Plasmid diversity among genetically related Klebsiella pneumoniae blaKPC-2 and blaKPC-3 isolates collected in the Dutch national surveillance

Carbapenemase-producing Klebsiella pneumoniae emerged as a nosocomial pathogen causing morbidity and mortality in patients. For infection prevention it is important to track the spread of K. pneumoniae and its plasmids between patients. Therefore, the major aim was to recapitulate the contents and diversity of the plasmids of genetically related K. pneumoniae strains harboring the beta-lactamase gene blaKPC-2 or blaKPC-3 to determine their dissemination in the Netherlands and the former Dutch Caribbean islands from 2014 to 2019. Next-generation sequencing was combined with long-read third-generation sequencing to reconstruct 22 plasmids. wgMLST revealed five genetic clusters comprised of K. pneumoniae blaKPC-2 isolates and four clusters consisted of blaKPC-3 isolates. KpnCluster-019 blaKPC-2 isolates were found both in the Netherlands and the Caribbean islands, while blaKPC-3 cluster isolates only in the Netherlands. Each K. pneumoniae blaKPC-2 or blaKPC-3 cluster was characterized by a distinct resistome and plasmidome. However, the large and medium plasmids contained a variety of antibiotic resistance genes, conjugation machinery, cation transport systems, transposons, toxin/antitoxins, insertion sequences and prophage-related elements. The small plasmids carried genes implicated in virulence. Thus, implementing long-read plasmid sequencing analysis for K. pneumoniae surveillance provided important insights in the transmission of a KpnCluster-019 blaKPC-2 strain between the Netherlands and the Caribbean.

. Distribution of K. pneumoniae bla KPC-2 and bla  isolates and resistance to meropenem. Based on the clinical breakpoints according to EUCAST, the isolates were classified as sensitive (S; < 2 mg/L), intermediate (I; ≥ 2 to 8 mg/L) and resistant (R; > 8 mg/L).
bla KPC Figure 1. Minimum spanning tree based on wgMLST of 478 sequenced K. pneumoniae isolates. Circles represent K. pneumoniae isolates, and the sizes of the circles indicate the numbers of isolates. Lines connecting the circles represent the genetic distance in numbers of alleles; the longer the connecting line, the larger the genetic distance. K. pneumoniae bla KPC-2 isolates were marked blue and K. pneumoniae bla KPC-3 were marked magenta. K. pneumoniae bla  or bla KPC-3 cluster isolates that were sequenced with TGS were marked green. Genetic clusters were indicated with either a blue or a magenta halo around the circles, if two or more isolates differ ≤ 20 alleles. A categorical coefficient was used for the clustering. Cluster names are indicated. Inset: genetic distance between the KpnClusters in which the allelic difference is indicated by numbers.
Antibiotic resistance genes among the genomic elements of the distinct genetic clusters. Long-read sequencing of seven isolates from six of the nine genetic K. pneumoniae bla KPC clusters with ≥ 3 isolates per cluster, revealed 22 plasmids with varying sizes (Fig. 3). Plasmids containing either the bla KPC-2 or bla KPC-3 allele were diverse in size. The large (≥ 150-250 kb) and medium (≥ 50-150 kb) sized plasmids contained one or two replicons from the incompatibility group IncFIB(K) and IncFII(K), IncHI2 and IncHI2a, or IncFIB(pQil) (Fig. 3). The small plasmids (< 50 kb) contained ColRNAI or IncX3/IncL/IncP6 type of replicons. The chromosomes of the analyzed isolates contained on average five acquired AMR genes, while the plasmids contained on average nine AMR genes. Sixteen of the 22 plasmids contained AMR genes from various classes and five plasmids from the isolate of KpnCluster-021 did not. The AMR genes conferring resistance to phenicol, trimethoprim and macrolide antibiotics were located only on medium or large sized plasmids. The small plasmids had one or two AMR genes conferring resistance to aminoglycosides or beta-lactams. Resistance genes for fosfomycin (fosA) and fluoroquinolones (oqxA and oqxB) were exclusively located on the chromosomes of the ARR-3  blaCTX-M-15   blaKPC-2  blaKPC-3  blaLEN12  blaOKP-A-5  blaOXA-1  blaOXA-9  blaSHV-101  blaSHV-11  blaSHV-182  blaSHV-26  blaSHV-28  blaTEM-1A  blaTEM-1B   catA1  catB3  dfrA12 dfrA14  Figure 2. Resistome of K. pneumoniae bla KPC-2 and bla KPC-3 cluster isolates. K. pneumoniae bla KPC-2 and bla KPC-3 cluster isolates were indicated on the y-axis and AMR genes on the x-axis. All isolates analysed contained the fosA, oqxA and oqxB AMR genes and were not included in this figure. The clustering was based on the presence (squares) and absence of AMR genes. Antibiotic classes are indicated above the AMR genes in different colors. Resistance genes in K. pneumoniae bla KPC-2 or bla KPC-3 cluster isolates that were sequenced with TGS were marked with green squares. Genetic relatedness was depicted in an UPGMA tree in which K. pneumoniae bla KPC-2 isolates were marked with blue branches, and K. pneumoniae bla KPC-3 were marked magenta. Dutch KpnCluster-019 isolates were marked with an *. A dotted line marks the 85% cut off. www.nature.com/scientificreports/ seven cluster isolates. KpnCluster-019 and KpnCluster-021 associated with the Caribbean contained plasmids encoding genes for phenicol and tetracyclin resistance. The KpnCluster-019 and KpnCluster-021 plasmids were not found in non-cluster isolates, whereas the plasmids of the other clusters were detected in a subset non-cluster isolates (Fig. 3). The plasmids of KpnCluster-003 and KpnCluster-005 were present in each of its cluster isolates, however, in isolates of the other clusters occasionally plasmids were lost, thereby impacting the composition of the resistome (Figs. 2 and 3). The bla KPC-2 KpnCluster-019 isolates were obtained from both the Caribbean and the Netherlands, while bla KPC-2 KpnCluster-021 isolates originated only from the Caribbean (Table 1, Fig. 3). In the KpnCluster-019 isolate RIVM_C014906, three copies of the bla KPC-2 gene were present, while other cluster isolates had only one bla KPC copy. One copy was located in the chromosome, one copy in the 200 kb plasmid pRIVM_C014906_1 and a third copy on the 16 kb plasmid pRIVM_C014906_3. All these three bla KPC-2 copies were located on a highly similar Tn4401a-derived ∆Tn4401a-like transposon of 5.6 kb in this strain. The chromosomes contained this ∆Tn4401a-like transposon in the exact same region. KpnCluster-003, KpnCluster-005, KpnCluster-008 and KpnCluster-025 consist of isolates that were obtained in the Netherlands and in these isolates the bla KPC allele was located on a Tn4401a transposon of 10 kb. comparison of the K. pneumoniae plasmid content among clusters. An UPGMA clustering based on the DNA sequence of the 22 plasmids from distinct genetic clusters revealed that the majority of the plasmids were unrelated (Fig. 4). The largest two plasmids pRIVM_C008981_1 from KpnCluster-003 and pRIVM_ C014947_1 from Kpncluster-021 carried the largest number of genes and this number decreased by the decreasing size of the plasmids. Most these plasmid located genes had unknown function. The large and medium sized plasmids contained the klcA gene, encoding an antirestriction protein implicated in the facilitation of bla KPC allele transfer 22 . None of the plasmids contained known virulence determinants such as rmpA, rmpA2, iroBC, or iucABC implicated in hypervirulence 23,24 . Comparison of the large plasmids revealed that pRIVM_C008981_1 and pRIVM_C015139_1 from KpnCluster-003 and KpnCluster-005 displayed 90% similarity (Fig. 4). Plasmid pRIVM_C014947_1 from KpnCluster-021 was not related to any other of the large plasmids. Despite the low similarity, these large plasmids from KpnClusters-003, -005, and -019, shared important clusters of genes among them. They all contained the silE and silP genes encoding a silver-binding protein and a silver exporting ATPase, cusSRCFB genes implicated in cation efflux, the copABCD-pcoE genes involved in copper resistance and the arsHACBAD arsenic resistance gene cluster. These large plasmids also contained fecIRABCDE implicated in Fe(3+)-dicitrate transport, the traIDSQCVAJM-ylpA plasmid conjugation gene cluster, and the higA-higA1 antitoxins, except pRIVM_C014947_1 from KpnCluster-021. In addition, the large plasmids also contained a proportion of plasmid-specific and thus K. pneumoniae cluster specific gene content (Suppl. Figure 2).  Figure 3. Antimicrobial resistance genes on chromosomes and plasmids. The presence of AMR genes among the 22 plasmids of seven TGS sequenced isolates is indicated with black squares and for the chromosomes using green squares. Chromosomes (cRIVM_C0xxxx) and plasmids (pRIVM_C0xxxx) are depicted on the Y-axis, and AMR genes on the x-axis. Antibiotic classes are indicated above the AMR genes in different colors.  pRIVM_C014947_1  pRIVM_C014906_1  pRIVM_C018535_1  pRIVM_C015139_1  pRIVM_C008981_1  pRIVM_C015139_2  pRIVM_C015451_1  pRIVM_C015274_1  pRIVM_C015451_2  pRIVM_C014906_2  pRIVM_C015274_2  pRIVM_C015139_3  pRIVM_C014947_3  pRIVM_C015274_3  pRIVM_C014906_3  pRIVM_C018535_2  pRIVM_C014947_2  pRIVM_C014947_4  pRIVM_C014947_6  pRIVM_C014947_5  pRIVM_C014947_7  pRIVM_C014947_8 KpnCluster 21 19 19 5 3 5 25 8 25 19 8 5 21 8 19 19 21 21 21 21 21 21 Gene product G ene Figure 4. K. pneumoniae plasmid gene content. An UPGMA clustering was performed based on the plasmid DNA sequence for the determination of the genetic relation among the 22 plasmids. Similarity is indicated on the y-axis using a scale from 0 (not similar) to 100% (identical). A similarity of ≥ 85 to 100% is regarded as the same plasmid. The plasmids are indicated on the x-axis. The presence (black squares) and absence is indicated of annotated genes among the 22 plasmids of seven TGS sequenced isolates. If a gene was present twice, blue squares were used and more than 2, red squares were used. Colors indicated different groups of genes with a specific function. In the UPGMA tree, large plasmids are indicated in red, medium plasmids in black and small plasmids in green color. www.nature.com/scientificreports/ Three medium-sized plasmids contained the virB virulence regulon transcriptional activator and the merAC mercuric reductase and transport protein. While pRIVM_C015274_1 from KpnCluster-008 and pRIVM_ C015451_1 from KpnCluster-025 contained a plasmid conjugation gene cluster, plasmids pRIVM_C014906_2, pRIVM_C015139_2, and pRIVM_C015451_2 contained truncated versions hereof. The more distantly related pRIVM_C014906_2 plasmid from KpnCluster-019 had in addition to the higA-higA1 antitoxins also a ccdA-ccdB toxin-antitoxin system. The small plasmids (< 50 kb) contained genes implicated in virulence. Plasmids pRIVM_C015139_3 and pRIVM_C015274_2 displayed 99% similarity and carried the virD4-B9-B8-B4-ptlH Type IV secretion system. pRIVM_C014947_3 contained a merPT mercuric transport system, while pRIVM_ C014906_3 and pRIVM_C015274_3 carried a ceaC colicin-E3. The plasmid pRIVM_C014947_5 contained the bdlA gene encoding a biofilm dispersion protein.

aac(3)-IIa aac(3)-IId aac(6')-Ib aac(6')-Ib-cr aadA1 aadA2 aph(3'')-Ib aph(3')-Ia aph(6)-Id
transposable elements in K. pneumoniae plasmids from distinct clusters. The large and medium sized plasmids contained the most transposase sequences, and each plasmid had its unique transposon signature (Fig. 5). The IS1, IS110 and IS3 transposase families dominated in the K. pneumoniae plasmids among genetic clusters. The IS1 family transposase was found most frequently among the plasmids and in most copies within plasmids. In the large and medium sized plasmids, the bla KPC allele was located on a Tn4401a transposon, except in pRIVM_C014906_1 and pRIVM_C01835_1 from KpnCluster-019. In the small plasmids carrying a bla KPC , the carbapenemase allele was located on a ∆Tn4401a-like transposon. The large plasmids pRIVM_C008981_1, pRIVM_C015139_1 and pRIVM_C014906_1 from KpnClusters-003, -005, and -019 harbored 37, 32 and 31 annotated tranposases, respectively. In contrast, the largest plasmid pRIVM_C014947_1 of 285 kb from Kpn-Cluster-021 contained only 16 transposons. The remainder of the plasmids from KpnCluster-021 also contained very few transposase sequences, in contrast to the other plasmids from the different clusters. The highly related pRIVM_C015139_3 and pRIVM_C015274_2 plasmids (99% similarity) from KpnCluster-005 and KpnCluster-008 had identical transposons. While IS66 and IS110 family transposase sequences also dominate in the large plasmids, the medium sized plasmids contained IS3 family type of transposases. The medium sized plasmids contained eleven to 23 transposases, and the small plasmids less than ten.
Similarity with previously reported plasmids. BLAST analysis of the K. pneumoniae plasmids identified in this study showed that 15 of the 22 plasmids were similar to previously reported plasmids in the NCBI sequence database (Table 2). These plasmids covered five distinct genetic clusters, except pRIVM_C008981_1 from KpnCluster-003. To date, none of these plasmids of K. pneumoniae were reported to be implicated in healthcare-associated outbreaks. Plasmids pRIVM_C008981_1, pRIVM_C014906_1, pRIVM_C014906_3 containing bla KPC-2 and pRIVM_C015274_1 harboring bla KPC-3 from distinct genetic clusters only had low sequence coverage 35-87% with plasmids present in the NCBI sequence database. The other bla KPC-2 and bla KPC-3 plasmids had high (93-99%) sequence coverage, indicating that these similar plasmids were detected previously by other researchers. Plasmids pRIVM_C014906_2, pRIVM_C015139_1, pRIVM_C015274_2 and pRIVM_C015274_3, not carrying a bla KPC allele, displayed 97-100% sequence coverage and 99-100% identity to plasmids isolated from K. pneumoniae from different countries ( Table 2). Plasmids pRIVM_C014947_5 and pRIVM_C014947_6 from KpnCluster-021 had 100% sequence coverage with 92.18 to 99.99% identity with plasmids isolated from Enterobacter hormaechei. Plasmids similar to the eight plasmids from KpnCluster-021 were detected previously in a variety of hosts, e.g. Salmonella enterica, K. pneumoniae, and E. hormaechei, suggesting these plasmids are broad-host range. The fact that 15 of the 22 plasmids analyzed in this study were found previously in distinct hosts, suggest international spread of these plasmids.
Prophage sequences in the K. pneumoniae cluster genomes. PHASTER analysis revealed that the majority of the large and medium-sized plasmids from different genetic clusters with IncFIB(K) or IncFIB(pQil) and IncFII(K) replicons contained one to four regions with prophage-related sequences e.g. genes encoding putative phage integrase, phage-like proteins, coat proteins, and/or tail shaft proteins ( Table 3). The size of the prophage sequence regions varied per plasmid. The most commonly found prophage-related sequence in large and medium-sized plasmids of cluster isolates was an Escherichia phage RCS47 (Table 3). This sequence entails the 14.2 kb ygbMLKJI-bla SHV -recF-lacY region flanked by IS26 elements and representing 12% of the RCS47 prophage genome. The small plasmids of < 50 kb lacked phage-related sequences. In contrast, the chromosomes of cluster isolates carried at least three to nine phage sequence regions covering 10-50% of the phage genome. These phage sequence regions covered a wide variety of distinct phages, including prophage sequences from Salmonella, Klebsiella, Cronobacter, Enterobacteria phages (Suppl. Table 1). The most commonly found prophage sequence in Klebsiella chromosomes was the Enterobacteria phage P4.

Discussion
We showed that a K. pneumoniae ST560 strain carrying bla KPC-2 from KpnCluster-019 was transmitted between the Netherlands and the Caribbean. This is based on the high genetic relatedness of the 13 isolates from KpnCluster-019 as assessed by wgMLST and their highly similar resistome and plasmidome. We found that one person lived in the Caribbean and migrated to the Netherlands. After migration, a KpnCluster-019 isolate was obtained from this person in a Dutch hospital. Possibly other transmissions by other persons could have occurred, but these were not confirmed in this study. By combining short-read with long-read sequencing data, we identified 22 plasmids of seven K. pneumoniae isolates from six distinct genetic clusters found in the Netherlands and the Caribbean and analyzed these plasmids for its AMR gene profile, bla KPC transposons, replicons, transposon families, and gene content. The plasmid composition varied among the genetic clusters. Some of the cluster isolates had unique MLST sequence types (ST144 and ST560) which were not published previously and differ from globally Scientific RepoRtS | (2020) 10:16778 | https://doi.org/10.1038/s41598-020-73440-2 www.nature.com/scientificreports/ circulating extensively drug-resistant (XDR) K. pneumoniae ST258 and ST307 strains 23,24 . KpnCluster-019 is unique compared to the other cluster isolates analyzed in this study for the following reasons. First, KpnCluster-019 harbors a unique and extensive set of AMR genes on the chromosome and in its plasmids. Secondly, KpnCluster-019 isolates were the only to contain three copies of the bla KPC-2 allele, two on two different plasmids and one in the chromosome. The localization of bla KPC-2 on the chromosome and additional bla KPC-2 copies have been reported previously and is further complicating the understanding of transmission of multidrug-resistant K. pneumoniae 25,26 . Thirdly, KpnCluster-019 and also KpnCluster-021 isolates from the Caribbean harbored the bla KPC-2 allele on a 5.6 kb ∆Tn4401a-like transposon, while the other isolates from the other genetic clusters from the Netherlands contained bla KPC on a 10 kb Tn4401a transposon. Most global descriptions of K. pneumoniae bla KPC the past decade have been associated with Tn4401a or isoforms hereof 9 . The traditional association of bla KPC with the Tn4401a transposon has possibly been eroded in K. pneumoniae isolates from the Caribbean   pRIVM_C014947_1  pRIVM_C014906_1  pRIVM_C018535_1  pRIVM_C015139_1  pRIVM_C008981_1  pRIVM_C015139_2  pRIVM_C015451_1  pRIVM_C015274_1  pRIVM_C015451_2  pRIVM_C014906_2  pRIVM_C015274_2  pRIVM_C015139_3  pRIVM_C014947_3  pRIVM_C015274_3  pRIVM_C014906_3  pRIVM_C018535_2  pRIVM_C014947_2  pRIVM_C014947_4  pRIVM_C014947_5  pRIVM_C014947_6  pRIVM_C014947_7  pRIVM_C014947_8   KpnCluster   21 19 19 5 3 5 25 8 25 19 8 5 21 8 19 19 21 21 21 21 21 21 bla KPC-2 Large ≥150-250 kb Medium ≥50-150 kb Small <50 kb www.nature.com/scientificreports/ to a smaller variant. This is the first report of identification of a 5.6 kb ∆Tn4401a-like bla KPC-2 transposon of K. pneumoniae in the Netherlands. Preliminary surveillance data analysis revealed that the ∆Tn4401a-like element carrying bla KPC-2 and smaller variants disseminated among E. cloacae, Serratia marcesens, K. oxytoca and E. coli in the Netherlands (unpublished data). Future work will seek to understand the dissemination of the ∆Tn4401alike bla KPC-2 element among CPE in the Netherlands. Lastly, the plasmids of KpnCluster-019 isolates contained also unique plasmid content, including a distinct transposon signature, two toxin-antitoxin systems and a ceaC colicin which possibly contribute to the success in survival, niche adaptation or transmission of this strain. The K. pneumoniae bla KPC-3 isolates had higher MICs for meropenem than the K. pneumoniae bla KPC-2 isolates, which is in line with a previous study 21 . The KPC-2 enzyme differs in a single amino acid substitution (Histidine 272 to Tyrosine) from KPC-3. Additional changes in KPC-3 can lead to increased resistance for ceftazidime and cephamycin 27 . The increase in meropenem resistance observed in our study is possibly correlated with improved ability of KPC-3 enzymes to hydrolyze the meropenem antibiotic 15 . Alternatively, additional beta-lactamase genes such as bla OXA-1 , bla  or bla TEM-1A may contribute to increased resistance for meropenem 28 .
Despite the limited number of long-read sequenced isolates, we have highlighted important new insights in the genomic surveillance of a notorious multi-antibiotic resistant nosocomial pathogen. In some clusters, the plasmidome varied as this was likely due to loss of a plasmid. Also, the resistome data suggest the presence of other plasmids in cluster isolates that were not present in the isolates that were sequenced using TGS. To overcome this limitation, all isolates used in this study should have been sequenced using long-read third generation www.nature.com/scientificreports/ sequencing. Nevertheless, we identified plasmids in K. pneumoniae bla KPC-2 and bla KPC-3 cluster isolates which vary in size from large, medium and small. The large and medium sized plasmids were enriched for a variety of transposons, conjugation transfer systems, cation efflux systems including Fe(3+)-dicitrate transport, and genes encoding for silver, copper and arsenic resistance. The small plasmids contained putative virulence determinants. The presence of these systems may contribute to the success of transmission of specific K. pneumoniae strains in the hospital setting or the community 13,29,30 . Escherichia RCS47 prophage sequences were found on medium and large plasmids in the cluster isolates analyzed. In contrast, the chromosomes contained a variety of prophagerelated sequences. RCS47 is a P1-like bacteriophage carrying the ESBL-encoding bla SHV-2 gene was isolated from a clinical E. coli strain 31 . The prevalence of P1-like prophages in animal and human E. coli strain collections was 12.6% 31 . The presence of P1-like phage sequences in plasmids of a snapshot of the K. pneumoniae population in the Netherlands and the Caribbean suggest that the role of P1-like phages in disseminating antibiotic resistance may be underestimated 32 .
In conclusion, long-read sequencing contributed to the understanding of the successful transmission of the KpnCluster-019 K. pneumoniae bla KPC-2 strain. Plasmid content such as conjugation machinery, transposons, virulence determinants and phages may contribute to diversification, and dissemination of plasmids containing AMR genes, and therefore represent important plasmid features that warrants future investigation. More longread plasmid sequencing efforts of CPE and K. pneumoniae in particular are required to identify the complete plasmid reservoir involved in the spread of antibiotic resistance determinants in the Netherlands and the Caribbean islands.

Methods
Bacterial isolates. For the Dutch National carbapenemase-producing Enterobacterales (CPE) Surveillance program, medical microbiology laboratories from the Netherlands and the Caribbean routinely send CPE isolates with a meropenem minimum inhibitory concentration (MIC) of ≥ 0.25 µg/ml and/or an imipenem MIC of ≥ 1 µg/ml or phenotypic (CIM-test) or genotypical evidence of carbapenemase production to the National Institute of Public Health and the Environment (RIVM) 16 . For this study, 84 carbapenemase-producing K. pneumoniae isolates carrying either the bla KPC-2 allele or the bla KPC-3 allele were included and collected in the period from January 1st 2014 until June 30th 2019. Only the first K. pneumoniae isolate per person in this study period was selected. The 84 isolates were obtained from 84 persons and from various isolation sites, i.e. rectum/perineum (n = 43), throat (n = 11), pus (n = 2), sputum (n = 4), urine (n = 10), wound (n = 5) and nine were from miscellaneous isolation sites. All bacterial strains were grown aerobically at 37 °C on Columbia sheep blood agar plates.
Antimicrobial susceptibility testing. Resistance to carbapenem was confirmed by assessing the MICs for meropenem for all the 84 isolates using an Etest (bioMérieux Inc., Marcy l'Etoile, France). Based on the clinical breakpoints according to EUCAST, the K. pneumoniae isolates were classified as sensitive (≤ 2 mg/L), www.nature.com/scientificreports/ intermediate (> 2 mg/L and ≤ 8 mg/L) and resistant (> 8 mg/L) to meropenem. In addition, all isolates were analyzed for the production of carbapenemase using the carbapenem inactivation method (CIM) as described previously 33 .
Next-generation sequencing, MLST and wgMLST. All 84 K. pneumoniae isolates were subjected to next-generation sequencing (NGS) using the Illumina HiSeq 2500 (BaseClear, Leiden, the Netherlands). The NGS data of the K. pneumoniae isolates were used for classical MLST and wgMLST analyses using the in-house wgMLST scheme in SeqSphere software version 6.0.2 (Ridom GmbH, Münster, Germany). The in-house K. pneumoniae wgMLST scheme was comprised of 4978 genes (3471 core-genome and 1507 accessory-genome targets) using K. pneumoniae MGH 78,578 (NC_009648.1) as a reference genome 21 . For classical MLST, the existing scheme was used and cluster nomenclature were depicted in Table 1 34 . The resulting data was imported into Bionumerics version 7.6.3 for subsequent comparative analyses (Applied Maths, Sint-Martens-Latem, Belgium). The antibiotic resistance gene profile and plasmid replicon compositions in all of the isolates were determined by interrogating the online ResFinder (version 3.1.0) and PlasmidFinder (version 2.0.2) databases available at the Center for Genomic Epidemiology website (https ://cge.cbs.dtu.dk/servi ces/) 35,36 . For ResFinder, a 90% identity threshold and a minimum length of 60% were used as criteria, whereas for PlasmidFinder, an identity of 95% was utilized.
Long-read third-generation sequencing. One K. pneumoniae isolate per genetic KpnCluster was sequenced using long-read third-generation Nanopore sequencing 18,37 . High molecular weight DNA was isolated using an in-house developed protocol. Bacteria were grown overnight in 1.5 ml Brain heart infusion broth and culture was spun down at 13,000 × g for 2 min. The pellet was washed and resuspended in 500 µl of 150 mM NaCl. The suspension was spun down at 5000 × g for 5 min and the pellet was resuspended in 100 µl of Quick-Extract DNA Extraction Solution (Lucigen) and 0.1 µl Ready-Lyse Lysozyme solution (Epicentre) and incubated for 1 h at 37 °C. Subsequently, 85 µl 10 mM Tris 1 mM EDTA pH = 8 (1 × TE), 10 µl proteinase K (> 600 mAU/ mL, Qiagen) and 5 µl 20% sodium dodecyl sulfate solution were added, and the mixture was incubated at 56 °C for 30 min. DNA was precipitated overnight at -20 °C by adding 0.1 × volume 3 M sodium acetate pH = 5.2 and 2.5 × volume ice cold 100% ethanol. Precipitated DNA was spun down at 13,000 × g for 15 min and pellets were washed with 1 ml 70% ethanol followed by another centrifugation at 13,000 × g for 5 min. After drying, the pellet was dissolved in 200 µl 1 × TE and diluted to 1 µg with Nuclease-free water. The Oxford Nanopore protocol SQK-LSK108 (https ://commu nity.nanop orete ch.com) and the expansion kit for native barcoding EXP-NBD104 was used. Briefly, a shearing step was performed using g-TUBE's (Covaris) to obtain an average DNA fragment size of 8 kb. The DNA was repaired using FFPE and end-repair kits (New England BioLabs) followed by ligation of barcodes with bead clean up using AMPure XP (Beckman Coulter) after each step. Barcoded isolates were pooled and sequencing adapters were added by ligation. The final library was loaded onto a MinION flow cell (MIN-106 R9.4.1). The 48-h sequence run was started without live base calling enabled on a MinION device connected to a desktop computer. After the sequence run, base calling and de-multiplexing was performed using Albacore 2.3.1 and a single FASTA file per isolate was extracted from the FAST5 files using Poretools 0.5.1 38 . Illumina and Nanopore data were used in a hybrid assembly performed by Unicycler v0.4.4 39 . The resulting contig files were annotated using Prokka and were subsequently loaded into BioNumerics for further analyses 40 . Minimum spanning tree and UpGMA analyses. The BioNumerics software was used to generate a minimum spanning tree (MST) or an UPGMA hierarchical clustering as described previously 16 . The MST was based on an in-house K. pneumoniae wgMLST scheme. The categorical coefficient was used to calculate the MST. wgMLST clusters were defined as a minimum of two isolates of which the genetic distance between the two isolates was ≤ 20 genes (20/4978 ≤ 0.4% different). An UPGMA clustering of K. pneumoniae bla KPC-2 and bla  isolates was performed based on the presence and/or absence of antibiotic resistance genes per isolate. plasmid reconstruction by read mapping. The CLC Genomics Workbench version 12.0 software (www.qiage nbioi nform atics .com) was used to reconstruct plasmids. For this, complete plasmids obtained by TGS were used as a scaffold to map the trimmed NGS reads of isolates that were from the same genetic wgMLST cluster. A plasmid was scored "present" in an isolate if reads mapped to a reference plasmid of interest and ≥ 85% of the consensus sequence size in kilo bases was reconstructed. Linear DNA fragments < 5 kb were omitted in this study. Nucleotide BLAST analyses on plasmid sequences were performed using the https ://blast .ncbi.nlm. nih.gov website and date from October 2019. plasmid content analysis. Bionumerics was used to extract and analyze annotated genes and tranposases in the 22 different plasmids. The data was plotted in Excel. Phaster, the PHAge Search Tool Enhanced Release website (https ://phast er.ca/) was used to determine the presence of phage sequences in the plasmids and searches date from October 2019 41 . ethics statement. The bacterial isolates used in this study belong to the medical microbiological laboratories participating in the Dutch National CPE Surveillance program and was obtained as part of routine clinical care in the past. Since no identifiable personal data were collected and data were analyzed anonymously, written or verbal patient consent was not required for this study and was therefore not obtained. According to the Dutch Scientific RepoRtS | (2020) 10:16778 | https://doi.org/10.1038/s41598-020-73440-2 www.nature.com/scientificreports/ Medical Research Involving Human Subjects Act (WMO) this study was considered exempt from review by an Institutional Review Board.

Data availability
The Illumina (NGS) and plasmid sequence data sets generated and analyzed in this study are freely available in the Sequence Read Archive (SRA) under BioProject ID PRJNA634885 and in Genbank under accession numbers as depicted in Supplementary Table 2.