Fifty seven non-duplicate CRKP clinical isolates from urine (n = 6), sputum or bronchoalveolar lavage fluid (n = 35), abscess (n = 2), blood (n = 5), pus (n = 2), abdominal fluid (n = 3), and other patient samples (n = 6) were collected from Shandong Provincial Hospital of China, between January 2017 and June 2020. Matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF/MS) (BioMérieux, Marcy-l’Étoile, France) was used to identify the isolates. Carbapenemases were detected using carbapenem inactivation method (CIM) and EDTA-modified CIM (eCIM).

Antibiotic susceptibility was analyzed with a VITEK-2 compact system (BioMérieux, France) for aztreonam (ATM), cefepime (FEP), ceftriaxone (CRO), ceftazidime (CAZ), ertapenem (ETP), imipenem (IMP), piperacillin-tazobactam (TZP), trimethoprim-sulfamethoxazole (SXT), ciprofloxacin (CIP), levofloxacin (LVX), gentamicin (GEN), and amikacin (AMK); broth microdilution for polymyxin B (POL), tigecycline (TGC); and agar dilution for fosfomycin (FOS), using Mueller–Hinton agar supplemented with 25 μg/ml of G6P (CLSI, 2020). Susceptibility assay results were interpreted by CLSI breakpoints (CLSI, 2020), except for TGC, which were defined by the European Committee on Antimicrobial Susceptibility Testing (EUCAST, 2020) guidelines (EUCAST, 2020). Fosfomycin susceptibility was interpreted according to CLSI breakpoints for E. coli urinary isolates. Phenotypic detection of carbapenemases was performed using the CIM and eCIM tests (CLSI, 2020).

Genomic DNA from clinical strains embedded in gel plugs was digested with QuickCut XbaI (Takara, Shiga, Japan), and restriction fragments, ranging from 50 to 500 kb, were separated using CHEF Mapper apparatus (Bio-Rad, Hercules, CA, United States) for 19 h with the pulse time switched from 6 to 36 s. Pulsed-field gel electrophoresis (PFGE) patterns were compared using Gel-J software, version 2.0 (Heras et al., 2015). Pulsotypes were assigned to the clusters with 80% similarity (Chen et al., 2019).

Conjugation experiments were carried out with sodium azide-resistant E. coli J53Azi^R being used as the recipient. Transconjugants harboring fosfomycin resistance genes were selected on Mueller–Hinton agar plates containing 64 mg/ml of fosfomycin, 100 mg/ml of sodium azide, and 25 μg/ml of G6P (Xiang et al., 2015). Transconjugants harboring carbapenemase resistance genes were selected on Mueller–Hinton agar plates containing 6 μg/ml of CAZ and 100 mg/ml of sodium azide. Antibiotic susceptibility test and PCR analysis were performed to confirm the fosA3 and/or carbapenemase gene transfer (Poirel et al., 2011; Xiang et al., 2015). Furthermore, PCR-based replicon typing (PBRT) was used to characterize the plasmid harbored by the transconjugants (Carattoli et al., 2005).

DNA from clinical isolates was extracted and sequenced using an Illumina Hiseq platform at Novogene Co., Ltd. (Beijing, China). Illumina sequences were assembled de novo using the SPAdes v3.10 (Nurk et al., 2013).

For the JNKPN52 and JNKPN30 isolates, genome sequencing was also performed on a PacBio RSII sequencer at Biozeron Biological Technology Co., Ltd. (Shanghai, China). The paired-end short Illumina reads were used to correct the long PacBio reads utilizing proovread, and then the corrected PacBio reads were assembled de novo utilizing SMARTdenovo.¹ Sequence annotation was conducted using RAST 2.0² combined with BLASTP/BLASTN searches against the UniProtKB/Swiss-Prot and RefSeq databases. Annotation of resistance genes and mobile elements was carried out using the online databases, including CARD³ and ISfinder.⁴

Antimicrobial resistance genes and multilocus sequence typing (MLST) were analyzed in silico by using Abricate software⁵ (Sherry et al., 2019). Virulence scores, capsular (K) serotypes, and lipopolysaccharide (LPS) O antigen serotype were predicted using Kleborate v0.3.0⁶ (Wyres et al., 2020). Single-nucleotide polymorphism (SNP) calling was performed using Snippy 3.1,⁷ and recombinant variants were excluded using ClonalFrameML 1.0 (Lu et al., 2019). Maximum likelihood phylogenetic trees were constructed with RAxML,⁸ from the recombination-free SNPs.

For our dataset, core-genome MLST (cgMLST) analysis was performed using SeqSphere+ software (8.0.2 version; Ridom, Münster, Germany) according to the ‘‘K. pneumoniae sensu lato cgMLST’’ version 1.0 scheme⁹ (Weber et al., 2019). A total of 2,358 target genes were used to characterize the gene-by-gene allelic profile of the K. pneumoniae strains. The resulting set of target genes was then used for interpreting the clonal relationship displayed in a minimum spanning tree using the “pairwise ignoring missing values” parameter during distance calculations.

The fosfomycin resistance-related proteins MurA, GlpT, and UhpT of the genomes were aligned with K. pneumoniae reference strain ATCC 700721 using local BLAST software. Multiple sequence alignments were performed by MAFFT, with the 4-kb fosA^kp gene-related fragment of JNKPN10 as a reference.

The complete sequence of pJNKPN52_KPC_fosA has been deposited in GenBank under accession number MZ709016. Eleven fully sequenced pCT-KPC-like plasmids harboring bla_KPC–2 were compared with pJNKPN52_KPC_fosA by BLAST Ring Image Generator,¹⁰ including pCT-KPC (GenBank accession no. KT185451), p69-2 (GenBank accession no. CP025458), p1068-KPC (GenBank accession no. MF168402), p20049-KPC (GenBank accession no. MF168404), p675920-1 (GenBank accession no. MF133495), pC2414-2-KPC (GenBank accession no. CP039820), pKP1034 (GenBank accession no. NZ_KP893385), pKSH203-KPC (GenBank accession no. CP034324), p16HN-263_KPC (GenBank accession no. CP045264), pEBSI036-2-KPC (GenBank accession no. MT648513), and pKP19-2029-KPC2 (GenBank accession no. CP047161). More detailed genome alignment between closely related plasmids was conducted by local BLAST and visualized with Easyfig.¹¹

The complete sequence of pJNKPN30_NDM has been deposited in GenBank under accession number OL389795. Four fully sequenced IncA/C type plasmids harboring bla_NDM–1 were compared with pJNKPN30_NDM by BLAST Ring Image Generator (see text footnote 10), including pNDM_KN (GenBank accession no. JN157804), pMS6198A (GenBank accession no. CP015835), p1605752AC2 (GenBank accession no. CP022126), and pT1 (GenBank accession no. KX147633).

Among the 57 tested CRKP strains, 40 were resistant to fosfomycin. High resistance (> 70%) was observed against β-lactam antibiotics, fosfomycin, and quinolones. The highest resistance (> 98%) was observed against TZP, CRO, FEP, and ETP. All the strains remained 100% susceptible to TGC, and only one isolate showed resistance to POL [minimal inhibitory concentration (MIC) ≥ 64]. The antibiotic susceptibility results are shown in Table 1 and Supplementary Table 4.

The 57 K. pneumoniae strains had 18 sequence types (STs), with ST11 being the most common (n = 36, 63%) (Figure 1 and Supplementary Table 2). The others were ST101 (n = 2), ST15 (n = 2), ST24 (n = 2), ST37 (n = 2), ST1031, ST133, ST152, ST1537, ST2246, ST25, ST258, ST29, ST323, ST392, ST3924, ST485, and ST528.

Among these strains, we detected 19 different K-loci, the most common being KL64 (n = 26, including 25 ST11) and KL47 (n = 10), together accounting for 63% of all the strains (Figure 1). Seven distinct O antigen encoding loci were detected among the strains, and the most common were O2 (n = 30), OL101 (n = 11), and O1 (n = 9).

According to PFGE profile, seven different clusters as A∼G clone groups and five singletons were identified. The phylogenetic tree revealed that CRKP strains could be broadly clustered into three major clades: clades 1 and clades 2 consisted of ST11 strains alone, while clade 3 consisted of ST11 and the other STs (Figure 2). A minimum spanning tree of the 57 K. pneumoniae isolates was constructed based on cgMLST allelic profiles, showing the presence of five cluster types (≤ 15 allele differences). ST11-KL64-wzi64-O2 isolates mainly belonged to Cluster 1 (n = 17) and Cluster 3 (n = 6), while ST11-KL47-OL101 isolates mainly belonged to Cluster 2 (n = 7) (Figure 3 and Supplementary Table 5).

All the 57 strains harbored at least one fosfomycin-modifying enzymes, including fosA2, fosA3, foskp96, and fosA^kp (Figure 2). fosA^kp was the most prevalent chromosomal-encoded enzyme, detected in 52 isolates, followed by foskp96, in five isolates. fosA^kp and fosA3 were detected in 23 isolates. foskp96 was detected among 5 isolates, though only one isolate showed fosfomycin-resistant phenotype in vitro.

Among the isolates, 3 isolates had Ser148Asn and Ser209Thr substitutions in MurA, and two isolates had MurA deletion. One variation, Val434Ile, in uhpT was detected in two isolates. Three substitutions in glpT, Ile260Val, Val337Ile, and Ile429Val, were detected in one isolate.

As illustrated in Figure 1 and Supplementary Table 1, 87.7% of the isolates produced carbapenemase in accordance with the results of CIM test. bla_KPC–2 was the main type of carbapenemase (n = 33). Five strains coharboring bla_KPC–2 and bla_NDM–1 were detected. Meanwhile, bla_NDM–1 (n = 5), bla_NDM–5 (n = 3), bla_NDM–3 (n = 1), bla_IMP–4 (n = 2), and bla_OXA–232 (n = 1) also contributed to the carbapenem resistance. Twenty transconjugants harboring carbapenemase genes were acquired. One transconjugant harbored bla_IMP–4, ten transconjugants harbored bla_KPC–2, and nine transconjugants harbored bla_NDM. ESBL resistance genes, such as bla_TEM, bla_CTX–M, bla_SHV, bla_CMY, and bla_SHV, were also detected, with bla_TEM, bla_CTX–M, and bla_SHV being the most prevalent. The strains coharboring bla_TEM, bla_CTX–M, and bla_SHV made 52.6% of all the CRKP strains (Figure 4 and Supplementary Table 1).

Plasmid-encoded fluoroquinolone resistance genes (QnrS and QnrB) were detected in 18 isolates, while chromosomal oqxA and oqxB were detected in 33 isolates (Figure 4 and Supplementary Table 1). A high prevalence of specific porin defects was detected, and only 14 isolates were without any mutation in ompK5 or ompK36. It was observed that ompK35 truncations and ompK36GD mutations coexisted in 59.6% of the strains (Figure 4 and Supplementary Table 2).

According to the Katholt criterion, 27 isolates were assigned a virulence score of 4, which were closely related to ST11 (n = 24) (Figure 4). Analysis of virulence genes showed that 41 isolates possessed yersiniabactin genes located on ICEKp3 (n = 38), ICEKp1 (n = 1), ICEKp5 (n = 1), and ICEKp12 (n = 1), with ICEKp3 being the most prevalent carrier. Of 25 ST11-KL64 isolates, 22 (88%) carried rmpA2 gene, but only three were positive for the string test, suggesting that rmpA2 was inactive in most isolates. Among 10 ST11-KL47 isolates, only one was positive for rmpA2 gene (Supplementary Table 2).

Four types of genetic environments existed in the 52 fosA^kp-positive CRKP isolates. The upstream genes of fosA^kp among the strains are identical, but those downstream are variable. The intergenic regions between fosA^kp and the downstream MocR gene could be DNA helicase-related genes, a Type I restriction–modification system, or a hypothetical Protein. As shown in Supplementary Figure 1, the genetic environment of foskp96 was similar to that of fosA^kp, with a backbone of YrkL-LysR-FosA-MocR-YjiS. The genetic environment of fosA3 was consistent, where fosA3 is flanked by IS26 at both ends, in a transposon-like structure.

The plasmids harboring fosA3 from strains JNKPN52, JNKPN54, JNKPN55, and JNKPN57 were successfully transferred into E. coli J53Azi^R by conjugation. All the four transconjugants were resistant to CRO, FEP, CAZ, ATM, TZP, ETP, IMP, AMK, GEN, and FOS but were susceptible to SXT, CIP, LVX, TGC, and POL (Supplementary Table 3). JNKPN52 and JNKPN54 were isolated from 2-month-old pediatric patients after cardiac surgery enrolled in the cardiac care unit in December 2019. JNKPN55 and JNKPN57 were isolated from premature babies enrolled in the neonatal intensive care unit in December 2019. A nosocomial outbreak caused by a clone of ST-KL47 KPC-KP strains was considered according to the high similarities of PFGE patterns of group G3 (> 85%) (Figure 1) and the high level of correlation within cgMLST Cluster 2 (up to 1 allele difference) (Figure 3).

The complete sequence of plasmid pJNKPN52_KPC_fosA from clinical strain JNKPN52 was determined to better characterize the self-transmissible plasmid coharboring fosA3 and bla_KPC–2. pJNKPN52_KPC_fosA is a 127,668-base pair (bp) multireplicon plasmid that belongs to the IncR:IncFII-type and shares a similar structure with pHN7A8/pKPC-LK30 hybrid plasmid pCT-KPC (Figure 5).

pJNKPN52_KPC_fosA contains two major accessory resistance regions, including the bla_KPC–2 region harboring bla_KPC–2 and bla_SHV–12, and the multidrug-resistant (MDR) region carrying rmtB (aminoglycoside resistance), fosA3, bla_TEM–1, and bla_CTX–M–65. The MDR region was generated from the insertion of ΔTn6377–bla_CTX–M–65, IS26–fosA3–IS26 unit, ΔTn2-rmtB element within IS1. The bla_KPC–2 region was organized in order of a truncated IS26–bla_SHV–12–IS26 unit, ΔTn6296, ΔTn21 with insertion of IS5075, an IS903B remnant, and ISKpn14 (Figure 5).

According to sequence alignment by BRIG, pJNKPN52_KPC_fosA showed 99% nucleotide identity with the previously reported plasmids p16HN-263_KPC and pKP19-2029-KPC2 isolated from China and pEBSI036-2-KPC isolated from Egypt (Ahmed et al., 2021) (Figure 5).

The MDR regions of the eleven plasmids were similar, with pJNKPN52_KPC_fosA, p675920-1, p69-2, and p20049-KPC slightly differing from one another. To determine the detailed structural differences between these plasmids, additional linear comparative genomics analysis was performed by BLAST. Compared with pJNKPN52_KPC_fosA, p675920-1 lacked a ΔIS1294 region, possibly because of recombination of IS26–fosA3–IS26 region. In p20049-KPC, the deletion of IS26–fosA3–IS26 region and insertion of partial plasmid backbone genes were observed within the MDR region (Figure 6).

The bla_KPC–2 region of the pJNKPN52_KPC_fosA was similar to that of p20049-KPC and pKP1034, with the inversion of ΔTn6296. Compared with pJNKPN52_KPC_fosA, the deletion of the truncated IS26–bla_SHV–12–IS26 unit was observed in p675920-1, probably due to IS26-mediated deletion (Figure 7).

The plasmids harboring bla_NDM from strains JNKPN23, JNKPN24, JNKPN26, JNKPN29, JNKPN30, JNKPN31, JNKPN46, JNKPN47, and JNKPN51 were successfully transferred into E. coli J53Azi^R by conjugation. All the nine transconjugants were resistant to CRO, CAZ, TZP, ETP, and IMP, but were susceptible to FOS, CIP, and LVX (Supplementary Table 6). All the plasmids harboring bla_NDM–1 were shown to belong to IncA/C2 type through plasmid typing (Supplementary Table 6).

The complete sequence of plasmid pJNKPN30_NDM from clinical strain JNKPN30 was determined to better characterize the self-transmissible IncA/C2 type plasmid harboring bla_NDM–1. pJNKPN30_NDM contains two main accessory resistance regions, including the IS1380–bla_CMY–6 region and the MDR region carrying Tn6196, class 1 integron structure bearing arr-3, dfrA1, AadA16, ErmE, and sul1, and partial of Tn125-bearing bla_NDM–1 interrupted by the insertion of ΔISKpn14, followed by the 16S rRNA methylase rmtC gene (Figure 8).

According to sequence alignment by BRIG, pJNKPN30_NDM was with a backbone similar to that of other IncA/C2 plasmids (84–99% query coverage, > 99% nucleotide sequence identity). The main regions of discontinuity were within the Int498 region and the ISKpn18 region (Figure 8).

Raw reads of all 57 isolates have been deposited in GenBank (BioProject PRJNA769451). The complete sequence of pJNKPN52_KPC_fosA and pJNKPN30_NDM has been deposited in GenBank under accession numbers MZ709016 and OL389795, respectively.