Genomic investigation of multispecies and multivariant blaNDM outbreak reveals key role of horizontal plasmid transmission

Objectives: New Delhi metallo-β-lactamases (NDMs) are major contributors to the spread of carbapenem resistance globally. In Australia, NDMs were previously associated with international travel, but from 2019 we noted increasing incidence of NDM-positive clinical isolates. We investigated the clinical and genomic epidemiology of NDM carriage at a tertiary-care Australian hospital from 2016 to 2021. Methods: We identified 49 patients with 84 NDM-carrying isolates in an institutional database, and we collected clinical data from electronic medical record. Short- and long-read whole genome sequencing was performed on all isolates. Completed genome assemblies were used to assess the genetic setting of bla NDM genes and to compare NDM plasmids. Results: Of 49 patients, 38 (78%) were identified in 2019–2021 and only 11 (29%) of 38 reported prior travel, compared with 9 (82%) of 11 in 2016–2018 (P = .037). In patients with NDM infection, the crude 7-day mortality rate was 0% and the 30-day mortality rate was 14% (2 of 14 patients). NDMs were noted in 41 bacterial strains (ie, species and sequence type combinations). Across 13 plasmid groups, 4 NDM variants were detected: bla NDM-1, bla NDM-4, bla NDM-5, and bla NDM-7. We noted a change from a diverse NDM plasmid repertoire in 2016–2018 to the emergence of conserved bla NDM-1 IncN and bla NDM-7 IncX3 epidemic plasmids, with interstrain spread in 2019–2021. These plasmids were noted in 19 (50%) of 38 patients and 35 (51%) of 68 genomes in 2019–2021. Conclusions: Increased NDM case numbers were due to local circulation of 2 epidemic plasmids with extensive interstrain transfer. Our findings underscore the challenges of outbreak detection when horizontal transmission of plasmids is the primary mode of spread.

assembly from our institution for each ST to use as a reference.Mobile genetic elements were excluded from these reference assemblies using PHASTER and IslandViewer 4 [13,14].A core chromosomal single nucleotide variant (SNV) alignment was generated using Snippy v.4.6.0 [15] and recombination was removed using Gubbins v3.3 [16].We then used this core genome alignment in IQtree v.2.0.3 to generate maximum likelihood phylogenies for each ST [17], with the best-fit model chosen using ModelFinder [18].For each ST, median SNV distances between isolates from our institution were then calculated.

Plasmid analyses
Using Abricate, we identified blaNDM-harboring contigs that were putative plasmids in our hybrid assemblies.We then used the MOB-typer v1.4.9 tool to determine plasmid replicons present, as well as to assign clusters [19].We identified possible mosaic plasmids resulting from fusion events by examining plasmid replicon content within MOB-typer cluster and identifying plasmids which had presence of additional plasmid replicons then manually inspecting the assemblies.
We then conducted analyses within all plasmid groups within our dataset with >1 plasmid.In order to identify SNVs in the plasmid backbone, we used Snippy v.4.6.0 [15] to create a core SNV alignment by mapping short reads to a reference plasmid from our institution from each plasmid group.We used progressiveMauve v2.4.0.r4736 to align all plasmids within a plasmid group and assess for structural re-arrangements [20].
We used Flanker v0.1.5[21] to identify and cluster flanking sequences around blaNDM.We performed clustering 5000 bp downstream of the blaNDM gene across windows in 500 bp increments.Clustering was only performed downstream due to prior findings from Acman et al. which showed very high diversity of flanking regions upstream of blaNDM [22].We extracted flanking regions and performed progressiveMauve to align all flanking sequences.Geneious v10.2.6 (https://www.geneious.com)was used to visualize and assess for structural re-arrangements, with subsequent manual annotation in Adobe Illustrator v2020.24.3.We compared key plasmid groups (IncX3, IncN and IncC) from the Alfred Hospital by using a representative plasmid from each group as a reference and conducting BLAST searches and noting the top 5 matches.For the best NDM plasmid matches, we visualized the comparisons between Alfred Hospital and publicly available plasmids in Easyfig v2.2.6 [23].Total no. of patients

Supp. Figure 2 -Comparison of Alfred Hospital and global New Delhi metallo-beta-lactamase plasmids
We conducted comparative analyses with global New Delhi metallo-beta-lactamase (NDM) plasmids that were closely related to NDM epidemic plasmids from our study.The location of blaNDM genes is shown in red, with the genetic context in light blue.Panel A shows that IncX3 plasmids from our study were almost identical to global NDM IncX3 plasmids (100% coverage and 99.96% identity).Panels B and C demonstrate IncN and IncC plasmids, respectively.Plasmids from our study had more significant differences to NDM global plasmids, with IncN plasmids having 95% coverage and 99.93% identity while IncC plasmids had 93% coverage and 99.99% identity.

Table 2 -Pairwise SNV distances of key New Delhi metallo-beta-lactamase bacterial host strains Bacterial strain Total genomes Median pairwise SNV distance Interquartile range Total patients
*Pairwise distance reported as only two genomes available for analyses Abbreviations: LV -locus variant; SNV -single nucleotide variant; ST -sequence type Supp.