Mukuru informal settlement is situated East of Nairobi city, about 15 km from the city centre. It is one of the largest slums in the city with a population of around 250,000 people KNBS KNBOS, 2009. The informal settlement is made up of improvised temporary dwellings often made from scrap materials, such as corrugated metal sheets, plywood, and polythene-sheets Olack et al., 2011. In addition to poverty, a number of factors associated with informal settlements, including overcrowding, substandard housing, unclean and insufficient quantities of water, and inadequate sanitation, contribute to a high incidence of infectious diseases and increased mortality among children under five years Kyobutungi et al., 2008; Mutisya and Yarime, 2011. Mukuru informal settlement is divided into eight villages; Mukuru Lunga-Lunga, Mukuru kwa Sinai, Mukuru kwa Ruben, Mukuru kwa Njenga, Mukuru Kayaba, Fuata Nyayo, Jamaica, and Mukuru North. This study was carried out in two of the large villages, Mukuru kwa Njenga and Mukuru kwa Ruben, with a combined population of 150,000. Spatial mapping of the two villages was conducted using the Universal Transverse Mercator system Tinline and Gregory, 1988, and patient details collected as described previously Mbae et al., 2020.

The two villages in the informal settlement are served by three outpatient clinics: Ruben Health Centre located in the Ruben village (zone named Simba cool, serves approximately 30% of the population), Missionaries of Mary Located in Kwa Njenga village (zone named Vietnam, serves approximately 45% of the population), and County Government Clinic in Kwa Njenga village (Zone named MCC and serves approximately 25% of the population). The fourth site, Mbagathi District Hospital, is located on the western side of Nairobi city, 5 km from city centre and was used as a referral facility. Participants living outside of the mapped demographic surveillance site (DSS) who came to seek medical services in any of the three study site health facilities or at Mbagathi District Hospital were included for the purpose of tracking typhoid cases and carriers treated at the facilities, but are reported separately in the Results section.

Typhoid fever cases and asymptomatic carriers presented in this study were identified and recruited as part of a larger study on surveillance and genomics of invasive Salmonella disease in children and young adults less than 16 years of age Mbae et al., 2020; Kariuki et al., 2020. Children presenting as outpatients at the three study clinics and Mbagathi District Hospital between August 2013 and November 2016 were triaged to identify those with fever, headache and/or diarrhoea for recruitment into the study as potential cases. Patients with current fever ( ≥ 38°C) and reportedly febrile for ≥3 days were considered potential typhoid cases and assessed via blood or stool culture. The primary typhoid case definition (data presented in Table 1) was children aged 0–16 years with ≥3 days fever ≥38°C and positive blood or stool culture for S. Typhi (see bacterial culture methods below).

During the study period, age-matched controls were recruited from children without current fever or diarrhoea attending the same health facilities for healthy mother and child clinics (e.g. for vaccination and nutritional advice). Those with S. Typhi-positive stool culture were designated as asymptomatic typhoid carriers as described previously Mbae et al., 2020. Hence, the inclusion criteria for asymptomatic typhoid carriers (data presented in Table 1) were children aged 0–16 years with no diarrhoea, no current fever, and no recent fever history, with stool culture positive for S. Typhi (see bacterial culture methods below). The total number of participants was computed on the basis of a 4% prevalence rate of typhoid from previous study Kariuki et al., 2010. (A structured questionnaire was used to collect demographic data for both cases and controls recruited into the study as described previously Mbae et al., 2020).

All isolates cultured from participants and identified as Salmonella were archived and later revived for WGS as detailed below. The sequence data revealed some mis-identification of Salmonella serotypes (Figure 1, Supplementary file 1), hence for genomic analyses we included all cases and controls whose cultures were found to be S. Typhi positive by WGS rather than those identified as S. Typhi positive by serotype in the microbiology laboratory.

For blood culture, 1–3 mL for children < 5 years of age and 5–10 mL for those 5–16 years of age was collected in a syringe, placed into Bactec media bottles (Becton-Dickinson, New Jersey, USA), incubated at 37°C in a computerised BACTEC 9,050 Blood Culture System (Becton-Dickinson), and subcultured after 24–48 hr onto blood, chocolate and MacConkey agar (Oxoid, Basingstoke, UK) plates. All isolates, whether from cases or carriers were cultured on selenite F (Oxoid) broth aerobically at 37°C overnight. Broth cultures were then subcultured on MacConkey agar and Salmonella-Shigella agar (Oxoid) and incubated at 37°C overnight. Blood and stool isolates were identified using a series of standard biochemical and serological tests as described previously Mbae et al., 2020. Briefly, colonies of S. Typhi identified from biochemical and PCR testing were subjected to serological identification as S. Typhi using the slide agglutination technique and applying monovalent antisera (Murex Diagnostics, Dartford, UK). A drop of the antisera to be tested was mixed with a bacterial smear on a slide to observe for the presence or absence of agglutination in two minutes.

Antimicrobial susceptibility testing was performed using the disk diffusion technique for ampicillin 10 µg, tetracycline 30 µg, co-trimoxazole 25 µg, chloramphenicol 30 µg, cefpodoxime 30 µg, ceftazidime 30 µg, ceftriaxone 30 µg, cefotaxime 30 µg, ciprofloxacin 5 µg, and nalidixic acid 10 µg as described previously Mbae et al., 2020. Results were interpreted according to the 2017 guidelines provided by the Clinical and Laboratory Standards Institute (CLSI) Patel et al., 2015.

All Salmonella isolated from cases and controls were subcultured at the end of the study for DNA extraction and WGS. These included 243 cultures identified as S. Typhi from cases (85 from blood and 63 from stool) and 95 from controls (all from stool) (see Figure 1, (1) Supplementary file 1), which are the subject of this study (non-typhoidal Salmonella data is reported elsewhere Kariuki et al., 2020). Twelve S. Typhi case isolates and 10 control isolates could not be revived and were not further analysed. DNA was extracted using the Wizard Genomic DNA Extraction Kit (Promega, Wisconsin, USA) and shipped on ice to the Wellcome Sanger Institute for sequencing using the Illumina platform as described previously Park et al., 2018. A total of 217 S. Typhi DNA samples were successfully sequenced (two were of insufficient quality to construct sequencing libraries, or failed sequencing). Non-S. Typhi bacterial DNA sequences were detected in 75 samples (34.6%; organisms detected are shown in Supplementary file 1), and 11 sequences originally identified as other Salmonella serotypes were later found to be S. Typhi with genomic data. Two sequences showed the presence of S. Typhi, but at low depth, and were subsequently omitted from further genomic analyses, leaving genome data for 153 S. Typhi isolates for further analysis.

The taxonomic identities of non-typhoidal Salmonella spp. were determined using Kraken (v0.10.6) Wood and Salzberg, 2014, multi-locus sequence typing (MLST) with SRST2 Inouye et al., 2014; Jolley and Maiden, 2010, as well as the Salmonella In Silico Typing Resource (SISTR) and Speciator (both available via Typhi Pathogenwatch; https://pathogen.watch/) Argimón et al., 2021.

For SNP analysis, paired-end reads from 153 S. Typhi isolates were mapped to the reference sequence of S. Typhi CT18 (accession number: AL513382) Parkhill et al., 2001 using the RedDog mapping pipeline (v1beta.10.3), available at https://github.com/katholt/reddog (copy archived at swh:1:rev:90707ace56d6189997526273cf97013d84b84d90; Holt, 2021a) and detailed in supplementary methods. Read alignments were used to assign isolates to previously defined lineages according to the extended genotyping framework Wong et al., 2016; Britto et al., 2018 with the GenoTyphi pipeline (available at https://github.com/katholt/genotyphi, copy archived at swh:1:rev:f489d0e004e30bdca683c76c25ac7f4d79162791; Holt, 2021b). Unique SNPs defining three novel lineages were identified from the genome-wide SNP allele table and added to the GenoTyphi scheme to facilitate easy identification of these lineages in future studies (details in supplementary methods and results).

Phylogenetic analyses were restricted to WGS-confirmed pure cultures of S. Typhi H58 (genotype 4.3.1, n = 128). For some analyses, an additional 1076 S. Typhi H58 genomes from previously published WGS studies of global and African isolates Wong et al., 2015; Park et al., 2018; Wong et al., 2016; Pham Thanh et al., 2016 were also included for context, along with 61 non-H58 genomes for phylogenetic outgroup rooting, using the same mapping approach detailed above (see Supplementary file 2 for full list of genomes analysed and their public data accessions). SNPs called in phage regions or repetitive sequences were filtered from the alignment (details in supplementary methods). Analysis with Gubbins (v2.3.2) identified two regions affected by recombination (coordinates 954,115–970,731 (genes STY0961-STY0976), and 1,438,676–1,467,273 (genes STY1485-STY1508) in the CT18 reference genome), which were excluded from the alignment Croucher et al., 2015. This resulted in a final set of 8635 SNPs. From this global alignment we extracted a separate SNP alignment for the set of 239 Kenyan S. Typhi 4.3.1 genomes (n = 128 from this study and n = 111 from published studies, see Supplementary file 3; Feasey et al., 2015; Yu et al., 2016), the resulting alignment of length 489 SNPs was used for temporal analyses (described below and in supplementary methods).

Maximum likelihood (ML) phylogenetic trees were inferred from SNP alignments using RAxML (v8.2.9) Stamatakis, 2014 (as detailed in supplementary methods) and the resulting trees were visualised using Microreact (interactive global H58 phylogeny available at: https://microreact.org/project/wViqmaRdZuFVEb6yk4i1jU) Argimon et al., 2016.

Pairwise SNP distances were calculated from the SNP alignment using the dist.dna() function in the R package ape (v5.4.1) Paradis et al., 2004. Terminal branch lengths were extracted from phylogenies using R package ggtree (v2.2.4) Yu et al., 2016. Non-synonymous mutations occurring in terminal branches were detected using SNPPar (v0.4.2dev) Edwards et al., 2020 and grouped by function based on the gene in which they were found, according to the functional classification scheme in the genome annotation of S. Typhi CT18 Thanh Duy et al., 2020; Parkhill et al., 2001.

To investigate temporal signal and date the introduction of S. Typhi H58 into Kenya based on the 239 available Kenyan genomes (n = 128 from this study, and n = 111 from previous studies Park et al., 2018; Wong et al., 2016), we used several methods. First, we used TempEst (v1.5.1) Rambaut et al., 2016 to assess temporal structure (i.e. clock-like evolution) by conducting a regression analysis of the root-to-tip branch distances of the ML tree as a function of sampling date, and later a date-randomisation test (full details of temporal signal assessment and model selection are provided in supplementary methods). To estimate divergence dates for the three S. Typhi H58 sublineages we detected in Kenya (EA1-3), we used BEAST (v1.10) Suchard et al., 2018 to fit a phylodynamic model to the SNP alignment and isolation dates as described in supplementary methods. The resultant MCC tree was visualised using ggtree (v2.2.4) Yu et al., 2016 and Microreact Argimon et al., 2016 (interactive phylogeny available at: https://microreact.org/project/I2KUoasUB).

The read mapping-based allele typer SRST2 (v0.2.0) Inouye et al., 2014 was used to detect the presence of plasmid replicons (PlasmidFinder database Carattoli et al., 2014) and antimicrobial resistance (AMR) genes (ARGannot database Gupta et al., 2014). Where AMR genes were observed without evidence of a known AMR plasmid, raw read data was assembled using Unicycler (v0.4.7) Wick et al., 2017 and then examined using Bandage (v0.8.1) Wick et al., 2015 to confirm the chromosomal location and composition of AMR-associated transposons. As the Tn2670-like composite transposon commonly associated with the acquisition of MDR genes in S. Typhi is mediated by IS1 translocation Wong et al., 2015, ISMapper (v2.0) Hawkey et al., 2015 was also used to identify the location of IS1 insertion sequences in the S. Typhi chromosome as described in supplementary methods. SRST2 was used to determine IncHI1 plasmid MLST (multi-locus sequence types) sequence types (pMLST), and minor alleles were identified by mapping to the plasmid pAKU1 reference sequence (accession number AM412236) in the same manner as described above for the S. Typhi chromosome (details in supplementary methods and Supplementary files 4-5). Point mutations located within the quinolone resistance determining region (QRDR) of genes gyrA, gyrB, and parC associated with reduced susceptibility to fluoroquinolones Pham Thanh et al., 2016 were detected using GenoTyphi Wong et al., 2016; Britto et al., 2018 as detailed in supplementary methods.

All statistical analyses unless otherwise stated were carried out using R (v4.0.2). Details of specific functions within R packages used for individual analyses are available in supplementary methods.

Raw Illumina sequence reads have been submitted to the European Nucleotide Archive (ENA) under accession PRJEB19289. Individual sequence accession numbers are listed in Supplementary file 1.

From August 2013-November 2016, a total of 4670 febrile children were recruited across the four study sites and subjected to blood and/or stool culture. S. Typhi was identified in cultures from 148 children (3.2%); the annual rate was steady over the study period but significantly higher amongst males (4.0% vs 2.3%, p = 0.0008, see Table 1). The odds of S. Typhi positive culture increased significantly with age (OR 1.08, p = 0.0005) but the effect was restricted to males (see Table 1), amongst whom the isolation rate was 1.3% in those ≤1 year, 2.0% in those aged 1–7 years, and 3.4% in those >7 years old (compared with 0.95%, 1.1%, and 0.94%, respectively amongst females). A total of 8549 age-matched control participants (with no current diarrhoea and no recent fever history) were recruited and subjected to stool culture. S. Typhi was identified in cultures from n = 95 (1.1%); these are considered asymptomatic carriers. S. Typhi culture positivity amongst controls was not significantly associated with age or sex and was stable over the study period (see Table 1 and Supplementary file 6). Significant associations identified from culture data were also observed when using WGS confirmed infections only, and no significant statistical association was found between phenotypic or genotypic AMR patterns and case/control status, age, or sex.

The presence of S. Typhi was confirmed by WGS in 94 cases (64%) and 50 controls (53%) that were originally identified as S. Typhi via microbiological culture (Figure 1, Supplementary file 1). S. Typhi genotype 4.3.1 (H58) was dominant throughout the study (n = 145, 95%), amongst both cases and controls (Table 2). Five other genotypes were detected: 2.2.2 (n = 1), 2.5.0 (n = 3), 3.0.0 (n = 3), and 4.1.1 (n = 1), see Table 2.

The few non-H58 isolates (Table 2) lacked any known AMR determinants. In contrast, the majority of H58 isolates were MDR (n = 116, 80%), often carrying acquired genes conferring resistance to ampicillin (blaTEM-1), chloramphenicol (catA1), co-trimoxazole (dfrA7 plus sul1 and/or sul2), and streptomycin (strAB). In 33 genomes (23% of H58), these genes were carried by a Tn2670-like complex transposable element inserted in the chromosome as reported previously in the region Wong et al., 2015; Feasey et al., 2015; Park et al., 2018. The remaining 83 MDR genomes (57% of H58) carried a closely related Tn2670-like transposon located within an IncHI1 plasmid, which in all but one isolate also carried an additional tetracycline resistance gene (tetB). The IncHI1 plasmids were genotyped as plasmid sequence type 6 (PST6), which is associated with MDR H58 in East Africa and South Asia Wong et al., 2015; Park et al., 2018; Holt et al., 2011. We compared single-nucleotide variant haplotypes for these plasmids with those from 534 IncHI1 PST6 plasmids sequenced previously from African and global studies (all of which were carried by H58 S. Typhi hosts, see Supplementary file 4). The wildtype PST6 plasmid haplotype was detected in S. Typhi hosts of all 6 H58 genotypes, whereas the derived plasmid haplotypes were each detected in a single S. Typhi host genotype (Figure 2—figure supplement 1). This is consistent with ongoing microevolution of the PST6 plasmid within S. Typhi lineages since the acquisition of the plasmid by the mrca of S. Typhi H58, but shows no evidence of transfer of plasmid haplotypes between S. Typhi lineages. Thus, the observed AMR phenotypes in our study site (n = 128 H58 and n = 8 non-H58 genome sequences) corresponded to the presence of known molecular determinants of AMR and mobile genetic elements. Estimates of sensitivity and specificity of AMR genotyping are presented in Supplementary file 7 and supplementary results. No statistical association was observed between the presence of MDR genes or QRDR mutations shown in Table 2 and case/control status, age, or sex.

S. Typhi H58 (genotype 4.3.1) can be subdivided into lineages I (genotype 4.3.1.1) and II (genotype 4.3.1.2). Lineage II was more common in this setting than lineage I: n = 90 (62.1% of H58) vs n = 55 (37.9%). Examination of the global phylogeny (Figure 2, and online interactive version https://microreact.org/project/wViqmaRdZuFVEb6yk4i1jU) revealed all H58 lineage I isolates from this study shared a most recent common ancestor (mrca) whose descendants form a monophyletic clade that exclusively comprised S. Typhi from East African countries (see Figure 2,), here defined as H58 sublineage EA1 (East Africa 1) with genotype designation 4.3.1.1.EA1 (labelled in Figure 2).

S. Typhi H58 lineage II (genotype 4.3.1.2) isolates from our study belonged to two distinct clades of the global phylogeny (Figure 2), which were each exclusively populated by East African isolates. The largest of these clades (n = 80 isolates, of which 81.3% derive from the current study) formed a monophyletic group nested within a deeper clade of diverse South Asian isolates (see Figure 2), and corresponds to the previously reported introduction of H58 lineage II into Kenya from South Asia Wong et al., 2015. This lineage, here defined as H58 sublineage EA2 (East Africa 2) is designated genotype 4.3.1.2.EA2 (labelled in Figure 2). The smaller East African H58 lineage II clade (n = 43 isolates) is designated genotype 4.3.1.2.EA3 (labelled in Figure 2) and comprised two sister clades, separated by ≥13 SNPs: one involving isolates from Kenya (n = 13, all from this study) and the other isolates from Uganda (n = 30), which accounted for 100% of the typhoid burden at the Ugandan site where they were identified (see Figure 2). All three East African H58 genotypes have been added to the GenoTyphi scheme using unique marker SNPs and further details on these are provided in supplementary results.

The three East African H58 subgroups circulating in our setting all had high rates of MDR (84%, 74%, and 94%, respectively); however, in EA2, MDR was exclusively associated with the PST6-IncHI1 plasmid, and in EA3 exclusively with the chromosomal insertion (see Table 2, Figure 2). In EA1, most MDR was associated with the PST6-IncHI1 plasmid. However, a subclade of isolates (associated with spread to Tanzania and Malawi) carried the chromosomal insertion instead (see Table 2, Figure 2, , supplementary methods).

No statistically significant differences in genotype distribution were observed between cases and controls (p = 0.077, using Chi-squared test, data in Table 2), or between males and females (p = 0.37, using Chi-squared test, data in Supplementary file 8), consistent with symptomatic and asymptomatic infections being drawn from the same general circulating pool of pathogens. The distribution of genotypes amongst cases varied by age group (p = 0.01, using Chi-square test), with the frequency of EA1 declining with age and the overall diversity increasing with age (Table 3). No significant differences in age groups was evident amongst controls (p = 0.9 using Chi-square test, see Table 3).

We examined the spatial and temporal distribution of all S. Typhi isolates collected at the study clinics (see methods; Figure 3—figure supplement 1 and Supplementary file 9), and the subset of 96 S. Typhi from cases and 67 from carriers living within the demographic surveillance site (DSS) (Figure 3, and Table 4). A number of peaks in monthly S. Typhi case and carrier numbers are apparent in both cohorts (Figure 3, Figure 3—figure supplement 1), with fewer cases and carriers observed in warmer months. Carrier counts remained relatively consistent throughout the study period. We tested for association between case or carrier peaks ( > 2 positives per month) and high rainfall or temperature in the same month, previous month, or 2 months prior to the month of observation (Table 4, and Supplementary file 9). For those S. Typhi from within the DSS, high temperatures were associated with lower case and carrier counts in the same month, and in the subsequent month (p < 0.05, using Fisher’s exact test), however no associations between high rainfall and elevated case or carrier counts was observed. Repeating these analyses with WGS-confirmed isolates, the association with temperature was no longer significant (although the trend remained; see Supplementary file 10, Supplementary file 11). No association was detected between patient numbers attending study clinics and climate variables (Supplementary file 12).

GPS coordinates were available for n = 139 (55%) S. Typhi isolates, and we endeavoured to look for geographic hotspots suggestive of major point-source single-genotype outbreaks in the informal settlement. However, our data revealed that the three H58 genotypes and non-H58 genotypes were co-circulating throughout the study area, with no evidence of geographic restriction of specific genotypes (see Figure 4—figure supplement 1). Further, we did not observe any spatially linked phylogenetic clusters of closely related sequences.

We applied Bayesian phylodynamic analysis to all available Kenyan H58 genomes to estimate the dates of emergence of each of the East African lineages. The data showed temporal structure (see methods and Figure 4—figure supplement 2), and we estimated a genome-wide substitution rate of 0.8 SNPs per genome per year (95% HPD, 0.1–1.0). This translates to a rate of 1.9 × 10^–7 genome-wide substitutions per site per year (95% HPD = 1.5 × 10^–7-2.2 × 10^–7). The novel EA1 isolates from this study (accounting for 35% of cases and 37% of controls) were intermingled with those sequenced previously from Kenya and were genetically diverse (median pairwise distance ~16 SNPs, interquartile range 12–27). This is indicative of a well-established EA1 S. Typhi population in Nairobi for which we estimate the mrca existed circa 1990 (95% HPD, 1981–1999) (see Figure 4a, Figure 4—figure supplement 3). The most common lineage was EA2 (48%), which also showed extensive diversity and we estimate emerged circa 1988–1990 (95% HPD, 1978–1997) (see Figure 4a, Figure 4—figure supplement 2), earlier than the first recorded H58 Lineage II isolation in Kenya in 2004 Kariuki et al., 2010. We estimate the MDR fluoroquinolone non-susceptible lineage EA3, which accounts for just 11 % of isolates, arrived much more recently (Kenyan mrca circa 2012, 95% HPD 2009–2014) (see Figure 4a, Figure 4—figure supplement 3). The topology of the global H58 tree (Figure 2) supports South Asia as the most likely origin for EA3, with EA3 strains spreading between Kenya and Uganda, probably through the shared transport systems.

The Bayesian tree of Kenyan H58 isolates (Figure 4a, Figure 4—figure supplement 3) shows intermingling of sequences from acute cases and asymptomatic carriers. Sequences from carriers appeared more deeply branched than those of cases which we tested by comparing the terminal branch lengths (estimated in units of time in the Bayesian phylogeny) and isolate-specific SNP counts, for high-quality H58 sequences from acute cases (n = 85) vs those of asymptomatic carriers (n = 43) (Figure 4b, ). The mean values were higher for carriers vs cases (Figure 4b–c, Figure 4—figure supplement 3), with the trend being more pronounced among samples from within the DSS but these trends were not statistically significant (p = 0.42 for unique SNPs and p = 0.57 for terminal branches for all samples, using one-sided Wilcoxon rank sum test; p = 0.051 for unique SNPs and p = 0.58 for terminal branches in DSS). The mean number of non-synonymous (NS) mutations detected in terminal branches was greater for carrier isolates than those from cases, but again this difference was not statistically significant (0.72 vs 0.54 for all sequences, p = 0.53 using Wilcox rank sum test; 0.81 vs 0.39, p = 0.20 inside the DSS; see Figure 4d, Figure 4—figure supplement 3). There was also no significant difference in terminal branch lengths or unique SNP counts between genomes carrying vs lacking MDR genes or QRDR mutations (data not shown).

Examination of the location of terminal-branch NS mutations revealed that certain functional categories of genes carried more NS mutations arising on terminal branches associated with carriage samples vs those from acute cases (Figure 5; Supplementary file 13). Notably, carriage samples were associated with significantly higher frequencies of terminal-branch NS mutations in genes responsible for the synthesis of surface polysaccharides and antigens (9.3% of carriers vs 1.2% of acute cases (all of which were blood isolates), p = 0.043, Fisher’s exact test). Notably, in the viaB operon (responsible for Vi capsule biosynthesis) we identified n = 2/43 carriage isolates that harboured NS mutations (tviD-R159C and tviE-P263S) compared with only n = 1/85 case isolate (tviB-V1M); and in the wba cluster genes (responsible for O-antigen biosynthesis) we identified n = 2/43 carriage isolates that harboured NS mutations (wza-V137G and wzxC-L26F) whilst none were detected among case samples (Supplementary file 13). Non-significant excesses of mutations in carriage isolates were also observed for periplasmic and exported lipoproteins (6.98% vs 1.92% in blood isolates from febrile cases and 3.03% in stool isolates from febrile cases, see Figure 5).

Our study is the first case-control study to identify and sequence both typhoid carriers and cases contemporaneously in an endemic community setting. High rates of AMR among both infection types in Kenya combined with high carriage and case rates, especially in the younger age groups, highlight the need for enhanced AMR and genomic surveillance in this region to inform both treatment guidelines and control strategies that keep pace with the local evolution and spread of AMR. Intervention strategies are urgently needed including the introduction of the new Vi conjugate vaccine in a programme that includes targeting of paediatric age groups in the short term, and improvements to WaSH infrastructure in the long term.