During March 2018–August 2019, seven kinds of samples including blood, ascitic fluid, sputum, bronchoalveolar fluid, wound secretion, urine, and ductus venosus of inpatients were collected from either public wards or intensive care units (ICUs) in a newly established hospital in Henan, China. The hospital that specializes in the treatment of cardiovascular diseases is a 1,000-bed tertiary hospital with 132 ICU beds and 34 public wards. Besides, the present study was approved by the Research Ethics Committee of Henan Provincial People’s Hospital.

Collected samples were subjected to standard bacterial isolation procedure. The samples were streaked directly onto 5% sheep blood agar plate. Colonies of different morphologies were selected to perform subsequent purification and stocked at −80°C. The detection of carbapenemase-encoding genes was conducted by multiplex polymerase chain reaction (PCR) (Poirel et al., 2011; Supplementary Table 1), and laboratory-stored strains carrying the corresponding carbapenemase-encoding genes were used as the positive control. Species identification of the strains and subsequent antimicrobial susceptibility testing were conducted by BD Phoenix100 (Becton, Dickinson and Company, Franklin Lakes, NJ, United States) and verified by disk diffusion method. The minimum inhibitory concentrations of ciprofloxacin, levofloxacin, aztreonam, chloramphenicol, ampicillin, ampicillin–sulbactam, piperacillin, piperacillin–tazobactam, amoxicillin–clavulanic acid, gentamicin, amikacin, cefazolin, ceftazidime, cefotaxime, cefepime, meropenem, imipenem, trimethoprim-sulfamethoxazole, and tetracycline were interpreted based on the standard of the Clinical and Laboratory Standards Institute [CLSI] (2018) except tigecycline and polymyxin B, which followed the criteria of European Committee on Antimicrobial Susceptibility Testing (version 11.0)³. Escherichia coli ATCC25922 was used as the quality control strain.

To preliminarily distinguish the STs and capsular types and confirm the presence of the virulence-associated genes including rmpA, rmpA2, iroN, and iutA, multiplex PCR analysis was performed as previously mentioned (Yu et al., 2018; Supplementary Table 1), and laboratory-stored strains carrying the corresponding genes were used as the positive control. Furthermore, the hypervirulence phenotype of K. pneumoniae was evaluated using string test and Galleria mellonella larvae infection assay. For string test, all isolates were inoculated onto 5% sheep blood agar and incubated at 37°C, and the cutoff criterion for positive was the viscous string longer than 5 mm (Shon et al., 2013). For G. mellonella larvae infection assay, larvae of approximately 300 mg were stored in a special box at 4°C until being used. Overnight cultures of K. pneumoniae were washed and adjusted to 10⁶ colony-forming units (CFU)/mL using phosphate-buffered saline (PBS). Ten larvae in each group were challenged with 10 μL of diluents, with ST11 clinical cKP HS11286 derivate YZ6 (Xie et al., 2018) used as the negative control. Infected larvae were incubated in sterilized Petri dishes at 37°C for 72 h, and survival rate was recorded every 24 h. All experiments were repeated in triplicate.

Transferability of carbapenem resistance phenotype was determined using conjugation assay with a filter mating method. Thirty-five CRKPs were used as donor strains, and K. pneumoniae YZ6 Hyg^r was served as the recipient strain. Transconjugants were selected on LB agar plates supplemented with hygromycin (200 mg/L) and meropenem (2 mg/L). The transconjugants harboring carbapenemase encoding genes were confirmed by PCR and antimicrobial susceptibility testing.

To investigate the fitness of CRKP isolates, growth curves of seventeen strains including 15 CRKP isolates in this study, YZ6 and ATCC700603 in LB broth were conducted according to standardized protocols using three technical replicates and three biological replicates. Klebsiella pneumoniae ATCC700603 and YZ6 were regarded as control strains (Schaufler et al., 2016). Growth rates were calculated as follows: μ = (ln(CFU/mL t1) – ln(CFU/mL t0))/t1 – t0.

We qualitatively detected siderophore secretion of CRKP isolates as previously described (Schwyn and Neilands, 1987). A single colony was transplanted into MKB solid medium for iron starvation treatment. After incubation for 24 h at 37°C, the bacterial suspension was adjusted to an OD600 of 0.6 by normal saline, and then 5 μL of suspension was placed on agar plates containing chrome azurol S-iron(III)-hexadecyltrimethylammonium bromide and incubated overnight at 37°C. The orange secretory ring around the colony indicated the production of siderophore. Klebsiella pneumoniae ATCC700603 and YZ6 were considered as control strains, and the experiment was repeated three times for each strain.

Biofilm formation assays were conducted as previously mentioned (Ma et al., 2020). Overnight cultures of tested isolates were adjusted to a cell density equivalent to a 0.5 McFarland standard. Two hundred microliters of culture per well were transferred to a 96-well plate. After incubation at 37°C for 2 days, cultures were discarded, and wells were washed twice with 200 μL PBS. The biofilms were fixed in methanol for 10 min. Subsequently, wells were stained with 1% crystal violet solution for 10 min and rinsed with PBS until colorless. Finally, biofilms were dissolved in 100 μL of 30% formic acid for 30 min, and biofilm formation was quantified by measuring the absorbance at OD590. Klebsiella pneumoniae ATCC700603 and YZ6 were used as control strains.

We evaluated the ability of human serum resistance as previously described (Heiden et al., 2020). Briefly, 5 μL of overnight culture was added to 495 μL LB fresh medium and incubated for 1.5 h at 37°C. Inoculum was resuspended with 1 mL of sterile 1 × PBS. Thirty microliters was mixed in triplicates with 270 μL 50% human serum in 96-well plates. Meanwhile, 30-μL mixture was sucked out from each well, serially diluted, placed on LB agar, and counted the next day. After incubation for 4 h at 37°C, 30 μL of mixture was subjected to the same procedure. Finally, the number of colonies of 0- and 4-h time points was compared to evaluate the survival ability of CRKP isolates in human serum. Klebsiella pneumoniae ATCC700603 and YZ6 were used as control strains.

Desiccation resilience assays were carried out according to previously methods (Heiden et al., 2020) with minor modified. Briefly, a single colony was cultured in LB broth until bacterial cells reached an OD600 value of 0.6–0.8. One hundred microliters of inoculum was serially diluted, plated on LB agar plates, and counted the next day. Meanwhile, another 100-μL inoculum was transferred to 96-well plates. Then, the plates were laid flat in a glass sterile dryer supplemented with desiccant and placed in a 37°C incubator. After 6 days of drying, 100 μL/well fresh LB broth was readded in 96-well plates; the prepared 96-well plates were cultured with 200-rpm shaking at 37°C for 3 h. At this point, 100 μL was collected, and the same procedure was performed to count the number of colonies. Klebsiella pneumoniae ATCC700603 and YZ6 were used as control strains.

The data were presented using GraphPad Prism 8.3.0. After ensuring that the data were non–normally distributed, the non-parametric Kruskal–Wallis test was utilized to perform multiple comparisons among different groups. Bonferroni adjustment was applied; the corrected p < 0.1 was considered significant.

Genomic DNA of the 35 CRKP strains was extracted using the TIANamp bacterial DNA kit (TianGen, Beijing, China). The plasmids of four ST11 bla_KPC–2-positive strains, which were selected based on virulence test (C13, C26, C31, and C38), were extracted using the Qiagen plasmid midi-kit (Qiagen, Germany). The extracted genomic DNA was evaluated by 1% agarose gel electrophoresis and quantified by the Qubit fluorometer and then subjected to short-read sequencing (2 × 150 bp) with the Illumina HiSeq 2500 platform. Subsequently, genomic DNAs of three bla_NDM-positive strains (C11, C39, and C20) from different branches and plasmids of four aforementioned strains were sequenced with the Oxford Nanopore Technologies MinION long-read platform with the RBK004 barcoding library preparation kit and MinION R9.4.1 flow cells as previously described (Wick et al., 2017; Li et al., 2018).

The short-read Illumina raw sequences of CRKP were quality filtered and assembled by SPAdes (Bankevich et al., 2012), and contigs less than 500 bp were discarded. The clone lineages, STs, insertion sequences, AMR determinants, and the virulence genes of CRKP were identified using online tools⁴ and Kleborate tool (Wick et al., 2018). The phylogenetic trees of the comparison within CRKP in this study and the comparison between CRKP in this study and other strains in GenBank were constructed using Roary and FastTree based on SNPs of core genomes (Price et al., 2009; Page et al., 2015), and further visualization and modification were performed in iTOL⁵. Combining the formed tree file and the gene presence and absence file, a phylogenetic tree with a matrix describing the presence and absence of core and accessory genes was constructed. The sequences of 35 CRKP were compared against the classical virulence plasmid pLVPK (GenBank accession AY378100), and representative plasmid sequences were further plotted by GView web server⁶ using pLVPK as reference sequence. The layout and output were edited in the GView Java stand-alone application obtained from results webpage. Genomic DNA with short-read Illumina and long-read Nanopore data was subjected to perform de novo hybrid assembly as described previously (Wick et al., 2017). The complete genome sequences were annotated using RAST⁷ automatically and modified manually. BRIG and Easyfig were used to generate the genetic comparison figures (Alikhan et al., 2011).

To analyze the risk factors responsible for the occurrence of CRKP, the clinical information of CRKP-carriers was compared to the non-carriers in terms of underwent different variables, which included gender, age, ICU, exposure to carbapenem during hospital stay, isolation season, and sample type. For all data, logistic regression analysis models were used to obtain odds ratios (ORs) and 95% confidence intervals (CIs) for analysis of independent risk factors associated with the occurrence of CRKP. Categorical variables were compared using χ² test or two-tailed Fisher exact test, with p < 0.05 considered statistically significant. All statistical analyses were processed in SPSS version 22.0.

The draft genome sequences of 32 CRKP isolates have been deposited in the GenBank database under BioProject accession no. PRJNA705380. The complete genome sequences of three bla_NDM–1-positive CRKP isolates obtained by hybrid assembly have been deposited in GenBank with accession numbers C11 (pending, deposited in figshare database temporarily), C20 (CP084103-CP084106), and C39 (CP061700-CP061702). The assembled plasmid sequences of four strains (C13, C26, C31, and C38) were deposited in the figshare database (https://doi.org/10.6084/m9.figshare.14199287.v5) for reference. Additional data that support the findings of this study are available from the corresponding authors upon reasonable request.

From March 2018 to August 2019, a total of 1,413 isolates were collected from different wards or ICUs of a newly established hospital in Henan province, China. In these isolates, K. pneumoniae (193, 14%) was the most prevalent species, followed by Acinetobacter baumannii [177 (13%)], Pseudomonas aeruginosa [158 (11%)], E. coli [116 (8%)], and Staphylococcus aureus [92 (7%)], which were the common nosocomial pathogens (Supplementary Table 2). Klebsiella pneumoniae isolates were from 18 different wards or ICUs in the hospital (Supplementary Table 3). PCR and Sanger sequencing identified 35 [of 193 (18.1%)] carbapenemase-producing K. pneumoniae. Among them, 30 isolates were positive for bla_KPC–2, whereas the remaining five isolates carried bla_NDM–1. All strains exhibited resistance to tested β-lactam antibiotics meropenem, imipenem, aztreonam, ampicillin, ampicillin–sulbactam, piperacillin, piperacillin–tazobactam, amoxicillin–clavulanic acid, cefazolin, ceftazidime, cefotaxime, and cefotaxime. Meanwhile, most strains were resistant to ciprofloxacin [30/35 (85.7%)], levofloxacin [30/35 (85.7%)], chloramphenicol [14/35 (40%)], gentamicin [33/35 (94.3%)], amikacin [25/35 (71.4%)], trimethoprim-sulfamethoxazole [21/35 (60%)], and tetracycline [20/35 (57.1%)], but remained susceptible to tigecycline [34/35 (97.1%)] and polymyxin B [35/35 (100%)] (Supplementary Table 4). To investigate the transferability of the carbapenemase-encoding genes, 35 strains were subjected to conjugation assay. However, only five isolates (C11, C12, C1, C29, and C21) including two bla_NDM–1-positive strains and three bla_KPC–2-positive strains could successfully transfer the carbapenem resistance phenotype to the recipient strain YZ6 Hyg^R, suggesting the carbapenemase-encoding genes of them were located on conjugative plasmids (Figure 1).

To evaluate the risk factors accounting for the emergence of CRKP, clinical information without individual identification of patients was compared (Supplementary Tables 5, 6). Among 35 CRKP colonization cases ranged from 3 days to 80 years old, 66% (23/35) were male and 71.4% (25/35) were older than 50 years, suggesting CRKP tended to invade middle-aged and elder patients. Furthermore, colonization was observed in 10 different wards or ICUs. Among them, 60% (21/35) of cases were from ICUs. Besides, 71.4% (25/35) of patients used to experience carbapenem treatment. Comparing with the CRKP non-carriers, no significant differences were detected in gender, age, isolation season, ICU patients, and sample source groups, but the correlation was detected between carbapenem treatment history and being CRKP positive (p = 0.009) [OR = 3.453 (95% CI = 1.369–8.707)].

Analysis of genomic characteristics revealed that the bla_KPC–2-positive isolates belonged to three known STs (ST11, ST15, and ST2237), with ST11 being the most dominant type [24/30 (80%)]. In contrast to the bla_KPC–2-positive isolates, five bla_NDM–1-positive strains exhibited diverse STs including ST37, ST1383, and ST304. To our knowledge, ST1383 and ST304 K. pneumoniae strains were not associated with carbapenem resistance; especially, the ST304 type has never been reported to carry bla_NDM–1. Analysis of wzi locus revealed that seven different types (KL64, KL110, KL10, KL25, KL47, KL15, and KL62) existed in ST11 isolates, whereas KL24 and KL19 were identified in ST15 and ST2237 isolates. In addition, bla_NDM–1-positive strains possessed three serotypes including KL15, KL110, and KL2. Roary identified a total of 9,610 genes in pangenome, including core genes (n = 3,951), soft core genes (n = 156), shell genes (n = 2,100), and cloud genes (n = 3,403) (Supplementary Figures 1, 2). A maximum likelihood phylogenetic tree demonstrated that all strains were clustered into five clades. ST11 bla_KPC–2-positive isolates were grouped into cluster I, whereas the remaining bla_KPC–2-positive isolates including three ST15 strains and three ST2237 strains were assigned to cluster V, suggesting the clonal expansion of bla_KPC–2-positive isolates dominated by ST11 K. pneumoniae along with ST15 and ST2237 K. pneumoniae that existed in this hospital. By contrast, five bla_NDM positive isolates were classified into cluster II (two ST37 strains), III (two ST1383 strains), and IV (one ST304 strain), respectively. The diversity of multilocus ST showed that bla_NDM-carrying strains had polyclonal spread. However, the epidemic features were distinguished between bla_KPC–2-harboring and bla_NDM–1-harboring isolates, as the bla_KPC–2-harboring isolates were detected in 10 different wards and ICUs, whereas the bla_NDM–1 strains were solely concentrated in children cardiac ICU. These findings suggested that the clonal spread of ST11 bla_KPC–2-positive isolates and local polyclonal spread of bla_NDM–1-positive isolates have existed in this hospital (Figure 1). Furthermore, the ST11 and ST15 CRKP strains reported in other studies (Liu et al., 2012; Li et al., 2020; Zheng et al., 2020) were also highly related to corresponding strains in this study. These strains derived from different hospitals in China, suggesting the CRKP involved in this study has been widely spread among the clinical settings (Supplementary Figure 3).

Resistome analysis revealed that the bla_KPC–2-positive strains harbored more types of AMR genes than those found in bla_NDM–1-positive strains. Moreover, bla_KPC–2-positive strains possessed almost all classes of genes conferring resistance to aminoglycoside, quinolone, sulfonamide, fosfomycin, tetracycline, and β-lactam, with the most prevalent being fosA, bla_TEM–1B, rmtB, aadA2, and bla_CTX–M–65 genes, which implied that bla_KPC–2 had potential risks of cotransmission with other AMR genes. However, the AMR profiles of bla_NDM–1-positive isolates were narrower than those of bla_KPC–2-positive isolates. They were resistant to β-lactam but still susceptible to other antibiotics including ciprofloxacin, levofloxacin, amikacin, polymyxin B, and tigecycline. Interestingly, three bla_NDM–1-positive strains carried rare ESBLs bla_SFO–1-like (with three bases mutation compared to bla_SFO–1) and bla_VEB–3, which were usually excluded from routine surveillance (Table 1).

Five bla_NDM–1-positive strains were separated to three clades based on phylogenetic analysis; therefore, three representative isolates from different branches were selected (C11, C20, and C39) for further exploration of genetic structures via MinION Nanopore long-read sequencing. The results showed that bla_NDM–1 was located on three distinct plasmids IncX3, IncFIB/IncHI1B, and IncHI5-like, respectively. In strain C11, bla_NDM–1 was found in typical IncX3 plasmid, which disseminated in human or animal sources worldwide and severed as the major vehicle of bla_NDM transmission to evolve with the generation of new NDM variants (Wu et al., 2019). In addition to bla_NDM–1, the plasmid also carried bla_SHV–12. In strain C20, bla_NDM–1-bearing plasmid pC20-394 kb with 52.1% G + C content and 458 predicted ORF was 394 kb in size and possessed IncFIB and IncHI1B replicons. Apart from bla_NDM–1, this latter plasmid harbored ESBL genes bla_CTX–M–14, bla_LAP–2, and bla_TEM–1B; tetracycline resistance gene tet(A); aminoglycoside resistance genes aadA2 and aac(3)-IId; sulfonamide resistance gene sul1; trimethoprim resistance genes dfrA1 and dfrA12; and macrolide resistance gene mph(A). Except bla_CTX–M–14, aadA2, and dfrA12 genes, the remaining AMR genes were in a 97-kb MDR region. Despite that the plasmid could not transfer by conjugation, the coselection of bla_NDM–1 may occur because of the existence of abundant AMR genes. Two integrons were found in different positions. The common genetic structure ΔISAba125-bla_NDM–1-ble_MBL-trpF-dsbC was embedded in downstream of In183, generating the complex class I integron In183-ISCR1-bla_NDM–1 structure. Another In1248-like integron with the genetic array intI1-dfrA12-aadA2-qacEΔ1-sul1 was flanked by IS26 and IS5075. BLASTn search of pC20-394 kb against the NCBI nr database showed that less homologous sequences were found between this plasmid and the known plasmids; the maximum similarity was 99% identical at 58% coverage (pAR-0161_plasmid_unnamed, CP028952) (Figure 2 and Supplementary Figure 4). The emergence of novel bla_NDM–1-bearing MDR plasmid in ST304 K. pneumoniae C20 implied that the novel plasmid mediated the transmission of bla_NDM–1 and expanded the host ranges of bla_NDM–1. Furthermore, the detailed analysis of IncHI5-like bla_NDM–1-bearing plasmid in C39 has been reported in another study; the plasmid was 334,893 bp in length and possessed a large MDR region, which contained abundant AMR genes and mobile elements (Liu et al., 2021).

A total of six classes of virulence factor analysis were conducted among these isolates. Fewer virulence genes were possessed by bla_NDM–1-positive strains than bla_KPC–2-positive isolates. In bla_NDM–1-positive strains, two isolates (C11 and C12) were not found to carry any virulence factors, whereas the remaining three isolates (C37, C39, and C20) harbored only one virulence factor yersiniabactin. Correspondingly, they also obtained lower virulence scores (Figure 3A). Analysis of the ybt locus revealed that 32 isolates were positive for the chromosomally encoded yersiniabactin, of which the dominant type was yersiniabactin lineage 9 within ICEkp3 element distributed in all ST11-bla_KPC–2 strains. Twenty-two bla_KPC–2-positive strains harbored aerobactin lineage iuc1 with aerobactin ST1. Besides, three salmochelin-producing strains, nine rmpA-positive strains, and 23 rmpA2-positive strains were also detected in bla_KPC–2-positive strains. To get further insight into virulence phenotype of CRKP, all isolates were subjected to string test. The positive results were observed in C6 (ST15/KL24) and C26 (ST11/KL64). However, a negative string test could not predicate low virulence (Russo and Marr, 2019). Therefore, eight representative bla_KPC–2-positive strains from clusters I and V, which contained all bla_KPC-carrying strains, were conducted with G. mellonella larvae infection assay. Seven strains (C6, C14, C31, C26, C19, C29, and C38) resulted in 0% survival at 24 h with an inoculum of 10⁶ CFU, and the survival rate was 20% after the infection of C13 at 72 h (Figure 3B). No deaths were observed in PBS treatment group, and the survival rate of negative control YZ6 was higher than the experimental group.

In order to gain further insights into the genetic basis of virulence and antibiotic resistance of plasmids harbored by ST11 bla_KPC–2-positive isolates, plasmids of four ST11 strains (C13, C26, C31, and C38) were sequenced by MinION Nanopore sequencing platform. As the results showed, bla_KPC–2-bearing plasmids of C13, C26, C31, and C38 shared similar backbone. These plasmids ranged from 99 to 148 kb and were classified as IncFII/IncR plasmid. Among them, the largest plasmid pC13_148 kb harbored by C13 was 148,462 bp and carried genes related to plasmid replicon, maintenance, conjugative elements, and AMR genes including bla_KPC–2, bla_CTX–M–55, bla_TEM–1B, and bla_SHV–12. BLASTn analysis demonstrated that it was similar to pSH2-85K-MDR (MH643792) and pKPC-L388 (CP029225) from K. pneumoniae, indicating the universal prevalence of this plasmid among K. pneumoniae (Figure 4). Besides, more detailed analysis of the remaining bla_KPC–2-bearing plasmids in ST11 strains was performed using pC13_148 kb as reference. All ST11 isolates possessed this type of plasmid, with the absence of some specific regions. The bla_KPC–2 gene was located on the same genetic context, flanked by genes belonging to the Tn3-based transposon family insertion sequences (ISKpn6 and ISKpn27). Obviously, the deficiency of conjugation transfer region was observed in the majority of plasmids, which may explain why the bla_KPC–2-bearing plasmids of most ST11 strains were non-conjugative. However, several bla_KPC–2-bearing plasmids (C33, C32, C23, C19, and C14) with intact conjugation transfer regions were unable to be transferred successfully; the underlying mechanism warranted further study (Supplementary Figure 5).

Three virulence plasmids carried by C13, C26, and C31 were aligned well with classical virulence plasmid pLVPK (GenBank accession AY378100), a 219-kb plasmid that harbors iroBCDN, iucABCD, rmpA, and rmpA2. Furthermore, we found the similar plasmid structure presented in other eight ST11 strains based on Illumina-based contigs analysis (Figure 5).

To evaluate indicators of survival in the clinical settings of CRKP isolates in this study, according to the previous literature (Heiden et al., 2020), the fitness, desiccant resilience, biofilm formation, human serum resistance, and siderophore secretion assays were performed. To facilitate the analysis of the results, a total of 17 strains covering 15 representative CRKP isolates in this study and two control strains were divided into five groups, including A group (nine bla_KPC–2-harboring ST11 strains), B (three bla_KPC–2-harboring ST15 or ST2237 strains), C (three bla_NDM–1-positive strains), D (K. pneumoniae ATCC700603), and E (clinical cKP HS11286 derivative YZ6).

We observed no significant difference in growth rates of the A, B, and C groups when compared to control groups. However, it was only found that the growth rate of the C group was significantly lower than that of group E at 2 h (p = 0.044 at 2 h) (Supplementary Table 7), which will be worth exploring further. Subsequently, we want to evaluate the performance of CRKP isolates under extreme dry environment, its capacity for biofilm formation, and the tolerance in human serum, which allowed us to assess the viability of CRKP in clinical settings and host. Desiccation resilience and human serum resistance experiments showed high survival rates of CRKP isolates under drying and human serum pressure. Comparable results were also obtained in biofilm formation, which suggested that these CRKP isolates will persist in this hospital for a long time, either under the clinical pressure or in patients (Figure 6).

Nevertheless, the siderophore secretion capacity of A group (bla_KPC–2-ST11) was significantly higher than C (bla_NDM–1-positive group, p = 0.005), D (standard K. pneumoniae ATCC700603, p = 0.002), and E (YZ6, p = 0.005), but no significant difference was found between the A and B groups. This might be the role of the presence of yersiniabactin, aerobactin, and salmochelin in ST11 bla_KPC–2-positive strains. However, the decreased siderophore secretion capacity was also observed in bla_KPC–2-harboring ST15 strain C1, as it exhibited quite smaller secretion zone in absence of those genes encoding siderophores (Figure 7).