Genetic variation in the MacAB-TolC efflux pump influences pathogenesis of invasive Salmonella isolates from Africa

The various sub-species of Salmonella enterica cause a range of disease in human hosts. The human-adapted Salmonella enterica serovar Typhi enters the gastrointestinal tract and invades systemic sites to cause enteric (typhoid) fever. In contrast, most non-typhoidal serovars of Salmonella are primarily restricted to gut tissues. Across Africa, invasive non-typhoidal Salmonella (iNTS) have emerged with an ability to spread beyond the gastrointestinal tract and cause systemic bloodstream infections with increased morbidity and mortality. To investigate this evolution in pathogenesis, we compared the genomes of African iNTS isolates with other Salmonella enterica serovar Typhimurium and identified several macA and macB gene variants unique to African iNTS. MacAB forms a tripartite efflux pump with TolC and is implicated in Salmonella pathogenesis. We show that macAB transcription is upregulated during macrophage infection and after antimicrobial peptide exposure, with macAB transcription being supported by the PhoP/Q two-component system. Constitutive expression of macAB improves survival of Salmonella in the presence of the antimicrobial peptide C18G. Furthermore, these macAB variants affect replication in macrophages and influence fitness during colonization of the murine gastrointestinal tract. Importantly, the infection outcome resulting from these macAB variants depends upon both the Salmonella Typhimurium genetic background and the host gene Nramp1, an important determinant of innate resistance to intracellular bacterial infection. The variations we have identified in the MacAB-TolC efflux pump in African iNTS may reflect evolution within human host populations that are compromised in their ability to clear intracellular Salmonella infections.

Introduction Salmonella infections continue to be a significant challenge for human health. With an estimated 95 million annual cases, non-typhoidal Salmonella (NTS) infection is typically characterized by severe but self-resolving gastroenteritis in otherwise healthy people [1][2][3]. Typhoid and paratyphoid fever cases number more than 14 million annually and are characterized by invasive, bloodstream infection by Salmonella serovars Typhi and Paratyphi, respectively [4]. Risk for typhoid disease remains high in geographic areas with inadequate sanitation infrastructure as the infectious cycle relies on human-to-human transmission. While some humans can become asymptomatic chronic carriers, untreated typhoid fever is often fatal [3]. With appropriate treatment, the 1% case mortality of typhoid fever is similar to that of gastroenteritis associated with NTS [2,4,5].
Invasive non-typhoidal Salmonella (iNTS) isolates belonging to the serovars Typhimurium and Enteritidis have caused major disease outbreaks within sub-Saharan Africa [6-9]. These African iNTS isolates are associated with systemic infections, and particularly high case mortality in children less than 5 years of age, the elderly, and those with comorbidities such as HIV and malaria [6,10]. Although typhoid infections outnumber iNTS infections by over 25-fold globally, iNTS caused nearly half as many deaths with overall mortality rates above 14% [5]. In Africa, iNTS was responsible for 49,600 deaths in 2017 [5]. There is great need to further understand iNTS pathogenesis and epidemiology in order to improve diagnosis and clinical outcomes for these increasingly antibiotic resistant infections [11][12][13].
By multi-locus sequence typing, many Salmonella enterica serovar Typhimurium (S. Typhimurium) gastrointestinal isolates are classified as sequence-type 19 ("ST19"), while African S. Typhimurium associated with invasive disease belong mainly to sequence-type 313 ("ST313") [7,9]. ST313 isolates have also been observed in the UK and Brazil, though African ST313 form a distinct lineage [14][15][16]. Comparative genomic analysis has identified numerous changes in African ST313 lineages. While they have acquired unique prophages, plasmids, and antibiotic resistance genes, African ST313 also display gene degradation events that impair the ability of these isolates to survive outside of mammalian hosts [13,[17][18][19][20][21][22][23]. Further gene degradation events have been shown to alter invasive and immune stimulating behavior in experimental animal models, supporting the view that African iNTS strains are evolving from causing strictly enteropathogenic disease to causing invasive disease in human hosts [15,[24][25][26].
In our comparative analysis of S. Typhimurium ST313 lineage isolates with other Salmonella genomes we observed specific changes at the macAB locus in ST313 lineages. In Gramnegative bacteria, MacAB forms a tripartite channel with the outer membrane protein TolC to efflux various antimicrobial compounds as well as endogenous molecules and toxins [27][28][29][30][31].
As an ABC-type efflux pump, hydrolysis of cytoplasmic ATP by MacB directly drives movement of MacA and TolC to translocate molecules from the periplasm into the extracellular space [32,33]. The naming of macA and macB (previously annotated as ybjY and ybjZ) references their ability to confer resistance to macrolide antibiotics when overexpressed together from a plasmid [34], though in standard laboratory culture conditions macAB is not expressed in Salmonella [35,36]. In clinical isolates of other bacteria, increased expression of macAB homologues increases resistance to antimicrobial peptides such as polymyxins [37]. The fact that MacAB has a virulence role in animal models of oral Salmonella infection [35,38] prompted us to explore how African ST313-associated macAB gene variants might influence pathogenesis.

Lineage-specific variation of the macAB locus in invasive African S. Typhimurium ST313 isolates
Comparative analysis of all currently available genomes of ST313 isolates of Salmonella enterica serovar Typhimurium identified several genomic changes at the macAB locus that differed from other gastroenteritis-associated S. Typhimurium (Fig 1A). Although some UK and Brazilian ST313 isolates carry an indel in the macA gene, this variant is not present in African isolates from ST313 lineages (1,2) or sublineages (2.1, 2.2) that are associated with invasive disease ( Fig  1A). Instead, all other ST313 lineage isolates in our analysis carry a C➝T non-synonymous SNP within macA (Fig 1A), which replaces Serine (S 174 ) with Leucine (L 174 ) (Fig 1B and 1C). Alignment of this sub-region of MacA shows that the Serine residue is highly conserved amongst other Gram-negative bacterial genera ( Fig 1D). The structure of the Escherichia coli (E. coli) MacAB-TolC complex has been solved [33]; assuming an analogous overall structure in S. Typhimurium, the hydrophilic Serine 174 residue faces the channel interior, residing beyond the proposed gating ring and in series with other hydrophilic residues that form the interior surface of the MacA channel [33] (Fig 1E). We hypothesized that the ST313-associated mutation of this conserved Serine residue to the hydrophobic amino acid Leucine altered the function of the MacAB-TolC channel, particularly since this mutation would be repeated around the interior of the fully-assembled hexameric MacA channel (Fig 1E, right).
S. Typhimurium ST313 lineage 1 isolates contain a nonsense mutation in macB [19] changing the codon for W262 of MacB to a stop codon; this change lies within the amphipathic helix that precedes the first transmembrane domain [32], leading to the production of a truncated MacB (Fig 1F) [32]. Furthermore, a two-nucleotide insertion (an indel) that results in a frameshift in the macB gene was found in about half of the S. Typhimurium ST313 lineage 2 isolates, including the reference strain D23580. This pseudogenization event created a stop codon that truncates the MacB protein ( Fig 1F). This truncation interrupts the N-terminal ATP-binding domain and prevents translation of the transmembrane domains that extend from the cytoplasm into the periplasmic space ( Fig 1F); assuming an analogous structure to E. coli MacB, the truncated MacB protein in ST313 Lineage 2 is predicted to be unable to interact with MacA in the context of the MacAB-TolC efflux pump [32]. The recently described African S. Typhimurium ST313 sublineage 2.1 [13] retains the macA C➝T SNP associated with other ST313 while also having an A➝G SNP in the 5' untranslated region (5'-UTR) of macA ( Fig 1B). Sublineage 2.1 and 2.2 isolates do not carry the macB mutations associated with lineage 1 or 2 isolates (Fig 1A).
The presence of multiple variations at the macAB locus of African ST313 suggested reductive or adaptive evolution has occurred [25]. We focused on characterizing macAB variants found in the most recently isolated African S. Typhimurium ST313 lineages 2 and sublineage 2.1 that are associated with invasive disease (Fig 1C). For clarity, these mutations are designated: macA C➝T for the non-synonymous SNP shared across all ST313 lineages; macB indel for the SNP present in ST313 lineage 2 that introduces multiple stop codons due to a frameshift and truncates the MacB protein; and 5'-UTR macA Lin2.1 for the A➝G SNP within the 5'-UTR of macA of ST313 sublineage 2.1 isolates. We use macA ST19 and macB ST19 to refer to the alleles that are carried by S. Typhimurium ST19.

Expression of macAB is promoted by PhoP in Salmonella Typhimurium
The macAB (ybjY-ybjZ) genes are operonic with the transcriptional start site (TSS) located 504 nucleotides upstream of the translational start of the macA gene [36]. This is a particularly long 5'-UTR, and regions of this type have previously been shown to play important regulatory roles in Salmonella [42,43].
Since previous studies suggested that macAB is important for S. Typhimurium virulence [35,38], we wanted to clarify how expression of this locus is regulated. In Salmonella the two-component system PhoP/Q senses low magnesium, acidic pH, and antimicrobial peptide disturbance of the inner membrane. PhoP up-regulates a set of genes that increase cellular resistance to antimicrobial peptides and promote survival in macrophages [44][45][46][47][48]. Previously, PhoP was shown to physically bind to a PhoP-box upstream of the macA coding sequence, with the authors concluding that PhoP represses macAB transcription [35].
We previously published the RNA-seq-based transcriptomic profiles of S. Typhimurium ST19 strain 4/74 during growth in multiple in vitro conditions and during intramacrophage replication [36,49]. The major pathogenicity locus SPI-2 is important for survival of Salmonella in mammalian phagocytic cells, and expression of SPI-2 genes can be induced by defined media that mimic some conditions of the vacuolar environment (InSPI2). Our published data show that S. Typhimurium 4/74 increases transcription of macA and macB in low magnesium InSPI2 medium, consistent with a role for stimuli sensed by PhoP/Q in promoting macAB expression (Fig 2A). More recently, we published the transcriptomic profiles of the macAB genes of S. Typhimurium ST313 strain D23580 using the same environmental conditions, including intra-macrophage replication [18], summarized at https://tinyurl.  Gene expression in InSPI2 media in response to environmental stimuli, fold-change relative to InSPI2 medium alone. (B) S. Typhimurium intramacrophage gene expression measured from RAW264.7 macrophages after 8 hours of infection, fold-change relative to expression in LB at early stationary phase. Panels A and B show RNA-seq data extracted from the SalComD23580, SalComMac, and SalComRegulon databases reported previously [18,36,49]. (C) ST19 4/74 and 4/74 phoP null mutants (phoP::Tn10) with chromosomally integrated lacZY transcriptional fusions for either phoN and macAB were grown to mid-exponential phase in N minimal medium, pH 7.4 and high (10mM) MgCl 2 before transfer to the same or low (10μM) MgCl 2 media and growth for 90 minutes. β-galactosidase production was measured by a kinetic Miller assay. (D) At mid-exponential phase, transcriptional fusion strains were shifted from pH 7.4 and 10mM MgCl 2 to media buffered to the indicated pH with high (10mM) or low (10μM) MgCl 2 and grown for 90 minutes. (E) Transcriptional fusion strains at mid-exponential phase were exposed to the antimicrobial peptide C18G (5μg/mL) in N minimal medium at pH 7.4 with 1mM MgCl 2 and grown for 90 minutes. Data are from three repeat experiments. ANOVA with Tukey post-test (C, D), or t-test (E); bar = mean; error = standard deviation.

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates com/macAB-SalCom474-D23. In data sets from both ST19 isolate 4/74 and ST313 lineage 2 isolate D23580, the macAB transcript is upregulated during replication in RAW264.7 macrophage-like cells, compared to early stationary phase (ESP) growth in LB medium ( Fig 2B). Furthermore, we previously observed that deletion of phoPQ reduced macAB transcription by about half in InSPI2 medium [50].
To examine macAB transcriptional regulation more closely, we created chromosomal lacZY transcriptional fusions in S. Typhimurium ST19 isolate 4/74 driven by the endogenous phoN or macAB promoters. We grew these reporter strains overnight in defined N minimal medium at pH 7.4 with high (10mM) magnesium, a condition that represses PhoP activity [51]. After subculture and growth to mid-exponential phase, we shifted cells to low (10μM) magnesium medium, a treatment that up-regulates expression of the PhoP-regulated gene phoN [52]. We found that β-galactosidase levels increased for both phoN::lacZY (positive control) as well as macAB::lacZY when cells were shifted to low magnesium medium ( Fig 2C). Furthermore, the up-regulation of these genes was impaired when phoP null mutants were shifted to low magnesium medium ( Fig 2C). In addition, when cells with the phoN and macAB transcriptional fusions were shifted to increasingly acidic conditions (from pH 5.8 to pH 4.9), levels of β-galactosidase activity were significantly increased ( Fig  2D), in agreement with previous reports that acidic pH stimulates expression of PhoP/Qdependent genes [53,54].
Cationic antimicrobial peptides induce expression of PhoP-regulated genes [52] and play bacteriostatic and bactericidal roles during Salmonella infection of macrophages [55]. To test whether macAB gene expression is upregulated by antimicrobial peptides, we grew the phoN and macAB transcriptional fusion strains in moderate (1mM) magnesium followed by treatment with a sub-inhibitory concentration of the salt-insensitive, cationic antimicrobial peptide C18G [52]. Exposure to C18G induced higher β-galactosidase activity in the phoN and macAB transcriptional fusion strains ( Fig 2E). These experiments collectively show that macAB transcription is facilitated by PhoP under biologically relevant conditions.

African S. Typhimurium ST313 macAB variants influence replication in macrophages
We next wanted to determine if the genetic changes in the macAB locus of the ST313 lineages commonly associated with invasive disease in Africa had functional consequences during infection. We focused our experiments on lineage 2 and sublineage 2.1 macAB variants, reasoning that the lineage 1 macB STOP would have effects that are similar to the lineage 2 macB indel . Furthermore, ST313 lineage 1 isolates are no longer causing a clinical problem in Africa [12]. To ensure otherwise native regulation of the macAB genes, we introduced marker-less nucleotide changes directly into the macAB locus of the ST19 isolate 4/74 or the ST313 lineage 2 isolate D23580. The 4/74 macA C➝T mutant was made to test the impact of this SNP alone, the 4/74 macA C➝T macB indel mutant to represent the lineage 2 macAB genotype, and the 4/74 5'-UTR macA Lin2.1 macA C➝T mutant to represent the lineage 2.1 macAB genotype ( Fig 1C). We additionally made a 4/74 5'-UTR macA Lin2.1 mutant to test whether this SNP alone affects pathogenesis. Similarly, we modified D23580 to test the role of individual macAB SNPs in modulating virulence. The ST313 lineage 2 macB indel was first removed, yielding D23580 macA C➝T macB ST19 . This was followed by alteration of the macA C➝T SNP, creating a D23580 strain with the full ST19 genotype (D23580 macA ST19 macB ST19 ). We confirmed that our engineered strains carried the desired nucleotide modifications, with no unintended mutations elsewhere in the chromosome, by whole genome sequencing (see S2 Table and Materials and Methods).

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates Previous studies have shown that S. Typhimurium ST19 strains 14028S and SL1344 with deletions of macAB replicate poorly within mouse macrophages [38,56]. We infected the murine RAW264.7 macrophage cell line with a range of our marker-less macAB mutants to assess Salmonella intracellular replication. As expected, we found that ST19 4/74 phoP and macAB null mutants showed significantly reduced replication in macrophages compared to parental 4/74 ( Fig 3A). Replication of the 4/74 macA C➝T mutant was significantly lower than the parental 4/74 and similar to replication of the macAB null mutant (Fig 3A) showing this ST313-associated SNP likely impairs MacAB-TolC functionality during intramacrophage replication. The 4/74 macA C➝T macB indel mutant (ST313 lineage 2 genotype) also showed lower replication than parental 4/74 ( Fig 3A). Interestingly, the 5'-UTR macA Lin2.1 SNP alone significantly reduced replication of 4/74, while the 4/74 5'-UTR macA Lin2.1 macA C➝T mutant with the full ST313 sublineage 2.1 genotype replicated at a level similar to the macAB null mutant ( Fig  3A). We also assessed replication of S. Typhimurium D23580, the reference ST313 lineage 2 African isolate, in murine RAW264.7 macrophages. D23580 replicated extensively in RAW cells to a higher level than 4/74 ( Fig 3A and 3B), as previously reported [17]. We made marker-less point mutations in D23580 to change the macA C➝T and macB indel SNPs to the respective ST19 alleles. The D23580 macA C➝T macB ST19 and D23580 macA ST19 macB ST19 mutants showed no change in fold-replication in this experimental setting when compared to the parent D23580 ( Fig 3B).
Taken together, these results show the macAB-associated SNPs of S. Typhimurium ST313 lineages do have phenotypic consequences in macrophage replication, though the replication effect may be epistatic. The ST19 strain 4/74 shows reduced levels of intra-macrophage replication when macAB is modified to the various ST313 genotypes, suggesting that these SNP mutations impair MacAB-TolC function. In contrast, an African ST313 lineage 2 isolate maintains its level of intra-macrophage replication irrespective of macAB genotype. This implies that D23580 has a macAB-independent mechanism for its enhanced replication phenotype in RAW macrophage-like cells.

ST19 macAB provides resistance to the cationic antimicrobial peptide C18G
Previous studies have shown that MacAB homologues can contribute to increased resistance to antimicrobial peptides in other bacteria [32]. In E. coli, the TolC-dependent secretion of the helical, amphiphilic peptide enterotoxin II is facilitated by MacAB, but not by other TolC-interacting partners [57]. Additionally, another ABC-type efflux pump (MtrCDE) improves Neisseria resistance to both macrolide antibiotics and antimicrobial peptides [58]. Since we showed macAB expression in S. Typhimurium is regulated by the PhoP/Q system that is important for antimicrobial peptide resistance, we hypothesized that MacAB of Salmonella would also support replication in the presence of amphiphilic cationic antimicrobial peptides.
We took a reductive approach to quantify the impact of macAB genotype on antimicrobial peptide resistance. We reasoned that the contribution of MacAB to antimicrobial peptide resistance could be obscured in vitro by the profound PhoP-induced membrane modifications that dramatically slow antimicrobial peptide interactions with the cell envelope [59]. Furthermore, the constitutively expressed AcrAB-TolC pump effluxes a wide variety of compounds and can mask the contributions of other efflux pumps like MacAB during in vitro tests of antibiotic resistance [34,60]. Thus, we compared the contributions of MacAB variants to Salmonella growth in the presence of the antimicrobial peptide C18G using 4/74 phoP acrAB macAB null mutants harboring low-copy plasmids constitutively expressing macA ST19 macB ST19 (the ST19 genotype), macA C➝T macB ST19 (the ST313 macA SNP alone), or macA C➝T macB indel (the ST313 lineage 2 genotype). We found that the macA ST19 macB ST19 plasmid facilitates growth of the 4/74 phoP acrAB macAB null mutant in minimal medium in the presence of 2μg/mL C18G (Fig 4A), while cells expressing macA C➝T macB ST19 , macA C➝T macB indel , or carrying empty plasmids show much longer lag times (Fig 4A). With a subinhibitory level of C18G treatment (1μg/mL), all strains grew equally well, indicating the differences in lag time are not due to toxicity from expression of MacAB variants ( Fig 4B). Comparison of lag times (Fig 4, inset table) suggests that the macA C!T SNP impairs resistance to C18G while the macB indel further disables the MacAB-TolC channel. These data demonstrate that the ST19 macAB genotype assists growth in the presence of inhibitory concentrations of C18G and can do so independently of other PhoP-induced genes.

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates

MacAB contributes to the ability of the S. Typhimurium ST19 isolate 4/74 to outcompete the iNTS ST313 isolate D23580 in the gut
Previous work has shown that MacAB promotes survival of Salmonella in the mouse gut after oral infection of C57BL/6 mice [38]. In addition, gut colonization is known to induce transcription of PhoP/Q regulated genes in Salmonella [61]. We thus hypothesized that the macAB mutations acquired by African S. Typhimurium ST313 isolates may impact fitness in the gut. To test this, we pretreated C57BL/6J mice with streptomycin and orally infected the next day with an equal mixture of 4/74 and the ST313 lineage 2 isolate D23580. We measured cecum and colon CFUs at day 2 after infection to calculate a competitive index (CI) and found that the ST19 strain 4/74 outcompeted D23580 by~20-fold ( Fig 5A). To test whether macAB contributed to this fitness difference, we competed macAB null mutants of 4/74 and D23580. The relative fitness of D23580 in gut tissues improved by approximately 5-fold when both isolates lacked macAB (Fig 5A), indicating that the ability of 4/74 to outcompete D23580 in the gut was partly macAB-dependent. To test whether the ST19 macAB genotype modulated D23580 fitness in the gut, we competed unmodified 4/74 with the D23580 macA ST19 macB ST19 mutant. We found that altering macAB locus SNPs to the macA ST19 macB ST19 genotype in D23580 improved its relative fitness in the cecum, colon, and feces when in competition with 4/74 ( Fig  5B). These data show that the lower fitness of lineage 2 isolate D23580 in the mouse gut is partly due to its macAB genotype.
For S. Typhimurium, the type-3 secretion systems (T3SS) encoded by the SPI-1 and SPI-2 pathogenicity islands are important for inducing colitis during gut infection [62,63]. Previous work has shown S. Typhimurium ST313 isolates from Africa induce less host inflammation due to lower SPI-1-mediated invasion activity and reduced levels of flagellin expression when compared to ST19 strains [19,26,64]. Since macAB was shown to play a role in the streptomycin pre-treatment model of colitis [38], we suspected that 4/74 could be using SPI-1 and/or  Table) were grown in N minimal medium, pH 7.4 and 1mM MgCl 2 . Overnight stationary phase cells were washed and normalized to OD600 = 1 before 1:200 final dilution into fresh N minimal medium with 2μg/mL C18G (A) or 1μg/mL C18G (B). OD600 was monitored over time using a BioTek Synergy HTX plate reader. Growth curves presented here are from one experimental run and representative of three independent experiments. Plotted data points are the geometric mean of quadruplicate or triplicate microplate wells. Lag time (inset table) was determined as time to reach OD600 = 0.15. https://doi.org/10.1371/journal.ppat.1008763.g004

Host genetics shape the utility of macAB genotypes for D23580 systemic infection
Since we saw no impact of the ST19 macAB genotype on D23580 replication in RAW264.7 cells, we predicted that ST313 macAB-associated SNPs might be dispensable or otherwise have

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates no impact on the ability of D23580 to spread to systemic tissues. To exclude the contributions of known differences between ST19 and African ST313 in their inflammatory and disseminating behaviors in vivo [19,24,26,64], we competed D23580 with isogenic D23580 macAB mutants. To calculate a competitive index, we created a marked D23580 strain by placing a kanamycin resistance cassette at an intergenic site on the D23580 chromosome (see Materials and Methods) that was chosen for its low transcriptional activity when assessed by RNA-seq under a variety of growth conditions. After oral infection of C57BL/6J mice with the marked D23580-Kan R strain and the parent D23580, we observed a competitive index of 1 in the gut and systemic sites, showing that the intergenic kanamycin marker insertion did not affect fitness in this model (Fig 6A). We found that the D23580 strain modified to the macA ST19-macB ST19 genotype was outcompeted more than 4-fold by the isogenic D23580-Kan R strain in systemic sites (liver and spleen), while showing no difference in the cecum (Fig 6A). This implies the macA C➝T macB indel genotype provides an advantage to D23580 during systemic infection.
One important hypothesis regarding the emergence of African iNTS lineages is that the immune status of certain human populations in sub-Saharan Africa provides a permissive niche for systemic S. Typhimurium infection, thus uniquely shaping Salmonella evolution [15,[65][66][67]. In our infection experiments with RAW264.7 macrophages (BALB/c origin) and C57BL/6J mice, the host gene Nramp1 is not functional. BALB/c and C57BL/6J mice have a Glycine to Aspartic Acid (G169D) mutation in Nramp1 (i.e. a genotype of Nramp1 D169/D169 ) which inactivates Nramp1 to yield a more permissive environment for Salmonella replication within macrophages [68]. Nramp1 dramatically restricts intracellular bacterial infection at systemic sites through the removal of magnesium and other divalent cations from the vacuolar environment [68][69][70][71][72]. However, Nramp1 has further influence on the host immune response. For example, Nramp1 +/+ mice have more rapid innate responses than isogenic Nramp1 -/mice, with higher levels of interferon γ (IFN-γ) and increased influx of neutrophils during the streptomycin pretreatment model of Salmonella-induced colitis [73] and in the dextran sodium sulfate (DSS)-induced colitis model [74]. Furthermore, Nramp1 +/+ dendritic cells produce more inflammatory cytokines than Nramp1 -/dendritic cells during Salmonella infection [75], an important route for the early, rapid dissemination of D23580 into the mesenteric lymph nodes [24]. These published data suggest that, in addition to the role of Nramp1 in control of intramacrophage replication of Salmonella, many relevant parameters of the immune response differ between Nramp1 -/and Nramp1 +/+ mice, especially in the amount of IFN-γ produced [76]. Furthermore, IFN-γ, in conjunction with other stimuli, induces maximal upregulation of Nramp1 transcription [9]. Thus, Nramp1 genotype has pleiotropic effects on the course of Salmonella infection. Given our results in a mouse background with defective Nramp1 (Fig 6A), we thus sought to test the effects of D23580 macAB genotypes during infection in mice with a more robust immune response.
Overall, our murine infection and competition experiments show that S. Typhimurium ST313-associated macAB SNPs and indels influence Salmonella fitness depending, in part, on

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates the host Nramp1 genotype. During gut infection, the ST19 macAB genotype provides a fitness benefit to D23580 in competition with an ST19 isolate in permissive Nramp1 -/mice (Fig 5). We also show that the impact of the ST313 lineage 2 macAB genotype upon systemic colonization by D23580 is Nramp1-dependent ( Fig 6A).

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates

Discussion
We have provided new insights into the function of the MacAB-TolC channel in Salmonella Typhimurium pathogenesis by exploring the consequences of variations in the 5'-UTR and coding sequence of macAB that are found within ST313 lineages associated with invasive disease in Africa. We show that macAB expression is influenced by PhoP and that the ST19 macAB genotype improves functional resistance to the antimicrobial peptide C18G. ST313-associated variations in macAB genotype can affect antimicrobial peptide resistance as well as replication in a permissive macrophage-like cell line. In mice, the ST313-associated macAB gene variants contribute to the lower relative fitness of ST313 lineage 2 isolate D23580 in competition with the ST19 strain 4/74 in inflamed and uninflamed settings in the gut.
Previously, macAB mutants of S. Typhimurium were reported to show impaired oxidative stress resistance [38]. However, S. Typhimurium macAB null mutants of 4/74 did not show survival defects after peroxide treatment when compared to the parent strain (S3A Fig). Furthermore, sub-lethal concentrations of hydrogen peroxide did not induce transcription of macAB after short-term in vitro exposure (Fig 2A) or in time course assays with the 4/74 macAB::lacZY transcriptional reporter strain (S3B Fig). Poly-specificity allows efflux pumps to take on various functional roles since net cellular resistance to toxic compounds results from redundancies in efflux pumps and dynamic properties of the outer membrane [77]. Given the poly-specificity of efflux channels like MacAB-TolC, it is likely that, in addition to antimicrobial peptides, some bacterially-generated molecules are exported to directly quench reactive oxygen species (ROS), as proposed by Bogomolnaya and colleagues [38,78]. However, this oxidative stress resistance function of MacAB may depend upon experimental conditions and the bacterial genetic background.
We thus do not exclude a direct role for macAB in oxidative stress resistance in Salmonella but propose that an additional function is to counteract the effects of antimicrobial peptides encountered in both the gut and systemic sites. Indeed, bacterial responses to oxidative stress and antimicrobial peptide exposure are tightly linked by the PhoP/Q two-component regulatory system [79,80]. Antimicrobial peptide exposure induces transcription of genes in the RpoS regulon that assist bacterial adaptation to oxidative stress [81,82]. We have shown that macAB variants confer distinct levels of antimicrobial peptide resistance independently from other PhoP-mediated effects. Notably, on its own the macA C➝T SNP which leads to the S 174 L mutation in MacA reduces the utility of MacAB in resistance to C18G (Fig 4). The role played by the PhoP/Q system in upregulating transcription of macAB is consistent with the MacAB-TolC channel of Salmonella countering antimicrobial peptide exposure, either through direct efflux or by assisting translocation of other bacterial factors important to cope with antimicrobial peptide stress.
Although one publication suggested PhoP repressed macAB transcription in Salmonella after low magnesium treatment [35], other published data suggest that macAB is positively regulated by PhoP. We have previously noted that phoPQ null mutants have lower expression of macAB when grown in InSPI2 medium [50], while others have seen that S. Typhi production of MacA protein under low magnesium conditions is PhoP-dependent [83]. Furthermore, macAB is important for S. Typhimurium replication in macrophages [38,56], a niche where PhoP/Q signaling is active and important for Salmonella survival [55,61]. In addition to the different Salmonella isolates used across studies, our results suggesting PhoP promotes macAB transcription may arise from differences in experimental conditions, the importance of regulatory kinetics such as PhoP turnover [38], changes in ribosomal RNA levels in response to low magnesium [84], or coordinated activity with other transcriptional regulators [59]. We conclude that PhoP-inducing signals generally serve to maximize macAB transcription in Salmonella.

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates The macAB locus contains additional regulatory complexity that requires further investigation. While we show that the 5'-UTR macA Lin2.1 SNP is sufficient to impair replication of 4/74 in macrophages (Fig 3A), the mutation did not affect the level of macAB transcription in low magnesium medium (S4 Fig). The mapped transcriptional start site (TSS) of macA is 504 nucleotides upstream of the start codon and is located within the divergently-transcribed gene ybjX (also known as somA) (see Fig 1B and Fig 1B). We observed that expression of MacAB from a high-copy plasmid (pUC19) or a plasmid with an arabinose-inducible promoter is toxic to cells, a lethality phenomenon reported by others [29]. Additionally, MacA exhibits high binding affinity for TolC in vitro, suggesting it may outcompete other TolCinteracting efflux pumps whenever MacAB is expressed [87]. Taken together, these observations support the conclusion that the tight transcriptional regulation of macAB functions to provide utility without compromising other cellular functions.
We have not demonstrated the direct structural consequences for the MacAB-TolC channel that result from the macAB SNPs present in ST313 isolates. Such changes might include assembly kinetics of the MacA hexamer, stability or mechanical function of the tripartite complex, or perhaps altered association with other important constituents of the Gram-negative cell envelope [37]. The E. coli MacA protein can bind tightly to rough LPS via residues on the N-terminus with affinity that is higher than that of polymyxin B, prompting the suggestion that rough LPS may be a cargo of the channel [29]. Given that PhoP is required for maximal macAB transcription and that MacAB-mediated resistance to C18G did not require other PhoP-regulated genes (Fig 2B and 2C), we speculate that rough LPS-binding by MacA is a structural feature important for its antimicrobial peptide resistance function. While the ST313 macA C➝T SNP resulting in the S 174 L mutation impairs MacAB-mediated resistance to C18G, it is possible this SNP variant improves efflux of other substrates not tested here and thus might serve a direct role in the pathogenic adaptation of African iNTS. Further structural investigation of the Salmonella MacAB-TolC channel will provide new insights into the role of this important efflux pump in Salmonella physiology and host-pathogen interactions.
It is not immediately clear why S. Typhimurium ST313-associated macAB gene variants lead to different infection outcomes depending on strain background; however, some possibilities may be proposed based on current knowledge. In the case of D23580 and other ST313 lineage 2 isolates, the mutation responsible for the marked increase in expression of the PhoPregulated protease PgtE [88] could sufficiently compensate for antimicrobial peptide resistance otherwise provided by the ST19 macAB genotype. In this case, the ATP-dependent MacAB--TolC channel might siphon energy or even directly interfere with other genes or behaviors unique to the D23580 strain background. Such loss-of-function mutations that favor the activity of other pathogenesis mechanisms have been noted in Salmonella evolution. For Salmonella Typhi, a stop codon that interrupts fepE prevents very-long O-antigen production [89], which in turn permits the horizontally-acquired Vi capsule to have maximal immune evasive effects during systemic infection [90]. Finally, it has been shown that point mutations in the highly expressed AcrAB-TolC efflux pump can alter expression of a variety of Salmonella pathogenic genes indirectly without affecting expression of other efflux pumps [91]. Further research will be required to identify other Salmonella genes that might interact with the macAB genotype to influence pathogenesis.

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates In single infections, S. Typhimurium D23580 and other ST313 isolates readily colonize the intestinal tracts of a variety of animals including primates, rodents, and chickens [23,24, [92][93][94]. A recent study of competitive fitness in the streptomycin-pretreatment model of mouse colitis showed that D23580 can outcompete the ST19 isolate IR715, a strain that is derived from the common laboratory reference isolate 14028S [23], further confirming the ability of ST313 to infect multiple animal species. We show that differences in macAB genotype are partly responsible for a relative fitness defect of D23580 during competition with the ST19 isolate 4/74 in the inflamed and uninflamed gut (Fig 5A, 5B and 5E). Several factors could contribute to these differences in competitive fitness of D23580 versus certain ST19 strains. Both D23580 and IR715 do not express sopE, a SopEF prophage-encoded virulence factor delivered by the T3SS of SPI-1 [19,26]. In contrast, sopE is present in 4/74 and the closely-related ST19 isolate SL1344 [95]. sopE is known to drive substantial cecal inflammation in the streptomycin pretreatment model of colitis [96]. Further, the ST19 strains SL1344 and 4/74 contain the pCol1b plasmid and can produce colicin Ib, potentially helping Salmonella outcompete other Enterobacteriaceae in the inflamed gut [97]. We found that ST19-associated macAB gene variants increase fitness of D23580 in scenarios with robust gut inflammation, such as in competition with a S. Typhimurium ST19 isolate that has sopE [96] and in mice with Nramp1 functionality [73]. However, when D23580 infects a more permissive host, such as Nramp1 -/-C57BL/6J mice, its ST313 lineage 2 macAB variant is advantageous for systemic infection ( Fig  6B). Thus, D23580 shows a degree of specialization for systemic infection that depends upon both its macAB genotype and parameters of the host innate immune response. Our experiments reinforce the value of testing Salmonella isolates in various host genotypes to identify potentially important host-pathogen gene interactions.
Although no studies have specifically linked NRAMP1/SLC11A1 polymorphisms to the susceptibility of humans to invasive Salmonella, several NRAMP1 polymorphisms found within West African populations are strongly linked to susceptibility to intracellular infection with Mycobacterium tuberculosis [98,99]. We note that numerous parameters in mammalian hosts that are related to susceptibility to invasive Salmonella disease are also compromised in Nramp1-deficient mice, as summarized below. For example, macaques that were previously infected with simian immunodeficiency virus have diminished IL-17 production that reduces the influx of neutrophils into the gut, thus compromising control of Salmonella dissemination from the gut [100]; similarly, neutrophil influx into the gut during Salmonella colitis is delayed in Nramp1 -/mice [73]. Nramp1 also regulates the oxidative burst capacity of phagocytes [101]. A reduced oxidative burst response is observed in Gambian children recovering from malaria infection [102] and is likely a consequence of malaria-induced hemolysis as revealed by experiments in mice [65].
In humans, STAT4 variants have been linked to increased risk for invasive Salmonella infection. Specifically, a genome-wide-association study identified a STAT4 polymorphism in children in Kenya and Malawi which correlated with reduced IFN-γ responses of ex vivo stimulated immune cells [103]. The importance of IFN-γ to control of Salmonella infection at both acute and chronic stages has been clearly demonstrated [104], and upregulation of NRAMP1 in macrophages is part of the response to IFN-γ [105]. Malaria, malnutrition, and HIV can compromise innate resistance to intracellular bacterial infection through a wide variety of mechanisms, including alterations in macrophage responsiveness to IFN-γ [106]. Based on outcomes of our competitive infections using Nramp1 -/and Nramp1 +/+ mice, we speculate that the S. Typhimurium ST313-associated macAB gene variants represent adaptive evolution for systemic infection when restriction of Salmonella dissemination is compromised in the host.
Oral infection of mice with Salmonella Typhimurium after streptomycin pretreatment causes substantial inflammation and colitis, which depends on the function of SPI1 and SPI2-encoded type three secretion systems [107]. Although D23580 exhibits less T3SS-1 mediated invasion in vitro and in animal infections [19,26], we found that D23580 was outcompeted by 4/74 in the mouse gut whether or not both isolates lacked T3SS-1 or T3SS-1/2, and this fitness defect was partly macAB-dependent (Fig 5D and 5E). Although all of these experimental models vary in degree and type of inflammation, they share a diverse and dynamic set of host-and microbe-generated antimicrobial peptides. We suggest that MacAB contributes to Salmonella pathogenesis by countering antimicrobial peptide stress as part of the bacterial response orchestrated by the PhoP/Q two-component system.
The S. Typhimurium ST313-associated macAB variants we have characterized suggest a pattern of evolutionary convergence toward a degraded function of the MacAB-TolC efflux pump. While MacAB remains functional amongst S. Typhimurium lineages associated with gastroenteritis, we conclude that inactivation of the MacAB system within African ST313 lineages represents a unique adaptation that may facilitate systemic infection of permissive hosts.
The macA 5'-UTR, macA, and macB sequences of strain 4/74 were used to generate BLAST databases with BLAST 2.9.0+ [115]. Raw reads of all the strains were assembled using Unicycler v0.4.8 [116]. The quality of assembly was checked by Quast v5.0.2 [117]. The N50 value of all assemblies was >20kb, and the number of contigs was <600. The genome assemblies were queried against the databases using the BLASTn algorithm.

Bacterial strains, plasmids, primers and growth conditions
See S2 Table for bacterial strains used in this study. Where applicable, P22 HT105/1 int-201 phage was used to move marked mutations from previously-generated S. Typhimurium
See S3 Table for plasmids and S4 Table for

Lambda (λ) red recombination for marked mutants
Marked mutants were made based on the λ red procedure [118] using the temperature sensitive pSIM5-Tet plasmid that contains a temperature shock inducible promoter driving recombinase expression [20]. Briefly, primers with 40bp homology targeting the flanking regions of gene to be deleted were used to amplify the FRT-flanked kanamycin cassette from pKD4. 4/74 or D23580 / pSIM5-Tet cells were grown with streptomycin and tetracycline at 30˚C overnight in LB Lennox before subculturing 1:100 into fresh mediumand growing to an OD of 0.4. The culture was then incubated at 42˚C shaking for 15 minutes before placing on ice make cells competent for electroporation. Cells were washed with chilled double-distilled water, followed by electroporation of 800ng of purified PCR product. Cells were recovered in SOC for 1.5 hours shaking at 30˚C before pelleting and plating on LB kanamycin plates. After overnight growth at 37˚C, individual colonies were struck across fresh plates to purify single colonies, followed by colony PCR to confirm correct insertion of the kanamycin cassette. For 4/74, marked mutants were moved into a clean 4/74 strain background using P22 HT105/1 int-201 phage, and the correct insertion was confirmed by PCR. For construction of D23580-Kan R we amplified the kanamycin resistance cassette from pKD4 using primers del_23_F and del_23_R for insertion by λ red recombination as described above. These primers target the Kan R cassette to the intergenic region between STMMW_41451 and STMMW_41461 between coordinates 4441510 and 4441511 in D23580 (GenBank: FN424405.1). This region is part of the remnant prophage Def4 [20] and not transcribed in RNA-seq datasets in variety of in vitro conditions [18]. Correct insertion of the KanR cassette was confirmed by primers Fw_STM4196-7_ext and Rv_STM4196-7_ext. D23580-Kan R was confirmed to be cured of temperature sensitive pSIM5-Tet after growth at 37˚C by testing isolated colonies for tetracycline sensitivity.

Construction of lacZY reporter strains
Chromosomal β-galactosidase (lacZY) transcriptional fusions were made using the method of Ellermeier and colleagues [119]. 4/74 marked mutants with FRT-flanked kanamycin cassettes derived from pKD4 were electroporated with 200ng pCP20, recovered for 1 hour in SOC at 30˚C before plating dilutions on LB streptomycin chloramphenicol plates, followed by growth at 30˚C overnight. Colonies were re-struck to purify for single clones, then patched onto LB kanamycin and LB streptomycin chloramphenicol plates to confirm loss of kanamycin by FLP-recombinase activity. Two kanamycin sensitive colonies were picked and grown in LB Lennox at 30˚C, followed by electroporation with 200ng pCE36 plasmid purified from the Salmonella pir+ strain JS198/pCE36 [119]. Cells were recovered in SOC at 37˚C and plated on LB

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates kanamycin to select for pCE36 integration. Purified colonies were screened by PCR using P1 and Lac primers for integration of pCE36 into the FRT site [119].

Scarless mutant generation
In order to study the influence of these nucleotide changes on Salmonella pathogenesis without otherwise altering macAB regulation or introducing other changes in the genome, we used scarless mutagenesis to make nucleotide changes in the genome of a given Salmonella isolate. We used the suicide plasmid pEMG system described by Martínez-García and de Lorenzo [120] and as applied to Salmonella Typhimurium by Owen and colleagues [20]. Genomic DNA from 4/74 or D23580 was used as template for Phusion (ThermoFisher) PCR of 1600bp flanking the polymorphism to be transferred. For marker-less deletions, 800bp flanking each side of the region to be deleted were amplified to generate two PCR products. Primers were designed to include~20nt overhangs to permit Gibson Assembly of a single 1600bp fragment (for marker-less nucleotide changes) or the two 800bp flanking regions (for marker-less deletions) into pEMG that was previously digested with XbaI and KpnI-HF (New England Biolabs, NEB). Purified PCR product (Qiagen) and digested pEMG backbone were assembled with the DNA HiFi Assembly MasterMix (NEB). The 5'-UTR macA Lin2.1 SNP was incorporated into pEMG::macA ST19 or pEMG::macA C➝T by PCR using primer pairs 1309,1310 and 1311,1312; the two PCR products were assembled with HiFi DNA Assembly MasterMix with pEMG previously digested with XbaI and KpnI-HF. A given pEMG::X plasmid was mobilized from E. coli S17-1 λpir into recipient Salmonella by conjugation, followed by selection for transconjugants on M9 minimal agar plates formulated with 0.2% glucose, 1mM MgSO 4 , and 40μg/mL kanamycin. Merodiploid transconjugants were resolved by electroporation of pSW-2 as previously described [20], followed by colony PCR and standard Sanger sequencing of PCR product to identify clones with the intended point mutations. Confirmed mutants were cured of unstable pSW-2 by several passages in LB before patch plating to confirm loss of pSW-2 by gentamicin sensitivity.
To confirm that key strains contained the intended engineered nucleotide(s) and no unintended mutations in other parts of the genome, the four strains 4/74 macA C➝T , 4/74 macA C➝T-macB indel , D23580 macA C➝T macB ST19 and D23580 macA ST19 macB ST19 were genomesequenced using Illumina paired-end sequencing (SNPsaurus, Oregon) as indicated in S2 Table and aligned to reference genomes (NCBI) using CLC Genomics Workbench (Qiagen).

Plasmid construction
The macAB sequence was amplified from genomic DNA templates with Phusion polymerase (ThermoFisher) using primers 1223 and 1224. The PCR product and pBAD33.1 (Addgene #36267) were digested with NdeI and HindIII-HF (New England Biolabs, NEB) followed by T4 ligation and transformation into NEB 10beta competent cells. The cassette inclusive of the T7 ribosome binding site, macAB, and the two transcriptional terminators was amplified with Phusion polymerase using primers 1272 and 1273 from pBAD33.1-macAB templates. The PCR products and destination pACYC177 were digested with ScaI and PstI, purified using the PCR purification kit (Qiagen), and ligated with T4 polymerase before transformation into NEB 10beta competent cells, selecting on kanamycin. The promoter for the Amp R gene thus drives expression of the macAB gene inserted at the ScaI site within the Amp R sequence of pACYC177.

Stimulation conditions for PhoP-regulated gene expression
The procedures here are based on methods described by Bader and colleagues [52]. For lowmagnesium treatment, cells were grown at 37˚C overnight in N minimal medium with

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates 100mM Tris-HCl pH 7.4, 0.2% deferrated Casamino acids (Chelex treated), 0.2% glycerol, and 10mM MgCl 2 (high magnesium) with relevant antibiotics. Cells were subcultured 1:100 into fresh medium and grown to OD600~0.2, or about 4 hours. Cells were washed 3x in the same medium or medium with 10μM MgCl 2 (low magnesium) and returned to the incubator, growing for a further 90 minutes. Where applicable, N minimal medium was buffered with 10mM MES at pH 5.8 or 4.9. For β-galactosidase assays, 1 mL of cell culture was then transferred to a microcentrifuge tube and placed on ice, followed by one wash in 1mL pre-chilled 100mM phosphate buffer, pH 7. A 150μL aliquot was removed and the OD600 was checked on a microplate reader (Synergy HTX, BioTek) and recorded. The aliquot was returned to the original tube and cells were pelleted again in a chilled centrifuge. The supernatant was removed and the pellet was frozen immediately on dry ice and transferred to -80˚C storage.
For C18G stimulation, cells were grown overnight in N minimal medium, pH 7.4 (see above) formulated with 1mM MgCl 2 . We titrated C18G (Anaspec) to confirm a concentration that did not inhibit growth, identifying 5μg/mL as optimal, similar to what was reported by Bader and colleagues [52]. Cells were subcultured 1:100 and grown to OD~0.2 before pelleting and resuspending in fresh media with or without 5μg/mL C18G. Cells were grown for a further 90 minutes and processed as described above.

β-galactosidase assays
We used a modified lysis protocol and a kinetic microplate assay to measure β-galactosidase activity based on the methods described by Schaefer and colleagues [121] and Thibodeau and colleagues [122]. Bacterial pellets were removed from -80˚C to room temperature, thawed briefly, then resuspended in 200μL of 100mM phosphate buffer, pH 7.0. A 200μL mixture of 20% PopCulture (Millipore) and 8U/μL of rLysozyme (Millipore #71110-4, Lot 3277983, 30U/ μL) was added to each sample and vortexed for 5 seconds. After 5 minutes at room temperature, the sample was vortexed again and incubated for a further 5 minutes. During lysis optimization tests, this freeze/thaw and lysis buffer treatment achieved a >95% reduction in OD405 within two minutes, and the suspension was visually clear within 20 seconds. In each well of a 96 well microplate, 70μL of Z-buffer with 0.05M β-mercaptoethanol was added. After lysis, 80μL of sample was added in quadruplicate wells, with control wells having lysis buffer only; a lysed pellet of wild-type 4/74 cells with no lacZY reporter served as an additional control. Using a multichannel pipette, 30μL of ONPG (Sigma) previously dissolved in 100mM phosphate buffer at 4μg/mL was added to each well. The microplate was transferred immediately to a Synergy HTX (BioTek) plate reader that was pre-equilibrated to 28˚C. OD420 was measured every minute for 90 minutes, incubating at 28˚C with shaking between reads. The change in absorbance at 420nm over time was calculated from the slope of the reads and used in the standard Miller unit calculation, where OD600 is the density of cells in 1mL before pelleting and freezing; 2.5 corrects for the pellet resuspension in a 400μL lysis volume; and 0.08 is the volume of sample (in mL) assayed per well.

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates 1 = 1x10 9 cells/mL, then diluted 1:100 into fresh medium without antibiotics. 75μL of fresh medium alone or 2x final concentration of C18G (Anaspec, Fremont, CA) in fresh medium were added to wells of a sterile, polypropylene 96-well flat-bottomed plate (Griener BioOne, Product 655261). 75μL (~7.5x10 4 cells) of the 1:100 bacterial suspension was added to wells in quadruplicate for a final 1:200 dilution of cells, with final μg/mL C18G as indicated. C18G was titrated previously, with higher concentrations (�3μg/mL) preventing growth of all strains while treatment at concentrations �1.5μg/mL showing no differences in growth kinetics compared to untreated wells. The plate was sealed with a Breathe-Easy gas permeable film (Diversified Biotech) and OD600 was measured every ten minutes while growing at 37˚C with linear shaking using a Synergy HTX plate reader (BioTek). Lag time was calculated as the time to reach OD600 = 0.150.

Macrophage infection assays
RAW264.7 cells were passaged in DMEM + 10% FBS with 4.5g/L glucose and 110mg/L sodium pyruvate and L-glutamine. For infection, cells were seeded at 2.5x10 4 cells per well in a 96 well plate starting 24 hours prior to infection. Single Salmonella colonies were picked from recently struck LB plates and grown overnight in LB Miller with appropriate antibiotics. After 16-18 hours of culture, cells were pelleted, washed once in PBS, and concentration was determined by OD600 of 1 = 1x10 9 CFU/mL. Bacteria were diluted to 1x10 8 /mL and added to DMEM complete medium for a final 2.5x10 6 CFU/mL concentration. Medium was aspirated from the RAW cells with a multichannel and replaced with 50μL of fresh media. 100μL of each inoculum (2.5x10 5 CFU, MOI = 10) was added in quadruplicate for each time point (0, 8 and 20 hours). Plates were spun at 300g for 10 minutes at room temperature (21-25˚C), then incubated at 37˚C with 5% CO 2 for 30 minutes to allow phagocytosis. Meanwhile, the inoculum was diluted 10-fold in PBS and plated to confirm the MOI. After 30 minutes, the medium was then aspirated and RAW cells were gently washed twice with 100μL medium, replacing finally with 100μL of DMEM complete with 100μg/mL gentamicin. The RAW cells were incubated at 37˚C for 60 minutes, then the high gentamicin medium was removed and replaced with medium with 20μg/mL gentamicin to suppress extracellular growth of Salmonella for the remainder of the culture period (this was t = 0). At each timepoint, medium was aspirated and RAW cells were washed 3x with 150 μL PBS before lysis in 30μL of 1% Triton X-100 at room temperature. After 5 minutes, the wells were pipetted vigorously with a multichannel, then each well was topped with 120μL of PBS. The well suspensions were mixed, serially diluted in PBS, and spot-plated to calculate CFUs per well.

Streptomycin pre-treatment model of colitis
Mice 7-12 weeks old were deprived of food briefly for 4 hours prior to gavage with 20mg streptomycin in sterile water. The next day, overnight stationary phase cultures of Salmonella strains grown with appropriate antibiotics were pelleted and washed twice in PBS before cell quantification by OD600. Twenty hours after streptomycin treatment, mice were again briefly deprived of food for 4 hours, followed by gavage of an equal mixture in PBS of 5x10 7 CFU of each strain for a total inoculum of 1x10 8 CFU. The inoculum was diluted and plated onto LB agar plates with appropriate antibiotics to differentiate strains and calculate the input ratio. Mice were euthanized by CO 2 asphyxiation at a given timepoint, tissues were removed and

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates homogenized in sterile PBS, then diluted and plated onto LB agar plates with different antibiotics to distinguish Salmonella strains. Total Salmonella CFUs were determined from streptomycin plates, while the D23580 proportion was calculated from streptomycin and chloramphenicol plates. In the isogenic D23580 competition, the addition of kanamycin identified the proportion of D23580-Kan R colonies from the total D23580 population. The output ratio in tissues was divided by the input ratio of the inoculum to compute the competitive index.

Statistical analysis
Data were analyzed with GraphPad Prism 8 (GraphPad Software, LLC). Statistical tests were performed as indicated in the figure legends.

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates S1

PLOS PATHOGENS
Variation in the MacAB-TolC efflux pump of African Salmonella isolates