolates were identified as blaNDM producers by polymerase chain reaction (PCR); the presence of plasmid-encoded blaNDM genes was confirmed by sequencing on both Illumina and Oxford Nanopore platforms. Plasmids were clustered using Pling and bacterial relatedness of host isolates was evaluated with split kmer analysis. Electronic health record data were used to identify shared unit-level spatiotemporal exposures and epidemiologic links within both plasmid and host clusters. Results: We identified 61 blaNDM-producing isolates collected from 54 patients sampled between November 2020 and July 2025. Isolates belonged to 15 Enterobacterales species; Enterobacter hormaechei was the most frequently sampled species (n=23, 37%), and blaNDM-5 was the most frequently observed blaNDM allele (n=36, 59%). We observed six clusters of genetically similar blaNDM-encoding plasmids each containing 2-28 isolates, and eight singleton plasmids. The two largest plasmid clusters consisted of a highly conserved 46 kb IncX3 family blaNDM-5-encoding plasmid (n=28 plasmids, 9 species) and a more variable 98-201 kb IncC family blaNDM-1-encoding plasmid (n=12 plasmids, 6 species). Epidemiologic investigation paired with whole genome sequencing identified spatiotemporal associations between shared patient exposures and putative plasmid and bacterial transmission clusters, suggesting that unit-level exposures contribute to plasmid dissemination. Finally, analysis of publicly available sequences showed that the most prevalent plasmids detected, IncX3(blaNDM-5) and IncC(blaNDM-1), also demonstrated high global prevalence. Conclusions: This study demonstrates the diversity of blaNDM carrying plasmids within a single hospital system and their capacity to cause prolonged, multispecies outbreaks. Integrating whole genome sequencing with epidemiologic data identified unit-level spatiotemporal overlap as a likely contributor to plasmid dissemination in the hospital.