A total of 80 clinical isolates of K. pneumoniae were collected from different inpatients between January 2018 and December 2018 at Aerospace Center Hospital, Beijing, China. The presumptive identification at the species level of K. pneumoniae strains was verified by rpoB gene sequencing (Drancourt et al., 2001; Rosenblueth et al., 2004; Saha et al., 2013). Of these strains, K. pneumoniae 94 and 256 showing hypermucoviscous phenotype were used to the characterization of Dpo42 and Dpo43. Antimicrobial susceptibility of these K. pneumoniae was determined using a Vitek 2.0 compact system (BioMerieux Clinical Diagnostics, Paris, France) to the following antimicrobial agents: ampicillin-sulbactam, piperacillin-tazobactam, cefazolin, cefuroxime, ceftriaxone, ceftazidime, cefepime, cefotetan, aztreonam, imipenem, meropenem, gentamicin, tobramycin, amikacin, trimethoprim-sulfamethoxazole, and ciprofloxacin. The minimum inhibitory concentrations (MICs) were determined by reference broth microdilution method and results were interpreted using the Clinical Laboratory Standard Institute breakpoints (Clinical and Laboratory Standards Institute [CLSI], 2018). Pseudomonas aeruginosa ATCC 27853 used as a quality control strain. In the present study, no human or animal subjects were involved, and therefore no ethics approval is required.

MLST of K. pneumoniae was performed according to Protocol 2 described on the MLST website¹. The seven housekeeping genes (gapA, infB, mdh, pgi, phoE, rpoB, and tonB) of K. pneumoniae were amplified by PCR with primers in Protocol 2. Sequence types (STs) were determined by using the K. pneumoniae MLST database of Institut Pasteur². To identify the capsular types of K. pneumoniae strains, the wzi allele was amplified, and the obtained sequences were aligned to the wzi sequences deposited in the database of Institut Pasteur (Brisse et al., 2013)².

The IME205 phage was isolated from a raw sewage sample collected at The Fifth Medical Centre of Chinese PLA General Hospital by using the hypermucoviscous K. pneumoniae 256 as host strain. Briefly, 4 mL of filtered sewage and 500 μL of bacteria culture in the exponential growth phase were added to 2 mL of 3 × Luria Bertani (LB) medium. After 4 h incubation at 37°C with shaking at 200 rpm, the culture was centrifuged and the supernatant was collected for double-layer agar plate assay to detect the presence of phage plaques (Kropinski et al., 2009).

The phage nucleic acids were extracted using the High Pure Viral RNA Kit (Rocha Diagnostics, Mannheim, Germany) and sequenced using the Ion Personal Genome Machine System (Thermo Fisher Scientific, Waltham, MA, United States). The obtained reads were assembled into a complete genome sequence by the Newbler 2.9.1 software (Rocha Diagnostics, Bloomington, IN, United States). Open reading frames (ORFs) were annotated by using the Rapid Annotation in Subsystem Technology (RAST) web-server³, and the putative function of coding sequences (CDSs) was predicted by NCBI BLASTP.

ORF42 (GenBank accession number ALT58497) and ORF43 (GenBank accession number ALT58498) were amplified by PCR with primers Dpo42p (5′-CAAATGGGTCGCGGATCCATGGACCAAGACATTAAAAC AG-3′ and 5′-GTGGTGGTGGTGCTCGAGTTACTGTTCGCC CCACTGCA-3′) and Dpo43p (5′-CAAATGGGTCGCGGATC ATGTTAAACAACCTGAACCAG-3′ and 5′-GTGGTGGTGG TGCTCGAGTTATGGACCAATAACCACACC-3′), respectively. The obtained PCR products were cloned into the pET28a vectors with restriction sites BamHI and XhoI (New England Biolabs, Ipswich, MA, United States) using the pEASY^® -Uni Seamless Cloning and Assembly Kit (TransGen Biotech, Beijing, China). The sequence of each insert in recombinant plasmids were checked by DNA sequencing before transformation into E. coli BL21(DE3). Cells carrying recombinant plasmids were cultured to the exponential phase and induced with 1 mM isopropyl β-D-1-thiogalactopyranoside (IPTG; Sigma-Aldrich, St. Louis, MO, United States) at 25°C overnight. After 5 min centrifugation at 10,000 rpm, the pellets were resuspended in lysis buffer (50 mM NaH₂PO₄, 300 mM NaCl, pH 8.0) and sonicated on ice (20 min with 3 s pulse and 4 s pause). Bacterial lysates were centrifuged and passed through a 0.45 μm filter. Then, the filtrates were loaded onto a gravity column with Ni-NTA resins (Sangon Biotech, Shanghai, China). The target protein was eluted with imidazole-containing buffer (50 mM NaH₂PO₄, 300 mM NaCl, 100 mM imidazole, pH 8.0) and dialyzed against a 1000-fold volume of lysis buffer over a 8–14 kDa molecular-mass-cutoff membrane (Viskase, Lombard, IL, United States). Protein molecular weight was estimated by 10% sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE), followed by Coomassie blue staining. The protein concentration was quantified with a fluorometer (Qubit 2.0, Thermo Fisher Scientific).

The Dpo42 or Dpo43 activity was qualitatively examined by the single-spot assay. Briefly, 500 μL of K. pneumoniae 94 or 256 culture in the exponential growth phase was mixed with 4.5 mL of molten soft nutrient agar to form a bacterial lawn on the plates. Then, 4 μL of diluted enzyme was dropped onto the plate and incubated overnight at 37°C. The enzymatic activity was monitored for the formation of halo zones.

The host range of phage IME205 was evaluated using the double-layer agar plate assay. Plates containing a mixture of IME205 and of each Klebsiella K47 strain were cultured overnight at 37°C. Phage-forming units (PFU) on bacterial lawns were counted. The experiment was independently conducted three times. Relative efficiency of plating (EOP) was calculated as the average PFU of the phage on the each tested bacteria divided by the average PFU on the host bacteria (K. pneumoniae 256) (Khan Mirzaei and Nilsson, 2015). The sensitivity of Klebsiella K47 strains to Dpo42 or Dpo43 (20 ng) was determined by the single-spot assay described above.

The extraction and purification of bacterial exopolysaccharides (EPS) containing both capsular polysaccharides and liposaccharides were performed via a modified hot water-phenol method as described previously (Hsieh et al., 2017). Briefly, 1 mL of K. pneumoniae 94 or 256 cultured overnight in LB with 0.25% glucose was centrifuged (10,000 rpm, 5 min) and resuspended in 200 μL of double distilled water (ddH₂O). An equal volume of water-saturated phenol (pH 6.6; Thermo Fisher Scientific) was added to the bacterial suspension. The mixture was vortexed and incubated at 65°C for 20 min, then extracted with chloroform to remove bacterial debris. The obtained EPS was lyophilized and stored at −20°C.

The enzymatic activity of Dpo42 or Dpo43 on bacterial surface polysaccharides was determined as described in Majkowska-Skrobek et al. (2016) with minor modifications. The EPS powder of K. pneumoniae 94 (or 256) was resuspended in ddH₂O (2 mg/mL), and mixed with Dpo42 or Dpo43 (10 μg/mL) to a final reaction volume of 1.0 mL. EPS or enzyme alone served as positive or negative controls, respectively. After 1 h incubation at 37°C, cetylpyridinium chloride (CPC, Sigma-Aldrich) was added to the mixture (final concentration of 5 mg/mL) which was further incubated at room temperature (RT) for 5 min. Absorbance was measured at 600 nm using a Synergy HT Multi-Detection Microplate Reader (BioTek, VT, United States). The experiment was repeated at least three times.

The CPS was detected by Alcian blue staining as previously described (Pan et al., 2013; Hsieh et al., 2017). The treated samples were separated by a 10% SDS-PAGE. The gel was then washed for three different time duration (5, 10, and 15 min) with the fix/wash solution (25% ethanol, 10% acetic acid in water) and stained by 0.1% Alcian blue (Sigma-Aldrich) dissolved in the fix/wash solution for 15 min in the dark. CPS was visualized blue after the gel was destained overnight in the fix/wash solution.

The EPS powder was dissolved in 100 mM citric acid-Na₂HPO₄ buffer (pH 3.0–8.0) or 100 mM Glycine-NaOH buffer (pH 9.0–10.0) to a final concentration of 2 mg/mL. The EPS solutions of K. pneumoniae 94 and 256 were mixed with Dpo42 and Dpo43 (10 μg/mL each) to a final reaction volume of 1.0 mL, respectively. After 1 h incubation at 37°C, the turbidity of residual EPS in various pH buffers was determined as described above. The effect of temperatures on the enzymatic activity was also determined. The EPS dissolved in citric acid-Na₂HPO₄ buffer (pH 5.0) was incubated with the depolymerase (10 μg/mL) at 20, 37, 50, 60, 70, 80, and 90°C for 1 h. EPS alone in citric acid-Na₂HPO₄ buffer (pH 5.0) at RT was served as the control. The enzymatic activity was determined as described above. The experiment was performed in triplicate.

The serum sensitivity assay was conducted as described previously with slight modifications (Pan et al., 2015; Lin et al., 2017). Briefly, the human blood sample from healthy donor was centrifuged at 1,000 rpm for 10 min to obtain serum and erythrocytes. K. pneumoniae 94 and 256 were used to test the serum sensitivity with Dpo42 and Dpo43, respectively. Overnight bacteria cultures were centrifuged, resuspended in PBS to give a bacteria concentration of 10⁹ cfu/mL, and treated with the depolymerase at an enzyme concentration of 10 μg/mL. Then, the treated bacteria were immediately mixed with active or inactive serum (heated at 56°C for 30 min) at a volume ratio of 1:3 to a final reaction volume of 100 μL for 1 h incubation at 37°C. The untreated bacteria were incubated with enzyme or active serum as controls. The reaction mixtures were serially diluted and plated for bacterial counting. The experiment was independently performed three times.

The hemolytic effect of depolymerases to erythrocytes was performed using previously described methods with minor modifications (Wang et al., 2015). The separated erythrocytes were washed and diluted to a concentration of 5% (v/v) with PBS. The erythrocytes were incubated with Dpo42 or Dpo43 (10 μg/mL) at 37°C for 1 h with gentle shaking at 60 rpm. The erythrocytes treated with PBS and 0.1% Triton X-100 were included as negative and positive control, respectively. After 10 min centrifugation at 1,000 rpm, 100 μL of supernatant was transferred to a 96-well microplate, and another 100 μL of PBS was added to the wells to a final volume of 200 μL. The amount of hemoglobin was measured at 540 nm. The experiment was conducted in triplicate.

All experimental data are presented as means ± standard deviation (SD), and the one-way analysis of variance (ANOVA) was used to compare multiple groups. The statistical analyses were conducted using Prism 7 (GraphPad Software, CA, United States), with P < 0.05 considered statistically significant.

During the study period, a total of 80 K. pneumoniae strains were identified to exhibit a carbapenem-resistant phenotype. The susceptibility rates of these bacteria toward different classes of antibiotics are given in Supplementary Table S1. These isolates were collected from sputum (66.25%, 53/80), urine (22.50%, 18/80), blood (6.25%, 5/80), wound secretion (2.50%, 2/80), anal swab (1.25%, 1/80) and central venous catheter (1.25%, 1/80). All strains belonged to ST11 and corresponded to four capsular types (Table 1), including K16 (1.25%, 1/80), K37 (2.50%, 2/80), K47 (70.00%, 56/80), and K64 (26.25%, 21/80). A lytic phage, IME205, was isolated using the K. pneumoniae 256 (K47) as the host bacterium. The phage formed clear plaques surrounded by translucent halos on the double-layer agar plate after overnight incubation at 37°C (Figure 1A). This observation suggested the phage IME205 produced depolymerases that could degrade bacterial surface polysaccharides.

The obtained genome sequence of IME205 was deposited in GenBank under accession number KU183006. The complete genome of phage was 41,310 bp, with GC content of 52.2%. Online BLAST analysis showed that the complete genomic sequence of phage IME205 shares high similarity (90.67–98.39%) with that of 40 Klebsiella phages, all of which belong to Podoviridae. Whole-genomic comparison indicated that these phage genomes have similar structures, but the regions encoding phage tail fiber proteins are significantly different. The annotation results of RAST showed that the phage IME205 included 49 ORFs, and 30 of which were predicted to have specific functions by NCBI BLASTP (Supplementary Table S2). According to Latka et al. (2019), phage IME205 belongs to Group A of KP32viruses containing two receptor binding proteins (RBPs). A first gene encoding the phage tail fiber protein (ORF42) with a phage T7 tail fiber protein domain (residues 1–154) in the N-terminus (anchor-branch domain) and a pectate_lyase_3 protein with a predicted beta-helix structure (residues 300–502) in its central domain, and a second down-stream gene coding for a hypothetical protein (ORF43) with 83% identity to the second RBP of Klebsiella phage KP32 (GenBank accession number YP_003347556.1) are predicted to have polysaccharide depolymerase activities (Latka et al., 2019). The predicted length of ORF42 and OFR43 was 793 and 641 aa, respectively. The ORF42 and ORF43 were closest homologous to a tail fiber protein of Klebsiella phage IME304 (GenBank accession number QDB73379.1; identity = 99.50%, query cover = 100%) and a hypothetical protein of Klebsiella phage SH-KP152226 (GenBank accession number QDF14645.1; identity = 99.22%, query cover = 100%), respectively. Unfortunately, the capsular specificities of these two homologous proteins were not reported. To determine whether these CDSs had depolymerase activities, the PCR products of ORF42 and ORF43 were cloned into the pET28a vector, respectively. Then, proteins with a hexahistidine tag were expressed and purified. As depicted in Figure 1B, Dpo42 and Dpo43 migrated as single bands on 10% SDS-PAGE corresponding to their predicted molecular weight of 85 and 68 kDa, respectively. The concentration of purified Dpo42 and Dpo43 was determined to be 0.5 mg/mL and 0.4 mg/mL, respectively.

The depolymerase activity was determined using the modified single-spot assay, with enzyme dilutions ranging from 1 to 100 ng. As presented in Figure 1C, both Dpo42 and Dpo43 could form semi-clear halos on its host bacteria lawns. The size and transparency of semi-clear circles decreased with decreasing enzyme concentration, with the halo disappeared at a protein level of 1 ng.

The lytic spectrum of phage IME205 was determined among 56 strains of Klebsiella capsular type K47, including the host bacterium K. pneumoniae 256 (Table 2). Phage IME205 could produce plaques on all tested K. pneumoniae strains, with the EOP was greater than 0.1. Moreover, the sensitivity of these strains to Dpo42 or Dpo43 was also tested (Table 2). The single-spot assay showed a total of 16 K. pneumoniae strains (16/56) were sensitive to Dpo42, whereas the rest 40 strains (40/56) were sensitive to Dpo43.

The depolymerase activity was evaluated by monitoring the turbidity of the residual EPS. The EPS was extracted from K. pneumoniae 94 or 256, which belonged to different subsets of capsular type K47. As shown in Figures 2A,C, the OD₆₀₀ values of EPS extracted from K. pneumoniae 94 were 1.156 ± 0.024, 1.154 ± 0.048, and 0.677 ± 0.033, after incubation with PBS, Dpo43 and Dpo42, respectively. For EPS extracted from K. pneumoniae 256, the OD₆₀₀ values after treating with PBS, Dpo42 and Dpo43 were 1.133 ± 0.016, 1.167 ± 0.052, and 0.434 ± 0.038, respectively. The turbidity of enzyme alone was very low (0.058 ± 0.003 or 0.064 ± 0.002). This result revealed the turbidity of EPS treated with its specific enzyme was significantly decreased as compared with EPS alone or EPS incubated with its non-specific enzyme (P < 0.0001, one-way ANOVA). The capsule degrading properties of Dpo42 and Dpo43 were further verified by Alcian blue staining (Figures 2B,D). Results of the SDS-PAGE gel stained with Alcian blue showed that the CPS of K. pneumoniae 94 or 256 was apparently degraded by Dpo42 or Dpo43, respectively. The weight of EPS alone was almost the same as the EPS mixed with its non-specific enzyme, and the enzyme alone was not stained by Alcian blue.

The enzymatic activity of Dpo42 or Dpo43 at pH 2 to 10 was determined (Figures 3A,C). The results showed that Dpo42 was active at pH 5.0 to 8.0, whereas the Dpo43 remained active at pH 4.0 to 8.0. The depolymerase activities of Dpo42 and Dpo43 were completely lost at pH outside their active ranges. The enzyme activities at temperatures ranging from 20 to 90°C in citric acid-Na₂HPO₄ buffer (pH 5.0) were determined (Figures 3B,D). Both Dpo42 and Dpo43 could maintain enzymatic activity at temperatures ranging from 20 to 70°C. The enzyme was completely inactive at 80°C.

To confirm Dpo42 or Dpo43 could sensitize K. pneumoniae to serum, the viable counts of bacteria in various treatment conditions were determined. The K. pneumoniae 94 and 256 were sensitize to Dpo42 and Dpo43, respectively. Figures 4A,B showed that the number of bacteria treated with its specific enzyme followed with serum was effectively decreased (≥3 log reduction) as compared with other groups (P < 0.0001, one-way ANOVA). While the serum alone slightly reduced the bacterial counts, the enzyme alone and the enzyme mixed with inactive serum did not affect the bacterial survival. Thus, both Dpo42 and Dpo43 could make bacteria more susceptible to the serum complement-mediated killing. To assess the safety of applying Dpo42 and Dpo43 as antimicrobial agents, the hemolysis of enzyme toward erythrocytes was assessed in vitro (Figure 4C). The Dpo42 and Dpo43 displayed no hemolytic activity to erythrocytes.