Clinical KP strains, including KPZ2 and its phage-resistant mutant KPZ2-R, were isolated from the clinical laboratory at Henan Provincial Hospital of Traditional Chinese Medicine and preserved in 25% glycerol at -80°C. The strains were identified using the IVD MALDI Biotyper mass spectrometer (Bruker Daltonik, Germany). Antimicrobial susceptibility testing was performed using the VITEK 2 Compact system (bioMérieux, France), and results were interpreted according to the Clinical and Laboratory Standards Institute (CLSI) guidelines. The strains were revived on blood agar plates and incubated overnight at 37°C with 5% CO₂. After that, one colony was added to Luria-Bertani broth and cultured at 37°C with shaking at 180 rpm until the OD600 reached 0.6–1.0, indicating the logarithmic growth phase.

Phages were isolated from untreated hospital wastewater using a modified enrichment protocol (Wang et al., 2024). To screen for lytic phages, several clinical KP isolates, including KPZ2, were randomly selected as candidate host strains. After being stored at 4°C for 2 h, untreated sewage samples were concentrated at 8000 rpm for 10 min and filtered through a 0.22 μm filter. The filtered sewage was then mixed with an equal volume of 2× LB broth and inoculated with a clinical isolate (OD600 = 0.5). The mixture was incubated overnight at 37°C with shaking at 120 rpm. The phage stock was obtained by collecting the supernatant after centrifuging and filtering the supernatant through a 0.22 μm filter. The double-layer agar plate approach was used to purify phages until a single type of phage morphology was observed.

The agarose gel blocks containing phages were incubated in PBS at 4°C for 6 h, then centrifuged and filtered through a 0.22 μm filter to obtain phage suspension. The phage suspension was concentrated with a 30 kDa ultrafiltration tube at 3000 × g and 4°C for 15 min. Phages were adsorbed onto the 200-mesh carbon-coated copper grid and stained with 2% phosphotungstic acid. Phage morphology was observed using a Hitachi HT7800.

The optimal MOI for phage HZJ31 was determined with a modified protocol (Wang et al., 2024) by incubating it with KPZ2 at MOI values of 0.001, 0.01, 0.1, and 1 at 37°C with shaking at 200 rpm for 4 h. The mixture of cultures was centrifuged at 6000 rpm for 15 min and the supernatant passed through a 0.22 μm filter to remove bacterial cells. Phage titers were determined using the double-layer agar method.

The one-step growth curve of phage HZJ31 was evaluated using the method (Fei et al., 2023). Briefly, phages HZJ31(MOI=0.001) and KPZ2 were incubated at 37°C for 15 min to allow adsorption. The mixture was then centrifuged at 12,000 rpm for 5 min at 4°C, and the supernatant was removed. The pellet was resuspended in LB broth and incubated at 37°C with shaking at 200 rpm for 100 min. The samples were collected at 10-minute intervals, and phage titers were quantified to generate the growth curve. The burst size = (Final phage titer - Initial free phage titer​)/Number of infected bacterial cells.

The host range of phage HZJ31 was determined by the spot assay according to the protocol described previously (Wang et al., 2024). Clinical CRKP isolates were adjusted to an OD600 of 0.5 using PBS, and the bacterial suspensions were evenly spread on LB agar plates. Then, 10 μL of phage HZJ31 was immediately spotted onto the surface, followed by overnight incubation at 37°C. The host range was evaluated through plaque formation: ++ denotes clear plaques, + signifies faint plaques, and - indicates the absence of plaque. All experiments were conducted in triplicate for consistency.

The stability of phage was carried out according to the published paper (Fei et al., 2023). The thermal stability of phage HZJ31 was tested by incubating the phage (1×10¹⁰ PFU/mL) at temperatures of 4, 25, 37, 50, 60, 70, and 80°C for 1 h. For pH stability, phage HZJ31 was mixed with SM buffer at pH values of 3, 5, 7, 9, 10, 11, and 13 in a 1:10 (v/v) ratio and incubated at 37°C for 1 h. Then, phage titers were determined.

The procedures were performed according to the previous study with minor modifications (Wang et al., 2024). The cells were plated at a density of 1×10⁴ cells per well in a 96-well plate and differentiated with 200 ng/mL of phorbol 12-myristate 13-acetate (PMA) overnight at 37°C with 5% CO₂. Phage HZJ31 was diluted from 10⁸ to 10¹⁴ PFU/mL in RPMI-1640 with 10% FBS and incubated with cells at 37°C, 5% CO₂ for 24, 48, and 72 h. Following incubation, 100 μL of a CCK-8 premix (Beijing Solarbio, China) with 90% PBS was added to each well after the cells were washed twice with PBS. After an additional incubation for 3 h, absorbance at OD450 nm was detected.

The Phage DNA was extracted utilizing the Viral Gene Extraction Kit (Takara Bio, China). A solution containing 200 μL Buffer VGB, 20 μL Proteinase K, one μL Carrier RNA, and 200 μL phage suspension (10⁸ PFU/mL) was incubated at 56°C for 10 min.200 μL of anhydrous ethanol was added and mixed before transferring to a spin column. The mixture was centrifuged at 12,000 rpm for 2 min, and the filtrate was discarded. Subsequently, 500 μL of Buffer RWA and 700 μL of Buffer RWB were added in sequence, centrifuged, and discarded, with Buffer RWB applied twice, followed by centrifugation at 12,000 rpm for 3 min. The spin column was then transferred to a new RNase-free tube, and 60 μL of RNase-free ddH2O was added to elute the DNA. This mixture was incubated for 5 min and centrifuged at 12,000 rpm for 5 min to collect the phage DNA.

The phage genome was sequenced using the Illumina HiSeq 2500 platform, yielding a high average sequencing depth of 24,068×. Raw reads were filtered and quality-checked using fastp, and de novo assembly was performed with SPAdes. The quality of the assembly was evaluated using QUAST, resulting in an N50 value of 40,624 bp, indicating good assembly continuity. The genome was uploaded to the National Center for Biotechnology Information (NCBI). It was then analyzed with BLASTn to identify similar phages in the GenBank database. The tRNA genes encoded by the phage were identified using tRNAscan-SE. Phage-encoded genes were annotated using the Prokaryotic Genome Annotation System (Prokka). Functional characterization of the predicted amino acid sequences of these genes was conducted using NCBI’s BLASTp. Virulence factors and antibiotic resistance genes were identified using the VFDB and ARDB databases (thresholds of ≥70% identity and ≥70% coverage). The genomic map was visualized with CGview.

The terminase large subunit sequences were used for phylogenetic analysis. The amino acid sequence of terminase large subunit of phage HZJ31 was entered into NCBI’s BLASTp, resulting in the identification of ten homologous sequences with notable similarity. The basis for generating a phylogenetic tree using the ClustalW alignment tool within the MEGA 11.0 platform. Furthermore, the genome sequences of phage HZJ31 and the selected phage were analyzed by Easyfig.

The MICs of tigecycline (TGC) against KPZ2 were conducted by the broth dilution method. TGC was serially diluted in LB broth in twofold dilutions.100 μL of each antibiotic dilution was added to the corresponding wells of a 96-well plate. Negative control wells contained only medium without bacteria, while positive control wells contained bacteria without antibiotics. The KPZ2 bacterial suspension was diluted in LB broth to 5 × 10⁵ CFU/mL from an initial McFarland 0.5 suspension. Then, 100 μL of the bacteria was added to each well. The plate was incubated for 18 h at 37°C. The MIC was determined as the lowest concentration of TGC that inhibited visible bacterial growth.

The assay was performed according to a previously reported method (Wang et al., 2024), with minor modifications. The experimental design included the following groups: The KPZ2 group, The phage HZJ31 group (MOI = 0.001), The 1/2 MIC TGC group, and the phage HZJ31 with 1/2 MIC TGC group. The KPZ2 was standardized to 1×10⁶ CFU/mL. 100 μL of the bacterial culture was added to the 96-well plate, followed by the addition of phage HZJ31 and/or TGC, according to group assignment. The plate was incubated at 37°C for 36 h, with OD600 measurements taken hourly to construct bactericidal kinetics curves to evaluate the inhibitory effects of phage and/or TGC on bacterial growth. Synergy was defined as a statistically significant enhancement of antibacterial activity in the combination group (phage + TGC) compared to each monotherapy group, based on OD600 values measured over time. Statistical analysis was performed using two-way ANOVA followed by Tukey’s multiple comparison test (P < 0.05).

The experiments were evaluated using a Galleria mellonella(G.mellonella)larvae infection model using a protocol modified from a previous study (Peng et al., 2025). The larvae, approximately 2–3 cm in length and weighing 250–300 mg, with a uniform milky white color and no gray spots, exhibiting high vitality, were selected for the experiments (purchased from Tianjin Huiyude Biotechnology Co., Ltd., China). To determine the optimal bacterial infection dose, we selected the concentration that resulted in approximately 80% larval mortality at 24 h post-infection as the target inoculum. Before experimentation, larvae were acclimated at 37°C for 1–2 h, and KPZ2 was cultured in LB broth to reach an OD600 of 0.8-1.0. The culture underwent centrifugation at 7,000 rpm for 10 min, followed by PBS washing and adjustment to concentrations of 10⁵ to 10⁹ CFU/mL. 10 µL bacterial suspension or PBS (as control) was administered into the last left proleg of the larvae. Larvae were incubated at 37°C with survival rates monitored every 12 h.

To evaluate the combined antibacterial efficacy of phage HZJ31 and TGC in the G.mellonella infection model, larvae were randomly assigned to five groups: the PBS control group, the KPZ2 group, the phage HZJ31 group, the TGC group, and the phage HZJ31 + TGC group. Except for the PBS control group (injected with an equal volume of PBS), each larva received 10 μL of KPZ2 suspension (at a concentration determined by preliminary experiments) into the last left proleg. After 1 h, 10 μL of the respective treatment was injected into the last right proleg of each larva: PBS (for the PBS control and KPZ2 group), phage HZJ31 (MOI = 0.001, HZJ31 group), 1/2 MIC TGC (TGC group), and a mixture of phage HZJ31 and 1/2 MIC TGC (combined treatment group). All larvae were incubated at 37°C and monitored every 12 h for 4 consecutive days. Survival status was recorded, and survival rates were calculated for each group to evaluate therapeutic efficacy.

The biofilm quantification was performed using a modified protocol

(Wang et al., 2024). A 96-well plate was incubated at 37°C for 24 h, 48 h, and 72 h with KPZ2 suspension (10⁶ CFU/mL) and phage HZJ31 (MOI =0.001). For biofilm disruption, another set of KPZ2 cultures was incubated for 48 h to form biofilms, which were then treated with phage HZJ31 (MOI = 0.01) for 6 h. After washing with PBS to remove planktonic cells, biofilms were fixed with methanol for 30 min and stained with 0.1% crystal violet solution (Shanghai Baisai, China) for 15 min. The residual stain was removed by PBS washing, and biofilm biomass was quantified by dissolving crystal violet in 200 μL of 30% glacial acetic acid. Absorbance at 595 nm was measured to assess biofilm biomass. All experiments were performed in triplicate.

The experiment was conducted with suitable modifications (Fei et al., 2023). KPZ2 suspension (10⁶ CFU/mL) and phage HZJ31 (MOI = 0.001) were co-incubated in a 96-well plate at 37°C for 24 h, 48 h, and 72 h. For biofilm disruption, preformed 48 h biofilms were treated with HZJ31 (MOI = 0.01) for 6 h. After incubation, PBS was used to wash the plate to remove planktonic bacteria. Subsequently, 200 μL of LB broth and 20 μL of XTT solution (Shanghai Yuanye, China) were introduced into each well. The plate was then incubated at 37°C in the dark for 3 h, followed by absorbance measurement at OD490 nm. The inhibitory rate was calculated using the following formula:

[ 1 − A ( Experimental group ) − A ( Blank control group ) A ( Control group ) − A ( Blank control group ) ] × 100 % .

The experiment was repeated three times.

The experiment was performed with some modifications based on a previous study (Wang et al., 2024). KPZ2 and phage HZJ31 (MOI =0.001) were incubated at 37°C with shaking at 200 rpm for 24 h. Cultures were streaked onto blood agar plates to isolate single colonies. The isolates were identified by the mass spectrometer. Phage resistance was confirmed by spot testing and double-layer agar assays. Resistant strains were designated KPZ2-R.

KPZ2 and KPZ2-R were incubated at 37°C with shaking at 220 rpm for 6 h. The cultures were then centrifuged at 8000 rpm for 10 min to collect the pellets, which were subsequently washed with PBS. Pellets were fixed in 2.5% glutaraldehyde at room temperature for 5 h, followed by PBS washing and 15-minute dehydration using graded ethanol concentrations (30%, 50%, 70%, 90%, 100%). After dehydration, the samples were dried, gold-coated, and observed under the SEM Hitachi SU8100 to capture morphology.

Antibiotic susceptibility of KPZ2 and KPZ2-R was determined using the VITEK-2 Compact system (BioMérieux, France). The KPZ2 and KPZ2-R were standardized to a 0.5 McFarland scale. The prepared bacterial suspension was loaded along with the AST-GN card into the VITEK-2 Compact system for automatic inoculation and determination. The results were analyzed following the guidelines of the Clinical and Laboratory Standards Institute (CLSI).

The assay was performed using a modified protocol (Wang et al., 2024). KPZ2 and KPZ2-R were incubated with phage HZJ31 (MOI =0.001) at 37°C, with samples collected at 0, 5, 10, 15, and 20 min. The samples were centrifuged at 11000 rpm for 8 min at 4°C. The supernatants were filtered (0.22 μm), and then the phage titer was determined. The Adsorption Efficiency = [(Initial Phage Titer - Phage Titer in Supernatant)/Initial Phage Titer] × 100%.

KPZ2 and KPZ2-R were adjusted to 1.0 × 10⁶ CFU/mL. The 96-well plate was prepared with 200 μL of bacterial suspension in each test well and 200 μL of LB broth in the control wells. The plate was incubated at 37°C for 24 h with hourly OD600 measurements.

The concentrations of KPZ2 and KPZ2-R were adjusted to 10⁶ and 10⁷ CFU/mL, respectively. Each G.mellonella larvae received an injection of 10 µL bacterial suspension (PBS as a control) into the last left proleg. Larvae were incubated at 37°C with survival monitored every 12 h over 4 days.

Genomic DNA of strains KPZ2 and KPZ2-R was extracted using a whole-genome extraction kit (Sangon Biotech, China) according to the manufacturer’s instructions. Briefly, 5 mL of overnight bacterial culture was centrifuged to collect the cells, followed by the addition of lysis buffer and incubation at 65 °C for 1 h to lyse the cells. Subsequently, Buffer PB was added, mixed thoroughly, and briefly incubated at −20 °C. After centrifugation, the supernatant was transferred, isopropanol was added, mixed well, and centrifuged again to pellet the DNA. The DNA pellet was washed with 75% ethanol, dried after removing the ethanol, and then dissolved in TE buffer. DNA concentration and integrity were assessed using a Qubit 4.0 fluorometer (Thermo Fisher Scientific, China). The extracted DNA was either used for subsequent experiments or stored at −20°C.

The genomes of strains KPZ2 and KPZ2-R were sequenced using the BGI MGISEQ 2000 platform. The extracted DNA was enzymatically fragmented, end-repaired, ligated with adaptors, and amplified by PCR to construct sequencing libraries. Library concentration was quantified using a Qubit fluorometer, and fragment size distribution was assessed using the Agilent 2100 Bioanalyzer; libraries with a main peak around 500 bp were considered qualified. The libraries were then denatured, circularized, and purified before sequencing. Raw sequencing data were subjected to quality control to remove adaptors and low-quality reads. The filtered reads were assembled using SPAdes (v3.13.0). Based on alignment results, ANGSD (v0.940) was used to reconstruct the genome sequences, and genome completeness and contamination were evaluated using CheckM (v1.2.2).

The K-locus types of KPZ2 and KPZ2-R were identified using the Kaptive Web tool (https://kaptive-web.erc.monash.edu/). Genome assemblies in FASTA format were uploaded, and typing was based on sequence coverage, identity, and completeness. Confidence levels were recorded for each result.

All statistical analyses were conducted using GraphPad Prism 9.2. One-way or two-way analysis of variance (ANOVA) followed by Tukey’s multiple comparisons test was employed to undertake multiple group comparisons. Survival curves for G.mellonella were analyzed using the log-rank (Mantel-Cox) test. Statistical significance was indicated by *P <0.05, **P <0.01, ***P < 0.001, and ****P <0.0001.

Phage HZJ31 was isolated using strain KPZ2 as the host bacterium from wastewater. TEM revealed the phage had an icosahedral head approximately 80 nm in diameter and a short tail measuring 10 nm ( Figure 1A ). The plaques were large, clear, and uniform, reaching up to 4.0 mm in diameter, with a halo of approximately 7.2 mm, which indicates the high efficiency ( Figure 1B ). Phage HZJ31 was classified within the Caudovirales based on its morphological characteristics.

The double-layer agar plate method was employed to determine the optimal MOI, one-step growth, and stability of phage HZJ31. Phage titer of HZJ31 remained largely stable across MOIs from 0.1 to 0.0001, peaking at an MOI of 0.001, which was chosen for further experiments ( Figure 1C ). The one-step growth curve showed a 10-minute latent period, followed by a rapid exponential rise in phage population between 10 and 70 min, culminating in a plateau phase. The estimated burst size was 374 PFU/cell, suggesting that HZJ31 possesses a highly efficient host cell lysis capability ( Figure 1D ).

The environmental stability of phage HZJ31 was further assessed under varying pH and temperature conditions. Phage HZJ31 exhibited stable titers between pH 5 and 9, with the highest stability observed around neutral pH. In contrast, the phage titer drastically decreased under highly alkaline (pH > 9) and acidic (pH < 5) conditions, becoming nearly undetectable at pH 3 ( Figure 1E ). The thermal stability of HZJ31 was robust between 4°C and 50°C, but its titer began to decrease beyond 50°C, and complete inactivation occurred at 80°C ( Figure 1F ). The stability of HZJ31 indicates its potential suitability for phage therapy applications. Phage HZJ31 demonstrates strong host specificity, with a lysis rate of 23.33% (7/30) against clinical CRKP isolates and only 13.33% (4/30) achieving complete lysis ( Table 1 ). Furthermore, CCK-8 assays showed that co-culturing THP-1 cells with HZJ31 at concentrations ranging from 10⁸ to 10¹¹ PFU/mL for 24 h and 48 h did not significantly affect cell viability compared to the untreated control (mean ± SD, n=6, Figure 1G ), confirming the cellular safety of HZJ31 for therapeutic applications.

The genome of Phage HZJ31 (GenBank: OR050820.1) was sequenced using Illumina HiSeq 2500 platform, identifying a circular dsDNA molecule of 40,624 bp with a GC content of 53.16%. The genomic map was constructed using GCview ( Figure 2A ). Annotation using Prokka identified 48 ORFs, of which 33 were functionally annotated and classified into four major categories: structural proteins, DNA replication and metabolism, lysis mechanisms, and genome packaging ( Table 2 ). In the structural module, ORF1 encodes a major capsid protein (98.84% identity with phage IME264), and ORF4 encodes a host adsorption protein (100% identity with phage K11), both essential for host recognition and infection. Tail-associated proteins such as ORF40 and ORF41 (tail fiber proteins) contribute to phage structural assembly. Replication-related genes include ORF12 (DNA polymerase I) and ORF16 (primase/helicase), while ORF30 encodes a DNA-directed RNA polymerase, suggesting the phage can replicate and transcribe independently of the host. Lysis functions are attributed to ORF20, encoding an amidase/lysin, and ORF39, encoding a holin, which together facilitates host cell wall degradation and phage release. Packaging proteins ORF36 and ORF38 encode the large and small terminase subunits, respectively, responsible for DNA cleavage and encapsidation. Overall, HZJ31 harbors a complete set of functional modules characteristic of a lytic phage, supporting its potential as a therapeutic agent against CRKP infections.

The absence of virulence or antibiotic resistance genes in the VFDB and ARDB databases highlights the phage’s biosafety for therapeutic use. Genomic analysis using BLASTn revealed that phage HZJ31 exhibits 94.84% sequence identity with Klebsiella phage 066012 (GenBank: MW042787.1) with 94% coverage. The terminase large subunit gene of HZJ31 was found to be highly conserved, and phylogenetic analysis based on this gene positioned HZJ31 in close relation to Klebsiella phage SH-KP152226 (QDF14649.1) ( Figures 2B, C ). These findings collectively indicate that HZJ31 represents a novel Klebsiella-specific phage, expanding our understanding of phage diversity and potential for therapeutic exploitation.

The schematic diagrams of the in vitro bactericidal kinetics assay and the G. mellonella infection model are shown in Figures 3A and 3B , respectively. In this study, the combined antibacterial effect of phage HZJ31 and TGC was first evaluated in vitro. Prior to combination testing, the MIC of TGC against strain KPZ2 was determined to be 1.0 μg/mL. Then the bactericidal kinetics of phage HZJ31(MOI = 0.001), TGC (1/2 MIC, 0.5 μg/mL), and their combination against strain KPZ2 were evaluated over a 36-h period ( Figure 3C ). During the initial 6 h, the bacterial growth among all groups was generally consistent, with no significant differences in OD600. In the HZJ31 group, although the OD600 values gradually increased after 7 h, they remained significantly lower than those of the KPZ2 group throughout the entire experiment. In the 1/2 MIC TGC group, bacterial growth was also delayed, with OD600 increasing at a slower rate than in the HZJ31 group, reaching approximately 0.45 at 36 h. In the HZJ31 combined with the 1/2 MIC TGC group, no significant increase in OD600 was observed during the first 10 h, followed by a slight increase thereafter. However, the OD600 of the HZJ31 with 1/2 MIC TGC group remained at a low level (< 0.3) throughout the 36-h period, significantly lower than those in all other groups (P < 0.05), indicating a synergistic antibacterial effect of the combined treatment.

Given the significant similarities between the innate immune responses of G.mellonella larvae and vertebrates, the G.mellonella model is widely used as an alternative to evaluating bacterial virulence and in vivo antimicrobial efficacy. We then evaluated the bactericidal efficacy of phage HZJ31 and TGC in the G.mellonella infection model. In this study, we established an infection model by injecting G.mellonella larvae with varying concentrations of bacteria, determining that an optimal inoculum of 10⁷ CFU/mL for KPZ2 resulted in an 80% mortality rate at 24 h ( Figure 3D ). Significant differences were observed in the survival curves among the groups (log-rank test, P < 0.05). Compared to the KPZ2 group, the survival rates at 96 h were 50% and 60% for the phage HZJ31 and TGC treatments, respectively, while the survival rate for larvae treated with the combination of phage HZJ31 and TGC was 70% ( Figure 3E ). In summary, phage HZJ31 effectively controlled KPZ2 infection in vivo.

The results showed that phage HZJ31 significantly inhibited KPZ2 biofilm formation at 24 h, 48 h, and 72 h (P < 0.05, Figure 4A ). The inhibitory effect of phage HZJ31 on bacterial growth within the KPZ2 biofilm was assessed, showing a 59.45% inhibition rate at 72 h (P < 0.05) ( Figure 4C ). When phage HZJ31 was co-cultured with preformed KPZ2 biofilm for 6 h, significant disruption of the biofilm was observed ( Figure 4B ), with bacterial inhibition within the biofilm exceeding 70% (P < 0.05, Figure 4D ). These findings demonstrate that phage HZJ31 effectively inhibits the formation of KPZ2 biofilms and disrupts preformed biofilms, exerting a substantial inhibitory effect on the bacteria within the biofilm.

After 24 h of co-culturing KPZ2 with phage HZJ31 (MOI = 0.001), the culture appeared visibly turbid, suggesting the emergence of phage-resistant mutants. Subsequent plating on LB agar and purification on blood agar yielded colonies that were no longer susceptible to phage HZJ31, as confirmed by both the double-layer agar method and spot assay, which showed no lysis zones. Mass spectrometry analysis identified all resistant colonies as Klebsiella pneumoniae, excluding the possibility of contamination. One phage-resistant isolate (designated KPZ2-R) was randomly selected from these colonies for further comparison with the wild-type strain. KPZ2 exhibited a slightly faster growth rate in LB broth than KPZ2-R, as measured by OD600 ( Figure 5A ). The phage adsorption efficiency of KPZ2 was 65.82% at 5 min, significantly higher than that of KPZ2-R (24.24%), and continued to increase over time, reaching 88.00% at 15 min. Conversely, the adsorption efficiency of KPZ2-R remained low throughout the incubation period ( Figure 5B ).

We also compared the virulence of KPZ2 and KPZ2-R using the G.mellonella infection model. At 10⁶ and 10⁷ CFU/mL of KPZ2, the survival rates of larvae at 48 h were 40% and 20%, respectively, whereas larvae injected with the same doses of KPZ2-R had survival rates of 70% and 50% ( Figure 5C ). Overall, KPZ2-R demonstrated lower virulence than KPZ2 in vivo. We further compared the antibiotic susceptibility of KPZ2 and KPZ2-R (shown in Table 3 ). Among all antibiotics tested, KPZ2 was only susceptible to TGC, polymyxin B, and ceftazidime/avibactam, while resistant to the others. In contrast, KPZ2-R showed increased susceptibility, shifting from resistance (R) to susceptibility (S) to piperacillin and gentamicin.

On blood agar, KPZ2 formed large, smooth, plump, and moist colonies with a milky white appearance ( Figure 5D ), while the phage-resistant mutant KPZ2-R formed smaller, gray, semi-transparent colonies, but no obvious difference in diameter ( Figure 5H ). SEM revealed that the surface of KPZ2 appeared smooth, dense, and uniform, exhibiting a full and thick structure without visible indentations ( Figures 5E, F ). In contrast, KPZ2-R displayed a rough, uneven surface with wrinkles or depressions, a comparatively thinner structure, and showed evident structural defects in the capsule ( Figures 5I, J ). TEM further confirmed these differences, revealing a prominent, well-defined capsule surrounding KPZ2 ( Figure 5G ), while KPZ2-R exhibited a significantly reduced and poorly defined capsule with only a faint or nearly absent surrounding layer ( Figure 5K ), indicating impaired capsule formation in the resistant mutant.

Consistent with these observations, Kaptive-based analysis revealed that both KPZ2(GenBank: CP178862.1) and its phage-resistant mutant KPZ2-R (GenBank: CP192283.1) matched the KL47 capsular locus with 99.67% sequence identity ( Figures 5L, M ). KPZ2 exhibited complete locus coverage (99.98%) and intact capsule biosynthesis genes. However, KPZ2-R was classified as “capsule null” due to a truncated wcaJ gene (51.58% identity), suggesting the loss of capsule synthesis capability is associated with phage resistance.