The definition and screening of MDR-hvKps are still controversial (Liu and Guo, 2019). In this study, clinical investigations and several factors were considered. K. pneumoniae isolates that resisted at least three antibiotic groups was considered MDR-Kp (Magiorakos et al., 2012). Furthermore, MDR-Kp isolates that showed and/or any of the following phenotypes, such as hyper mucoid phenotype (carried rmpA and/or rmpA2) (Yu et al., 2006; Russo and Marr, 2019), similar growth pattern with hypervirulent isolate NTUH-K2044 (Hu et al., 2022), hypervirulence phenotype in Galleria mellonella infection model (Shen et al., 2019), were considered as MDR-hvKps.

Considering the above factors, 23 MDR-hvKp isolates were collected from the clinical microbiology laboratory of the first affiliated hospital of University of Science and Technology of China (USTC), Anhui, China. These critically infected patients were admitted between August 2022 and January 2023 (Supplementary Table S1). Isolates were streaked on 5% sheep agar plates and cultured at 37°C overnight to isolate the pure bacterial clones, followed by standard procedures. A VITEK 2 Compact System (bioMérieux, France) was utilized to identify the correct isolates. All isolates are stored in 40% (v/v) glycerol broth at a cryo-refrigerator (−80°C) until further experiments.

Antibiotic susceptibility of all clinical isolates was determined by previously described methods with slight modifications (He et al., 2022). The MICs of routinely used antibiotics (Supplementary Table S2), were determined by the VITEK 2 Compact System and the results of the antimicrobial susceptibility tests were interpreted according to the Clinical Laboratory Standards Institute (CLSI, 2022).

K. pneumoniae was cultured overnight in LB liquid medium at 37°C and 220 rpm. 0.5 μL of overnight cultured K. pneumoniae was streaked on LB agar plates and then incubated at 37°C for 24 h. Several monoclonal strains were selected to adjust the concentration of the bacteria in the MH (Mueller-Hinton Broth) medium (Wiegand et al., 2008). The final inoculum size for broth dilution was 5 × 10⁵ colony forming units (CFUs)/well after inoculating into MH medium with various concentrations of polymyxin B. Each concentration gradient was divided into three parallel groups and grown at 37°C and 220 rpm with shaking for 24 and 48 h. The experiment was independently repeated three times, and three technical replicates were included each time.

Multi-locus sequence typing (MLST) and capsular K-typing of all clinical hvKp isolates were determined by previously described PCR amplification methods with some modifications (Zhao et al., 2022). First, extract the whole genome from overnight bacterial culture following the manufacturer’s protocol with minor changes (GenStar). For MLST and K typing, genomic DNA from all clinical isolates was used as templates and performed PCR using 7 pairs of housekeeping genes (infB, phoE, pgi, tonB, mdh, gapA, rpoB) and a pair of wzi primer, respectively (Table 1). The PCR conditions and parameters were followed by the manufacturer’s protocol 2 × Rapid Taq Master Mix (Vazyme). Finally, the PCR product was sequenced through sanger sequencing using gene specific primers (Table 1), by General Biotechnology, China. The sequenced outcome was analyzed by SnapGene software version 5.2. ST sequences and capsular typing are available for comparison in the database.¹

Regulator of mucoid phenotype bio-markers (rmpA and rmpA2) has been widely recognized till now to determine the K. pneumoniae hypermucoviscous phenotype. First, the whole genome was extracted from overnight bacterial culture, followed by the manufacturer’s protocol with minor modifications (GenStar). Genomic DNA from all clinical isolates was used as templates and performed PCR using two pairs of primers (rmpA and rmpA2) (Table 1). The PCR conditions and parameters were followed by the manufacturer’s protocol 2 × Rapid Taq Master Mix (Vazyme). Finally, the PCR product was run in 1% agarose gel. Isolates with target PCR band indicated hypermucoviscous phenotype positive.

The viscosity of K. pneumoniae was determined using the string test. Strains that formed strings 5 mm or longer after stretching with the tip of a sterile inoculation loop were considered to have a hypermucoviscous phenotype (Shon et al., 2013). K. pneumoniae was cultured overnight in LB liquid medium at 37°C and 220 rpm. The cultures were diluted the following day to an OD₆₀₀ of 1 and centrifuged at 2,350 × g for 5 min, and the OD₆₀₀ of the supernatant was measured every minute by Gen5 Microplate Reader and Imager Software (BioTek Instruments, https://www.biotek.com/). Experiments were performed using three technical replicates and three biological replicates.

The virulence of K. pneumoniae isolates was evaluated using the Galleria mellonella infection model (Shen et al., 2019). Larvae (0.3–0.4 g) were stored in the dark and used within 3 days after shipment (Tianjin Huiyude Biotechnology Co., Ltd.). Before injection, the bacterial pellet was washed with sterile saline and diluted to 1 × 10⁸ CFU/mL. Using a 1 mL insulin syringe, 10 μL of the bacterial suspension was injected into the center of the second gastropod of the larvae. A group of 10 larvae was randomly selected for injection. A group of larvae was injected with 0.9% NaCl solution as the negative control. Another group of larvae without injection was also included parallel. After injection, the larvae were incubated at room temperature, and survival rate was monitored daily up to 72 h. Death was recorded when the larvae no longer responded to touch. In all cases, no dead larvae were observed in the negative control groups. This experiment was repeated three times independently.

The growth curves of K. pneumoniae were established in LB medium manually, followed by standard protocol with minor modifications. Overnight cultures were diluted to an OD₆₀₀ of 0.02 and grown in 96-well plates at 37°C and 220 rpm with shaking. The absorbance of the culture solution at OD₆₀₀ was measured every 0.5 h until it reached the peak and became flat. The OD₆₀₀ was measured by Gen5 Microplate Reader and Imager Software (BioTek Instruments, https://www.biotek.com/). Experiments were performed using three technical replicates and three biological replicates.

A total of 23 K. pneumoniae isolates were sequenced by second-generation technology. Briefly, whole genome sequencing of K. pneumoniae was performed using an Illumina Hi-Seq 4,000 platform at Nuosai Jiyin Zu Research Center Limited Company (Beijing, China). The fastp is used to filter the raw sequencing data (Chen et al., 2018). Filtered sequence data were assembled using Unicycler v0.4.8 (Wick et al., 2017) and annotated with the rapid prokaryotic genome annotation tool, Prokka 1.14.6 (Seemann, 2014).

ABRicate version 1.0.14 was implemented to detect acquired antimicrobial resistance genes (ARGs) by aligning genomic sequences to the ResFinder and NCBI databases (Zankari et al., 2012). Kleborate and ABRicate were utilized to determine the isolates’ virulence factors by aligning their genomic sequences to the VFDB database (Zankari et al., 2012; Wyres et al., 2016). Multi-locus sequence typing (MLST) was conducted by MLST 2.15 (Jolley and Maiden, 2010). Capsule typing was undertaken by Kleborate (Wyres et al., 2016; Lam et al., 2021). The HarvestTools kit (Parsnp, Gingr, and HarvestTools) and BacWGSTdb were utilized to perform a comparative genomic and phylogenetic analysis of the different isolates to build phylogenetic trees based on the maximum likelihood method (Treangen et al., 2014; Feng et al., 2021). The sequence alignment of mutant cps gene cluster was visualized using Easyfig 2.2.5 (Sullivan et al., 2011). The interactive tree of life (iTOL) v5 was used to construct a phylogenetic tree (Letunic and Bork, 2021). SNPs among various strains were investigated using Snippy.² Heatmaps were illustrated with the ComplexHeatmap R package (Gu et al., 2016).

Whole genome sequencing data were deposited in the NCBI database under BioProject: PRJNA971333.

All the statistical analyses were performed using GraphPad Prism 7.0.³ Error bars represent SEM. All the experiments were performed at least three times with three technical replicates.

A total of 23 MDR-hvKp isolates were collected. Most of the patients’ age is above 60 years (n = 16), ≥40- ≤ 60 (n = 4) and young individuals (n = 3), where 78.3% were male (n = 18), and 21.7% were female (n = 5). Interestingly, in clinical investigation, the majority of patients were hospitalized for other than lung diseases or its associated pneumonia (n = 7) and most of them were hospitalized due to old age diseases or others (n = 16) that indicates nosocomial infection (Supplementary Table S1). Clinically, most of the isolates were collected from sputum (n = 20) and remaining from blood (n = 3). These 3 patients were admitted in ICU for rather than K. pneumoniae associated diseases. In antibiotic susceptibility test, all isolates showed resistance to Cefuroxime axetil, Cefuroxime, Ceftriaxone, and intermediate resistance to Polymyxin B (Supplementary Table S2). In multi-locus sequence typing (MLST) and capsular typing (K-typing), nine isolates belonged to ST11 and all were KL64 except one belonged to KL47, five to ST23 (KL1), two to ST307 (KL102), and remaining belonged to an individual ST e,g. ST15 (KL19), ST 86 (KL2), ST 147 (KL64), ST485 (NA) and ST3132 (KL24) (Supplementary Table S3). In addition, the WGS results indicated that one isolate belonged to Klebsiella quasipneumoniae (22100407).

The presence of regulator of mucoid phenotype A (rmpA) and A2 (rmpA2) positively determines hyper-mucoid phenotype and hypermucoviscosity. Of 23 isolates, nine isolates showed rmpA positive and 14 isolates showed rmpA2 positive where nine isolates showed both rmpA and rmpA2 positive, and 7 isolates showed negative to both (Supplementary Figures S1, S2). Isolates with both genes showed hyper-mucoid phenotype and mucoviscosity in centrifuge at 2350 × g (Supplementary Figure S3). Interestingly, two isolates, 22122315 (ST15) and 22091569 (ST307), that lacked both genes showed hyper-mucoid phenotype. Meanwhile, isolate 22091569 showed the highest string length in the blood agar plate after overnight incubation at 37°C, whereas isolate 22122315 showed stickiness, but the string length was not significant (≤ 5 mm) (Figure 1).

We then investigated some phenotypic experiments, e.g., mucoviscosity assay, growth curve, and including G. mellonela infection model to assess its virulence. NTUH-K2044 was used as a reference hypervirulent isolate and positive control.

In the mucoviscosity assay, all the isolates, including NTUH-K2044, were incubated overnight, adjusted OD (OD₆₀₀ = 1) and centrifuged at low speed (2,350 × g) for 5 min, where measured the OD₆₀₀ after every single minute. The supernatant of NTUH-K2044 became absolutely transparent after 3 min. Alternatively, isolate 22112350 remained turbid, and no pellet was formed after 5 min, whereas isolate 22122315 showed a significant mucoviscosity up to 3 min but still exhibited a mucoid pellet (Figure 2A and Supplementary Figure S3A). In addition, 22111436 and 22111439, and 22091569 showed hypermucoviscosity and remained turbid; no pellet was formed after 5 min (Figure 2B and Supplementary Figure S3B). In isolates 22091650 and 22111913, it remained hyper-mucoid and showed hypermucoviscosity compared to the control (Figure 2C and Supplementary Figures S3B,C). Furthermore, isolates X22083021 and X22083165 showed slight stickiness compared to the control (Figure 2D and Supplementary Figure S3C). Here, except isolates 22122315 and 22091569, all the hyper-mucoid isolates had both rmpA and rmpA2 gene.

In G. mellonella infection model, all the clinical MDR-hvKp isolates demonstrated more virulence compared to NTUH-K2044 where 40% (±10) of insects survived after 72 h post-infection (Figure 3 and Supplementary Table S4). Specifically, isolates 22122315 (and 22091650) showed the highest virulence and all larvae were died after 12 h post-infection (mean survival rate, 0% (±0)). Meanwhile, isolate 22091569 showed higher virulence compared to control, where only 10% (±10) of injected larvae remained alive after 72 h post-infection. Moreover, for isolates 22111605 and X22083021, all the larvae died after 24 h post-infection. All the larvae were died after 36 h post-infection in case of isolates, X22121850, 22112350, 22111719, and X22083165. This infection model indicated both rmpA/ rmpA2 lacked isolates 22122315 and 22091569 showed hypervirulence compared to control (Supplementary Table S4).

In growth curve, NTUH-K2044 showed the fastest growth rate, and all the clinical isolates showed similar growth patterns except isolate X22083021, which reached the peak at 8.5 h and declined at 9 h and then remained stable (Supplementary Figure S4).

We then performed WGS to explore functional genomics and better understanding the reasons behind the hyper-mucoid phenotype and mucoviscosity. In comparative genomics, we analyzed the distribution of antibiotic resistance genes, plasmid, and virulence factors among the isolates (see Figure 4).

We identified 46 MDR genes across the MDR-hvKp clinical isolates, 17 of which encoded β-lactamases, including carbapenemase-encoding genes. Moreover, the MDR-hvKp isolates carried β-lactamase-encoding genes bla_{TEM − 1B} (n = 14), bla_{LAP − 2} (n = 8) and bla_{OXA − 1} (n = 5), ESBL encoding genes, bla_{CTX − M − 15} (n = 8), bla_{CTX − M − 65} (n = 8), bla_{SHV − 190} (n = 5) and carbapenemase-encoding genes bla_{KPC − 2} (n = 11). In addition, we noticed all the MDR-hvKp isolates carried the fosfomycin resistance gene, fosA.

Meanwhile, both isolates 22122315 and 22091569 showed MDR phenotype in antibiotic susceptibility testing and the comparative genomic analysis of antibiotic resistant genes also confirmed this phenomenon. Both isolates carried MDR genes such as β-lactamase-encoding genes bla_{TEM − 1B}, and bla_{OXA − 1}, carbapenemase-encoding genes, bla_{KPC − 2}, ESBL encoding genes, bla_{CTX − M − 15}, bla_{SHV − 106}, quinolone efflux pump coding genes, oqxAB, tetracycline resistance genes, tet(A) (Figure 4). However, none of these isolates carried any polymyxin-resistant gene, mcr, and antibiotic susceptibility test also confirmed it.

Furthermore, majority of isolates carried IncFIB(K) plasmid. In addition, some hypervirulent isolates also carried IncFII(pHN7A8), IncFIB(Mar), ColRNAI, and IncHI1B plasmids. Meanwhile, isolate 22091569 carried IncFIB(Mar), IncFII, and IncFIB(K) plasmid, whereas 22122315 carried FII(pBK30683) only.

In case of virulence factors, all the MDR-hvKp isolates carried outer membrane protein-coding genes (ompA), fimbriae encoding genes (ecpABCDER), enterobactin coding genes (entAB, and fepC). Besides, the majority of isolates carried yersiniabactin coding genes (fyuA, ybtE, ybtT, ybtU, irp1, irp2, ybtA, ybtP, ybtQ, ybtX, ybtS) (69.5%). In addition, aerobactin coding genes (iutA, iucABC), salmochelin coding genes (iroBCD, and iroN) were also present in a few isolates. However, the virulence-defined genes were not prevalent in all phenotypically observed hypervirulence isolates in G. mellonella infection models. We found that isolates 22122315 and 22091569 lacked virulence-defined genes, aerobactin and salmochelin. However, the 22122315 isolates alone had yersiniabactin coding genes.

Among the 23 MDR-hvKp isolates, 22122315 and 22091569 showed hypervirulence and hyper-mucoid phenotype without regulators of mucoid phenotype gene, rmpA and rmpA2. Usually, these two biomarkers are positively responsible for hypermucoviscosity in K. pneumoniae. We performed single nucleotide polymorphism (SNP) analysis using WGS to reveal its mystery.

When analyzed SNPs in 22122315 (ST15), it was compared with recently reported ST15 isolates (Zhao et al., 2022). Considering all SNPs in the coding DNA sequence (CDS) region, we identified 36 common mutations, including 1 stop lost. Moreover, we identified 8 missense novel mutations in unidentified locus among the 36 repetitive missense mutations, and all the unknown mutant genes produce hypothetical proteins except FEBNDAKP_03184 (Supplementary Table S5). In 2084th nucleotide position, adenine(A) mutated to cytosine(C), resulting in a change of 695th amino acid from Asparagine (Asn) to Threonine (Thr) (Figure 5). Interestingly, this mutant gene is responsible for putative tyrosine-protein kinase in the cps region.

Similarly, to detect SNPs in 22091569, it was compared with ST307 isolates (He et al., 2022). Here, we identified 17 mutations were same in each ST307 isolates. Among them, 4 mutations were found in unknown locus, namely, EOFMAFIB_00699 (1186C > T, Arg396Cys), EOFMAFIB_00523 (179A > T, Asn60Ile), EOFMAFIB_00116 (310G > T, Ala104Ser), and EOFMAFIB_00796 (c. 719A>T. p. Asn240Ile) that produce a hypothetical protein (Supplementary Table S6). Apart from this, we identified a novel mutation in EOFMAFIB_02276 loci where the 1930th nucleotide was mutated from cytosine (C) to adenine (A) which resulted in a change of 644th amino acid Proline (Pro) to Threonine (Thr) (Figure 6). Interestingly, this mutation is also responsible for putative tyrosine-protein kinase in the cps region, the same as in isolate 22122315.

Furthermore, we examined the mutation of both loci, FEBNDAKP_03184 and EOFMAFIB_02276, in the cps gene cluster. This cluster ranges from galF (UTP--glucose-1-phosphate uridylyltransferase) to ugd (UDP-glucose dehydrogenase). Isolate 22122315 contained a set of 20 genes within the cps gene cluster while 22091569 had 17 genes. NCBI blast analysis showed that both mutant loci carried 720 amino acids and were located in the wzc of the cps gene cluster. Besides, wzc is also responsible for the tyrosine auto-kinase involved in polysaccharide biosynthesis (Figure 7).

We conducted molecular phylogenetics to construct a phylogenetic tree based on the maximum likelihood approach. Moreover, the trees were evolutionary illustrated into four subgroups based on highly similar homology where, two hypervirulent isolates, 22122315 and 22091569 were originated from different sub-groups (Figure 8). The tree showed that all the ST11, ST23, ST307 isolates showed similar homology and were positioned in the same sub-groups. Interestingly, isolates 22111719 and 22110212 showed similar homology, but their ST and K-typing were not identical; they belonged to ST147 and ST273, respectively.