3.1. Phage Isolation
To isolate a diverse pool of phages, we used water samples collected in the UK and Malawi, close to areas where Salmonella
or other related bacteria would be expected to coexist, such as sewage plants or outlets. We enriched for phages that plated on S.
Typhimurium or S.
Enteritidis hosts (Table 2
), and we selected plaques with different morphologies. The RAPD PCR assay was used to differentiate the phages, and 32 distinct phages were selected for whole genome sequencing (Figure S2
). The 32 bacteriophages included 20 phages isolated from Malawi and 12 phages from the UK.
The majority of the phages (25 phages, 78%) were isolated using the phage-susceptible S.
Typhimurium D23580 ΔΦ Δbrex
mutant as a host. This mutant was constructed by removing all known ‘anti-phage’ encoding genetic features, which include all five prophages [8
] and BREX bacteriophage exclusion defence system [32
], from a well-characterised S.
Typhimurium ST313 associated bloodstream infection in Malawi [4
The remaining seven phages were isolated with several other host strains, including the S.
Typhimurium ST19 representative strain 4/74, S.
Typhimurium D23580 ΔΦ [8
] (with the BREX system intact), and S.
3.2. Host Range Analysis
The host range of all 32 phages was tested using bacterial strains representing S.
Typhimurium ST19 and ST313, and S.
Enteritidis Global Epidemic and African clades (Figure 1
). No significant difference in host range was observed between phages isolated from the UK and those from Malawi (Figure 1
). A small proportion of phages (9 phages, 28%) that originated from UK or Malawi were able to infect S.
Enteritidis from the African clades, of which only two (BPS5 and BPS6) were isolated in Liverpool with S.
Enteritidis as the host. BPS5 was the only phage capable of infecting all hosts tested, including mutants. The phages BPS3, BPS6 and BPS7 were able to infect all wild-type Salmonella
strains tested. These results indicate that phages isolated in the UK had a broader host range than phages isolated from Malawi.
Of the 32 phages isolated, seven were unable to infect S.
Enteritidis. The Global Epidemic S.
Enteritidis strains P125109 and A1636 were susceptible to infection by the same phages, with the exception of phage ER8 that only infected A1636 (Figure 1
). This result reflects the fact that these S.
Enteritidis strains were very similar at the genomic level, differing by ~40 SNPs. S.
Enteritidis from the African clades were resistant to infection of most (>70%) of the 32 phages. S.
Enteritidis strain 4030/15, a representative from the West African clade, was susceptible to infection of the same phages as D7795 and CP255, representatives of the Central/Eastern African clade, with the addition of phages BPS1, ER2, ER4, ER5 and ER10.
Enteritidis strains, especially from the African clades, are also resistant to being infected by other commonly used bacteriophages, such as the Salmonella
phages 9NA, BTP1 and P22 (Figure 1
). The P22 transducing phage is commonly used for genetic manipulation of Salmonella.
However, P22 infects S.
Enteritidis from the Global Epidemic clade at a lower level than S
. Typhimurium, and did not cause clearing of infected S.
Enteritidis cultures due to an unknown mechanism. Interestingly, Salmonella
phage Det7 was capable of infecting both S
. Typhimurium and S.
Enteritidis hosts, sharing a similar host range profile to phages BPS3, BPS5, BPS6, and BPS7.
More phages infected S.
Typhimurium ST19 (Figure 1
) strains 14028s (32 phages) and LT2 (29 phages), than S.
Typhimurium 4/74 (eight phages), possibly due to prophage-encoded “anti-phage” defence mechanisms. Host range was tested on a S
. Typhimurium 4/74 mutant that lacked prophages (4/74 ΔΦ), and the number of infecting phages increased from eight to 13 (Figure 1
All isolated phages, except for ER21, were able to infect S.
Typhimurium D23580 (Figure 1
). However, in comparison to the original host D23580 strains that were deleted for either prophages (D23580 ΔΦ) or prophages and BREX (D23580 ΔΦΔbrex
), some plaque sizes were smaller on D23580 wild type (Figure 1
). This suggested phages ER22, ER23, ER25, BPS3, BPS6 and BPS7 are susceptible to a resistance mechanism encoded by one of the D23580 prophages. BREX did not appear to impact these phages (Figure 1
Phage BTP1 is one of the key prophages carried by S.
Typhimurium ST313 that has very high spontaneous induction (~109
]. BTP1 is active and highly conserved in both the lineages 1 and 2 of ST313 that are responsible for about two-thirds of the iNTS cases in Sub-Saharan Africa [53
]. Here, we used a S.
Typhimurium D23580 mutant (D23580 ΔBTP1) to show that BTP1 conferred resistance against phage ER21, possibly due to the removal of the novel BstA defence system [10
] or the GtrAC LPS-modification enzymes [11
Cell-surface properties frequently mediate phage-resistance by blocking attachment of tailed-bacteriophages, either by modification of the lipopolysaccharide (LPS), outer membrane proteins or flagella [54
]. The LPS oligo-polysaccharide residues (O-antigen) are common phage receptors [55
]. To determine whether the O-antigen was required for phage infection, a S.
Typhimurium D23580 mutant lacking all prophages, the BREX system, and galE
(D23580 ΔΦ Δbrex
) was tested. The gene galE
encodes the UDP-glucose 4-epimerase component of the galactose biosynthetic pathway, and is required for the synthesis of the LPS O-antigen of Salmonella
. Phages ER25, BPS3, BPS6, BPS7, ER6 and ER18 were unable to infect the mutant D23580 ΔΦ Δbrex
), suggesting that LPS is the receptor for attachment of these bacteriophages.
3.3. Comparative Genomics and Phylogeny of Isolated Phages
We used whole genome sequencing to investigate the taxonomy and relatedness of all the Malawian and UK phages. Pairwise genome comparison of nucleotide similarity between the 32 isolated phages revealed three main clusters (Figure S3
). The relatedness of the bacteriophages was also assessed by alignment of the gene encoding the terminase large sub-unit (terL
) that was present in all phages (Figure 2
Cluster 1 contained phages isolated in the UK that were able to infect all wild-type Salmonella strains tested. Cluster 2 included a mixture of bacteriophages isolated in the UK and Malawi that were only capable of infecting S. Typhimurium (except phage BPS1). Cluster 3 was divided into two sub-clusters: 3.a contained phages isolated from the UK, and 3.b from Malawi. Generally, the Cluster 3 phages infected S. Typhimurium 14028s, LT2 and D23580, and S. Enteritidis from the Global Epidemic clade, with a few of the phages infecting the representative West African clade strain 4030/15. Phage ER25 was the only one that did not belong to Cluster 1, 2 or 3. Phages belonging to Cluster 1 had a relatively large genome of ~158 kb, compared to the ~60 kb genomes of Cluster 2 phages, and ~43 kb genomes of Cluster 3 phages.
Comparative genomic analysis of Cluster 1 phages (Figure 3
A) showed that BPS3, BPS5 and BPS6 were closely related to S117 (>98% nucleotide identity; GenBank MH370370.1), a phage originally isolated on S.
Typhimurium LT2 [56
]. Figure 3
B shows that most phage-encoded genes are conserved throughout Cluster 1. BPS3 lacked one of the tail-spike genes present in BPS6, BPS5, BPS7, and reference phage S117, a gene that is upstream of the predicted virulence protein VriC, also found in other Salmonella
]; however, it is not clear whether VriC is a functional effector protein. Other genes were missing from Cluster 1 phages such as BPS7, that encode hypothetical proteins preventing any inference of their function. The comparative genomic analysis identified 26 SNP differences between BPS5 and BPS6, all of which are in intergenic regions upstream hypothetical proteins and tRNA genes. None of these nucleotide differences explained why BPS5 is capable of infecting all hosts tested, including the mutant S.
Typhimurium D23580 strain with a short LPS (D23580 ΔΦ Δbrex
), but BPS6 could not. Interestingly, BPS3 is the only phage in Cluster 1 with a different tail-spike protein (tail-spike 1).
Cluster 2 (Figure 4
A) included phages isolated from Malawi and the UK that were closely related (>93% identity) to S.
Typhimurium phages iEPS5 (GenBank KC677662.1) [58
] and øχ (Chi, GenBank NC_025442.1) [59
], which uses the flagellum as the receptor, and phage 37 [60
] isolated from sewage in India. Based on the comparative genomic analysis (Figure 4
B), seven structural genes including a tape measure protein, head vertex, and a portal protein were identified. Additionally, genes encoding putative endolysins proteins A and B and terminase large and small subunits (terSL)
Cluster 3 was the largest and most diverse cluster, with different host range and genome composition. Genome comparison (Figure 5
A) indicated a clear division between sub-clusters 3.a and 3.b, with sub-cluster 3.b only containing phages isolated from Malawi. Phages ER21, ER22 and ER23 (sub-cluster 3.a) (Figure 5
B) were closely related (>96% identity) to phage S134 (GenBank MH370381) [56
], originally isolated using S
. Enteritidis PT1 strain E2331, which might indicate their ability to infect S.
Enteritidis from the Global Epidemic clade. Sub-cluster 3.b included very closely related phages, with differences only in two hypothetical proteins (Figure 5
C). Several structural genes were identified in phages from sub-cluster 3.a, however no structural genes were found in sub-cluster 3.b isolates, highlighting the novelty of these phages. Genes related to replication, like DNA polymerases and topoisomerases, and recombination were identified in phages from sub-cluster 3.b. Additionally, the lysozyme encoded by gene rrrD
was identified; however, most genes were identified as hypothetical proteins.
Finally, phage ER25 did not belong to any of the identified clusters. Comparative genomic analysis demonstrated that phage ER25 is very similar (>99% identity) to P22 (GenBank AF217253.1) [61
], differing by only 49 SNPs (Figure S4
). This finding, complemented by analysis with PhageAI, suggests that ER25 was the only temperate phage identified, whereas all other isolated phages were likely to have a virulent life cycle.
Four bacteriophages were selected as representatives of each Cluster for morphological analysis by transmission electron microscopy. All phages were from the order Caudovirales, with two main morphotypes representing different phage families. Phages from Clusters 2 and 3 had a long (>100 nm) non-contractile tail and belonged to the Siphoviridae family, whereas the phages from Cluster 1 had a contractile tail, belonging to the Ackermannviridae family, confirmed by genome comparison with phage Det7 that also groups in Cluster 1.