A complete list of strains and plasmids used in this study and construction details can be found in Table 1. Lysogeny broth (Lennox formulation; LB) contained 10 g tryptone, 5 g yeast extract and 5 g NaCl per liter (pH 7.0). All strains were cultured in LB broth for 16 h at 37°C with shaking at 220 rpm, unless otherwise stated. For chemical phage induction, mitomycin C (Sigma-Aldrich M4287) suspended in water was sterilized using a 0.22 μm filter and added to exponentially growing cultures (OD₆₀₀ = 0.3) at a final concentration of 2 μg/ml. When required, LB was supplemented with the following antibiotics: ampicillin (Ap), 100 μg/ml; chloramphenicol (Cm), 25 μg/ml; kanamycin (Km), 50 μg/ml; gentamicin (Gm), 20 μg/ml; oxytetracycline (OxyTc), 10 μg/ml; tetracycline (Tc), 20 μg/ml.

Phage enumeration and plaque isolation were carried out using the double layer agar technique. One milliliter of bacterial culture containing phage was filtered through a 0.22 μm filter to remove bacterial cells, and diluted appropriately in sterile LB. Dilutions were applied in 10 μl drops to 4 ml 0.4% LB agar lawns seeded with 100 μl indicator strain (∼10⁸ CFU) and incubated overnight at 37°C.

All oligonucleotide primers used in this study are listed in Supplementary Table S1. For cloning and recombineering procedures, PCR reactions were performed with Q5 High-Fidelity DNA polymerase (New England Biolabs M0491S) according to the manufacturer’s instructions. The PCR fragments obtained from the constructed strains and the DNA inserts cloned into plasmids were sequenced by GATC Biotech (Konstanz, Germany). Recombineering procedures were carried out using the λ Red recombination plasmids pSIM5-tet and pKD46 in D23580 and 14028s/LT2 backgrounds, respectively (Datsenko and Wanner, 2000; Koskiniemi et al., 2011). Electro-competent cells of Salmonella carrying pSIM5-tet or pKD46 were prepared and transformed with the indicated PCR fragment as previously reported (Yu et al., 2003; Datta et al., 2006; Lemire et al., 2011) and recombinants were selected on LB agar plates supplemented with antibiotic. Generalized transduction was performed with the highly efficient transducing phage P22 HT 105/1 int-201 (Schmieger, 1972). As D23580 is highly resistant to P22 infection (Kintz et al., 2015), high numbers of P22 transducing particles were necessary to obtain transductants. Therefore, 200 μl of stationary phase culture of the recipient strains were mixed with 100 μl of the P22 lysates. After 1 h of incubation at 37°C, transductants were selected on LB-Km agar plates. Transductants were grown to stationary phase in LB containing 10 mM EGTA (Sigma-Aldrich, E3889) and Km. Cultures were streaked on indicator Green Plates and P22-free colonies were selected for further experiments (Maloy, 1990).

The prophages of S. Typhimurium D23580 were cured according to the scarless genome editing technique developed in the de Lorenzo Lab (Martínez-García and de Lorenzo, 2011). The phage attachment site (attB) of Gifsy-1^D23580, Gifsy-2^D23580, BTP1, BTP5, and ST64B^D23580 were PCR amplified from S. Typhimurium prophage-cured strains or naturally naive strains with the appropriate primers and cloned into the suicide plasmid pEMG, as summarized in Supplementary Table S2. Escherichia coli S17-1 λpir was used as host strain for the resulting suicide plasmids pNAW15 (attB^Gifsy-1), pNAW16 (attB^Gifsy-2), pNAW17 (attB^BTP1), pNAW18 (attB^BTP5) and pNAW42 (attB^ST64B). The suicide plasmids were mobilized from S17-1 λpir into the recipient S. Typhimurium strains by conjugation. The donor S17-1 λpir strains harboring the pEMG derivatives and the Salmonella recipient strains were grown overnight in LB-Km and LB, respectively. Cells were washed once with LB, 500 μl of both cell suspensions were mixed and bacteria were concentrated in 40 μl of LB. The mixtures were dropped on LB agar plates and incubated for mating for 6 h at 37°C. Bacteria were suspended in LB and Salmonella transconjugants having integrated the pEMG derivative plasmids by homologous recombination were selected on LB agar containing Km and Cm. The inserted pEMG encompasses two I-SceI restriction sites that are used as counter-selective markers in the presence of the I-SceI restriction enzyme. Therefore, transconjugants were transformed with the pSW-2 plasmid and the transformants were selected on LB agar containing Gm and 1 mM m-toluate (prepared with m-toluic acid, Sigma-Aldrich T36609 and NaOH to pH 7.0), stimulating the production of the I-SceI enzyme from pSW-2 (Martínez-García and de Lorenzo, 2011). To optimize the transformation yield, pSW-2 was purified from D23580/pSW-2. The transformants having lost the integrated pEMG by a second homologous recombination were identified by their Km sensitive phenotype (Km^S). The excision of the prophages was assessed in Km^S clones by PCR using the primers used to amplify the attB of the corresponding prophages (Supplementary Table S2). Finally, the unstable pSW-2 was eliminated from the clones of interest by two passages in LB.

The galE496 allele (premature stop codon TAG on the 326^th codon of galE) of MA6684 was transferred into D23580 derivative strains by a two-steps procedure. An aph (Km^R) cassette was amplified from pKD4 with primers NW_86 and NW_87 and was inserted by λ Red recombination into the intergenic region STM0777-galE496 of MA6684. Due to the defective O-antigen of the galE mutants, the resulting strain was grown in LB supplemented with 0.2% of glucose and 0.02% of galactose for the preparation of a P22 lysate (Butela and Lawrence, 2012). The aph-galE496 module was subsequently transduced into D23580 derivative strains and transductants were screened by PCR with the primers NW_61 and NW_82 and checked by sequencing to confirm the presence of galE496. The functional promoter of the antirepressor gene (gfoA) in Gifsy-1 of S. Typhimurium 14028s was inserted in place of the corresponding promoter of D23580 (Lemire et al., 2011). Therefore, a single nucleotide substitution (C→T position 2790162) was introduced into the promoter controlling dinI(STMMW_26401)-gfoA(STMMW_26391) transcription (P_dinI-gfoA) of D23580 by a protocol based on the principle described by Blank et al. (2011). Due to the MDR phenotype of D23580, a plasmid harboring an I-SceI-aph module was first constructed: the I-SceI restriction site was inserted upstream of aph in pKD4, using the Site-Directed Mutagenesis method with primers NW_161 and NW_162 (Laible and Boonrod, 2009). The resulting plasmid (pKD4-I-SceI) was used as template to insert an I-SceI-aph module into P_dinI-gfoA by λ Red recombination, using primers NW_214 and NW_215. This module was transduced into a clean D23580 background and the resulting strain, D23580 I-SceI-aph-dinI-gfoA, was transformed with pSIM5-tet for a second λ Red recombination. The functional P_dinI-gfoA of S. Typhimurium 14028s was PCR-amplified using the primers NW_194 and NW_195 and the resulting fragment (319 bp) was electroporated into D23580 I-SceI-aph-dinI-gfoA carrying pSIM5-tet. To select the recombinants, new electro-competent cells were directly prepared after the electroporation, and bacteria were transformed with pSW-2. After selection on LB agar containing Gm and m-toluate, colonies were screened by PCR with primers NW_196 and NW_197 and the resulting PCR fragments were sequenced to confirm the mutation. Two passages in LB were used to eliminate pSW-2. To construct a RecA^- LT2 strain (LT2 derivative MA8508 was used, Table 1), the aph cassette was first amplified from pKD13 with primers RecA_pKD13_Fw and RecA_pKD13_Rev, and then inserted by λ Red recombination into MA8508 pKD46. Kanamycin resistance was subsequently removed using the Flp recombinase encoded in the pCP20 plasmid. The same methodology was applied to D23580 to obtain the D23580 RecA^- strain, but using the pKD46 bla::Tn10 and pCP20-TcR plasmids instead, as ampicillin resistance in D23580 (Kingsley et al., 2009) prohibits the use of the original pKD46 and pCP20 plasmids.

Exponential phase cultures of susceptible host strains and BTP1 phage were mixed and plated using the double layer agar technique as described above. BTP1 containing plaques were identified using primers for the pid gene, pid_Fw, and pid_Rev (Supplementary Table S1). Plaques that were positive for pid were further picked and purified on Green indicator plates (Maloy, 1990). This medium contains glucose as a carbon source, and a pH indicator dye that turns dark green at sites where phage infection causes cell lysis and the concomitant release of organic acids, while the lysogens remain pale green. Candidate LT2 [BTP1] lysogens were picked onto fresh Green Plate and cross streaked against BTP1 phage to confirm resistance of the lysogens to superinfection by BTP1.

Viable phages were detected in the supernatant of D23580 WT after 16 h culture. Where chemical induction was required, mitomycin C (Sigma-Aldrich M4287) was added to exponentially growing cultures (OD₆₀₀ = 0.3) at a final concentration of 2 μg/ml. Phages were enumerated using the double layer agar technique as described above using specific indicator strains to detect each phage (Table 2). Prophage viability was defined as the ability to form infective phage particles with or without chemical induction.

Burst size determination was achieved using the single step growth curve protocol previously described (Hyman and Abedon, 2009). Briefly, 50 μl of exponentially growing D23580 ΔBTP1 cells (OD₆₀₀ = 0.3) at a density of 10⁸ CFU/ml were mixed with 50 μl BTP1 phage at a density of 10⁷ PFU/ml to achieve a multiplicity of infection (M.O.I.) of ∼0.1. After 5 min adsorption time at 37°C, 1 μl of the cell-phage mixture was diluted 1000X by addition of 999 μl sterile LB, in order to prevent further adsorption. Free (non-adsorbed) phages were enumerated by plaque assay on a susceptible indicator strain as described above. Number of infected cells was calculated as: [(number of cells x M.O.I.) – number of unabsorbed phage]. Subsequently, the number of free phage was enumerated by sampling and filtration every 10 min. Difference between the phage titer pre- and post-burst divided by the number of infected cells was determined as the burst size. Duration of time between infection and burst was determined as the latent period.

The percentage of the population undergoing spontaneous induction was estimated as [(100/average number of cells in 16 h culture)/(average number of phage in 16 h culture/burst size)].

A pure BTP1 phage stock was produced by picking and resuspending a BTP1 plaque in sterile LB and plating on indicator strain D23580 ΔΦ (JH3949). This process was repeated three times to ensure purity. A high titer BTP1 lysate (∼10¹¹) was produced by mixing 10 μl of BTP1 plaque suspension with 5 ml of exponential phase (OD₆₀₀ = 0.3) D23580 ΔΦ. BTP1 virions were concentrated by PEG precipitation. Briefly, 1 ml of PEG-8000 20% NaCl 2.5 M was added to 4 ml BTP1 lysate and incubated on ice for 2 h. Subsequently the mixture was centrifuged at 15,000 g for 1 h at 4°C and liquid decanted. The pellet was resuspended in 50 μl molecular grade water and treated with DNase I for 1 h at 37°C to remove any contaminating bacterial DNA. Subsequently the sample was heated to 90°C for 10 min to simultaneously deactivate DNase I and release BTP1 DNA from capsids. BTP1 DNA was purified using the Zymo Research Quick-DNA^TM Universal Kit (cat# D4069) as per the Biological Fluids and Cells protocol and eluted in 15 μl molecular grade water. Sequencing libraries were prepared using the Illumina Nextera XT DNA sample kit per the manufacturer’s protocol (Illumina, USA) with 1 ng of input DNA. Sequencing was performed on an Illumina MiSeq instrument using V2 chemistry (2 × 250 bp). Sequencing reads were quality trimmed using Sickle (Joshi and Fass, 2011) and assembled using Spades (Bankevich et al., 2012). The assembly was manually reordered to begin at the small terminase. Functional annotation was transferred from the D23580 (FN424405) prophage BTP1 sequence using prokka 1.11 (Seemann, 2014). Read data, assembly and functional annotation were deposited at the DDBJ/EMBL/GenBank database under the sample accession LT714109¹ and study accession PRJEB18919.

Prophage locations in the D23580 genome were extracted from the Genbank annotation (Accession number: FN424405) confirmed by submitting the D23580 genome sequence to PHAST (Zhou et al., 2011) and for BTP1 and BTP5, precise prophage boundaries were determined by analysis of the att sites. The genome of S. Typhimurium strain 4/74 (Accession number: CP002487) was used as a comparator to identify D23580-specific prophages. Genome comparison was done using BLASTn alignment with default parameters and visualized using the Artemis Comparison Tool (ACT; Carver et al., 2008). Prophage ORFs were characterized by uploading individual prophage sequences to the RAST server, and visualized using the SEED viewer in the context of the closest phage relatives in the RAST database (Aziz et al., 2008; Overbeek et al., 2014). Protein functions were inferred based on homologies to known proteins by the BLASTp and psiBLAST server of NCBI using the UniProt/Swiss-Prot database. Morphological classification was made based on similarity to other well-studied phages. Clustering of functional gene groups was based on a similar analysis by Zou et al. (2010). Assessment of the conservation of the functionally degradative mutations identified in Gifsy-1^D23580, ST64B^D23580, and Gifsy-1^D23580 across other ST313 isolates was achieved using BLASTn. Genome assemblies were downloaded from Enterobase².

To prepare samples for Transmission Electron Microscopy (TEM), 10 μl of the supernatant of a 16 h D23580 culture was pipetted onto a carbon/Pioloform-coated copper 200-mesh grid, and allowed to adhere for 10 min. To image phage infections, the supernatant of a 16 h D23580 culture containing virions was added to mid-exponential phase S. Typhimurium 4/74 (OD₆₀₀ = 0.3) cells in LB media, and incubated at 37°C for 20 min before TEM sample preparation. The grids were subsequently washed twice in distilled water for 2 min, negative stained with 2% uranyl acetate for 1 min and examined using a FEI 120 kV Tecnai G2 Spirit BioTWIN transmission electron microscope at the EM Unit, University of Liverpool.

Six prophages were originally annotated in the genome of ST313 representative strain D23580 (Kingsley et al., 2009). In fact, D23580 contains a total of nine phage-derived sequence islands (Figure 1A; Table 3). Four short prophage-remnants were named Def1-4, using the nomenclature defined for S. Typhimurium strain LT2 (Casjens, 2011). Def1-4 are found in all Typhimurium complete genomes currently available on NCBI (accessed 28th July 2016) and are likely to be ancestral to the serovar (Casjens, 2011). One feature designated ‘BTP6’ was included in the original annotation of D23580 (accession FN424405; Kingsley et al., 2009) and is located at the position of prophage remnant Def4. The remaining five prophage regions, designated BTP1-5, are full-length prophages (Table 3) that contain all the functional gene clusters required for phage function such as integration/excision, immunity, lysis, capsid and tail gene clusters. Three represent well-characterized prophages, commonly found in S. Typhimurium genomes: BTP2 is Gifsy-2^D23580, BTP3 is ST64B^D23580 and BTP4 is Gifsy-1^D23580 (Figure 1A) (Figueroa-Bossi et al., 2001; Tucker and Heuzenroeder, 2004). BTP1 and BTP5, however, are novel prophages that are only found in ST313 and not present in the 187 complete Salmonella genomes available in the NCBI complete genome database (accessed 28th July 2016). Together, the phage-derived sequence islands harbor many characterized virulence-related genes in S. Typhimurium including genes encoding effector proteins and O-antigen modification genes (gtr genes) (Table 3).

The gene content and synteny of BTP1 resembles P22, suggesting that BTP1 belongs to the Podoviridae family of short-tailed phages (Ackermann, 2009). Functional annotation of BTP1 genes identified a number of putative cargo loci on the prophage: independent transcriptional units that are not required for the prophage lytic or lysogenic cycles (Cumby et al., 2012). Like P22, BTP1 contains a gtrAC operon (STMMW_03911-03921) (Supplementary Figure S1A) which encodes a glycosyltransferase enzyme that modifies the O-antigen of the bacterial lipopolysaccharide (Kintz et al., 2015). This glycosyltransferase mediates the acetylation of the rhamnose residue of the O-antigen and is classified as a family 2 GtrC protein, which are typically associated with invasive Salmonella serovars such as Typhi and Paratyphi A (Kintz et al., 2015). Previously it has been shown that a BTP1 lysogen is highly immune to superinfection by P22, suggesting the two phages share a superinfection immunity system, likely mediated by the gtr glycosyltransferase. The st313-td gene (STMMW_03531) is encoded immediately downstream of the CI repressor locus, and its deletion decreases survival within murine macrophages, and results in attenuation in a mouse infection model (Herrero-Fresno et al., 2014). At the time of description of the st313-td gene, prophage BTP1 had not been characterized and therefore the fact that st313-td is BTP1-encoded has not been previously explored. Additionally, the pid ORFan locus (STMMW_03751) is present in BTP1. The P22-encoded Pid regulatory protein de-represses the host dgo operon in S. Typhimurium LT2 (Cenens et al., 2013). The complete functionally annotated map of BTP1 is shown in Supplementary Figure S1A. The chromosomal context of the BTP1 attB is shown in Supplementary Figure S2A, and the core att sequence for BTP1, is given in Supplementary Table S3.

A comparative genomic analysis assigned putative function to 43/64 (62.5%) of the annotated open reading frames (ORFs) of BTP1. A comparison with model phages P22 and Lambda revealed that BTP1 ORFs can be classified into 10 functional blocks (Figure 2A, full details of the sequences used in this analysis can be found in Supplementary Table S4). The functional blocks show mosaic conservation in other phage and prophage genomes suggesting functional overlap and recombination. Similarity to other phages extends beyond phages infecting the Salmonella genus, as BTP1 shares sequence identity with P22-like phages of E. coli and Shigella (CUS-3, HK620, and Sf6). Some BTP1 functional blocks are rarer among the sequences sampled in this analysis than others. For example, the BTP1 gtr block is only found in Salmonella phage 𝜀34. Others, such as the capsid and tail block, int/xis block and recombination block, are found in the majority of sampled sequences. We found that the immunity regions of the P22-like phages and prophages also show mosaicism, with some phages sharing an almost identical immunity region, and other phages completely lacking homology to the immunity region of BTP1. For example, there is no significant nucleotide homology between the immunity regions of BTP1 and P22. In contrast, prophage ST104 contains a very similar immunity region to BTP1 (Supplementary Figure S3). The differences between the immunity regions suggest the regulation of lysogeny may differ between these prophages. We note that the sieB gene, encoding an asRNA-based superinfection exclusion system (Ranade and Poteete, 1993), is present in the immunity region of P22 and ST104 and is absent from the immunity region of BTP1.

The BTP5 structural genes suggest it is a member of the Myoviridae family of long contractile-tailed phages and it shares sequence similarity with P2-like phages including P2, WΦ, and 186 which infect E. coli. P2 has been described as the prototype for the non-inducible class of temperate phages, as its repressor protein lacks an identified RecA cleavage site (Kalionis et al., 1986), and P2-like phages are known to be associated with satellite phages (Calendar, 2006). However, BTP5 contains two genes that show homology to the SOS operon of coliphage 186 (Brumby et al., 1996): tum (STMMW_32041) and orf97 (STMMW_32031), which have been shown to function as an antirepressor system and facilitate induction of coliphage 186 (Shearwin et al., 1998). No function was able to be inferred based on sequence homology to other studied phages for 7/42 of BTP5 ORFS. The complete functionally annotated map of BTP5 is shown in Supplementary Figure S1B. The chromosomal context of the BTP5 attB is shown in Supplementary Figure S2B, and the core att sequence for BTP5 is given in Supplementary Table S3.

Comparison of BTP5 to other P2-like phages and prophages suggests that there is less genetic mosaicism than in the P22-like group (Figure 2B, full details of the sequences used in this analysis are given in Supplementary Table S5). The majority of the P2-like phages and prophages show conservation of nearly the entire capsid and tail gene clusters, which make up ∼60% of the phage chromosome. Structural gene homology is not restricted to the genus Salmonella, and BTP5 shares >70% sequence identity to the structural genes of phages and prophages from other Enterobacteriaceae including Escherichia, Klebsiella, Serratia, and Enterobacter. However, outside of the structural gene clusters, there is limited homology to other phages or prophages, suggesting that BTP5 may represent novel phage biology.

Prophages were defined as viable if functional (infectious) phage particles could be produced, with or without induction. In order to test viability of each of the D23580 prophages, we created a series of naive indicator strains of D23580 using scarless genome editing to delete each prophage (see Materials and Methods). This resulted in a collection of five isogenic D23580 derivative strains, each lacking a single prophage (Table 2) and therefore harboring intact attB. As Gifsy phages have previously been shown to yield visible plaques only on hosts with short LPS, due to the receptor OmpC being blocked by the O-antigen (Ho and Slauch, 2001), an additional galE496 mutation was introduced to facilitate detection of Gifsy-phage plaques. Specific indicator strains used to detect each phage are shown in Table 2.

The supernatant from a 16 h LB culture of strain D23580 yielded circular plaques of ∼1 mm in diameter with turbid centers on the specific indicator strain D23580 ΔBTP1 at an unusually high abundance of ∼10⁹ PFU/ml (plaque-forming units per ml). Mitomycin C induction increased the phage titer 10-fold to ∼10¹⁰ PFU/ml (Table 2).

We propose that BTP1 represents a new species designated “Salmonella virus BTP1” within the P22virus genus. This is consistent with the International Committee on Taxonomy of Viruses (ICTV) criterion that members of a proposed species differ from those of other species by more than 5% at the DNA level as confirmed with the BLASTN algorithm. The BTP1 phage genome was sequenced and re-ordered to begin at the small terminase subunit, to conform to the reference sequence for the P22virus genus, Salmonella virus P22 (accession: BK000582). No sequence differences were found between the BTP1 prophage and phage sequences. The annotated genome and sequencing reads for the BTP1 phage have been deposited at the DDBJ/EMBL/GenBank database under the accession LT714109³ and study accession PRJEB18919. The proposed species “Salmonella virus BTP1” has been assigned the taxon ID 1934252.

The supernatant from a 16 h LB culture of strain D23580 did not form plaques on the specific indicator strain D23580 ΔBTP5, even after mitomycin C induction. Therefore, due to the inability to observe plaques, and lack of obvious functional degradation of the BTP5 prophage sequence, we were unable to verify the viability of prophage BTP5.

Supernatant from non-induced and mitomycin C-treated D23580 16 h cultures did not form plaques on specific indicator strains for Gifsy-2^D23580, ST64B^D23580, and Gifsy-1^D23580 (D23580 ΔGifsy-2^D23580 galE496, D23580 ΔST64B^D23580, and D23580 ΔGifsy-1^D23580 galE496, respectively) (Table 2). Closer inspection of the genome revealed evidence for functional inactivation of Gifsy-2^D23580, Gifsy-1^D23580, and ST64B^D23580 by mutation.

All coding genes remain intact in Gifsy-1^D23580 compared to Gifsy-1¹⁴⁰²⁸, where Gifsy-1 is known to be functional (Figueroa-Bossi and Bossi, 1999), with no evidence for functional degradation via pseudogene formation. However, a single nucleotide polymorphism (SNP) was detected in the -10 promoter region of the dinI gene of Gifsy-1^D23580 compared to the same region in S. Typhimurium 14028s (Gifsy-1^14028s) (Bossi et al., 2003). It has previously been shown that the anti-repressor gene of Gifsy-1, gfoA, is located immediately downstream of dinI and is co-transcribed from the dinI promoter (P_dinI-gfoA), making gfoA essential for the induction of Gifsy-1 phage (Lemire et al., 2011). This finding prompted us to test if the promoter SNP is responsible for the lack of detectable Gifsy-1 in the D23580 supernatant (Table 2). Single nucleotide allelic exchange of P_dinI-gfoA^D23580 with P_dinI-gfoA^14028s yielded viable (infectious) Gifsy-1 viruses that formed plaques on D23580 ΔGifsy-1. D23580 P_dinI-gfoA^14028s produced a Gifsy-1 titer consistent with that reported for 14028s, both with and without mitomycin C induction (Figueroa-Bossi and Bossi, 2004) (Figure 1B). The fact that a single modified nucleotide caused the production of functional phage shows that the P_dinI-gfoA SNP is responsible for lack of Gifsy-1 viability in D23580.

ST64B is defective in S. Typhimurium 14028s due to a frame-shift mutation in a tail assembly gene that prevented assembly of viable phage particles, and created two short gene fragments annotated sb21 and sb22 (Figueroa-Bossi and Bossi, 2004). The tail gene frame-shift mutation is conserved in ST64B^D23580 (D23580 genes STMMW_19871 and STMMW_19861 are the homologs of sb21 and sb22, respectively). Together with the failure of D23580 supernatant to produce plaques on a ST64B-sensitive host strain, we conclude that ST64B is similarly defective in S. Typhimurium ST313.

We noticed that the gene encoding the O replication protein of Gifsy-2^D23580 was a pseudogene; a 71 bp deletion in the middle of the coding sequence caused a frame shift at position 1,102,739 on the D23580 chromosome, and the generation of a truncated O protein. The O protein of lambdoid phages is required for replication of the phage chromosome during lytic growth and contains the phage origin of replication (Taylor and Wegrzyn, 1995). As the replication of the phage chromosome is an essential part of the lytic cycle, we suggest that this pseudogene is likely to be responsible for the lack of viable Gifsy-2^D23580 viruses when D23580 is grown with and without mitomycin C.

The SNPs associated with the loss of functionality of Gifsy-1^D23580 and ST64B^D23580 described above were conserved across 182 genomes of ST313 isolates from Malawi (representing both lineage I and lineage II of ST313) (Supplementary Table S6) suggesting that Gifsy-1 and ST64B are defective in all ST313 isolates surveyed. However, the deletion responsible for creating the O replication pseudogene of Gifsy-2^D23580 was only present in lineage II strains, with ST313 lineage I isolates encoding a full-length protein. Therefore, Gifsy-2 may still be viable in ST313 lineage I.

Prophages may contain superinfection exclusion systems that can interfere with other prophages. To control for the possibility that other D23580 prophages may be excluding the target phage in the specific indicator strains, we constructed a strain with all five prophages deleted (designated D23580 ΔΦ). When supernatant from D23580 ΔBTP1 was plated onto a lawn of D23580 ΔΦ, no plaques were observed, supporting the conclusion that of the five D23580 prophages, only BTP1 is capable of plaque formation (Table 2).

Comparison of the BTP1 and P22 late genes using tBLASTx shows that the structural regions share extensive homology, with the notable absence of the P22 immunity region (imml) in BTP1 (Figure 3A). Due to the high level of sequence identity and the detailed three-dimensional structural information that is available for many P22 proteins, it was possible to map the protein coding genes of BTP1 onto a structural model of P22 (Figure 3B). The capsid and tail genes of BTP1 show a high level of similarity to P22, with the majority of proteins sharing >90% amino acid identity. The genes involved in DNA injection (STMMW_03871-91) show less homology, suggesting subtle differences between the mechanism of DNA entry of the phages. The tailspike protein of P22-like phages consists of a short head-binding domain (HBD) which attaches the tailspike to the capsid, and a receptor-binding domain (RBD) responsible for recognition and binding to the bacterial receptor (Steinbacher et al., 1997). The HBD of the BTP1 tailspike shares 97% sequence identity with P22, however, the RBD is more divergent and is just 65% identical. This may indicate differences in the enzymatic activity of the tailspike protein, which in P22 is known to have endorhamnosidase activity (Iwashita and Kanegasaki, 1973).

The morphology of the BTP1 phage was determined by TEM. The purified BTP1 phage preparation contained short-tailed phages with icosahedral capsids of ∼65 nm in diameter (Figure 4B). The addition of purified BTP1 phages to susceptible S. Typhimurium 4/74 cells allowed the same short-tailed phages adsorbing to the surface of the bacterial cells (Figure 4A). Empty ‘ghost’ virions were also seen, suggesting that some phages had already adsorbed and injected DNA into the host cell at 20 min post-infection.

To determine whether the unusually high level of spontaneous induction (Table 2) is an intrinsic characteristic of the BTP1 prophage, a prophage cured derivative of S. Typhimurium LT2 (MA8505) was lysogenized with BTP1, and compared with LT2 lysogenized with P22. The phage titer of non-induced culture supernatants was determined. The titer of BTP1 in culture supernatants was 10,000-fold greater than the titer of P22 (Figures 5A,B) showing that a high frequency of spontaneous induction is an inherent characteristic of phage BTP1, independent of the D23580 host. The burst size of BTP1 in D23580 was around 600 (Supplementary Figure S4), resembling the burst size of phage P22 (Aramli and Teschke, 1999). The titer of 10⁹ PFU/ml in the D23580 non-induced 16 h culture (Table 2) suggests that the fraction of the cellular population undergoing phage-mediated lysis is ∼0.2%. In LT2 [BTP1], the fraction of the population undergoing phage-mediated lysis is ∼0.02%.

To determine if the high-level of BTP1 spontaneous induction was dependent on the SOS-response, we measured the spontaneously induced phage titer in a RecA^- BTP1 lysogen, using a RecA^- P22 lysogen as a comparator (Figure 5B). In host strain LT2 (MA8505), P22 spontaneous induction is virtually abolished in a ΔrecA background, and there is a 3-log reduction to the order of only 10 PFU/ml. In an LT2 (MA8505) lysogen, BTP1 spontaneous induction was also dramatically reduced in the ΔrecA background by 5-log but remained 50-times greater than that of P22 at around 500 PFU/ml. Considering that LT2 is not the native host of BTP1, we also assessed the role of RecA on spontaneous induction of BTP1 in D23580, and found a 4-log reduction in PFU/ml in a ΔrecA background. These results show that the majority of BTP1 spontaneous induction is RecA-dependent. However, there is still considerable RecA-independent induction of the BTP1 prophage, and we note that there is 1 log more RecA-independent induction in the native D23580 background than in an LT2 (MA8505) background.