The GenBank Genome database (Benson et al., 2018) collect all the plasmids belonging to K. pneumoniae.⁵ A total of 4,451 plasmids (without duplicates) of K. pneumoniae (Supplementary Table S1) were downloaded on April 26th, 2022. Files in FASTA DNA format of the 4,451 plasmids were downloaded in batches into our Linux-based server.

The β-lactamase genes of the plasmids of K. pneumoniae were identified applying the ResFinder software, standalone version 4.1 (Bortolaia et al., 2020), with the minimum coverage of 60%, minimum identity of 90%, and species of “Klebsiella.” The term “bla_NDM” was used to search in the “Resistance gene” list of the ResFinder results in order to judge the bla_NDM-harboring plasmids of K. pneumoniae and identify the variants of the bla_NDM genes. For some bla_NDM-harboring plasmids, the variants of bla_NDM were not determined by the ResFinder software; instead, they were submitted to the CARD database⁶ (Alcock et al., 2020) and the Beta-lactamase database (BLDB; Naas et al., 2017) for further analysis.

Replicon typing of the bla_NDM-harboring plasmids was executed via the PlasmidFinder software (Carattoli and Hasman, 2020). Then, selecting the database “Enterobacteriales,” the FASTA-formatted DNA files were analyzed and classified in batches by using the PlasmidFinder tool version 2.0.1, with a minimum coverage cut-off of 60% and minimum identity cut-off of 95%. The database version was updated on November 29th, 2021.

The files of the bla_NDM-harboring plasmids of K. pneumoniae, in GenBank format, were downloaded in batches using two Bioperl modules (Bio::SeqIO and Bio::DB::GenBank). Plasmid files containing protein sequences were compiled from the plasmid files in GenBank format through the Bioperl/Bio::SeqIO module. Phylogenetic cladogram based on the presence/absence of orthologous gene families of all the bla_NDM-harboring plasmids of K. pneumoniae were constructed. First, a binary gene presence/absence matrix was built using OrthoFinder software (Emms and Kelly, 2019), and subsequently a hierarchical cluster result was generated by PAST3 (Hammer et al., 2001) and eventually displayed by iTOL (Letunic and Bork, 2016).

Information about geographic location of bla_NDM-harboring plasmids and its host strains were extracted from the files of the bla_NDM-harboring plasmids in GenBank format. Table containing the correspondence between strains and plasmids of K. pneumoniae were downloaded from the GenBank.⁷ The bla_NDM-harboring plasmid-matched host K. pneumoniae strains were collected, and their DNA FASTA sequences were downloaded in batch using the Bioperl. The MLST software (Larsen et al., 2012) version 2.0.9 was downloaded from the website⁸ and installed on the Linux platform. The genomes of K. pneumoniae strains were analyzed in batch using MLST software.

Files in GenBank format of the bla_NDM-harboring plasmids in K. pneumoniae were analyzed in batches using oriTfinder software (local version; Li et al., 2018) to identify the presence/absence of oriTs, relaxase-coding genes, T4CP-coding genes, and gene clusters for T4SS. Furthermore, the types of oriTs, relaxases, T4CPs, and T4SSs toward the plasmids were determined based on the exhibition of the oriTDB database⁹ (Li et al., 2018). In addition, the types of gene clusters for T4SS were classified based on the SecReT4 database¹⁰ (Bi et al., 2013).

The bipartite network was constructed based on all the bla_NDM-harboring plasmids of K. pneumoniae using the AccNet software using default parameters (Lanza et al., 2017). The obtained network files including nodes, edges and clusters were then imported into the Cytoscape software (Shannon et al., 2003) for visualization. We displayed the relative genomic content of each plasmid by making the diameter of each node proportional to its degree.

Based on the results analyzed by ResFinder, 171 (3.84%) bla_NDM-harboring plasmids (Supplementary Table S2) were identified from 4,451 plasmids of K. pneumoniae, which were downloaded from the GenBank Genome database. Among the 171 bla_NDM-harboring plasmids of K. pneumoniae, nine different variants of bla_NDM were identified (Figure 1A). Among the nine variants of bla_NDM, bla_NDM-1 was found to be the predominant variant, accounting for 73.68% (126 bla_NDM-1-harboring plasmids), followed by bla_NDM-5, accounting for 16.37% (28 bla_NDM-5-harboring plasmids) (Figure 1A).

Replicon typing of the 171 bla_NDM-harboring plasmids of K. pneumoniae was executed using PlasmidFinder. Of the 171 plasmids, 154 were successfully identified with their replicon types, including 103 single-replicon plasmids and 51 multi-replicon plasmids (44 plasmids with two replicons, three plasmids with three replicons, and four plasmids with four replicons; Figure 1B; Supplementary Figure S1). For the 103 single-replicon plasmids harboring bla_NDM in K. pneumoniae, the TOP5 prevalent replicons (in descending order) were IncX3 (29 plasmids), IncC (15 plasmids), IncFIB(pQil) (11 plasmids), IncFII (11 plasmids), and IncFII(Yp) (11 plasmids). Of the 44 bla_NDM-harboring plasmids with two replicons, 25 contained replicons IncFIB(pNDM-Mar) and IncHI1B(pNDM-MAR), which were the most prevalent two-replicon plasmids harboring bla_NDM in K. pneumoniae (Figure 1B; Supplementary Figure S1).

In summary, 21 of the 171 bla_NDM-harboring plasmids of K. pneumoniae were found to carry the replicon of IncC, accounting for 12.28% of all the bla_NDM-harboring plasmids of K. pneumoniae in this study (Figure 1B; Supplementary Figure S1). Notably, 91 of the 171 bla_NDM-harboring plasmids in our study were found to be the IncF plasmids, including IncFI and IncFII plasmids, accounting for 53.22% of all the bla_NDM-harboring plasmids of K. pneumoniae in this study (Figure 1B; Supplementary Figure S1).

To get the comprehensive overview of bla_NDM-harboring plasmids in K. pneumoniae, we created a phylogenetic cladogram of the 171 bla_NDM-harboring plasmids (Figure 2). Based on phylogenetic patterns of the 171 plasmids, combined with the replicon types, conjugative transfer regions, and genome sizes of the bla_NDM-harboring plasmids, most of the 171 bla_NDM-harboring plasmids were clustered into 10 main clades (clades I–X), representing 10 plasmid patterns carrying bla_NDM genes in K. pneumoniae (Table 1).

Clade I: A total of 29 IncX3 plasmids were found in the clade I cluster, mainly bla_NDM-1 and bla_NDM-5 (Figure 2). For the 29 IncX3 plasmids harboring bla_NDM, their genome sizes varied from 45.1 to 159.3 kb (25th percentile = 46.2 kb; 75th percentile = 57.3 kb), with a median size of 53.1 kb (Supplementary Figure S2). For the conjugative transfer regions, all the plasmids belonging to clade I were found to carry genes encoding relaxases of the MOB_P family characterized by the domain “Relaxase (Pfam: PF03432),” T4CPs of the VirD4/TraG subfamily characterized by the domain “T4SS-DNA_transf (Pfam: PF02534),” and VirB-like T4SS gene clusters (Figure 2; Supplementary Figure S3). Members of clade I were mainly geographically distributed in China (Figure 3; Table 1; Supplementary Table S3). No predominant ST types of isolates were found in the plasmids harboring bla_NDM in K. pneumoniae (Table 1; Supplementary Table S3).

Clade II: Ten bla_NDM-positive IncN plasmids were clustered into clade II, mainly carrying bla_NDM-1 (Figure 2). The genome sizes of the 10 bla_NDM-harboring IncN plasmids varied from 38.4 to 63.1 kb (25th percentile = 47.2 kb; 75th percentile = 59.8 kb), with a median size of 52.0 kb (Supplementary Figure S1). Almost all the IncN plasmids carried the NW-type oriTs and genes encoding relaxases of MOB_F family characterized by the domain “TrwC (PF08751).” All the 10 bla_NDM-positive IncN plasmids carried the genes encoding T4CPs of the TrwB/TraD subfamily characterized by the domain “TrwB_AAD_bind (PF10412)” and Trw-like T4SS gene clusters (Figure 2; Supplementary Figure S4). The members of clade II were mainly geographically distributed in China (Figure 3; Table 1; Supplementary Table S3). The 10 bla_NDM-positive IncN plasmids were distributed in seven ST types of K. pneumoniae strains, and four plasmids were distributed in K. pneumoniae ST15 (Table 1; Supplementary Table S3).

Clade III: Eleven bla_NDM-1-positive IncF plasmids with the IncFIB(pQil) replicon were grouped into clade III, and most were 54-kb plasmids (Figure 2). Moreover, no conjugative transfer regions were identified in the 11 plasmids of clade III, indicating that the 11 plasmids should be non-transferable. Plasmids belonging to clade III were mainly geographically distributed in Italy and United States (Figure 3; Table 1; Supplementary Table S3). All the members of clade III were harbored by the strains of K. pneumoniae ST147 (Table 1; Supplementary Table S3).

Clade IV: Five bla_NDM-positive untypeable plasmids were clustered into clade IV, involving three bla_NDM-1-positive plasmids and two bla_NDM-4-positive plasmids (Figure 2). These five untyped plasmids, with lengths ranging from 75.3 to 86.0 kb, all carried the genes encoding relaxases of MOB_F family, genes encoding T4CPs of TrwB/TraD subfamily, and Tra_F-like T4SS gene clusters (Figure 2; Supplementary Figure S5). For the five plasmids, three were found in Viet Nam, one was found in China, and one was found in South Korea (Figure 3; Table 1; Supplementary Table S3). The STs of K. pneumoniae host strains containing the clade IV plasmids were distributed into ST395 and ST16 (Table 1; Supplementary Table S3).

Clade V: Thirteen plasmids with the IncFII replicon, mainly carrying bla_NDM-5, were classified into the clade V (Figure 2). For the 13 IncFII plasmids harboring bla_NDM, genome sizes varied from 75.3 to 140.6 kb (25th percentile = 88.8 kb; 75th percentile = 101.4 kb), with a median size of 96.2 kb (Supplementary Figure S1). They all carried the F-type oriTs and Tra_F-like T4SS gene clusters (Figure 2; Supplementary Figure S6). Most of the plasmids clustered into clade V were found to carry genes encoding relaxases of the MOB_F family and genes encoding T4CPs of the TrwB/TraD subfamily (Figure 2). The members of clade V were widely distributed in India, Southeast Asia, North America, East Asia, and Europe, with the highest prevalence in India (Figure 3; Table 1; Supplementary Table S3). The STs of K. pneumoniae host strains containing all Clade V plasmids were mainly distributed in ST16, ST147, and ST2096 (Table 1; Supplementary Table S3).

Clade VI: Four bla_NDM-1-positive plasmids, including one IncM2 plasmid and three untyped plasmids, were classified into a small cluster named clade VI in our study (Figure 2). These four plasmids, with lengths ranging from 75.6 to 100.2 kb, all carried the L/M-type oriTs, genes encoding relaxases of MOB_P family, genes encoding T4CPs of TrwB/TraD subfamily, and Tra_I-like T4SS gene clusters (Figure 2; Supplementary Figure S7). The four plasmids were sporadically discovered in Canada, Germany, United Kingdom, and China (Figure 3; Table 1; Supplementary Table S3). No prevalent STs of K. pneumoniae host strains containing all clade VI plasmids were found (Table 1; Supplementary Table S3).

Clade VII: Four bla_NDM-1-positive plasmids with the IncFII(pKPX1) replicon were classified into clade VII (Figure 2). The genome sizes of the four IncFII(pKPX1) plasmids varied from 96.8 to 250.4 kb. Similar to the conjugative transfer regions of plasmids belonging to clade VI, they all carried the L/M-type oriTs, genes encoding relaxases of the MOB_P family, genes encoding T4CPs of the TrwB/TraD subfamily, and Tra_I-like T4SS gene clusters (Figure 2; Supplementary Figure S8). The four plasmids were sporadically discovered in United States, India, Switzerland, and China (Figure 3; Table 1; Supplementary Table S3). No obvious common STs of strains were found (Table 1; Supplementary Table S3).

Clade VIII: Eighteen IncF plasmids, mainly carrying bla_NDM-1, were grouped into the clade VIII cluster (Figure 2). Most of the IncF plasmids contained the IncFII(Yp) or IncFII(K) replicon in their genomes. For the 18 bla_NDM-harboring plasmids of clade VIII, genome sizes varied from 94.4 to 316.2 kb (25th percentile = 106.8 kb, 75th percentile = 150.1 kb), with a median size of 110.6 kb (Supplementary Figure S1). Most of the plasmids of clade VIII were found to contain the F-type oriTs. They all carried the genes encoding relaxases of the MOB_F family, genes encoding T4CPs of the TrwB/TraD subfamily, and Tra_F-like T4SS gene clusters (Figure 2; Supplementary Figure S9). Notably, K. pneumoniae strain JNQH116 plasmid pJNQH116-2 (NZ_CP070900), belonging to the clade VIII cluster, was found to contain both Tra_F-like and Trw-like T4SS gene clusters in its genome. For clade VIII, its members were widely geographically distributed all over the world, including China, India, Southeast Asia, Middle East, North America (Canada and United States), South Africa, and some European countries (e.g., Germany, Romania, and the United Kingdom; Figure 3; Table 1; Supplementary Table S3). No prevalent STs of K. pneumoniae host strains containing all clade VIII plasmids were found (Table 1; Supplementary Table S3).

Clade IX: A total of 21 IncC plasmids carrying bla_NDM-1 were grouped into the clade IX cluster of the phylogenetic cladogram (Figure 2). Their genome sizes varied from 140.1 to 329.2 kb, with the 25th percentile, median size, and 75th percentile being 144.3, 147.9, and 178.7 kb, respectively (Supplementary Figure S1). For the conjugative transfer modules, all the plasmids belonging to clade IX carried the A/C-type oriTs, genes encoding relaxases of the MOB_H family characterized by the domain “TraI_2 (Pfam: PF07514),” mostly genes encoding T4CPs of TrwB/TraD subfamily, and Tra_F-like T4SS gene clusters (Figure 2; Supplementary Figure S10). The IncC plasmids harboring bla_NDM-1 were widely geographically distributed all over the world, mainly in the United States, Viet Nam, and China (Figure 3; Table 1; Supplementary Table S3). ST11 and ST1967 were the common STs strains containing bla_NDM-1-harboring IncC plasmids (Table 1; Supplementary Table S3).

Clade X: A total of 40 mega plasmids, where the length range varied from 238.0 to 401.6 kb (25th percentile = 293.5 kb; median = 327.3 kb; 75th percentile = 355.1 kb), mainly carrying bla_NDM-1, were grouped into a large cluster, named clade X in our study (Figure 2). Of the plasmids belonging to clade X, 27 (67.5%) were found to contain both replicons IncFIB(pNDM-Mar) and IncHI1B(pNDM-MAR), seven (17.5%) were unable to be typed, and four (10.0%) were IncQ1 plasmids. Moreover, all the plasmids of clade X carried genes encoding T4CPs of the TrwB/TraD subfamily and Tra_F-like T4SS gene clusters (Figure 2; Supplementary Figure S11). Most of the plasmids belonging to clade X were found to have no oriT and harbored the genes encoding relaxases of the MOB_H family. For the clade with the largest number, clade X, its members were widely distributed all over the world, mainly in China and the United States (Figure 3; Table 1; Supplementary Table S3). ST14, ST11, and ST147 were the common STs strains containing the plasmids of clade X (Table 1; Supplementary Table S3).

We also perform a bipartite network analysis with the 171 bla_NDM-harboring plasmids of K. pneumoniae. The bipartite network consisted of two classes of nodes: 171 plasmid units (PUs) and 2,502 homologous protein clusters (HPCs, protein families according to amino acid sequence identity, coverage, and E-value; Figure 4). Edges connected every PU with the HPC that it contained. The PUs of the bipartite network clearly showed distinct clustering phenomena. Overall, one homologous PU clusters (PUCs) contained the almost the same members of Clade X in the analysis above (Figure 4), which was clearly distinct from other PUs. One large region including clades III, VI – IX was also identified, which were mostly IncF and IncC plasmids. In addition, clades I, II, IV, and V were also found their corresponding PUCs in the PU-HPC bipartite network.