Population structure of bla_KPC-harbouring IncN plasmids at a New York City medical centre and evidence for multi-species horizontal transmission

Abstract

Background

Carbapenem-resistant Enterobacterales (CRE) are highly concerning MDR pathogens. Horizontal transfer of broad-host-range IncN plasmids may contribute to the dissemination of the Klebsiella pneumoniae carbapenemase (KPC), spreading carbapenem resistance among unrelated bacteria. However, the population structure and genetic diversity of IncN plasmids has not been fully elucidated.

Objectives

We reconstructed bla_KPC-harbouring IncN plasmid genomes to characterize shared gene content, structural variability, and putative horizontal transfer within and across patients and diverse bacterial clones.

Methods

We performed short- and long-read sequencing and hybrid assembly on 45 CRE isolates with bla_KPC-harbouring IncN plasmids. Eight serial isolates from two patients were included to assess intra-patient plasmid dynamics. Comparative genomic analysis was performed to assess structural and sequence similarity across plasmids. Within IncN sublineages defined by plasmid MLST and kmer-based clustering, phylogenetic analysis was used to identify closely related plasmids.

Results

Comparative analysis of IncN plasmid genomes revealed substantial heterogeneity including large rearrangements in serial patient plasmids and differences in structure and content across plasmid clusters. Within plasmid sublineages, core genome content and resistance gene regions were largely conserved. Closely related plasmids (≤1 SNP) were found in highly diverse isolates, including ten pST6 plasmids found in eight bacterial clones from three different species.

Conclusions

Genomic analysis of bla_KPC-harbouring IncN plasmids revealed the presence of several distinct sublineages as well as substantial host diversity within plasmid clusters suggestive of frequent mobilization. This study reveals complex plasmid dynamics within a single plasmid family, highlighting the challenge of tracking plasmid-mediated transmission of bla_KPC in clinical settings.

Introduction

Carbapenem-resistant Enterobacterales (CRE) are highly concerning MDR pathogens that were identified as an urgent antibiotic threat by CDC in 2019.¹ Included among their concerns was the potential for mobile genetic elements (MGEs) encoding carbapenemases to spread readily between different bacteria. Production of the Klebsiella pneumoniae carbapenemase (KPC) has emerged as a leading mechanism for carbapenem resistance in Enterobacteriaceae worldwide, detected in >50% of CRE isolates in a large global epidemiologic study.²  bla_KPC and other clinically prominent carbapenemases including bla_NDM, bla_VIM, bla_IMP, and bla_OXA-48 are commonly found on plasmids in Enterobacteriaceae, often within transposons or other MGEs, providing multiple potential mechanisms for their dissemination.^3–7 Recent studies have shown that within clinical settings, multiple different bacterial species and bla_KPC-harbouring plasmids may be present, contributing to complex patterns of both clonal and multi-species CRE dissemination.^5,7–9 Advances in next-generation sequencing technology have enabled increasingly comprehensive genomic surveillance of CRE, including detailed characterization of evolution and transmission mechanisms.

Here we aimed to characterize bla_KPC-producing IncN plasmids from clinical CRE isolates, which we and others previously found to be present in CRE isolates belonging to different bacterial species and sequence types (STs), suggestive of horizontal plasmid transfer among unrelated bacteria.^10–15 We took advantage of single molecule, long-range sequencing, which has been shown to efficiently resolve complex and repetitive genetic structures often found on plasmids to enable detailed analysis of plasmid structures.^16–18  De novo assembly of plasmid genomes using a hybrid assembly pipeline enabled comprehensive analysis of plasmid genomes including identification of conserved genes and structural changes in isolates from both serial and unique patient samples. In addition to elucidating their genomic diversity, we sought to further corroborate IncN plasmid transmissibility by assessing plasmid relatedness in diverse bacterial hosts and measuring in vitro conjugation efficiencies.

Materials and methods

Isolate selection

CRE isolates included in this collection were derived from retrospectively collected and catalogued meropenem-resistant clinical cultures (meropenem MIC >2 mg/L) from the clinical microbiology laboratory at a New York City medical centre. As a part of ongoing genomic surveillance efforts,^14,19–21 CRE isolates collected at our institution underwent WGS on an Illumina platform followed by determination of the isolate MLST and antibiotic resistance gene (ARG) carriage using SRST2.^22–24 For the purposes of this study, the first CRE isolate harbouring bla_KPC collected from each patient between 2012–15 was screened for the presence of an IncN plasmid allele using the PlasmidFinder²⁵ database. In an attempt to include consecutive bla_KPC-harbouring IncN plasmids in this study, all available isolates with a detectable IncN rep gene underwent long-range sequencing and hybrid assembly as described below; isolates found to have bla_KPC located on an IncN plasmid were included in the analysis. To investigate within-patient plasmid variability, we also sequenced six additional CRE isolates collected longitudinally from two different patients. From among three K. pneumoniae ST258 isolates with available long-range sequencing data, three isolates with bla_KPC-harbouring IncF plasmids were randomly selected for inclusion in conjugation studies (see below).

Ethics

Study procedures were approved by the Columbia University IRB (protocol #AAAP3617).

Complete plasmid sequencing

Our Illumina sequencing approach has been described previously.²⁰ For long-range sequencing, genomic DNA was manually extracted using the QIAamp DNA Blood Mini Kit (Qiagen) with modifications to reduce shearing. The Rapid Barcoding Kit (RBK004) was used to prepare DNA libraries for nanopore sequencing using the MinION (Oxford Nanopore Technologies). Base-calling and demultiplexing were performed by MinKNOW; reads then underwent additional trimming and filtering using Porechop and Mothur, respectively.^26,27 We generated hybrid assemblies from Illumina and Nanopore reads using Unicycler,²⁸ which uses long reads to correct and bridge short-read assemblies followed by additional polishing with accurate Illumina reads. Resulting assembly graphs were visually reviewed using Bandage,²⁹ including use of integrated BLAST searches to identify nodes co-harbouring plasmid replicon genes and bla_KPC.³⁰ If necessary, isolates were re-sequenced to achieve assemblies with closed, circular bla_KPC-harbouring plasmid sequences without ambiguities. At a minimum, average Illumina read depths of ≥20× for housekeeping genes and ≥10 000 MinION reads with mean lengths of ∼5 kb were targeted to achieve adequate assemblies. Predicted coding regions were annotated using Prokka.³¹ Transposable elements (IS and transposons) were identified and annotated from ISfinder output.³² Genomes were deposited in Genbank (https://www.ncbi.nlm.nih.gov/genbank/) under BioProject number PRJNA759273.

Plasmid clustering and comparative sequence analysis

To characterize differences across bla_KPC-harbouring plasmid genomes, sequences were first aligned and visually compared using ProgressiveMauve³³ in Geneious (Biomatters). For plasmids derived from same-patient cultures, we also compared percentage identity across collinear regions using MUMmer4,³⁴ where plasmid genomes were mapped against the earliest available isolate. IncN plasmids from both serial and different patient isolates were genotyped using plasmid MLST typing (pMLST)³⁵ and subdivided into similarity groups using MOB-cluster from the MOB-suite.³⁶ Within each MOB-cluster, SNPs were identified and aligned using Snippy,³⁷ where the earliest available plasmid was used as the reference sequence. Maximum likelihood trees were generated from concatenated SNPs using RAxML v8.2.11 with GTR-GAMMA-based likelihood estimation³⁸ and rooted to the reference plasmid. Support for nodes was assessed using 100 rapid bootstrap inferences and thereafter by searching for best-scoring maximum-likelihood trees. After assessing genetic differences across same-patient plasmids, the presence of ≤1 SNP in shared gene alignments was used to define relatedness in plasmids derived from different patients.

Conjugation analysis

We performed conjugation assays to compare conjugation efficiencies among IncN plasmids from the same patient, among IncN plasmids belonging to different plasmid STs (pSTs), and to clinical bla_KPC-harbouring IncF plasmids. IncF plasmids were selected for this analysis because of their frequent association with bla_KPC carriage, in particular in isolates belonging to the epidemic clone K. pneumoniae ST258.³⁹ IncN and IncF-carrying CRE donors and rifampicin-resistant Escherichia coli recipients were streaked onto LB agar plates with antibiotic selection (carbenicillin and rifampicin, respectively) and grown overnight at 37°C; single colonies were then used to inoculate 2 mL of LB broth and placed in a shaking incubator for 16 h at 37°C. Cells were collected and resuspended in M9 media (M9CA with 0.4% glucose), combined in a 1:1 mixture, and incubated at 25°C for 1 h. To quantify transconjugants, serial 10-fold dilutions of the cell mixture was plated onto LB agar plates with dual carbenicillin and rifampicin selection. Parent densities were also determined by plating serial 10-fold dilutions of liquid culture onto respective single selective media. Conjugation efficiencies were calculated by dividing the number of transconjugants [colony-forming units (cfu)/mL] by the number of recipients (cfu/mL) times the number of donors (cfu/mL). All experiments were performed in quadruplicate and mean values were compared using one-way ANOVA for individual plasmid means and t-tests for group comparisons in GraphPad Prism v9.0.2.

Results

Characteristics of IncN plasmids and host bacteria

Among 73 unique patient CRE isolates with a detectable IncN allele, 39 isolates were found to have bla_KPC encoded by an IncN plasmid and were included in the analysis (Table 1). This represents 14% of 279 patients with bla_KPC-positive CRE clinical isolates collected between 2012–15 that were available for Illumina sequencing. We also identified and included six additional isolates from two of these patients with serial isolates (Patients A and B; Table 2). This collection of 45 CRE isolates was highly diverse, comprising 25 different sequence types (STs) from five different bacterial species [K. pneumoniae (n = 30), Enterobacter cloacae complex (n = 8), E. coli (n = 4), Klebsiella oxytoca (n = 2), and Enterobacter asburiae (n = 1)] collected from a variety of culture sites (Tables 1 and 2). Both patients with serial isolates had IncN plasmids identified in at least two different bacterial species. Only six STs included two or more isolates, K. pneumoniae ST17 (n = 4), E. cloacae ST78 (n = 2), E. cloacae ST133 (n = 2), K. pneumoniae ST258 (n = 13), E. coli ST617 (n = 2), and K. pneumoniae ST1471 (n = 4, all isolated from the same patient).

Table 1.

bla _KPC-harbouring IncN plasmid characteristics

Isolate number	Year	Organism	Culture type	MLST	rep gene allele	bla KPC subtype	Tn4401 subtype	pMLST	MOB-clustera
pKP0011	2012	K. pneumoniae	Blood	1224	IncN1	KPC-2	Tn4401b	6	1
pKP0028	2012	K. pneumoniae	Blood	17	IncN1	KPC-2	Tn4401a	9	5
pNR5436	2012	K. pneumoniae	Respiratory	258	IncN1/R	KPC-2	Tn4401a	9	4
pKP0063	2012	K. pneumoniae	Blood	36	IncN1	KPC-3	Tn4401bb	6	1
pKP0925	2012	E. cloacae	Blood	93	IncN1	KPC-3	Tn4401b	6	1
pKP0177	2012	K. pneumoniae	Blood	258	IncN1/R	KPC-2	Tn4401a	9	4
pNR5451	2012	K. pneumoniae	Abdominal fluid	258	IncN1/R	KPC-2	Tn4401a	9-like	6
pKP0222	2012	K. pneumoniae	Blood	1495	IncN1/R	KPC-2	Tn4401b	6	S
pKP0247	2013	K. pneumoniae	Abdominal fluid	258	IncN1	KPC-2	Tn4401a	9	5
pKP0265	2013	K. pneumoniae	Blood	17	IncN1	KPC-2	Tn4401a	9	5
pNR5466	2013	K. pneumoniae	Respiratory	258	IncN1/R	KPC-2	Tn4401a	9-like	6
pKP0713	2013	K. pneumoniae	Blood	258	IncN1/R	KPC-2	Tn4401a	9-like	S
pNR3086	2013	E. cloacae	Urine	78	IncN1	KPC-3	Tn4401b	6	1
pNR0242	2014	E. cloacae	Respiratory	269	IncN1	KPC-3	Tn4401b	6	1
pNR0276	2014	E. cloacae	Respiratory	78	IncN1	KPC-3	Tn4401bb	6	1
pNR0421	2014	E. coli	Urine	617	IncN1	KPC-2	Tn4401bc	6	1
pNR0411	2014	E. coli	Urine	617	IncN1	KPC-2	Tn4401bc	6	1
pNR0694	2014	K. pneumoniae	Urine	2054	IncN1	KPC-3	Tn4401bb	6	1
pNR0704	2014	K. pneumoniae	Rectal swab	841	IncN1	KPC-3	Tn4401b	6	1
pNR0791	2014	E. cloacae	Wound	252	IncN1	KPC-3	Tn4401b	7	2
pNR0852	2014	E. cloacae	Wound	133	IncN1	KPC-3	Tn4401b	6	1
pNR0976	2014	K. pneumoniae	Respiratory	3375	IncN1	KPC-3	Tn4401b	6	1
pNR0970	2014	K. pneumoniae	Respiratory	307	IncN1	KPC-3	Tn4401b	7	2
pNR1202	2015	K. pneumoniae	Urine	258	IncFIB(pQil)/FII/N1	KPC-3	Tn4401a	9	3
pNR1246	2015	K. pneumoniae	Rectal swab	17	IncN1	KPC-3	Tn4401b	7	2
pNR1295	2015	K. oxytoca	Urine	363	IncFIB(pQil)/N2	KPC-3	Tn4401bc	NA	S
pNR1597	2015	K. pneumoniae	Urine	17	IncN1	KPC-3	Tn4401b	7	2
pNR1633	2015	K. pneumoniae	Urine	340	IncFIB/FII/N1	KPC-3	Tn4401a	9	S
pNR1686	2015	K. pneumoniae	Respiratory	258	IncN1	KPC-3	Tn4401a	9	5
pKP1676	2015	K. pneumoniae	Blood	258	IncN1	KPC-3	Tn4401a	9	5
pNR1932	2015	K. pneumoniae	Urine	42	IncFII/N1	KPC-2	Tn4401b	9	3
pNR2074	2015	K. pneumoniae	Urine	258	IncN1	KPC-3	Tn4401a	9	5
pNR2064	2015	E. cloacae	Respiratory	133	IncN1	KPC-3	Tn4401b	6	1
pNR2144	2015	K. oxytoca	Urine	138	IncN1	KPC-3	Tn4401b	6	1
pNR2679	2015	K. pneumoniae	Respiratory	111	IncN1	KPC-3	Tn4401b	6	1
pNR2956	2015	E. coli	Respiratory	3168	IncN1	KPC-2	Tn4401a	9-like	S
pNR3568	2015	E. coli	Wound	4110	IncN1	KPC-3	Tn4401b	6	1
pNR3537	2015	K. pneumoniae	Respiratory	76	IncA/C/N1	KPC-3	Tn4401b	6	S
pNR4047	2015	K. pneumoniae	Urine	258	IncN1	KPC-2	Tn4401a	9	5

Abbreviations: MLST, multilocus sequence type; kb, kilobases; pMLST, plasmid multilocus sequence type; S, singleton.

^aRefers to plasmid similarity groups determined using MOB-cluster.³⁶

^b210 bp deletion in tnpA.

^cTwo copies of Tn4401b present.

Table 2.

IncN plasmids from serial patient isolates

Isolate number	Patient	Culture site	Collection month/year	Organism	MLST	rep gene type	bla KPC subtype	Tn4401 subtype	Plasmid assembly length (kb)
pNR2064	A	Respiratory	May 2015	E. cloacae	133	IncN1	KPC-3	Tn4401b	70.8
pNR2296	A	Respiratory	June 2015	K. pneumoniae	1471	IncN1	KPC-3	Tn4401b	71.1
pNR2767	A	Pleural fluid	June 2015	K. pneumoniae	1471	IncN1	KPC-3	Tn4401b	72.2
pKP1700	A	Blood	June 2015	K. pneumoniae	1471	IncN1	KPC-3	Tn4401b	60.9
pNR4045	A	Respiratory	November 2015	K. pneumoniae	1471	IncN1	KPC-3	Tn4401b	72.3
pNR1246	B	Rectal swaba	January 2015	K. pneumoniae	17	IncN1	KPC-3	Tn4401b	60.6
pNR1247	B	Rectal swaba	January 2015	E. cloacae	454	IncN1	KPC-3	Tn4401b	60.5
pNR1280	B	Respiratory	January 2015	E. asburiae	252	IncN1	KPC-3	Tn4401b	59.7

Abbreviations: MLST, multilocus sequence type; kb, kilobases.

^aHost bacteria were isolated from the same culture.

All 45 isolates underwent short- and long-read sequencing and de novo hybrid assembly. The resulting assemblies were highly contiguous and consisted of 2–26 contigs (mean 7.5, SD 4.8). All IncN plasmids were successfully reconstructed into single, circular contigs and had no structural ambiguities seen on assembly graphs. As described further below, they varied substantially in length (54.1–222.1 kb) and structure; 10/45 plasmids harboured additional rep genes identified as IncR (n = 6/10), IncF (n = 3/10), or IncA/C (n = 1/10) replicons. Most plasmids harboured bla_KPC-3 (n = 30/45, 63%), which was primarily associated with Tn4401b, although bla_KPC-2 and other Tn4401 subtypes were also detected (Tables 1 and 2). IncN plasmids from two E. coli ST617 isolates encoded two complete copies of Tn4401b. In addition to the IncN plasmid, most isolates harboured between one and seven additional plasmids from diverse plasmid families, including IncA/C, IncF, IncHI2/HI2A, IncL/M, IncR, IncY, and Col.

To confirm the reproducibility of our sequencing and assembly approach, one randomly selected isolate from this collection (KP0925) underwent repeat long-read sequencing. De novo assemblies of KP0925 sequences from independent runs showed similar chromosome length (4 829 373 bp versus 4 829 368) and had no structural differences. KP0925 IncN plasmids also had identical length (69 239 bp) and structure, and differed only by one SNP in an intergenic region, supporting high reproducibility of hybrid assemblies.

IncN population structure and shared genome content

To better understand IncN plasmid and host diversity, we analysed the 39 bla_KPC-harbouring IncN plasmid genomes found in CRE isolates collected from different patients (for patients with multiple plasmids, only the first plasmid was considered; see Table 1). Almost all plasmids (38/39) harboured the AY046276 repA allele consistent with the IncN1 subgroup whereas one plasmid belonged to the IncN2 subgroup (JF785549 allele).²⁵ Further genotyping of IncN1 plasmids using the pMLST scheme further subdivided them into three main plasmid sequence types (pSTs), pST6 (n = 18), pST7 (n = 4), and pST9 (n = 12) (Table 1). Four additional plasmids lacked at least one complete pMLST allele but were most similar to ST9 plasmids. The IncN2 plasmid backbone lacks sequence homology with IncN1 plasmids and cannot be typed using pMLST.

To define and compare shared regions across groups of IncN1 plasmids, we identified genes common to all plasmid sequences using comparative sequence analysis (Figure 1a). In addition to repE, almost all IncN1 plasmids had an intact tra gene locus, which encodes the conjugal transfer system.⁴⁰ Additional genes shared by IncN1 plasmids are thought to be involved in plasmid stability and included genes encoding the anti-restriction proteins ardA and ardB;⁴¹ the stbA, stbB, and stbC genes thought to promote efficient partitioning;⁴² the umuC and umuD genes encoding subunits of the error-prone DNA polymerase V;⁴³ a recD2-like helicase which may participate in double-stranded DNA break repair and homologous recombination;⁴⁴ and dcm and ecoRII, genes encoding a DNA methyltransferase and restriction enzyme comprising a type II restriction-modification system (Figure 1a).⁴⁵ For most plasmids, these genes were subdivided into two main regions largely comprising the plasmid backbone, although overall plasmid structure varied across pST types. Notably, the tra gene locus was absent in 3/6 IncN1/R plasmids, one of which (pKP0713) also lacked many other backbone genes, including ardA and ardB, stbABC, and recD2.

Figure 1.

Comparative genomic analysis of IncN plasmids to identify related clusters. (a) In comparing the annotated genome of the first plasmid to be collected belonging to each pMLST in this collection, differences in overall plasmid structure were noted, although core plasmid regions were highly conserved. (b) pST6 plasmids shared a large transposon-rich region harbouring multiple ARGs conferring resistance to multiple drug classes, shown in detail here for representative plasmid pNR2064 (outlined in light orange). This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

In IncN plasmids belonging to all pST types, conserved loci were interspersed with variable transposon-rich regions such as the ARG-encoding region including Tn4401. Overall, the number of ARGs varied across pST, including differences in the number of additional β-lactamase genes and genes conferring resistance to other drug classes. pST6 plasmids had the highest number of ARGs, with most clustered in an approximately 12 kb region immediately downstream of Tn4401 encompassing a portion of the Tn1331 transposon, in which Tn4401 was embedded,⁴⁶ and Tn5393 (Figure 2b). Both pST6 and pST7 plasmids encoded the trimethoprim resistance gene dfr; pST7 and pST9 plasmids lacked additional ARGs.

Figure 2.

Comparative alignment of patient A and B IncN plasmids. Five IncN plasmids ranging in length from 60.9–72.3 kb were isolated from serial bla_KPC-producing Enterobacterales isolates collected from patient A. (a) A linear alignment of four patient A IncN plasmid sequences compared with the initial available plasmid, pNR2064, was generated using MUMmer4 to highlight structural differences. Aligned forward sequences are shown in pink and inversions in blue. Resulting plots demonstrated large inversions (shown in blue) in pKP1700 and pNR4045. (b) In contrast, three IncN plasmids from patient B were similar in size (59.7–60.6 kb) and demonstrated similar structure. Serial plasmids from both patients were found in more than one bacterial species. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

IncN plasmids from serial patient isolates

To investigate within-patient plasmid variability, we characterized serial bla_KPC-harbouring IncN plasmids isolated from two different patients. IncN plasmids were identified in five clinical CRE isolates that were collected from Patient A over the course of 6 months and belonged to two different bacterial species (K. pneumoniae, n = 4; E. cloacae, n = 1; Table 2). These five plasmids had highly similar sequences over aligned regions spanning most of the plasmid genomes (99.4%–99.5% identity and 99.5%–100% coverage compared with the first available isolate, pNR2064). Very few SNPs were observed within protein coding regions including none in genes involved in plasmid stability and conjugation (see above). However, alignment of plasmid genomes revealed deletions as well as large structural rearrangements (Figure 2a) resulting in differences in plasmid sizes, which ranged from 60.9–72.3 kb. Notably, large inversions in pKP1700 and pNR4045 were flanked by transposon-rich regions. Patient B had IncN plasmids in three isolates belonging to three different bacterial species (K. pneumoniae, E. cloacae, and E. asburiae; Table 2). Two of these plasmids were derived from the same rectal swab culture, and the third isolate was collected 11 days later. These plasmids were similar in size (59.7–60.6 kb) and had high pairwise identity (98.4%–99.8% identity, 100% coverage compared with pNR1246). One plasmid (pNR1280) had a 411 bp deletion in an intergenic region as well as several smaller deletions in a hypothetical protein-coding region immediately downstream of repE. Alignment of plasmid genomes demonstrated similar structure with no large genomic rearrangements (Figure 2b). Taken together, these results support the potential for horizontal transmission to occur within single patients leading to multispecies transmission of bla_KPC. We also found that large rearrangements leading to structural changes in plasmid genomes may also occur in this setting.

Phylogenetic analysis of IncN plasmid clusters

To better assess plasmid relatedness, we further subdivided IncN1 plasmids into similarity groups using MOB-cluster and performed phylogenetic analysis within each group (here referred to as MOB-clusters, see Table 1). Serial patient plasmids were included in phylogenetic analyses to corroborate our comparative sequence analysis and define criteria for plasmid relatedness. Overall, we identified six different MOB-clusters largely corresponding to pST types. Most pST6 and pST7 plasmids each formed a single cluster while pST9 and pST9-like plasmids comprised four clusters including two or more plasmids. Plasmids from all six MOB-clusters were found in multiple different clonal backgrounds. Five additional plasmids, including three multireplicon plasmids and the IncN2 plasmid pNR1295, did not cluster with other plasmids and were termed singletons.

The largest MOB-cluster (MOB-cluster 1) consisted of 20/22 pST6 plasmids including five patient A plasmids (Figure 3a). MOB-cluster 1 plasmids were highly similar, differing by 0–6 core genome SNPs, and may have recently diverged from a shared ancestor. Many of these SNPs occurred in non-coding regions or hypothetical proteins (7/13, 54%); within plasmid backbone genes, SNPs were limited to 1–2 plasmids (ecoRII, n = 1; dcm, n = 2; ardB, n = 1). Interestingly, three MOB-cluster 1 plasmids had an SNP detected in the bla_KPC gene conferring a change in bla_KPC subtype from bla_KPC-3 to bla_KPC-2. Conversely, host bacteria were highly diverse, comprising 14 different STs and four bacterial species. Phylogenetic analysis of MOB-cluster 1 plasmid genomes revealed the presence one dominant sublineage of 13 closely related plasmids, including all serial Patient A plasmids (≤1 SNP; Figure 3a), which likely belonged to the same transmission network. These plasmids were found in three different bacterial species and eight STs, supporting high transmissibility among unrelated bacteria. Interestingly, BLAST analysis of MOB-cluster 1 plasmid genomes against the NCBI non-redundant nucleotide database showed the best hit as a bla_KPC-3-harbouring IncN plasmid from K. pneumoniae S12 collected at the same New York City medical centre in 2003 (GenBank accession FJ223605.1).⁴⁷ Alignment of the first available MOB-cluster 1 plasmid (pKP0011) with the S12 plasmid demonstrated similar plasmid structure and high sequence identity (Figure 3b). MOB-cluster 1 also consisted of two additional pairs of closely related plasmids (Figure 3b). One pair was composed of plasmids from two E. coli ST617 isolates, suggesting clonal spread. The second pair were derived from unrelated hosts, potentially reflecting a second horizontal transfer event.

Figure 3.

Phylogenetic analysis of Group 1 IncN plasmids and comparison with a historical plasmid. (a) Phylogenetic analysis of largest cluster of plasmids (MOB-Cluster 1) within pST6 showed a dominant group of closely related plasmids derived from unrelated bacteria, corroborating transmission of the IncN plasmid among multiple unrelated bacterial clones. (b) Comparison of a MOB-Cluster 1 plasmid (pKP0011) with a plasmid derived from K. pneumoniae S12 (here designated pS12), a bla_KPC-harbouring plasmid collected at the same New York City medical centre in 2003 (GenBank accession FJ223605.1), demonstrated similar structure and high sequence identity. Differences in plasmid length were primarily due to insertion of an arsenic resistance cassette and associated transposases in the tra gene locus of pS12 and other smaller insertions (denoted by grey triangles). Interestingly, an SNP in the bla_KPC gene conferred differences in subtype; however, there was otherwise no evidence that Tn4401 underwent independent targeted insertions into this region. This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

All pST7 plasmids belonged to the same cluster (MOB-cluster 2, n = 6), which differed by 0–15 SNPs. Phylogenetic analysis revealed one branch of closely related plasmids (≤1 SNP; n = 4) consisting of Patient B IncN plasmids and one additional plasmid found in the same clonal background as a Patient B plasmid. Conversely, pST9 and pST9-like plasmid consisted of four different MOB-clusters. Only one cluster of pST9 plasmids (MOB-cluster 5) included two or more plasmids (n = 7). Overall, MOB-cluster 5 plasmid genomes differed by 0–25 SNPs. Within MOB-cluster 5, we identified two sublineages each consisting of three plasmids differing by ≤1 SNP. One of these branches was limited to K. pneumoniae ST258 and may represent clonal spread, while the other was derived from two K. pneumoniae ST258 isolates and one K. pneumoniae ST17 isolate. Taken together, our findings supported the potential for IncN plasmids to spread by horizontal transfer within and between patients, although this seemed to vary across IncN clusters and subsequent clonal spread seemed to be limited.

IncN plasmid conjugation efficiencies

To further assess plasmid transmissibility, we determined conjugative transfer efficiencies of ten different IncN plasmids including five plasmids from patient A and other representative plasmids from pST6 (including an additional plasmid from MOB-cluster 1), pST7, and pST9 (Figure 4a). Mean conjugation efficiencies ranged from 1.79 × 10⁻¹⁴ (SD 1.31 × 10⁻¹⁴) to 8.61 × 10⁻¹³ (SD 1.08 × 10⁻¹²) for 9/10 isolates. One patient A IncN plasmid (pKP1700) demonstrated substantially lower transferability (mean conjugation efficiency 1.32 × 10⁻¹⁷, SD 9.33 × 10⁻¹⁸). Analysis of the pKP1700 plasmid genome revealed the presence of a large genetic rearrangement leading to deletion of multiple genes from the tra gene locus required for conjugation (Figure 4b). While other large rearrangements were seen in other patient A-derived plasmids, e.g. NR4045, this did not result in loss of genes in this region. Conjugation efficiencies were also compared with those of three randomly selected IncF plasmids from clinical K. pneumoniae ST258 isolates harbouring bla_KPC. These IncF plasmids had different genotypes based on rep gene typing as well as demonstrating differences in size and genetic structure [NR5461, IncFIB(K)/FII(K), 229.561 kb; NR5462, IncFIB(pQil)/FII(K), 113.640 kb; NR5486, IncFII(pBK30683), 174.666 kb]; however all were predicted to be conjugative by MOB-typer³⁶ and appeared to have intact tra gene loci. Compared with IncN plasmids with the exception of pKP1700, IncF plasmids demonstrated lower conjugation efficiencies (Figure 4a). Overall, a statistically significant difference in mean conjugation efficiency was detected across all plasmids using one-way ANOVA (F = 3.24, DF = 12, P = 0.003); after adjusting for multiple comparisons only the individual mean differences in conjugation efficiency between the MOB-cluster 1 IncN plasmid pNR0276 and pKP1700 as well as between pNR0276 and the IncF plasmids pNR5461, pNR5462, and pNR5486 were statistically significant (P = 0.04 for all four comparisons). Using a t-test, the mean difference in conjugation efficiency between IncN and IncF plasmids was also statistically significant (2.92 × 10⁻¹³ versus 4.05 × 10⁻¹⁶, P = 0.02). These studies support the potential for highly efficient plasmid transfer among IncN plasmids in this collection with intact transfer gene regions.

Figure 4.

In vitro conjugation of IncN and IncF plasmids into an E. coli recipient. (a) Conjugation efficiencies of bla_KPC-harbouring plasmids into a rifampicin-resistant E. coli recipient were compared for ten IncN plasmids, including five plasmids derived from patient A isolates and representative plasmids from other pSTs, as well as three IncF plasmids from K. pneumoniae ST258 isolates. Mean conjugation efficiencies for IncN plasmids were statistically significantly higher than those of IncF plasmids (2.93 × 10⁻¹³ versus 4.05 × 10⁻¹⁶, P = 0.02; statistical significance is denoted with an asterisk). The mean conjugation efficiency was notably diminished for pKP1700. Comparative analysis of pST6 plasmid genomes (b) revealed that pKP1700 had a large structural rearrangement resulting in the deletion of multiple tra genes, which are required for conjugation, as shown in a Mauve alignment. Inversion of a large region of the plasmid genome (shown in green) resulted in an approximately 11 kb deletion in pKP1700 resulting in loss of most of the tra gene locus (shown in yellow; the junction at which the deletion occurred is indicated by a red arrow). This figure appears in colour in the online version of JAC and in black and white in the print version of JAC.

Discussion

Taking advantage of long-range sequencing, we were able to comprehensively characterize plasmid genomes from clinical CRE isolates including consecutive plasmids from the broad-host-range IncN plasmid family in order to assess their genomic contents and diversity. Overall, IncN plasmids belonged to six different MOB-clusters distributed among three pST types. Substantial differences in genome organization and content were seen across plasmid sublineages, while a set of core genes appeared to be highly stable across all plasmids. Moreover, IncN plasmid genomes demonstrated the potential for structural heterogeneity, including large genetic rearrangements affecting conserved plasmid regions and transmissibility. Within plasmid clusters, phylogenetic analysis provided evidence for plasmid transmission within single patients as well as across highly diverse Enterobacterales. Taken together, our findings support a need to further investigate multispecies spread of bla_KPC and other ARGs while highlighting the challenge of determining plasmid relatedness given their genetic diversity and plasticity.

In several recent genomic studies, IncN plasmids were identified as important drivers for multispecies transmission of bla_KPC, with evidence for transmission found between bacteria co-colonizing individual patients and environmental surfaces.^10–13,15 While the most common host species was K. pneumoniae, they also appear be important contributors to the spread of bla_KPC among diverse carbapenem-resistant E. cloacae.¹⁴ This included its acquisition by the widespread ESBL clone ST78, with evidence for subsequent clonal spread. However, unlike the relationship between K. pneumoniae ST258 and bla_KPC-harbouring IncF plasmids,⁴⁸ IncN plasmids have not been consistently associated with specific CRE clones in prior studies. Here we found that most IncN bacterial hosts did not belong to established MDR clones, with E. cloacae ST78 and K. pneumoniae ST258 and ST17 each representing a small proportion of overall bacterial isolates, suggesting limited ongoing propagation of the bla_KPC-harbouring IncN plasmid in these lineages. Remarkably, plasmids from MOB-cluster 1, which were identified among multiple unrelated bacteria, shared high sequence identity with an IncN plasmid first isolated and characterized at our institution in 2003.⁴⁷ These findings suggests that these plasmids may be able to persist without being closely associated with a successful clonal lineage.

In the subset of IncN plasmids tested here and found to have an intact tra gene locus, in vitro conjugation frequencies were relatively high in IncN plasmids, supporting efficient transmission. Although host bacteria may have a variety of mechanisms that allow them to resist uptake of plasmids, we also identified several genes present in all or almost all IncN genomes including across different sublineages that may further facilitate their dissemination, such ardA and ardB encoding anti-restriction proteins and genes that may encode a type II restriction-modification system. Finally, we hypothesize that plasmid plasticity including structural changes seen in some plasmids may contribute to rapid adaptation to different hosts. In IncN plasmids, we identified ubiquitous genes such as umuC and umuD and recD, which may play a role in promoting stress-related mutagenesis and contribute to double-stranded DNA break repair and homologous recombination, respectively. Additional functional studies are needed to further explore the relationship between these genes and plasmid uptake and adaption to different bacterial hosts.

Our study adds to recent literature demonstrating substantial diversity of bla_KPC genetic backgrounds using long-range sequencing. While plasmid typing approaches may be used for tracking of other ARGs with more consistent genetic backbones,^7,49 we and others found hybrid assembly essential for correctly assigning bla_KPC to the correct plasmid backbone while maintaining high fidelity.^8,9 Here we were able to not only fully characterize bla_KPC genetic context but also assess plasmid relationships. By first assessing intra-patient plasmid dynamics, we were able to determine the amount of genetic diversity that would be expected to accumulate as a result of transmission between different bacterial hosts and culture sites and over time, better informing our assessments of inter-patient plasmid relatedness. Using the SNP distance between plasmids from the same patient as a conservative threshold cutoff for defining plasmid relatedness, we felt we were able to identify plasmids belonging to the same MOB group that were most likely to belong to the same transmission network, although further epidemiological analysis would be needed to confirm recent transmission. Many additional gaps in knowledge regarding mechanisms of plasmid evolution and transmission remain. Our study was limited to a collection of plasmids belonging to a single family and resistance gene profile with very limited comparison with other plasmid types. While we found evidence for both long-term stability of specific plasmid backgrounds and short-term potential for rapid plasmid evolution, the dynamics of plasmid genome plasticity and stability in nosocomial settings and with respect to different host populations remain largely unknown.

In summary, in this collection of CRE isolates from a single hospital, we found evidence for both substantial heterogeneity among bla_KPC-harbouring IncN plasmid backbones as well as potential for high transmissibility among diverse bacterial hosts. While this genetic plasticity adds substantial complexity to the detection of plasmid-mediated spread of ARGs, next-generation sequencing technologies are increasingly enabling comparative plasmid genomic studies. Novel tools will be needed to streamline assessments of plasmid transmission in hospitals leading to the spread of bla_KPC and other ARGs among unrelated bacteria.

Funding

This study was supported by the National Institute of Allergy and Infectious Diseases at the National Institutes of Health (RO1 AI116939 to A.C.U. and K23 AI137316 to A.G.S.), the National Institute of General Medical Sciences (R15 1R15GM143694-01 to A.J.L.), and the National Sciences Foundation (2040697 to A.J.L.). The funders of the study had no role in the study design, data collection, data analysis, data interpretation, or writing of the report.

Transparency declarations

A.C.U. has received research funding from Merck unrelated to the current study. All other authors have none to declare.