Regional Dissemination of a Carbapenemase-encoding Plasmid: Plasmidome analysis Reveals Diverse Adaptations of Carbapenemase-Producing Enterobacteriaceae

Background. The global dissemination of carbapenem-resistant Enterobacteriaceae (CRE) threatens human health by limiting the range of usable antibiotics even against common bacterial infections. The spread of CRE is primarily due to the transmission of carbapenemase genes located on plasmids. However, few studies have comprehensively identied regionally spreading carbapenemase-encoding plasmids because of the diculty to determine the complete sequence of a plasmid encoding carbapenemases. In a CRE surveillance study of 1,507 patients from 43 hospitals in northern Osaka, Japan, we previously found that 12% of the patients carried CRE and 95% of CRE isolates were IMP-6 producers. This result suggested a vast horizontal spread of a clonal plasmid carrying bla IMP-6 among Enterobacteriaceae in this region. In the current study, we aimed to describe the dynamics of this regional horizontal plasmid transmission. to trace only by the comparisons of the whole genomes. A seemingly clonal horizontal dissemination of the predominant plasmid had embraced heterogenous subpopulations that contribute to diverse adaptations including covert transmission, stable chromosomal integration of bla IMP-6 , or broadened antimicrobial resistance patterns, ultimately leading to treatment failure.


Introduction
The rapid global dissemination of multidrug-resistant Enterobacteriaceae threatens healthcare systems worldwide [1]. Carbapenem-resistant Enterobacteriaceae (CRE) are of major concern because alternative treatment options are limited [2]. Carbapenem resistance is primarily conferred by carbapenemases, which are enzymes that hydrolyse carbapenem [3]. KPC, NDM, and OXA-48 are the most commonly detected carbapenemases [3]. Carbapenemase genes are generally plasmid-encoded and are frequently transmitted across species [4]. Genetic tracking of plasmids encoding carbapenemase genes has allowed monitoring the spread of CRE isolates. For example, structural similarities among plasmids from isolates obtained in a single hospital outbreak allowed elucidating links between patients carrying the isolates [5][6][7], and plasmid data accumulated globally revealed the worldwide spread of an epidemic plasmid carrying bla KPC . [8]. However, most regional surveillance studies compared the whole genomes of CRE isolates without analysing the clonality of the spreading carbapenemase-encoding plasmids, and few studies have comprehensively analysed regionally spreading carbapenemase-encoding plasmids in order to reveal the modes of horizontal plasmid transmission in a certain region [9].
We previously conducted a surveillance study of CRE in 1,507 patients from 43 hospitals in northern Osaka (population: 1,170,000, area: 307 km 2 ), Japan [10], and we reported that 12% of the patients carried CRE and 95% of CRE isolates harboured bla IMP−6 , the predominant carbapenemase in Japan. The predominance of this particular carbapenemase gene might have resulted from vigorous horizontal spreading of a speci c plasmid carrying bla IMP−6 , pKPI-6 [11], in this region. The aim of the current study was to pro le the mode of carriage of carbapenemase genes, primarily bla IMP , to unveil their diversity within a de ned geographical region.

Results
Dissemination of pKPI-6 All bla IMP -positive CRE isolates of Escherichia coli (n = 135) and Klebsiella pneumoniae (n = 95) were classi ed into seven groups based on the results of S1-PFGE followed by Southern blot hybridization with probes for the bla IMP and repA genes ( Figure 1). Ninety-nine of the 135 E. coli isolates (73%) and 88 of the 95 K. pneumonia isolates (93%) carried plasmids classi ed as Group pKPI-6 based on plasmid size and replicon type (Supplementary Figure S1). pKPI-6 was the predominant plasmid responsible for the transmission of bla IMP-6 (187 out of 230 bla IMP -positive CRE isolates, 85.6%). Next, we compared the similarity between pKPI-6 and 39 representative plasmids categorized as Group pKPI-6 based on WGS data. The overall sequence identity was 99 ± 0.28%, and the sequence coverage was 98 ± 4.0% (mean ± standard deviation) (Supplementary Figure S1). This analysis con rmed that pKPI-6 is the predominant plasmid carried by CRE isolates in the study area.
Genomic Analysis of Derivatives of the Predominant Plasmid, pKPI-6 During the characterization of the bla IMP-6 plasmids mentioned above, nine E. coli isolates and three K. pneumoniae isolates possessed bla IMP-6 plasmids categorized as Group IncN (Figure 1). Group IncN bla IMP-6 plasmids were characterized by replicon type IncN and ranged from 35 to 264 kbp in size, which was different from the pKPI-6 plasmid of 50 kbp (Supplementary Figure S1). The complete sequences of these plasmids indicated that they had preserved the nearly complete locus of pKPI-6 and typically were multi-replicon plasmids that had integrated IncF-type plasmids framed by insertion sequences (Supplementary Figure S2 and Table S1). Additionally, two isolates (E208 and E328) of K. pneumoniae harboured plasmids categorized as Group Non-IncN KP ( Figure 1B). These plasmids comprised a cassette carrying bla IMP-6 without IncN-type repA of the pKPI-6 plasmid integrated into another plasmid (Supplementary Figure S3). Interestingly, E. coli isolate E119 and K. pneumoniae isolate E206 coharboured two distinct bla IMP-6 -encoding plasmids of different sizes and were categorized as Group double bla IMP-6 ( Figure 1 and Supplementary Figure S4). Barring occasional isolations of strains coharbouring different carbapenemase genes [12,13], few studies have shown the coexistence of two identical carbapenemase genes on different plasmids within an isolate [14]. WGS revealed that isolate E119 carried pKPI-6 and an IncF-type plasmid (pEC743_1) that had a bla IMP-6 cassette from pKPI-6 integrated (Supplementary Figure S5).
Characterization of IncF Plasmids Encoding bla  In addition to the K. pneumoniae isolates carrying Group Non-IncN KP plasmids, E. coli isolates carrying plasmids without IncN replicon were found in a single hospital (hospital D; Figure 1A). WGS of these isolates revealed that they harboured nearly identical bla IMP-6 -encoding plasmids with an IncFIA-type replicon (categorized as Group IncF) (Supplementary Figure S6A and Table S1). These plasmids were generated by integration of a cassette carrying bla IMP-6 on pKPI-6 into another IncF plasmid at IS26. This IncF plasmid (pEC302/04; Supplementary Figure S6B) has been reported to transmit antimicrobial resistance since 1965 [15].
The minimum inhibitory concentrations (MICs) of meropenem for the E. coli isolates carrying Group IncF plasmids were low when compared with those of E. coli isolates harbouring other bla IMP-6 -encoding plasmids, such as pKPI-6 (Supplementary Figure S7). Mutations or deletions in the porin (OmpF) gene in E. coli have been reported to enhance resistance to β-lactams [16]. However, all E. coli isolates carrying Group IncF plasmids had a premature termination codon within ompF, whereas the other isolates carried wild-type ompF (Supplementary TableS2 and S3). MICs of meropenem were low for these Group IncF plasmid-carrying isolates, despite them being OmpF-de cient. To investigate carbapenem resistance in the same genetic background, plasmids from representative isolates in each bla IMP-6 carriage group were transformed into the E. coli TOP10 strain and MICs for the transformants were determined. Transformant T305 carrying pE305_IMP6 single of Group IncF from E. coli isolate E305 was more susceptible to meropenem than transformants carrying bla IMP-6 -harbouring plasmids of groups (Supplementary Table   S4). bla IMP-6 transcription in the pE305_IMP6 single transformant was signi cantly lower than that in the pKPI-6 transformant (Supplementary Figure S8A), although the plasmid copy numbers in the bacterial cells were comparable (Supplementary Figure S8B). These results indicated that the lower MICs of meropenem in E. coli isolates carrying Group IncF plasmids were due to reduced transcription of bla IMP-6 .
WGS of E305 and E318 revealed the complete sequence of pE318_IMP6; however, it failed to determine the complete sequence of pE305_IMP6. Therefore, to analyse the structure of pE305_IMP6, we used a combination of WGS, Southern blotting, and qPCR analysis. The length and depth of each contig of pE305_IMP6 deduced from WGS are shown in the de-novo assembly graphs generated using the Bandage software [18] in Figure 2A. The total length of pE305_IMP6 deduced from WGS data was approximately 149 kbp. However, according to Southern blotting results, pE318_IMP6 and pE305_IMP6 were ~145 kbp and ~200 kbp in size, respectively ( Figure 2B). Based on the depth of each contig, the copy number of each contig was predicted as follows: Contig3, 1 copy; Contig2 and Contig5, 6 copies; Contig1 and Contig6, 3 copies; Contig4, 5 copies ( Figure 2A). Therefore, pE305_IMP6 was predicted to have a ~19-kbp repeat region consisting of triplication of Contig1 and Contig6, sextuplication of Contig2 and Contig5, and quintuplication of Contig4 ( Figure 2C). Except for the repeat region, pE305_IMP6 and pE318_IMP6 exhibited high sequence similarity (identity; 99.27%, coverage; 100%) ( Figure 2D). The bla IMP-6 gene was located on Contig6 and was predicted to be triplicated. qPCR analysis corroborated that pE305_IMP6 carried three copies of bla IMP-6 , whereas pE318_IMP6 harboured a single copy (appendix p10). bla IMP-6 transcription was signi cantly higher in isolate E305 than in isolate E318 ( Figure 2E), even though the bla IMP-6 -carrier plasmid copy numbers in the cells of these isolates were not signi cantly different (Supplementary Figure S9B). Triplication of bla IMP-6 in tandem resulted in a higher transcription level in E305, resulting in a higher level of resistance to meropenem.
Subculture of the clonal isolate E305 in broth medium revealed a mixture of subpopulations of bacteria carrying a plasmid with multiple bla IMP-6 copies (which represented the majority) and bacteria carrying a plasmid with a single bla IMP-6 copy. In Southern blotting for bla IMP-6 , a faint band at ~145 kbp was observed in addition to the major band at ~200 kbp ( Figure 2B). It was also found that T305 (transformant of pE305_IMP6 single extracted from E305) carried a ~145-kbp plasmid without bla IMP-6 ampli cation due to recA de ciency in the recipient E. coli TOP10 strain (Supplementary Figure S10) [19]. qPCR analysis con rmed that T305 carried one bla IMP-6 copy on its plasmid (Supplementary Figure S9C).
These results indicated the existence of a subpopulation carrying a plasmid with one bla IMP-6 copy within E. coli isolate E305, whereas the majority of the population carried a plasmid harbouring three copies of bla IMP-6 .
Comparison of CRE Isolates Carrying pKPI-6 with Those Carrying Other Groups of Plasmids Harbouring bla IMP-6 bla CTX-M-2 , which is an ESBL gene located distant from bla IMP-6 on pKPI-6, compensated for the narrow range of hydrolysis of β-lactams by IMP-6 [11,17]. However, toweverhhhhese two β-lactamase genes were not always transferred together from pKPI-6 to another plasmid. Plasmids categorized as Group Non-IncN KP and Group IncF did not carry ESBL genes (Supplementary Table S6) and rarely conferred resistance to penicillins, in contrast to pKPI-6, which confers broad resistance to β-lactams ( Figure 1). We next measured the conjugation e ciency of representative plasmids in each group (Supplementary Table  S7). pKPI-6 plasmids and Group IncN plasmids, which had the entire pKPI-6 plasmid incorporated, showed a higher conjugation e ciency than Group Non-IncN KP/IncF plasmids. These characteristics may have facilitated the vast horizontal dissemination of pKPI-6 in the study area.
Compared with the chromosomal diversity among E. coli isolates bearing pKPI-6, K. pneumoniae isolates carrying pKPI-6 exhibited higher clonality as indicated by XbaI-PFGE analysis (Figure 1). This may be explained by the presence of the kikA gene on pKPI-6, the product of which reportedly promotes cell death of K.pneumoniae following conjugation [20]. The conjugation e ciency of pKPI-6 into K. pneumoniae ATCC13883 was considerably lower than that into E. coli TUM3456 (3.3 × 10 -4 and 3.7 × 10 -1 , respectively). Maybe only "kikA-resistant" K. pneumoniae are able to acquire pKPI-6, leading to clonal similarity among the K. pneumoniae isolates bearing pKPI-6.
Chromosomal Integration of bla  Unlike most CRE isolates, which carried the predominant pKPI-6 or other bla IMP-6 -encoding plasmids, three out of 135 E. coli isolates (E138, E300, and E302) harboured bla IMP-6 on their chromosomes as indicated by S1-PFGE followed by Southern blotting with bla IMP-6 probes ( Figure 1A and Figure 3A). I-CeuI-PFGE followed by Southern blotting with probes for the bla IMP-6 and 16S rRNA genes con rmed chromosomally located bla IMP-6 ( Figure 3B). WGS revealed that the chromosome of isolate E138 had a cassette harbouring bla IMP-6 integrated, framed by a set of IS15 ( Figure 3C). The chromosomes of E300 and E302 had IncFIA plasmids carrying bla IMP-6 integrated ( Figure 3D,E). While these plasmids were essentially identical to pE301_IMP6 (E. coli, Group IncF), these isolates were phylogenetically distinct on the XbaI-PFGE phylogenetic tree (Figure 1).
Emergence of pKPI-6-like Plasmid Harbouring bla IMP-1 One K. pneumoniae isolate, E105, harboured bla IMP-1 , which is a single-nucleotide variant of bla IMP-6 , within a clonal cluster of pKPI-6 carriers ( Figure 1B). Due to this mutation, E105 was resistant to imipenem, whereas most isolates carrying bla IMP-6 were susceptible to this antibiotic. WGS revealed that plasmids pKPI-6, pE013_IMP6 (plasmid group pKPI-6), and pE105_IMP1 were 99.8% identical, with a coverage of 100% (query: pE013_IMP6) (Figure 4). The only difference was the presence of a 714-bp region bracketed by a set of homologous regions in pE013_IMP6.

Discussion
IMP-producing Enterobacteriaceae have been reported sporadically on a global basis [2]. IMP-4-producing Enterobacteriaceae are endemic to Australia [21], and IMP-1-, 4-, and 8-producers have been occasionally detected in China [22]. Our study revealed the exclusive dissemination of IMP-6 producers (95% of CRE isolates) in northern Osaka, Japan, consistent with ndings in previous studies [11,23,24]. By tracking plasmids carrying bla IMP−6 , we clari ed the relationships between bla IMP -harbouring isolates that seemed diverse based on XbaI-PFGE analysis or comparison of short-read WGS results.
The current study revealed predominant dissemination of pKPI-6 in the study area, which may have resulted in the emergence of heterogeneous subpopulations. Group IncF plasmids possessed similar genomic structures, consisting of the globally disseminated IncF plasmid and a bla IMP−6 cassette cointegrated on the pKPI-6 genome, without accompaniment of bla CTX−M−2 ( Figure S6). Our analysis revealed that bla IMP−6 transcription was lower from Group IncF plasmid (pE305_IMP6 single ) than from pKPI-6 in E. coli cells of the same genetic background (Supplementary Figure S8). Low carbapenemase gene transcription is considered as one of the reasons for reduced resistance to meropenem [25]. Therefore, CRE isolates carrying Group IncF plasmids might have a reduced tness cost for the carriage of bla IMP−6 , leading to further environmental dissemination of bla IMP−6 [26].
Unlike for other plasmids in Group IncF, the complete sequence of pE305_IMP6 could not be obtained by long-read or short-read sequencing because of a signature 19-kbp repeat sequence unit. Based on combined WGS, Southern blotting, and qPCR data, we proposed a hypothetical structure of pE305_IMP-6 ( Fig. 2C). Our results indicated that, despite its clonal origin, CRE isolate E305 comprised two different populations: a major population carrying pE305_IMP-6 with multiple bla IMP−6 copies and a minor population carrying pE305_IMP-6 single with a single bla IMP−6 copy (Supplementary Figure S10). Moreover, the ampli cation of bla IMP−6 on the IncF plasmid enhanced the transcription of bla IMP−6 (Fig. 2E), resulting in increased resistance to meropenem (Supplementary Table S5). These results are consistent with previous studies reporting higher resistance to carbapenem through ampli cation of bla OXA−58 [27] and bla NDM−1 [19].
All E. coli isolates carrying Group IncF plasmids were found to possess ompF with a premature termination codon (Supplementary Table S3). When an isolate producing wild-type OmpF carries this plasmid with a single copy of bla IMP−6 , it is di cult to detect due to weaker resistance to meropenem.
However, when an isolate with a porin mutation acquires a Group IncF plasmid with multiple bla IMP−6 copies, it may abruptly exhibit strong resistance to meropenem without any direct trace of horizontal transfer. These types of plasmids may act as "hidden transmitters" of bla IMP−6 .
Moreover, we demonstrated chromosomal integration of Group IncF plasmids in some E. coli isolates. Carbapenemase genes have been reported to be transmitted primarily through plasmid conjugation [4], and chromosomal integration has been reported in a limited number of strains [28]. In our study, three out of 135 E. coli isolates (2.2%) exhibited chromosomal integration of bla IMP−6 , which presumably occurred during the vast horizontal spread of pKPI-6. Compared with bla IMP−6 on plasmids, chromosomal bla IMP−6 was not readily transmissible to another patient. However, these isolates may stably possess bla IMP−6 within a patient and not lose carbapenem resistance through the elimination of plasmids harbouring bla IMP−6 .
In the early 1990s, some metallo-β-lactamases were reported in Japan [29,30], followed by the identi cation of IMP-1 [31]. Since then, these β-lactamases have been frequently identi ed in Japan [32]. The single amino-acid variant, IMP-6, was identi ed in 2001 [17]. IMP-1 producers have disseminated mainly in eastern Japan, including Tokyo [23,24,33], whereas IMP-6 producers have been almost exclusively found in western Japan, including Osaka [7,10,11,24]. Consistent herewith, in this study, only one K. pneumoniae isolate carrying bla IMP−1 , E105, was isolated in hospital A, where CRE carrying pKPI-6 were dominant. The patient carrying CRE isolate E105 was hospitalized for 512 days with other inpatients carrying CRE with pKPI-6, and the isolate showed ~ 83% similarity with a cluster of K. pneumoniae isolates carrying pKPI-6 in the XbaI-PFGE phylogeny (Fig. 1B). In addition, WGS of the plasmids revealed that a 714-bp region bracketed by 32-bp homologous regions was the only difference between pE105_IMP1 and pE013_IMP6 (Fig. 4A). This very small fragment appeared to have been removed by homologous recombination in pE105_IMP1 (Fig. 4B). Our results suggest that bla IMP−6 had disseminated via the transmission of pKPI-6, and spontaneous mutation may have generated the bla IMP−1 -encoding plasmid providing broader antimicrobial resistance, resulting in increased tness in the clinical setting.

Conclusions
This multi-institutional surveillance study uncovered the clonal dissemination of a plasmid encoding a speci c carbapenemase IMP-6 and demonstrated that a seemingly clonal horizontal dissemination of CRE isolates had embraced heterogeneous minor subpopulations, which exhibited broadened antimicrobial resistance, stable carriage of bla IMP−6 through chromosomal integration, or heteroresistance related to covert bla IMP transmission. Such diverse gene adaptations might also be common among CRE isolates carrying other carbapenemase genes. By focusing on the modes of carbapenemase gene carriage, this study revealed the clonal dissemination of a carbapenemase-encoding plasmid, along with the presence of diverse subpopulations that would ensure and facilitate the dissemination of carbapenemase genes in various environments, resulting in serious complications in clinical settings.

Materials And Methods
CRE Isolates and PFGE Phylogenetic Analysis. We performed a CRE surveillance study of 1,507 patients hospitalized in 43 hospitals located in northern Osaka between December 2015 and January 2016 [10]. In the current study, we analysed 230 CRE isolates carrying bla IMP obtained in the surveillance study, including 135 E. coli isolates and 95 K. pneumoniae isolates. All isolates were subjected to XbaI-digested PFGE for phylogenetic analysis [34]. Dendrograms were generated from PFGE patterns by the UPGMA method using BioNumerics software (version 6.6) (Applied Maths NV, Sint-Martens-Latem, Belgium).
Classi cation of bla IMP Carriage by PFGE and Southern Blotting. The size and replicon type of bla IMPharboring plasmids were determined by S1-nuclease-digested PFGE followed by Southern hybridization (S1-nuclease was obtained from Takara Bio, Shiga, Japan). S1-PFGE and Southern blot hybridization for the bla IMP−6 and repA genes encoded on the IncN-type plasmid were performed as described in our previous study [35]. The sizes of bla IMP -encoding plasmids were determined using BioNumerics software (version 7.5) (Applied Maths NV). The modes of bla IMP carriage were classi ed into seven groups based on the sizes and replicon types of the plasmids carrying bla IMP . The groups and their associated characteristics are as follows: Group pKPI-6, a pKPI-6-like bla IMP−6 -encoding plasmid (~ 50 kbp, encoding repA for IncN plasmid); Group IncN, a bla IMP−6 -encoding plasmid (not ~ 50 kbp, encoding repA for IncN plasmid); Group Non-IncN KP, a bla IMP−6 -encoding plasmid (without repA for IncN plasmid) harboured by K. pneumoniae isolates; Group IncF, a bla IMP−6 -encoding plasmid (without repA for IncN plasmid) harboured by E. coli isolates; Group Double bla IMP−6 , multiple plasmids with bla IMP−6 harboured by a single isolate; Group Chromosome, chromosomal bla IMP−6 ; Group Non-Typeable, a bla IMP−6 -encoding plasmid of unknown size; Group IMP1, a bla IMP−1 -carrier plasmid.
Isolates classi ed as chromosomal bla IMP carriers were further analysed to identify the location of bla IMP .
In brief, I-CeuI endonuclease-digested PFGE followed by Southern blotting using probes for bla IMP−6 and 16S rRNA genes was performed to con rm the location of the bla IMP gene in three E. coli isolates E138, E300, and E302, as previously described [28].
Whole-Genome Sequencing and Genomic Analysis. Genomic DNA for long-and short-read sequencing was extracted using the DNeasy PowerSoil Kit (Qiagen, Hilden, Germany). Short-read sequencing was conducted on an Illumina HiSeq 3000 sequencer using the KAPA library preparation kit (Kapa Biosystems, Woburn, MA, USA) or on an Illumina MiSeq sequencer using the KAPA HyperPlus Library Preparation Kit (Kapa Biosystems). Long-read sequencing was conducted on a Nanopore GridION sequencer (Oxford Nanopore Technologies, Oxford, UK) using the SQK-LSK109 1D Ligation Sequencing Kit and the EXP-NBD103 Native Barcoding Kit. The reads were assembled and polished using Unicycler [37]. In cases where the complete plasmid sequences could not be constructed, sequences were assembled with CANU (version 1.8) [38] or ye [39] and improved using Pilon [40] or Racon [41]. The PlasmidFinder [42] and ResFinder [43] databases were used to identify antimicrobial resistance genes and plasmid replicon types, respectively. A detailed analysis of the insertion sequence was performed using IS nder [44]. The sequences were annotated with RASTtk [45], and the genomic structures were compared with EasyFig [46]. Plasmids similar to those found in this study were identi ed using BLAST.
Bacterial conjugation assays were performed using the transformants as donors and the sodium azideresistant E. coli strain TUM3456 [47] as a recipient. After mixing overnight cultures of donors and recipients at a 1:10 volumetric ratio, the mixture (10 µL) was incubated on LB agar for 24 h at 37 °C. Transconjugants were selected on LB agar containing cefotaxime (2 µg/mL) and sodium azide (150 µg/mL). The conjugation frequency was calculated from the CFU as the number of transconjugants divided by the number of donors plus transconjugants.
Determination of the Plasmid Copy Number per Host Bacterial Cell. DNA of E. coli isolates E305 and E318, and E. coli transformants with plasmids pE188_IMP6 and pE305_IMP6 single (T188 and T305, respectively) was extracted using the DNA Mini Kit (Qiagen). Using qPCR, the copy numbers of the repA2 gene on plasmids pE305_IMP6 and pE318_IMP6 and the bla IMP−6 gene on pE188_IMP6 were compared with the copy number of the rrsA gene encoding 16S ribosomal RNA on the chromosome. qPCRs were carried out using THUNDERBIRD SYBR qPCR Mix (TOYOBO Life Science, Osaka, Japan) on a LightCycler 96 System (Roche Life Science, Penzberg, Germany). Primers used for this assay are list in Table S8. qPCR analysis was performed using data from repeated experiments (n = 6), and the plasmid copy number per cell was calculated from Ct values using the comparative Ct method [48].
Determination of the Copy Number of bla IMP−6 per Plasmid. Plasmids of E. coli isolates E305 and E318 were extracted using the Plasmid Miniprep Kit (Qiagen). Using qPCR, the copy numbers of the bla IMP−6 gene were compared with those of the repA2 gene on plasmids pE305_IMP6 and pE318_IMP6. qPCRs were carried out using THUNDERBIRD SYBR qPCR Mix on a LightCycler 96 System. Primers used for this assay are listed in Table S8. qPCR analysis was performed using data from repeated experiments (n = 5), and the bla IMP−6 copy number per plasmid was calculated from Ct values using the comparative Ct method.
Transcription of bla IMP−6. E. coli isolates E305 and E318, and E. coli transformants T188 and T305 were incubated in LB broth until the optical density at 600 nm (OD 600 ) reached 0.3-0.4. Total RNA was extracted using the RNeasy Mini Kit (Qiagen). RNA was treated with ReverTra Ace qPCR RT Master Mix with gDNA Remover (TOYOBO Life Science) to remove contaminating DNA and to reverse-transcribe the RNA into cDNA. For quality control, DNase-treated RNA that had not been reverse-transcribed was subjected to a DNA contamination test by qPCR. The rrsA gene encoding 16S ribosomal RNA served as an endogenous control for normalization. qPCRs were carried out using THUNDERBIRD SYBR qPCR Mix on a LightCycler 96 System. Primers used for this assay are listed in Table S8. qPCR analysis was performed using data from repeated experiments (n = 7), and transcript levels were calculated from Ct values using the comparative Ct method. Figure 1 Phylogenetic trees based on XbaI-PFGE and classi cation of blaIMP carriage and antimicrobial resistance patterns. blaIMP carriage of (A) E. coli and (B) K. pneumoniae isolates was classi ed according to the size and replicon type of the blaIMP-carrier plasmids, determined by S1-PFGE and Southern blotting for blaIMP-6 and repA on the IncN plasmid. blaIMP carriage was classi ed designated Contig2 connects Contig1 with Contig3 or Contig4, and Contig5 connects Contig6 with Contig3 or Contig4. (B) Sizes of plasmids pE305_IMP6 and pE318_IMP6. PFGE of S1-digested genomic DNA from E. coli isolates E305 and E318, followed by Southern blotting with a blaIMP-6 probe indicated the size of each plasmid. M, DNA size marker (lambda ladder; Bio-Rad). (C) Hypothetical structure of pE305_IMP6.

Declarations
The colours correspond to the colours of contigs in (A). (D) Genomic comparison of pE318_IMP6 and hypothetical pE305_IMP6single. According to the overlap between contigs of pE305_IMP6, we assembled the hypothetical sequence shown and compared it with the sequence of plasmid pE318_IMP6. Except for the repeating, pE305_IMP6single and pE318_IMP6 were highly similar. Block arrows indicate con rmed or putative open reading frames (ORFs), and their orientations. Arrow size is proportional to the predicted ORF length. The colour code is as follows: red, carbapenem resistance gene; yellow, other antimicrobial resistance gene; light blue, conjugative transfer gene; blue, mobile element; purple, toxin-antitoxin. Putative, hypothetical, or unknown genes are represented as grey arrows. The grey-shaded area indicates regions with high identity between the two sequences. Accession numbers of the plasmids are indicated in brackets. The colours under arrows of pE305_IMP6single correspond to the colours of contigs in (A). (E) Transcript levels of blaIMP-6 in E. coli isolates E305 and E318. qPCR revealed signi cantly higher transcription of blaIMP-6 in isolate E305 than in isolate E318. The bar chart represents the mRNA transcript ratio of blaIMP-6 to the housekeeping gene rrsA, which was used as a reference gene. Bars indicate the mean ± standard deviation, calculated from sextuplet experiments. The p-value was calculated by the Mann-Whitney U test.