For phage isolation, a ST29 CR-hvKP clinical strain with K54 serotype, SCNJ1 (Yuan et al., 2019), which was recovered from the sputum of a patient with acute bronchiolitis in a hospital in Sichuan Province in November 2018, was employed as the host strain. vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, and vB_KpnM_SCNJ1-Y were isolated from the sewage water of the Southwest Medical University’s Affiliated Hospital for Traditional Chinese Medicine, Yangtze River, and Yudai River in Luzhou, Sichuan, China, respectively. 100 mL of sewage or river water was centrifuged for 20 min at 12,000 rpm, 4°C to remove debris, and a 0.22-μm filter (JETBIOFIL, China) was used to filter the supernatant. To enrich phage, 30 mL filtered supernatant with 15 mL 3×LB broth was inoculated with 200 µL exponential host cells and cultured for 12 h at 37°C with shaking at 220 rpm. The mixture was centrifuged at 12,000 rpm, 4°C for 10 min, and the supernatant was collected to isolate the phage using the double-layer agar method (Adams, 1959). Single plaques were purified in 5 passages until the formation of uniform plaques was obtained. The plaque-purified phages were stored in SM buffer at -80°C for further experiments.

The TEM method was performed as previously described by Taruna Anand et al. with slight variations (Anand et al., 2020). Briefly, the purified phages were placed on carbon-coated copper grids, allowed to stand for 5 min, and stained with 2% phosphotungstic acid for 30 seconds. The morphology of the phages was observed under a Hitachi transmission electron microscope HT7820 (Japan) at 80 kV.

The method of optimal multiplicity of infection is similar to those described by Mingfang Pu et al. with slight changes (Pu et al., 2022). Briefly, exponential host cells were mixed with the phages with different multiplicities of infection (MOI: 0.0001,0.001,0.01,0.1,1,10,100), then shaken for 12 h at 37°C (220 rpm). The phage titers were determined using the double-layer agar method. The proportion with the highest phage titer was the optimal multiplicity of infection. One-step curves were performed as previously described (Hyman and Abedon, 2009). Briefly, 1mL of phage (1×10⁸ PFU) was mixed with 9 mL of the exponential host cells (1×10⁹ CFU) at the MOI of 0.1. The mixture was incubated at 37°C with shaking (150 rpm), 5 min after mixing phages and bacteria, the mixture was 1000 fold diluted in LB medium to lower the bacterial concentration and avoid reinfection. The diluted mixture was continued to shake and 100 μL sample was collected at 2-min intervals during 30 min for titration. The assay was performed in triplicate.

The method of host range determination was performed as previously described by Yuting Shang et al. with slight variations (Shang et al., 2021). To investigate the host range of the phages, spot-test assays (Tan et al., 2020) were performed on 50 clinical strains ( Supplementary Table S1 ). Briefly, 5 µL filtered phage suspension (~10¹⁰ PFU/mL) was dropped onto plates that were covered with different bacterial lawns. The plates were incubated at 37°C for 24 h to observe plaque formation.

The methods of pH and thermal stabilities were performed as previously described by Narges Torkashvand et al. with slight changes (Torkashvand et al., 2023). In brief, for the pH stability test, 100 µL phage (~10¹⁰ PFU/mL) was added to 900 µL SM buffer of different pH (1-14) and incubated for 1 h at 37°C. For the thermal stability test, 100 µL phage suspension (~10¹⁰ PFU/mL) was mixed with 900 µL SM buffer, and incubated for 1 h at temperatures ranging from 20 to 80°C. The phage titers were determined by the double-layer agar method. All assays were performed in triplicate.

To determine whether there are lipids in the viral capsid, the sensitivities of the phages to chloroform were examined. The method of chloroform susceptibility test was performed as previously described by Yubing Chen et al. with slight changes (Chen et al., 2023). In brief, 1 mL phage suspension was mixed with 100 µL chloroform, the mixture was vigorously shaken for 1 min and then incubated at room temperature for 30 min. The mixture was centrifuged at 10,000 rpm for 10 min and the supernatant was collected and plated for phage titer using the double-layer agar method.

The inhibitory effects of the phages on biofilm formation were evaluated as previously described with slight modifications (Mulani et al., 2022). Overnight grown host cells were diluted 1:100 with LB broth to approximately 10⁷ CFU/mL and mixed with phages with different MOI (100, 10, 1, and 0.1). 200 µL of the mixture, 200 µL of LB as negative controls, and 200 µL of bacterial suspensions as positive controls were added individually in triplicate into a sterile 96-well microplate. The 96-well microplate was incubated for 24 h at 37°C, then, washed three times with sterile phosphate-buffered saline (PBS). After 15 min of methanol fixation, all wells were added with 200 µL of 0.1% ammonium oxalate crystal violet solution, stained for 20 min, washed three times with ultrapure water, dried, and solubilized with 200 µL of 33% acetic acid for 10 min. Optical density was determined at 595 nm using a microplate reader.

The method of antibacterial activity in vitro was performed as previously described by Jingyun Fu et al. with slight changes (Fu et al., 2023). Briefly, the host cells were cultured to exponential phase, diluted to 10⁷ CFU/mL (10 mL), and infected with phages (10 mL) with an MOI of 1, 10, and 100. The mixture was incubated at 37°C, 150 rpm. For the control group, 10 mL LB and 10 mL bacteria (10⁷ CFU/mL) were mixed and cultured under the same conditions. 100 µL of each mixture was removed at 0, 1, 2, 3, 4, 5, and 6 h, diluted, and plated on LB plated to count CFU per milliliter. Each experiment was carried out in triplicates.

The method of DNA extraction was performed as previously described by Na Li et al. with slight changes (Li et al., 2021). In brief, filtered phages (16 mL) were incubated for 1 h at 37°C with 10 µL RNaseA (25 mg/mL) and 20 µL DNaseI (1 U/µL) to remove host bacterial nucleic acid, followed by the addition of the 10% (w/v) polyethylene glycol 8000 (PEG8000) and sodium chloride (NaCl) solution. The mixture was precipitated at 4°C overnight and centrifuged at 11,000 rpm for 30 min at 4°C. The pellet which contained phage particles was resuspended with 1 mL of SM buffer and mixed with an equal volume chloroform to remove the proteinaceous material. The process was repeated twice and the aqueous phase was used to extract phage DNA with the omega BIO-TEK Viral DNA Kit D3892 (omega, Norcross, GA, USA) following the manufacturer’s protocol.

The genomic DNA of three phages was sequenced by Shanghai Majorbio Bio-Pharm Technology Co., Ltd. (Shanghai, China) using the HiSeq 2000 (Illumina, San Diego, CA, USA) Sequencer. The raw sequenced reads were quality controlled and trimmed with FastQC 0.11.2 (https://www.bioinformatics.babraham.ac.uk/projects/fastqc/) and Trimmomatic 0.36 (http://www.usadellab.org/cms/?page=trimmomatic), respectively. The high-quality reads were assembled into a single raw contig via SPAdes 3.5.0. (http://cab.spbu.ru/software/spades/). Genes were predicted and annotated using Rapid Annotation using Subsystem Technology (RAST, http://rast.nmpdr.org/). Circular genome maps of the phage genomes were drawn using CGView (https://paulstothard.github.io/cgview/). The antimicrobial resistance genes (AMRs) and the virulence genes were screened in the Comprehensive Antibiotic Resistance Database (CARD) (https://card.mcmaster.ca/analyze) and the VirulenceFinder (https://cge.food.dtu.dk/services/VirulenceFinder/), respectively. Phylogenetic trees based on large terminase and major capsid protein were conducted by the Neighbor-joining method using MEGA11.0 with 1000 bootstrap replications. The complete genome sequence of the three phages was aligned with other phages using the BLASTn tool in the NCBI database, and complete genome sequence similarity between the three phages and other phages was visualized using the Circoletto program (http://bat.ina.certh.gr/tools/circoletto/).

The method of mouse experiments was performed as previously described by Mingfang Pu et al. with slight changes (Pu et al., 2022). Briefly, 91 female specific-pathogen-free (SPF) BALB/c mice, aged 4-5 weeks and weighed 18-20 g, were purchased from Beijing Huafukang Biotechnology Co., Ltd. All animal experiments were followed National Guidelines for Experimental Animal Welfare (Ministry of Science and Technology of China) and approved by the Animal Welfare and Research Ethics Committee at Southwest Medical University (swmu20230036). To construct immunodeficient mouse models, all mice were given cyclophosphamide (LKT labs, America) at a dose of 300 mg/kg for three days before the experiment.

Thirteen mice were randomly assigned to each group and anesthetized by inhalation of isoflurane. Four experimental groups and the positive control group were infected with 50 µL (1 × 10⁶ CFU/each) of SCNJ1 strain via nasal drip, two hours after infection, 50 µL of phages (1 × 10⁸ PFU/each) of vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, vB_KpnM_SCNJ1-Y, a cocktail (vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, and vB_KpnM_SCNJ1-Y in equal proportion) or 50 µL PBS was administered via nasal drip, respectively. The negative group and the safe test group were given 50 µL PBS first and treated with either the 50 µL of PBS or the cocktail after 2 hours. The mice were observed for survival curves every 24 h for seven days. 30 hours after phage treatment, three mice were randomly selected from each group for dissection. The left lung, liver, and spleen were removed under sterile conditions, and homogenized, and the bacterial load of each organ was determined using the bacterial dilution plate method. The right lung was fixed in 10% formalin, embedded, sectioned, and stained with hematoxylin and eosin (H&E) to observe pathological changes.

All data were analyzed using GraphPad Prism 9.4.0 (GraphPad Software, Inc., San Diego, CA, USA) and expressed as means and standard deviation values. Survival curve analyses were performed using Kaplan–Meier survival analysis with the log-rank test. Student’s t-test was utilized to compare two groups and one-way analysis of variance (ANOVA) was used to compare multiple groups. Statistical significance was set at p<0.05.

Three novel lytic phages, named vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, and vB_KpnM_SCNJ1-Y, formed circular translucent plaques on lawns of K. pneumoniae SCNJ1 with haloes distributed around the plaque center. The plaque images taken every 24 h for 3 days revealed the haloes expanding in size with time and indicated the presence of depolymerase which can digest bacterial capsules (Adams and Park, 1956) ( Figure 1A ). The transmission electron microscopy (TEM) showed that vB_KpnA_SCNJ1-Z had an icosahedral head, with a head diameter of 64.8 ± 2.0 nm and a tail length was 12 ± 2.5 nm ( Figure 1B ); vB_KpnS_SCNJ1-C had an icosahedral head of 68.9 ± 1.9 nm and long contractile tail of 143.0 ± 1.0 nm ( Figure 1C ); vB_KpnM_SCNJ1-Y revealed an icosahedral head with a diameter of 77.5 ± 6.1 nm and connected tail of 89.0 ± 8.8 nm in length ( Figure 1D ). Under the TEM, vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, and vB_KpnM_SCNJ1-Y exhibited podovirus/Autographiviridae, siphovirus, and myovirus morphologies, respectively.

The complete genome sequences of the three phages were obtained through whole genome sequencing (WGS). vB_KpnA_SCNJ1-Z (Accession no. OQ689084), vB_KpnS_SCNJ1-C (Accession no. OQ718882), and vB_KpnM_SCNJ1-Y (Accession no. OQ689083), possess double-stranded DNA genomes with sizes of 43,428 bp, 46,039 bp, and 50,360 bp, and GC contents of the phages are 54.04%, 47.63%, and 48.52%, respectively.

The genome of vB_KpnA_SCNJ1-Z contains 57 putative open reading frames (ORFs), with 26 annotated and 31 uncharacterized ORFs ( Figure 2 ; Supplementary Table S2 ). The annotated ORFs are grouped into three major functional modules: (i) structure and packaging (8 ORFs), mainly including head and tail-related proteins (ORF48, 49, 52, 53, 56), major capsid (ORF50), and depolymerase (ORF57); (ii) host lysis (3 ORFs), encompassing Rz-like spanin (ORF5), holin (ORF6), and endolysin (ORF7); and (iii) nucleotide metabolism and genome replication (15 ORFs), mainly including terminase (ORF1, 2), endonuclease (ORF9, 19, 41), DNA primase (ORF22), DNA helicase (ORF23), DNA polymerase (ORF28), nucleotidyltransferase (ORF29, 30), and RNA polymerase (ORF45). The genome sequence alignment showed that vB_KpnA_SCNJ1-Z had the highest coverage (91%) and identity (93.07%) with Klebsiella virus KpV2883 (Accession no. MT682065) ( Supplementary Figure 1A ), followed by Klebsiella phages vB_KpnP_SU552A (Accession no. NC_028870), KP-Rio/2015 (Accession no. NC_047779), and VLC6 (Accession no. MT197176), with identity ranging from 92.46% to 92.8% (coverage 85%) ( Supplementary Figure 2 ).

The genome of vB_KpnS_SCNJ1-C contains 83 putative open reading frames (ORFs), of which 35 are annotated and 48 are uncharacterized ( Figure 3 ; Supplementary Table S3 ). These annotated ORFs can be classified into four functional modules: (i) structure and packaging (11 ORFs), mainly comprising head and tail-related proteins (ORF17, 18, 34, 37,44, 47, 49), capsid protein (ORF19, 33), restriction alleviation protein (ORF13, 23), membrane protein (ORF29), and depolymerase (ORF50); (ii) host lysis (3 ORFs), which includes lysozyme (ORF1), Rz-like spanin (ORF3), and endolysin (ORF83); (iii) nucleotide metabolism and genome replication (19 ORFs), mainly including terminase (ORF69), endonuclease (ORF54, 56), single-stranded DNA-binding protein (ORF51), exonuclease (ORF53), and recombinase (ORF52); (iv) tRNA (2 ORFs): tRNA-Met (ORF35) and tRNA-Arg (ORF36). Genome sequence alignment showed that vB_KpnS_SCNJ1-C shared the highest coverage (70%) and identity (94.03%) with Klebsiella phage vB_KpnS_MK54 (Accession no. NC_071146) ( Supplementary Figure 1B ), followed by Vibrio phage pYD38-A (Accession no. NC_021534), Aeromonas phage pIS4-A (Accession no. NC_042037), and Klebsiella phage VLCpiS13d (Accession no. NC_071142), with identities ranging from 91.6% to 94.34% (65% coverage) ( Supplementary Figure 3 ).

The genome of vB_KpnM_SCNJ1-Y contains 80 putative open reading frames (ORFs), with 52 annotated and 28 unannotated ORFs ( Figure 4 ; Supplementary Table S4 ). The annotated ORFs are organized into three functional modules: (i) structure and packaging, consisting of 21 ORFs mainly including head and tail-related proteins (ORF13, 20, 25, 26, 31, 33-35, 42), capsid protein (ORF7, 15, 24), baseplate protein (ORF37-39), and depolymerase (ORF40); (ii) host lysis, consisting of 2 ORFs: Rz-like spanin (ORF44) and endolysin (ORF45); and (iii) nucleotide metabolism and genome replication, consisting of 29 ORFs mainly including terminase (ORF1, 2), endonuclease (ORF3, 6, 16, 17, 27, 78), transcriptional regulator (ORF68), DNA helicase (ORF64, 69), DNA polymerase (ORF50, 52), exonuclease (ORF57), and polynucleotide kinase (ORF75). The genome sequence alignment showed that vB_KpnM_SCNJ1-Y shared the highest coverage (69%) and identity (91.24%) with Klebsiella phage vB_KpnM_JustaPhage (Accession no. NC_071134.1) ( Supplementary Figure 1C ), followed by Klebsiella phage BUCT_49532 (Accession no.NC_071129.1), Klebsiella phage 1611E-K2-1 (Accession no.NC_071138.1), and Escherichia phage ZCEC13 (GenBank: NC_071140.1), which shared 94.04%, 94.16%, and 92.82% identity with coverage of 71%, 69%, and 69%, respectively ( Supplementary Figure 4 ).

Phylogenetic trees were constructed using the amino acid sequences of the major capsid and large terminase of representative virus species of the family Autographiviridae and unclassified Caudoviricetes. In both phylogenetic trees, vB_KpnA_SCNJ1-Z was clustered with viruses within the family Autographiviridae, while both vB_KpnS_SCNJ1-C and vB_KpnM_SCNJ1-Y were clustered with unclassified Caudoviricetes within the class Caudoviricete ( Figure 5 ). According to the genome alignment result, vB_KpnA_SCNJ1-Z shared the highest similarity (91% coverage and 93.07% identity) with Klebsiella virus KpV2883, vB_KpnS_SCNJ1-C shared the highest similarity (70% coverage and 94.03% identity) with Klebsiella phage vB_KpnS_MK54, and vB_KpnM_SCNJ1-Y shared the highest similarity (69% coverage and 91.24% identity) with Klebsiella phage vB_KpnM_JustaPhage. The main species demarcation criterion for bacterial and archaeal viruses is set at overall DNA sequence homology of 95% according to the International Committee on Taxonomy of Viruses (ICTV) (Adriaenssens et al., 2020). Thus, all three phages are designated novel species within the class Caudoviricetes. No virulence factors, lysogenic, integrase, or AMRs were detected in the genomes of the three phages, suggesting their potential as virulent phages for use in biological safety applications.

The optimal MOI of vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, and vB_KpnM_SCNJ1-Y was determined to be 0.0001, with the highest titers observed at 3.6×10⁹ PFU/mL, 4.3×10¹⁰ PFU/mL, and 8.2×10⁹ PFU/mL, respectively ( Supplementary Table S5 ). Analysis of their one-step growth curves revealed that vB_KpnS_SCNJ1-C and vB_KpnM_SCNJ1-Y had a shorter latent period of 7 min, while the latent period of vB_KpnA_SCNJ1-Z was 9 min. The burst size of vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, and vB_KpnM_SCNJ1-Y was determined to be 9.06 ± 0.81 PFU/cell, 6.25 ± 1.58 PFU/cell, and 13.26 ± 2.53 PFU/cell, respectively ( Figure 6A ).

For pH stability testing, both vB_KpnA_SCNJ1-Z and vB_KpnM_SCNJ1-Y exhibited a survival rate of 50%-100% within a wide pH range (pH 5-10), while they were inactivated at pH levels below 3 and above 13. vB_KpnS_SCNJ1-C exhibited a survival rate of 40%-60% within the pH range of 4-14 ( Figure 6B ). For heat resistance testing, approximately 50% of phage vB_KpnA_SCNJ1-Z remained viable at 60°C, while vB_KpnS_SCNJ1-C and vB_KpnM_SCNJ1-Y exhibited survival rates of only 5%-10%. At a higher temperature of 70°C, no viable phages were detected for any of the three phages ( Figure 6C ). Furthermore, incubation with 10% chloroform for 30 min resulted in survival rates of 87%, 78%, and 100% for vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, and vB_KpnM_SCNJ1-Y, respectively.

The host range of the phages was assessed against a panel of 50 clinical bacterial strains, as presented in Supplementary Table S1 . We found that the three phages did not lyse any other strains except the K. pneumoniae NJ09 which is a clone of the host bacterium K. pneumoniae SCNJ1.

To assess the ability of phages to impede biofilm formation by K. pneumoniae SCNJ1, different MOIs (100 - 0.1) of the mixture were incubated for 24 h using 96-well plates. The results showed that biofilm formation was considerably reduced at different MOIs in the phage-treated wells compared to the control group. Specifically, the absorbance value of the vB_KpnA_SCNJ1-Z-treated group (MOI = 100) ( Figure 7A ) was substantially reduced by 80% in comparison to the control group. Similarly, the absorbance value of the vB_KpnS_SCNJ1-C-treated group and vB_KpnM_SCNJ1-Y-treated group (MOI = 0.1) were significantly reduced by 90% and 89% ( Figures 7B, C ), respectively, in comparison to the control group. The results suggested that all three phages have the potential to inhibit the biofilm formation of SCNJ1.

To evaluate the efficiency of phage infection, K. pneumoniae SCNJ1 was infected with either a single phage or a phage cocktail consisting of vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, and vB_KpnM_SCNJ1-Y in equal proportions, with MOIs of 1, 10, and 100. The inhibitory effects of bacterial growth were assessed by comparing bacterial counts in the presence or absence of phage. All the single phage-treated groups showed a sharp decline in bacterial counts in the first hour at different MOIs, followed by regrowth of the resistant population during the remaining 5 hours of incubation, resulting in bacterial counts of 0.8×10⁴-1.5×10⁷ CFU/mL. However, after 6 hours of incubation, the bacterial counts of all single phage-treated groups were much lower than that of the control group which had bacterial counts of ~ 4×10⁸ CFU/mL. The cocktail-treated group showed a similar distribution of growth inhibition compared to that of every single phage ( Figure 8 ).

After 30 hours post-infection, the bacterial counts in the phage-treated groups showed a significant reduction in organs compared to those in the positive control group. For lung, the bacterial loads were approximately 10² CFU/0.1g, 10² CFU/0.1g, 10⁶ CFU/0.1g, and 10⁴ CFU/0.1g in the vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, vB_KpnM_SCNJ1-Y, and cocktail groups, respectively, whereas the bacterial load in the positive control group was approximately 10⁸ CFU/0.1g. For the liver, bacterial count was 10⁵ CFU/0.1g in the positive control group while those in the vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, vB_KpnM_SCNJ1-Y, and cocktail-treated groups was ~10² CFU/0.1g, ~10 CFU/0.1g, ~10 CFU/0.1g, ~10³ CFU/0.1g, respectively. Furthermore, compared to the bacterial load of the positive control group (10⁷ CFU/0.1g), all the phage-treated groups exhibited a substantial reduction, with the bacterial count reaching ≤ ~10² CFU/0.1g in the spleen ( Figure 9 ). vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, vB_KpnM_SCNJ1-Y, and cocktail-treated groups demonstrated a survival rate of 80%, 70%, 70%, and 80% within 7 days, respectively ( Figure 10 ). All mice in the positive control group succumbed to infection, whereas mice in the negative group and the safe test group survived. The histological examination of lung tissues of mice treated with vB_KpnA_SCNJ1-Z, vB_KpnS_SCNJ1-C, vB_KpnM_SCNJ1-Y, and cocktail showed significantly alleviated lesions with less inflammatory cell infiltration around the blood vessels ( Figures 11A–E ), in contrast, the positive control group showed thickened alveolar septum with infiltration of inflammation cells ( Figure 11F ). The lung tissue in the safe test group showed a small number of inflammatory cells in the alveolar ( Figure 11G ).