Strains used for prophage induction and phage isolation were listed in Supplementary Tables S1 and S2, respectively. Strain K88 came from a swine source. In addition, the 101 avian pathogenic E. coli strains used in these experiments were isolated from the brains of ducks with clinical signs of septicemia and neurological symptoms at different times and in different areas in the east of China, as previously described (Ma et al., 2013) (Supplementary Table S2). All E. coli strains were grown in Luria-Bertani (LB) medium at 37°C. When necessary, the LB medium was supplemented with appropriate antibiotic: ampicillin (100 μg/mL), kanamycin (50 μg/mL), chloramphenicol (30 μg/mL), or tetracycline hydrochloride (5 μg/mL). The plasmids used in this study were listed in Supplementary Table S3. The Red- and I-SceI-expressing plasmid pWRG99 was a derivative of pKD46 (Datsenko and Wanner, 2000; Blank et al., 2011).

Bacteriophages were obtained following mitomycin C (Sigma, St. Louis, MO, USA) induction of 54 lysogenic strains (Supplementary Table S1). Each strain was cultured for 5 h in LB medium and then diluted 1:100 in 5 mL of fresh medium. When the optical density at 600 nm (OD600) of the cultures reached 0.2, mitomycin C was added to a final concentration of 500 ng/mL. The cultures were then incubated at 37°C for 10 h. The lysates were centrifuged for 10 min at 3,500 × g at 4°C, and the supernatants were collected for follow-up experiments. The double-layer agar plate method (Adams, 1959) was used to find the host of the induced bacteriophage from the 108 E. coli strains (Supplementary Table S2).

The phage filtrate was applied to a copper grid before negative staining with phosphotungstic acid (PTA, 2% w/v). Electron micrographs were observed using an H_7650 transmission electron microscope (TEM; Hitachi, Japan).

Phage DNA extraction was performed as previously described (Ma and Lu, 2008) with some modifications. The phage pellet was digested with 2 μg/mL DNase and 10 μg/mL RNase for 30 min at 37°C. Phages were purified with NaCl-polyethylene glycol (PEG) 8000, and the DNA was isolated using SDS-Proteinase K. To analyze the quality of the DNA fragments, the molecules were separated by 0.8 % (w/v) agarose gel electrophoresis in TAE buffer (40 mM Tris-HCl, 500 mM sodium acetate, 50 mM EDTA, pH 7.2). The DNA was then suspended in ultrapure water.

The lysogenic bacteria K88, host bacteria DE048, and phage P88 were analyzed by high-throughput sequencing to confirm if the phage P88 genome is located in the K88 genome. The phage P88 and pro147 genomes were sequenced by Roche 454 Sequencing, and reads were assembled into contigs using the Newbler (version 2.8). Strain K88 and the DE048 genomes were sequenced by Illumina MiSeq Sequencing. Reads were assembled into contigs using the Newbler.

RAST (Aziz et al., 2008; Overbeek et al., 2014; Brettin et al., 2015) was used to predict the open reading frames (ORFs) of sequences and to analyze their corresponding functions. BLASTp was used to analyze the putative ORFs against the NCBI non-redundant proteins (NR) database. The phylogenetic trees of several selected genes were constructed with MEGA5.2 (Tamura et al., 2011) using the Neighbor-Joining algorithm. The structural model was generated using the Phyre¹ server¹ (Kelley et al., 2015).

Nucleic acid sequences of 10 P2-like phages [P88, pro147, P2 (NC_001895.1), Wphi, PsP3, 186, L-413C, Fels-2, fiAA91-ss, and P2 (KC618326.1)] were downloaded from GenBank. Comparisons of nucleic acid sequences were carried out with BLASTn. Comparisons of nucleic acid sequences of P88 (or pro147) with other P2-like phages were performed. Comparisons of nucleic acid sequences of P88 with pro147 were also performed. Phylogenetic trees based on the amino acid sequences (the lysins, the replicase, the tail fiber protein and the ATPase subunit of the terminase) were constructed with MEGA to analyze their evolutionary relationships.

Amino acid sequences of the tail fiber proteins of 10 P2-like phages [P88, pro147, P2 (NC_001895.1), Wphi, PsP3, 186, L-413C, Fels-2, fiAA91-ss, and P2 (KC618326.1)] were downloaded from the NCBI database and comparisons of amino acid sequences were carried out with ClustalX (version 2.1). The phylogenetic tree of the tail fiber proteins showed that P88 had a close evolutionary relationship with pro147, P2 (NC_001895.1), and Wphi. The amino acid sequences of the tail fiber proteins of P88, pro147, P2 (NC_001895.1), and Wphi were then compared to analyze which segments of encoding DNA could be selected for gene manipulation. The amino acid sequences of the tail fiber proteins of P88 and pro147 were compared to select the replacement sections.

The recombineering method of scarless mutagenesis was carried out according to the protocol described previously (Blank et al., 2011) with some modifications. The recombineering method of scarless mutagenesis was divided into two steps. In the first step, the target gene was replaced with a chloramphenicol resistance cassette (cat) gene and the I-SceI recognition site, which was PCR-amplified from plasmid pKD3 (Supplementary Table S3). The primers (Supplementary Table S3) were homologous to the flanking region of the corresponding target gene. The PCR products were then transformed by electroporation into K88 containing the lambda Red recombinase expression plasmid pKD46 (Datsenko and Wanner, 2000; Wang et al., 2011). After electroporation, samples were incubated at 30°C for 90 min in SOC broth (Datsenko and Wanner, 2000) and plated on LB agar with chloramphenicol to select for target gene mutants. Mutants were confirmed by PCR and sequencing using primers k1 and k2 (c1 and c2) (Supplementary Table S3). In the second step, the PCR products containing regions homologous to the flanking regions of the resistance gene for homologous recombination were transformed by electroporation into K88 containing the lambda Red recombinase expression and I-SceI endonuclease expression plasmid pWRG99 (Supplementary Table S3). After electroporation, samples were incubated at 30°C for 90 min in SOC broth and plated on LB agar with tetracycline hydrochloride. Selection of successful recombinants was mediated by the sequential expression of I-SceI endonuclease after Red-mediated recombination. If recombination occurred, the 200–500 bp dsDNA fragment harboring the same homologous region would replace the corresponding sequence of K88. If no recombination occurred, the unique I-SceI restriction site within the genome would still be present and the I-SceI endonuclease expressed from plasmid pWRG99 would kill unsuccessful recombinants by induction of double-strand breaks. Finally, pWRG99 was cured by streaking colonies and incubated at 37°C. Mutants were confirmed by PCR and sequencing using primers check1 and check2 (Supplementary Table S3).

Each K88 mutant strain was cultured for 5 h in LB medium and then diluted 1:100 in 5 mL fresh medium. When the culture reached an OD600 of 0.2, mitomycin C was added and then was incubated at 30°C for 10 h. The lysates were centrifuged and used to test their lytic capacity by the double-layer agar plate method with DE048, the host of P88.

Phage P88 could not infect the bacteria if the phage suspension was dropped onto the LB agar plate spread with bacteria, so the host range analysis of each P88 mutant strain constructed by the scarless mutagenesis method was carried out by the double-layer agar plate method with DE048, MC1061, DH5a and BL21.

Phage induction of 54 E. coli strains was investigated by the addition of mitomycin C. After induction, cell lysates were prepared and tested for their ability to lyse various E. coli hosts in plaque assays using the 108 strains listed in Supplementary Table S2. Out of the 54 E. coli strains used, the lysates of E. coli strain K88 and DE147 produced phages that could lyse the clinical isolates of avian pathogenic E. coli (APEC) strain DE048. These two phages were designated as P88 and pro147. After growth on DE048 in double agar LB plates at 37°C for 12 h, P88 formed clear and round plaques of approximately 0.06 cm in diameter (Figure 1A); pro147 formed clear and round plaques of approximately 0.13 cm in diameter (Figure 1B). Among the 108 E. coli strains tested, P88 specifically lysed DE048, while pro147 specifically lysed DE048, DH5a, BL21, and MC1061.

After incubation for 3.5 h, P88 had the highest concentration (1.1 × 10¹⁰) at the ratio of 0.01 (pfu/cfu); therefore, the ratio of 0.01 (pfu/cfu) was considered as the optimal MOI. The one-step growth curve obtained for P88 showed a latent period of about 20 min, a rise period of 25 min, and an average burst size of about 65 (Figure 1C).

Electron microscopy showed that the heads of P88 and pro147 were icosahedrons, and the tail was long and could shrink. In P88, the diameter of the capsid head was estimated at 55 nm and the tail length was estimated at 137 nm (Figure 1D). For pro147, the diameter of the capsid head was estimated at 54 nm and the tail length was estimated at 131 nm (Figure 1E). Phage P88 and pro147 belong to Myoviridae phages.

The P88 genome exhibited low similarity with pro147, covering 6% (ident 92%); however, the tail fiber protein sequence of P88 shared high similarity with that of pro147, covering 100% (ident 70%). In addition, the P88 tail fiber protein shared high similarity with that of P2 (NC_001895.1), covering 100% (ident 65%) and Wphi, covering 100% (ident 62%). Alignment of the P88 tail fiber protein with those of pro147, P2 and Wphi showed that there were two hypervariable regions of 178aa (517–695 in P88) and 31 aa (716–746 in P88) in the C-terminal region (Figure 5A). Alignment of the P88 tail fiber protein with that of pro147 showed that there were two segments of hypervariable regions of 120aa (576–695 in P88) and 31 aa (716–746 in P88) in the C-terminal region (Figure 7A).

Structure prediction and homology modeling of the tail fiber proteins of P88 and pro147 using the Phyre2 server² showed that the C-terminal domain (amino acids 426–612) of the P88 tail fiber (Figure 5B2) had the same folding topology as that of the domain (amino acids 937–1127) of T4 phage gp34 (Figure 5B1), with confidence of 97.17%; the C-terminal domain (amino acids 423–546) of the pro147 tail fiber (Figure 5B4) had the same folding topology as that of the domain (amino acids 984–1108) of T4 phage gp34 (Figure 5B3), with confidence of 97.17%. Despite the high degree of structural similarity, the tail fiber domains of P88 (or pro147) and T4 phage gp34 shared low amino acid identity over the relevant sequences with identity of 18 and 21%, respectively (Figure 5B).

Construction of gene mutants was carried out by modifying the phage genome in the genome of lysogenic bacteria and then inducing the production of the prophage (Figure 2B). The scarless mutagenesis method (Figure 6) was used to produce mutants of P88. We modified the tail fiber protein of P88 by replacing different parts of it with the corresponding parts of pro147. After tail fiber comparison of P88 and pro147, hypervariable regions of amino acids 576–695 (HR1) and 716–746 (HR2) (Figure 7A) in the C-terminus of the P88 tail fiber protein were replaced with the corresponding parts of pro147. After replacement of HR2, the phage mutant P88HR2 displayed a different host range to that of P88; therefore the 12 different amino acids in the region of HR2 [amino acid sets SA (716), SB (718), SC (719), SD (721), SE (722), SF (725), SG (729), SH (730), SI (734), SJ (736), FC (744), and FB (746)] (Figure 7B) and amino acid set FA (757) (Figure 7B) of P88 were replaced with the corresponding amino acids of pro147. Finally, nine recombinant bacteria of K88HR1 (576–695), K88HR2 (716–746), K88SD (721), K88SE (722), K88SH (730), K88SI (734), K88FA (757), K88FB (746) and K88FC (744) were constructed (Table 2).

The mutant bacteria were then induced by mitomycin C to evaluate their lysis capacity. Nine mutants of P88 were investigated. Eight mutants of K88HR1(576–695), K88HR2 (716–746), K88SD (721), K88SE (722), K88SH (730), K88SI (734), K88FA (757), and K88FB (746) could be induced to produce phages (P88HR1, P88HR2, P88SD, P88SE, P88SH, P88SI, P88FA, and P88FB) those were able to infect DE048 (Table 2; Figure 8), the host of P88; however, the bacteria mutant K88FC (744: glutamic acid to proline) could not be induced to produce phage that infected DE048 (Table 2). Meanwhile, K88FC could not be induced to produce phage to infect MC1061, DH5a, and BL21. The phage particles induced from the mutants that could lyse the E. coli strain DE048 were enriched for the subsequent experiment of host range determination and electron microscopy.

The host ranges of P88, pro147 and P88 mutants (P88HR1, P88HR2, P88SD, P88SE, P88SH, P88SI, P88FA, and P88FB) were analyzed. After incubation at 37°C for 10 h, on the lawn of DE048, all the phages could form clear plaques; on the lawn of MC1061 and DH5a, phages pro147, P88HR2 and P88SH (730: leucine to isoleucine) could form clear plaques; on the lawn of BL21, phages pro147 and P88HR2 could form clear plaques; pro147 and P88HR2 had the same host ranges (Figure 8). P88, pro147 and all the P88 mutants could not infect DE048 when incubated at 28°C.

Electron micrographs showed that compared with those of the original P88, the head and the tail of the phage mutants with altered host ranges (P88HR2 and P88SH) and with no change in host range (P88HR1, P88SE, and P88FB) did not have visible differences (Figure 8).