Klebsiella quasipneumoniae A708 and K. pneumoniae A324 were isolated in 2014 from the blood specimens of two different patients from a teaching hospital in Guangzhou City, China. K. pneumoniae 13190 and 12208, and K. variicola 13450 were recovered in 2013 from the sputum, sputum and blood specimens of three different patients from a teaching hospital in Hangzhou City, China, respectively. K. pneumoniae 11219 was isolated in 2013 from a sputum specimen of a patient from a teaching hospital in Hefei City, China. K. pneumoniae 19051 was recovered in 2011 from a urine specimen of a patient from a public hospital in Ningbo, China.

Activity of Ambler class A/B/D carbapenemases in bacterial cell extracts was determined by a modified CarbaNP test (Wei et al., 2016). Bacterial antimicrobial susceptibility was tested by BioMérieux VITEK 2 and interpreted as per the 2017 CLSI guidelines (CLSI, 2017).

Conjugal transfer experiments were carried out with the rifampin-resistant Escherichia coli EC600 used as a recipient and the bla_IMP-positive A324 isolate as a donor. Three milliliters of overnight cultures of each of donor and recipient bacteria were mixed together, harvested and resuspended in 80 μL of Brain Heart Infusion (BHI) broth (BD Biosciences). The mixture was spotted on a 1 cm² hydrophilic nylon membrane filter with a 0.45 μm pore size (Millipore) that was placed on BHI agar (BD Biosciences) plate and then incubated for mating at 22°C for 24 h. Bacteria were washed from filter membrane and spotted on Muller-Hinton (MH) agar (BD Biosciences) plates containing 2500 μg/mL rifampin together with 4 μg/mL meropenem for selecting an E. coli transconjugant carrying bla_IMP (pA324-IMP).

To prepare competent cells for electroporation, 200 mL of overnight culture of E. coli TOP10 in Super Optimal Broth (SOB) at an optical density (OD₆₀₀) of 0.4 to 0.6 was washed three times with electroporation buffer (0.5 M mannitol and 10% glycerol) and concentrated into a final volume of 2 mL. One microgram of DNA were mixed with 100 μL of competent cells for electroporation at 25 μF, 200 Ω and 2.5 Kv. The resulting cells were suspended in 500 μL of SOB and an appropriate aliquot was spotted on SOB agar plates containing 4 μg/mL meropenem for selecting of an electroporant carrying bla_IMP (pA324-IMP).

Genomic DNA was isolated from each of the A708, 13190, 11219, 12208, 13450, and 19051 isolates using a Qiagen blood & cell culture DNA maxi kit. Genome sequencing was performed with a sheared DNA library with average size of 15 kb (ranged from 10 to 20 kb) on a PacBio RSII sequencer (Pacific Biosciences, Menlo Park, CA, United States), as well as a paired-end library with an average insert size of 400 bp (ranged from 150 to 600 bp) on a HiSeq sequencer (Illumina, San Diego, CA, United States). The paired-end short Illumina reads were used to correct the long PacBio reads utilizing proovread (Hackl et al., 2014), and then the corrected PacBio reads were assembled de novo utilizing SMARTdenovo¹.

Plasmid DNA was isolated from the A324-IMP-TOP10 electroporant using a Large Construct Kit (Qiagen, Germany) and then sequenced from a mate-pair library with average insert size of 5 kb (ranged from 2 to 10 kb) using a MiSeq sequencer (Illumina, San Diego, CA, United States). Quality control, removing adapters and low quality reads, were performed using Trimmomatic 0.36 (Bolger et al., 2014). The filtered clean reads were then assembled using Newbler 2.6 (Nederbragt, 2014), followed by extraction of the consensus sequence with CLC Genomics Workbench 3.0 (Qiagen Bioinformatics). Gapfiller V1.11 (Boetzer and Pirovano, 2012) was used for gap closure.

Open reading frames and pseudogenes were predicted using RAST 2.0 (Brettin et al., 2015) combined with BLASTP/BLASTN searches (Boratyn et al., 2013) against the UniProtKB/Swiss-Prot database (Boutet et al., 2016) and the RefSeq database (O’Leary et al., 2016). Annotation of resistance genes, mobile elements, and other features was carried out using the online databases including CARD (Jia et al., 2017), ResFinder (Zankari et al., 2012), ISfinder (Siguier et al., 2006), INTEGRALL (Moura et al., 2009), and Tn Number Registry (Roberts et al., 2008). Multiple and pairwise sequence comparisons were performed using MUSCLE 3.8.31 (Edgar, 2004) and BLASTN, respectively. Gene organization diagrams were drawn in Inkscape 0.48.1².

The backbone regions of indicative plasmids were aligned using MUMmer 3.0 (Kurtz et al., 2004). Inference of homologous recombination was performed using ClonalFrameML (Didelot and Wilson, 2015) to remove recombination-associated single-nucleotide polymorphisms (SNPs). A maximum-likelihood tree was constructed from recombination-free SNPs using MEGA7 (Kumar et al., 2016) with a bootstrap iteration of 1000.

The complete nucleotide sequences of plasmids p11219-IMP, p12208-IMP, p13190-VIM, p13450-IMP, p19051-IMP, pA324-IMP, and pA708-IMP, and those of the A708, 12208, 13450, 11219, 13190, and 19051 chromosomes were submitted to GenBank under Accession Nos. MF344561 to MF344567, CP030171 to CP030174, CP026017, and CP022023, respectively.

The seven plasmids sequenced in the present work varied in size from about 238 kb to nearly 345 kb with variation in the number of predicted ORFs from 261 to 379 (Table 1 and Supplementary Figure S1). All these plasmids belonged to the IncHI5 group, because each contained a conserved IncHI5 backbone especially including the IncHI5-type replication gene repHI5B and an additional repFIB-like gene (Liang et al., 2017). Table 1 also lists the features of the previously four sequenced IncHI5 plasmids. Further comparative genomics of all these 11 plasmids revealed that the IncHI5 backbones were interrupted by various accessory modules (defined as acquired DNA regions associated and bordered with mobile elements) inserted at different sites. In addition, while pKOX_R1 (Huang et al., 2013), the first sequenced IncHI5 plasmid, was used previously as the IncHI5 reference (Liang et al., 2017), the newly sequenced pA324-IMP seemed a more appropriate reference in this analysis because it contained the most complete IncHI5 backbone (Supplementary Figure S2).

This large collection of IncHI5 plasmids allowed us to accurately distinguish backbone and accessory modules. This has allowed us to gain a deeper understanding of the evolution and diversification of IncHI5 plasmids.

Pairwise sequence comparison using BLASTN showed that these 11 plasmids had >99% nucleotide identity across >73% of their backbone sequences (Supplementary Table S1). The major IncHI5 backbone genes or gene loci (including repHI5B together with its iterons and repFIB-like for replication, parABC for partition, and tra1 for conjugal transfer) were conserved among all these 11 plasmids. Two conjugal transfer regions tra1 and tra2 were found in IncHI5 plasmids, but some of these plasmids lost tra2, which would impair their self-transferability (Supplementary Table S3). A total of 574 core SNPs (among them 115 were recombination-free) were identified from the backbone regions of these 11 plasmids. A maximum likelihood phylogenetic tree was constructed using these 115 recombination-free SNPs, and accordingly these 11 plasmids could be assigned into three clades I, II, and III (Figure 1).

These 11 plasmids harbored different profiles of accessory modules (Table 1) and different collections of resistance genes (Supplementary Table S2). The accessory modules were further divided into resistance (containing resistance genes) and non-resistance (containing no resistance genes) ones (Table 1). The resistance accessory modules included Tn10 (an IS10-composite transposon carrying class B tetracycline-resistance genes) and two antibiotic resistance islands designated ARI-A and ARI-B. The presence and modular organization of ARI-A and ARI-B islands in the seven newly sequence plasmids and pYNKP001-dfrA (Liang et al., 2017) were validated by a set of PCR amplifications (see Figures 3, 4 for location of PCR primers and expected amplicons) that targeted various key jointing fragments of these islands and their surrounding backbone regions, using the genomic DNA of each corresponding wild-type isolate as template. The non-resistance accessory modules were composed of 12 different insertion sequences (ISs), four distinct IS-related regions, and two cryptic transposons Tn6344 (Supplementary Figure S3) and Tn1722.

At least 10 major events of gene acquisition/loss accounted for modular diversity of these 11 plasmids across their genomes (Figure 2). First, IS903 was inserted at a site upstream of orf444 in five plasmids, additionally resulting in a 42.5-kb deletion (containing the whole tra2 region) in p19051-IMP. Second, ISEc33, ISEc33 and IS102 were inserted at different sites within the tra2 regions of pKpNDM1, p12208-IMP and pKOX_R1, respectively. Third, IS903B in pA708-IMP, ISEc33 in six plasmids, and ISEc33:IS10L in p12208-IMP were inserted within hnhc (HNH endonuclease), splitting it into two separate parts Δhnhc-5′ and Δhnhc-3′. Fourth, in pKpNDM1, Tn10 was inserted at a site within orf648, splitting it into two separate parts Δorf648-5′ and Δorf648-3′. Fifth, compared to the backbone region from orf633 to hokG in pA324-IMP as a prototype structure, various insertions occurred in all the other plasmids: (i) insertion of ISKpn28 upstream of hokG resulted in deletion of a 34.3-kb region in pA708-IMP and that of a 24.8-kb region in pYNKP001-dfrA, respectively; however, upstream-of-hokG insertion of ISKpn8 in pA324-IMP, that of ISKpn28 in six plasmids, and that of ISKpn8–ISKpn28 in p11219-IMP did not cause deletions; and (ii) insertion of ARI-B at a site within xerC2 occurred in the following nine plasmids, which led to a 3.2-kb deletion in seven plasmids, a 14.0-kb deletion in pKpNDM1, and no deletion in pA324-IMP. Sixth, six plasmids had complete tra1 regions; by contrast, four additional plasmids had undergone insertion of an IS4-related region at a site downstream of tivF3 (resulting in a 3.1-kb deletion), and ISkox1 was inserted at a site between tivF3 and orf171 in pA708-IMP (no further deletion occurred). All the above insertions and/or deletions within tra1 and tra2 might impair self-transferability of corresponding plasmids. Seventh, ISKpn21 in five plasmids, ISKox3 in pKP04VIM, and ISKpn21:ISKpn38 in p12208-IMP were inserted at a site downstream of orf294. Eighth, Tn1722 was inserted at a site between orf333 and orf261 in pA324-IMP. Ninth, IS5 or Tn6344 was inserted at a site between luxR and orf183 in all the night plasmids except for p11219-IMP and p13450-IMP, and additionally the Tn6344 insertion resulted in a 1.8-kb deletion in p13450-IMP. Tenth, the ARI-A islands were inserted at a site downstream of orf342 in all the 11 plasmids, which resulted in the 12.6-kb deletion (covering terE-5′ and terABCDZW) in pA708-IMP and the 15.1-kb deletion (covering the complete ter gene cluster) in p11219-IMP; additionally, ISKpn37 was inserted at a site within terF in six plasmids. The above insertions would impair tellurium resistance gene expression.

In conclusion, massive gene acquisition and loss were found in IncHI5 plasmids: a wealth of accessory modules were integrated at various sites of IncHI5 backbones, which resulted in various deletions of surrounding backbone regions and thus considerable diversification of IncHI5 backbones.

The ARI-A islands (Figure 3) were found in all the 11 plasmids analyzed and identified as Tn1696 derivatives. Tn1696, a unit transposon belonging to the Tn21 subgroup of Tn3 family, was generated from insertion of a class 1 integron In4 into the resolution (res) site of a primary backbone structure: IRL (inverted repeat left)–tnpA (transposase)–tnpR (resolvase)–res–mer (mercury resistance locus)–IRR (inverted repeat right) (Partridge et al., 2001).

Being similar to Tn1696, the ARI-A islands from six plasmids had paired terminal 38-bp IRL/IRR and were further bracketed by 5-bp direct repeats (DRs; target site duplication signals for transposition), and thus they were identified as unit transposons designated Tn6338, Tn6401, Tn6400, Tn6384, Tn6382, and Tn6383, respectively. The remaining five ARI-A islands carried only IRLs (interrupted by IS5075 that was a hunter of terminal IRL/IRR of Tn21 subgroup transposons (Partridge and Hall, 2003)) but did not harbor IRRs (due to truncation at 3′-terminal regions of these islands), and thus they were identified as transposon-like structures rather than intact transposons.

In conclusion, the ARI-A islands was inserted at a site downstream of orf342 in all 11 plasmids, and further discriminated as six intact transposons (among them Tn6384, Tn6382, and Tn6383 were novel) and five transposon-like structures. These 11 islands were derived from Tn1696 but differed from it mainly by insertion of distinct integrons or integron-related regions instead of In4 in Tn1696. These integrons could be divided into concise integrons each containing a single gene cassette (GC) array, and complex integrons each harboring one or more variable regions (VRs) in addition to the GC array. These GCs and VRs commonly carried antibiotic resistance genes.

The ARI-B islands (Figure 4) as found in nine plasmids were identified as the derivatives of a prototype arsenic-resistance (ars) unit transposon Tn6535. As observed in the chromosome (Accession No. CP009706) of Hafnia alvei FB1 (Tan et al., 2014), Tn6535 was assembled from integration of an ars region with a Tn3-family core transposition module tnpA–res–tnpR.

The ARI-B island from pA324-IMP or pKpNDM1 was identified as an unit transposon designated Tn6381, with presence of paired terminal 38-bp IRL/IRR and 5-bp DRs. Tn6381 differed from Tn6535 by insertion of a Tn3-family unit transposon Tn6399b at a site within arsB. Tn6399a (it was previously designated Tn6901 because it was 6901 bp in length) was found in plasmid Rts1 from Proteus vulgaris (Murata et al., 2002) and carried several alcohol-metabolism genes (Chen et al., 2013), while interruption of tnpA by IS1618 insertion turned Tn6399a into Tn6399b. Compared to Tn6535 and Tn6381, all the other ARI-B islands contained only IRLs but not IRRs and identified as transposon-like structures. Various types of insertion events occurred within these ARI-B islands, leading to truncation of prototype regions as found in Tn6535 and Tn6381 as well as integration of foreign resistance markers and associated mobile elements.

ARI-B_{pKOX_R1} contained at least 10 resistance loci, including IS26–fosA3–IS26 unit, IS26–bla_{SHV -12}–IS26 unit, a 3-kb Tn6029 remnant, ΔTn6292, In797, ΔTn1548, ΔTn6535, a truncated IS26–catA2–IS26 unit, a residual aacC2–tmrB region, and a mer region (Liang et al., 2017). The resistance markers from ARI-B_p13450-IMP and ARI-B_p19051-IMP differed from pKOX_R1 by replacement of In797 with In823 and In792, respectively; moreover, Tn6339 (containing bla_TEM-1B and bla_CTX-M-3) was inserted at a site between In792 and ΔTn1548 in p19051-IMP. The resistance markers from ARI-B_p11219-IMP were composed of IS26–fosA3–IS26, IS26–bla_{SHV -12}–IS26, a Tn6029 remnant, ΔTn6292, In792, Tn6339, and ΔTn1548. Similarly, IS26–fosA3–IS26, IS26–bla_{SHV -12}–IS26, ΔTn6339, and ΔTn1548 were present in ARI-B_p12208-IMP. The resistance markers from ARI-B_p13190-VIM consisted of IS26–bla_{SHV -12}–IS26, IS26–mph(A)–mrx–mphR(A)–IS6100, chrA region, a In37 remnant and Tn6402, while those from ARI-B_pKP04VIM were composed of IS26–fosA3–IS26, IS26–mph(A)–mrx–mphR(A)–IS6100 and Tn6402. Tn6402 was an IS26-composite transposon (delimited by 4-bp DRs at both ends) derived from Tn1548. Tn6402 differed from Tn1548 by deletion of arsB-3′–ISEc29–msr(E)–mph(E)–repAciN and inversion of the 3′-end copy of IS26.

In conclusion, the ARI-B islands, integrated at a site within xerC2 in nine plasmids, could be further identified as Tn6381 and eight transposon-like structures. These nine ARI-B islands were derived from Tn6535 but differed from it by insertion of various collections of mobile elements and associated resistance genes into the original Tn6381 backbone.

All the transposon-like structures of ARI-A or ARI-B could not be annotated as intact transposons because they lacked paired terminal inverted repeats, and complex transposition and homologous recombination events accounted for assembly and diversification of these transposons and transposon-like structures. The ARI-A or ARI-B islands from different IncHI5 plasmids carried distinct profiles of resistance markers and associated mobile elements, promoting accumulation and spread of antimicrobial resistance among bacterial species.

Compared to the intact transposons Tn6338, Tn6401, Tn6400, Tn6384, Tn6382, and Tn6383 (corresponding to ARI-A), and Tn6381 (corresponding to ARI-B), two kinds of translocation events across ARI-A and ARI-B occurred in each of the following five plasmids (Supplementary Figure S4): (i) exchange of the ΔTn1548–to–IS26 region (finally observed in ARI-B) and the ΔTn1548–to–Tn6535 region (finally observed in ARI-A) in each of p12208-IMP and p11219-IMP; and (ii) movement of the Δorf6–mer region from ARI-A to ARI-B in each of pKOX_R1, p13450-IMP, and p19051-IMP. These two kinds of translocation might be mediated by the common regions ΔTn1548 and IS6100, respectively.

As for the seven plasmids sequenced in this study, the carbapenemase genes bla_IMP-4, bla_IMP-38, and bla_VIM-1 were identified within In823_{p11219-IMP/p12208-IMP/p19051-IMP}, or In1377_pA708-IMP, In1376_pA324-IMP, and In916_p13190-VIM respectively, from the ARI-A islands, while a bla_IMP-4 gene was found within In823_p13450-IMP from the ARI-B island. Of all the integrons identified in these seven plasmids, In1376 and In1377 were novel.

As a representative IncHI5 plasmid, pA324-IMP could be transferred from the A324 isolate into E. coli EC600 and TOP10 through conjugation and electroporation, respectively, generating the A324-IMP-EC600 transconjugant and the A324-IMP-TOP10 electroporant, respectively. This was consistent with the presence of two complete sets of tra1 and tra2 genes in pA324-IMP, making it self-transferable. All the above three strains had class B carbapenemase activity (data not shown), and were resistant to all the cephalosporins and carbapenems tested (Table 2), which were resulted from production of IMP or VIM enzymes in these strains.