Introduction

Carbapenem-resistant Pseudomonas aeruginosa (CRPA), most commonly seen in healthcare-associated infections, constitutes a serious challenge for global public health1,2. The carbapenemase distribution and related sequence types (STs) among clinical CRPA isolates varied worldwide. In most geographical regions outside China, subclass B1 metallo-β-lactamases were the predominant carbapenemase types, which were mainly identified in high-risk clones ST235 and ST1113,4,5. However, in China, clinical CRPA isolates exhibited distinctly different genomic characteristics.

Over the past few years in China, Klebsiella pneumoniae carbapenemase (KPC) associated with the emerging epidemic high-risk clone ST463 has rapidly expanded in P. aeruginosa and far outweighed other carbapenemases5,6,7. According to a multi-center epidemiological study, 40.4% of total 374 CRPA isolates from China possessed blaKPC-2 genes, and 70.9% of these KPC-2-producing P. aeruginosa (KPC-2-PA) isolates belong to ST4638. Carriage of KPC-2 enzymes rendered the hypervirulent clone ST463 with an additional carbapenem-resistant phenotype. These convergent phenotypes could worsen patient outcomes due to the high possibilities of both inappropriate empirical treatments and severe clinical infections. According to a retrospective cohort study focusing on bloodstream infections from China9, compared to non-ST463 CRPA isolates, ST463 CRPA isolates showed higher resistance rates to antipseudomonal agents and were also associated with significantly higher mortality rates. Thus, from the clinical aspect, KPC-2 prevalence should be considered a significant contributor to the evolutionary success of clone ST463 in China.

Plasmids are the most important vehicles for the spread of KPC enzymes among P. aeruginosa7,10. Globally, KPC-PA isolates have predominantly been reported in countries such as Colombia, Chile, and Argentina, where the ST463 clone is rarely observed11,12,13. These global KPC plasmids vary in size, ranging from 3.5 to 480 kb, and belong to various incompatibility (Inc) groups, such as IncU, IncP-6, and IncP14. In contrast, the previous study in China has identified two main types of KPC-related plasmids: type I and type II, which together account for ~80% of the KPC plasmid pool8. Unlike global KPC plasmids, the type I and type II plasmids in China are not associated with any existing Inc group14. Furthermore, while isolates from other STs in China usually harbor only one of these plasmid types, both type I and type II plasmids have been detected in the ST463 clone8. This indicates a unique evolutionary trajectory for ST463 KPC-PA, with more complex plasmid characteristics than those observed from other STs.

To better understand the evolution processes of KPC-related plasmids in clone ST463, we collected 1258 clinical P. aeruginosa isolates over a decade in a single tertiary hospital from China, proposed a multi-stage evolutionary trajectory for KPC-related plasmids in ST463 strains based on whole-genome analysis results, and finally verified our proposals under clinical ST463 genetic backgrounds.

Results

Genomic characteristics of ST463-PA

To explore the global distribution and general characteristics of clone ST463, we first analyzed the genomic data of 580 ST463 strains from NCBI datasets. As shown in Fig. 1, the majority of these isolates were collected from China (463/580) and the United States (105/580), with sporadic isolates identified in Singapore, Israel, and several European countries. The earliest known ST463 isolate was identified in 2000 in the United States, where ST463 strains exhibited a more diverse evolutionary distribution compared to the more concentrated distribution observed in China.

Fig. 1: Core genome phylogenetic tree of 580 ST463-PA isolates from NCBI datasets as of August 27, 2024.
figure 1

The concentric rings, from inside to outside, represent plasmid types, blaKPC gene carriage, separation year, and country/region of isolation for each strain. The tree was constructed using IQ-TREE version 2.1.433 with the GTR + F + ASC + R5 model and was midpoint rooted.

Full size image

There are significant differences between Chinese and United States ST463 strains regarding the blaKPC gene and plasmid carriage. In China, nearly all ST463 isolates carried type I and/or type II plasmids (446/463), with the majority also harboring the blaKPC gene (433/463). In contrast, only three ST463 isolates from the United States carried either type I or type II plasmids, and none of the United States isolates carried the blaKPC gene. These findings suggest that ST463 may have originated in the U.S. and subsequently spread to China, where it evolved to acquire KPC plasmids, facilitating its rapid dissemination across multiple hospitals in China.

Given the rapid spread and clinical relevance of ST463 in China, in the following sections, we further analyzed the genomic characteristics of ST463 strains collected in our study over the past decade. In total, 102/839 CRPA isolates and 4/419 non-CRPA isolates in this study were assigned to ST463. The detection rate of ST463 isolates in CRPA increased over time, especially from 2015 to 2020 (Fig. 2a). The general genomic characteristics of these ST463-PA strains are shown in Fig. 2b. Apart from ten non-CPPA isolates, other 96 ST463-PA isolates all harbored blaKPC genes, including the blaKPC-2 allele (95/96, 99.0%) and the blaKPC-33 allele (1/96, 1.0%). Blast results also revealed two kinds of KPC-related plasmids within clone ST463. The type I plasmid was identified in all ST463-PA isolates, while the type II plasmid was only identified in five ST463 isolates from 2011-2014. In other words, 101/106 ST463-PA isolates harbored only the type I plasmid, while 5/106 ST463-PA isolates harbored both the type I and type II plasmids.

Fig. 2: Genomic characteristics of 106 ST463-PA isolates included in this study.
figure 2

a The distribution of type I plasmid coverage (blasted against pS11–17) in ST463-PA isolates (depicted as red dots); and the prevalence of clone ST463 among CRPA isolates (depicted as blue line). b Core genome phylogenetic tree of ST463-PA isolates. Carbapenemase type, presence of type I plasmid, presence of type II plasmid, and isolation year are indicated from the inside out. c Comparison of the typical blaKPC-2 genetic context from the type I and type II plasmids in ST463-PA isolates. Gray shading indicates identical sequences. d Alignments of plasmid contigs from ST463-PA isolates showing decreased coverage to pS11–17. CDS coding sequence, CRPA carbapenem-resistant Pseudomonas aeruginosa, TR transcriptional regulator.

Full size image

Genetic contexts and locations of blaKPC-2 genes in ST463-PA were analyzed based on both Illumina and Nanopore sequencing, with the representatives shown in Fig. 2c. In isolates co-carrying the type I and type II plasmids, the blaKPC-2 genes were located in the type II plasmids with IS6100-related structure (IS6100-ISKpn27-blaKPC-2-ΔISKpn6-korC-klcA-Tn1403). This structure was basically identical to the typical KPC structure in type II plasmids from other STs7,15. Instead, in ST463 isolates not carrying the type II plasmids, the blaKPC-2 genes were located on the type I plasmids and/or the chromosomes. Under these circumstances, IS6100-related structure was not identified; instead, the core blaKPC-2 platforms were found to be surrounded by multiple copies of IS26 (typically IS26-ISKpn27-blaKPC-2-ΔISKpn6-korC-klcA-IS26). These distinct differences in blaKPC-2 genetic contexts suggested that blaKPC-2 translocations within type I and type II plasmids would be hard to occur in clone ST463.

To further analyze the type I plasmids from ST463-PA, a non-KPC-encoding type I plasmid pS11–17 was chosen. The plasmid pS11–17 was detected in S11–17, the first ST463 strain isolated in this study and the only ST463 strain isolated in 2011. With this ancestor plasmid as the reference, plasmid coverage was calculated for all type I plasmids harbored by ST463-PA. As shown in Fig. 2a, in 2011–2016, all ST463-PA strains shared 94.2–100.0% coverage with plasmid pS11–17; in 2017, two out of five ST463 isolates started to show decreases in pS11–17 coverage (56.2%-60.0%). This plasmid pattern rapidly became the majority during 2018-2020, indicating that large fragment deletions occurred during the type I plasmid evolution in clone ST463.

Plasmid contigs from 68 ST463 strains showing decreased coverage were then extracted and mapped to pS11–17 (Fig. 2d). In most cases, these ST463-PA strains exhibited losses of two plasmid fragments. The first 4.9 kb fragment consisted of genes encoding DNA methyltransferase, transcriptional regulators, and hypothetical proteins. The second 16 kb fragment mainly consisted of genes encoding type IV secretion system (T4SS) VirB proteins and transcriptional regulators. Notably, the rapid expansion of clone ST463 in CRPA from 2017 was consistent with the prevalence of fragment-deleted type I plasmids in ST463-PA (Fig. 2a). Thus, we supposed that fragment deletions in the type I plasmid might provide evolutionary advantages for clone ST463.

Proposal on the evolution trajectory of KPC-related plasmids in ST463-PA

Based on the genomic characteristics described above, a total of six plasmid patterns were concluded for ST463 strains in our study (Fig. 3a). Specifically, harboring non-KPC-encoding original type I plasmid (pT1 plasmid) was referred to as pattern a; while harboring KPC-encoding original type I plasmid (pT1-KPC plasmid) was referred to as pattern A. Harboring pT1 plasmid and non-KPC-encoding type II plasmid (pT2 plasmid) was referred to as pattern b; while harboring pT1 plasmid and KPC-encoding type II plasmid (pT2-KPC plasmid) was referred to as pattern B. Harboring non-KPC-encoding fragment-deleted type I plasmid (pT1del plasmid) was referred to as pattern c; while harboring KPC-encoding fragment-deleted type I plasmid (pT1del-KPC plasmid) was referred to as pattern C.

Fig. 3: Evolution processes of KPC-related plasmids in ST463-PA isolates.
figure 3

a Time distribution of different KPC-related plasmid patterns from ST463-PA isolates in our study. The dominant and subordinate plasmid patterns every year are marked by opaque color with solid lines, and transparent color with dash lines, respectively. b Proposal on the multi-stage evolution trajectory of KPC-related plasmids in ST463-PA isolates. Created in BioRender. Chen, Y. (2025) https://BioRender.com/l76o033.

Full size image

The timeline distribution of these plasmid patterns in ST463-PA is also depicted in Fig. 3a. At first in 2011, only pattern a was detected. Then in 2012, patterns A and B both emerged. Over the next few years (2013–2016), strains with pattern B gradually disappeared; meanwhile, strains with pattern A expanded and became the dominant subclone. In 2017, pattern C arose and started to prevail among ST463 isolates, finally superseding the former pattern A.

Based on these findings, we proposed three stages in the evolution trajectory of KPC-related plasmids in ST463-PA (Fig. 3b). In stage 1 (2011–2012), original ST463 isolates (pattern a) succeeded in overcoming carbapenem pressure, either by integrating the blaKPC-2 gene into the original pT1 plasmid (pattern A) or receiving the pT2-KPC plasmid (pattern B). In stage 2 (2012-2017), owing to the higher fitness associated with pT1-KPC plasmid over pT2-KPC plasmid, strains with pT1-KPC plasmid (pattern A) became prevalent in ST463 CRPA. In stage 3 (2017–2020), pT1del-KPC plasmid (pattern C) rendered ST463 isolates with more fitness than pT1-KPC plasmid (pattern A), thus replacing pattern A and making ST463 the dominant ST among CRPA.

To prove these hypotheses, we chose a rifampin-resistant S11–17 mutant (namely S11-17R) for further experiments. S11–17, isolated in 2011 with plasmid pattern a, was the ancestor ST463 strain in our study. As shown in Fig. 4, by separately delivering representative plasmids from clinical strains into S11-17R, we constructed four additional plasmid patterns under the chromosome background of S11-17R. A derivative of plasmid pattern B, namely pattern Bvar (harboring only pT2-KPC plasmid), was also constructed to simplify comparison. Among these KPC-producing isolates, the lowest minimum inhibitory concentrations (MICs) were 128 mg/L for imipenem and 16 mg/L for meropenem. Therefore, we used 1/2 MICs (64 mg/L imipenem and 8 mg/L meropenem) and 1/16 MICs (8 mg/L imipenem and 1 mg/L meropenem) as the carbapenem selection pressure in the following experiments. Specifically, to simulate the in-vivo carbapenem concentration in competition experiments, only 8 mg/L meropenem and 8 mg/L imipenem were used.

Fig. 4: Characteristics of experimental ST463 strains used in this study.
figure 4

Schematic representation of experimental strains and their corresponding minimum inhibitory concentrations of carbapenems are shown in the figure. The pT1 (GenBank accession: CP138579), pT2-KPC (GenBank accession: CP137894), pT1-KPC (GenBank accession: CP138390), and pT1del-KPC (GenBank accession: CP138392) plasmids in these strains were originated from clinical isolates S11–17, R11-08, R13-28, and R20-48, respectively. Created in BioRender. Chen, Y. (2025) https://BioRender.com/h33e664.

Full size image

Fitness costs associated with different plasmid patterns in ST463 genetic backgrounds

Fitness costs associated with various plasmid patterns under different selection pressures were evaluated by the growth curve (Fig. 5). From the comparison results in ST463 genetic backgrounds, several features could be summarized. First, to obtain carbapenem-hydrolyzing ability, ST463-PA acquired the blaKPC-2 gene at the expense of sacrificing strain fitness under no antibiotics (Fig. 5a). Compared with the strain harboring pT1 plasmid (pattern a), strains harboring pT1-KPC plasmid (pattern A) or an additional pT2-KPC plasmid (pattern B) exhibited decreased area under the curve (AUC), with or without reduced maximum growth rates.

Fig. 5: Fitness costs associated with different KPC-related plasmids in experimental ST463 strains.
figure 5

Various conditions were tested: a no antibiotics, b 8 mg/L imipenem, c 64 mg/L imipenem, d 1 mg/L meropenem, and e 8 mg/L meropenem. Each data point in the panels corresponds to one biological replicate calculated from three technical replicates (n = 3 biologically independent samples). Error bars represent the mean ± SD and two-tailed P values are denoted. Statistical significance was assessed by unpaired t-test.

Full size image

Second, although type I and type II plasmids could both mediate carbapenem resistance in ST463, fitness costs accompanying them were different. As displayed in Fig. 5a–e, the fitness properties of KPC-producing strains with only type I plasmid (patterns A and C) all outweighed those with type II plasmid (patterns B and Bvar), whatever the antimicrobial pressure. These results demonstrated that the KPC-encoding type I plasmid fitted better in ST463 genetic backgrounds than the KPC-encoding type II plasmid.

Third, fitness impacts exerted by fragment deletions in type I plasmid varied with antimicrobial conditions. When imipenem or 1 mg/L meropenem was applied, strains with pT1del-KPC plasmid (pattern C) or pT1-KPC plasmid (pattern A) presented comparable growth characteristics (Fig. 5b–d). However, with no antimicrobial or 8 mg/L meropenem, a higher fitness level was noted in strain with pattern C than in strain with pattern A (Fig. 5a, e). Overall, the pT1del-KPC plasmid was always non-inferior to the pT1-KPC plasmid in the aspect of ST463 fitness.

Copy numbers of plasmid genes under different carbapenem pressures in ST463 background

To figure out whether ST463-PA would regulate the copy number of plasmid genes under carbapenem pressure, the genomic DNA of carbapenem-pre-exposed strains was analyzed by quantitative PCR. Generally, the ST463 isolates with KPC-encoding type II plasmid responded poorly to carbapenem exposure. As shown in Fig. 6a, the strain harboring pT2-KPC plasmid showed no elevation in the blaKPC-2 gene copy number, regardless of carbapenem stress types and concentrations. In contrast, different strategies were adopted by strains harboring different KPC-encoding type I plasmids. The details are shown as follows.

Fig. 6: Gene copy number changes under different carbapenem pressure.
figure 6

a–c Gene copy number of pT2-KPC, pT1-KPC, and pT1del-KPC plasmid. d Comparisons between pT1-KPC and pT1del-KPC plasmids. Each data point in the panels corresponds to one biological replicate calculated from three technical replicates (n = 3 biologically independent samples). Error bars represent the mean ± SD, and two-tailed P values are denoted. Statistical significance was assessed by unpaired t-tests. CAMHB cation-adjusted Mueller–Hinton broth, IPM imipenem, MEM meropenem, I_repA the repA gene of type I plasmid, II_repA the repA gene of type II plasmid.

Full size image

In strain harboring pT1-KPC plasmid, changes in gene copy number varied by conditions (Fig. 6b). When 8 mg/L imipenem or 8 mg/L meropenem was applied, copy numbers of the pT1-KPC plasmid, the IS26 gene, and the blaKPC-2 gene all remained unchanged. When 64 mg/L imipenem was applied, the pT1-KPC plasmid copy number decreased, which further lowered blaKPC-2 and IS26 copy numbers. On the contrary, when 1 mg/L meropenem was applied, copy numbers of blaKPC-2 and IS26 genes both elevated. This trend was accompanied by the unaltered plasmid copy number, suggesting IS26-mediated blaKPC-2 amplification was achieved in pT1-KPC plasmid under low meropenem concentration.

The strain harboring pT1del-KPC plasmid responded to carbapenem stress in a somewhat different way (Fig. 6c). Specifically, neither 8 mg/L nor 64 mg/L imipenem raised copy numbers of the pT1del-KPC plasmid, the IS26 gene, and the blaKPC-2 gene. Instead, 1 and 8 mg/L meropenem both upregulated the copy numbers of the pT1del-KPC plasmid, the IS26 gene, and the blaKPC-2 gene. Under this circumstance, the IS26 and blaKPC-2 copy numbers relative to the plasmid copy number remained stable. This demonstrated plasmid-mediated blaKPC-2 amplification in pT1del-KPC plasmid under both low and high meropenem concentrations.

The type I plasmid copy number was further compared between strains harboring pT1-KPC plasmid or pT1del-KPC plasmid (Fig. 6d). Without carbapenem pressure, the copy number of pT1del-KPC plasmid was 41.5% lower than that of pT1-KPC plasmid. After being exposed to imipenem, copy number fold changes of pT1-KPC and pT1del-KPC plasmids were similar. Nevertheless, after being exposed to 1 or 8 mg/L meropenem, copy number fold change of pT1del-KPC plasmid was much higher than that of pT1-KPC plasmid. Taken together, these data suggested that fragment deletions provided type I plasmid with more copy-number flexibility in clone ST463.

Stability of KPC-related plasmids and bla KPC-2 genes in ST463 background

The stability of KPC-related plasmids (pT1, pT1-KPC, pT1del-KPC, pT2-KPC) and plasmid-carrying blaKPC-2 genes was tested via 12-day serial passage without antibiotics (Table 1). All four types of plasmid backbone were 100% preserved by the corresponding ST463 population, yet blaKPC-2 genes were only 100% preserved by the population harboring pT2-KPC plasmid. For populations carrying pT1-KPC or pT1del-KPC plasmids, the blaKPC-2 gene loss rates at day 12 were 14.9 and 2.8%, respectively (two-sided P = 0.0005, Fisher’s exact test). Based on our analysis above, we supposed that the IS26-flanked structure in type I plasmid provided the blaKPC-2 gene with more mobility than the IS6100-related structure in type II plasmid. Moreover, fragment deletions in type I plasmid compensated the fitness costs brought by blaKPC-2 gene, thus loss rate of blaKPC-2 gene was significantly lower in pT1del-KPC plasmid than in pT1-KPC plasmid.

Table 1 Stability of KPC-related plasmids and plasmid-carrying blaKPC-2 genes in ST463 background
Full size table

Compensation for impaired pT1del-KPC plasmid transferability by pT1 plasmid from recipients

Compared with the pT1-KPC plasmid, deleted fragments in the pT1del-KPC plasmid were largely composed of T4SS VirB proteins involved in the mating-pair formation process during conjugation16. Therefore, we investigated the differences in their transferability from rifampin-resistant PAO1 donors to tetracycline-resistant PAO1 recipients. As shown in Fig. 7a, the calculated conjugation frequencies were 3.77 × 10−1 for pT1-KPC plasmid and <3.71 × 10−4 for pT1del-KPC plasmid, demonstrating an at least 1016-fold decrease in the transferability of KPC-encoding type I plasmid resulting from the fragment loss.

Fig. 7: Calculated conjugation frequency under different conditions.
figure 7

a Conjugation frequency differences between pT1-KPC and pT1del-KPC plasmids. b Different transferability of pT1del-KPC plasmids in the presence and absence of pT1 plasmid in the recipient cells. The conjugation efficiency shown in the figure represents the mean of three independent biological replicates, with each experiment conducted in three technical replicates (n = 3 biologically independent samples). Created in BioRender. Chen, Y. (2025) https://BioRender.com/r61n031.

Full size image

Interestingly, although no transconjugant carrying pT1del-KPC plasmid was obtained during PAO1-PAO1 conjugation, pT1del-KPC plasmid could still transfer horizontally between ST463 isolates under specific conditions. As shown in Fig. 7b, clinical ST463 strain R20-48 carrying the pT1del-KPC plasmid was chosen as the donor strain; meanwhile, the previously-mentioned S11-17R (with pT1 plasmid) and S11-17R ΔpT1 (without pT1 plasmid) were chosen as the recipient strains. The transconjugants were obtained at frequencies of 3.95 × 10−3 in R20-48-S11-17R conjugation and <2.20 × 10−5 in R20-48-S11-17R ΔpT1 conjugation, respectively. Hence, an at least 180-fold increase in pT1del-KPC plasmid transferability was achieved. In this context, transmission constraints of pT1del-KPC plasmid resulting from fragment loss were significantly weakened with the help of pT1 plasmid from the recipient.

Competition advantage of pT1-KPC plasmid over pT2-KPC plasmid in clone ST463

Since the pT1-KPC and pT1del-KPC plasmids possessed the same repA gene and exhibited high sequence similarity, in-vitro competition experiments were only performed for pT1-KPC and pT2-KPC plasmids under the chromosome background of S11-17R. As shown in Fig. 8, in clone ST463, the pT1-KPC plasmid displayed obvious competitive fitness over the pT2-KPC plasmid, regardless of carbapenem stress types and concentrations. As a result, the proportion of pT1-KPC plasmid increased averagely 19.7–24.2% faster than pT2-KPC plasmid per generation, which was consistent with both the experimental findings above and clinically observed phenomena.

Fig. 8: Selection coefficients of pT1-KPC plasmid over pT2-KPC plasmid under various conditions in clone ST463 background.
figure 8

Each data point in the figure corresponds to one biological replicate (n = 3 biologically independent samples). Error bars represent the mean ± SD. CAMHB cation-adjusted Mueller–Hinton broth, IPM imipenem, MEM meropenem.

Full size image

Discussion

Given the successful dissemination of KPC-producing ST463 isolates in China, an exploration of the underlying evolution mechanisms is urgently needed. In this study, we investigated the changes in KPC-related plasmids within clone ST463 over ten years. Our result revealed that the overwhelming advantages of type I plasmids over type II plasmids, and further fitness compensation by fragment deletions in type I plasmids were two major milestones to the success of clone ST463.

Amplification of plasmid antimicrobial resistance (AMR) genes serves as a crucial method for bacteria to adapt to antibiotic stress17,18,19. To achieve the amplification, both the plasmid-driven strategy and the IS-driven strategy have been demonstrated to be adopted by plasmid-bearing bacteria20,21,22,23. Nevertheless, in most cases, bacteria only utilized a single tactic, and strategy shifts along with the change of plasmid size have not been reported before. Interestingly, in our study, we observed a transition from IS26-driven blaKPC-2 amplification in pT1-KPC plasmid to plasmid-driven blaKPC-2 amplification in pT1del-KPC plasmid. This strategy transition was likely for the purpose of minimizing plasmid fitness costs. Compared to the pT1-KPC plasmid, the pT1del-KPC plasmid was about 20 kb smaller and lost the bulk of T4SS. Thus, plasmid-driven strategy would produce much lower fitness costs in pT1del-KPC plasmid than in pT1-KPC plasmid, which facilitated the transition from the IS26-driven strategy. Remarkably, pT1del-KPC plasmids at the late evolution stage possibly reserved the blaKPC-2 duplications from earlier pT1-KPC plasmids. Under this condition, blaKPC-2 gene dosage effects would be very high if the pT1del-KPC plasmid copy number further increased.

In our study, we observed that meropenem and imipenem exposure led to different trends in plasmid gene copy number, which may be attributed to differences in their chemical structure, mechanism of action, and associated resistance pathways24. Specifically, meropenem’s higher affinity for PBP-3 and its regulation by the MexAB-OprM efflux pump, in contrast to imipenem, could explain the differential impact on KPC plasmids4,25. Additionally, clone-specific factors in the high-risk ST463 strain may further contribute to these variations. This finding highlights the complex interplay between antibiotic pressure and plasmid dynamics in clinical settings. Moreover, it is likely that additional factors beyond antibiotic resistance also play a role in shaping the fate of KPC-related plasmids. Future studies could further investigate the roles of other plasmid-encoded genes, such as those involved in metabolism, stress response, or virulence, in shaping the evolutionary trajectories of KPC-related plasmids. Combining these aspects with antibiotic pressure will provide a more comprehensive understanding of KPC plasmid evolution in clinical and environmental contexts.

Plasmid-encoding blaKPC-2 genes could be transferred within ST463 isolates either vertically through bacteria division or horizontally via plasmid conjugation. In our study, both gene transfer methods contributed to the evolutionary success of pT1del-KPC plasmid over pT1-KPC plasmid in clone ST463. From the perspective of vertical transmission, our analysis demonstrated that the pT1del-KPC plasmid has significant fitness advantages compared to the pT1-KPC plasmid. This can be attributed to two key factors. First, the pT1del-KPC plasmid carries a deletion of ~20 kb, removing conjugative and accessory elements that likely impose metabolic burdens on the bacterial host. By losing these elements, the pT1del-KPC plasmid reduces the energy demand on the host, thereby lowering fitness costs and improving stability. Second, the blaKPC-2 gene on the pT1del-KPC plasmid shows increased copy-number flexibility under carbapenem pressure, allowing the host bacteria to better respond to antibiotic stress. These factors together likely contribute to the superior growth of strains harboring the pT1del-KPC plasmid. From the perspective of horizontal transmission, although the pT1del-KPC plasmid lost the capacity of self-transmission, it could still be mobilized with the help of conjugative plasmids pT1 or pT1-KPC from the recipients. This phenomenon has been similarly observed in Enterobacteriaceae26. In their study, mobilization of a non-conjugative virulence plasmid from Klebsiella pneumoniae to Escherichia coli J53 could only be achieved when J53 was pre-transformed with a self-transferable plasmid26. Thus, this conjugation process resulted in the co-existence of both the non-conjugative plasmid and the self-transferable plasmid in transconjugants. However, in our study, the non-conjugative plasmid (pT1del-KPC) and the self-transferable plasmid (pT1 or pT1-KPC) could not stably co-exist in a strain due to the plasmid incompatibility27. Consequently, a pT1del-KPC plasmid from the donor cell first utilized the T4SS encoding by pT1 or pT1-KPC plasmids from the recipient cell to achieve mobilization, then replaced these incompatible plasmids by its competitive advantages. Using this tricky scheme, ST463 strains with pT1del-KPC plasmids successfully expanded their own advantages by shifting the T4SS synthesis burden onto other ST463 rivals.

The proposed evolution trajectory of KPC-related plasmids in clone ST463 from our study was based on single-hospital data, yet their generality could be verified by ST463 datasets from other hospitals, both cross-sectionally and longitudinally. In a cross-sectional dataset (BioProject accession number PRJNA672835) recently deposited8, a total of 104 ST463 isolates were identified from three other East China hospitals (except the hospital included in our study and a hospital with only five ST463 strains). Our analysis of these genome sequences revealed that all these ST463 isolates carried type I plasmids. Meantime, the majority of them (85/104, 81.7%) shared 52.6%-58.1% coverage with pS11–17, and exhibited fragment loss in an almost identical manner to the pT1del-KPC plasmid in our study (Supplementary Table 1). In addition, we also analyzed a longitudinal ST463 dataset (BioProject accession number PRJNA648026)7. This dataset comprised 77 ST463 strains isolated from 2010 to 2018 in their hospital. Similarly, the co-existence of type I and type II plasmids in these ST463 strains was only found before 2014 (Supplementary Table 2). In 2011-2015, nearly all ST463-PA isolates shared 90.3–100.0% coverage with pS11–17. However, from 2016, ST463 strains with decreased pS11–17 coverage (56.3–58.1%) started to emerge and became the majority (9/11, 81.8%) in 2018 (Supplementary Table 3). Moreover, their fragment loss patterns were also basically the same as the pattern described in our study. These data supported our findings that pT1del-KPC plasmid finally dominated in clone ST463 due to its competitive advantages associated with the characterized fragment deletions. Taken together, the wide applicability of our proposals on clone ST463 could be concluded.

Our study has important implications for clinical practice and infection control. First, the rapid spread of KPC-PA isolates, particularly the high-risk ST463 clone, underscores the need for routine carbapenemase testing for CRPA strains resistant to traditional β-lactam antibiotics in epidemic areas. Early detection of KPC-PA strains can help guide the appropriate use of novel β-lactam/β-lactamase inhibitor combinations, such as ceftazidime/avibactam, imipenem/relebactam, or cefiderocol, thereby optimizing treatment and reducing unnecessary antibiotic exposure. Second, the recent surge of the ST463 clone in East China highlights the critical need for regular molecular epidemiological surveillance in other regions. Although the global dissemination of KPC-related plasmids within ST463 strains remains limited, the rapid spread observed in China could potentially be mirrored in other regions if similar plasmid acquisition events occur. Timely monitoring is essential in preventing the spread of multidrug-resistant strains, controlling hospital outbreaks, and minimizing the risk of further nosocomial infections. Beyond clinical relevance, our study also provides key insights into bacterial plasmid evolution. The stepwise evolution of KPC-related plasmids in ST463 underscores the critical role of horizontal gene transfer and plasmid adaptation in shaping bacterial populations, offering new perspectives on microbial evolution and resistance management.

In conclusion, with whole-genome sequencing data, we deciphered the genomic characteristics of high-risk ST463 P. aeruginosa isolates from China over a decade. Based on these analyses, we concluded various representative plasmid patterns within clone ST463, and proposed a multi-stage evolutionary trajectory for KPC-related plasmids in ST463 strains. Under clinical ST463 genetic backgrounds, we demonstrated the fitness and competitive superiority of original type I plasmids over type II plasmids, and illustrated the further evolutionary benefits obtained by type I plasmids due to large fragment loss. Overall, our data highlighted the plasmid dynamics of clone ST463, which deepened the understanding of its rise and prevalence in China. Considering its rapid adaptations and tight correlations with KPC-2 carbapenemase, clinicians should be alert to this emerging high-risk clone and prevent its hospital transmission in clinical practice. Besides, further genomic surveillance on clone ST463 is also needed to trace its evolution in time.

Methods

Strain collection and antimicrobial susceptibility testing

Clinical P. aeruginosa strains were collected from 2011 to 2020 in a tertiary teaching hospital in Hangzhou, China. In total, 839 nonduplicated CRPA isolates and 419 nonduplicated non-CRPA isolates were included. Strain distribution by year is depicted in Supplementary Fig. 1. We defined nonduplicated strain as one CRPA strain and/or one non-CRPA strain for each patient during a single hospitalization. With P. aeruginosa ATCC 27853 as quality control, MICs of bacterial strains to carbapenems were determined using broth microdilution methods described in CLSI M0728. CLSI M100 breakpoints were adopted for MIC results interpretation29.

Whole-genome sequencing and plasmid type analysis

The genomic DNA of all clinical P. aeruginosa isolates was extracted using QIAamp DNA Mini Kit (Qiagen, Hilden, Germany) and subjected to Illumina short-read sequencing. De novo genome assembly was performed using shovill version 0.9.0 (https://github.com/tseemann/shovill). Sequence types were assigned by mlst version 2.19.0 (https://github.com/tseemann/mlst), and AMR genes were detected by ABRicate version 1.0.0 (https://github.com/tseemann/abricate). For further genomic analysis and experiments, we selected representative P. aeruginosa isolates for Nanopore sequencing. Their assemblies were first generated using Nanopore long reads with Raven version 1.1.10 (https://github.com/lbcb-sci/raven), and then polished using Illumina short reads with Polypolish version 0.5.030.

To clarify the presence and types of KPC-related plasmids in clinical ST463 isolates, we blasted the assembled contigs against five representative KPC-encoding plasmids previously reported (GenBank accessions: KY296095, type I plasmid; CP064404, type II plasmid; CP078008, type III plasmid; CP064400, type IV plasmid; CP078010, type V plasmid)8,31. Before calculating plasmid coverage, contigs with plasmid hit lengths less than 2000 bp were filtered out. For each strain, plasmid coverage was calculated as the total plasmid hit length from all filtered contigs divided by the total plasmid sequence length. The presence of KPC-related plasmid in a strain was confirmed by a 45% coverage cutoff (for type I–IV plasmids) or a 50% coverage cutoff (for type V plasmid). Specifically, an ST463 strain in our study showed only 31% coverage with KY296095, and it was manually confirmed to possess the type I plasmid based on Nanopore sequencing results. The rationale behind the cutoff values is explained in detail in Supplementary Note 1.

Pangenome analysis and phylogenetic tree construction for ST463-PA

We retrieved all available P. aeruginosa genome sequences (a total of 31,836 GCA sequences) from NCBI as of August 27, 2024. The identification of ST463 isolates was performed using mlst version 2.19.0 (https://github.com/tseemann/mlst) with “--scheme paeruginosa” option. Duplicate genomes were removed by checking biosample accession numbers, and genomes with excessive contig numbers or abnormal gene counts were excluded to ensure data quality.

For both the ST463-PA strains in our study and those from the NCBI database, we performed separate pangenome analyses and phylogenetic tree constructions. Using P. aeruginosa PAO1 (GenBank accession: NC_002516) as the reference genome, the pangenome of ST463-PA strains was annotated with Prokka version 1.14.6 (https://github.com/tseemann/prokka), and core gene alignments were obtained through Panaroo version 1.2.932. The phylogenetic trees were constructed using IQ-TREE version 2.1.433, applying the best-fit models TVM + F + ASC + R3 and GTR + F + ASC + R5 for the respective datasets. Visualization of the phylogenetic trees was completed using iTOL version 6.834.

Plasmid delivery and plasmid curing

Plasmid delivery into P. aeruginosa isolates was performed using electroporation or conjugation35. Conjugation frequency was calculated as the transconjugant cell number per recipient cell from three independent experiments. To facilitate selection during conjugation, rifampin-resistant mutants were obtained through one-step resistance selection, while tetracycline-resistant mutants were generated using CRISPR/Cas9-based tetA gene insertion. The rifampin and tetracycline concentrations used during conjugation selection were 800 and 300 mg/L, respectively. Plasmid curing in P. aeruginosa isolates was achieved with an established CRISPR/Cas single-plasmid system8. The primers used for PCR screening during these processes are listed in Supplementary Table 4.

Growth curve

To figure out the fitness of different P. aeruginosa isolates under different antibiotic concentrations, growth curves were obtained. Briefly, 1.5 μL overnight cultures were diluted into 998.5 μL cation-adjusted Mueller–Hinton broth (CAMHB) with or without antibiotics. The diluted samples were incubated at 37°C with continuous agitation, and OD420-580 values were measured every five minutes for 1300 min using a Bioscreen C MBR machine (Oy Growth Curves Ab Ltd., Finland). The AUC calculated using GraphPad Prism version 9.0.0, and the relative maximum growth rate obtained through linear regression, were adopted for fitness comparison. Growth curve analysis was performed in biological triplicates. Each biological triplicate was analyzed in technical triplicates.

Quantitative PCR

To evaluate the change trends in gene copy number under different antibiotic concentrations, genomic DNA was extracted in biological triplicates for each P. aeruginosa strain using overnight culture as described above. Quantitative PCR was then performed in technical triplicates using the LightCycler 480 Instrument II Real-Time PCR system (Roche, Switzerland). The reaction mixture consisted of 10 µL of TB Green Premix Ex Taq (Takara, Beijing, China), 5 pmol of each primer, 2 µL of template DNA (reverse transcription product from 40 ng of total RNA), and nuclease-free water, making up a final volume of 20 µL. The thermal cycling protocol consisted of an initial denaturation step at 95°C for 30 seconds, followed by 40 cycles of denaturation at 95 °C for 5 seconds and annealing at 60 °C for 30 s. A melt curve analysis was also performed at the end of the quantitative PCR to confirm the specificity of the amplified products. The housekeeping rpoD gene was selected as the endogenous reference gene. Gene copy numbers were determined from threshold cycles using a standard formula36. The primers used for quantitative PCR and the qPCR efficiency for each primer pair are listed in Supplementary Table 5.

Stability of KPC-related plasmids and bla KPC-2 gene

To evaluate the stability of KPC-related plasmids and the blaKPC-2 gene in P. aeruginosa isolates, serial passage without antibiotics was carried out for 12 days. Each P. aeruginosa isolate was assayed in biological triplicates. At day 0, single colonies were inoculated into Luria-Bertani (LB) broth and cultured overnight. From day 1 to day 11, overnight stationary-phase culture from the day before was diluted (1:1000) into fresh LB medium and grown overnight again. On day 4, day 8, and day 12, overnight cultures (47 colonies randomly selected for each biological replicate) were evaluated for plasmid loss and blaKPC-2 loss. For strains harboring KPC-encoding plasmids, colonies that grow on Mueller–Hinton agar (MHA) plates with 16 mg/L imipenem were considered not losing the KPC-encoding plasmids; instead, colonies that did not grow on MHA plates with 16 mg/L imipenem were PCR screened for the blaKPC-2 gene and the plasmid repA gene. For strains harboring non-KPC-encoding plasmids, all colonies obtained on MHA plates were PCR screened for the plasmid repA gene. Primers used for PCR screening are listed in Supplementary Table 4.

Competition experiments

To assess the relative fitness of different KPC-encoding plasmids, we conducted pairwise competition experiments in biological triplicates under the same genetic backgrounds. In brief, at day 0, the separate overnight cultures of two different strains were adjusted with 1X PBS buffer to reach an OD600 of 0.5. Bacterial suspensions were then equally (50 μL each) added to 4.9 mL fresh CAMHB medium (with or without antibiotics), allowing a 24-h period of in-vitro competition. From day 1 to day 3, 50 μL overnight mixed culture from the day before was diluted (1:100) into fresh CAMHB medium and competitively grown for 24 h again. The genomic DNA of overnight mixed culture was daily extracted as described above, and plasmid quantification was performed by multiplex digital PCR with primers targeting different plasmid repA genes. Selection coefficients were estimated by fitting data to an established formula37,38.

Statistics and reproducibility

Statistical analysis was performed using GraphPad Prism version 9.0.0. Growth curve and quantitative PCR data were analyzed using unpaired two-tailed t-tests; plasmid stability experiments were analyzed using Fisher’s exact test. Statistical significance was considered at P < 0.05.

For reproducibility, all experiments were performed with three independent biological replicates, each consisting of three technical replicates. The number of biological replicates and technical replicates for each experiment is provided in the figure legends. Biological replicates refer to independent cultures of the same strain under identical experimental conditions.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.