During February 2022 to July 2024, 17 nonduplicated CRPA strains co-harboring bla _VIM-1 and bla _IMP-45, collected from 10 patients, were enrolled in the study. Demographics and relevant clinical data of the patients were obtained through review of medical records. Susceptibility to ticarcillin-clavulanic acid, ceftazidime, cefoperazone-sulbactam, cefepime, meropenem, imipenem, amikacin, tobramycin, ciprofloxacin, levofloxacin, and colistin were measured using VITEK^®2 system (BioMérieux, Marcy l’Étoile, France) following the manufacturer’s instructions. Minimal inhibitory concentrations (MICs) were determined for cefiderocol, piperacillin-tazobactam, aztreonam (AZT) and ceftazidime/avibactam (CZA) by broth microdilution method and interpretation was according to recommendations of the CLSI (Lewis and James, 2022), P. aeruginosa ATCC27853 served as quality control strains.

To better understand the features of genetic environments surrounding bla _IMP-45 and bla _VIM-1 genes, we searched the Genbank database on NCBI and obtained 15 intact plasmids harboring bla _IMP-45 worldwide from China as of October 1, 2024.

As of October 1, 2024, 53 P. aeruginosa genomes assigned to ST313 were downloaded from the RefSeq database on NCBI, We excluded strains with unknown country of origin and collection time, retaining 46 ST313 P. aeruginosa isolates and 17 ST313 VIM-1-IMP-45 CRPA strains for phylogenetic analysis. The phylogenetic tree for all ST313 P. aeruginosa genomes was constructed according to their SNPs by using Snippy version 4.6.0 (https://github.com/tseemann/snippy). ST313 P. aeruginosa genomes’ SNPs were filtered to remove recombination using Gubbins v3.3.5 (Croucher et al., 2015). The phylogenetic tree of the IncP-2 plasmids carrying blaIMP-45 and the IncP-2 plasmids co-harboring bla _VIM-1 and bla _IMP-45 were also constructed based on their SNPs using Snippy version 4.6.0. All maximum likelihood phylogenetic trees were inferred from a core genome alignment by using Raxml v8.2.12 (Stamatakis, 2014). The specific parameters for Snippy were: alignment coverage threshold –minfrac = 0.9 and minimum depth –mincov = 10. For Gubbins, we used maximum iterations = 5 and P-value threshold = 0.05. All trees were visualized in iTOL (https://itol.embl.de).

To induce cefiderocol resistance in P. aeruginosa, we followed a previously reported method (Nurjadi et al., 2022) with slightly modifications. Briefly, The amikacin-sensitive clinical single colony of P. aeruginosa AS01 strain grown on Columbia agar supplemented with 5% sheep blood was inoculated into 5 mL of cation-adjusted Mueller-Hinton broth (CA-MHB) (BD Diagnostics, Germany) and incubated at 37°C with constant shaking at 200 rpm for 18 hours. After overnight incubation, 100 μL of the culture was transferred into a fresh 5 mL of CA-MHB, and cefiderocol (Shionogi, Japan) was added to achieve a final concentration of 0.5 mg/L. The culture was then incubated under the same conditions as the initial culture. This process was repeated daily, with the cefiderocol concentration doubled at each passage, until no visible turbidity or bacterial growth was observed after overnight incubation, or until the cefiderocol concentration reached 64 mg/L. For population analysis profiling, 10 μL of the bacterial suspension from the overnight culture was plated onto Columbia blood agar, and 10 single colonies were randomly selected for cefiderocol susceptibility testing using broth microdilution. Conjugation experiments were performed with strain AS02, which exhibited resistance to cefiderocol but remained susceptible to amikacin and was selected as the recipient strain, while strain CJ05 served as the donor strain. Donor and recipient bacteria were separately cultured in LB broth at 37°C until reaching logarithmic growth phase (OD600 = 0.6-0.8), then mixed in a 1:1 ratio and incubated at 37°C overnight for conjugation. The conjugation mixture was subjected to 10-fold serial dilution and plated on Mueller-Hinton agar containing cefiderocol (4 mg/L) and amikacin (16 mg/L), followed by incubation at 37°C for 18–24 hours. Putative transconjugants were confirmed by whole-genome sequencing. Additional conjugation experiments were conducted using Escherichia coli J53 as an alternative recipient strain. Each conjugation experiment was repeated three times. The antimicrobial susceptibility tests mentioned above were performed on strains AS01, AS02, and transconjugant TC01 using the VITEK^®2 system and broth microdilution.

Genomic DNA from VIM-1-IMP-45-CRPA strains CJ01-CJ17, and transconjugant strain TC01 was extracted and were sequenced using an Illumina NovaSeq PE150 at the Beijing Novogene Bioinformatics Technology Co., Ltd. Raw reads were filtered to remove low-quality sequences and adaptors. De novo assembly was conducted using SOAP de novo 2.04 (Li et al., 2008; Li et al., 2010), SPAdes 3.10.0 (Bankevich et al., 2012), and ABySS 1.3.7 (Simpson et al., 2009). The assembly results were integrated with CISA 1.3 software. The gap in preliminary assembly results was filled using Gapcloser 1.12. In addition, whole-genome sequences for CJ05, AS01 and AS02 were obtained through Illumina NovaSeq PE150 and nanopore sequencing on MinION flow cells. Hybrid assemblies of Illumina short reads and MinION long reads were prepared with Unicycler version 0.4.8 (Wick et al., 2017). Multilocus sequence typing (MLST) was performed by using an MLST tool (https://github.com/tseemann/mlst). Antimicrobial drug resistance and virulence genes were identified with ABRicate version 1.0.0 (https://github.com/tseemann/abricate) according to the National Center for Biotechnology Information (NCBI) AMRFinderPlus (https://www.ncbi.nlm.nih.gov) and virulence factor (http://www.mgc.ac.cn/VFs) databases. Pairwise single-nucleotide polymorphism (SNP) distance was evaluated with Snp-dists (https://github.com/tseemann/snp-dists). Sequence comparisons were performed by using Easyfig version 2.2.3 and annotated by Prokka and Bakta (Seemann, 2014; Schwengers et al., 2021). Alignments and visualization of plasmids were generated by BLAST Ring Image Generator (BRIG) software version 0.95 (https://sourceforge.net/projects/brig). Linear alignments of bla_IMP-45-bearing structures were generated using ggplot2 and gggenes in R-4.4.2. The heatmap was translated using TBtools-II (Chen et al., 2023). Platon (Schwengers et al., 2020) was used to predict plasmids in transconjugant TC01 and all ST313 VIM-1-IMP-45-CRPA except CJ05.

Retrieving the RCSB Protein Data Bank (PDB), the crystal structures of PBP3 complexed with aztreonam (PDB DOI: https://doi.org/10.2210/pdb3PBS/pdb) was applied as the structure template in the present study (Han et al., 2010). After PSI-BLAST (Altschul et al., 1997) and MUSCLE alignment (Edgar, 2022), the homology models of PBP3(P527S) were both built in Schrödinger. The protein structure was refined and preprocessed using the OPLS4 force field (Lu et al., 2021), following the ligands preparing for diverse ionization states and isomers. In the covalent docking, the aztreonam of PBP3 complexed with aztreonam was set as the centroid of grid box, and the β-lactam was chosen as reaction type for pose prediction. After completion of docking, the Prime MM-GBSA scores of the top one output poses for each ligand was calculated for comparing the binding affinities of the ligands (Li et al., 2011).

During 2022-2024, 17 ST313 P. aeruginosa strains co-harboring bla _IMP-45 and bla _VIM-1 were identified, including 2 isolates from 2022, 11 from 2023, and 4 from 2024. Following comprehensive disinfection measures implemented in the ward, no further occurrences of VIM-1-IMP-45-CRPA strains were detected. The VIM-1-IMP-45-CRPA strains were isolated from 10 immunocompromised patients, all of whom had undergone lung or kidney transplants ( Table 1 ). The cohort consisted of 9 males and 1 female, aged between 40 and 70 years, all of whom presented with complex medical histories. Among the 10 patients, 8 had VIM-1-IMP-45-CRPA detected in their bronchoalveolar lavage fluid (BALF), of whom 7 received ceftazidime/avibactam combined with aztreonam (CZA+AZT) treatment. Another 2 patients had VIM-1-IMP-45-CRPA detected in their blood but did not receive CZA+AZT treatment. Among patients who received CZA+AZT treatment, the mortality rate was 12.5% (1/8), while among patients who did not receive CZA+AZT treatment, the mortality rate was 100% (3/3), with 66.7% (2/3) of the deaths associated with bloodstream infections ( Table 1 ). Through phylogenetic analysis of 63 ST313 genomes, (46 from the NCBI Reference Sequence database and 17 from this study), using the midpoint rooting method, we found that ST313 P. aeruginosa can be divided into two subgroups ( Figure 1 ), revealing contrasting evolutionary trajectories and transmission dynamics. Subgroup 2 demonstrated tight clonal clustering with multiple closely related strains predominantly originating from China during 2017-2024, indicating recent clonal expansion and suggesting a single-source outbreak with rapid person-to-person transmission capabilities. In contrast, subgroup 1 exhibited greater phylogenetic diversity with strains distributed across multiple countries and spanning a broader temporal range (1997-2022), reflecting longer evolutionary timescales and endemic circulation patterns. The transmission timeline analysis revealed two distinct epidemiological phases: a recent epidemic emergence in subgroup 2 concentrated within a 2-year period, and historical global circulation in subgroup 1 spanning over two decades. Notably, the evolutionary relationship analysis demonstrated that while Subgroup 1 maintained ancestral ecological flexibility with 27.5% (11/40) environmental or animal isolates ( Supplementary Table S3 ) distributed throughout the phylogeny, subgroup 2 exhibited marked anthropocentric specialization with only a single animal isolate among an otherwise exclusively human-derived population.

All 17 VIM-1-IMP-45-CRPA strains isolates shared almost identical resistance genes, differing by 0–192 SNPs. Except for strain CJ05, the differences among the remaining strains range from 0 to 79 SNPs, demonstrating a high degree of relatedness among those isolates and clonal transmission of -producing P. aeruginosa ST313 in China ( Supplementary Figure S1 ). Notably, all 17 ST313 VIM-1-IMP-45-CRPA strains possessed virulence genes similar to the highly virulent strain PA14, including the type III secretion system genotypes exoU+, exoS+ and exoY+( Supplementary Figure S2 ).

Drug susceptibility testing showed that, except for the CJ13 and CJ04, intermediate and sensitive to aztreonam, respectively, all ST313 VIM-1-IMP-45-CRPA strains were resistant to carbapenems (meropenem and imipenem), several β-lactams (ceftazidime, cefepime, piperacillin/tazobactam, cefoperazone/sulbactam, ceftazidime/avibactam, aztreonam), quinolones (ciprofloxacin, levofloxacin), and aminoglycosides (amikacin, tobramycin). But all strains remained sensitive to colistin and cefiderocol ( Table 2 ).

Interestingly, almost all aztreonam-resistant VIM-1-IMP-45-CRPA strains (14/15) harbored a mistranslation mutation (P527S) in PBP3. The non-mutated PBP3 protein was consistent with the PBP3 protein of the standard strain PAO1.Based on the PBP3(PAO1) structure in the RCSB PDB database, we used Schrödinger to predict the structure of the point-mutated PBP3 (P527S) to analyze interactions with aztreonam. Subsequently, molecular docking of aztreonam with both PBP3(PAO1) and PBP3(P527S) was performed using Schrodinger software ( Supplementary Figure S4 ). The docking score of the PBP3(P527S) protein decreased (docking score: -9.911kcal/mol to -8.770kcal/mol). Subsequently the MM/GBSA binding free energy from the top docking pose of aztreonam to PBP3 was calculated. From our analysis, PBP3(P527S) exhibited less negative binding free energy values for aztreonam (-21.36 kcal/mol for PBP3(PAO1), -20.06 kcal/mol for PBP3(P527S)), suggesting the weakened binding affinity to PBP3(P527S).

To assess plasmid transferability, we selected strain CJ05 for further experiment. CJ05 strain contains an IncP-2 megaplasmid co-harboring bla _VIM-1 and bla _IMP-45 by whole-genome sequencing. Given all VIM-1-IMP-45-CRPA strains were resistance to amikacin but sensitive to cefiderocol, we chosen a clinical amikacin-susceptible P. aeruginosa strain AS01 for cefiderocol induction resistance experiments. An cefiderocol induced resistance P. aeruginosa strain, AS02, was used for conjugation experiments. The AS01 strain was identified as ST463 P. aeruginosa. The whole-genome sequencing revealed that it harbors a plasmid of 41,101 bp in size carrying the bla _KPC-2 gene. Other resistance genes, including aph(3’)-IIb, bla _PAO, bla _OXA-486, fosA, catB7, and crpP, are located on the chromosome. Antibiotic susceptibility testing showed that the strain is resistant to carbapenems (meropenem and imipenem), several β-lactams (ceftazidime, cefepime, piperacillin/tazobactam, cefoperazone/sulbactam, ceftazidime/avibactam, aztreonam), and quinolones (ciprofloxacin, levofloxacin). However, it remains sensitive to aminoglycosides (amikacin, tobramycin), colistin, and cefiderocol ( Supplementary Table S2 ). By analyzing the genome of the AS02 strain, we found that the bla _KPC-2 gene on its plasmid had evolved into bla _KPC-33, while other resistance genes remained identical to those in the AS01 strain. The antibiotic susceptibility profile of AS02 was similar to that of AS01, except that AS02 exhibited resistance to cefiderocol ( Supplementary Table S2 ). Additionally, we observed mutations in certain genes in the AS02 strain, including hasR. Furthermore, conjugation experiments were conducted using E. coli J53 as the recipient strain. Conjugation experiments showed that the IncP-2 plasmid could be transferred within P. aeruginosa, but not to E. coli I53. Using Platon to analyze the plasmid sequence of the transconjugant TC01 revealed the presence of resistance genes, including aac(6’)-Ib3, armA, bla _KPC-33, bla _IMP-45 and bla _VIM-1. Drug susceptibility test showed that it was resistant to amikacin and cefiderocol, but only sensitive to colistin ( Supplementary Table S2 ).

Using Platon predict plasmids in transconjugant TC01 and all ST313 VIM-1-IMP-45-CRPA strains except CJ05. Phylogenetic analysis of the system shows high homology in the IncP-2 plasmids of ST313 VIM-1-IMP-45-CRPA and transconjugant TC01 in the VIM-1-IMP-45-CRPA plasmids ( Supplementary Figure S3 ).

Third-generation sequencing reveals that the genome of the CJ05 strain contains a chromosome and an IncP-2 megaplasmid ( Figure 2A ). This plasmid carries bla _VIM, bla _IMP, and other resistance genes like tmexCD3-toprJ3. The bla _VIM-1 gene is located on a Tn3 family transposon ( Figure 2C ), including a complete Tn402 module. The genetic background of bla _IMP-45 is similar to previously reported IncP-2 plasmids carrying bla _IMP-45( Supplementary Table S1 ), situated on a Tn6485-like transposon ( Figures 2B and 3 ). The bla _IMP-45 gene is found directly downstream of the transposable element TnAs1 and within a class 1 integron, In786, with the gene arrangement intI1-aacA4-blaIMP-45-blaOXA-1-catB3. In786 is located within a Tn6485-like transposon. Similar genetic contexts have been detected or reported in several other IncP-2 plasmids, including pPA166-2-MDR, pR31014-IMP, pNF143349, and pPAHT-1 ( Figure 2B ). Subsequent snp-based phylogenetic analysis of the bla _IMP-45-carrying IncP-2 plasmid revealed that pCJ05 is homologous to other bla _IMP-45-carrying IncP-2 plasmids ( Figure 3 ). Based on this observation, we hypothesize that pCJ05 acquired bla _VIM-1 through horizontal gene transfer (HGT) from a bla _IMP-45-carrying IncP-2 plasmid. Further analysis of the genetic context surrounding bla _VIM-1 revealed direct evidence of transposition. Specifically, we identified both direct repeat (DR) and inverted repeat (IR) sequences flanking bla _VIM-1. We discovered that bla _VIM-1 is carried by a Tn402-like transposon, which subsequently inserted into a Tn3-family transposon harboring partial mer operon genes, resulting in the formation of a novel Tn3-family transposon ( Figure 4 ).The Tn402-like transposon was found to carry a complete tniABQR transposition module, while the Tn3-family transposon with partial mer operon contained tnpA but lacked tnpR. Comparative analysis using the NCBI database identified two plasmids carrying these transposons, pVIM-1-ZDHY316(CP064944) and pTTS12(CP009975), respectively. Based on these findings, we propose an evolutionary trajectory for the co-harboring bla _VIM-1 and bla _IMP-45 IncP-2 plasmid. First of all, the Tn402-like transposon carrying bla _VIM-1 transposed into the Tn3-family transposon containing partial mer operon genes, forming a novel bla _VIM-1-carrying Tn3-family transposon. This composite transposon was subsequently mobilized onto the bla _IMP-45-harboring IncP-2 plasmid, ultimately resulting in the emergence of an IncP-2 plasmid co-harboring both bla _VIM-1 and bla _IMP-45 ( Figure 4 ).