All 188 sequenced strains of hypervirulent K. pneumoniae were available in the Genbank database (as of May 2022), and the characteristics of these genomes are listed in Table S1 . Additionally, 530 K. pneumoniae clinical isolates were previously collected from nine hospitals in seven provinces between January 2017 and February 2018. We selected strains of the three main hypervirulent serotypes (K1, K2, and K64) from these K. pneumoniae isolates, for a total of 48 strains; the details of these strains are also provided in Table S1 . MLST analysis was conducted using the Pasteur Institute MLST website for K. pneumoniae (http://bigsdb.pasteur.fr/klebsiella/klebsiella.html). Additionally, K serotypes were determined using Kaptive 2.0 (https://github.com/katholt/kaptive). Virulence genes (iro and iuc; rmpA and rmpA2) were identified using the blastn tool (https://blast.ncbi.nlm.nih.gov/). Isolates were classified as the hvKp genotype if they contained iuc and/or iro, and rmpA and/or rmpA2.

PCR and sequencing were carried out to screen for the wcaJ/wbaP gene in isolates obtained from the Genbank dataset and from clinical sources. The primers employed are shown in Table S2 . The wcaJ/wbaP sequences were aligned using the Clustal W software package, and maximum-likelihood phylogenetic trees were constructed using the MEGA v7.0.26 software package with 1000 bootstrap values.

Plasmids pACYC-wcaJ^K1, pACYC-wcaJ^K2, and pACYC-wcaJ^K64 were constructed using the NEBuilder HiFi DNA Assembly Cloning Kit (NEB), following the manufacturer’s recommended protocol. Additionally, the pACYC184 plasmid was electrotransferred into NTUH-K2044 as a negative control.

The gene knockout strains K2044^ΔwcaJ and K2044^ΔrmpA were constructed using the λ-Red homologous recombination system, as previously reported (Datsenko and Wanner, 2000). The primer sequences employed are presented in Table S2 . When necessary, appropriate antibiotics were added at the following final concentrations: chloramphenicol (chlo): 60 μg/mL for pACYC184; kanamycin (kana): 100 μg/mL for K2044^ΔwcaJ and K2044^ΔrmpA.

The effect of wcaJ on the growth of K. pneumoniae was measured. In brief, overnight cultures of all K. pneumoniae strains were diluted at 1:100 in broth supplemented with different appropriate antibiotics. The cultures were incubated at 37°C and A₆₀₀ was measured every 30 min.

K. pneumoniae bacteria in the mid-log growth phase were collected and fixed with 2.5% glutaraldehyde overnight at 4°C, then washed twice in 0.1 M phosphate buffer. Next, bacteria were post-fixed with 1% osmium tetroxide for 1 h, washed twice in 0.1 M phosphate buffer water, and washed twice in water. The samples underwent subsequent dehydration in increasing grades of alcohol (30%, 50%, 70%, 80%, 95%: 15 minutes; 100%: 2 × 10 min). Next, infiltration and embedding were performed using Spurr’s resin and samples were polymerized at 60°C for 48 h. Samples were examined using a Tecnai G2 Spirit Twin transmission electron microscope and corresponding images were recorded.

The mucoviscosity of the Klebsiella strains was measured as previously described (Bachman et al., 2015). An overnight culture grown in LB was diluted at 1:100 in media and grown at 37°C. At 6 h, the culture was centrifuged at 1000 × g for 5 min, followed by immediate measurement of the OD₆₀₀ of the supernatant.

Uronic acid was extracted and quantified as described previously (Mike et al., 2021); the test strains were cultured for 6 h. Subsequently, 500 µL culture was mixed with 100 µL of 1% Zwittergent 3–12 detergent and heated for 20 min at 50°C, then centrifuged for 5 min at 13,000 × g. Next, 300 µl of supernatant was mixed with 1.2 mL absolute ethanol and centrifuged for 5 min at 13,000 × g. The pellet was dried and resuspended in 200 µl of sterile water, to which 1.2 mL of tetraborate solution (12.5 mM sodium tetraborate in sulfuric acid) was added. This was incubated for 5 min at 100°C, followed by immediate cooling on ice for at least 10 min, which was then followed by addition of 20µl of hydroxyphenyl reagent. After 5 min incubation at room temperature, OD was determined at 520 nm.

Biofilm assay was performed, following previously published methods, in 96-well plates. Briefly, 200 μL of mid-log phase bacteria cells (1.5x10⁷CFU/ml) was added to a 96-well plate and incubated at 37°C for 24 h. All cultures were removed and the wells were washed twice with phosphate-buffered saline (PBS), then stained with 200 μL of 0.1% crystal violet for 20 min and rinsed twice with PBS. Stained biofilms were solubilized with 95% ethanol and quantified by measuring the OD₅₉₅ value.

RAW 264.7 murine macrophages were grown in DMEM medium supplemented with 10% FBS. Macrophages (1x10⁵/well) were seeded into 24-well plates and infected at a MOI of 50 (bacteria/cell). Subsequently, plates were centrifuged at 200 × g for 5 min and incubated at 37°C with 5% CO₂. After 1.5 h, cells were washed three times with PBS. Next, cells were further incubated for 1.5 h with meropenem to kill extracellular bacteria. Cells were then rinsed again three times with PBS and lysed with 0.2% Triton X-100 for 20 min. Serial dilutions of the lysate were plated for LB agar plates to determine the number of CFUs per unit volume. Three biological replicates per strain were used per experiment.

Mid-log-phase K. pneumoniae in DMEM medium containing 10% FBS were added to the wells (at MOI 50), incubated for 2 h, and washed three times with PBS. Cells were then incubated for 2 h with fresh medium containing meropenem. Finally, the number of intracellular bacteria was determined by the number of CFUs on the LB agar.

For serum survival assay, the mid-log phase bacteria cells (1.5x10⁶ CFU/ml) were mixed with normal human serum at a 1:3 ratio and incubated at 37°C for 2 h. Bacteria were grown overnight in LB for enumeration of viable bacteria.

G. mellonella killing assays were carried out on the six isolates. Ten larvae weighing approximately 300 mg were randomly selected for each strain. For each of these, 10 μl of bacterial suspension (1×10⁶ CFU/ml) in PBS or 10µl of PBS (control group) was injected into the last left proleg. The insects were incubated at 37°C in the dark and observed for 96 h. Larvae were considered dead when they repeatedly failed to respond to physical stimuli.

Total RNA was extracted from the bacteria in the logarithmic growth phase. Following this, the RNA was converted to cDNA using the PrimeScript™ RT Reagent Kit. Real-time PCR was performed using a LightCycler^® System. Relative gene expression levels were measured in terms of Ct values and analyzed using the 2^-ΔΔCt method. The primer sequences are presented in Table S3 . RpoB was used as an internal control. Three biological replicates per strain were used.

All experiments were performed with biological replicates, and graphical and numerical data analysis was carried out using Graphpad Prism 9.0. Analyses included two-tailed Student’s t-tests, one-way analysis of variance (ANOVA), two-way ANOVA, and log-rank tests. Statistical significance was defined as p < 0.05.

Multilocus sequence typing (MLST) of 226 hvKp strains showed that ST11 was the most prevalent sequence type (ST), followed by ST23, ST48, and ST86 ( Figure 1A ). Additionally, the most prevalent capsule types were the K64, K1, and K2 serotypes ( Figure 1B ). These results are consistent with previous findings. Previously, hvKP strains have usually been found to be associated with sequence types ST23, ST65, and ST86 and with serotype K1/K2. However, in recent years, virulence plasmids have been acquired in CRKP strains to an increasing extent (ST11), creating hypervirulent and carbapenem-resistant K. pneumoniae (CR-hvKP) and resulting in widespread epidemics (Lan et al., 2021; Tian et al., 2022).

We also analyzed the expression of wcaJ/wbap in different serotypes. Considering the transfer of Gal-1-P from UDP-galactose or UDP-glucose to und-P (Whitfield, 2006), we analyzed wcaJ and wbap sequences by constructing evolutionary trees separately. Firstly, as shown in Figure 1C , wcaJ was present in the K1, K2, K5, K16, K23, K24, K25, K30, K39, and K64 serotypes, whereas wbaP was present in the K19, K20, K21, K43, K47, K51, K57, K62, and K112 strains ( Figure S1 ). The results showed that wcaJ and wbaP sequences of the same serotype were clustered in the same branch, indicating that wcaJ or wbaP sequences in the same serotype are relatively conserved. However, different nucleotide mutations also appeared in the wcaJ sequence of the same serotype, such as the K1 serotype ( Figure S2 ).

To date, two main categories of hvKp serotype appear to have been identified: 1) K1 and K2; and 2) K64. Therefore, we next focused on studying the wcaJ sequences in the three common hvKp serotypes: K1, K2, and K64. To this end, a total of 48 hvKp isolates were collected from six hospitals, including K1 (n = 20), K2 (n = 13), and K64 (n=15). The results, as presented in Figure 2 , showed that the wcaJ sequences were identical within strains belonging to the same serotype (including K1, K2, and K64 clinical strains). Furthermore, 75% of the K1 serotype was composed of ST23, while the K2 serotype comprised a greater variety of STs, such as ST65, ST86, and ST25. Finally, the K64 serotypes were mostly composed of ST11. Interestingly, a majority of the K1/K2 strains were hypermucoid, whereas most of the K64 strains were not. It is unknown whether different wcaJ sequences play an important role in hypercapsule formation.

The effect of wcaJ on capsule formation was analyzed by constructing various mutants: K2044^K1wcaJ, K2044^K2wcaJ, and K2044^K64wcaJ. We found that the K2044^K1wcaJ and K2044^K2wcaJ recombinants exhibited significantly slower rates of growth compared to other mutants ( Figure 3A ). However, the presence of pACYC184 plasmid did not affect the growth of the strain. The growth defects in K2044^K1wcaJ and K2044^K2wcaJ require further exploration. Next, electron microscopy results showed that K2044, K2044^K1wcaJ, and K2044^K2wcaJ (but not K2044^ΔwcaJ or K2044^K64wcaJ) produced a loose and thick mucus substance ( Figure 3B ), suggesting that the K2044^K64wcaJ mutants had impaired capsule formation abilities, similar to those of the K2044^ΔwcaJ mutants.

Viscosity and uronic acid were subsequently measured. The results indicated that knockout of wcaJ significantly reduced capsular viscosity, and the complement of p-K1wcaJ or p-K2wcaJ caused hypercapsulation; these were even more hypermucoid than the K2044 parent strains ( Figure 3C ). In addition, capsule formation was quantified by measuring uronic acid. As shown in Figure 3D , decreased uronic acid was observed in K2044^ΔwcaJ, but K2044^K1wcaJ and K2044^K2wcaJ produced uronic acid to an extent similar to K2044. The biofilm, an important bacterial protective mechanism, is another substance existing on the surface of cells. There was an inverse correlation between capsule production and biofilm formation ( Figure 3E ), as the K2044^ΔwcaJ and K2044^K64wcaJ capsule-null mutants were better able to form biofilms than K2044, K2044^K1wcaJ, or K2044^K2wcaJ. Furthermore, there was no significant difference between K2044^ΔwcaJ and K2044^K64wcaJ in terms of capsular viscosity, uronic acid, or biofilm production.

To summarize, the plasmids carrying K64 wcaJ exhibited impaired capsule formation, whereas the complement of K2 wcaJ did not. The associated molecular mechanism was unclear. A comparison was made of the composition of cps loci among K1, K2, and K64 serotypes. The results revealed the presence of conserved genes within the three serotypes, including galF, wzi, wza, wzb, wzc, wcaJ, gnd, manC, manB, and ugd ( Figure 4A ). Additionally, wcaJ protein alignment showed that F244-Y467 were identical in K1 and K2, but not in the K64 serotype ( Figure 4B ).

Previous studies have demonstrated that the hypercapsule of hvKp is essential for resistance to phagocytosis. Compared with K2044^ΔwcaJ, the bacterial invasion of RAW264.7 and Caco2 cell lines was equally significantly suppressed in the K2044, K2044^K1wcaJ, and K2044^K2wcaJ groups. However, the number of bacterial invasions did not differ between the K2044^ΔwcaJ and K2044^K64wcaJ groups ( Figure 5 ). The above results indicate that the presence of the capsule reduced the invasion and phagocytosis of hvKp strains.

Since the capsule acts as an important virulence factor, we wondered whether wcaJ sequence variations could affect the virulence of hvKp isolates. As shown in Figure 6A , K2044, K2044^K1wcaJ. and K2044^K2wcaJ exhibited significantly higher levels of serum resistance than K2044^ΔwcaJ. As shown in Figure 6B , when the larvae were infected with the same bacterial inoculum, K2044^ΔwcaJ strains exhibited significantly decreased virulence levels compared to K2044, K2044^K1wcaJ, and K2044^K2wcaJ. K2044^K1wcaJ and K2044^K2wcaJ exhibited similar virulence effects to K2044, but K2044^K64wcaJ did not. The above results instead confirm that the synthesis of hypermucoviscous capsules significantly enhanced the virulence of a given K. pneumoniae strain.

To examine whether wcaJ mutants induced different inflammatory responses in vitro, RAW 264.7 cells were treated with different recombinants and cytokine production was measured by ELISA. The results showed that release of IL-1β reached a peak at 8 h ( Figure 6C ), IL-6 at 16 h ( Figure 6D ), IL-8 at 16 h ( Figure 6E ), and IFN-γ at 8 h ( Figure 6F ). Compared to K2044^ΔwcaJ, the K2044, K2044^K1wcaJ, and K2044^K2wcaJ strains significantly promoted the production of IL-6 and IL-8, but K2044^K64wcaJ did not. This phenomenon suggests that the presence of the capsule promoted the inflammatory response. Additionally, it is worth noting that no significant production of IL-1β or IFN-γ was observed ( Figure 6F ).

Studies have demonstrated there is a strong correlation between p-rmpA/rmpA2 and hypermucoviscosity/hypervirulence. Therefore, p-rmpA/rmpA2 is among a set of genes proposed as biomarkers for detection of hvKp strains (Russo et al., 2018). A poly-G frameshift leads to the early termination of p-rmpA2 transcription. The resulting truncated product is 99 amino acids long and lacks the C-terminal DNA-binding region. We therefore next focused on the regulatory effect of p-rmpA on capsule production and virulence.

Phylogenetic tree results indicated that rmpA sequence was conserved regardless of serotype ( Figure 7 ). The CPS biosynthesis capacity of the K2044^ΔrmpA mutant was then determined. Compared to the K2044 strains, mucoviscosity and uronic acid were markedly decreased in the K2044^ΔrmpA mutants ( Figures 8A, B ). The cps locus contains three characterized promoters, which are located upstream of galF, wzi, and manC ( Figure 8C ). We found that the expression of galF, wzi, and manC was decreased in the K2044^ΔrmpA strains ( Figure 8D ). Thus, the capsule alterations described above may have been caused by the reduced expression of genes associated with CPS biosynthesis. In consideration of the changes affecting the capsule, the virulence of K2044^ΔrmpA mutants was evaluated via G. mellonella testing. The results demonstrated that the p-rmpA-mediated reduction in capsule production significantly impaired the virulence of K. pneumoniae in G. mellonella ( Figure 8E ).