Four bla_KPC−2- and bla_NDM−1-coharboring CRCF strains were collected during our surveillance of the prevalence of carbapenem-resistant Enterobacteriaceae isolates. Species identification was determined using matrix-assisted laser desorption ionization–time of flight mass spectrometry (MALDI-TOF MS) (Bruker Daltonik, Bremen, Germany), as described previously (Lu et al., 2021).

In vitro susceptibility tests were performed using N335 susceptibility cards and the Vitek-2 system (bioMérieux, France), including amikacin, minocycline, doxycycline, ceftazidime, cefepime, piperacillin/tazobactam, aztreonam, levofloxacin, ciprofloxacin, and sulfamethoxazole/trimethoprim. The minimal inhibitory concentrations (MICs) of imipenem, meropenem, tigecycline, colistin, ceftazidime–avibactam, and aztreonam–avibactam were determined using the microdilution broth method (bio-KONT, Ltd. China) with E. coli ATCC 25922 as the quality control strain, as we described previously (Zhang et al., 2023). The breakpoint of tigecycline was defined by the U.S. Food and Drug Administration (FDA) (Marchaim et al., 2014). The results of other antimicrobial agents were interpreted following the standards of the Clinical Laboratory Standards Institute (CLSI, 2021) (2021). The production of carbapenemases was determined using the modified carbapenem inactivation method (mCIM) and the EDTA-modified carbapenem inactivation method (eCIM), as recommended by the CLSI 2021 (2021).

As we described previously (Zhang et al., 2023), WGS was performed using the Illumina HiSeq 2500 platform and the nanopore sequencing method on MinION flow cells (for CPCF strains CF2075, CF2084, CF2085, and CF26019 and the transconjugant E. coli J53_K-N, respectively). Raw reads were filtered to remove the low-quality sequences and adaptors using skewer (Jiang et al., 2014) and Porechop (https://github.com/rrwick/Porechop), respectively. De novo assembly was performed via the SPAdes Genome Assembler v3.13.1 (Prjibelski et al., 2020) and Unicycler (Wick et al., 2017). Gene prediction for CRCF genomes, including four from this study and three retrieved from the NCBI genome database, was performed using Prokka 1.12 (Seemann, 2014). Insertion sequences were identified using the ISfinder database (Siguier et al., 2006). The antimicrobial resistance genes, multilocus sequence types (MLSTs), and plasmid replicon were analyzed via the CGE server (https://cge.food.dtu.dk/services/). Single-nucleotide polymorphisms (SNPs) were determined using Snippy (https://github.com/tseemann/snippy). Homology analysis was performed using BLASTn, FastANI, and SNP. Linear alignments of bla_KPC−2- and bla_NDM−1-bearing structures were generated using genoPlotR and gggenes in R-4.1.2. The transferability of bla_KPC−2 and bla_NDM−1-carrying plasmids in four CRCF strains was evaluated using oriTfinder (Li et al., 2018).

Plasmid conjugation experiments were conducted for four CRCF strains, as described previously (Zhang et al., 2023). Azide-resistant E. coli J53 and amikacin-resistant K. pneumoniae KP54 were used as the recipient strains. In brief, the four CRCF strains and recipient KP54 were adjusted to a McFarland standard of 0.5 and mixed at a ratio of 1:3, and a 0.1-milliliter aliquot of mixture was transferred into LB broth without antibiotics. After an 18-h incubation at 37°C, 200-ml cultures were streaked onto China blue agar (CBA, addition of rosolic acid as the pH indicator) plates containing both amikacin (16 mg/L) and meropenem (1 mg/L) to screen the bla_KPC−2− or bla_NDM−1-carrying transconjugants. Similarly, KP54_CF2075K-N and recipient E. coli J53 were mixed, and transconjugants carrying bla_KPC−2 or bla_NDM−1 were also selected on CBA plates containing both azide (150 mg/L) and meropenem (1 mg/L). The above transconjugants were confirmed by antimicrobial susceptibility testing (AST), polymerase chain reaction (PCR), and pulsed-field gel electrophoresis (PFGE), respectively (Zhang et al., 2023).

The plasmid stability and fitness cost were assessed as previously described but with slight modifications (Gao et al., 2020). The stability of CF2075, CF2084, CF2085, CF26019, and the transconjugants was evaluated by a passage experiment. In brief, the aforementioned strains were grown in LB broth and transferred at a 24-h interval for 10 consecutive days (approximately 200 generations), at a 1:1000 dilution, into fresh LB broth. The cultures at the 2nd, 4th, 6th, 8th, and 10th days were serially diluted and streaked onto the antibiotic-free LB agar. Approximately 50 colonies were randomly selected to identify the retention of bla_KPC−2 and bla_NDM−1 using PCR. All the above experiments were conducted in triplicate on different days.

Fitness cost was evaluated by growth curves. In brief, CF2085, CF2085ΔNDM (obtained in the above passage assay), CF26019, CF26019ΔNDM, the recipient KP54, and its transconjugants were cultured and shaken at 200 rpm overnight at 37°C in 10 mL LB broth. The overnight cultures were diluted and incubated at 37°C for 25 h to measure the optical density values (OD₆₀₀). The experiment was repeated two times. The growth curves were estimated using Tukey's multiple comparison tests with a one-way analysis of variance (ANOVA).

To better unravel the plasmidome of bla_KPC−2- or bla_NDM−1-harboring plasmids in C. freundii, we searched the RefSeq database on NCBI and obtained the intact plasmids harboring bla_KPC−2 (42) or bla_NDM−1 (21) in C. freundii worldwide as of 20 November 2022. Blasting was performed with BLASTn and illustrated with the R ggplot2 package.

A phylogenetic tree from relevant plasmids was built using PhyML (Guindon et al., 2010), under model GTR running 1,000 bootstrap replicates, from the alignment results generated by mafft (Kuraku et al., 2013). The maximum-likelihood phylogenetic tree of bla_KPC−2- and bla_NDM−1-coharboring CRCF strains in this study with three complete genomes was built using RaxML (Stamatakis, 2014), running 1,000 bootstrap replicates under model GTR-G, from the alignment generated by SNPs and filtered to remove recombination using Gubbins v2.4.1 (Croucher et al., 2015). Visualization was performed using iTOL (https://itol.embl.de).

Data analyses were performed using GraphPad Prism 8.2.1. One-way analysis of variance was used for assessing significant differences, with a P-value of < 0.05 being considered statistically significant.

From 11 September 2020 to 12 October 2020, three bla_KPC−2- and bla_NDM−1-coharboring CRCF strains (CF2075, CF2084, and CF2085) were recovered from urine specimens of three inpatients with congenital hydronephrosis post-operation or urinary infection (aged 8 months to 9 years) at the same urological ward of a tertiary hospital in Henan, China (Table 1). The CRCF strain CF26019 was isolated from the renal perfusion fluid of Patient 4 with chronic kidney disease at another tertiary hospital in Beijing, China, after kidney transplantation (Table 1). The four patients all finally recovered from urological diseases.

Multilocus sequence typing (MLST) revealed that all CRCF strains, CF2075, CF2084, and CF2085, belonged to ST523. The strains CF2084 and CF2085 had >99% coverage and identity to CF2075 by BLASTn and FastANI, >99% vice versa. Further analysis revealed that there were only 6 and 18 single-nucleotide polymorphisms (SNPs) between CF2084 and CF2075 and CF2085 and CF2075, respectively. Moreover, Xbal and S1-PFGE also revealed that CF2075, CF2084, and CF2085 had identical bands (Supplementary Figures S1A, B). Overall, PFGE and WGS both indicated that there was a nosocomial outbreak of ST523 CRCF strains through clonal dissemination.

The CRCF strains CF2075, CF2084, CF2085, and CF26019 showed resistance to imipenem, meropenem, ceftazidime, ceftazidime–avibactam, aztreonam, piperacillin/tazobactam, and cefepime but were susceptible to amikacin, tigecycline, colistin, and aztreonam/avibactam for limited treatment options (Table 2). WGS revealed they all carried bla_KPC−2 and bla_NDM−1, in line with resistance phenotypes. Moreover, CF26019 also harbored aac(3)-IIa and aac(6′)-Ib-cr, resulting in resistance to tobramycin.

Due to the high homology and clonal spread of the strains of CF2075, CF2084, and CF2085, only CF2075 was chosen for further analysis in detail. WGS and S1-PFGE revealed that CF2075 consisted of a chromosome (5,013,357 bp and 52% GC content) and two plasmids designated as pCF2075-1 (46,049 bp and 56% GC content; identical plasmids pCF2084-1 and pCF2085-1 were identified in CF2084 and CF2085, respectively) and pCF2075-2 (81,274 bp and 53% GC content; identical plasmids pCF2084-2 and pCF2085-2 in CF2084 and CF2085, respectively), classified as IncR and IncFII incompatibility groups, respectively (Supplementary Figures S1A, B). Interestingly, pCF2075-1 concomitantly carried bla_KPC−2 and bla_NDM−1, and the genetic surroundings of the two genes were also contiguous. The bla_KPC−2-harboring genetic environment was Tn1721-klcA-korC-ΔISKpn6-bla_KPC−2-ISKpn27-Tn3 (13,760 bp), which is particularly prevalent in China (Figure 1A). It is likely that the bla_NDM−1-harboring genetic context was ISCR1-dsbD-trpF-ble-bla_NDM−1-ΔISAba125-Δsul1 (5,250 bp). The bla_NDM−1 gene flanked by ISCR1 was rarely reported before (Figure 1A). No resistance gene was detected in pCF2075-2.

To our knowledge, we have first reported the co-existence of bla_KPC−2 and bla_NDM−1 located on one plasmid (pCF2075-1) in C. freundii. The plasmid pCF2075-1, an IncR plasmid, exhibited low coverage (< 82.0%) compared to these plasmids on NCBI. Therefore, pCF2075-1 is a previously unreported fusion plasmid. To decipher the forming mechanism of pCF2075-1, its structure was analyzed in detail. It consisted of pA1732-KPC (63.5%, MT810373.1), the pM218A-NDM-1 gene cluster (11.4%, KX687851.1), and pB72B (22.4%, CP069803.1). The pA1732-KPC was contained in K. pneumoniae A1732 and harbored the same replicon IncR and bla_KPC−2-harboring genetic context as pCF2075. Consequently, pA1732-KPC could be considered the backbone of pCF2075-1. However, the genetic environment of bla_KPC−2 of pCF2075-1 did not harbor IRL-2 (Tang et al., 2022), which could embed bla_KPC−2 into transposon Tn1721 for transfer. This transposon is non-transferable by carrying bla_KPC−2. The pM218A-NDM-1 in Escherichia coli harbored bla_NDM−1, flanked by ISCR1 of a class 1 Integron. ISCR1 can generate a transposable element (TE) mediating the transfer of bla_NDM−1. It implied that the bla_NDM−1-harboring TE was integrated into the backbone of pCF2075-1 to generate the bla_KPC−2− and bla_NDM−1-co-carrying cointegrate (Figure 1B). Furthermore, the other two convergences of bla_KPC−2 and bla_NDM−1 in the K. pneumoniae strain CHS5 chromosome (CP110688.1) and the K. michiganensis strain K254 plasmid pK254-KPC_NDM (OM938013.1) were retrieved from NCBI (Figure 2A). The bla_KPC−2-harboring genetic contexts of the CHS5 chromosome and pK254-KPC_NDM were similar to Tn6296, which originated from the Tn1722-based transposon inserted by the core bla_KPC−2 region. Notably, the bla_NDM−1-harboring genetic contexts of both the CHS5 chromosome and pK254-KPC_NDM were contained in a class 1 Integron, in which the core bla_NDM−1 region was flanked by ISCR1. Collectively, a propensity for the co-existence of bla_KPC−2 and bla_NDM−1 was overwhelmingly dominated by the ISCR1-mediating transfer of bla_NDM−1 and the Tn1721/Tn6296-based-transposon-mediating transfer of bla_KPC−2.

In addition, CF26019 belonged to ST118 and contained a chromosome (5,177,371bp and 52% GC content) and four plasmids. Unlike CF2075, bla_NDM−1 and bla_KPC−2 of CF26019 were separately located on plasmid pCF26019-1 (44,169 bp and 56% GC content) and pCF26019-3 (35,420 bp and 57% GC content), belonging to IncFII/IncFIB and IncP6 incompatibility groups, respectively. Furthermore, pCF26019-1 showed a maximum query coverage of 73% with a nucleotide identity of 100% to the plasmids p205880-NDM of K. pneumoniae MH909345.1 and pK218-NDM of C. portucalensis OL988824.1. The genetic surrounding of bla_NDM−1 on pCF26019-1 was IS5-ΔISEhe3-groL-groS-cutA-dsbD-trpF-ble-bla_NDM−1-ΔISAba125(8,208 bp), which is a common genetic surrounding observed in bla_NDM−1-harboring Enterobacteriaceae strains. The pCF26019-3 exhibited 91% coverage and 100% identity to eight plasmids on NCBI, including CF121SC21 plasmid1 (LT992437.1) of C. freundii and plasmid pA1705-KPC (MH909348.1) of K. pneumoniae A1705. The genetic context of bla_KPC−2 on pCF26019-3 was ΔISEc33-Tn3-ISApu1-hp-ISApu2-ISKpn27-Δbla_{TEM − 1}-bla_KPC−2-ΔISKpn6-korC-klcA (13,066 bp), which commonly exists on different plasmids in Enterobacteriaceae (MN539620.1, MN477223.1, and MH909348.1) and Pseudomonas species (CP040685.1).

The transferability of bla_KPC−2 and bla_NDM−1 was evaluated by conjugation assays. The results revealed that bla_KPC−2 and bla_NDM−1 of CF2075, CF2084, and CF2085 could be concomitantly transferred to the recipient K. pneumoniae KP54, a clinical strain isolated from urine samples, and the transconjugants were designated as KP54_CF2075K-N, KP54_CF2084K-N, and KP54_CF2085K-N, respectively (Supplementary Figure S1B). Further conjugation experiments revealed that bla_KPC−2 and bla_NDM−1 of KP54_CF2075K-N could also be simultaneously conjugated into recipient E. coli J53, designated as E. coli J53_K-N (Supplementary Figure S1C). The above transconjugants were verified by resistance phenotype using PCR and PFGE. However, no transconjugant was obtained after co-culturing CF26019 with KP54, even though we repeated the conjugation assay dozens of times.

Unexpectedly, S1-PFGE showed that all the transconjugants, KP54_CF2075K-N, KP54_CF2084K-N, and KP54_CF2085K-N, contained a novel fusion plasmid co-carrying bla_KPC−2 and bla_NDM−1, named pCfr_tK-N, which is different from any plasmids of CF2075, CF2084, and CF2085 in size. However, the size of the pCfr_tK-N is almost equal to the sum of the size of pCF2075-1 and pCF2075-2; that is to say, pCfr_tK-N was probably generated by the recombination of pCF2075-1 and pCF2075-2. To unravel the forming mechanism of pCfr_tK-N, WGS was performed. Sequence analysis revealed that pCfr_tK-N was indeed the recombination result of pCF2075-1 and pCF2075-2 (Figures 3A, B), and the junctions of these two plasmids were two insertion sequence IS6100s. Of note, two 8-base-pair direct repeat (DR, ATGCTCAG) sequences were adjacent to the left inverted repeat sequence (IRL) of one IS6100 and the right inverted repeat sequence (IRR) of another IS6100, which implied that the recombination was mediated by IS6100 (Figure 3C). The plasmid stability of four CRCF strains and transconjugants was evaluated by passage experiment (Supplementary Figure S1E). Excisions of bla_NDM−1 strains were obtained from CF2085 and CF26019 on the 8th day, named CF2085ΔNDM and CF26019ΔNDM, respectively. However, S1-PFGE and the sequence analysis showed that bla_NDM−1-harboring genetic surroundings were excised from the plasmids of CF2085 and CF26019, which were inconsistent with the loss of bla_NDM−1 usually caused by the removal of plasmids harboring bla_NDM−1 in most cases (Figure 2B and Supplementary Figure S1B). A similar phenomenon was also observed in CF2075 and CF2084 on the 10th day of the passage experiment. However, the transconjugants (KP54_CF2075K-N, KP54_CF2084K-N, KP54_CF2085K-N, and E. coli J53_K-N) maintained 100% retention during the process of 10-day continuous passage, showing that the hybrid plasmids harboring bla_KPC−2 and bla_NDM−1 could be stably inherited. In summary, bla_KPC−2- and bla_NDM−1-coharboring CRCF strains and transconjugants can retain the stable inheritance of bla_KPC−2- and bla_NDM−1-co-carrying plasmids (>90% retention in the 10-day passage) that might slow the clearance of resistance genes and facilitate clonal dissemination.

Growth rates were measured to assess the impact of the acquisition of bla_KPC−2- and bla_NDM−1-coharboring plasmids on biological fitness cost. A significant growth difference (P < 0.0001) was shown between CF2085 and CF2085ΔNDM, KP54 and KP54_CF2075K-N, and E. coli J53 and E. coli J53_K-N, whereas no significant difference (P > 0.5) was observed in the growth rates between CF26019 and CF26019ΔNDM. Taken together, the bla_KPC−2- and bla_NDM−1-coharboring plasmids exhibited punishment of fitness in CF2075, CF2084, CF2085, and their transconjugants.

A total of 46 bla_KPC−2-harboring plasmids contained in C. freundii strains, including 42 retrieved from NCBI and 4 in this study, were analyzed via phylogenetic tree (Figure 4). These 42 C. freundii strains were recovered from urine, sediment, rectal swab, and blood and also from wastewater in the hospital environment and were gathered in Spain, the Czech Republic, China, and USA. Heterogeneous STs were identified in bla_KPC−2-harboring C. freundii, including ST8, ST18, ST22, ST65, ST118, ST257, ST259, ST523, and ST632. The C. freundii strains belonging to the same STs might carry both closely related plasmids (such as pCF2075-1, pCF2084-1, and pCF2085-1 or CP037739.1 and CP054297.1), largely implying a clonal spread, and distantly related ones (CP011608.1 and CP011656.1). Plasmid incompatibility groups revealed that the above closely related plasmids had the same replicon, including IncP6, IncR, and IncN. Furthermore, bla_KPC−2-harboring plasmids showed ≤ 52% and ≤ 59% coverage compared with the bla_KPC−2- and bla_NDM−1-co-carrying plasmids pCF2075-1 and pK254-KPC_NDM, respectively. Notably, most bla_KPC−2-harboring plasmids (43/46, 93.5%) exhibited ≥82.0% coverage at 100% identity to the plasmids contained in non-C. freundii Enterobacteriaceae. Similarly, bla_KPC−2-harboring plasmids (34/46, 74.0%) also exhibited ≥82.0% coverage to the plasmids of C. freundii.

Four bla_NDM−1-harboring plasmids in this study and twenty-one on NCBI were comprehensively analyzed (Figure 5). Their STs also showed diversity, including ST18, ST19, ST88, ST98, ST116, ST252, ST257, ST396, and ST523. Incompatibility groups mainly focus on IncX3 (10/25, 40.0%), IncC (5/25, 20.0%), and IncR (4/25, 16.0%), while identical replicons showed an obvious cluster (MK101346.1 and MH995506.1; CP097106.1 and CP055250.1; and pCF2075-1, p-CF2084-1, and pCF2085-1). However, a similar core bla_NDM−1 region was not always contained in a cluster. The bla_NDM−1-harboring plasmids were isolated from clinical species and environments in line with bla_KPC−2-harboring plasmids and were mainly from China (16/25, 64.0%) since 2010. In contrast with pCF2075-1 and pK254-KPC_NDM, all bla_NDM−1-harboring plasmids exhibited ≤ 55% coverage. To better evaluate the origin of bla_NDM−1-harboring plasmids in C. freundii, these plasmids were compared with those on NCBI. The results revealed that 72.0% (18/25) plasmids in C. freundii and 92.0% (23/25) plasmids in other Enterobacteriaceae shared >80% of their length with the aforementioned bla_NDM−1-containing plasmids, respectively.

To decipher the underlying mechanism in forming bla_KPC−2- and bla_NDM−1-coharboring CRCF strains, seven bla_KPC−2- and bla_NDM−1-coharboring CRCF strains having complete genomes, including three reported previously (Wu et al., 2016; Ouyang et al., 2018; Li et al., 2022) and four in this study, were analyzed (Figure 6). All were recovered in China after 2013 and could be divided into three clusters based on genome (cluster 1: CF26019; cluster 2: P10159 and WCHCF65; cluster 3: CF2075, CF2084, CF2085, and SCLZS47). STs showed an extensive distribution, including ST88, ST116, ST118, ST252, and ST523.

The bla_KPC−2-harboring plasmids in both P10159 and SCLZS47 were identical, although not grouped into the same cluster, and CF2075, CF2084, and CF2085 also contained the same plasmids. Therefore, four different bla_KPC−2-harboring plasmids were identified in the seven CRCF strains. These different plasmids, other than the type of pCF2075-1, have a high coverage of similar sequence (bi-directional ≥90%) in Enterobacteriaceae. For bla_NDM−1-harboring plasmids, three different plasmid types were identified. It is likely that the coverage of similar sequences among these plasmids was also distantly related (4.0–16.0%). Only the type of pP10159-1 (MF072961.1) had similar plasmids (bi-directional coverage and identity ≥99.9%) reported before in Enterobacteriaceae, and others were novel ones. Furthermore, P10159 and SCLZS47 were isolated from Chongqing and Sichuan in China, which are geographically close, and had identical bla_KPC−2- and bla_NDM−1-harboring plasmids, but the chromosomes of P10159 had 81.0% coverage of similar sequence at 95.94% to SCLZS47 and 80.0% the other way around, hinting at a horizontal gene transfer (HGT) by plasmids.