Klebsiella variicola strain JNQH579 was recovered from a sputum sample of a 64-year male patient in intensive care units (ICUs) at a tertiary hospital in Jinan City, Shandong Province, in March 2021, who had been hospitalized for 17days due to severe pneumoniae and renal failure. The patient had received tigecycline treatment for 7days at a dosage of 100mg (IV) q12h before the isolation of the strain. Based on the antibiotic susceptibility testing results, antibiotic therapy was switched to aztreonam 0.5g (IV) q8h in combination with tigecycline. The respiratory symptoms improved after antibiotic treatment and the patient continued to be hospitalized for 3months due to cardiovascular and renal disease. Klebsiella quasipneumoniae strain JNQH473 was recovered from a urine sample of a 1-month-old infant with sepsis at a tertiary hospital in Xiamen City, Fujian Province, in August 2019, who was hospitalized at the department of pediatrics for 7days. During her hospitalization, the neonate received multiple antimicrobial treatments, including cefotaxime, vancomycin, and ceftazidime. The patient was fully recovered after antimicrobial and supportive treatment and was discharged on hospital day 13. Klebsiella michiganensis strain JNQH491 was recovered from a blood culture of a 48-year male patient at the department of oncology, who had received chemotherapy due to nasopharyngeal carcinoma in the same hospital as JNQH473. The patient was discharged home after a cycle of chemotherapy on hospital day 11. The overall strain features of the three strains are listed in Table 1. All the patients reported no recent travel abroad.

Antibiotic susceptibility testing (AST) was performed using the VITEK 2 (bioMérieux, Nürtingen, Germany) system. Minimum inhibitory concentrations (MICs) for tigecycline were performed by broth microdilution according to the Clinical and Laboratory Standards Institute (CLSI) guidelines. ATCC 25922 (Escherichia coli) and ATCC 27853 (Pseudomonas aeruginosa) were used as quality control strains for susceptibility testing. All the tests were performed in duplicate in different days. The breakpoints were interpreted according to CLSI guidelines except for tigecycline, of which the EUCAST epidemiological cutoff value >2μg/ml (for K. pneumoniae) was used.¹

The combination Oxford Nanopore (MinION system) and Illumina sequencing (NovaSeq) were used to achieve a high-quality genome assembly. First, we derived fastq read sequences from MinION raw electric signal fast5 files using guppy 3.2.2 with the high accuracy flip-flop algorithm. Adapters were trimmed out with Porechop.² Low-quality reads were filtered out using trimmomatic 0.38 (Bolger et al., 2014). The filtered nanopore reads were de novo assembled with Flye 2.8.3 (Kolmogorov et al., 2019). Then, the obtained assemblies were polished using Nanopore reads by Racon 1.4.3 (1–4 iterations; Vaser et al., 2017). Next, the polished sequences were additionally corrected using Illumina reads by Pilon 1.23 until no changes occur (Walker et al., 2014). The whole-genome sequences were annotated by Prokka (Seemann, 2014) and RAST (Brettin et al., 2015), followed by manually curations.

The in silico multilocus sequence typing (MLST) was carried out using MLST v. 2.19.0,³ and antibiotic resistance/plasmid replicon gene detections were carried out using ABRicate v. 0.9.9⁴ using CARD (Jia et al., 2017) and PlasmidFinder (Carattoli et al., 2014) database, respectively. The intrinsic variants of oqxAB, chromosomal ampH and fosA, which confer resistance to quinolones, β-lactams and fosfomycin, respectively, in Enterobacterales rather than Klebsiella species, are therefore not reported (Henderson et al., 1997; Lam et al., 2021). Kleborate v. 1.0.0 (Lam et al., 2021) was used for Klebsiella K locus and O locus typing. The comparative analysis of TmexCD2-ToprJ2 harboring plasmids was done by BLASTn and illustrated using CGView Server (Grant and Stothard, 2008). Easyfig (Sullivan et al., 2011) was used to visualize the genetic context comparisons. Selected plasmids were compared with Mauve 2.3.1 (Darling et al., 2010), followed by visualization using genoPlotR (Guy et al., 2010). Plasmid distance trees were generated using Mashtree (Katz et al., 2019). Alignment between three TmexCD-ToprJ variants was done by Clustal Omega (Madeira et al., 2019). ISFinder⁵ was used to identify ISs. Tn number was identified using Tn Number Registry (Roberts et al., 2008). In order to examine the distribution and relationships of TmexCD2-ToprJ2 and its variants (TmexCD1-ToprJ1, TmexCD2-ToprJ2), plasmid sequences were downloaded from NCBI⁶ and compared.

Conjugation experiments were performed by plate mating using E. coli J53 (AziR) as the recipient as described in our previous study (Hao et al., 2021). Briefly, overnight cultures of the donor strains and the recipient strain E. coli J53 were mixed (1:1) and applied onto an LB agar plate, followed by overnight culture at 37°C. Transconjugants were selected on Mueller-Hinton (MH) agar containing sodium azide (200μg/ml) and tigecycline (0.5μg/ml). The presence of TmexCD2-ToprJ2 and IncHI1B replicons, as well as bla_NDM-1 and sul1 resistance genes in transconjugants, was confirmed by PCR (Supplementary Table S1). Conjugation frequency was calculated by dividing the number of transconjugants by the number of recipient cells. AST of the E. coli J53 transconjugants was performed as described above.

The MICs for JNQH473, JNQH491, and JNQH579 were shown in Table 2. The MICs of tigecycline in JNQH473 and JNQH579 were both 4μg/ml. However, JNQH491 was susceptible to tigecycline with a MIC value of 1μg/ml. These strains were resistant to almost all tested β-lactam antibiotics except that JNQH491 and JNQH579 were susceptible to aztreonam. All the strains were susceptible to amikacin with an MIC value of less than 2μg/ml.

Strain JNQH473 was classified as sequence type 571 (ST571) based on the in silico MLST, and it belonged to KL64 capsule and O5 lipopolysaccharide serotypes (Table 1). It harbored a 5.39-Mb chromosome and four plasmids, designated pJNQH473-1 (229.2-Kb), pJNQH473-2 (70.9-Kb), pJNQH473-3 (297.9-Kb), and pJNQH473-4 (46.1-Kb), respectively (Table 3). Two metallo-β-lactamase (bla_MBL) genes, bla_IMP-8 and bla_NDM-5, were located on incompatible FIB type plasmid pJNQH473-1 and IncX3 type plasmid pJNQH473-4, respectively. bla_IMP-8 was located in class 1 integron In655 carrying the gene cassette of bla_IMP-8-aacA4 (Jiang et al., 2017), and bla_NDM-5 was located in an ΔTn125-like region containing ble_MBL and IS26 genes downstream and IS3000, IS30, and IS5 family transposase genes upstream. The TmexCD2-ToprJ2 gene cluster was located on an IncHI1B type plasmid, pJNQH473-3. In addition to TmexCD2-ToprJ2 genes, nine antimicrobial resistance genes were found on the same plasmid, including two copies of sulfonamide resistance gene sul1 and single copy of aminoglycoside resistance gene aac(3)-IId and aadA16, quinolone-resistant gene qnrS1, macrolides resistance gene mphA, trimethoprim resistance gene dfrA27, the β-lactamase gene bla_TEM-1D, and rifampicin resistance ribosyltransferase gene arr-3.

Strain JNQH491 belonged to ST109 type (Table 1). The K and O locus could not be classified according to the currently available K and O loci database. The genome contained a 6.04-Mb chromosome and two plasmids, pJNQH491-1 (205.5-Kb) and pJNQH491-2 (307.4-Kb; Table 3). The TmexCD2-ToprJ2 gene cluster was located on pJNQH491-2, which is an IncHI1B type plasmid, co-harboring multiple resistant genes, including the bla_MBL and bla_NDM-1 (Figure 1). The pJNQH491-1 plasmid had an IncFIB replicon and did not carry any known resistance genes.

Strain JNQH579 belonged to ST2013 and harbored an OL103 serotype (Table 1). The genome included a 5.58-Mb chromosome and two plasmids, designated pJNQH579-1 (197.4-Kb) and pJNQH579-2 (369.3-Kb), respectively (Table 3). The TmexCD2-ToprJ2 gene cluster was carried by a 34,827bp mobile region, which was located on pJNQH579-2 plasmid (Figure 2). pJNQH579-2 is also an IncHI1B type plasmid, co-harboring multiple resistance genes [aac(6′)-Ib, aadA16, qnrS1, mphE, msrE, catB3, catII.2, arr-3, sul1, dfrA27, bla_OXA-1, and bla_NDM-1; Table 3]. Notably, the bla_NDM-1 carbapenem-resistant gene was co-existed with TmexCD2-ToprJ2 in pJNQH579-2 plasmid. bla_NDM-1 was found within a truncated transposon Tn125 (Figure 3), with the structure of “ΔISAba125-bla_NDM-1-ble_MBL-tat-dvt-groESL- tnpAISCR21” (Poirel et al., 2012). The pJNQH579-1 plasmid had an IncFIB replicon but did not carry any known resistance genes.

The TmexCD2-ToprJ2 cluster in the three strains had almost 100% nucleotide identities to the cluster of the originally reported TmexCD2-ToprJ2 (Wang et al., 2021a). Plasmid sequence comparison demonstrated that pJNQH473-3 and pJNQH491-2 had highly conserved plasmid synteny and structure, with over 99.9% nucleotide identities (Figure 3). pJNQH491-2 had a blast query coverage of 74% and over 99% nucleotide identities with pJNQH579-2. Further plasmid BLAST query against GenBank database showed that pJNQH473-3 and pJNQH491-2 were closely related to plasmid pKOX-R1 (NC018107), isolated from Taiwan in K. michiganensis E718 (Figure 3). In comparison with pKOX-R1, a 16,846bp region, containing the TmexCD2-ToprJ2, was inserted into the umuC gene of pJNQH473-3, generated a 3bp (CAT) direct repeats (Figure 2). The three pJNQH plasmids contain the same TmexCD2-ToprJ2 structure of “hp1-hp2-tnpA-tnfxB2-IS881-TmexCD2-ToprJ2,” with an average GC content of 42.3%. pJNQH579-2 was found to be closely related to plasmid pC2315-2-NDM (CP039829). The putative “hp1-hp2-tnpA-tnfxB2-IS881-TmexCD2-ToprJ2” transposon was surrounded by XerC, XerD, and additional hypothetical genes. This region was also inserted into the umuC gene therefore constituted a putative larger transposon unit of 34,827bp (Figure 2).

TmexCD2-ToprJ2 cluster shares a high similarity to TmexCD1-ToprJ1, of which the variants of TmexC, TmexD, and ToprJ genes had 98.02, 96.75, and 99.93% nucleotide identities and 97.67, 97.61, and 99.79% amino acid identities between each other. Compared with TmexCD3-ToprJ3, the variants of TmexC, TmexD, and ToprJ genes shared 94, 96.72, and 99.86% nucleotide identities and 97.12, 98.08, and 99.79% amino acid identities. ToprJ gene shares the highest similarity at the amino acid level among the three variants, of which ToprJ2 differs from ToprJ1 and ToprJ3 by a single amino acid substitution (Ala47Thr; Supplementary Figure S1).

The analysis of plasmid database revealed that a total of 43 plasmids carried TMexCD1-ToprJ1 (n=23), TmexCD2-ToprJ2 (n=8), and TmexCD3-ToprJ3 (n=12) gene clusters (with 100% gene coverage and over 99.97% identity), which were distributed mainly in Enterobacteriaceae and P. aeruginosa (Figure 4). TmexCD1-ToprJ1 and TmexCD2-ToprJ2 were mostly identified in K. pneumoniae, whereas TmexCD3-ToprJ3 was more frequently found in P. aeruginosa. In addition to Klebsiella, the TmexCD2-ToprJ2 harboring plasmids were also found in R. ornithinolytica (n=1), Citrobacter freundii (n=1). Most TmexCD2-toprJ2 gene clusters were associated with IncFII (n=3), IncHI1B (n=3), and IncQ (n=2) plasmids. Of note, six TmexCD2-ToprJ2 harboring plasmids also co-harbored carbapenem-resistant genes, including four bla_NDM-1 and two bla_KPC-1 genes.

The TmexCD2-ToprJ2 harboring IncHI1B plasmids were successfully transferred into E. coli J53 from three JNQH strains. Further PCRs confirmed that other plasmids in the donor strains were not co-transferred along with the TmexCD2-ToprJ2 harboring IncHI1B plasmids into the recipient E. coli J53 strain (Supplementary Figure S2). The conjugation frequency was 10⁻⁶, 10⁻⁶, and 10⁻⁵ for JNQH473, JNQH491, and JNQH579 per recipient cell, respectively. The MICs for the E.coli J53 transconjugants are shown in Table 2. The MICs of tigecycline against E. coli J53 transconjugants were 1μg/ml, which were 8-fold higher than that of the E. coli J53 itself. In addition, the transfer of bla_NDM-1 together with the TmexCD2-ToprJ2 genes for JNQH491 and JNQH579 in the E. coli J53 transconjugants conferred resistance to all tested β-lactams (ampicillin, amoxicillin-clavulanate, piperacillin/tazobactam, cefoxitin, ceftriaxone, cefepime, imipenem, and ertapenem) except for aztreonam. All transconjugants exhibited resistance to quinolones (ciprofloxacin and levofloxacin, ≥4μg/ml), which was likely due to the combinational effects of both quinolone resistance determinant qnrS1 and the TmexCD2-ToprJ2. Further, E.coli J53 transconjugants for JNQH491 and JNQH579 exhibited resistance to gentamycin, while it was susceptible for JNQH473. Aac(3)-IId is likely the main source of the disparity on the basis that aac(3)-IId was co-harbored with TmexCD2-ToprJ2 in the same plasmids for JNQH473 and JNQH491 strains, while it was absent on the pJNQH579-2 plasmid.