Klebsiella sp. isolates were isolated and collected from 2018 to 2020 (Wang et al., 2020) and preserved in the laboratory, from which 56 Klebsiella sp. isolates were selected, including 21 isolates from fresh fecal samples of animals and 35 isolates from human clinical samples (Supplementary Table 1). The pure isolates were revived in Mueller–Hinton broth. The Klebsiella sp. isolates were subjected to antimicrobial susceptibility testing using a broth microdilution kit (BIO-KONT, China) to determine the minimum inhibitory concentrations of ampicillin (AMP, 64–2 μg/ml), cefuroxime (CXM, 64–2 μg/ml), cefazolin (CZO, 16–0.5 μg/ml), ceftriaxone (CRO,8–0.25 μg/ml), cefepime (FEP, 64–2 μg/ml), ampicillin/sulbactam (SAM, 64/32–2/1 μg/ml), piperacillin/tazobactam (TZP, 256/4–8/4 μg/ml), meropenem (MEM, 8–0.25 μg/ml), gentamicin (GEN, 32–1 μg/ml), amikacin (AMK, 128–8 μg/ml), chloramphenicol (CHL, 64–2 μg/ml), levofloxacin (LVX, 16–0.5 μg/ml), cotrimoxazole (SXT, 8/152–0.25/4.75 μg/ml), and tigecycline (TGC, 16–0.5 μg/ml); the broth microdilution method was conducted according to guidelines of the Clinical and Laboratory Standards Institute (CLSI). The susceptibility results were interpreted according to the CLSI 2019 guidelines (Clinical and Laboratory Standards Institute [CLSI], 2019), whereas the breakpoint for polymyxin E and TGC followed the European Committee on Antimicrobial Susceptibility Testing V6.0 (The European Committee on Antimicrobial Susceptibility Testing, 2020). Escherichia coli ATCC25922 was used for quality control.

Genomic DNA extracted from 56 Klebsiella sp. isolates was subjected to whole-genome sequencing using the Oxford Nanopore Technologies MinION platforms (Biomarker Technologies Co., Ltd., Beijing, China; Ashton et al., 2015; Loman et al., 2015). Sequencing reads, including short-read and long-read data, were assembled using Unicycler 0.4.4 via the hybrid strategy (Wick et al., 2017; Li et al., 2018). The plasmid sequences were initially annotated using the RAST server¹ and corrected manually. The plasmid replicon genotype was identified using PlasmidFinder.² IS elements were identified using ISfinder³; comparative analysis was performed, and plasmid maps were generated using Easyfig and BRIG (Alikhan et al., 2011; Sullivan et al., 2011). The phage sequences were screened using PHASTER (Arndt et al., 2016).

The Multiple Locus Sequence Typing (MLST) assay was performed as previously described (Clinical and Laboratory Standards Institute [CLSI], 2012). Briefly, seven K. pneumoniae housekeeping genes (infB, tonB, pgi, gapA, phoE, rpoB, and mdh) were extracted from genomes. Alleles and STs were analyzed using the K. pneumoniae MLST database.⁴

The similarity between the sequenced plasmids and known plasmids was compared using PlasmidFinder (see text footnote 4) (Carattoli et al., 2014; Clausen et al., 2018): The parameter is Max.p-value = 0, Max.distance = 0.04, and Per.Ident ≥ 60%. The incompatible population data sets of similar plasmids were confirmed, and the reporting year and global frequency of similar plasmids were determined according to the PLSDB database. Using the proprietary database of ARGs (card)⁵, the sequencing plasmids were annotated, and the movable elements and other characteristics were annotated according to the results of genome component analysis and functional annotation (NR, GO, UniProt, COG, SwissProt, Pfam, and KEGG public databases, and ISfinder, integral, and Tn number Registry proprietary database). The SNAPGENE software was used to draw the plasmid maps of some plasmids, and pM1026-3Ar.1 was selected as the representative plasmid whose data were used to construct the plasmid fusion pattern map.

The complete sequences of the chromosome and four plasmids have been submitted to GenBank under accession numbers (Supplementary Table 1).

The results of antimicrobial susceptibility testing in this study showed that 80.36% (45/56) of Klebsiella sp. isolates were resistant to ampicillin. Most isolates (45/56) were multiple drug-resistant (MDR) (resistant to three or more antimicrobial classes; Magiorakos et al., 2012). The isolates from the M. pangolin have a high drug resistance rate for cephalosporins, chloramphenicol, levofloxacin, and cotrimoxazole. Only six isolates, M1023-4Ar, N1059-5At, M1026-3Ar, M297-1, S141, and S90-2, from other sources (yak, cow, white-lipped deer, red kangaroo, scarlet breasted parrot, and grouse) were highly resistant to the five drugs mentioned earlier. Among the isolates from human sources, only BS375-3 and S183-1 showed strong resistance to cefuroxime sodium, cefazolin, ceftriaxone, and cefepime (Figure 1). It should be noted that all Klebsiella sp. strains showed strong resistance to beta-lactams but were sensitive to MEM and TGC. The phenomenon of the inconsistency between bacterial drug resistance phenotype and genotype does exist, mainly because the mechanism of bacterial drug resistance is extremely complex. For example, M164-1 from M. pangolin isolate in this study does not carry the resistance gene of sulfonamides but shows resistance to sulfamethoxazole (Supplementary Table 2). Bacterial drug resistance is a very complex process, and the drug resistance pump system may also play an important role (Supplementary Table 1).

According to the results of 16S rRNA and rMLST identification, of the 56 Klebsiella sp. isolates, 29 isolates were identified as K. pneumoniae, among which 27 isolates (27/29, 93.10%) carried plasmids; 9 isolates were identified as Klebsiella quasipneumoniae, among which 8 isolates (8/9, 88.89%) carried plasmids, and 18 isolates were identified as Klebsiella variicola with 5 of these isolates (5/18, 27.78%) carrying plasmids. MLST analysis typing showed there were 33 different ST types, among which 16 ST type isolates carried multiple replicon plasmids, including ST1565 (four isolates), ST1 and ST23 (two isolates each), ST36, ST86, ST101, ST111, ST133, ST231, ST290, ST791, ST1035, ST1662, ST1910, ST2558, and ST3972 (one isolate each) (Supplementary Table 1). In general, the chromosomal DNA of the Klebsiella sp. isolates in this study usually carried at least two of the β-lactam resistance genes bla_DHA, bla_SHV, bla_LEN, and bla_OKP–B. In addition, some drug resistance efflux pump genes, including efflux pump genes belonging to the ABC family, MFS family, SMR family, or MATE family, were also carried on chromosomal DNA. Thus, there was no significant correlation between the multiple-replicon plasmids and ST type.

In the PlasmidFinder database, incompatible group (Inc.) typing is based on the replicon (REP) sequence in plasmid DNA (Carattoli et al., 2014). In this study, 30 of the 69 identified plasmids were single-replicon plasmids, 28 were multiple-replicon plasmids, and 11 were unknown or novel replicon-type plasmids (Supplementary Table 1). Further analysis of the relationship between replicon types and resistance genes showed that the number of resistance genes carried by multiple-replicon plasmids was significantly greater than that carried by non-multiple-replicon plasmids (Figure 2). The proportion of single-replicon plasmids carrying more than five resistance genes was only 26.67%, that of the novel or unknown plasmids was only 9.09%, and that of multiple-replicon plasmids was as high as 75%. The proportion of multiple-replicon plasmids carrying more than 10 resistance genes (n ≥ 10) was 35.71%, and a few plasmids carried more than 25 resistance genes. Multiple replicon plasmids mainly carried β-lactamase genes such as bla_CTX–M–15, bla_CTX–M–14, bla_CTX–M–55, bla_OXA–1, and bla_TEM–1 and aminoglycosidase genes such as aac (6′)–ib cr, aph (6)–id, and aph (3″)–ib. Some plasmids also carried the sulfonamide resistance genes sul1, sul2, and sul3, the tetracycline resistance genes tetA and tetG, members of the fluoroquinolone resistance qnr gene family, and the floR gene for chloramphenicol resistance (Supplementary Table 1). Thus, the ability of Klebsiella sp. isolates to successfully obtain and carry multiple-replicon plasmids may be an important survival strategy for coping with the environmental pressure of antibiotics.

To evaluate the relationship between Klebsiella sp. and multiple-replicon plasmids, the plasmids and drug resistance genes carried by K. pneumoniae, K. variicola, and K. quasipneumoniae were compared and analyzed. Overall, it was found that multiple-replicon plasmids were widely distributed and were mostly carried by K. pneumoniae and were rarely carried by K. quasipneumoniae and K. variicola. The phylogenetic tree of K. pneumoniae constructed using 16S rRNA showed that the K. pneumoniae isolates were divided into five branches. The MDR K. pneumoniae isolates carrying multiple-replicon plasmids were on the same evolutionary branch, suggesting that these isolates showed close homology (Figure 3). It is worth noting that most of the isolates carrying multiple-replicon plasmids came from wild animals rather than from humans (Figure 3), suggesting that wild animals may be important reservoir hosts for MDR bacteria. Therefore, K. pneumoniae is considered to be an important host of multiple-replicon plasmids, and wild animals are also important hosts of multidrug-resistant K. pneumoniae.

In this study, 28 multiple-replicon plasmids were further analyzed, including eight IncFIB/IncFII/(IncQ1), five IncFIB/IncHI1B/(IncQ1), four IncFII/IncR, four IncHI1B/RepB, two IncFII/IncQ1, one IncFIA/IncFIB/IncFII/IncFII/IncFII, one IncFIA/IncFII, one IncFIA/IncR, one IncFII/IncHI1B/IncR/IncR/RepB, and one IncN/IncN plasmid (Supplementary Table 1). With the exception of the IncFII/IncR and IncFIA/IncFII plasmids, all other multiple-replicon plasmids listed earlier were found to carry many resistance genes. When the number of resistance genes on plasmids increased, the base number of plasmids also increased significantly, but the total G + C content of the plasmids did not increase significantly. As mentioned previously, these plasmids mainly come from natural hosts consisting of a variety of ST-type K. pneumoniae strains (Supplementary Table 1).

The plasmid skeleton is usually composed of a replication regulatory region, distribution system, binding transfer system, and plasmid maintenance region (Novick, 1987; Couturier et al., 1988). Of the 28 multiple-replicon plasmids, five were selected for BLAST analysis, and the five plasmids with high homology (homology ≥ 93%) and the multiple-replicon plasmids with a high abundance of resistance genes were screened for the further analysis of plasmid recombination (Table 1). The plasmid skeleton genes rep, sop/par, tra/trb, and the antibiotic resistance and mobile elements genes were used for comparative structural analysis. For the multiple-replicon plasmid pM1026-3Ar.1 [belonging to IncFIB/IncFII/(IncQ1)], the upstream and downstream gene structure of the repB gene was IS609 insQ-sopB-sopA-repB-intI-parD-ybdN-parB-IS110 tnp, but the upstream and downstream gene structure of the repA gene of the IncFII replicon was tn5393 tnpA-repA-repA2-pld-IS1 insB. The incomplete repA gene of the IncQ1 replicon occurs within a complex structure composed of 18 resistance genes and 5 mobile elements, and the upstream and downstream gene structure is repC-repA-IS431mec tnp-Tn21 tnpM-intI1.

The upstream and downstream structure of the repB gene of the pS174-1.3 plasmid [IncFIB/IncHI1B/(IncQ1)] replicon is intM-int-repB-sopA-sopB, the upstream and downstream structure of repA gene of IncHI1B replicon is IS431mec tnp-other-repA-yadA-other-Tn903 tnp, and the upstream and downstream structure of repA gene of IncQ1 replicon is repC-repA-IS431mec. The tnp-Tn3 tnpR is also surrounded by a complex structure composed of four resistance genes and two transposase genes. The upstream and downstream gene structure of the repA1 gene of the pBS361-1 IncFII plasmid (belonging to IncFII/IncR type) replicon is IS431mec tnp-yedK-repA1-pld-aer-Tn4653 tnpR-Tn1721 tnpA. The upstream and downstream gene structure of the repB gene of the tnpA and IncR replicons is parM-parA-other-repB-other-resD, and the four IncFII/IncR plasmids show high similarity (≥99%), suggesting that they may be different copies of the same plasmid, carrying two resistance genes, mexT and tetG (Supplementary Table 1).

The upstream and downstream gene structure of the repA gene of the pS161-2.2 IncHI1B plasmid (belonging to the IncHI1B/repB type) replicon is Tn1721 tnpA-other-other-repA-other-other-Tn903 tnp, and the upstream and downstream gene structure of the repA gene of the repB replicon is sopB-sopA-repA-int-intM. This type of plasmid also has a strong ability to carry resistance genes and has two distribution systems, sopA-sopB/parA-parB. The physical distance between the repA gene of pM297-1.2 IncFII plasmid and the repA gene of IncQ1 is very short (4,038 bp), represented by the structure Tn903 tnp-other-repA(IncFII)-pld-aer-IS431mec tnp-other-repA(IncQ1)-repC. Notably, although the repA gene of IncQ1 is not included in the complex structure of the resistance gene and mobile element, it is always followed by the sulfonamide resistance gene sul2 (Figure 4).

The pM279-1.2, pM1026-3Ar.1, and pS174-1.3 plasmids all carried the IncQ1 replicon, and the numbers of resistance genes were 17, 27, and 8, respectively. The ability of these plasmids to carry genes encoding for resistance to β-lactams, aminoglycosides, and sulfonamides was stronger (Table 1), suggesting that the resistance genes-carrying ability may be related to the IncQ1 plasmid. In addition, the pM1026-3Ar.1, pBS361-1, and pM297-1.2 plasmids all exhibited the structure traABCDEFGHIGKLMNOPQRSTUVWXY/trbABCDEFGHIG KLMN-finO, and pM297-1.2 even contained two complete conjugate transfer systems. Mobile elements (insert sequences, transposons, or integrase genes) were found in at least one upstream and downstream of the rep gene in the five plasmids, possibly because mobile elements can act as key fusion sites by mediating the fusion of plasmid fragments to form a complex structure.

The genomic DNA of the multiple-replicon pM1026-3Ar.1 plasmid contains three replicons IncFIB, IncFII, and IncQ1. An analysis based on BLAST and the PLSDB database showed that three single-replicon plasmids were highly homologous with the pM1026-3Ar.1 replicon (homologous region ≥97%), among which the IncFIB replicon region (32,883–96,121 bp) presented high homology with plasmid p203 of K. pneumoniae obtained from human samples in 2018 (GenBank accession number: NZ)_ Cp021166.1, query coverage = 40%); the IncFII replicon region (1,421–94,997 bp) was highly homologous with the pGSU10-3-2 plasmid of K. pneumoniae producing the KPC-2 enzyme isolated from a wastewater treatment plant in Japan in 2018 (GenBank accession number: NZ)_ Ap018673.1, query coverage = 31%); and the IncQ1 replicon region (451–5,462 bp) showed high homology with the pSRC15 plasmid of Salmonella enterica isolated in 2012 (GenBank accession number: NC)_ 013104.1, query coverage = 1%). These results suggested that the multiple-replicon pM1026-3Ar.1 plasmid may have been formed by the fusion of these three plasmids. According to the structural information of these plasmids, the possible plasmid fusion pattern was drawn (Figure 5).

The full length of plasmid pM1026-3Ar.1 was 262,519 bp. It was speculated that the p203-like plasmid provided ∼63,239-bp gene fragment as the basic plasmid structure, whereas the pGSU10-3-2-like plasmid provided ∼40,295-bp gene fragment of the conjugated transfer region, and the p203- and pGSU10-3-2-like plasmids provided a large number of resistance genes and transportable genes for the pM1026-3Ar.1 mobile element (Figure 5A). The pSRC15-like plasmid provided ∼5,012-bp gene fragment, including ARGs such as aph(6)-Id, aph(3″)-Ib, sul2, bla_OXA–1, and aac(6′)-Ib-cr, and a small number of efflux pump genes (Figure 5B). Therefore, under antibiotics pressure, a transformation from multiple single-replicon plasmids to multiple-replicon fusion plasmids is necessary to reduce the cost of survival.