Stepwise evolution of Salmonella Typhimurium ST313 causing bloodstream infection in Africa

Bloodstream infections caused by nontyphoidal Salmonella are a major public health concern in Africa, causing ~49,600 deaths every year. The most common Salmonella enterica pathovariant associated with invasive nontyphoidal Salmonella disease is Salmonella Typhimurium sequence type (ST)313. It has been proposed that antimicrobial resistance and genome degradation has contributed to the success of ST313 lineages in Africa, but the evolutionary trajectory of such changes was unclear. Here, to define the evolutionary dynamics of ST313, we sub-sampled from two comprehensive collections of Salmonella isolates from African patients with bloodstream infections, spanning 1966 to 2018. The resulting 680 genome sequences led to the discovery of a pan-susceptible ST313 lineage (ST313 L3), which emerged in Malawi in 2016 and is closely related to ST313 variants that cause gastrointestinal disease in the United Kingdom and Brazil. Genomic analysis revealed degradation events in important virulence genes in ST313 L3, which had not occurred in other ST313 lineages. Despite arising only recently in the clinic, ST313 L3 is a phylogenetic intermediate between ST313 L1 and L2, with a characteristic accessory genome. Our in-depth genotypic and phenotypic characterization identifies the crucial loss-of-function genetic events that occurred during the stepwise evolution of invasive S. Typhimurium across Africa.

Although a causal relationship is unproven, inactivation of ratB is suggested to have reduced the enteric potential of ST313, resulting in more systemic infections 9 .

katE (STM1318)
Stationary phase catalase involved in protecting high density bacterial communities in the environment from oxidative stress.
Reduction in catalase activity across all ST313 2, 10 . We found that ST313 L3 had lower catalase activity than ST19 (this study). L-tartaric acid and dihydroxyacetone cannot be used as sole carbon sources by ST313. Results were originally generated with Biolog phenotype microarrays 9 . We found that ST313 L3 was also unable to grow on tartaric acid as a sole carbon source (this study).
ST313 L2: F311L ST313 L3: F311L An additional synonymous SNP was also identified at base position 462 in ST313 L3. ST313 L2 cannot grow on melibiose as a sole carbon source 11 . Assessment of the alpha-galactosidase activity of ST19 and ST313 wild-types and mutant strains showed that melR controlled the melibiose utilisation system 12 . We found that ST313 L3 was unable to grow on melibiose as a sole carbon source (this study). Although a causal relationship is unproven, the pipD gene contributes to macrophage persistence in murine models 9,13 . Of further interest, the Salmonella pathogenicity island SPI-5 gene sopB encodes a SPI-1 effector and has an associated chaperone pipC. In ST19 strains, these genes are upregulated in the gastrointestinal tract, which facilitates invasion into epithelial cells. However, in ST313 strains sopB and pipC are also upregulated in SPI-2 media and macrophages 12 . It is possible that this regulatory change is due to the loss of orfX, STM1093 and the 3' end of pipD.

bcsG (STM3624)
Cellulose biosynthetic enzyme involved in biofilm formation. Studies in mice have demonstrated that an accumulation of SNPs within sseI in ST313 L2 leads to an increased ability of ST313 L2 bacteria to disseminate rapidly from the gut to the draining lymph nodes in a murine model 14 .
In D23580, lpxO is a pseudogene due to the presence of one SNP that introduces an early stop codon 3 . By mass spectrometry, we confirmed that the lipid A of D23580 was not modified by LpxO during growth in InSPI2 medium, while the lipid A of 4/74 contained this modification when grown in this medium. To confirm that LpxO-mediated modification of lipid A, we established that the mass spectrum of the lipid A from a 4/74 ΔlpxO::aph (KmR) mutant grown in InSPI2 was similar to the spectrum of D23580 grown in the same medium.
MacAB is involved in oxidative stress resistance 15 and mediates resistance to macrolide antibiotics 17 . Recently, ST313 L2associated variants of the MacAB-TolC tripartite efflux pump have been shown to affect replication in macrophages, and influence fitness during colonisation of the murine gastrointestinal tract 16 .
Supplementary Discussion 1: pAnkS characterisation 3,18,19 Of interest we identified an 8,274 bp novel plasmid in ST313 L1 which shares 99.27% sequence identity with plasmid pAnkS 31  To understand the role of the novel plasmid repertoire associated with ST313 L1, we performed functional experiments. In E. coli, the LsoA endoribonuclease acts as a toxin involved in resistance to the Dmd negative mutant of the E. coli bacteriophage (phage) T4 (Myoviridae family) 18,19 . This observation suggests a broader role of the LsoAB TA system in phage immunity and we investigated the role of pAnkS ST313-L1 in Salmonella phage exclusion. We transferred pAnkS ST313-L1 from S. Typhimurium ST313 L1 strain A130 3 into S. Typhimurium strain LT2 (ST19), a strain commonly used in phage biology. We used a double-layer plaque assay to assess the phage susceptibility of strain LT2, in comparison with LT2 carrying pAnkS ST313-L1 . We assessed the phage susceptibility of LT2 with and without pAnkS ST313-L1 using a range of temperate and virulent phages. We discovered that pAnkS ST313-L1 conferred resistance to the virulent phages Det7 (Akermannviridae family) and S16 (Myoviridae family). Specifically, the phage efficiency of plating and plaque size were drastically reduced by the presence of pAnkS ST313-L1 (Supplementary Discussion 1 Figure   1). To confirm the involvement of LsoAB in the phage exclusion phenotype, we excised the lsoAB from pAnkS ST313-L1 , generating the plasmid pAnkS ST313-L1 ∆lsoAB. As expected, pAnkS ST313-L1 ∆lsoAB did not confer resistance to phages Det7 and S16 in LT2 (Supplementary Discussion 1, Figure 1). Taken together, these results demonstrate a protective role of pAnkS ST313-L1 against phage predation in S. Typhimurium ST313 L1.
This ability to resist phage killing could enhance survival of ST313 L1 in certain environmental niches. To determine AMR trends, we examined the 680 S. Typhimurium isolates for genomic determinants of resistance. To confirm genome-based predictions experimentally, we calculated the sensitivity and specificity of our genome predictions phenotypically using 528 available isolates ( Sub-sampling of a comprehensive archive allowed us to capture the full range of AMR profiles in S. Combining the available details on antimicrobial usage with fluctuating AMR profiles allows us to speculate on trends in circulating Salmonella lineages, particularly the switch from MDR ST313 L2 to the pan-susceptible ST313 L3. We hypothesise that the phased removal of chloramphenicol from clinical practice following local policy changes has opened a window of opportunity for the emergence of fully susceptible ST313 L3 as a clinical problem in Malawi. A comprehensive epidemiological study on antimicrobial usage would be required to properly investigate this hypothesis.

Supplementary
Recently, an XDR ST313 L2 sub-lineage was identified in the Democratic republic of Congo 6 . We did not identify any isolates displaying an XDR genotype in our collection. However, we did identify five isolates that have a single gyrA mutation associated with reduced susceptibility to ciprofloxacin, which were ST313 L2 strains isolated in Malawi (n=3)

Catalase assay
To compare catalase activity of ST313 lineage-representatives, a catalase assay was performed based on methods outlined previously 2,10 . Briefly, 10 μL of 6% aqueous H2O2 was added to 1 cm diameter glass test tubes containing 1ml of bacterial overnight cultures grown in LB (Lennox). Tubes were photographed after 5 minutes incubation at room temperature and bubble height was compared.

Melibiose and tartrate utilisation
To determine melibiose and tartrate utilisation as sole carbon sources, a Salmonella culture was grown overnight in LB (Lennox) broth at 37°C. A 1 mL aliquot was centrifuged at 13,000 xg for 5 min and the supernatant discarded. The cell pellet was washed twice with 1 mL of PBS and streaked on minimal were also supplemented with 20 mM Trimethylamine N-oxide dihydrate (TMAO) as an electron acceptor, and incubated for 5 days at 30°C under anaerobic conditions using gas packs (AnaeroGen 2.5 L sachets, Thermo Scientific) and a resazurin indication (Thermo Scientific, Oxoid).

RDAR morphotype
Red Dry and Rough (RDAR) phenotype was assessed to determine production of extracellular matrix components curli and cellulose 10,28 . A Salmonella culture was grown overnight in LB (Lennox) broth at 37°C, and an aliquot of 10 μL was dropped onto LB agar without NaCl, supplemented with 40 μg/mL Congo red and 20 μg/mL Coomassie Brilliant Blue. Plates were incubated at room temperature for 7 days.

Bacterial growth and sample preparation for lipid A analysis by mass spectrometry
Bacteria were grown in LB (Lennox) broth or SPI-2-inducing (InSPI2) growth condition at 37°C 220 rpm in a water bath. For growth in Lennox broth, two growth conditions were tested: early stationary phase (ESP), growth to OD600nm 2; and late stationary phase (LSP), growth to OD600nm 2 followed by a further 6 h growth in the same incubation conditions. Lennox broth consisted in 10 g/L tryptone, 5 g/L yeast extract, and 5 g/L NaCl. PCN medium. The InSPI2 growth condition involved growth in phosphate carbon nitrogen (PCN) minimal medium 32 at pH 5.8 and 0.4 mM Pi to OD600nm 0.3 33 . Pellets from bacterial cultures were washed twice with phosphate buffer (10.14 mM Na2HPO4, 1.76 mM KH2PO4, adjusted to pH 7.4) and lyophilized.

Mass Spectrometry Analysis
Lipid A molecules were extracted using an ammonium hydroxide-isobutyric acid method 24   to blunt-end the BglII cohesive-ends. After purification, the fragment was self-ligated, using T4 DNA ligase and the ligation reaction was transformed into E. coli Top10. After selection on ampicillin, the resulting plasmid pAnkS ST313-L1 ∆lsoAB (6,719 bp) was extracted and verified by restriction analysis with BamHI. Plasmids pAnkS ST313-L1 and pAnkS ST313-L1 ∆lsoAB were transferred by electroporation 26

Bacteriophage manipulation and double-layer plaque assay
All the bacteriophage (phage) stocks were prepared in LB (Lennox) with the prophage-cured strain S.