Intracellular niche-specific profiling reveals transcriptional adaptations required for the cytosolic lifestyle of Salmonella enterica

Salmonella enterica serovar Typhimurium (S. Typhimurium) is a zoonotic pathogen that causes diarrheal disease in humans and animals. During salmonellosis, S. Typhimurium colonizes epithelial cells lining the gastrointestinal tract. S. Typhimurium has an unusual lifestyle in epithelial cells that begins within an endocytic-derived Salmonella-containing vacuole (SCV), followed by escape into the cytosol, epithelial cell lysis and bacterial release. The cytosol is a more permissive environment than the SCV and supports rapid bacterial growth. The physicochemical conditions encountered by S. Typhimurium within the epithelial cytosol, and the bacterial genes required for cytosolic colonization, remain largely unknown. Here we have exploited the parallel colonization strategies of S. Typhimurium in epithelial cells to decipher the two niche-specific bacterial virulence programs. By combining a population-based RNA-seq approach with single-cell microscopic analysis, we identified bacterial genes with cytosol-induced or vacuole-induced expression signatures. Using these genes as environmental biosensors, we defined that Salmonella is exposed to oxidative stress and iron and manganese deprivation in the cytosol and zinc and magnesium deprivation in the SCV. Furthermore, iron availability was critical for optimal S. Typhimurium replication in the cytosol, as well as entC, fepB, soxS, mntH and sitA. Virulence genes that are typically associated with extracellular bacteria, namely Salmonella pathogenicity island 1 (SPI1) and SPI4, showed increased expression in the cytosol compared to vacuole. Our study reveals that the cytosolic and vacuolar S. Typhimurium virulence gene programs are unique to, and tailored for, residence within distinct intracellular compartments. This archetypical vacuole-adapted pathogen therefore requires extensive transcriptional reprogramming to successfully colonize the mammalian cytosol.


Introduction
There are two major niches in which intracellular bacteria survive and proliferate after internalization into host cells, confined within a membrane-bound vacuole or free-living within the cytosol. Different pathogenic mechanisms are required to occupy these diverse environments. While bacterial pathogens have been historically categorized as being either vacuolar or cytosolic, it has recently been realized that some bacteria can occupy both niches, often in a celltype specific manner. Examples are Salmonella enterica, Mycobacterium tuberculosis and Listeria monocytogenes [1][2][3]. What is unclear is how bacteria that are adapted to survive within one intracellular niche can successfully colonize a distinct cellular compartment.
Of the foodborne bacterial, protozoal and viral diseases, non-typhoidal Salmonella enterica (NTS) cause the largest burden of illness and death worldwide [4]. Infection can cause either a self-limiting gastroenteritis or a life-threatening, invasive disease (invasive non-typhoidal salmonellosis) in immunocompromised individuals, which is particularly a public health problem in sub-Saharan Africa and south-east Asia [5]. Upon ingestion of contaminated food or water, S. enterica enters epithelial cells lining the gut and resides within a membrane-bound compartment derived from the endocytic pathway [6,7], the Salmonella-containing vacuole (SCV). Entry into non-phagocytic cells is largely governed by a type III secretion system, T3SS1, that is encoded by Salmonella pathogenicity island (SPI) 1. The T3SS acts as a molecular syringe to inject bacterial effector proteins across the eukaryotic cell plasma membrane to modulate host signaling networks that induce rearrangement of the actin cytoskeleton, leading to the formation of plasma membrane "ruffles" and engulfment of bacteria (reviewed in [8]). Establishment and maintenance of the SCV is dependent upon effectors being translocated across the vacuolar membrane by T3SS1 and another T3SS, T3SS2, which is specifically induced by environmental cues sensed by bacteria within the SCV lumen [9,10]. T3SS2 is also important for survival within phagocytic cells, which S. enterica encounters in the lamina propria during an enteric infection or in the mesenteric lymph nodes, reticuloendothelial tissues (liver and spleen) and circulating blood during invasive disease.
S. enterica resides within a membrane-bound vacuole within epithelial cells, fibroblasts, macrophages and dendritic cells (reviewed in [11]). However, the intracellular lifestyle of S. enterica differs between cell types, with a proportion of the total bacterial population living freely in the cytosol of epithelial cells, a phenomenon described for Salmonella enterica serovar Typhimurium (S. Typhimurium) and S. Typhi infections in vitro and/or in vivo [12][13][14][15][16][17]. The eventual outcome of epithelial cells harboring cytosolic bacteria is their expulsion from the monolayer into the lumen of the gut and gall bladder [12,[16][17][18][19], serving as a mechanism for bacterial spreading within and between hosts. Notably, the cytosol of macrophages is not permissive for S. Typhimurium replication [20,21], possibly due to nutrient limitation or enhanced host cell innate immune defenses such as autophagy and/or inflammasomes [21]. Considering that the site of intracellular replication is cell-type restricted, we propose that S. enterica is an "opportunistic" cytosolic pathogen.
A specialized form of autophagy, called xenophagy, protects the mammalian cytosol by targeting intracellular pathogens to autophagosomes for their eventual degradation in lysosomes (reviewed in [22]). In fact, many of the components of the autophagic machinery have been identified using S. Typhimurium as a model pathogen. In the first report of autophagic recognition, Brumell and colleagues showed that a sub-population of internalized S. Typhimurium damage their nascent vacuole in epithelial cells in a T3SS1-dependent manner and these bacteria are decorated with the autophagy marker, microtubule-associated light chain-3 (LC3) [23]. A recently identified type III effector, SopF, limits the decoration of SCVs with LC3 [24,25]. Damage exposes host glycans restricted to the vacuole lumen to the cytosol, which are then recognized by β-galactoside-binding lectins, specifically galectin-3 (GAL3), GAL8 and GAL9 [26]. Galectin binding acts as an "eat me" signal that initiates a cascade of receptor binding, phagophore formation and tethering to the bacterium, culminating in autophagosome formation [26,27]. However, autophagic control of Salmonella is temporal and incomplete, at least in epithelial cells [12,23]. Furthermore, S. Typhimurium can also use autophagy to promote replication in the cytosol of epithelial cells [28] and repair damaged SCVs in mouse embryonic fibroblasts [29]. Autophagy therefore serves both a pro-and anti-bacterial role in S. enterica infections.
Previous studies have largely investigated the infectious cycle of S. Typhimurium in epithelial cells by population-based analyses, which do not account for the heterogeneous population of intracellular bacteria. Only by determining the distinct responses of intracellular Salmonella to the cytosolic and the vacuolar niche can the infection biology of this important pathogen be accurately defined. Whilst the distinct milieus encountered within a vacuole versus the cytosol provide site-specific cues for Salmonella gene induction, little is known about what these cues might be, or which bacterial genes drive replication/survival in the cytosol. Using a combination of RNA-seq-based transcriptomics and single-cell microscopic analysis, here we describe niche-specific environments encountered by S. Typhimurium in epithelial cells and identify S. Typhimurium genes that are required for bacterial proliferation in the cytosolic compartment.

Modulation of autophagy affects bacterial proliferation in the epithelial cell cytosol
We hypothesized that the modulation of autophagy in epithelial cells would selectively perturb the cytosolic, and not vacuolar, proliferation of S. Typhimurium. Autophagy levels can be manipulated by pharmacological or genetic means (reviewed in [30]). To test our hypothesis, we used nutrient starvation to upregulate autophagy and the class III phosphoinositide 3-kinase (PI3K) inhibitor, wortmannin (WTM), to inhibit autophagy ( Fig 1A). To enumerate cytosolic bacteria after autophagy activation/inhibition, we used the digitonin permeabilization assay [12,13] to label the bacteria accessible to cytosol-delivered anti-S. Typhimurium lipopolysaccharide (LPS) antibodies at the early stages of infection (15 min-3 h p.i., Fig 1B). Wild-type bacteria were constitutively expressing mCherry from a plasmid, pFPV-mCherry. In untreated cells (basal levels of autophagy), 6.8% of bacteria were accessible to the cytosol as early as 15 min p.i. This proportion increased to~20% by 45 min p.i. and remained at a steady-state thereafter (Fig 1B). Treatment with Earle's balanced salt solution (EBSS), i.e. starvation-induced autophagy, reduced the fraction of bacteria that were accessible to the cytosol at 45 min, 90 min and 180 min p.i. (Fig 1B), consistent with enhanced autophagic capture of bacteria in, or repair of, damaged vacuoles. In contrast, inhibition of autophagy (WTM) increased the proportion of cytosolic bacteria at all timepoints examined (Fig 1B), in agreement with previous findings [14,31].
To verify cytosolic localization, we used a transcriptional reporter plasmid, pNF101, that only expresses gfp-ova in the sub-population of S. Typhimurium that are in damaged vacuoles and/or free in the cytosol. Epithelial cells were infected with mCherry-S. Typhimurium (S. Typhimurium wild-type bacteria constitutively expressing mCherry on the chromosome) harboring pNF101. In untreated cells, 35% and 13% of infected cells harbored GFP-positive (cytosolic) bacteria at 2 h and 8 h p.i., respectively (Fig 1C). EBSS treatment lowered this proportion Epithelial cells seeded on glass coverslips were pretreated with EBSS or 100 nM WTM as depicted in (A) and infected with S. Typhimurium wild-type bacteria harboring pFPV-mCherry for plasmid-borne, constitutive expression of mCherry. The proportion of bacteria accessible to anti-Salmonella LPS antibodies delivered to the mammalian cell cytosol was determined by digitonin permeabilization assay and fluorescence microscopy. n�3 independent experiments. (C) Epithelial cells seeded on coverslips were pretreated as in (A) and infected with mCherry-S. Typhimurium (wild-type glmS::Ptrc-mCherryST::FRT bacteria constitutively expressing a chromosomal copy of mCherry) harboring the fluorescent reporter plasmid, PuhpT-gfpova (pNF101). At the indicated times, cells were fixed and the proportion of infected cells containing cytosolic bacteria (GFP-positive) was scored by fluorescence microscopy. The mean from each experiment is represented as a large dot (n = 3). (D) Cells seeded on coverslips were pretreated as in (A) and infected with wild-type bacteria harboring pFPV-mCherry. The number of bacteria in each infected cell was scored by fluorescence microscopy. Cells with �100 bacteria contain cytosolic S. Typhimurium. Each small dot represents one infected cell. n = 2 (1 h) or 3 ( at both timepoints to 23% and 8%, respectively. Conversely, WTM treatment increased the level of infected cells with cytosolic bacteria at both 2 h and 8 h p.i. (79% and 48%).
To quantify the niche-specific effects of EBSS and WTM treatments on bacterial replication, epithelial cells were infected with S. Typhimurium harboring pFPV-mCherry and the number of bacteria per cell was scored. Cytosolic S. Typhimurium that evade autophagic clearance in epithelial cells replicate at a much faster rate than vacuolar bacteria [12,13,32], eventually occupying the entire cytosolic space (�100 bacteria/cell). By contrast, vacuolar bacteria display low to moderate replication (2-40 bacteria/cell). Comparing the three treatment conditions, the incidence of infected cells with �100 bacteria at 8 h p.i. (8.2±1.6% in control, n = 3 experiments) was reduced by EBSS treatment (1.2±2.0%) and promoted by WTM treatment (16 ±5.4%) (Fig 1D). A similar trend was seen with the chloroquine (CHQ) resistance assay, which quantifies the proportion of cytosolic bacteria in the total population-51±8.7%, 28±14% and 65±10% (n = 4 experiments) of intracellular bacteria were cytosolic at 8 h p.i. in untreated, EBSS-treated and WTM-treated cells, respectively. Consequently, total bacterial proliferation was restricted and promoted by EBSS and WTM treatment, respectively, compared to untreated cells ( Fig 1E). Importantly, neither bacterial invasion at 1 h p.i. nor vacuolar replication at 8 h p.i. was overtly affected (Fig 1D). Collectively, these data confirm that modulation of autophagy selectively affects the proliferation of S. Typhimurium in the cytosol of epithelial cells.

Identification of cytosol-and vacuole-induced genes by RNA-seq analysis
To define the transcriptional signature of cytosolic S. Typhimurium, we isolated bacterial RNA from epithelial cells that had been treated with EBSS or WTM, conditions that depleted or enriched the cytosolic population, respectively. Treated cells were infected with wild-type bacteria (Fig 1A), samples were collected and processed at 8 h p.i. and bacterial RNA was extracted using an enrichment protocol described previously ( [33]; S1 Fig). Importantly, EBSS and WTM treatments were restricted to the early stages of infection (�90 min p.i.) ( Fig 1A) to limit any effects on bacterial gene expression at the time of sample collection (8 h p.i.). Four cDNA libraries were derived from two biological replicates and analyzed by RNA-seq (see Materials and Methods; [34]). The entire dataset is available in S1 Dataset and includes the expression profiles of the >5,000 coding and non-coding genes identified previously in S. Typhimurium [34,35]. Genes for which the adjusted p-value was 0.05 or less (DeSeq2 analysis) were deemed as preferentially induced in the cytosol or vacuole, resulting in a shortlist of 216 "up cytosol" genes and 443 "up vacuole" S. Typhimurium genes (Fig 2A and S1 Dataset).
Of the twelve chromosomally-encoded pathogenicity islands (PAIs) in S. Typhimurium, only four were up-regulated in the cytosolic environment (S2 Fig). Most obvious was the abundance of SPI1 genes that were up-regulated, consistent with the reported induction of prgH in the cytosolic population at later times in epithelial cells [12,36]. SPI4, which encodes a giant non-fimbrial adhesin and its cognate type I secretion system and is co-regulated with SPI1 [37,38], was also up-regulated in the cytosol. Lastly, genes encoding for two T3SS1 effectors within SPI5 and SPI11, SopB [39,40] and SopF [24,25,41], respectively, were up-regulated in cytosolic bacteria. The remaining SPI5 and SPI11 genes were either up-regulated in the vacuolar population, or had unchanged expression, highlighting the mosaic nature of PAIs. Most SPI2 genes were up-regulated in the vacuolar population, as were genes encoding two type III effectors translocated by T3SS2 (pipB in SPI5 and sspH2 in SPI12). SPI3 genes were up-regulated in the SCV of epithelial cells, like in macrophages [42,43]. Similarly, most genes within SPI11, a PAI that is important for macrophage survival [44], and SPI12 were up-regulated in the epithelial SCV. The majority of SPI6 genes (which encode a type VI secretion system; [45]), all SPI9 genes (encode a type I secretion system; [46]) and SPI16 genes were not expressed in bacteria residing within either niche in epithelial cells. Overall, the transcriptional signatures of cytosolic and vacuolar bacteria at the PAI level (S2 Fig) validates the basis of our RNA-seq analysis.
To put the observed gene expression changes into the context of the extensive S. enterica literature, we used an innovative approach that involved the curation of 80 custom pathways from previously published regulons that had been defined by RNA-seq, ChIP-seq or microarray-based analysis. We combined these pathways with KEGG and GO gene sets to annotate the subsets of "up cytosol" and "up vacuole" S. Typhimurium genes ( Fig 2B and S2 Dataset). The list of "up vacuole" genes included an abundance (e.g. SPI2, pipB, pgtE, PinT, InvS, steB) that are positively regulated by the two-component systems that govern SPI2 induction in the SCV, namely PhoPQ, SsrAB and OmpRZ. Conversely, genes that are negatively regulated by these two component systems were in the "up cytosol" category. The SPI1-encoded transcription factors, HilA, HilD, and InvF, are required for the expression of SPI1 genes [47], and accordingly many of the "up cytosol" genes/sRNAs (e.g. sipBCDA, siiCD, sopF, sopB) grouped with these regulons. Our analysis also revealed several specific gene signatures (Fig 2B). For example, genes associated with iron transport were solely up-regulated in the cytosol. Up-regulation of the tricarboxylic acid (TCA) cycle, lysine degradation and molybdate transport was evident in vacuolar bacteria, suggesting increased reliance of Salmonella on these energy sources in the SCV. Interestingly, genes/sRNAs that are repressed by Ferric uptake regulator (Fur) in iron-replete conditions and activated by Hfq, BarA/SirA and FliZ were up-regulated in both the vacuole and cytosol.
We used the expression profiles of S. Typhimurium grown in 22 distinct infection-relevant in vitro growth conditions [34] to predict the environmental conditions experienced by S. Typhimurium when in the cytosol and vacuole. For the 216 genes up-regulated in the cytosol, three main patterns emerged: (i) up-regulation by bile shock (3% bile for 10 min) and/or low Fe 2+ shock (200 μM 2,2-dipyridyl (DPI) for 10 min) (e.g. mntH, sitABCD, entCEBA), (ii) upregulation by NaCl shock (0.3 M NaCl for 10 min) (e.g. proVW, ilvX), (iii) up-regulation during late exponential/early stationary phase of growth (e.g. sopE2, sipBCDA, siiABCDEF) (S1 Dataset). This suggests that cytosolic bacteria are growing fast and exposed to a low iron, high osmolarity physicochemical environment. The rapid growth of cytosolic S. Typhimurium has been previously observed [12,13,32], validating the former prediction. A common theme observed for the 443 genes up-regulated in the vacuole was induction by: (i) in vitro SPI2 growth conditions and/or (ii) internalization into macrophages (S1 Dataset).

Cytosolic iron-limitation is a cue for S. Typhimurium gene induction
To independently assess the cytosolic expression of genes predicted by RNA-seq analysis, we used green fluorescent protein (GFP)-based transcriptional fusions. This fluorophore is an ideal reporter to study differential gene expression in individual bacteria in infected host cells and tissues [12,18,48,49]. Previously gfpmut3 transcriptional fusions have demonstrated the vacuole-specific induction of SPI2 genes [50], for example. Due to the low abundance of many transcripts in the "up cytosol" shortlist (according to TPM; S1 Dataset), we also chose to use gfpmut3, which encodes for a stable GFP variant, to increase the likelihood of detecting fluorescent bacteria in epithelial cells. We initially focused on those genes that are induced when Salmonella are exposed to low Fe 2+ and/or bile shock in broth [34] (S1 Dataset). Epithelial cells were infected with mCherry-S. Typhimurium carrying individual gfpmut3 reporters and bacterial GFP fluorescence intensity was assessed qualitatively and quantitatively at 8 h p.i. The following genes were up-regulated in the cytosol: entC (entCEBA operon), fepA (fepA-entD operon), fepB, fhuA (fhuACBD operon), fhuE, iroB (iroBCDE operon), iroN, mntH, nrdH (nrdHIEF operon), sitA (sitABCD operon), SL1802, SL3990 (SL3990-SL3989 operon), sufA (sufABCDSE operon), yjjZ (SL1344_4483), STnc3080 and STnc3250 (Figs 3 and S3). Up-regulation of STnc4000, ilvC and SL1344_2715 was not observed by fluorescence microscopy (S4 All of the confirmed S. Typhimurium genes are up-regulated by "low Fe2+ shock" in broth (addition of 200 μM DPI for 10 min [34]). DPI chelation is not specific to iron, however [51,52], raising the possibility that other metals also regulate the expression of these genes. We used iroN, sitA, yjjZ, sufA, fepA and STnc3250 promoters to determine metal cation specificity in vitro. The effect of increasing concentrations (0.1-100 μM) of Co 2+ , Fe 3+ , Mn 2+ , Ni 2+ or Zn 2 + on GFP fluorescence for bacteria grown in defined minimal media (M9) was measured (S5 Fig). For all reporters, maximal expression was observed in the absence of any added metal. Upon addition of Fe 3+ and Co 2+ , GFP expression of S. Typhimurium carrying PiroN-gfpmut3, PyjjZ-gfpmut3 and PfepA-gfpmut3 reporters decreased in a dose-dependent manner. Notably, the promoter responses were 10-to 100-fold more sensitive to Fe 3+ than Co 2+ . PsitA-gfpmut3 expression was modulated by Fe 3+ , Mn 2+ and Co 2+ , although Fe 3+ and Mn 2+ were clearly the most potent repressors by an order of magnitude, in agreement with a previous study using a sitA::lacZ reporter [53]. Lastly, expression of PSTnc3250-gfpmut3 and PsufA-gfpmut3 were only reduced by increased concentrations of Fe 3+ . PiroN-gfpmut3 was the most sensitive iron reporter, with a >50% reduction in fluorescence at 0.1 μM Fe 3+ . Collectively, the in vitro repression of these genes/sRNAs by 0.1-1 μM Fe 3+ hinted that the induction cue in the cytosol was iron limitation (<0.1 μM).
To test the response of reporters to iron concentrations during infection, we used ferric ammonium citrate (FAC) to increase free iron levels in mammalian cells [54,55]. Epithelial cells were incubated overnight in growth media containing increasing concentrations (10-300 μM) of FAC. Untreated and FAC-treated cells were infected with mCherry-S. Typhimurium carrying transcriptional reporters, fixed at 8 h p.i. and the MFI of GFP signal for cytosolic and vacuolar bacteria quantified by fluorescence microscopy and ImageJ (Fig 4). The fluorescence intensity of cytosolic bacteria harboring PiroN-gfpmut3, PsitA-gfpmut3 or PyjjZ-gfpmut3 reporters was reduced in a dose-dependent manner over the FAC concentration range. Specifically, the average MFI was reduced by 57-, 36-and 15-fold, respectively, upon the addition of 300 μM FAC to growth media (Fig 4A), indicating repression of iroN, sitA and yjjZ expression in the cytosol under iron-replete conditions. The low MFI for vacuolar bacteria in untreated cells limited our analysis of the iron-dependent repression of iroN, sitA and yjjZ in the SCV.
We used the lipophilic, cell-permeable chelator, DPI, as an independent way to assess the effects of metals on S. Typhimurium gene expression in the mammalian cytosol. Because DPI

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella has a high affinity for various metal cations, we assessed the specificity for iron by means of an "add-back" experimental design [51,56,57] whereby an equimolar amount of Fe 3+ was added to the DPI-treated media. Epithelial cells were treated overnight with 200 μM DPI, 200 μM DPI plus 200 μM FAC, or 200 μM FAC alone and infected with mCherry-S. Typhimurium harboring PiroN-gfpmut3, PsitA-gfpmut3 or PyjjZ-gfpmut3 reporters. Cells were fixed at 8 h p. i. and the intensity of GFP signal for cytosolic and vacuolar bacteria quantified via fluorescence microscopy. Addition of DPI increased the expression of iroN, sitA and yjjZ for both cytosolic and vacuolar bacteria (Fig 4B), indicating that DPI accesses both cellular compartments. However, the effects of divalent cation chelation were more pronounced for the vacuolar population (13-to 43-fold increase in average MFI) than the cytosolic population (1.7-to 2.2-fold increase in average MFI). Add-back of iron restored the MFI to that of untreated cells for cytosolic bacteria harboring PiroN-gfpmut3 and PsitA-gfpmut3 reporters (Fig 4B), demonstrating that iron limitation is the primary stimulus for the up-regulation of these genes in the cytosol. Add-back of iron had a lesser effect on the average MFI of bacteria harboring the PyjjZ-gfpmut3 reporter, regardless of their intracellular niche (Fig 4B), possibly due to a lower sensitivity of yjjZ to changes in free iron levels ( Fig 4A) or a responsiveness to deprivation of metals other than iron in the intracellular environment (S5 Fig). Add-back of iron only partially restored the expression of sitA and iroN in vacuolar bacteria ( Fig 4B) suggesting that sitA and iroN respond to multiple divalent cations in the SCV or FAC treatment is less effective at increasing free iron levels in endocytic-derived compartments compared to the cytosol. Taken together, our data show that iron limitation in the mammalian cytosol is a major cue sensed by S. Typhimurium.
Using the same treatment conditions, we determined the role of iron limitation on bacterial replication in epithelial cells. As assessed by gentamicin protection assay (Fig 4C), total bacterial replication was limited by treatment with DPI (4.0-fold replication over 8 h) and DPI plus FAC (5.6-fold), but not FAC alone (8.8-fold), compared to untreated cells (10.0-fold). The CHQ resistance assay showed a similar profile ( Fig 4D). Compared to untreated cells (50 ±3.4% cytosolic bacteria), treatment with DPI reduced the levels of cytosolic bacteria to 28 ±6.7% at 7 h p.i. The proportion of cytosolic bacteria was increased to 42±11% upon add-back of iron (DPI+FAC), whereas it was indistinguishable from untreated cells for FAC alone (52 ±5.5%). Notably, DPI treatment (200 μM) also retards the growth of S. Typhimurium in complex media (S6 Fig).
It has been shown that epithelial cells contain either cytosolic bacteria, vacuolar bacteria or a mix of both populations at later times p.i. [32]. We used single-cell analysis to monitor the effect of metals on the distribution of these intracellular populations. Using mCherry-S. Typhimurium harboring pFN101 as a fluorescent reporter for cytosolic access, we analyzed the intracellular populations at 8 h p.i., in addition to the number of cytosolic (GFP+, mCherry+) and vacuolar (GFP-, mCherry+) bacteria within each infected cell. There was no significant difference in the frequency of epithelial cells that only contained cytosolic bacteria (0-1.2% of . Add-back of iron restored vacuolar replication (11.7 bacteria/cell) whereas FAC alone (11.6 bacteria/cell) was not significantly different to untreated cells. In summary, iron restriction affects S. Typhimurium proliferation in both the cytosol and vacuole in epithelial cells but has a much more profound effect in the cytosol.

Zinc and magnesium are more limiting in the vacuole than the cytosol
We investigated whether metals other than iron affected bacterial gene expression in epithelial cells. S. Typhimurium encodes multiple transporters for divalent cations [58]. The up-regulation of sitABCD and mntH in cytosolic S. Typhimurium (Figs 3 and S3), which are induced by iron and manganese limitation (S5 Fig) [53,56] and encode high affinity Mn 2+ transporters [59,60], suggests that Mn 2+ availability is lower in the cytosol than SCV. Zn 2+ import is dependent on a high affinity transport system in S. Typhimurium, ZnuABC [61], and its accessory protein, ZinT (formerly known as YodA) [62,63], and a low affinity ZIP family metal permease, ZupT [64,65]. zupT is constitutively expressed [64], in agreement with our RNA-seq analysis (S1 Dataset). By contrast, znuA and zinT are induced when zinc is limiting [66] and both

Manipulation of cellular iron levels primarily affects cytosolic bacteria. (A) Epithelial cells seeded on coverslips
were left untreated (control) or treated overnight in growth media supplemented with 10 μM, 30 μM, 100 μM or 300 μM ferric ammonium citrate (FAC). Cells were infected with wild-type mCherry-S. Typhimurium harboring PiroN-gfpmut3, PsitA-gfpmut3 or PyjjZ-gfpmut3 reporters, fixed at 8 h p.i. and the MFI of bacterial GFP signal was quantified by fluorescence microscopy and Image J. Bacteria were designated as being cytosolic (Cyto) or vacuolar (Vac) if residing within cells with �100 bacteria or 2-40 bacteria, respectively. Small dots represent individual bacteria; large dots indicate the mean of each experiment; horizontal bars indicate the average of two experiments. Acquisition parameters (exposure time and gain) were set-up using PiroN-gfpmut3 (the highest GFP intensity) and these same parameters were applied throughout.

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella genes are in the "up vacuole" shortlist (Fig 2A and S1 Dataset), consistent with increased ZnuA production upon S. Typhimurium internalization into mammalian cells [61]. We found that a PzinT-gfpmut3 reporter was exquisitely sensitive to zinc concentrations, with a 54% and 99% reduction in GFP fluorescence detected upon the addition of 0.1 μM and 1 μM Zn 2+ to minimal media, respectively (S5 Fig). The fluorescence intensity of both vacuolar and cytosolic mCherry-S. Typhimurium harboring PzinT-gfpmut3 increased at 8 h p.i. compared to t 0 , indicating the intracellular induction of zinT. However, GFP signal was considerably greater for vacuolar bacteria (Fig 5), suggesting that zinc is more limiting in the SCV (<0.1 μM). Quantification of the average MFI of GFP fluorescence confirmed the qualitative findings (5.8-fold increase for vacuolar bacteria/cytosolic bacteria; Fig 5, lower panel). A recent study utilizing a znuA transcriptional reporter to measure intracellular Zn 2+ levels reached a similar conclusion [67].
S. Typhimurium also has multiple transporters for Mg 2+ uptake. CorA is a constitutively expressed Mg 2+ transporter, and MgtA and MtgB are inducible Mg 2+ transporters [68]. Consistent with its constitutive expression, expression of corA was similar in both vacuolar and cytosolic bacteria (S1 Dataset). Low Mg 2+ induces mgtA and mgtB transcription [68,69]. Expression of mgtA was not induced in the vacuole, however (S1 Dataset). Acidic pH abolishes the mgtA (but not mgtB) transcriptional response to Mg 2+ concentration [69], which could explain why mgtA is not induced in the SCV of epithelial cells. In contrast, all three genes within the mgtCBR operon were up-regulated in vacuolar bacteria according to the RNA-seq dataset (S1 Dataset). Using mCherry-S. Typhimurium PmgtC-gfpmut3, we confirmed mgtCBR is responsive to Mg 2+ levels (S5 Fig) and the operon is up-regulated after bacterial internalization ( Fig 5). Furthermore, the average MFI of GFP signal was increased by 6.2-fold for vacuolar bacteria/cytosolic bacteria (Fig 5, lower panel) indicating that Mg 2+ is more limiting in the SCV than cytosol, in agreement with a recent study [67]. Altogether, our studies of metal-regulated gene expression reveal that S. Typhimurium encounters distinct metal ion availabilities in the SCV and cytosol; specifically, iron and manganese are more limiting in the cytosol whereas magnesium and zinc are limiting in the vacuole.

Identification of growth-phase related, cytosol-induced genes
Our transcriptomic analysis highlighted another set of cytosol-induced genes that are highly expressed during late exponential/early stationary phase of in vitro growth in broth, and in some cases responsive to oxygen shock (Fig 2 and S1 Dataset). Many of the genes encode for T3SS1 structural proteins (orgA, spaQ, sipB, sipC, sipD) and effectors known to be translocated by T3SS1 (sipA, sopB, sopD, sopE, sopE2, sopF [SL1344_1177]). We have previously shown that prgH (prgHIJK-orgABC operon), encoding a structural T3SS1 component, is up-regulated in cytosolic bacteria [12], in agreement with our RNA-seq results (S1 Dataset). Using GFP transcriptional reporters, we confirmed the increased expression of sicA (sicA-sipBCDA operon), sopE2 and sopF (Fig 6) in cytosolic bacteria. Of note, the average MFI of cytosolic bacteria was comparable to t 0 cultures (Fig 6, lower panel), which are highly T3SS1-induced [70]. PipB, a T3SS2 effector encoded on SPI5 [71], was used as a control for the vacuole-specific induction of a type III effector. In accordance, the average MFI of GFP fluorescence was 9.8-fold higher for vacuolar bacteria harboring a PpipB-gfpmut3 reporter compared to those in the cytosol (Fig 6). We confirmed that siiA (siiABCDEF operon) and lpxR were more highly expressed in bacteria residing within the cytosol compared to the SCV (Fig 6). The siiABCDEF operon in SPI4 (S2 Fig) encodes for a type I secretion system for the secretion of SiiE, a giant non-fimbrial adhesin [72,73]. Immunostaining with anti-SiiE antibodies detected the adhesin on the surface of cytosolic bacteria (S7 Fig, upper panel), and diffusely spread throughout the cytosol

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella (S7 Fig, lower panel). Vacuolar bacteria were negative for SiiE staining (S7 Fig). We have previously shown by electron microscopy that cytosolic bacteria have extensive filamentous material on their surface [12]. Taken together, our findings suggest the filaments could be SiiE. Like the sii operon, lpxR (SL1344_1263) is also co-regulated with SPI1 [74] and encodes an outer membrane 3'-O-deacylase of lipid A [75,76]. Transcriptional reporters did not verify cytosolic up-regulation of ygbA (SL1344_2840), which encodes a conserved hypothetical protein required for S. Typhimurium growth during nitrosative stress [77], or asnA, which is involved in asparagine biosynthesis [78] We conclude that there is a generalized up-regulation of "SPI1 associated" genes in the mammalian cytosol. However, we do not observe uniform up-regulation of these genes within the cytosolic population (Fig 6), in agreement with previous reports for prgH [12,14,36]. Based on average MFIs, all genes were more highly expressed in cytosolic bacteria than vacuolar bacteria (Fig 6,

Extraneous genes that are induced in the epithelial cytosol
Several osmotically-sensitive genes that are up-regulated when S. Typhimurium is exposed to NaCl shock [34] were identified as candidate cytosol-induced genes, namely cysP (cysPUWA operon), soxS and proV (proVWX operon) (S1 Dataset). We qualitatively and quantitatively confirmed the cytosol-specific expression of cysP and soxS using transcriptional reporters (S8 Fig). However, the proV-gfpmut3 transcriptional fusion was not up-regulated in the cytosol (S4 Fig). The cysPUWA operon encodes a sulfate/thiosulfate permease and periplasmic binding protein [81]. SoxS is a member of the AraC/XylS family of transcriptional regulators that is required for bacterial resistance to oxidative stress [82,83].

Identification of S. Typhimurium genes required for optimal cytosolic proliferation
Having identified S. Typhimurium genes/sRNAs that are up-regulated within the epithelial cytosol, we next tested whether any of these genes/sRNAs are required for colonization of this niche. We did not focus on the SPI1-associated effectors as their individual contributions to the cytosolic stage of the intracellular cycle have been reported previously [24,32,87,88]. Twenty-nine single or multiple gene deletion mutants were constructed and their ability to access and replicate within the cytosol was first assessed in a population-based assay by CHQ resistance (Fig 7A). Four single gene deletion mutants had a significantly reduced proportion of cytosolic bacteria at 7 h p.i. compared to wild-type bacteria (50.2±7.2%), namely ΔentC Plasmid-borne complementation of ΔentC (with entCEBA), ΔgrxAΔsoxS (with soxS), ΔmntHΔsitA (with sitABCD) and ΔznuA (with znuA) mutants restored their respective phenotypes to wild-type levels in the CHQ resistance assay (Fig 7B). Chromosomal complementation by Tn7-based integration at the glmS site with fepB also restored the ΔfepB defect to wild-type levels ( Fig 7B). Importantly, ΔentC, ΔfepB, ΔgrxAΔsoxS, ΔmntHΔsitA, ΔmntHΔsitAΔentC, ΔsoxS and ΔznuA mutants were comparable to wild-type bacteria for their proportion of cytosolic bacteria at 90 min p.i. (Fig 7C), indicating that the observed defects at 7 h p.i. are not simply due to delayed vacuole lysis.
We extended this population-based assay with single-cell analysis using S. Typhimurium harboring the dual fluorescence reporter plasmid, pCHAR-Duo(ASV), in which constitutive expression of mCherry is driven by the synthetic promoter ProB and destabilized GFP (gfpmut3.1(ASV)) is under the control of the S. Typhimurium uhpT promoter. All bacteria harboring this plasmid are mCherry-positive and only those bacteria in the cytosol will be GFP-positive. Epithelial cells were infected with wild-type, ΔentC, ΔfepB, ΔmntHΔsitA, ΔmntHΔsitAΔentC, ΔsoxS and ΔznuA bacteria harboring pCHAR-Duo(ASV) and the number  (Fig 7D). However, ΔentC, ΔfepB, ΔmntHΔsitA, ΔmntHΔsi-tAΔentC and ΔsoxS mutants were all defective for proliferation in the cytosol compared to wild-type bacteria as demonstrated by a decreased number of infected cells containing >50 cytosolic bacteria at 8 h p.i. (Fig 7E). The decreased proportion of cytosolic bacteria for the ΔznuA mutant (Fig 7A) was explained by an increased number of vacuolar bacteria/cell at 8 h p.i. (Fig 7E) rather than a defect in cytosolic proliferation per se (Fig 7D).
Overall, our data show that of the 27 bacterial genes we identified as being up-regulated in the epithelial cytosol (Figs 3, 6, S3, and S8), five are specifically required for the optimal proliferation of S. Typhimurium in this intracellular niche-entC, fepB, mntH, sitA and soxS.

Discussion
Compared to an endocytic-derived vacuole, the host cell cytosol could naïvely be viewed as a nutrient-rich milieu that allows for the efficient growth of bacteria. However, not all pathogens can survive within this niche, and S. Typhimurium can only replicate in the cytosol of epithelial cells and certain embryonic fibroblast lines [12,14,21]. Our finding that vastly different virulence gene programs are activated when S. Typhimurium colonizes the SCV and cytosol (Fig  8) confirm previous suggestions that the mammalian cytosol is a complex environment that requires pathogen-specific adaptations for efficient bacterial survival and replication [89].
Here we report the "cytosol transcriptional signature" of bacteria by integrating our data with the available gene expression profiles of Gram-negative pathogens that colonize the cytosol (Table 1). We focus on the prototypical cytosolic pathogen, S. flexneri [90][91][92][93], and uropathogenic Escherichia coli (UPEC) [94,95], which forms intracellular bacterial communities (IBCs) in the cytosol of bladder epithelial cells. The most striking feature that emerges is the enrichment of genes that mediate the acquisition of iron (Table 1). Salmonella acquires ferric (Fe 3+ ) iron by secreting two siderophores, enterobactin and salmochelin, a C-glucosylated form of enterobactin [58,96,97]. Iron-loaded siderophores then transit back into bacteria by first binding the outer membrane receptors, FepA and IroN [98,99], then FepB in the periplasm, and finally across the cytoplasmic membrane via the ABC-type transporter, FepDGC [100]. FhuE, FhuA and FoxA also function as outer membrane receptors to exploit the utilization of ferric siderophores produced by other bacteria [101]. All of the siderophore synthesis the CHQ resistance assay. n�4 independent experiments for each strain. Asterisks indicate data significantly different from WT (ANOVA with Dunnett's post-hoc test). (B) Genetic complementation. Epithelial cells were infected with WT bacteria or deletion mutants harboring an empty vector (-), pWSK29 or pWKS30, or the respective vector encoding the indicated gene(s), except for the ΔfepB mutant, which was complemented with a chromosomal copy of fepB at the glmS site. The proportion of cytosolic bacteria at 7 h p.i. was quantified using the CHQ resistance assay. Large dots indicate the mean of each experiment (n�3). Asterisks indicate data significantly different from WT (ANOVA with Dunnett's post-hoc test). (C) Epithelial cells were infected with WT bacteria or the indicated gene deletion mutants and the proportion of cytosolic bacteria at 90 min p.i. was quantified using the CHQ resistance assay. Large dots indicate the mean of each experiment (n�3). (D) Epithelial cells seeded on coverslips were infected with WT bacteria or gene deletion mutants harboring pCHAR-Duo(ASV), a plasmid-borne dual reporter-constitutive mCherry expression is driven by the synthetic ProB promoter and gfpmut3. 1

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella and transport loci are up-regulated in cytosolic S. Typhimurium (Table 1 and Fig 2). S. flexneri uses a similar iron-scavenging strategy involving the synthesis of another siderophore, aerobactin [102], and the genes required for aerobactin synthesis and binding (iucABCD operon) are highly up-regulated in the cytosol ( Table 1). The limited data available for UPEC IBCs are consistent with an important role of ferric iron acquisition for this bacterium as well (Table 1). Genes encoding the ferrous (Fe 2+ ) iron transport systems of S. flexneri (feoB, sitABCD and mntH) and S. Typhimurium (sitABCD and mntH) are also up-regulated, but not feoB ( Table 1). The Sit and MntH systems in S. Typhimurium transport both ferrous iron and manganese, but likely function as manganese transporters under physiological conditions [59,60]. We found that a S. Typhimurium ΔsitAΔmntH mutant is compromised for growth in the epithelial cytosol (Fig 7), indicating a specific requirement for Mn 2+ acquisition in this compartment, but the same transport systems do not contribute to the growth of S. flexneri in epithelial cells [92]. The plethora of iron acquisition strategies used by enteric bacteria means that genetic deletions in multiple pathways are required to affect intracellular proliferation. For SCV-resident bacteria are translocating type III effectors via T3SS2, devoid of flagella and exposed to limiting zinc and magnesium concentrations. By contrast, cytosolic bacteria highly express genes implicated in iron uptake/storage, manganese and sugar transport suggesting that accumulation of these nutrients is important for bacterial proliferation in this compartment. Up-regulation of a subset of genes associated with oxidative stress resistance further indicates that bacteria are exposed to reactive oxygen species. Cytosolic bacteria are also T3SS1-active, flagellated and decorated with the MUC1-binding adhesin, SiiE, and therefore primed to adhere to, and enter, naïve cells upon their release. https://doi.org/10.1371/journal.ppat.1009280.g008

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella example, single gene deletions of sitA, iucB or feoB in S. flexneri have no effect on intracellular growth but a mutant in all three genes is unable to grow [103]. Similarly, S. Typhimurium ΔentC (defective in the synthesis of enterobactin and salmochelin; [104]) and ΔfepB (defective for the import of enterobactin, salmochelin and catecholate breakdown products containing iron; [96]) mutants have impaired cytosolic replication abilities (Fig 7).

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella hydrogen donors and functions under aerobic conditions, but only during iron restriction [106,107]. The suf operon encodes for an iron-sulfur cluster system and is induced by oxidative stress and iron starvation [108], like the nrdHIEF operon [109,110]. Notably, genes encoding for antioxidant proteins involved in bacterial resistance to hydrogen peroxide (H 2 O 2 ), including dps, katG, ahpC and ahpF [111], were not induced in the epithelial cytosol. This finding suggests that superoxide anion generation from H 2 O 2 by epithelial cells is not a major contributor to the oxidative stress-driven responses of S. Typhimurium in the cytosol. While the function of YjjZ remains unknown, yjjZ (SL1344_4483) is part of the Fur regulon in S. Typhimurium [112] and UPEC [113]; Fur is the master regulator of iron homeostasis in bacteria [114]. Transcription of nrdHIEF, sufABCDSE and yjjZ is therefore controlled by common factors, iron deprivation and Fur. Two genetic loci involved in sugar uptake or conversion, fruBKA and uhpT, were also identified as part of the cytosol transcriptional signature (Table 1 and Fig 8). uhpT and fruBKA are not induced by iron deprivation or oxidative stress but rather by sugars present in the eukaryotic cytosol, specifically glucose-6-phosphate [90] and fructose [84], respectively.
The combination of available transcriptome data revealed bacterial species-specific cytosolinduced genes. Specifically, bioA, lysA, pstSC, phoRB, proVWX and mgtA were induced in Shigella but not in S. Typhimurium (Table 1 and S1 Dataset; [91,103]). The differential expression profile suggests variance in the metabolic and phosphate acquisition strategies used by these two pathogens in the epithelial cytosol. The contribution of Mg 2+ transport also seems to differ between the two bacteria, as mgtA-encoded Mg 2+ acquisition is induced by the cytosolic environment for S. flexneri and not for S. Typhimurium. Unlike S. flexneri, S. Typhimurium has a second Mg 2+ transport system encoded by mgtB, which is up-regulated in the vacuole (Fig 5  and S1 Dataset). Furthermore, the third Mg 2+ /Ni 2+ /Co 2+ transporter (CorA) is constitutively expressed in S. Typhimurium [115,116] and is required for optimal proliferation in the cytosol [117]. The importance of Mn 2+ acquisition also seems to differ (see above).
We identified S. Typhimurium-specific coding and non-coding genes that serve as signatures of the cytosolic lifestyle. Examples include STnc3080, STnc3250, SL1802 and siiABCDEF, as well as eight type III effectors that are translocated by T3SS1 (Figs 3, 6 and S3 and S1 Dataset). STnc3080, STnc3250 and SL1802 are all negatively regulated by Fur [35,112,118] but their role in Salmonella pathogenesis remains unknown. Intracellular expression of the siiABCDEF operon has not been reported before, prompting us to establish that the SiiE protein, which mediates adhesion to the MUC1 on the apical surface of intestinal epithelial cells [119], is both on the surface of cytosolic bacteria and secreted into the cytosol of epithelial cells (S7 Fig). The combination of findings from this study, and others [12,36], shows that cytosolic S. Typhimurium are induced for T3SS1, secrete T3SS1-associated effectors, are flagellated and decorated with SiiE. We conclude that cytosolic S. Typhimurium are "primed" to enter naïve enterocytes following cellular release.
The major environmental factor that controls the transcriptional program of cytosol-residing bacteria is clearly iron-limitation. Within mammalian cells, the "labile iron pool" refers to the pool of chelatable and redox-active iron [120,121]. In resting cells, the labile iron pool is estimated to be~1 μM, representing only a minor fraction of the total cellular iron (50-100 μM) [120,121]. At a subcellular level, measurements have shown that the concentration of labile iron is lower in the cytosol than the nucleus, mitochondria or endosomes/lysosomes [122,123], in agreement with our findings and those of others [67]. We believe that consumption of free iron from the cytosol by rapidly growing S. Typhimurium further depletes the pool compared to uninfected/resting cells.
The uniform induction of iron-responsive S. Typhimurium genes/sRNAs in the cytosol (Figs 3 and S3) indicates a generalized bacterial response to iron deprivation. Fur is the major

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella iron-responsive transcriptional regulator that controls iron homeostasis and oxidative stress defenses in bacteria [114]. When bound to iron (i.e. in iron-replete conditions), Fur acts as both a repressor and activator of gene expression, either directly or indirectly, and modulates expression of~10% of the S. Typhimurium genome [112,124]. Except for SL3990, all the low Fe 2+ /bile shock-regulated "up cytosol" genes/sRNAs (S1 Dataset) that were confirmed using transcriptional reporters (Figs 3 and S3) are negatively regulated by Fur [101,112,125] (Fig 2). Many have a Fur box in their promoter regions, the regulatory site to which iron-bound Fur binds, suggesting direct regulation. Of note, sitABCD and mntH belong to the MntR (SL1344_0810) mini-regulon that responds to manganese abundance, and are also regulated by Fur [53,56,126]. An unanswered question is why all Fur-repressed S. Typhimurium genes are not induced in the cytosolic population ( [112]; S1 Dataset). It is also unclear why SPI1-associated genes are induced in a subset of cytosol-exposed bacteria in an iron-limited environment, when expression of the SPI1 regulon is typically activated by iron-bound Fur under iron-replete conditions via HilD and H-NS [127,128]. The regulation of SPI1 and its associated genes is complex [47], with numerous environmental signals feeding into the regulatory network including osmolarity, oxygen and bacterial growth phase, as well as iron. Bacterial cellto-cell variation in the integration of all the relevant signals could explain the phenotypic heterogeneity observed for SPI1-associated gene expression in cytosolic bacteria (Fig 6).
We have identified some of the site-specific cues encountered by S. Typhimurium in the distinct milieus of the SCV versus the cytosol. S. Typhimurium residing within the epithelial cytosol have limited access to iron and manganese, are experiencing oxidative stress and utilize sugars as a carbon source. Interestingly, S. Typhimurium genes involved in iron uptake and storage, manganese uptake and Fe-S cluster synthesis are upregulated in the lag-phase of growth [129], which suggests a requirement for these particular genes as bacteria prepare for exponential growth in the mammalian cytosol [12]. Other studies of S. flexneri and L. monocytogenes colonization further describe the mammalian cytosol as a nutrient-rich environment of neutral pH with low concentrations of Na 2+ , Mg 2+ , Ca 2+ and PO 4 3ions, but a high concentration of K + [89,91,130,131]. In contrast, the mature SCV is acidic [7,132,133], has high levels of K + and oxygen, but is limiting in Mg 2+ , PO 4 3-, and Zn 2+ and S. Typhimurium are experiencing oxidative stress but not amino acid starvation (Figs 3, 5 and 8) [42,43,67,91,134].
Based on single-cell analysis, we can refine previous conclusions drawn from populationbased studies of S. Typhimurium and epithelial cells. For example, in the landmark study of Hautefort et al. (2008) [33], the simultaneous induction of SPI2, SPI1 and flagella within epithelial cells was reported, which at the time was puzzling. Single-cell microscopic analysis has solved this mystery by showing that the cytosolic population is solely responsible for the intracellular expression of both SPI1 and flagella ( [12,14,16,36]; this study). Based on the intracellular induction of iroA-and mgtB-lacZ reporter fusions in MDCK (epithelial) cells, concentrations of Fe 2+ and Mg 2+ were thought to be low in the mature vacuole [135]. Additionally, a recent proteomic analysis of S. Typhimurium isolated from infected epithelial cells at 6 h p.i. suggested that bacteria are starved for numerous metals, including iron, manganese, and zinc [136]. Our findings show that metal limitation experienced by intracellular bacteria is niche-dependent; levels of iron and manganese are lower in the cytosol, whereas zinc and magnesium are more limiting in the vacuole (Fig 8), which demonstrates the importance of verifying results derived from population-based studies at the single-cell level. We suggest that the many studies that have identified individual S. Typhimurium or mammalian genes that influence bacterial proliferation in epithelial cells should be revisited, especially if the distinct replication niches were not considered. As heterogeneity in terms of bacterial proliferation and intracellular colonization site clearly influences the outcome of infection [137,138], it is crucial that the study of host-pathogen interactions incorporates single-cell analysis.

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella
For genetic complementation, the deleted gene was either restored on a low copy number plasmid or in the chromosome. Specifically, ΔentC bacteria were transformed with pWSK29-entCEBA (provided by Steve Libby and Joyce Karlinsey), ΔgrxAΔsoxS bacteria with plasmidborne soxS under the control of its native promoter (pWKS30-soxS), ΔmntHΔsitA bacteria with pWKS30-sitABCD [145] and ΔznuA bacteria with pWKS30-znuA, which contains a DNA fragment encompassing the znuA promoter region and coding sequence. The ΔfepB mutant was complemented with a chromosomal copy of fepB by transposition at the attTn7 site [146]. Oligonucleotides used to construct pWKS30-soxS, pWKS30-znuA and pGP-Tn7-Cm-fepB are listed in S2 Table. SL1344 wild-type bacteria constitutively expressing mCherry from the chromosome (glmS:: Ptrc-mCherryST::FRT, denoted as mCherry-S. Typhimurium) or a plasmid (pFPV-mCherry) have been described previously [7,147]. mCherry expression is driven by the constitutive trc promoter or S. Typhimurium rpsM promoter, respectively. The PuhpT-gfpova plasmid, pNF101 [24], was used as a biosensor for bacterial exposure to the mammalian cytosol. Expression of the unstable GFP variant, gfp_ova (protein half-life <60 min, [148]), is under the

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella control of the glucose-6-phosphate responsive uhpT promoter from S. flexneri. Alternatively, bacterial access to the cytosol was assessed using the bidirectional reporter plasmid, pCHAR--Duo(ASV), whereby constitutive mCherry expression is driven by the synthetic ProB promoter and gfpmut3.1(ASV) expression is under the control of the S. Typhimurium uhpT promoter. To create pCHAR-Duo(ASV), the coding sequence for the stable GFP variant, gfpmut3.1 (protein half-life >24 h), was excised from pCHAR-Duo [32] by XmaI/ApaI digestion and replaced with the coding sequence for a destabilized GFP variant, gfpmut3.1(ASV) (protein half-lifẽ 110 min), which had been released from pGFP(ASV) (Clontech) by XmaI/ApaI digestion. To construct fluorescent transcriptional reporters, we extracted the precise transcriptional start sites from the online SalComMac database (http://tinyurl.com/SalComMac; [34,43]). Approximately 500 bp of sequence upstream of the ATG start codon of each gene, or transcript start site of each sRNA, was amplified from SL1344 genomic DNA with the oligonucleotide pairs listed in S3 Table. Amplicons were digested with XbaI/SmaI and ligated into the corresponding restriction sites of pGFPmut3.1 (Clontech). XbaI/ApaI digestion released the gene/ sRNA promoter-gfpmut3 fragment, which was then ligated into the corresponding sites of the low copy number plasmid, pMPMA3ΔPlac [70]. The pipB and sopB promoters were excised from pMPMA3ΔPlac-PpipB-GFP(LVA) and pMPMA3ΔPlac-PsopB-GFP(LVA), respectively [70], by digestion with XbaI/KpnI and ligated into the corresponding sites of pGFPmut3.1 and then transferred to pMPMA3ΔPlac as described above. The pMPMA3ΔPlac-PuhpT-gfpmut3 construct has been described previously [149]. GFP reporter plasmids were electroporated into wild-type mCherry-S. Typhimurium.

Measurement of bacterial growth in broth
Bacterial cultures were grown overnight in LB-Miller broth, shaking (220 rpm) at 37˚C, then subcultured 1:100 into 10 ml LB-Miller broth in 125 ml Erlenmeyer flasks with shaking (220 rpm) at 37˚C. Samples were collected at hourly intervals for OD 600 readings.

Mammalian cell culture
HeLa human cervical adenocarcinoma epithelial cells were purchased from American Type Culture Collection (ATCC, CCL-2) and used within 15 passages of receipt. Cells were

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella maintained at 37˚C and 5% CO 2 in the growth medium (GM) recommended by ATCC i.e. Eagle's minimum essential medium (EMEM, Corning) containing sodium pyruvate, L-glutamine and 10% (v/v) heat-inactivated fetal calf serum (FCS; Gemini Bio Products). Tissue culture plasticware was purchased from Thermo Scientific Nunc.

Bacterial infection of mammalian cells
HeLa cells were seeded 24 h prior to infection at the following densities: (i) 5x10 4 cells/well in 24-well tissue culture plates, (ii) 6x10 4 cells/well on acid-washed glass coverslips in 24-well plates, or (ii) 2.6x10 6 cells/15 cm tissue culture dish. T3SS1-induced bacterial subcultures were prepared in LB-Miller broth (Difco) [150] and cells were infected for 10 min with bacterial subcultures (MOI~50) as described [150]. DPI was prepared as a 100 mM stock in 95% ethanol and stored at -80˚C for a maximum of 2 weeks. FAC was prepared extemporaneously in MilliQ water. Gentamicin protection and CHQ resistance assays (chloroquine diphosphate salt, Sigma) were as described previously [150]. Monolayers were solubilized with 0.2% (w/v) sodium deoxycholate (Sigma) and serial dilutions were plated on LB-Miller agar (Remel) for enumeration of colony forming units (CFUs).

Bacterial RNA enrichment from infected cells, cDNA synthesis and Illumina sequencing
S. Typhimurium was enriched from infected HeLa cells (12 x 15 cm dishes for EBSS treatment, 6 x 15 cm dishes for WTM treatment) as described previously by Hautefort et al. (2008) [33] with minor modifications. Briefly, infected HeLa cells were washed once with 30 ml chilled phosphate-buffered saline (PBS), then lysed on ice for 10-20 min in 0.1% SDS, 1% acidic phenol, 19% ethanol in water (15 ml per 15 cm dish). Pooled lysates were collected into 50 ml conical tubes and centrifuged at 3,220 xg for 30 min, 4˚C. After three washes in ice-cold wash buffer (0.1% phenol, 19% ethanol in water) (30 min, 3,220 xg, 4˚C each wash), the remaining bacterial pellet was resuspended in <1 ml of wash buffer, transferred to an RNase-free microcentrifuge tube and centrifuged at 16,000 xg for 2 min. The supernatant was discarded, and the pellet resuspended in 1 ml TRIzol reagent (Life Technologies) by gently pipetting up-anddown (~60 times), then transferred to -80˚C for storage. Total RNA was extracted as described [34] and the RNA concentration quantified using a Nanodrop 2000 spectrophotometer (Thermo Scientific). RNA quality was analyzed using an Agilent Bioanalyzer 2100.
RNA samples isolated from two independent experiments were sent to Vertis Biotechnologie AG (https://www.vertis-biotech.com) for cDNA synthesis and RNA sequencing. The RNA samples were fragmented with ultrasound (4 pulses of 30 sec at 4˚C) followed by a treatment with antarctic phosphatase and re-phosphorylated with polynucleotide kinase. Afterwards, the RNA fragments were poly(A)-tailed using poly(A) polymerase 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. The resulting cDNA was PCR-

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella amplified to about 20-30 ng/μl using a high-fidelity DNA polymerase. The cDNA was purified using the Agencourt AMPure XP kit (Beckman Coulter Genomics) and analyzed by capillary electrophoresis. The primers used for PCR amplification were designed for TruSeq sequencing according to the instructions of Illumina. The four cDNA samples were pooled in equimolar amounts and size fractionated in the size range of 150-500 bp using a differential clean-up with the Agencourt AMPure kit, then the cDNA samples were sequenced on an Illumina HiSeq 2000 machine with 50 bp read length.

Pathway enrichment analysis
A collection of 80 custom, Salmonella-specific pathways was generated from previous RNAseq, ChIP-seq, and microarray studies of S. Typhimurium regulons [112,124,128,151,152], as well as transcriptional profiling of S. Typhimurium 4/74 under infection-relevant in vitro growth conditions [34] and inside murine macrophages [43]. A detailed description of the pathway curation process and all relevant scripts can be found at https://github.com/apredeus/ salmonella_pathways. Briefly, custom curated gene sets of SPI1 plus SPI1-translocated effectors, SPI2 plus SPI2-translocated effectors, and iron transporters and siderophores were compiled using many literature sources that are listed in the "Salmonella custom pathways.xlsx" table in the repository above (supp_tables subdirectory). For data from [112,151,152] [ 124,128], supplementary tables from the publications with differentially expressed genes were used; up-and down-regulated genes were used to create separate pathways. Whenever the authors used a different reference strain, gene IDs were converted to 4/74 IDs using the provided ortholog table. For single-replicate RNA-seq experiments profiling 4/74 and D23580 under the seventeen in vitro conditions, the following strategy was used. Tables of TPM-normalized expression values (see supp_tables/474.tsv, D23.tsv) were used to calculate fold change (FC) of a particular condition with regard to all 16 others. Genes with log2FC of 2 or more were selected as "marker" genes for a particular in vitro condition; only up-regulated genes

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella were considered for this specific dataset. The lists of genes up-regulated in each single-replicate experiment were generated using the calculate_fc_tpm.pl script available from the same repository. Additionally, KEGG pathways for S. Typhimurium SL1344 were obtained using the KEGGREST R package v1.28.0, and gene ontology (GO) mapping of SL1344 genes was extracted from SL1344 protein annotation by InterProScan v5.41-78. See S2 Dataset for the entire dataset. The combined pathways were used to annotate gene sets up-regulated in the vacuole and cytosol using the clusterProfiler R package v3.16.1. The resulting overlaps were visualized using ggplot2 v3.3.2.

Fluorescence microscopy
HeLa cells were seeded on acid-washed, 12 mm glass coverslips (#1.5 thickness, Fisher Scientific) in 24-well plates. Infected HeLa cells were fixed with 2.5% (w/v) paraformaldehyde in PBS for 10 min at 37˚C. The immunostaining procedure has been described previously [24]. Rabbit anti-SiiE serum (kindly provided by Michael Hensel) was used at a dilution of 1:200. Digitonin permeabilization to deliver anti-Salmonella LPS antibodies (Salmonella O-Antiserum Group B Factors 1, 4, 5, 12; Difco; 1:300 dilution) directly to the cytosol was as described previously [14]. Cells were stained with Hoechst 33342 (1:10,000 in DDH 2 O, Life Technologies) to label DNA and coverslips were mounted onto glass slides using Mowiol.
Samples were visualized on a Leica DM4000 upright fluorescence microscope for scoring the number of bacteria per cell, the proportion of cytosolic bacteria and quantification of transcriptional reporter activity. ImageJ software was used to quantify the activity of transcriptional reporters at the individual bacterium level. Images in the red (mCherry) and green (GFP) channels were acquired sequentially under 63x magnification on a Leica DM4000 upright fluorescent microscope using the same acquisition settings (exposure and gain) for each group of transcriptional reporters (determined by the reporter with the highest intensity in the green fluorescence channel). A minimum of 100 cytosolic and 100 vacuolar bacteria, from cells with �100 bacteria and 2-40 bacteria, respectively, acquired from �5 random fields of view were quantified for each transcriptional reporter from each experiment. From grayscale images, well-defined, individual bacteria were arbitrarily chosen on the mCherry channel (represents all bacteria) and manually outlined, converted to a binary image and the pixel intensity associated with each identified particle (i.e. one bacterium) quantified from the corresponding GFP channel image. GFP fluorescence/bacterium was subsequently plotted as the mean fluorescence intensity (MFI).
Images were acquired using confocal microscopy (Leica SP8) in sequential acquisition mode through an optical section of 0.3 μm in the z-axis. Images are maximum intensity projections of z-stacks.

Statistical analysis
All experiments were conducted on at least three separate occasions, unless otherwise indicated, and results are presented as mean ± SD. Except for the RNA-seq data (see separate paragraph on RNA-seq data analysis), statistical analyses were performed using: (i) one-way analysis of variance (ANOVA) with Dunnett's post-hoc test, or (ii) one-way ANOVA on ranks (Kruskal-Wallis test) for data with a non-normal distribution (i.e. number of bacteria/cell, MFI) (GraphPad Prism). A p-value of <0.05 was considered significant. Acquisition parameters (exposure time and gain) were set-up using PcysP-gfpmut3 (the highest GFP signal intensity) and these same parameters were applied throughout. Dashed lines indicate the range of background fluorescence in the GFP channel measured for mCherry-S. Typhimurium (no reporter). (TIF) S1  Table. Oligonucleotides used to construct gfpmut3 transcriptional reporters. (DOCX) S1 Dataset. RNA-seq data set. Tab 1 shows the raw counts. Raw sequencing reads have been deposited at Gene Expression Omnibus (GEO) database under series ID GSE179103. A total of 42 to 57 million sequence reads were generated from each sample. Tab 2 shows transcripts per million (TPM). All genes with mapped reads were reported as TPM and a threshold TPM value of 10 was used as a cut-off to distinguish gene expression from background. Tab 3 shows differential sequence analysis via DESeq2 v1.30.1. Tab 4 shows genes that were deemed differentially expressed (adjusted p-value of 0.05 or below). Tab 5 and Tab 6 shows the expression profiles of "up cytosol" and "up vacuole" genes that passed statistical significance in 22 distinct infection-relevant in vitro growth conditions plus in macrophages (http://bioinf.gen.tcd.ie/cgibin/salcom.pl?db=salcom_mac_HL). (XLSX)

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella S2 Dataset. The entire dataset used for the custom pathway enrichment analysis of "up cytosol" and "up vacuole" genes/sRNAs (relates to Fig 2B)

PLOS PATHOGENS
Transcriptional profiling of cytosolic Salmonella