Salmonella enterica serovar Typhimurium ST313 sublineage 2.2 has emerged in Malawi with a characteristic gene expression signature and a fitness advantage

Abstract Invasive non-typhoidal Salmonella (iNTS) disease is a serious bloodstream infection that targets immune-compromised individuals, and causes significant mortality in sub-Saharan Africa. Salmonella enterica serovar Typhimurium ST313 causes the majority of iNTS in Malawi. We performed an intensive comparative genomic analysis of 608 S. Typhimurium ST313 isolates dating between 1996 and 2018 from Blantyre, Malawi. We discovered that following the arrival of the well-characterized S. Typhimurium ST313 lineage 2 in 1999, two multidrug-resistant variants emerged in Malawi in 2006 and 2008, designated sublineages 2.2 and 2.3, respectively. The majority of S. Typhimurium isolates from human bloodstream infections in Malawi now belong to sublineages 2.2 or 2.3. To understand the emergence of the prevalent ST313 sublineage 2.2, we studied two representative strains, D23580 (lineage 2) and D37712 (sublineage 2.2). The chromosome of ST313 lineage 2 and sublineage 2.2 only differed by 29 SNPs/small indels and a 3 kb deletion of a Gifsy-2 prophage region including the sseI pseudogene. Lineage 2 and sublineage 2.2 had distinctive plasmid profiles. The transcriptome was investigated in 15 infection-relevant in vitro conditions and within macrophages. During growth in physiological conditions that do not usually trigger S. Typhimurium SPI2 gene expression, the SPI2 genes of D37712 were transcriptionally active. We identified down-regulation of flagellar genes in D37712 compared with D23580. Following phenotypic confirmation of transcriptomic differences, we discovered that sublineage 2.2 had increased fitness compared with lineage 2 during mixed growth in minimal media. We speculate that this competitive advantage is contributing to the emergence of sublineage 2.2 in Malawi.


Introduction
Non-typhoidal Salmonella (NTS) is a k e y bacterial pathogen that thr eatens people acr oss the world.Typhim urium and Enteritidis are the two serovars of Salmonella enterica responsible for the highest le v els of self-limiting gastr ointestinal disease in Eur ope, the USA, and other high-income countries (Zhang et al. 2003 ).In the industrialized world, NTS has been associated with intensive food production, animal husbandry, and global distribution systems (Majowicz et al. 2010 ).The S. Typhimurium sequence types r esponsible for gastr oenteritis globall y include ST19, ST34, and monophasic 1,4, [5],12: i:-variants (Branchu et al. 2018 ).The diarrhoeal NTS disease is termed dNTS, and is mainly foodborne, posing a significant burden to public health with ∼153 million cases and 57 000 deaths per ann um world wide (Kirk et al. 2015, Chirwa et al. 2023 ).
In contrast, a lethal systemic disease called invasive nontyphoidal Salmonella (iNTS) disease has emerged in recent decades in low-and middle-income countries in sub-Saharan Africa.iNTS tar gets imm unocompr omised individuals suc h as adults with HIV, and c hildr en under 5 years of age with malaria, malnutrition or se v er e anaemia (Feasey et al. 2012 ).In some countries of sub-Saharan Africa, Salmonella causes more cases of communityonset bloodstream infections than any other bacterial pathogen (Marchello et al. 2019 ).In 2017, 535 000 cases of iNTS disease were estimated worldwide, with ∼80% of cases and 77 000 deaths occurring in sub-Saharan Africa (Stana wa y et al. 2019 ).
Clinicall y, the tr eatment of iNTS is complicated by multi-drug r esistance (MDR) whic h limits ther a peutic options (Crump et al. 2015 ).Widespr ead r esistance of iNTS pathogens to first-line drugs such as chloramphenicol, ampicillin, and cotrimoxazole has been seen in many countries (Kariuki et al. 2006 ).This MDR phenotype may be one of the reasons the case fatality rate associated with iNTS is amongst the highest in comparison to any infectious disease (15%) (Marchello et al. 2022).Resistance to second-line drugs suc h as ceftriaxone, cipr ofloxacin, and azithr omycin has been reported in a few African countries (Tack et al. 2020 ).Clearly, the challenge posed by MDR Salmonella must be addressed urgently (Gilchrist and MacLennan 2019 ).
The African iNTS epidemic is mainly caused by two Salmonella patho variants , S. Typhimurium sequence type 313 (ST313) and specific clades of S. Enteritidis (Kingsley et al. 2009, Ok or o et al. 2012, 2015, Feasey et al. 2016 ). S. Typhimurium ST313 is responsible for about tw o-thir ds of clinical iNTS cases that have been reported in Africa (Gilchrist and MacLennan 2019 ).
It is not certain how these pathogens ar e tr ansmitted, but ther e is incr easing e vidence fr om case-contr ol studies that ST313 str ains ar e human-associated but not animal-associated within households (Post et al. 2019, Koolman et al. 2022 ).A recent summary concludes that the available data are consistent with iNTS disease being transmitted person-to-person (Chirwa et al. 2023 ).Global efforts to combat iNTS infections are currently focused on vaccine development, which has now progressed to Phase 1 clinical trials (Piccini andMontomoli 2020 , Skidmore et al. 2023 ).
Since 1998, continuous sentinel surveillance for fever and bloodstream infections among adults and children has been undertaken at Queen Elizabeth Central Hospital (QECH).This tertiary r eferr al hospital in Blantyr e , Mala wi, serves an urban population of ∼920 000 with a high incidence of malaria, HIV, and malnutrition (Musicha et al. 2017 ).Following blood culture of samples collected from patients of all ages presenting with fever, whole genome sequencing identified the ST313 variant of S. Typhimurium (Kingsley et al. 2009 ).Phylogenetic analysis revealed that the c hlor amphenicol-sensitiv e ST313 lineage 1 was clonally replaced in Malawi by the chloramphenicol-resistant lineage 2 (Ok or o et al. 2012 ).Mor e r ecentl y, an ST313 sublinea ge II.1 (2.1) emer ged fr om linea ge 2 in the Democr atic Republic of Congo (DRC) in Centr al Africa.Sublinea ge 2.1 had alter ed phenotypic properties including biofilm formation and metabolic capacity and resistance to azithromycin (Van Puyvelde et al. 2019 ).An elegant genomic analysis that provides insight into the diversity of S. Typhimurium ST19 clades in the context of ST313 lineage 2 clades is also available (Van Puyvelde et al. 2023 ).
The initial suggestion that ST313 lineage 2 was undergoing evolutionary change in East Africa came from a small study that identified se v er al S. Typhim urium ST313 Mala wian isolates , dated between 2006 and 2008, that differed from lineage 2 by 22 coregenome single nucleotide pol ymor phisms (SNPs) (Msefula et al. 2012 ).
To examine the e volutionary tr ajectory of S. Typhimurium in Malawi at a large scale, we conducted a compar ativ e genomic analysis study focused on 680 isolates dating between 1996and 2018(Pulford et al. 2021 ) ).We pr e viousl y r eported that ST313 lineage 1 (L1) was replaced by lineage 2 (here designated L2.0), and discov er ed an antibiotic-sensitive lineage 3 (L3) that emerged in 2016 (Pulford et al. 2021 ).We have now performed a more intensive phylogenetic analysis of the same collection of S. Typhimurium ST313 isolates, most of which caused bloodstream infections in Malawi over two decades.We discov er ed two novel sub-lineages named 2.2 (L2.2) and 2.3 (L2.3) that emerged in 2006-2008, and have been replacing L2.0.
Here we present a comprehensive comparative genomic analysis of the most pr e v alent ST313 L2.2 sublineage and report the results of a functional genomic a ppr oac h that identified k e y phenotypic c har acteristics that distinguish L2.2 fr om L2.0.

Bacterial strains
To investigate the evolutionary dynamics of S. Typhimurium ST313 L2 in Malawi over a 22-year period, we focused on the large collection of 8 000 S. Typhimurium isolates derived from bloodstream infection in hospitalized patients at the QECH, Blantyre, Mala wi (F easey et al. 2015 ).The collection was assembled by the Malawi-Liv er pool-Wellcome Trust Clinical Research Programme (MLW) between 1996 and 2018; the precise annual numbers of isolates are shown in Fig. 1 C.A random sub-sampling strategy was used to select 608 isolates for whole-genome sequencing, which included 549 S. Typhimurium ST313 isolates (Pulford et al. 2021 ).
The two S. Typhimurium ST313 strains that are the focus of this study are D23580 and D37712.D23580 was isolated from a Malawian 26-month-old child with malaria and anaemia in 2004.D37712 was isolated from the blood of an HIV-positive Malawian male child in 2006.These two African Salmonella str ains hav e been deposited in the National Collection of Type Cultures (NCTC).The D23580 (lineage 2.0) strain is available as NCTC 14677.The ST313 sublinea ge 2.2 str ain D37712 is av ailable as NCTC 14678.All bacterial strains are detailed in Table S8 .

Genome sequencing
The assembled genome and annotation of D23580 (Canals et al. 2019b, Kingsley et al. 2009 ) (L2.0) was obtained from the European Nucleotide Archive (ENA) repository (EMBL-EBI) under accession PRJEB28511 ( https:// www.ebi.ac.uk/ ena/ data/ view/ PRJEB28511 ).For genome sequencing of D37712 (L2.2),DN A w as extracted using the Bioline mini kit, and quality was assessed using gel electr ophor esis (0.5% a gar ose gel, at 30 volts for 18 h).The genome was generated by a combination of long-read sequencing with a PacBio RS II and short-read sequencing on an Illumina HiSeq machine at the Center for Genome Researc h, Univ ersity of Liv er pool, United Kingdom.
The assembled genome of S. Typhimurium SDT313 L2.2 strain D37712 was deposited in Genbank (GCA_014250335.1,assembly ASM1425033v1).Raw sequencing reads were deposited for both PacBio and Illumina, under BioProject ID PRJNA656698.Sequence Read Arc hiv e (SRA) database IDs are SRR12444880 for Illumina and SRR12444881 for PacBio.
The assembled genome and annotation of S .Typhimurium ST19 r epr esentativ e str ain 4/74 (Ric har dson et al. 2011 ) w ere obtained from GenBank (Accession number GCF_000188735.1), while the raw sequencing data of 27 S .Typhimurium ST313 strains described in a pr e vious study (Msefula et al. 2012 ) were downloaded from the EMBL-EBI database ( https://www.ebi.ac.uk , accession number ERA015722).T he ra w reads were assembled using Unicycler v0.4.8 (Wick et al. 2017 ).The quality of the assemblies was assessed by Quast v5.0.2 (Gur e vic h et al. 2013 ).The N50 value of all assemblies was > 20 kb, and the number of contigs was < 600.
To identify SNPs, Snippy v4.4.0 ( https:// github.com/tseemann/ snipp y ) w as used to ma p the r aw r eads a gainst the 4/74 genome.To detect pseudogene-associated SNPs/indels in each sub-lineage, the SNPs/indels that caused nonsense or frameshifted mutations wer e filter ed.The identifications and names of the disrupted genes were summarized, then the wild-type gene sequences were extr acted fr om the 4/74 genome.To v alidate the pseudogeneassociated SNPs/indels, the wild-type gene sequences were used to make a BLAST database with BLAST 2.9.0 + (Camacho et al. 2009 ).The 29 genome assemblies were queried against the databases, using the BLASTn algorithm to confirm the nonsense and frameshifted mutations in all isolates.

Phylogenetic analysis of African Salmonella Typhimurium isolates dating from 1966 to 2018
To examine the ov er all population structur e of Salmonella Typhim urium r esponsible for blood infection in Malawi (Fig. 1 A,B and Fig. S1 ), the r aw r eads of 707 published genome sequences were downloaded ( T able S7 ).T rimmomatic v0.36 (Bolger et al. 2014 ) was used to trim adapters and Seqtk v1.2-r94 ( https:// github.com/lh3/ seqtk ) was used to trim low-quality regions using the trimfq flag.Fastqc v0.11.5 ( https:// www.bioinformatics.babraham.ac.uk/ projects/ fastqc/ ) and multiqc v1.0 ( http://multiqc.info ) were used to pass sequence reads according to the following criteria: passed basic quality statistics, per base sequence quality, per base N content, adapter content and an av er a ge GC content of between 47% and 57%.Only high-quality reads were used in the downstream analysis.Sequence reads were aligned to the S .Typhimurium D23580 genome using Snippy v4.4.0 with parameter '--mincov 5'.The recombination sites of the alignment were removed by Gubbins (Cr ouc her et al. 2015 ), and the phylogenetic tree was built with Raxml-ng (K ozlo v et al. 2019 ) using GTR_G models ad 100 bootstr a ps .T he tree was rooted on Salmonella Typhi strain CT18 (GCA_000195995.1)as the outgr oup.The tr ee was visualized with the iToL online tool (Letunic and Bork 2006 ).The sub-lineages were identified with rHierBAPS (Tonkin-Hill et al. 2018 ).The stac ked-ar ea c hart and the bar c hart sho wing the per centage and number of isolates from each sub-lineage were made in MS Excel.

RN A purifica tion and gro wth conditions
Initiall y, a scr een of tr anscriptomic gene expr ession was performed without biological replicates.Total RN A w as purified using TRIzol fr om S .Typhim urium D37712 grown in 15 different conditions as described pr e viousl y (Kröger et al. 2013 ).To gener ate statisticall y r obust gene expr ession pr ofiles, total RNA was subsequently purified using TRIzol fr om S .Typhim urium D37712 grown in four in vitro growth conditions (ESP, anaerobic growth, NonSPI2, and InSPI2) with three biological replicates as described pr e viousl y (Kröger et al. 2013 ).RN A w as isolated fr om intr a-macr opha ge D37712 following infection of RAW264.7 m urine macr opha ges using our published pr otocol (Srikumar et al. 2015 ).

RNA-seq of S . Typhimurium strain D37712 using Illumina technology
For tr anscriptomic anal yses, cDN A samples w er e pr epar ed fr om S .Typhimurium RN A b y Vertis Biotechnologie AG (F reising, Germany).RN A w as first treated with DNase and purified using the Agencourt RNAClean XP kit (Beckman Coulter Genomics).RNA samples were sheared using ultrasound, treated with antarctic phosphatase and re-phosphorylated with T4 polynucleotide kinase.RNA fr a gments wer e pol y(A)-tailed using pol y(A) pol ymerase and an RNA adapter was ligated to the 5 -phosphate of the RNA.First-strand cDNA synthesis was performed using an oligo(dT)-adapter primer and M-MLV reverse transcriptase .T he resulting cDN A w as PCR-amplified to ∼10-20 ng/ μl.The cDN A w as purified using the Agencourt AMPure XP kit.The cDNA samples were pooled using equimolar amounts and size fractionated in the size range of 200-500 bp using pr epar ativ e a gar ose gels .T he cDN A pool w as sequenced on an Illumina NextSeq 500 system using a 75 bp read length.
For the biological replicates of the four growth conditions (ESP, anaer obic gr owth (abbr e viated as NoO2), NonSPI2, and InSPI2) and the intr a-macr opha ge RN A, cDN A samples w er e gener ated as above with some improvements in library preparation.First, after fr a gmentation with ultrasound, an oligonucleotide adapter was ligated to the 3 end of the RNA molecules.Second, first-strand cDN A synthesis w as performed using M-MLV r e v erse tr anscriptase and the 3 adapter as primer, and, after purification, the 5 Illumina TruSeq sequencing adapter was ligated to the 3 end of the antisense cDNA.Sequencing of the cDN A w as performed as described abo ve .All ra w sequencing reads were deposited to the Gene Expression Omnibus (GEO) database under accession GSE161403.

RNA-seq and dRNA-seq read processing and visualization
RNA-seq data from S .Typhimurium 4/74 and D23580 were extr acted fr om pr e viousl y published experiments (Canals et al. 2019b, Kröger et al. 2013, Srikumar et al. 2015 ;GEO dataset GSE119724).A combined r efer ence genome was generated that contained the D23580 c hr omosome plus plasmids pBT1, pBT2, pBT3, pSLT-BT (from D23580) and the D37712 plasmid pCol1B9 D37712 .All reads were aligned and quantified using Bacpipe v0.8a ( https:// github.com/a pr edeus/m ulti-bacpipe ).Briefly, basic read quality control was performed with FastQC v0.11.8.RNA-seq reads were aligned to the genome sequence using STAR v2.6.0c using '-alignIntronMin 20 -alignIntronMax 19 -outFilterMultimapNmax 20' options.A combined GFF file was generated by Bacpipe, where all features of interest were listed as a 'gene', with each gene identified by a D37712 locus tag.Subsequentl y, r ead counting was done by featureCounts v1.6.4,using options '-O -M -fraction -t gene -g ID -s 1'.For visualization, scaled gedGr a ph files were generated using bedtools genomecov with a scaling coefficient of 10 9 /(number of aligned bases), separ atel y for sense and antisense DNA str ands.Bedgr a ph files were converted to bigWig using bedGr a phToBigWig utility ( http://hgdownload.soe.ucsc.edu/admin/ exe/ linux.x86_ 64/ ).Cov er a ge tr ac ks, annotation, and genome sequence were visualized using JBrowse v1.16.6.TPM were calculated for all samples and used as absolute expression values ( Table S5 ).A conservative cut-off was used to distinguish between expressed (TPM > 10) and not expressed (TPM ≤ 10), as we pr e viousl y described (Kröger et al. 2013 ).Relative expr ession v alues wer e calculated by dividing the TPM value for one condition in one strain by the TPM value for the same condition in a different strain.Before the calculation, all TPM values below 10 were set up to 10.A conservative fold-change cut-off of 3 was used to highlight differences in expression between strains.

Differential gene expression analysis with multiple biological replicates
For differential expression analysis of S .Typhimurium strains 4/74, D23580, and D37712, the raw counts ( Table S4 ) from 3 to 5 biological replicates in four growth conditions were used (ESP, anaer obic gr owth (abbr e viated as NoO2), NonSPI2, and InSPI2).Differ ential expr ession anal ysis was done using DESeq2 v1.24.0 with default settings.A gene was considered to be differ entiall y expressed if the absolute value of its log2 fold change was at least 1 (i.e.fold change > 2), and the adjusted P value was < .001.

The SalComD37712 community data resource, and the associated Jbrowse genome browser
SalCom provides a user-friendly Web interface that allows the visualization and comparison of gene expression values across multiple conditions and between str ains.P articular genes can be selected thr ough pr e-defined lists of inter est, suc h as all sRNAs or all genes belonging to a specific pathogenicity island.The resulting heatmap-style display highlights expression differences and provides access to the ric h, manuall y cur ated annotation of strains D37712 and D23580.The actual values behind the display can be downloaded for further processing, and a link connects the curr ent vie w to a genome br owser interface.
Visualization of all the RNA-seq and dRNA-seq (TSS) cov er a ge tr ac ks in JBr o wse 1.16.6 sho ws sequence r eads ma pped a gainst the combined r efer ence genome described abo ve .Overall, the genomic distance between strains 4/74 and D23580 ( ∼1000 SNPs, or ∼1 SNP per 5000 nucleotides), and between D37712 and D23580 ( ∼30 SNPs, ∼1 SNP per 150 000 nucleotides) allo w ed the alignment of RNA-seq reads to the simplified combined r efer ence genome without significant loss of reads .T he combined reference genome facilitated a direct comparison of gene cov er a ge as well as TSS.The unified browser is hosted at http:// hintonlab.com/jbrowse/ index.html?data=Combo_ D37/data .

Phenotypic and mixed competiti v e growth experiments
T he s wimming motility of S. Typhimurium strains D37712, D23580 and 4/74 was determined by a plate assay (Canals et al. 2019b ), whic h involv ed spotting 3 μl ov ernight cultur e onto 0.3% LB a gar.Relative motility of the three strains was assessed by migration diameter after 4 h and 8 h of incubation at 37 • C.
Relativ e expr ession of the ssaG SPI2 promoter in strains D23580 and D37712 was measured at the single cell le v el via GFP fluorescence.Following the construction of a kanam ycin-sensiti ve deri vati ve of D23580 (strain JH4235), a P ssaG::gfp + transcriptional fusion was incor por ated into the c hr omosome of JH4235 and D37712 by inserting the gfp + gene downstream of the ssaG gene, under the control of the P ssaG promoter.The P ssaG::gfp + D23580 deri vati ve (JH4692), and the P ssaG::gfp + D37712 deri vati ve (JH4693) are listed in Table S8 .
The strains JH4692 and JH4693 were genome sequenced to confirm the integrity of the transcriptional fusions, and to verify that unintended nucleotide changes had not arisen.Following growth in 25 ml non-inducing NonSPI2 media in a 250 ml flask at 37 • C with shaking at 220 rpm for ∼8 h until OD 600 = 0.3, fluorescence was determined with a BD FACSAria Flow Cytometer.The relativ e fluor escence of the two str ains JH4692 and JH4693, and the numbers of individual fluorescent bacteria that expressed the P ssaG::gfp + pr omoter, wer e determined with FlowJo VX software.
The r elativ e fitness of S. Typhim urium str ains D37712 and D23580 was assessed in two independent mixed-growth experiments.First, kanam ycin-resistant deri vati ves of each strain were constructed by inserting the aph kanamycin resistance gene into the c hr omosome at the inter genic r egion between the STM4196 and STM4197 genes , a region that we have previously shown to be tr anscriptionall y silent (Canals et al. 2019b ).The str ains wer e designated D23580::Km R JH3794 and D37712::Km R , JH4232.Mixed cultures of wild-type or kanamycin-resistant deriv ativ es of eac h str ain wer e gr own ov ernight in LB, InSPI2, and NonSPI2 media in a 250 ml flask at 37 • C with shaking at 220 rpm.Following plating on LB agar or LB + kanamycin, colonies were counted and the ratio of bacterial strains was determined.To confirm that the insertion of kanamycin resistance at the intergenic region between STM4196 and STM4197 did not impact upon fitness, a mixed-growth experiment was done in both LB and NonSPI2 media ( Fig. S7 ).

Identification of S. Typhimurium ST313 sublineages 2.2 and 2.3 in Malawi
The emergence of the ST313 lineage 2 genotype in Malawi in 2002 prompted us to hypothesize that subsequent evolution would select variants with increased fitness, leading to the clonal expansion of one or more sublineages by outcompeting pr e viousl y dominant genotypes .We in vestigated this hypothesis by conducting a detailed core-gene SNP-based maximum likelihood (ML) phylogenetic analysis to investigate the population structure of S. Typhimurium ST313 L2.0 ( Fig. S1 ).As well as identifying members of the antibiotic-sensitive lineage 3, reported previously (Pulford et al. 2021 ), we discov er ed that ST313 L2 comprised thr ee phylogeneticall y distinct sublinea ges that differ ed by a total of 39 SNPs.The S. Typhimurium ST313 reference strain D23580 (Kingsley et al. 2009 ) belongs to the ST313 L2.0 lineage (Fig. 1 A).As ST313 sublineage L2.1 had been defined previously (Van Puyvelde et al. 2019 ), the new sublineages which belonged to different hierBAPS level 2 clusters were designated L2.2 and L2.3 (Fig. 1 A and Fig. S1 ).In total, we identified 151 L2.2 isolates, 74 L2.3 isolates, and 350 L2.0 isolates.
In Blantyre , Mala wi, S. Typhimurium ST313 L2.2 was first detected in 2006, and L2.3 was initiall y observ ed in 2008 (Fig. 1 B,C).Both L2.2 and L2.3 increased in prevalence at the QECH in Blantyre in subsequent years.By 2018, L2.2 and L2.3 had lar gel y r eplaced L2.0 (Fig. 1  To understand the accessory gene complement of L2.2 and L2.3, we compared the genomes of se v en L2.2 isolates and four L2.3 isolates with 17 L2.0 isolates, ST313 L1 and ST19 and the results are shown in Fig. 1 A and Table S1 .S. Typhimurium strain D23580 is the r epr esentativ e str ain of L2.0 (Kingsley et al. 2009 ), for whic h we pr e viousl y used long-r ead sequencing and other a ppr oac hes to thor oughl y c har acterize the c hr omosomal and plasmid complement (Canals et al. 2019b ).

Antimicrobial resistance
MDR variants of S. Typhimurium with resistance to ampicillin and cotrimoxazole were detected at an early stage of the iNTS epidemic, from 1997 onw ar d (Gor don et al. 2008 ).Multidrug-resistant v ariants of S. Typhim urium ST313 that wer e no longer susceptible to c hlor amphenicol, ampicillin and cotrimoxazole subsequentl y emerged in Malawi (Gordon et al. 2008 ) and have been reported else wher e in sub-Sahar an Africa by the Global Enteric Multicenter Stud y (GEMS) stud y (Kasumba et al. 2021 ).The S. Typhimurium ST313 L2.0, L2.2, and L2.3 isolates shared the same MDR profiles (resistance to chloramphenicol, ampicillin, and cotrimoxazole), and carried identical IS21-associated antimicrobial gene cassettes within the pSLT-BT plasmid (Fig. 2 B).

Compar a ti v e genomics of S. Typhimurium ST313 sublineage 2.2
Because S. Typhimurium ST313 L2.2 was the predominant novel sublineage in Blantyre , Mala wi in 2018, we focused on L2.2 for the remainder of this study.We used the phylogeny (Fig. 1 A) to select strain D37712 as a r epr esentativ e of L2.2.D37712 was isolated from the blood of an HIV-positive Malawian male child and has been deposited in the National Collection of Type Cultures as NCTC14678.The draft genome sequence of D37712 was obtained in 2012 with Illumina technology, an assembly that comprised 27 individual contigs (Msefula et al. 2012 ).To generate a r efer encequality genome, we resequenced D37712 with both long-read PacBio and Illumina short-read technologies.Our hybrid strategy generated a complete genome assembly that included one circular c hr omosome and thr ee plasmids (see Materials & Methods; GenBank CP060165, CP060166, CP060167, and CP060168).This high-quality genome sequence allo w ed us to conduct a detailed comparison between the genomes of L2.2 strain D37712 and L2.0 strain D23580 (accession number FN424405), summarized in Fig. 2 and Table S2 .
Ov er all, the gene content of the two strains was lar gel y equivalent.The D23580 annotation contains 4823 protein-coding and pseudogenes and 287 small RN A (sRN A) genes that w e identified pr e viousl y (Canals et al. 2019b ), while D37712 contains 4821 protein-coding and pseudogenes and the same 287 sRNAs.In total, the D37712 and D23580 genomes shared 4729 orthol-ogous protein-coding genes and pseudogenes .T he 104 protein genes that differ are encoded by the pSLT D37712 , pBT1 D37712 , and pCol1B9 D37712 plasmids.

Comparison of D23580 and D37712 chromosomes
The detailed genomic comparison of D37712 with D23580 sho w ed that the sizes of the two c hr omosomes v aried by onl y 3342 bp.Ov er all, the onl y differ ences between the genomes of the L2.0 and L2.2 str ains wer e 26 c hr omosomal SNPs and small indels, plus one large deletion, and an inversion of the hin switch.In-depth annotation of the nucleotide variants identified three putative lossof-function mutations (two stop mutations, one frameshift insertion), one disruptiv e in-fr ame deletion, four synon ymous m utations, 13 missense m utations, and fiv e inter genic v ariants, summarized in Fig. 2 A. None of the SNP differences that distinguished D37712 from D23580 were located within 150 nucleotides of a Transcriptional Start Site (TSS) (Canals et al. 2019b ), and so would not be predicted to modulate gene expression.
The 3358 bp-long deletion of a Gifsy-2 pr opha geassociated region that spanned the sseI pseudogene of D23580 (STMMW_10 631) r emov ed two coding sequences (STM1050-51; STMMW_10611-STMMW_10 631), and substantially truncated the STM1049 (STMMW_10 601) gene (Fig. 2 E).The sseI gene encodes a cysteine hydrolase effector protein that modulates the directional migration of dendritic cells during systemic infection (Brink et al. 2018 ).In strain D23580, the insertion of an IS26 transposable element inactivated the sseI gene (Kingsley et al. 2009 ), causing increased dendritic cell-mediated dissemination of strain D23580 during infection (Carden et al. 2017 ).We used an independent Pol ymer ase Chain Reaction (PCR)-based a ppr oac h to confirm that the 3358 bp deletion had r emov ed the sseI gene fr om the c hr omosome of strain D37712 ( Fig. S2 ).
In summary, strain D37712 carried the pSLT-BT, pBT2, and pCol1B9 plasmids as detailed below.Both D23580 and D37712 strains carried a variant of the pSLT-BT virulence plasmid (Kingsley et al. 2009 ) that contains a Tn21-like transposable element with five antibiotic resistance genes .T he D37712 version of pSLT-BT onl y differs fr om the pSLT-BT of D23580 in two important t ways (Fig. 2 B).Firstly, the Tn21-like element is inserted in the opposite direction with regards to the rest of the plasmid, suggesting that the transposable element remains active.Secondly, three nucleotide v ariants wer e identified in the pSLT-BT carried by D37712, two deletions in noncoding regions, and one frameshift insertion that generates a pseudogene of spvD.
Plasmid pCol1B9 was of particular interest because it was absent from D23580, but was present in S. Typhimurium ST19 strain 4/74 (Richardson et al. 2011 , Fig. 1 A).4/74 is the parent of S. Typhimurium SL1344, a strain that has been used extensively for the study of S .Typhimurium pathogenesis and gene regulation in recent decades (Rankin andTaylor 1966 , Kröger et al. 2012 ).Our new annotation of the pCol1B9-like plasmid identified 95 distinct protein-coding genes, while the previously published annotation of pCol1B9 4/74 assigned 101 protein-coding genes.Some of these r epr esent annotation discr epancies, while others r epr esent true genetic differences ( Fig. S3 ).
Following careful examination, we identified 14 genes unique to pCol1B9 D37712 , and 20 genes unique to pCol1B9 4/74 .There were 81 genes carried by both plasmids.Inter estingl y, pCol1B9 D37712 lacked the colicin toxin-antitoxin system that both gave pCol1B9 its name, and provides Salmonella with a competitive adv anta ge in the gut (Nedialk ov a et al. 2014 ).The pCol1B9 D37712 plasmid carried a locus that was absent from pCol1B9 4/74 , namely the impC-umuCD operon ( Fig. S3 ) which encodes the error-prone DNA polymerase V responsible for the increased mutation rate linked to the SOS str ess r esponse in E. coli (Sikand et al. 2021 ).

Comparison of pseudogene status of D23580 and D37712
Our compar ativ e genomic anal ysis focused on the pseudogenes found in strains 4/74, D23580, and D37712 (Fig. 2 F and Table S3 ).The pseudogenization of se v er al D23580 genes, compared with strain 4/74, has been linked to the invasive phenotype of African Salmonella ST313 (Kingsley et al. 2009 ).We found that the pseudogene complement of D23580 was lar gel y conserv ed in D37712, consistent with inheritance from a common ancestor.We have recentl y r eported the r ole of the MacAB-TolC macr olide efflux pump in the virulence of S. Typhimurium ST313 and sho w ed experimentally that macB was an inactive pseudogene in D23580 (Honeycutt et al. 2020 ).Inter estingl y, the macB gene is functional in D37712.Compar ed with D23580, thr ee additional D37712 genes were pseudogenized ( spvD, yadE, and STMMW_42 692, as detailed in Table S3 ).YadE is a predicted polysaccharide deacetylase lipoprotein.The functional impact of these pseudogenes on L2.2 remains to be established.
Ov er all the c hr omosomes of ST313 lineage 2 and sublineage 2.2 wer e highl y conserv ed and differ ed by just 29 SNPs/small indels, and a 3-kb deletion in the Gifsy-2 pr opha ge r egion.The ST313 lineage 2 and sublineage 2.2 have distinct plasmid profiles.

Transcriptional landscape of S. Typhimurium ST313 sublineage L2.2
Pr e viousl y, we c har acterized the primary tr anscriptome of two other S. Typhimurium strains, 4/74 and D23580, using a combination of multi-condition RNA-seq and differential RNA-seq (dRNAseq) techniques (Canals et al. 2019b, Kröger et al. 2013 ).To identify the TSS of strain D37712, we analysed a pooled sample containing RNA from 15 in vitro conditions by dRNA-seq and RNA-seq as detailed pr e viousl y (Kröger et al. 2013 ).The high similarity between the D23580 and D37712 c hr omosomes allo w ed us to map the curated set of TSS that wer e pr e viousl y defined for D23580 (Hammarlöf et al. 2018) onto a combined D37712/D23580 reference genome.To allow individual TSS to be examined in particular c hr omosomal or plasmid r egions, data fr om both the dRNAseq and pooled RNA-seq experiments can be visualized in our on-line genome browser ( http:// hintonlab.com/jbrowse/ index.html?data=Combo _ D37/data ).

Preliminary gene expression profiling of S. Typhimurium ST313 sublineage L2.2
Given the high level of similarity between the genomes of L2.2 and L2.0, w e w ent on to identify differ ences at the tr anscriptional le v el.We performed a multi-condition RNA-seq-based transcriptomic analysis of gene expression profiles of L2.2 strain D37712 without biological replicates.
This compar ativ e tr anscriptomic scr een was based on our published a ppr oac h (Canals et al. 2019b ).Specifically, we used 15 individual infection-r ele v ant in vitro conditions (Kröger et al. 2013 ) and did intr a-macr opha ge tr anscriptome pr ofiling using the pr otocol pr e viousl y established for S. Typhim urium ST19 (Srikumar et al. 2015 ).The RNA-seq samples were mapped to a combined r efer ence genome, whic h included the annotated D23580 c hr omosome (Canals et al. 2019b ), as well as all the plasmids described earlier (pSLT-BT, pBT1, pBT3, and pCol1B9; see Methods).The initial RNA-seq assessment (detailed in Methods) involved 2-4 M non-rRNA/tRNA reads per sample, allowing gene signatures specific for each in vitro condition to be identified.Although single-replicate RNA-seq experiments of this type cannot be used for statisticall y r obust differ ential gene expr ession analysis , they do pro vide a useful scr eening a ppr oac h for identifying growth conditions to be used for follow-up experiments .T he individual RNA-seq experiments showed broad condition-specific similarities in gene expression between strains 4/74, D37712, and D23580 (Fig. 3 A).The gene expression values from each profiled condition are available as raw counts and Transcripts Per Million (TPMs) in Tables S4 and S5 .
To select the ideal environmental conditions to use for subsequent experiments, we assessed the expression profiles of known Salmonella pathogenicity islands which were broadly similar in strains D37712, and D23580.Although the expression profile of the SPI2 pathogenicity island was br oadl y similar between D37712, D23580 and 4/74 in most growth conditions, the SPI2 genes of D37712 were highly up-regulated in a single growth condition, NonSPI2 (Fig. 3 B,C).NonSPI2 is a minimal medium with a neutral pH and a relatively high level of phosphate, in which S. Typhimurium does not usually express the SPI2 pathogenicity island (Löber et al. 2006, Kröger et al. 2013 ).This intriguing observation prompted us to perform a more discriminating set of transcriptomic experiments, as described below.

Differential gene expression analysis of S . Typhimurium D37712 versus D23580 in four in vitro conditions with multiple biological replicates
To define the transcriptional signature of strain D37712 more accur atel y, we gener ated RNA-seq data from D37712 grown in four in vitro conditions that stimulate expression of the majority of virulence genes: Early Stationary Phase (ESP), anaerobic growth, NonSPI2, and InSPI2, with multiple (3-4) biological replicates .T he combination of acidity (pH 5.8) and low phosphate (0.4 mM Pi) in the InSPI2 media stimulates transcription of SPI2 genes in S. Typhimurium (Löber et al. 2006, Kröger et al. 2013 ).The NonSPI2 condition is based on the same PCN media recipe as InSPI2 media, but is neutral (pH 7.4), and contains higher levels of phosphate (25 mM Pi) (Löber et al. 2006, Kröger et al. 2013 ).
We compared the results with our published transcriptomic data for S. Typhimurium strains 4/74 and D23580 (Canals et al. 2019b, Kröger et al. 2013 ).Differential expression analysis with DEseq2, with conserv ativ e cut-offs (fold change ≥ 2, FDR ≤ 0.001), sho w ed that the gene expr ession pr ofiles of D37712 and D23580 wer e br oadl y similar, and shar ed k e y differ ences to the tr anscriptional profile of strain 4/74 under each of the four in vitro conditions (Fig. 4 A).The differ ential expr ession r esults ar e summarized in Table S6 .
We specificall y inv estigated tr anscription of the pgtE gene, which encodes the outer-membrane protease previously linked to the ability of African Salmonella ST313 to resist human serum killing (Hammarlöf et al. 2018 ).Compared to 4/74, the pgtE gene of both the D23580 and D37712 strains sho w ed a similar pattern of up-regulation by a factor of seven to 18 across all conditions .T his finding is consistent with the fact that D37712 carries the same T nucleotide in the -10 region of the pgtE promoter that is responsible for increased expression of the pgtE transcript in strain D23580 (Hammarlöf et al. 2018 ).
Ther e wer e no statisticall y significant c hanges in the expr ession of the majority (92%) of the 4729 orthologous coding genes shared by D37712 and D23580.We identified a total of 364 genes that were differentially expressed in at least one growth condition between D37712 and D23580 as follows: ESP (69 differ entiall y expr essed genes), anaer obic gr owth (214 differ entiall y expr essed genes), NonSPI2 (88 differ entiall y expr essed genes), and InSPI2 (17 differ entiall y expr essed genes; Fig. 4 B).
Ov er all, the differ entiall y expr essed genes that distinguished D37712 from D23580 only sho w ed expression differences in a single growth condition rather than across all conditions .T he differ entiall y expr essed genes included fla gellar genes (downr egulated), SPI2-associated genes(up-r egulated), and genes involved in general and anaerobic metabolism (down-regulated).
SPI2 pathogenicity island genes play a k e y role in the intracellular replication of S. Typhimurium, and encode the type III secretion system that is responsible for the translocation of k e y effector proteins into mammalian cells (Jennings et al. 2017 ).The RNA-seq data sho w ed that SPI2 genes w er e expr essed at similarl y high le vels in both D37712 and D23580 strains following induction (InSPI2 media; Fig. 4 B), and confirmed that the k e y SPI2 expression difference was only seen in strain D37712 under non-inducing growth conditions (NonSPI2 media).It is important to put this differential SPI2 expression into context.D37712 expresses SPI2 genes at about a 10-fold higher le v el than D23580 during growth in noninducing NonSPI2 media, but the actual le v el of expr ession was 20-fold less than the le v el stim ulated by growth in SPI2-inducing conditions (InSPI2 medium).
The up-regulation of fljA and fljB and the down-regulation of fliC in D37712, compared to D23580 in all four growth conditions likel y r eflects the opposite orientation of the hin switch in the D37712 genome compared to D23580.This type of hin inversion occurs frequently in S. Typhimurium (Johnson and Simon 1985 ).
Another gene that was up-regulated in D37712 across all profiled conditions was the c hr omosomall y encoded cysS chr , whic h encodes c ysteine-tRN A synthetase.Pr e viousl y, we r eported that transcription of the cysS chr of strain D23580 was uniformly downr egulated compar ed to 4/74.This down-r egulation was compensated by the presence of a pBT1 plasmid-encoded cysteine-tRNA synthetase (Canals et al. 2019a ).Accordingly, the incr eased expr ession of the c hr omosomal cysS gene in D37712 was consistent with the absence of the pBT1 plasmid.Our compar ativ e tr anscriptomic anal ysis sho w ed that expr ession le vels of c ysS w ere similar in D37712 and 4/74 under all growth conditions.
(B) (A) Numerous virulence genes and operons were differentially expressed between D23580 and D37712.The SPI-16-associated gtrABCa operon (STM0557, STM0558, and STM0559) is responsible for adding glucose residues to the O-antigen subunits of LPS that enhance the long-term colonization of the mammalian gastr ointestinal tr act by S. Typhim urium ST19 (Bogomolnaya et al. 2008 ).We found that the gtrABCa genes were significantly up-regulated in several conditions in D37712, compared to both D23580 and 4/74.
The spvABCD operon of D37712 was up-regulated under non-SPI2-inducing gr owth conditions, compar ed to D23580.A signature pseudogene of ST313 L2.2 is the frameshift insertion in the spvD gene that generates a truncated version of the SpvD protein.The H199I mutation at position 199 and the associated 17 amino acid truncation is predicted to ablate the activity of the SpvD cysteine protease (Grabe et al. 2016 ).SpvD negatively regulates the NF-B signaling pathway and promotes virulence of S. Typhimurium in mice .T he functional consequences of the spvD variant of ST313 L2.2 strain D37712 and the up-regulation of expression of the spvABCD operon remain to be established experimentally.

The SalComD37712 community transcriptional data resource
To allow scientists to gain their own biological insights from the analysis of this rich transcriptomic dataset, the transcriptomic and gene expression data generated in this study are presented online in a ne w comm unity r esource, SalComD37712 ( https:// tinyurl.com/SalcomD37712).The data resour ce sho ws the expression le v els of all D37712 coding and non-coding genes, including both c hr omosomal and plasmid-encoded transcripts .T he Sal-ComD37712 website complements our existing SalComD23580 ( https:// tinyurl.com/SalComD23580 ) resource, and adds an interstrain comparison of gene expression profiles between D37712 and D23580 as well as normalized gene expression values (TPM), using an intuitive heat map-based approach.SalComD37712 included our published RNA-seq data (Canals et al. 2019b ), reanalysed with an updated bioinformatic pipeline and a combined r efer ence genome (see Methods).This online resource facilitates the intuitiv e interr ogation of tr anscriptomic data as described pr e viousl y (Per ez-Sepulv eda and Hinton 2018 ).
Additionall y, we gener ated a unified genome-le v el br owser that provides access to the S. Typhimurium L2.2 D37712 transcriptome, in the context of our pr e viousl y published RNA-seq data for the L2.0 strain D23580 and the ST19 strain 4/74.T his no vel 'combo' browser is available at http:// hintonlab.com/jbrowse/ index.html?data=Combo _ D37/data .

Identification of phenotypes that distinguish ST313 sublineage L2.2 from L2.0
To explore the phenotypic impact of the transcriptomic signature of L2.2 (D37712), we performed a series of motility experiments, fluor escence-based gene expr ession experiments and mixed-growth assa ys .D33712 sho w ed a significantl y decr eased le v el of motility on NonSPI2 minimal media, compared with both the ST19 strain 4/74 and the L2 D23580 strain (Fig. 5 A).This finding was consistent with the transcriptomic data, which sho w ed do wnregulation of D37712 flagellar genes compared with D23580 in the NonSPI2 condition (Fig. 4 ).In contrast, no differential expression of flagellar genes was seen between D33712 and D23580 in the In-SPI2 growth condition (Fig. 4 ).The decreased motility phenotype may be linked to the inversion of the hin element detailed abo ve .The flagella system encodes a distinct type III secretion apparatus responsible for the dual functions of bacterial motility and activation of the mammalian innate immune system via TLR5 (Lai et al. 2013 ).
A k e y transcriptomic finding for strain D33712 was the expression of SPI2 genes during growth in an unusual environmental condition (NonSPI2) (Figs. 3 B,C and 4 B).NonSPI2 media differs from InSPI2 media by having a higher pH (pH7.4 versus pH5.8) and a higher le v el of phosphate (Löber et al. 2006 ).This a ppar ent differ ential expr ession of SPI2 genes at the transcriptomic level under non-inducing conditions led us to investigate the expression of SPI2 at a single-cell le v el using fluor escence tr anscriptional fusions.First, we introduced an ssaG -GFP + transcriptional fusion into the c hr omosome of str ains D33712 and D23580 (Methods; Table S8 ) to interrogate the expression of the k e y SPI2 operon with flow cytometry.Fig. 5 B shows that in NonSPI2 media, the ssaG promoter was expressed at a 62% higher level in D33712 than in D23580 confirming the results of the transcriptomic analysis.
Because only a proportion of S. Typhimurium cells express certain pathogenicity island-encoded genes during in vitro growth (Hautefort et al. 2003, Ackermann et al. 2008 ), we determined whether the increased level of expression of SPI2 genes (Fig. 4 B) w as caused b y a higher proportion of D33712 cells expressing SPI2 than D23580 cells.Using deri vati ves of the two strains that carried the ssaG -GFP + construct, we determined the numbers of fluorescent and non-fluorescent cells with flow cytometry (Methods).Under non-inducing conditions, slightly more D37712 cells expressed the ssaG SPI2 promoter than D23580 cells (65% vs. 60%, respectiv el y) (Fig. 5 C).Although this small difference was statistically significant ( t -test: P < .001,n = 3), it did not account for the 62% incr eased le v el of non-induced SPI2 expr ession seen in Fig. 5 B.
SPI2 expression is controlled by a complex regulatory system that operates at both a negative and positiv e le v el, involving silencing via H-NS (Lucchini et al. 2006 ), activation by SlyA and SsrB (Fass andGroisman 2009 , Walthers et al. 2011 ) as well as input from OmpR and Fis under non-inducing conditions (Osborne and Coombes 2011 ).The mechanistic basis of the aberrant SPI2 expr ession in str ain D37712 is worthy of further study.Possible explanations include the incomplete silencing of SPI2 transcription or the partial activation of the SPI2 virulence genes under noninducing growth conditions by an unknown regulatory factor.

Increased fitness of S. Typhimurium ST313 sublineage L2.2 compared with L2.0 in minimal media
It has become incr easingl y clear that distinct Salmonella pathovariants have evolved particular phenotypic properties that confer fitness adv anta ges during infection of particular avian or mammalian hosts (Br anc hu et al. 2018 ).Because S. Typhimurium ST313 L2.2 a ppear ed to hav e displaced S. Typhim urium ST313 L2.0 in Malawi, we speculated that S. Typhimurium ST313 L2.2 might have a competitive edge in some situations.Accor dingly, w e determined bacterial fitness using a mixed-growth competition assay (Wiser andLenski 2015 , Lian et al. 2023 ).The competitive index was calculated in thr ee differ ent gr owth media using pair-wise combinations of strains D37712 and D23580.Two independent appr oac hes wer e used to phenotypicall y distinguish the two str ains, one based on antibiotic resistance (Fig. 5 D) and the other based on fluorescent tagging ( Fig. S5 ).
To confirm that strains engineered to be kanamycin-resistant or gentamicin-resistant did not impact on fitness (Methods), we first verified that the tagged variants of D37712 or D23580 did not confer a gr owth adv anta ge in LB or NonSPI2 media ( Fig. S7 ).Next, we used a mixed-growth assay to investigate fitness of S. Typhimurium ST313 L2.0 strain D23580 or S. Typhimurium ST313 L2.2 strain D37712 during growth in LB, or InSPI2 or NonSPI2 minimal media.The data show that both str ains gr e w at similar le v els following overnight mixed growth in nutrient-rich LB media, but D37712 had a competiti ve ad v anta ge during mixed growth in In-SPI2 media (CI = 1.79;P < .05)and a gr eater competitiv e edge in NonSPI2 media (CI = 2.20; P < .0001).
We then used an independent fluor escence-based a ppr oac h to assess the fitness of strains D23580 and D37712 during mixed growth in NonSPI2 media.This time, the strains were engineered to carry either mScarlet or sGFP2 proteins and the mixedgro wth experiments inv olved pair-wise comparisons of reciprocall y ta gged str ains .T he flo w c ytometric data sho w ed that in both cases D37712 had a significant competitive adv anta ge in NonSPI2 media ( Figs.S5 and S6 ).
This combination of antibiotic resistance-based and fluor escence-based competitiv e index experiments lead us to conclude that S. Typhimurium ST313 L2.2 strain D37712 had a clear fitness adv anta ge ov er S. Typhim urium ST313 L2.0 str ain D23580 during mixed growth in two formulations of minimal media.The molecular basis of this fitness adv anta ge r emains to be established.

Discussion
Her e, we r eport that S. Typhim urium ST313 L2.0 has been clonall y r eplaced by the ST313 sublineages L2.2 and L2.3 as a cause of bloodstream infection in Blantyre , Mala wi.In 2018, L2.2 r epr esented the majority of the ST313 strains isolated from hospitalized patients in Malawi at the QECH.Our compar ativ e genomic analysis of ST313 L2.3 identified 30 c hr omosomal alter ations, one of which generated a deletion of the sseI effector gene.
Our RNA-seq-based analysis of ST313 L2.2 involved a detailed comparison versus ST313 L2.0 which revealed a k e y difference involving SPI2 expression.Following initial observations at the transcriptomic le v el in the ST313 L2 and L2.2 strains grown in a pHneutral minimal medium (NonSPI2), the increased expression of SPI2 was confirmed at the single-cell le v el using an ssaG transcriptional fusion.A series of experiments sho w ed that the ST313 L2.2 strain D37712 had a competitive adv anta ge ov er L2 strain D23580 during mixed growth in minimal media.We propose that this increased fitness of S. Typhimurium ST313 L2.2 has contributed to the replacement of ST313 L2.0 in Malawi in recent years.
Pr e viousl y, we compar ed thr ee virulence pr operties of the S .Typhimurium ST313 L2.0 D23580 and ST313 L2.2 D37712 strains.First, experiments involving Mucosal Invariant T (MAIT) cells sho w ed that both D37712 and D23580 fail to elicit the high le v el of activ ation of MAIT cells that c har acterizes infection by S .Typhimurium ST19 4/74 (Preciado-Llanes et al. 2020 ).Second, the D37712 and D23580 strains stimulate similar levels of up-regulation of IL10 gene expression upon infection of human dendritic cells (Aulicino et al. 2022 ).Thir d, w e sho w ed that both D37712 and D23580 express similarly high levels of the PgtE virulence factor that is responsible for the ability of S .Typhimurium ST313 to survive human serum-killing (Hammarlöf et al. 2018 ).These findings lead us to conclude that the compar ativ e genomic and transcriptomic differences that distinguish S .Typhimurium ST313 L2.0 strain D23580 from ST313 L2.2 D37712 (Fig. 4 ) do not modulate the ability of the pathogens to activate human MAIT cells or dendritic cells or to influence the PgtE-mediated serum survival phenotype of S. Typhimurium ST313.
Ideally, the implications of the competiti ve ad vantage of ST313 L2.2 would be determined in the context of pathogenesis.Howe v er, we lac k an informative infection model for S .Typhimurium ST313 (Lac harme-Lor a et al. 2019 ), and it is not yet possible to experimentally determine whether the impr ov ed fitness of L2.2 significantly enhances the success of ST313 during infection of humans.
We ha ve in vestigated the intricate interplay of gene function that underpins the success of S .Typhimurium ST313 L2.2.It is hoped that our findings will contribute to future therapeutic or pr ophylactic str ategies for combatting iNTS infections in the African setting.

Figure 1 .
Figure 1.Emergence of S. Typhimurium ST313 sublineages L2.2 and L2.3 in Malawi.(A) Evolutionary dynamics of S. Typhimurium lineages in Blantyre, Malawi from 1996 to 2018.A maximum likelihood tree constructed with 1000 bootstraps using the GTRGAMMA model in RaxML rooted on ST19, LT2.The genomes of 549 S .Typhimurium ST313 isolates from bacteraemic patients at the Queen Elizabeth Hospital in Blantyre , Mala wi were used for this analysis .T he proportions of the five lineages/sublineages are shown.(B) The total number of isolates of each lineage/sublineage per year.(C) Phylogenetic comparison between r epr esentativ e str ains of S. Typhim urium ST19 and four ST313 linea ges/sublinea ges (L1, L2.0, L2.2, and L2.3) showing the presence and absence of plasmids, prophages and the spvD pseudogene .T he complete phylogenetic analysis of 707 S. Typhimurium genomes is shown in Fig. S1 .

Figure 2 .
Figure 2. Key genetic similarities and differences between the chromosome and plasmid profiles of D23580 (lineage 2) and D37712 (L2.2). (A) A comparison of the D23580 (L2.0) and D37712 (L2.2) c hr omosomes .T he dots around the chromosome are different kinds of SNPs identified.Phages and Salmonella pathogenicity islands are shown in blue and red respectively.(B) Plasmid profile of D37712 versus D23580.The pSLT-BT virulence plasmid is present in both D37712 and D23580 and carries the Tn-21 transposable element; (C) pCol1B9 is present in D37712 and absent from D23580; and (D) pBT3 is present in both D37712 and D23580.(E) Absence of sseI gene and the STM1050 coding sequence in L2.2 (D37712), as compared to S.Typhimurium ST19 4/74 and S. Typhimurium ST313 L2.0 (D23580).(F) List of pseudogenes in D37712 and D23580, with r efer ence to 4/74.The colour blue means pseudogene/disrupted gene while grey indicates functional genes.macB is a pseudogene in D23580 (L2.0) but not in L2.2, while spvD is a pseudogene in L2.2 but not in L2.0.All L2.2 strains share similar pseudogenes.

Figure 4 .
Figure 4. Differential gene expression of S. Typhimurium 4/74, D37712, and D23580 under four in vitro conditions.(A) Boxplots indicating the number of differ entiall y expr essed genes identified in the following in vitro gr owth conditions: earl y stationary phase, ESP; anaer obic gr owth, NoO2; SPI-2 inducing medium, InSPI2; SPI-2 non-inducing minimal medium, NonSPI2.Multiple (3-5) biological replicates were used for comparison.DESeq2 was used for differential analysis; only genes with | log2FC | ≥ 1 and with the adjusted P value ≤.001 were retained.(B) Heatmap of the genes differentially expressed between D23580 and D37712.Functional groups and operons of interest are highlighted on the right of Panel B.

Figure 5 .
Figure 5. Phenotypes that distinguish ST313 L2.2 from ST313 L2.0.(A) Swimming motility assay of strains D23589, D37712 and 4/74, with a r epr esentativ e plate shown on the left.Av er a ge migr ation diameters wer e measur ed after 4 and 8 h.Eac h bar r epr esents the mean of thr ee biological replicates, with error bars showing standard deviation.A significant difference ( * * * ) indicates P value ( t -test) < .001.In Panels B & C, comparison of ssaG expression by flow cytometry using D23580 and D37712 deri vati ves containing a chromosomal ssaG -GFP + transcriptional fusion, strains SZS008 and SZS032, r espectiv el y.Cells wer e collected at 8 h after inoculation in NonSPI2 media.Ten thousand e v ents wer e acquir ed for eac h sample.(B) Mean fluorescent intensity signal of ssaG -GFP + for D23580 (SZS008, dark grey) and D37712 (SZS032, grey) grown in NonSPI2 media.A significant difference ( * * * ) indicates P value ( t -test) < .001.(C) The proportions of bacterial cells that expressed ssaG -GFP + during growth in NonSPI2 were determined.The percentage of GFP-expressing (green) and cells (white) for D23580 (SZS008) and D37712 is shown.Each bar the mean of three biological replicates, error bars show standard deviation.A significant difference ( * * * ) P value ( t -test) < .001.(D) Relative fitness of wild-type D23580 and D37712 and their kanamycin-resistant deriv ativ es.Bacterial numbers were after overnight culture of a 1:1 mixture (wild versus Km R ) in LB (left), InSPI2 (middle) and NonSPI2 (right) media.Each dot represents the log-transformed mean competitive index of three biological replicates with error bars r epr esenting a 95% confidence interval from the standard deviation.A log number higher than 0 reflects the increased fitness of kanamycin-resistant deri vati ves.P values were determined t -test ( * * * : P < .001;* * : P < * : P .05;ns: not significant).