Bacterial isolates and whole genome sequencing

163 DT8 and 34 DT30 strains were isolated during surveillance and obtained from the Animal and Plant Health protection Agency (APHA), and Public Health England (PHE) United Kingdom (now UKHSA). Whole genome sequence data for these isolates was produced by Illumina Mi-Seq. Three DT8 clade long read reference genomes were (S04527-10 DT8, L01157-10 DT8, S03645-11 DT30) generated by Pacific Biosciences SMRT sequencing technology [40]. Information on the source, genome sequence accession and phage type for each strain described in this study are reported in S1 Table.

Phylogenetic analysis

Phylogenetic trees from whole genome sequence data were constructed using SNPs identified in the whole genome sequences by aligning reads using BWA-MEM [41], variant calling with Freebayes [42] and SNP filtering using vcflib/vcftools [43], combined as a pipeline using Snippy v4.3.6 [44]. Maximum-likelihood trees were constructed using a general time-reversible substitution model with gamma correction for amongst-site rate variation with RAxML v8.0.20 with 1000 bootstraps using scripts and parameters (S1 File) [45].

Bacterial growth condition and recombinant strain construction

Bacterial strains were routinely cultured in Miller Lysogen Broth (LB) containing 10g NaCl, 10g tryptone and 5g yeast extract per litre (Formedium, Swaffham, UK) with shaking at 200 revolutions per minute in atmospheric conditions at 37⁰C. For culture on solid medium, LB was supplemented with 1.5% agar (Oxoid). Construction of recombinant strains was using the lambda red recombineering with primers GTGTAGGCTGGAGCTGCTTCG and CATATGAATATCCTCCTTAGT to amplify the cat gene or aphII gene from plasmids pKD13 or pKD14, respectively [46]. Variable non-priming sequence at the 5’ end of primers was used to direct allelic replacement in the chromosome (S1 Table). This approach was used to knock out the wzy gene (replaced by aphII), insertion of the cat gene into the wzy-thrS intergenic region, the attL^wzy (replaced by aphII gene) region and to insert a kanamycin resistance cassette in the neutral intergenic region adjacent to the iciA gene in strain L01157-10 [47]. The Lambda red recombinase encoding plasmid, pSIM18 was used for the recombineering step [48]. This plasmid uses a temperature sensitive replicons and promotor to activate recombination machinery proteins exo, beta, and gam when exposed to 42⁰C, and cure the lambda red containing plasmids at 37°C. Knock in of a wzy-tet cassette, and tet cassette alone was carried out as described previously with minor modifications [49]. A modified pDOC-GG-glmS plasmid with ampicillin selection marker was used as a delivery vector. The plasmid encodes a λ-red recombinase which allows sequence specific homologous recombination. The constructs were designed to insert the cassettes in the glmS intergenic region which was previously reported as a neutral locus for S. Typhimurium fitness [49]. All the restriction enzyme digestion steps were using BsaI type II restriction enzyme (NEB). The Golden gate assembly [50] of the modified pDOC-GG was using primers in S1 Table. The tet cassette was obtained by digestion of the muABbla tet 5’ plasmid. The cassettes were inserted in different S. Typhimurium strains as described in the text.

Estimation of ancestral phenotype histories

We generated various models, and tested model fit by using the likelihood ratio test (LRT) in R. We compared scaled maximum likelihood generated ancestral probabilities at bifurcating nodes between models generated using an equal rate for transition (Q) or allowing different rates for Q. We also generated models using a Markov-Chain with Monte-Carlo (MCMC) sampling approach using equal rates for Q and another with a predefined distribution of Qs. The likelihood ratio distribution asymptotically converges toward a χ² distribution. We determined which model was significant using the LRT and identify where the result sits within a χ² probability table, using one degree of freedom for an extra parameter of different rates of Q (S1 File). The tested models were scaled ML with equal rate (Q = -143.6473) or different rate (Q = -90.9481) using ace from the R package ape [52] and MCMC [51] with equal rate (Q = -140.3482) or different rate (Q = -185.748) using the make.simmap function from phytools [55]. This last model (MCMC with different rate) had the greatest log-likelihood and was therefore accepted as the most likely model.

The ancestral histories across a reconstructed phylogeny of 165 DT8 isolates and 34 DT30 isolates were assessed to determine the probable phenotype of each ancestor at each node in the tree. This was then used to determine a likely history of changes from the phage sensitive to phage resistant phenotype (DT8 to DT30). The MCMC approach was conducted with discrete character mapping using posterior sampled maps from SIMMAP [52]. One thousand sampled stochastic character maps were constructed after a burn-in period of 1000 iterations for Q, followed by 1,000,000 Markov-Chain steps and sampling for the posterior every 1000 generations using the pre-computed distribution of Q [53]. As a complementary approach, pastML [53] was also used to estimate switches from DT8 to DT30 at each node, using the MPPA (marginal posterior probabilities approximation) model and estimating the character variation from tips. Resulting data was interpreted and viewed using R package phytools [53], iTOL [54], and pastML [55]. Scripts and parameters are available (S1 File).

For the MCMC approach, a permutation test was used to assess if the estimated character state at a node was due to a skewed result from sample bias, with 20 permutations. Tip labels were randomly assigned to the tree through the ‘sample’ R command and a starting tree constructed using the maximum likelihood-based function ACE software, with a subsequent tree sampled every 100 iterations of MCMC, 100 times, resulting in a distribution of 100 trees with ancestral states for each permutation. To assess whether the permutated data was from the same distribution as that estimated from actual tip data, pairwise, Mann-Whitney-Wilcoxon tests were performed.

Bacterial genome wide association

Genome wide association was conducted to discern potential genetic polymorphisms associated with a trait of phage resistance for 34 DT30 isolates, or susceptibility to 10 phages for 164 DT8 isolates. This was undertaken through DNA of length K (k-mer) based analysis, where k-mers were identified from draft genome assemblies of each of the 196 isolate sequences using frequency based string mining algorithm FSM-lite [56]. The approach adjusts the probability of association of k-mers based on phylogenetic structure. The population structure was estimated using mash and converted to a three dimensional distance matrix [57]. Subsequently, a mixed linear model approaches were used for testing k-mer significance implemented with Sequence Element Enrichment (SEER) using scripts and parameters (S1 File) [58]. This was done initially with no significance filtering of k-mers and then the top 1% of k-mers in the range of p-values determined a likelihood ratio test (LRT) p-value cutoff of 1x10^-3. k-mers with a likelihood ratio test (LRT) p-value < 1x10^-1 were plotted.

Determination of specific sequence by mapping and local assembly

For sequence read mapping to the wzy locus to determine read-depth, a genomic region of 6kb encompassing the wzy locus and flanking genes was extracted from reference sequence L01157-10 (RefSeq Accession no. GCF_902500305.1), and short-read Illlumina WGS data for 196 DT8 complex isolates was mapped to this region using Bowtie2 without filtering secondary mapped reads to ensure that any reads mapping to the region would be included [59]. Bedtools-2.26.0 was used to extract the read depth per nucleotide [60]. These were split into 250 bp bins and the mean average of each section used as raw data for heatmap using R package gheatmap of ggtree [61]. The read data was normalised by dividing the raw value for 250 bp bins by the average read depth over the chromosome. This enabled visualisation of each wzy region in the context of each sequencing run. SNPs were identified in wzy regions using snippy v4.3.6 as previously mentioned with DT8 isolates S04527-10 and L01157-10 as references for variant calling [40]. For detection of the wzy locus in representative S. Typhimurium genomes, ARIBA was used along with a custom database containing the sequence of the att, wzy, nucA and thrS loci from the strain L01157-10 whole genome sequence.

Determination of phage sensitivity and phage type

Phage typing was carried out as described in Public Health England’s phage typing protocol for S. Typhimurium using typing phages. STMP8, 10, 18, 20, 29, and 32 were a kind gift from UKHSA. Briefly, a single colony of the bacteria to be tested was incubated in 4 ml of nutrient broth with static, atmospheric conditions at 37°C for 2 hours. Subsequently a nutrient agar plate was flooded with the culture before drying. 10 μL of each phage suspension at recommended titre dilution was spotted onto the plate and incubated for 16 hours. Phage typing was used to assess that strains of DT8 and DT30 were susceptible/ resistant to the corresponding phage. Constructed mutants were also subject to challenge with the typing phage to assess the phenotype. A colony picked from the clearance zone in an experiment in which strain L01157-10 was exposed to phage STMP10 was used in reversion experiments (L01157-10 Δwzy).

Determination of genotype and wzy circularisation by PCR amplification

Presence of the wzy region was routinely tested using primer pair TGCGACTATCAGGTTACCGT and GTTAGCGTGCGGTCAAGATC that anneal in the nucA and thrS genes. qPCR was utilized to quantify the wzy^- genotype in clonal populations with and without phage selection. Specific amplification of the circularised wzy locus was using primers AAGCCGAGACTCAGAGTGAC and CTCCGCCCTAATCCACATCT, and the same primers were used for the determination of the nucleotide sequence of the amplicon using TubeSeq (Eurofins). The consensus sequence of the PCR product and a circularised version of the wzy region (S2 File) were aligned using Muscle [62].

Quantification of wzy genotype

The relative and absolute quantification of wzy and Δwzy genotypes was determined by qPCR using an Applied Biosystems StepOnePlus. The strain to be tested was grown for 16–18 hours at 37°C in LB broth with shaking atmospheric conditions. This was sub-cultured in 10 mL of LB broth with 10⁴ cells per mL using a guide of an OD₆₀₀ of 3.5 equalling 10⁹ cells per mL before challenge with 3.34x10⁴ or 3.34x10⁵ pfu/mL with STMP10. Then at 2-hour intervals for 8 hours, and a sample 24 hours post inoculum, 400 µL of culture was taken and stored at -20°C. Once all samples had been frozen genomes were extracted using Promega Maxwell. L01157-10 was used as a positive control for growth, L01157-10 wzy^- for control of growth of a resistant mutant under phage pressure, and a pure lysate with 3.44x10⁵ phage to control for any effect of phage DNA increasing the Ct values. qPCR was undertaken using extracted genomic DNA with rpoD as a housekeeping control gene, and wzy was assessed using primer pairs wzy_qPCR_absent_F, wzy_qPCR_absent_R and wzy_qPCR_present_F, wzy_qPCR_present_R. A limit of detection for wzy^- was established to the nearest log₁₀–1 in 10 dilution of culture.

Transfer of the wzy locus during co-culture

Co-culture and PCR amplification of the wzy locus from suspected revertant colonies was used to assess the reversion of wzy. A donor strain of S. Typhimurium L01157-10 wzy⁺ was constructed by the insertion of a cat gene conferring resistance to chloramphenicol into the intergenic region between wzy and attR^wzy. A recipient strain of S. Typhimurium L01157-10 Δwzy picked from the otherwise cleared plaque formed by phage STMP32 in an agar overlay assay and determined to have the deletion by analysis of whole genome sequence was further engineered by inserting an aphII gene conferring resistance to kanamycin in the 5’ intergenic region of iciA and subsequent selection of a nalidixic acid resistant variant by culture on LB agar containing 100 mg/l nalidixic acid. Two genetically unlinked selectable markers were used in the recipient to discount the possibility of transfer from the recipient strain to the donor. We reasoned that potential transfer of the aphII gene by transduction and selection for nalidixic acid-resistant variants of L01157-10 wzy⁺ are both rare events and therefore the frequency of both events occurring together is extremely unlikely. Donor and recipient strains were each cultured for 18 hours in LB and diluted to OD_600nm of 0.05 and 2.5 ml of each mixed and incubated for 24 hours with shaking. The CFU of donor and recipients were enumerated by culture of serial dilutions on LB agar plates supplemented with 0.025mg/ml chloramphenicol (donor), 0.05mg/ml kanamycin and 0.03mg/ml nalidixic acid (recipient) and all three antibiotics to enumerate revertants. The presence of the wzy region was tested using primer pair TGCGACTATCAGGTTACCGT and GTTAGCGTGCGGTCAAGATC that anneal in the nucA and thrS genes, presence of wzy was tested using primers GCCTGAAGATTTTGGCGCAT, TGCGCTGACTTTGTTTCCTG that both anneal within the wzy gene and presence of the cat cassette with ACAAACGGCATGATGAACCT and GCACAAGTTTTATCCGGCCT that both anneal within the cat gene. Six revertants were subject to WGS with Illumina HiSeq 2500 to check the genotypes of the mutants and reversion of wzy within the same chromosomal location.

The S. Typhimurium DT30 strains are sporadically distributed within the DT8 clade

Isolates typed as DT8 and DT30 are predominantly isolated from ducks and form a distinct clade within the S. Typhimurium population structure [28,36,40]. Of 2213 S. Typhimurium strains isolated from ducks between 1992 and 2016, 68% (n = 1509) were DT8 and 23% (n = 352) were DT30. To further investigate the phylogenetic relationship of DT8 and DT30 strains we established a convenience collection of 176 whole genome sequences of S. Typhimurium isolated between the years 1993 and 2013 during surveillance by the Animal and Plant Health Agency (APHA), in the UK (S1 Table). These whole genome sequences were supplemented with 41 isolates from human infection and two genomes described previously [63]. A maximum likelihood phylogenetic tree based on sequence variation in the core genome of 200 isolates (Fig 1A) indicated that strains of DT30 strains were distributed widely on the phylogenetic tree of the clade among DT8 with little evidence of clonal expansion.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Phylogenetic relationship and association of sequence polymorphisms of S. Typhimurium DT8 and DT30 strains.

(A) Maximum-likelihood phylogenetic tree constructed using 2,297 core genome variable sites from 162 DT8 strains and 34 DT30 strains. The heatmap indicates the relative level of reads mapped to the wzy locus of S. Typhimurium DT8 reference strain L01157-10 in 250 bp bins for each strain. Coloured outlines denote the approximate location of polymorphisms that may disrupt Wzy function. (B) Bacterial genome wide association (GWAS) to identify sequence polymorphisms associated with phage resistance phenotypes of DT8 (susceptible) and DT30 (resistant). Manhattan plot of all significant k-mers position and likelihood ratio test (LRT) probability values of significant k-mers mapped to in S. Typhimurium DT8 reference strain L01157-10 chromosomal. The density of k-mers at a particular genomic coordinate is indicated by colour (inset key). (C) Location and depth of k-mers mapping to the wzy locus. Horizontal lines are k-mers with significant association with phage sensitivity defining DT8 and DT30.

https://doi.org/10.1371/journal.pgen.1011688.g001

The wzy locus is polymorphic in the DT8-DT30 clade and is flanked by two direct perfect repeats.

To test the hypothesis that genome sequence polymorphisms account for decreased sensitivity to phage predation of DT30 compared to DT8 strains, a genome-wide association study (GWAS) was undertaken to identify k-mers associated with these two phagetypes. 2,617 k-mers that had a significant (likelihood ratio test, LRT p < 0.1) association with the DT8 or DT30 trait after adjusting for the phylogenetic signal, mapped to four loci in the DT8 L01157-10 chromosome. A minority of k-mers (n = 602) had an LRT p-value ranging from 0.1 and 0.01 and mapped to yjbR (n = 200) encoding an uncharacterised DNA-binding protein encoding gene, fucO (n = 201) encoding lactaldehyde reductase or gss (n = 201) encoding a bifunctional glutathionylspermidine synthetase/amidase. The majority of k-mers, (n = 2,015) mapped to wzy gene which encodes an O-antigen polymerase and had a likelihood ratio of association p-value < 0.0001 (Fig 1B). Quantification of whole genome sequence reads of the S. Typhimurium DT8 and DT30 strains that aligned to sequence within and flanking the wzy locus of S. Typhimurium DT8 strain L01157-10 whole genome sequence indicated a region of zero mapped reads in 14 of 34 DT30 isolates, consistent with a deletion of this region (Fig 1A). Like other S. Typhimurium strains, the wzy gene was present in the rfc locus in the strain L01157-10 whole genome sequence and the wzy gene in the rfb locus was absent [64]. A further 20 strains that were typed as DT30 had sequence coverage across the wzy locus. Of these, three DT30 strains with sequence coverage had mutations resulting in a premature nonsense mutation in wzy and likely to abrogate function of the o-antigen polymerase, two SNPs and a small indel. Another strain had a non-synonymous SNP predicted to result in a lysine to asparagine substitution with unknown impact on function (Fig 1A).

The assembled sequence of S. Typhimurium DT8 strain L01157-10 had 2816 bp between the stop codon of nucA and the start codon of thrS compared to 259 bp in the assembled whole genome sequence of all 14 strains with the deletion at the wzy locus. Therefore, the deletion was 2558 bp. The genome sequence flanking the wzy gene contained two 113 bp identical direct repeats that coincided with the deletion boundaries observed in the DT30 strains in which the deletion at the wzy locus contained a single copy of the repeat sequence (S1 Fig). The direct repeats flanking the wzy gene were therefore designated attL^wzy (nucA - wzy intergenic) and attR^wzy (wzy - thrS intergenic) and attB^wzy (∆wzy) due to their resemblance to att repeat sequences that flank prophage in the bacterial chromosome and that mediate excision [65]. The att sequence was present in all 134 S. Typhimurium genome short-read sequence of representative strains (S3 Table), described previously [40]. We did not investigate the distribution in other serotypes of S. enterica. Together these observations were consistent with a site-specific mechanism for deletion of this region that is likely to be widely present in S. Typhimurium (Fig 1C).

Deletion of the wzy gene results in resistance to typing phage

To investigate the role of wzy in the sensitivity to predation by typing phages, a series of genetically engineered variants of DT8 strain L01157-10 in which the wzy gene was either deleted or replaced in an alternative genomic location for complementation were constructed. Strain L01157-10 Δwzy::aphII in which wzy was deleted and replaced by the aph gene was resistant to phage STMP32 tested DT8-lysing phage (Fig 2). Reintroduction of the wzy gene into the 3’ intergenic region of the glmS gene, linked to a tet gene to aid selection, resulted in return of sensitivity to phage STMP32 while the tet gene alone did not result in restored sensitivity (Fig 2). The phage sensitivity phenotype of the L01157-10 Δwzy::aph strain was therefore the same as DT30 strains. However, it is known that mutation other than those in wzy can result in changes in LPS expression that can also affect sensitivity to phage predation. We therefore tested if insertion of the wzy gene linked to a tet gene could also complement the phage sensitivity phenotype of strain S03645-11 DT30 that was resistant to typing phage STMP32 (Fig 2). Consistent with mutation of wzy in this strain being responsible for the loss of sensitivity to typing phage and the DT30 phenotype strain S03645-11 wzy, tet was sensitive to the typing phage. But introduction of tet alone resulted in no change to sensitivity (Fig 2). Since a large proportion of DT30 strains had this specific deletion of the wzy gene and flanking sequence, we therefore concluded that this was a common mechanism for the switch from DT8 to DT30 phage sensitivity phenotype.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Sensitivity of S. Typhimurium DT8, DT30 and recombinant derivative strains to typing phage 32.

The genotype of each strain is indicated with shaded arrows indicating genes as labelled (right) and images of serial dilutions of STMP32 overlaid on agar containing each strain.

https://doi.org/10.1371/journal.pgen.1011688.g002

The Δwzy genotype variants are present in cultures of S. Typhimurium DT8 that increase in frequency during phage predation.

Since the wzy locus was flanked by attL^wzy and attR^wzy direct repeats was consistent with a site-specific mechanism of deletion we tested the hypothesis that excision of the locus occurs spontaneously during culture by PCR amplification across the wzy locus. PCR of genomic DNA prepared from stationary phase cultures of two DT8 strains (L01157-10 and S04527-10) using oligonucleotide primers that flanked the wzy locus including the attL^wzy and attR^wzy repeats, resulted in two amplicons of 2.8 kb and 150 bp consistent with a mixture of the wild type wzy locus and the deletion variant. In contrast, a naturally occurring variant of L01157-10 in which the wzy locus had been deleted, picked from resistant bacteria in the zone of clearance due to lysis by phage STMP10 placed over a lawn of L01157-10 (L01157-10 Δwzy), resulted in only the 150 bp amplicon consistent with deletion of the wzy locus (Fig 3A). Using qPCR with the same oligonucleotide primers we estimated that in stationary phase LB broth culture of L01157-10, the ratio of wzy: Δwzy was approximately 500:1. To investigate the impact of phage predation on the ratio of wzy: Δwzy, we determined the change in frequency of each genotype during culture of strain L01157-10 in the presence and absence of phage STMP10. In the absence of phage STMP10, the frequency of each genotype was stable during 24 hours of culture in LB (Fig 3B). In the presence of phage STMP10, the frequency of the wild type wzy locus decreased and the frequency of the Δwzy genotype increased, consistent with selection of the deletion variant in response to phage predation. The genotype of a strain in which wzy was deleted and replaced by the aph gene remained constant in the presence of phage STMP10. Together these data suggest that the wzy locus is spontaneously excised during culture and the deletion variant is maintained at a constant frequency. Further, upon predation by a phage that uses O-antigen as a receptor, the frequency of the deletion variant increases relative to the wild type resulting in a change in ratio of wzy: Δwzy to approximately 30:1. These data are consistent with deletion being a mechanism of evasion of predation by phage that use O-antigen as a receptor.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Determination of wzy locus genotype during culture and selection in the presence of phage predation.

(A) Genetic map of the wzy locus indicating the position of oligonucleotide primers used for genotyping using PCR and qPCR. Genes are drawn to scale but arrows indicating the position of primers is not drawn to scale. B) PCR amplification of the wzy locus of two S. Typhimurium DT8 strains L01157-10 and S04527-10, and a spontaneous phage resistant variant of L01157-10 using primer pair A with a long extension time to detect Δwzy and wild type wzy genotypes. (B) Quantitative PCR of wzy genotypes of S. Typhimurium DT8 with and without pressure from STMP10 using primer pair A with a short extension time to detect Δwzy genotype and primer pair B to detect the wild type wzy genotype.

https://doi.org/10.1371/journal.pgen.1011688.g003

Deletion of the wzy locus is through an attL-dependent excision mechanism

To investigate whether the direct repeats attL^wzy and attR^wzy were required for excision of wzy locus, strain L01157-10 recombinants in which either attL^wzy or attR^wzy were replaced by the aphII gene were constructed by allelic replacement. An L01157-10 DattR^wzy strain had a growth defect during culture in LB as indicated by small colonies on LB agar, potentially due to the deletion affecting the thrS promotor region and therefore was not investigated further. However, an L01157-10 DattL^wzy strain had a similar growth characteristic as the wild-type parent strain. PCR amplification across the wzy locus of strain L01157-10 resulted in a 2.8 kb band and a 150 bp amplicons but only the larger amplicon in L01157-10 DattL^wzy (Fig 4A). This was consistent with attL^wzy being essential for deletion of the wzy locus. Prophage excision that is also dependent on attL and attR repeats results in the formation of a circular molecular of the phage genome [65]. To investigate if the wzy locus also formed a circular molecule, we PCR amplified genomic DNA prepared from strain L01157-10 using oligonucleotide primers that annealed within the excised region of the wzy locus in an outward facing orientation. This reaction resulted in a 0.5 kb fragment consistent with a circularisation of the wzy locus. The nucleotide sequence of the amplicon aligned to the sequence that would be expected following att-mediated circularisation (S2 and S3 Files). We did not detect the 0.5 kb amplicon in L01157-10 Δwzy. Taken together, the observation suggested that loss of wzy is driven by excision of the region using the flanking attL repeat and likely the attR repeat and although a circular form is present long enough to be detected by PCR amplification it is was not retained in a strain in which the wzy locus had been excised in the past (Fig 4B).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Dependency of wzy locus deletion and formation of a circularised wzy locus molecule on attL^wzy repeat sequence.

(A) PCR amplification of the wzy locus from S. Typhimurium DT8 strain L01157-10, a spontaneous phage resistant variant of L01157-10 (L01157-10 Δwzy) and a variant of L01157-10 in which the attL^wzy repeat sequence was deleted. Inward facing oligonucleotide primers (green arrows) flanking the wzy locus (top panel) and outward facing oligonucleotide primers (purple arrows) from within the deleted region (lower panel) were used in separate amplification reactions. Note: these are different primers than those used in Fig 3. (B) Model for the genotypes resulting from excision and circularisation of the wzy locus.

https://doi.org/10.1371/journal.pgen.1011688.g004

Δwzy deletion reversion to wzy⁺ during co-culture with a wild-type strain.

For Salmonella Typhimurium to successfully evade predation by deleting wzy without suffering a fitness cost during its lifecycle, the wzy gene must be restored once predation ceases. We therefore investigated if Δwzy can revert to wild type wzy⁺ during co-culture of a Δwzy recipient strain with a donor strain encoding a wild type wzy gene. A donor L01157-10 strain was constructed in which the cat gene conferring resistance to chloramphenicol was inserted between the 5’ end of wzy and attR^wzy (Fig 5A, L01157-10 wzy, cat) so that transfer of the wzy locus could be selected by culture on chloramphenicol. A recipient L01157-10 Δwzy strain that was spontaneously resistant to nalidixic acid due to a non-synonymous (D87Y) in the gyrA gene, and with an aphII gene conferring resistance to kanamycin inserted in the 5’ region of the iciA gene by allelic exchange was used/constructed (L01157-10 Δwzy, aphII, nal^r, Fig 5B). A mixture of donor and recipient strain cultured in LB for 24 hours resulted in chloramphenicol, kanamycin and nalidixic acid triple-resistant L01157-10 at a frequency of 1x10^-8 per CFU. The culture of the donor and recipient strain in LB supplemented with 0.5 mg/ml mitomycin C, which is known to induce prophage activation, increased the frequency by 100-fold to 1x10^-6 per CFU (Fig 5B). No triple-resistant L01157-10 recipient strains emerged when this strain was cultured on its own in the presence or absence of mitomycin C supplementation. The assembled whole genome sequence of six arbitrarily selected candidate revertant strains was determined. All had the aphII gene, a non-synonymous (D87Y) in the gyrA gene and the cat gene inserted in the 5’ intergenic region of wzy, expected for reversion (Fig 5C).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Transmission of wzy locus from S. Typhimurium wildtype L01157-10 to S. Typhimurium L01157-10 Δwzy during co-culture.

(A) Genotypes and genomic context of selectable markers and the wzy locus in L01157-10 wzy::cat donor, L01157-10 Δwzy aph, Nal^r recipient and L01157-10 wzy::cat aph, Nal^r revertant strains. (B) Reversion frequency in the presence or absence of mitomycin C supplementation.

https://doi.org/10.1371/journal.pgen.1011688.g005

Ancestral state reconstruction of DT8/ DT30 clade is consistent with bidirectional switching between states.

Our laboratory data was consistent with a bidirectional excision and reversion of the wzy locus. To investigate the dynamics of phage sensitivity changes in natural populations of S. Typhimurium, we investigated the dynamics of DT8 and DT30 phenotype by estimating ancestral state at each bifurcating ancestral node and along each branches of the ML phylogenetic tree of DT8 and DT30 strain constructed previously in this study. We used two independent and complementary methods, a maximum likelihood approach and a Bayesian Markov-Chain with Monte-Carlo sampling (MCMC). The maximum likelihood estimation at each node inferred switches from DT8 to DT30 and from DT30 to DT8 (Fig 6A). Most ancestral nodes were predicted to be DT8, but there were multiple predictions of transitions from DT8 to DT30 and back to DT8 (Fig 6B). A total of 45 changes in phage resistance phenotype were predicted, consisting of 26 switches from DT8 to DT30 and 19 switches from DT30 to DT8.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Ancestral state reconstruction of phage type for S. Typhimurium strains within the DT8/ DT30 clade.

A) Maximum likelihood phylogenetic reconstruction of marginal ancestral states with empirical posterior probabilities used to infer an estimated history along the phylogeny. The outer ring indicated phage type DT8 (green) or DT30 (orange). The colours indicated at points in the phylogenetic tree are estimates of whether the state is bacteriophage resistant (DT30, orange) or bacteriophage susceptible (DT8, green). B) Collapsed-branch node graph to visualise the possible changes in state along lineages in (A). Filled squares indicate the state within a particular collapsed branch section of a tree, and how many tips track back to the section for DT8 (green) and DT30 (orange). Arrows indicate the direction of evolution, with the root node at the top and descendant nodes at intervals moving downwards. C) Probability density map indicating the probability that a state across the reconstructed phylogeny is DT8 from 1000 stochastic maps with periodic sampling from a distribution of possible trees with all DT8 probabilities shown as colours. Red indicates a probability of 1, and blue a probability of 0 that the ancestor was DT8.

https://doi.org/10.1371/journal.pgen.1011688.g006

Using a complementary Bayesian approach of ancestral estimation, the common ancestor was also predicted to be DT8 (Fig 6C). A mean of 39.659 changes in phenotype state between DT8 and DT30 was estimated based on 1000 simulated trees. Again changes were predicted to be bidirectional with an average of 25.754 switches from DT8 to DT30, and 13.905 for DT30 to DT8. To investigate whether the highest likelihood model where phage type transition rates were unequal was significant due to increased sampling of DT8 strains, rather than based on phylogenetic structure, we performed a permutation test. A total of 18 permutations out of 20 were significantly different from the data (S2A Fig), indicating 90% confidence that the observed data did not occur by chance. The Bayesian MCMC and ML approaches produced models that were highly correlated as indicated by an R² of 0.675 for the probability of outcomes for each common ancestor being DT30 from scaled maximum likelihood ancestral states and stochastic mapping (S2B Fig).

Together these analyses were consistent with multiple state changes from DT8 to DT30, back to DT8 consistent with a reversible mechanism consistent with the phage-resistant DT30 phenotype occurring sporadically throughout the population.