In July 2021, AHSWKP25 was isolated from the blood of a patient diagnosed with acute myeloid leukemia with a bloodstream infection at the Affiliated Hospital of Southwest Medical University (Luzhou, China). The patient had previously been tread with ceftazidime, meropenem, moxifloxacin, tigecycline, and voriconazole for fungal infection and repeated fever. The clinical microbiology laboratory identified AHSWKP25 as K. pneumonia that was resistant to carbapenems, tigecycline, fluoroquinolones, and other common antimicrobial agents. The patient ultimately died of multiorgan failure. To better understand the antimicrobial resistance mechanisms and virulence characteristics of AHSWKP25, we undertook a series of experiments described below.

The antimicrobial minimum inhibitory concentrations (MICs) of ceftazidime, cefepime, gentamicin, amikacin, chloramphenicol, ciprofloxacin, levofloxacin, tetracycline, polymyxin B, ceftazidime/avibactam, fosfomycin (agar dilution method), and tigecycline were determined according to the (Clinical and Laboratory Standards Institute (CLSI), 2020) standards. The MIC breakpoints of imipenem, ertapenem, aztreonam, levofloxacin, nitrofurantoin, and trimethoprim-sulfamethoxazole were determined using the MicroScan Walk Away System (Siemens, Germany). The breakpoint of tigecycline was interpreted according to the guidelines of the United States Food and Drug Administration (FDA¹) on Antimicrobial Susceptibility Testing. The modified carbapenemase inactivation (mCIM) and EDTA-modified carbapenemase inactivation methods (eCIM) were also used following the CLSI 2020 standards.

To evaluate the efflux pump activity of AHSWKP25, we first measured the changes in the MICs of several antimicrobial agents in the presence of 1-(1-Naphthylmethyl)-piperazine (NMP, 100 mg/L). A fourfold or greater reduction in MIC was considered an indicator of overexpression in the efflux pumps (Schumacher et al., 2006). Polymerase chain reaction (PCR) was performed to detect whether the isolate carried the bla_KPC or bla_NDM genes, and the primers used are listed in Supplementary Table 1.

The hypermucoviscous phenotype was determined using a string test. A single purified colony on a blood agar plate incubated at 37°C for 18-24 hours was picked with an inoculation loop for string, and the string test was considered positive if the length of the viscous string was longer than 5 mm (Yan et al., 2021). A serum killing assay was performed as described previously. In brief, 25 μL of the bacterial suspension (concentration of ∼1 × 10⁶ CFU/mL) was added to 75 μL of healthy human serum for co-culture. After 0, 1, 2, and 3 h, the plates were inoculated with the mixture and the numbers of viable bacteria were calculated. The strains were classified as “highly sensitive,” “moderately sensitive,” or “resistant” to serum, depending on the results (Liu et al., 2017).

The pathogenicity of AHSWKP25 was assessed using a G. mellonella larvae infection model as previously described (Kim et al., 2021). A total of 15 healthy vigorous larvae were inoculated with the bacterial suspension of AHSWKP25 isolate at a dose equivalent to 10⁶CFU. When the larvae were inactive and black, they were considered dead. The numbers of larval deaths were recorded every 12 h. K. pneumoniae NTUH-K2044 and ATCC 700603 were used as the hypervirulent control and negative control, respectively. The bacterial suspension was serially diluted and injected into G. mellonella larvae for 3 days of incubation, and the lethal dose 50 (LD₅₀) was calculated using the probit model (Shi et al., 2018).

Luria-Bertani (LB) broth was inoculated with the selected purified Klebsiella colonies, and the bacteria were cultured to log phase. Extraction of bacterial genomic DNA using a magnetic bead-based kit (Qiagen, Germany). The extracted bacterial DNA was purified and sequenced using the Illumina NovaSeq PE150 and Oxford nanopore platforms. The bacterial genome was assembled de novo using Canu (v 1.7). Prokka (v 1.10) was used to predict and annotate the coding genes, tRNAs, and rRNAs in the assembled genome. The plasmid replicon types, acquired resistance genes, sequence types, and virulence genes were determined using the online services of the Center for Genomic Epidemiology² and VFDB (virulence factor database³). IS finder⁴ was used to identify the insertion sequences. The genomes were compared using the Basic Local Alignment Search Tool (BLAST⁵), BRIG v0.95, Mauve⁶, and the OAT software (Lee I. et al., 2016).

We used K. pneumoniae UCI 38 (accession number: JCMB01) as a reference strain to identify point mutations in the rcsAB and lon (Lon protease) genes of AHSWKP25 to evaluate whether AHSWKP25 was associated with hypercapsule production (Ernst et al., 2020). The rscD and pal mutants were identified using K. pneumoniae ATCC13883 (accession number: JOOW01) as the reference genome.

A goeBURST full MST analysis based on multilocus sequence typing data was performed by the Phyloviz 1.1a software to infer the phylogeny of the sequence types⁷. The Orthologous average nucleotide identity (OrthoANI) was calculated by comparing homologous genes between the genomes to construct a phylogenetic tree (Lee I. et al., 2016), which was visualized by iTOL⁸. The accession numbers of the genomes/sequences obtained from the NCBI database were listed in Supplementary Table 2.

To evaluate the role of chromosomal point mutations in the antimicrobial resistance of AHSWKP25, we used RT-PCR to measure the transcriptional levels of efflux pump-encoding genes (acrA, acrB, marA, marR, rarA, and ramA) involved in conferring tigecycline resistance. In brief, single purified colonies of AHSWKP25 were inoculated into 5 mL of LB broth and the bacteria were grown to log phase. Their RNA was extracted according to the recommendations of the reagent’s manufacturer (Magen, China). The RT-PCR primers used in the present study are listed in Supplementary Table 1. The rpoB gene was used as the internal reference, and the tigecycline-susceptible K. pneumoniae strain NTUH-K2044 was used as the control strain. All experiments were performed in triplicate independently, with three biological replicates per experiment. The gene expression results were compared to the expression levels of rpoB to calculate the relative expression of the target genes (2^–ΔΔCt method).

Sodium azide-resistant Escherichia coli J53 was used as the recipient and AHSWKP25 as the donor in the conjugation experiments. The McFarland (McF) standard turbidity of the bacterial suspension was adjusted to 0.5 McF. The recipient bacteria (200 μL) and donor bacteria (400 μL) were inoculated into LB broth (800 μL) (Xiang et al., 2020). After 16–18 h of culturing at 35°C, MH plates containing 180 μg/mL sodium azide and antimicrobial agents [4 μg/mL meropenem; 4 μg/mL meropenem + 5 mM EDTA (final concentration); 0.25 μg/mL ciprofloxacin] were inoculated with 100 μL of the culture solution to screen for transconjugants. The transferred genes in the transconjugants were confirmed by PCR using the primers listed in Supplementary Table 1.

The stability of the self-transferred plasmids in the positive transconjugants was calculated using the plate count method, as described previously with some modifications (Nang et al., 2018). Briefly, antibiotic-free LB broth was inoculated with the positive transconjugants at a ratio of 1,000:1, and the bacteria were passaged continuously for 10 days. The mixed culture was sampled every day by inoculating aliquots (100 μL) of the mixed culture into meropenem-containing (4 μg/mL), antibiotic-free LB agar plates. The plates were incubated for 24 h to calculate the plasmid retention rate.

The antimicrobial susceptibility tests indicated that AHSWKP25 was resistant to common antibacterial agents, such as meropenem, tetracycline, ciprofloxacin, levofloxacin, and gentamicin. Notably, AHSWKP25 was also resistant to fosfomycin, tigecycline (16 μg/mL), and ceftazidime-avibactam, as shown in Table 1. The mCIM and eCIM results suggested that AHSWKP25 carries serine carbapenems. Subsequently, PCR and Sanger sequencing confirmed that AHSWKP25 was a bla_NDM–1 and bla_KPC–2 co-producing K. pneumonia isolate.

The string test of AHSWKP25 was negative, as the mucoid string length was < 5 mm. AHSWKP25 was highly pathogenic in the G. mellonella infection model. Of the 15 G. mellonella larvae that were inoculated with 10⁶ CFU of AHSWKP25, 15 died within 24 h, corresponding to a mortality rate of 100%, which was similar to the rate for the hypervirulent control strain NTUH-K2044 (P > 0.05) but different from that of K. quasipneumoniae ATCC700603 (negative control, P < 0.0001) (Figure 1B). The LD₅₀ of AHSWKP25 was 4.44 ± 0.23 (log10 CFU), and the LD₅₀ of the NTUH-K2044 was 4.21 ± 0.17 (log10 CFU) (P > 0.05, Supplementary Figure 1). Both NTUH-K2044 and AHSWKP25 showed resistance to the serum, whereas ATCC 700603 was highly sensitive to the serum (Figure 1A and Supplementary Table 3). To further investigate the virulence characteristics and resistance mechanisms of AHSWKP25, we sequenced its whole genome to identify mutations in the genes that regulate the resistance mechanisms.

The full length of the AHSWKP25 genome was 5,874,657 bp, and it consisted of a chromosomal backbone carrying bla_SHV–5 and seven circular plasmids. Of these, the plasmid incompatibility (Inc) groups for pSW25NDM1, pSW25HRG, and pSW25tet(A) were X3, FIB, and FII/FIA, respectively. The bla_KPC–2 and bla_NDM–1 genes were located on plasmids pSW25KPC2 and pSW25NDM1, respectively. The AHSWKP25 genome was longer than K. pneumoniae MGH78578 (5,694,894 bp, GenBank accession number: CP000647-652) and ATCC13883 (5,545,784 bp, accession number: JOOW01). The capsular serotype of AHSWKP25 was identified as K53 based on the wzc locus, which was consistent with the serotype of K. quasipneumoniae ATCC700603 (GenBank accession number: CP014696.2). The wzi allele (wzi outer membrane protein of cluster) of AHSWKP25 was most similar to allele 534 but it differed by one base pair. Therefore, we uploaded the wzi allele of AHSWKP25 into a public database⁹ and assigned a novel profile to it defined as wzi 725. The mrkABCDFHIJ, fimABCDEFGHIK, rcsAB, entABCDEFS, and iroE genes were all identified on the chromosomal backbone of AHSWKP25, whereas the iucA and rmpA/rmpA2 genes were not detected (Supplementary Table 4). AHSWKP25 shared an average nucleotide identity (ANI) of 98.97 and 98.96% with K. pneumonia MGH78578 and ATCC 13883 (Supplementary Figure 2), respectively. The multilocus sequence type of AHSWKP25 was ST464, which differed from ST4292 and ST2439 by only one allele, whereas ST464 differed from ST11 by five alleles (Supplementary Figure 3). The amino acid substitutions in ramR, acrR, and ompK36/ompK37 genes of AHSWKP25, as shown in Table 2. Furthermore, we also identified Ser35Asn, Glu259Ala, and Gln110His missense mutations in the rcsA, lon, and csrD genes of AHSWKP25, respectively. However, no missense mutations were found in the pal gene. A phylogenetic tree generated from orthoANI data revealed that AHSWKP25 and the K. pneumoniae strains co-producing NDM-1 and KPC-2 previously isolated from other regions of China (Liu et al., 2019; Gao et al., 2020; Xu et al., 2020) had obvious regional and temporal differences and were significantly separated phylogenetically (Figure 2).

Among the seven circular plasmids carried by AHSWKP25, we focused on four of them, namely pSW25KPC2, pSW25HRG, pSW25tet_A, and pSW25NDM1. pSW25KPC2 was a circular plasmid with a length of 248,847 bp, but we failed to identify its incompatibility group with Plasmid Finder 2.1. This plasmid carried multiple resistance determinants, including bla_KPC–2 (beta-lactam resistance), aadA16 (aminoglycosides resistance), aac(3)-IId (aminoglycosides resistance), sul1 (sulfamethoxazole resistance), dfrA27 (trimethoprim resistance), qnrB4 (quinolones resistance), ARR-3 (rifampicin resistance), merR [Hg(II)-responsive transcriptional regulator] and qacEΔ1 (quaternary ammonium compound resistance) (Figure 3A). We searched the National Center for Biotechnology Information (NCBI) database and found that pSW25KPC2 shared a 99.97% identity (with an 92% query coverage) compared to the plasmid pYNKP001-dfrA isolated from R. ornithinolytica (GenBank accession number: KY270853.1). Sequence alignment indicated that the backbone of pSW25KPC2 was highly similar to the plasmids pKP04VIM (K. pneumoniae), pYNKP001-dfrA (R. ornithinolytica), and pRJA166a (K. pneumoniae), but there were multiple Local Colinear Blocks (LCBs) that were inverted (Figures 3A, 4A). Notably, pKP04VIM, pYNKP001-dfrA (K. pneumoniae), and pRJA166a (K. pneumoniae) did not carry the bla_KPC–2 gene. Further analysis revealed that the bla_KPC–2 gene of pSW25KPC2 was located in an LCB of about 6 kb that was highly homologous with pK55602_2 (GenBank accession number: CP042976.1) and pKPC2_095132 (GenBank accession number: CP028389.3) and was highly collinear with no inversions or rearrangements. The mapping of this LCB to Figure 4B revealed that pSW25KPC2, pK55602_2, and pKPC2_095132 all contained a conserved region carrying korC (encoding transcriptional repressor protein), ISKpn6, bla_KPC–2, ISKpn27, Tn3, and a truncated bla_TEM. Transposase TnAs1 (Tn3-like element) and IS4321 were also present on the pSW25KPC2 plasmid flanking a region encoding bla_KPC–2. Additionally, a region encoding aac(3)-IId was inverted in pSW25KPC2 compared to pKPC2_095132 (Figure 4A). The Tn4401 transposon and its variants were not identified in pSW25KPC2. Lastly, pSW25KPC2 also contained a region connecting ISCR1 and intl1, that contained ARR-3-dfrA27-aadA16- qacEΔ1-sul1 genes (Figure 3A).

pSW25HRG was a FIB plasmid that did not carry any acquired antimicrobial resistance genes. There were several heavy metal resistance genes in the backbone of pSW25HRG: pcoABCDERS (copper resistance), silBCEFGPRS (silver resistance), and terABCDEWXYZ (tellurium resistance). Sequence alignments constructed with BLAST (see text footnote 5) revealed that most plasmid sequences similar to pSW25HRG have been isolated from K. pneumoniae strains. pSW25HRG shared a 99.93% identity (with an 80% query coverage) with a plasmid isolated from the K. pneumoniae strain FDAARGOS_1322 (GenBank accession number: CP070038.1) and a 99.90% identity (with an 83% query coverage) with a plasmid isolated from K. pneumoniae strain NICU_2_P7 (GenBank accession number: CP060050.1). Notably, pSW25HRG had a 23-kb homologous region compared to the well-known virulence plasmid pK2044 (GenBank accession number: NC_006625.1), wherein the above mentioned heavy-metal-resistance genes mainly occurred (Figure 3B).

The pSW25tet_A carried two replicon types, FII and FIA, and was 120,990 bp in length. Similar to pSW25KPC2, pSW25tet_A carried many genes encoding type IV conjugative transfer systems and multiple antimicrobial resistance determinants, including tet(A) (tetracycline resistance), dfrA14 (trimethoprim resistance), sul2 (sulfamethoxazole resistance), qnrS1 (quinolones resistance), and bla_lAP–2 (penicillins resistance). We identified that tet(A) in the pSW25tet_A was a type I mutant with a mutation profile of I5R, V55M, I75V, T84A, S201A, F202S, and V203F. This tet(A) mutant was previously demonstrated to be associated with reduced tigecycline susceptibility (Chiu et al., 2017). By searching the NCBI database, we found that pSW25tet_A was also similar to several other plasmids isolated from K. pneumoniae strains. Of these, pSW25tet_A shared a 99.97% identify (with a 100% query coverage) with a plasmid pRGT34-2-122k (GenBank accession number: CP075310.1) that was also isolated from K. pneumoniae. The homology between pSW25tet_A and pRGT34-2-122k was confirmed by comparing genomic circles (Figure 3C). pSW25tet_A shared only 28 and 29% coverage compared with pYUSHP2-2 (GenBank accession number: CP073773.1), a plasmid isolated from E. hormaechei, and pWP8-W19-CRE-01_3 (GenBank accession number: AP022271.1), a plasmid isolated from R. ornithinolytica, respectively.

Lastly, pSW25NDM1 was identified as an X3 plasmid carrying bla_NDM–1 and bla_SHV–12 genes. It was 59,349 bp in length and had a GC ratio of 49.06%. The genetic structure surrounding bla_NDM–1 included IS3000, IS5, ble_MBL, trpF, tat, dct, groES and groEL. pSW25NDM1 shared a 100% identity (with an 89% query coverage) with pECL-14-60-NDM-1 (GenBank accession number: MN061454.1), a plasmid isolated from E. cloacae. Sequence alignment revealed that pSW25NDM1 was very similar to plasmids isolated from K. pneumoniae, E. coli, E. cloacae, and R. ornithinolytica (Figure 3D). pSW25NDM1 contained an additional region of about 6 kb, which was not present in other plasmids carrying NDM-1, and it was highly homologous to a plasmid isolated from Leclercia sp. (GenBank accession number: CP026168.1) (Figure 3D). Further analysis showed that this region could be inserted by ISRo7 (marked black in Figure 3D). The bla_SHV–12 was located in a 4-kb inserted region that could be inserted by IS26 at both ends.

Among the three plasmids carrying the acquired antimicrobial resistance genes mentioned above, we only demonstrated the horizontal transferability of pSW25NDM1 in conjugation experiments. The pSW25NDM1 can be transferred to E. coli J53 at a transfer frequency of about 2.03 × 10^–3. Recipient strain E. coli J53 showed significantly reduced susceptibility to meropenem and cefepime after the acquisition of plasmid pSW25NDM1 (Table 1). E. coli J53 carrying pSW25NDM1 displayed high plasmid stability during its continuous passage in an antibiotic-free environment for 10 days, and more than 95% of the recipient strains cells still retained pSW25NDM1 on day 10 (Supplementary Figure 4).

The MICs of ceftazidime and meropenem for AHSWKP25 did not change significantly after the addition of NMP (100 μg/mL). However, there were ≥ 4-fold reductions in the MICs of ciprofloxacin and chloramphenicol in the presence of NMP (100 μg/mL). And NMP restored the susceptibility of AHSWKP25 to tigecycline (Table 1). Furthermore, the RT-PCR results indicated that the expressions of acrA and acrB genes were up-regulated by 1.4-fold (1.43 ± 0.05) and 1.5-fold (1.54 ± 0.12), respectively, in AHSWKP25 compared to that in NTUH-K2044. The expressions of rarA, marA, and marR were up-regulated approximately twofold in AHSWKP25 (2.25 ± 0.24, 2.82 ± 0.43, 1.94 ± 0.19, respectively), while ramA was up-regulated 19-fold (19.18 ± 1.04) (Figure 5).