Ecological niche adaptation of a bacterial pathogen associated with reduced zoonotic potential

The emergence of new bacterial pathogens is a continuing challenge for agriculture and food safety. Salmonella enterica serovar Typhimurium (S. Typhimurium) is a major cause of foodborne illness worldwide, with pigs a major zoonotic reservoir. Two variants, S. Typhimurium phage type U288 and monophasic S. Typhimurium (S. 4,[5],12:i:-) ST34 emerged and have accounted for the majority of isolates from pigs in the UK in the past two decades, but have distinct host range and risk to food safety. ST34 accounts for over 50% of all S. Typhimurium infections in people while U288 less than 2%. U288 and ST34 form distinct phylogenetic clusters within S. Typhimurium, defined by approximately 600 SNPs within their 5 Mbp genomes. Evolution of the U288 clade from an LT2-like ancestor was characterised by the acquisition of AMR genes, insertions and deletions in the virulence plasmid pU288-1, and the accumulation of polymorphisms, some of which resulted in truncation of coding sequences. U288 isolates exhibited lower growth rate and viability following desiccation compared to ST34 isolates, characteristics that could affect transmission through the food chain. U288 and ST34 isolates exhibited distinct outcomes of infection in the murine model of colitis, and colonised pigs in a manner that affected the disease symptoms and distribution in organs. U288 infection was more disseminated in the lymph nodes while ST34 were recovered in greater numbers in the intestinal contents. These data are consistent with the evolution of S. Typhimurium U288 adaptation to pigs that may determine their reduced zoonotic potential. Importance Bacterial pathogens continually evolve to exploit new ecological niches as they emerge due to human activity including agricultural, medical or societal practice. The consequences of the emergence of new pathogens may affect outcome of infection and risk to human or animal health. Genome sequence can resolve the population structure, identify variants that are evolving as they enter a new niche, and pinpoint potential functional divergence. We report a variant S. Typhimurium that adapted to a unique niche distinct to that occupied by a second S. Typhimurium variant circulating in the same pig populations. Adaptation was accompanied by phenotypic and genotypic changes consistent with a more invasive lifestyle and a decreased zoonotic potential observed in the epidemiological record. Our findings suggest that pathogen genotypic variation may be useful in estimating zoonotic potential and threat to livestock welfare.


Introduction
Emergence of infectious diseases presents new challenges for the management of human and livestock health, with substantial human and economic costs through excess mortality and morbidity, and lost productivity. The emergence of 335 human infectious diseases between 1945 and 2004 was dominated by zoonoses and bacterial aetiological agents, in-part due to the emergence of new variants of drug resistant pathogens (1). Emergent niches are rapidly occupied by bacteria from the vast genetic diversity in the biosphere, that are best suited to its exploitation. The plastic genomes, due to horizontal gene transfer and errors during replication, and relatively short replication time, facilitate rapid evolution by the selection of variants with increased survival and replication (2).
The genus Salmonella consists of over 2500 different serovars, such as Typhimurium, Enteritidis and Choleraesuis, that have diverse host ranges, pathogenicity and risk to human health. Prevalent serovars differ between host species and in different geographical locations.
The prevalence of each serovar may also vary over time, although certain serovars are often characteristic of specific host species and maintained for decades. For example, S.
Typhimurium (including monophasic variants) has consistently been a dominant serovar in pigs globally and accounts for around two thirds of isolates in the UK (3). Despite the apparently stable prevalence of S. Typhimurium in pig populations over decades, the epidemiological record of variants identified by phage type, indicates a dynamic process where distinct phage types increase and decrease in prevalence over time (3). The  (3,4). These two variants co-existed in the pig population and together accounted for around 80% of isolates (5). Despite, approximately half of all pork consumed in the UK being from UK pig herds (6), throughout this period U288 were rarely isolated from human infections (7). In contrast, monophasic S. Typhimurium ST34 isolations from human infections closely reflected its prevalence in pig populations and by the year 2013, over half of all S. Typhimurium isolates from human non-typhoidal salmonellosis (NTS) in the UK were monophasic S. Typhimurium ST34 (8,9). A similar epidemiology has been reported in other European countries (10), where the remaining pork consumed in the UK is sourced.
Non-typhoidal Salmonella (NTS) causes up to 78 million human infections each year, globally (11) and pigs are a major zoonotic reservoir, with 10-20% of human salmonellosis in Europe attributable to pigs (12,13). A baseline survey reported prevalence of 21.2% and 30.5% in mesenteric lymph nodes and caecal contents, respectively, for UK slaughter pigs (14,15). It is believed that contamination of pig carcasses with faeces and gut contents at slaughter, and the ability of Salmonella to spread from the gut to other organs, results in contamination of meat products that enter the food chain and pose a risk to humans if improperly handled or cooked. However, the relative risk from contamination of meat by gut contents during slaughter or from tissue colonised by Salmonella prior to slaughter is not known. Survival in food depends upon adaptive response to environmental stresses including osmotic pressure from preservatives and desiccation, antimicrobial activity of preservatives and fluctuating temperatures during storage or cooking. In order to cause disease, Salmonella may also need to replicate to achieve a population size able to overcome the colonisation resistance of the host.
Multiple pathovariants of S. Typhimurium evolved from a broad host range ancestor resulting in distinct host range, outcome of infection and risk to food safety (16)(17)(18). An understanding of the molecular basis of risk to food safety of pathovariants is critical to improve assessment of risk and devise intervention strategies aimed at decreasing the likelihood that Salmonella is present in food intended for human consumption. Moreover, analysis of Salmonella genome sequences by machine learning has enabled the prediction of host of origin for specific variants, which may aid source attribution in outbreak investigations (18)(19)(20). We therefore investigated the population structure of S. Typhimurium U288 and the genomic evolution leading to the clonal expansion of U288 associated with porcine infection by analysis of whole genome sequences. We compared the interaction of representative strains of the S. Typhimurium U288 and the monophasic S. Typhimurium epidemic clades with the environment and in vivo in the pig host to gain insight into the phenotypic consequences of their distinct evolutionary trajectories.

Materials and Methods
Bacterial strains and culture. Salmonella Typhimurium U288 and ST34 isolates used in this study were isolated from human clinical infections during routine diagnostic testing by Public Health England (PHE), or from animals during routine surveillance or epidemiological investigation by Animal and Plant Health Agency (APHA) (Supplementary Table 1). All sequence data generated in this study is available in SRA database under BioProject accession number PRJNA641292. 131 genotypically diverse S. Typhimurium isolates used to place U288 and ST34 in phylogenetic context have been described previously (21,22 broth or MacConkey containing 5% Agar, and supplemented with chloramphenicol (30 mg/l) or kanamycin (50 mg/l) as appropriate.

Construction of wild type isogenic tagged strains and single knock out.
A modified recombineering method based on the Lambda Red system was used to construct knockout mutations in S. Typhimurium SL1344 and wild type isogenic tagged strains (WITS) (23). Briefly, primers where used to amplify the kanamycin resistance cassette from pKD4 and recombination into the genome was directed by the inclusion of 50 nucleotide sequence flanking the insertion site (Supplementary Table 2). For the construction of WITS, insertion was directed to the intergenic region of iciA and yggE at position 3,247,245 in SL1344 (24).
Each forward primer included a unique 10 nucleotide tag which allowed identification of the tagged strain through whole genome sequencing.
Determination of growth rate, biofilm formation and desiccation survival. For determination of growth rate, bacterial cultures in LB broth were diluted to approximately 1x10 5 CFU per ml and incubated at 37°C with atmospheric aeration or an anaerobic environment (10% CO 2 , 5% H 2 , %% O 2 and 80% N 2 ) and viable bacteria in colony-forming units (CFU) enumerated by serial dilution and culture on LB agar at 1, 3, 5 and 7 hours post inoculation. Doubling time was calculated in the exponential range of growth using the mean from three biological replicates. Determination of survival after desiccation was based on a method previously described (25). Briefly, bacteria were cultured in LB broth, harvested by centrifugation, washed with phosphate-buffered saline pH7.4 (PBS) and re-suspended in PBS and adjusted to OD 600nm of 1. Next, 0.05 ml of cell suspension were added to polystyrene 96 well plate (Nunc) and desiccated at 22°C, 36% relative humidity (RH  and amplified by incubating the sample for 72˚C for 3 minutes, followed by 14 PCR cycles   consisting of 95 o C for 1 minute, 65 ˚C for 20 seconds and 72 ˚C for 3 minutes. 20µl of   amplified DNA was added to 20µl of Kapa beads and incubated at room temperature for 5 minutes to precipitate DNA molecules >200bp onto the beads. The beads were then pelleted on a magnetic particle concentrator (MPC), the supernatant removed, and two 70% ethanol washes performed. After removal of the final 70% ethanol wash, the beads were left to dry for 5 minutes at room temperature before the beads were re-suspended with 20µl of 10mM Tris-HCl, pH8. This was then incubated at room temperature for 5 minutes to elute the DNA molecules. The tube was then placed back on the MPC, the beads allowed to pellet, and the aqueous phase containing the size selected DNA molecules transferred to a new tube. The size distribution of each purified library was determined on a PerkinElmer GX by diluting 3 µl in 18 µl 10mM Tris-HCl, pH8. Using the PerkinElmer GX software, the smear analysis function was used to determine the amount of material in the 400 to 600bp size range and this information used to equimolar pool purified libraries. Once pooled the samples were then subjected to size selection on a Sage Science 1.5% BluePippin cassette recovering molecules between 400 and 600bp. QC of the size selected pool was performed by running 1µl aliquots on a Life Technologies Qubit high sensitivity assay and an Agilent DNA High Sense BioAnalyser chip and the concentration of viable library molecules measured using qPCR. The genome sequence of S. Heidelberg (NC_011083.1) was used as an outgroup in the analysis to identify the root and common ancestor of all S. Typhimurium strains.
To infer the time of nodes on phylogeny, we used the BactDating software package implemented in R (28), in which sequence variation in the core genome was analysed with potential regions of recombination removed using Gubbins (29). The resulting sequence alignments were used to construct a maximum likelihood phylogenetic tree using RAxML and the true root was estimated using S. Typhimurium SL1344 genome as an outgroup. The MCMC was run for 1 million iterations and the convergence and mixing of chains were 113.3, 129.5, 145.8 for ( , , , respectively) calculated using the R package corda (30).

Pan-genome analysis and
In-silico genotyping. The pan-genome of U288, ST34 and representative strains of S. Typhimurium was determined as described previously (18). The presence of antibiotic resistance, virulence and plasmid replicon genes in short-read data was determined by the mapping and local assembly of short reads to data-bases of candidate genes using Ariba (31). The presence of candidate genes from the ResFinder (32), VFDB (33) and PlasmidFinder (34) databases was determined. Reads were mapped to candidate genes using nucmer with a 90% minimum alignment identity. This tool was also used to determine the presence of specific genes or gene allelic variants. The results of the ARIBA determination of the presence or absence of the lpfD gene were confirmed using SRST2 (35) setting each alternative form of the gene as a potential allele. SRST2 was also used to verify the ARIBA findings of the VFDB data set, as the presence of orthologous genes in the 1 0 Metabolic profiling using OMNIlog microarray system. To assess utilisation of carbon sources, isolates were cultured on LB agar overnight at 37 o C, inoculated into IF-0 medium containing tetrazolium dye, and added to a PM-1 plate (with 95 different carbon sources), according to the manufacturer's instructions (IBiolC). Accumulation of purple indicator dye as a consequence of redox activity was measured every 15 minutes for 48 hours. Raw absorbance data were processed using R software and the opm package (36). The area-underthe-curve was a metric for total respiration of the indicated carbon sources and plotted as a heatmap. were incubated for 30 minutes before washing 5 times with PBS and re-suspending in DMEM + gentamicin (100 mg/l) and incubating for a further 30 minutes at 37 o C in 5% CO 2 atmosphere to kill extracellular bacteria. The medium was then replaced with DMEM + gentamicin (10 mg/l) and the plates were incubated for 60 minutes at 37 o C in 5% CO 2 atmosphere. Plates were then incubated at 37 o C in 5% CO 2 atmosphere for 90 minutes, before a final wash with PBS + 0.1% Triton-X100. Plates were left for 2 minutes and cells were disrupted by vigorous pipetting for one minute followed by serial dilution and culture on LB agar to determine viable counts.
Streptomycin pre-treated mouse infections. Groups of five female 6-9 week old C57bl/6 mice were treated with 20 mg of streptomycin sulphate by oral gavage 24 hours prior to 1 1 LB broth and approximately 5x10 6 to 1x10 7 CFU were inoculated orally in 0.2 ml of PBS pH7.4 by gavage. On day 3 post-inoculation mice were humanly euthanised and the cecum was aseptically removed. An approximately 3mm section of the cecum half way down the organ from the ileal junction were removed and fixed in formalin for histopathology examination of 5um thin sections stained with hemotoxylin and eosin. Approximately 5 mg of cecum tissue was removed and placed in RNAlater and stored at -80 o C (Thermo Fisher).
The remaining cecum (approximately two thirds) was homogenised in sterile PBS pH7.4 and serial dilutions plated on LB agar containing 0.05 mg/ml kanamycin, incubated at 37 o C for 18 hours and viable counts enumerated.
Determination of Nos2 and Cxl1 expression in mouse cecum tissue. RNA was prepared using Guanidinium thiocyanate-phenol-chloroform extraction with Tri Reagent (Merck).
Tissue was homogenized in 1 ml of Tri reagent and disrupted using 1.4mm zinc oxide beads in a bead beater and tissue debris removed by centrifugation. 0.2 ml of nuclease free water and 0.2ml of chloroform were added to the supernatant, mixed and centrifuged for 15 minutes at 12,000 x g. 0.5 ml of isopropanol was added to the upper aqueous phase and centrifuged for 15 minutes at 12,000 x g. The resulting RNA pellet was washed twice with 70% ethanol, briefly dried and resuspended in 0.02 ml of RNase free water. The relative abundance of Nos2 and Cxl1 mRNA was determined by quantitative RT-PCR using primers specific to the test genes and the Gapdh house-keeping gene as the control (Supplementary Table ?) as described previously (37).

Mixed-strain infection of pigs. Four 6-week-old Landrace x Large White x Durock pigs
were challenged orally with a mixed-strain inoculum as previously described (38). Animal experiments were conducted according to the requirements of the Animals (Scientific Procedures) Act 1986 (project licence PCD70CB48) with the approval of the local ethical review committee. Pigs were confirmed to be culture negative for Salmonella before inoculation by enrichment of faecal samples in Rappaport-Vassiliadis broth at 37°C for 18 hours, followed by plating on MacConkey agar at 37°C for 24 hours. A mixed-strain inoculum was prepared by mixing equal volumes of individual cultures of the six strains grown statically at 37°C for 16 hours in LB broth supplemented with 50 mg/l kanamycin, which were optical density (OD 600nm ) standardized to contain 8.9 log 10 CFU/ml. This was confirmed retrospectively by plating 10-fold serial dilutions on MacConkey agar. Aliquots of the inoculum were stored at -20°C for DNA extraction. Five ml of the mixed-strain inoculum was mixed with 5 ml of antacid [5% Mg(SiO 3 ) 3 , 5% NaHCO 3 , and 5% MgO in sterile distilled water] to promote colonization and administered orally by syringe before the morning feed. Pigs were fed as normal following challenge. Rectal temperatures were recorded every 24 hours and faecal samples were collected at 24 and 48 hours post-infection.
At 72 hours post-infection, a section of distal ileal mucosa, mesenteric lymph nodes (MLNs) draining the distal ileal loop, a section of spiral colon, colonic lymph nodes (CLNs) and a section of liver were collected. Lymph nodes were trimmed of excess fat and fascia, and the sections of distal ileum and spiral colon were washed gently in PBS to remove nonadherent bacteria. One gram of each tissue was homogenized in 9 ml of PBS in gentleMACS M tubes using the appropriate setting on the gentleMACS dissociator (Miltenyi Biotec). Homogenates were filtered through 40-μm-pore-size filters and an aliquot was used to determine viable counts. The remaining homogenate was spread onto 10 MacConkey agar plates (500 μ l per plate) and incubated overnight at 37°C. The bacterial lawns recovered from each sample were collected by washing with PBS, and the pellets were stored at -20°C for DNA extraction.
Genomic DNA (gDNA) was extracted from the pellets using the NucleoSpin tissue kit (Macherey-Nagel), according to the manufacturer's instructions. The quality and quantity of DNA were assessed initially by NanoDrop 3300 (Thermo Scientific), and samples with an A 260/280 of <=1.8 were considered suitable for library preparation. These were confirmed further by using the DNA ScreenTape (Agilent Technologies) and the Qubit double-stranded DNA (dsDNA) BR assay kit (Life Technologies), respectively. One microgram of gDNA with a DNA integrity number (DIN) of <=6 was used for library preparation using the TruSeq PCR-free library preparation kit (Illumina) according to the manufacturer's protocol. Whole genome sequencing on the HiSeq system (Illumina) followed by bioinformatics analysis were performed as previously described (38), with the exception that strains were quantified by mapping sequence data to the unique WITS tag sequence incorporated chromosomally into each strain. Sequence data was submitted to the NCBI SRA database (Supplementary Table   3). For each strain, the percentage in a population was calculated as the average WITS frequency x 100. Data are presented as the mean ± standard error of the mean (SEM). Raw sequence data were deposited in the Biosample database at NCBI (Supplementary Table 3).

Single-strain infection of pigs.
From the strain phenotypes identified in the mixed-strain infection, one representative strain of each DT193 and U288 was selected for in vivo phenotype validation, S04698-09 and 11020-1996, respectively. The strains were grown statically at 37°C for 16 hours in LB broth supplemented with 50 mg/l kanamycin and the optical densities (OD 600nm ) were standardized to contain 9.2 log 10 CFU/ml, which was confirmed retrospectively by plating 10-fold serial dilutions on MacConkey agar. Groups of four Salmonella-free pigs were challenged orally with 5 ml of each strain as described above. (GraphPad Software). The viable counts of bacteria in each case is presented as mean ± SEM, and differences between strains were analysed using the Mann-Whitney test of significance.
Area under the curve analysis followed by the Mann-Whitney test was used to analyse the cumulative clinical scores of the infected pigs during single-strain infections. P values of ≤ 0.05 were considered to be statistically significant. The main U288 clade was closely related to 13 human clinical isolates, of various phage types, but predominantly U311; none were U288.

S. Typhimurium U288 and monophasic
To investigate the relationship of contemporaneous S. Typhimurium U288 in the UK pig population and human clinical isolates, we determined the whole genome sequence of 134 S.
Typhimurium U288 isolated from animals in years the 2014 and 2015 (APHA collection).
To place these in the phylogenetic context of S. Typhimurium, we included 131 isolates that represented diverse phage types, including 12 isolates from the current monophasic S.
Typhimurium ST34 epidemic (40). We also included the 33 human clinical isolates from 2014 and 2015 from the main U288 clade and 13 closely related isolates, and a U288 isolate (CP0003836), reported previously from Denmark in 2016 (41), for context. The phylogenetic structure of S. Typhimurium was consistent with that described previously (42), with a number of deeply rooted lineages, some of which exhibited evidence of clonal expansion at terminal branches ( Figure 2). All S. Typhimurium U288 isolates from pigs were present in a single phylogenetic clade together with the 33 isolates from human clinical infections (U288 clade, green lineages, Figure 2). The U288 clade was closely related to thirteen S.
Typhimurium isolates of various other phage types but none were phage type U288. Most of these were isolated from human clinical infections, and just two from animal hosts, both 1 6 from avian hosts ( Figure 2). Of note, S. Typhimurium strain ATCC700720 (LT2) differed by fewer than 5 SNPs from the common ancestor of the U288 clade and the 13 related non-U288 strains. S. Typhimurium strain ATCC700720 (LT2) was originally isolated from a human clinical infection at Stoke Mandeville hospital, London in 1948, and subsequently has been used for studying the genetics of Salmonella worldwide (43). The three U288 isolates from human clinical infections in the minor U288 clade clustered together with isolates of other non-U288 phage types.
S. Typhimurium U288 are rare and monophasic S. Typhimurium ST34 are common in human clinical cases. Epidemiological surveillance based on phage-typing indicated that S.
Typhimurium U288 is rarely isolated from clinical cases of salmonellosis in the UK, while monophasic S. Typhimurium is commonly reported (7). A similar epidemiology has also been noted in other European countries, including Denmark (10). However, estimation of relative risk of human infection based on phage-typing is potentially misleading, due to potential polyphyletic clusters of common phage-types. Our phylogenetic analysis indicated that the majority of U288 isolates are from a single clonal group, but that this clade also contained a number of isolates that were not identifiable by phage typing as U288, therefore, the contribution to human infection may be underestimated. Conversely, a proportion of the S. Typhimurium U288 isolates from human clinical infections were only distantly related to the clonal group of S. Typhimurium U288 associated with pigs in the UK, and therefore, contribute to an over estimation of the pig-associated U288 genotype to human infection. We The thrW locus was variably occupied by either the mTmV prophage in monophasic Typhimurium ST34 strain S04698-09 that carries the sopE gene (21), or ST104 in U288 strain S01960-05, a prophage previously described in S. Typhimurium DT104 strain NCTC13384 (44). U288 strain S01960-05 had a complete Fels-2 prophage which was absent from S04698-09. The S04698-09 genome also harboured two additional prophages related to HP1 and SJ46, that were absent from U288 strain S01960-05.
A notable difference in coding capacity affecting the core genome of the two strains resulted from hypothetically disrupted coding sequences (HDCS) due to introduction of a premature nonsense codon from small insertions or deletions (indels) resulting in a frame shift or single nucleotide polymorphisms (SNP). The monophasic S. Typhimurium ST34 strain S04698-09 genome contained three HDCS outside of prophage, with reference to S. Typhimurium SL1344. In contrast, S. Typhimurium U288 strain S01960-05 contained 19 HDCS outside of prophage (Table 1). Nine U288 HDCS encoded hypothetical proteins of unknown function (ygbE, yciW, SL2283, yfbK, SL2330, yhbE, SL0337, ybaO, SL1627, yfbB, and yqaA), while ten had predicted functions based on sequence similarity or known functions (assT5, assT3, dtpB, hutU, cutF, oadA, pncA, sadA, tsr, oatA, and rcoR). Several of these may be important for colonisation of the caecum of pigs following oral inoculation based on an initial screen using transposon insertion library (46). In particular, several insertion mutants in assT3 and assT5 indicated a fitness score of -2 and -5.76.
Analysis of the pangenome revealed a similar sized core and accessory genome, with 3962 and 4056 core gene families, out of pangenomes of 5501 and 5578 gene families, for U288 and ST34, respectively (Supplementary Figure 2). Lineage-specific differences in the accessory genome of U288 and ST34 were largely due to distinct prophage repertoires, the presence of SGI-4 in ST34, and the presence of pU288-1 (pSLT related) in U288. The prophage repertoire was particularly variable within the ST34 lineage, while plasmid sequence, including pU288-1 was highly variable in U288. Notably, U288 lacked lineagespecific genes present on the chromosome, with the exception of prophage.
The S. Typhimurium U288 clade evolved from an S. Typhimurium LT2-like common ancestor by genome degradation and acquisition of AMR genes. To investigate the evolutionary events associated with the emergence of the S. Typhimurium U288 clade, using short read sequence data of U288 clade isolates and related S. Typhimurium isolates we determined the distribution of AMR genes, plasmid replicons and plasmid sequence (pSLT, pU288-1, pU288-2 and pU288-3), and identified allelic variants of genes identified as HDCS in S. Typhimurium U288 reference strain S01960-05 ( Figure 4).
The incF replicon was present in all but one isolate from the U288 clade and closely related isolates, consistent with the presence of all or part of the pSLT plasmid-associated sequence ( Figure 4). Deletions of large parts of the pSLT-associated sequence were evident in the U288 clade isolates. Furthermore, in 58 of 133 U288 clade isolates, deletions affected two or more of the spv genes, previously implicated in virulence in the mouse model of infection (47).
The pattern of AMR gene presence was consistent with acquisition in two distinct evolutionary events. First, an incQ1 plasmid (pU288-2) was acquired concurrent with initial clonal expansion of the clade, followed by subsequent acquisition of a composite transposon on the pSLT-like plasmid, pU288-1 ( Figure 4). The incQ1 replicon of the pU288-2 plasmid was present in 97 of 133 U288 clade isolates, including a cluster of six most deeply rooted isolates in the U288 clade, and was associated with the strA, strB, sul2 and tetA genes, encoding resistance to streptomycin, tetracycline and sulphonamide antibiotics. AMR genes cml, sul3, dfrA, bla TEM and aadA that confer resistance to chloramphenicol, sulphonamides, trimethoprim, -lactam and aminoglycoside antibiotics respectively were present on pU288-1 in all but 16 U288 clade isolates. These AMR genes were absent from six of the most deeply rooted isolates in the S. Typhimurium U288 clade.

S. Typhimurium U288 isolates have a longer doubling time and exhibit greater
sensitivity to desiccation compared to ST34. Growth and survival in stress conditions encountered in food is likely to be an important factor in risk to food safety presented by Salmonella Typhimurium. We therefore compared U288 and ST34 isolate replication rate, motility, biofilm formation and ability to survive desiccation ( Figure 6). Strains from the U288 clade exhibited a longer aerobic and anaerobic doubling time and increased sensitivity to desiccation, but similar motility and capacity to form biofilm, compared to monophasic S. Typhimurium ST34 isolates. The mean doubling time for in the U288 isolates was 0.6 hours and 0.54 hours in aerobic and anaerobic environments, respectively, compared to 0.52 hours and 0.47 hours for three ST34 isolates ( Figure 6A).
Since strains of each clade exhibited distinct replication rates, we compared respiration for three isolates of ST34 and four isolates of U288, utilizing a range of substrates as the sole carbon source for metabolism. All of the strains tested were able to use the majority of 95 carbon sources tested, but there was variation among approximately a quarter of substrates ( Figure 6B). The pattern of carbon source utilisation of U288 and ST34 isolates was distinct from the commonly used lab strains S. Typhimurium 4/74 (histidine prototroph variant of strain SL1344). The inability or diminished ability of strain 4/74 to utilise m-Tartaric, Tricarballylic acid and D-xylose was a major factor that distinguished this strain from U288 and ST34 strains. Three ST34 strains clustered together, but the U288 isolates exhibited considerably greater diversity in carbon source utilization. Utilization of myo-inositol as a sole carbon source was the most pronounced phenotype that distinguished the two clusters of U288 isolates. Of note, strain S05968-02 and 11020-1996 that were able to use myo-inositol were isolated earlier and were more deeply rooted than strains S01960-09 and S07292-07 that were unable to use this source of carbon.
Desiccation is a common stress associated with the food chain. We observed a clade-specific variation in tolerance to desiccation by comparing ten U288 strains and three ST34 strains ( Figure 6C). Following desiccation for 24 hours, approximately 2% of the initial inoculum remained viable for all three ST34 strains. In comparison the mean viability of U288 strains was 0.1%, but varied between 0.0001% and 0.3% among the ten U288 strains tested.
The loss of ability to form biofilm is a common feature of some host-adapted variants of Salmonella enterica (48,49) (Figure 6D). S. Typhimurium strain SL1344 formed moderate biofilm that was dependent on expression of the csgD gene as previously described (48). The mean biofilm formation for ten U288 strains was not significantly different from that of three ST34 strains. However, considerable variation was observed especially for the U288 strains, and two strains of U288 had a statistically significant difference in biofilm formation compared to ST34 strain S04698-09, with U288 strain S01960-05 produced significantly less biomass and strain 10584-1997 significantly greater biomass.

S. Typhimurium U288 and monophasic S. Typhimurium ST34 isolates exhibit distinct
interactions with the host. We initially evaluated the interaction of representative strains of ST34 and U288 with tissue culture cells. No difference in the ability of U288 or ST34 isolates to invade human (T84) or porcine (IPEC-J2) epithelial cells in culture was observed ( Figure 6E and 6F). Although, several strains of U288 and ST34 had a small but significantly lower invasion compared to strain SL1344 in T84 cells, and two U288 isolates (S01960-09 and H091520254) exhibited decreased invasion of IPEC-J2 cells.
In an initial experiment to investigate the host-pathogen interactions of U288 and ST34 isolates, we used the streptomycin pre-treated C57bl/6 mouse that is commonly used as  Figure 3C). However, in each case, U288 strains colonised the mouse cecum to a greater level, induced higher levels of KC and iNOS transcripts and triggered a more severe pathology, compared to that resulting from infection with ST34 strains.
To compare the ability of six strains to colonise pigs, three U288 and three ST34 were modified by insertion of unique sequence tags in the chromosome to facilitate identification by sequencing, and four pigs were inoculated orally with an equal mixture of all six isolates.
Colonisation was investigated after 48 hours by sequencing cultured homogenates of faeces and tissue and enumeration of the sequence reads for the strain-specific tags ( Figure 7A).  Figure 7C). The mean Salmonella CFU present in faeces was also lower at all time points in pigs inoculated with U288, compared with ST34 ( Figure 7D). At 48 hours post-inoculation approximately 100-fold more ST34 than U288 were present in the faeces. Similar colonisation for each isolate was observed in the distal ileum and spiral colon, but U288 was present in higher numbers in the mesenteric and colonic lymph nodes ( Figure   7D). These data were consistent with the mixed inoculum experimental infections.

Discussion
The majority of S. Typhimurium U288 isolates from livestock and human infections were present in a phylogenetic clade that evolved from a common ancestor that was closely related to strain LT2. LT2 was isolated at Stoke Mandeville hospital in 1948 and has been used in a large number of studies on the genetics and biochemistry of S. Typhimurium (50). with U288, is not known (50). Evolution of successive descendants of the LT2-like hypothetical ancestor gave rise to multiple lineages evident from our analysis, including one that gave rise to the U288 epidemic clade in pigs. In our strain collection, non-U288 clade isolates that were also direct descendants of the LT2-like hypothetical ancestor, were isolated from human infections or from avian hosts (red lineages in Figure 2). Comparative genome analysis revealed that the evolution of U288 was characterized by the step-wise the acquisition of genes involved in resistance to multiple antibiotics, and the accumulation of genome sequence polymorphisms, some of which resulted in interruption of coding sequences.
The acquisition of resistance to antimicrobials is a key factor in the emergence of bacterial pathogens over the past 50 years , and is often the evolutionary event immediately preceding the spread and clonal expansion of a new clone (22,51). A U288 isolate was previously reported to encode antimicrobial resistance genes on two plasmids, a pSLT-like plasmid pU288-1 with tetA, sulII, strA and strB genes, and an incQ plasmid pU288-2 (45). Resistance genes on pU288-1 resulted from acquisition of two mobile genetic elements, bla TEM associated with an IS26 element, and sulIII, aadA, cmlA and dfrA associated with a class I integron in a composite Tn21-like element (45). The pU288-2 plasmid may have been acquired first as it is present in isolates from throughout the U288 clade. Insertions carrying AMR genes in pU288-1 may have occurred later than the acquisition of pU288-2, since a group of seven U288 isolates that form a more deeply rooted, basal clade, lacked the pU288-1-associated AMR genes. The majority of the U288 isolates in our analysis were direct descendants of the hypothetical ancestor that acquired AMR genes on pU288-1, and few from that of the basal clade that lacked these genes, suggesting pU288-1 evolution was an important event in the success of the U288 epidemic clade. However, despite a number of examples of apparent loss of AMR genes associated with both pU288-1 or loss of the pU288-2 plasmid (loss of AMR genes and the incQ replicon), there were just three isolates that had lost both concurrently, highlighting the importance of MDR.
Sequence polymorphisms in S. Typhimurium U288 strain S01960-05 resulted in truncated coding sequences affecting 26 genes, with reference to strain SL1344, a commonly used lab strain. Sixteen of these polymorphisms were predicted to have occurred before the hypothetical LT2-like ancestor of the U288 and related clades (red and green lineages in Figure 2). Six coding sequences (SL0337, assT3, yhbE, cutF, yciB and yciW) were truncated in all descendants of the hypothetical LT2-like ancestor, and one other (oadA) in all but one deeply rooted lineage. However, eight genes (pncA, yqaA, ybaA, sadA, ygbE, tsr, oatA and rcoR) were truncated in either the hypothetical ancestor of all U288 clade isolates, or subsequently in a subset of descendant lineages, in a stepwise manner. The first gene to be truncated was pncA, an event on an internal lineage on the phylogenetic tree that coincided with the acquisition of pU288-2, encoding AMR genes. Truncation of pncA is therefore, characteristic of the U288 clade. PncA is a nicotinamidase, a component of one of the pyridine nucleotide cycle (PNC) pathways, involved in the recycling nicotinamide adenine dinucleotide (NAD). The PncA dependent PNC pathway is probably more active in scavenging of pyridine compounds present in the environment (52). NAD is central to metabolism in all living systems, participating in over 300 enzymatic oxidation-reduction reactions (52), and therefore, conditions where de novo synthesis of NAD is limited by the availability of tryptophan or aspartate, the inability to use exogenous pyridines may limit metabolism. Perhaps significantly, the pncA gene is also truncated in S. Choleraesuis, a serotype that is highly host-adapted to pigs and replicates more slowly than S. Typhimurium in the pig intestinal mucosa (53). Chronologically, the next events were the truncation of the yqaA, ybaA, sadA and ygbE genes. SadA is a surface localised adhesin that contributes to cell-cell interactions and therefore multicellular behaviour (54). Truncation of tsr, that encodes a methyl accepting chemotaxis protein involved in energy taxis and colonisation of Peyer's patches in the murine model of infection, was acquired by a common ancestor of the majority of the U288 clade. This polymorphism occurred on an internal branch of the tree that coincided with the acquisition AMR genes inserted on pU288-1 and a clone that spread successfully through the pig population.
Accumulation of SNPs by isolates since the hypothetical LT2-like common ancestor was approximately linear facilitating estimation of dates for nodes within the phylogenetic tree.
This indicated that the truncation of the pncA gene, the earliest identified evolutionary event since the hypothetical LT2-like common ancestor specific to the U288 clade likely occurred in the 1970s, and culminated with the acquisition of AMR genes on pU288-1 and truncation of the tsr gene around 1995. The U288 clade may therefore have been evolving in the pig population before its first detection and rapid spread around the year 2003. The slow emergence of the U288 clade may have been associated with a gradual adaptation to a unique niche in the pig population, but it may also reflect a lag in time from emergence to detection by surveillance, as was proposed for other epidemics such as S. Enteritidis in poultry layer flocks, after the eradication of S. Gallinarum (55).
S. Typhimurium U288 and monophasic S. Typhimurium ST34 exhibited important differences in the way it interacted with the non-host environment, that could affect the likelihood that it survives in food and is transmitted on to consumers. Considerably more viable monophasic S. Typhimurium ST34 bacteria were recovered following desiccation for 24 hours, compared to U288. Many foodborne disease outbreaks due to Salmonella have been traced back to low moisture, ready-to-eat (RTE) foods (56), including dried pork products (57). Resistance to desiccation may be of particular significance because low oxygen tension is associated with increased resistance to a number of secondary stressors such as pH, salt, alcohol and heat (56). Monophasic S. Typhimurium ST34 also replicated at a significantly higher rate than U288 in culture, a characteristic that may result in a greater risk were it to apply in relevant food matrices. The infective dose of Salmonella enterica required to produce disease symptoms in humans? is in the range of 10 5 to 10 9 CFU, while naturally contaminated meat samples typically contained around 10 3 CFU/g (58), suggesting that replication in food may be important for transmission. Enterobacteriaceae replicate in food stored at 7 o C, increasing up to 10,000-fold in sausage meat in 2 weeks (59).
S. Typhimurium U288 and monophasic S. Typhimurium ST34 are associated with distinct risk to human health, despite circulating in the same pig population in the UK (3). Our data suggest that this may be due to the differences in tissue tropism and levels of each variant in the pig host that has the potential to impact the likelihood that Salmonella enters the food during the slaughter and butchering process. The greatest risk to Salmonella entering pork products is the contamination of the carcass at slaughter due to careless evisceration processes and inadequate cleaning of polishing machines (60). The relative contribution of Salmonella present in the faeces and tissues to contamination of food is not known, but increased colonisation of Salmonella in the caecum due to longer time in lairage correlated with contamination of the carcass at slaughter (61). Our data is consistent with a greater role for faecal contamination, since the mean viable counts of U288 isolates were up to two orders of magnitude lower in the faeces, but 10-fold greater in mesenteric lymph nodes, compared to monophasic S. Typhimurium ST34. The reason for the difference in colonisation of pigs is not known. However, monophasic S. Typhimurium ST34 lack the pSLT virulence plasmid, encoding the spv locus that is required for invasive disease in mice and severe gastroenteritis in cattle (62)(63)(64). Polymorphisms in some U288 strains also affect this locus. Acquisition of the IS26 element appears to have been accompanied by deletion of the spvR and spvA genes of pSLT, likely affecting expression of spvB and spvC that are pU288-1 also contained large regions of duplicated sequence of pSLT origin and sequence similar to other plasmids previously reported in Salmonella and E. coli, suggesting a complex history of horizontal gene transfer and deletion. The pU288-2 plasmid in U288 isolates acquired resistance genes in two distinct events, on an incQ plasmid pU288-2, and by acquisition of genes located on the pSLT-like plasmid pU288-1.
Taken together, our data contribute to a better understanding of the evolutionary history and phenotypes associated with the emergence of a new bacterial pathovariant. In the case of S.
Typhimurium U288, emergence in pigs appears to have been associated with a decreased risk to food safety for the consumption of pork or cross-contaminated food products by the human population. However, the consequences to the health and productivity of pigs as a result of a more invasive disease is not known, but may be an important consideration for the pork production industry. 3 1

SL1344 locus
Gene  truncation is indicated as insertion (ins), deletion (del) or non-sense codon (nsSNP) with reference to sequence (S04698-08). * Fitness score was derived from the screening of S. Typhimurium strain ST4/74 mutants by oral inoculation of pigs and recovery from the colon using transposondirected insertion site sequencing (46). The fitness score is the log 2 -fold change in the number of sequence reads across the boundaries of a transposon insertion in the gene between the input and output pools, after normalisation to account for variations in the total number of reads obtained for each sample. Zero indicates no change in relative abundance, positive values indicate mutants that are more abundant in the output pool and negative values suggest attenuation. Where multiple mutants for a given gene were screened, the mean score is presented with the number of mutants shown.