Salmonella typhimurium discreet-invasion of the murine gut absorptive epithelium.

Salmonella enterica serovar Typhimurium (S.Tm) infections of cultured cell lines have given rise to the ruffle model for epithelial cell invasion. According to this model, the Type-Three-Secretion-System-1 (TTSS-1) effectors SopB, SopE and SopE2 drive an explosive actin nucleation cascade, resulting in large lamellipodia- and filopodia-containing ruffles and cooperative S.Tm uptake. However, cell line experiments poorly recapitulate many of the cell and tissue features encountered in the host’s gut mucosa. Here, we employed bacterial genetics and multiple imaging modalities to compare S.Tm invasion of cultured epithelial cell lines and the gut absorptive epithelium in vivo in mice. In contrast to the prevailing ruffle-model, we find that absorptive epithelial cell entry in the mouse gut occurs through “discreet-invasion”. This distinct entry mode requires the conserved TTSS-1 effector SipA, involves modest elongation of local microvilli in the absence of expansive ruffles, and does not favor cooperative invasion. Discreet-invasion preferentially targets apicolateral hot spots at cell–cell junctions and shows strong dependence on local cell neighborhood. This proof-of-principle evidence challenges the current model for how S.Tm can enter gut absorptive epithelial cells in their intact in vivo context.

Introduction remain rare. In the intestine of permissive mice, gut absorptive epithelial cells constitute prominent targets for S.Tm invasion [29][30][31]. By sharp contrast to tumor-derived cell lines, the absorptive intestinal epithelium in vivo exhibits i) the signaling properties of primary non-transformed cells, ii) a strictly polarized columnar cell arrangement, iii) dense apical microvilli, iv) a high degree of cellular interconnectedness, v) a heterogeneous set of neighboring epithelial cell types, and vi) a luminal barrier comprised of antimicrobial peptides, mucus, and commensal microbes [32]. As cell line infection assays were used to establish the prevailing model for S.Tm epithelium invasion, the implications of most of these physiological features remain ill-defined.
Experiments in in vivo models of salmonellosis have addressed the impact of bacterial virulence factors and host defenses on mucosal tissue pathology. In the frequently used streptomycin-pretreated mouse model [33,34], per-oral infection studies have revealed a contribution of the TTSS-1 apparatus, the TTSS-1 effectors SopE, SopE2, and SipA as well as the SiiE adhesin encoded on SPI-4 to S.Tm-inflicted mucosal pathology at~1-3 days post-infection (p.i.) [35][36][37][38]. Multiple TTSS-1 effectors, including in addition SopA, SopB, and SopD, contribute to diarrheal symptoms also during bovine infection [39]. Some morphological work has also been conducted to probe the epithelial invasion step in vivo, in mucosal tissue explants, or in ligated loops from e.g. guinea pig, calf, mouse or pig [40][41][42][43][44]. These studies have highlighted S. Tm invasion of both microfold cells (M cells) and absorptive epithelial cells in Peyer's patch regions of the mucosa, and identified a variety of morphological host cell features connected to S.Tm entry [41][42][43][44]. Additionally, recent work in neonate mice has revealed that the effectors SipA and SopE2 work redundantly to promote S.Tm invasion of, and traversal through, immature gut epithelial cells during the first days of the infection [45]. Nevertheless, comprehensive studies of how S.Tm invades the mature gut epithelium of adult hosts during the first critical hours of acute infection remain scarce.
Our earlier work in the streptomycin mouse model established that S.Tm targets cecal absorptive epithelial cells in substantial numbers during the initiating phase of acute infection [29,30]. Here, as well as in the small intestine of infected neonate mice, S.Tm rapidly forms intraepithelial microcolonies [30,31]. These microcolonies could largely be explained by replication, i.e. not by entry of several bacteria into the same host cell [30,31]. Since co-invasion of multiple bacteria is a hallmark of ruffle-mediated entry in cultured cell lines [5,18,19], these observations hinted that the model for S.Tm epithelial cell invasion might not be transferable to the intact gut epithelium.
Here, we have undertaken a comparative analysis of S.Tm invasion into common epithelial cell lines and the murine gut absorptive epithelium in vivo. Our work reveals that S.Tm invades gut absorptive epithelial cells through "discreet-invasion", a mode that differs markedly from ruffleinvasion of epithelial cell lines. Discreet-invasion critically requires TTSS-1 translocation of the primordial effector SipA, induces modest lengthening of local microvilli in the absence of expansive ruffles, does not support cooperative S.Tm entry, targets the apicolateral region of infected epithelial cells in a neighbor-dependent manner, and results in swift normalization of the epithelial cell brush border. These findings challenge the accepted model for how S.Tm enters gut epithelial cells and prompt further in vivo studies across the diversity of Salmonella strains and host species.
dispensable for entry into non-polarized cell lines (S1A-S1C Fig) [47]. In S.Tm wt (SL1344), the TTSS-1 effectors SopB, SopE, and SopE2 have been shown to predominantly drive rufflemediated entry, while the actin-binding effector SipA has a less prominent role [8,9,17,27]. We used automated microscopy to explore the generality of these findings in non-polarized epithelial cell lines of human (HeLa), canine (sub-confluent MDCK), and murine origin (m-ICc12). Joint deletion of the ruffle-inducers (i.e. ΔsopBEE2) essentially abolished S.Tm invasive behavior in all cases, similar to deletion of all four effectors (i.e. deletion of sopBEE2sipA;"Δ4") ( Fig 1A). By contrast, we were unable to detect a significant invasion defect upon deletion of SipA (ΔsipA) in either cell line (Fig 1A). In fact, at high pathogen densities, S.Tm ΔsipA invaded HeLa and m-ICc12 marginally better than S.Tm wt . These data, supported by bacterial plating assays (S1D Fig), validate the ruffle-inducers SopBEE2 as the key drivers of S.Tm invasion into diverse non-polarized epithelial cell lines.
Due to its well-defined colonization kinetics, we chose the streptomycin mouse model of Salmonella gut infection to study S.Tm invasion of the intestinal epithelium in vivo. In this model, luminal colonization and epithelial cell invasion begins in the cecum [34]. We investigated if the first wave of invasion shows a similar timing throughout this entire gut segment. C57BL/6 wild-type mice were infected (by oral gavage; 5x10 7 CFU) with S.Tm wt harboring a pssaG-GFP reporter (turns GFP+ subsequent to host cell entry [36]). Microscopy scoring of tissue-lodged S.Tm-GFP+ revealed equivalent numbers of intracellular bacteria across the entire cecal length (S1E and S1F Fig). As expected, the vast majority of S.Tm invasion foci (~95%) localized to EpCam-positive absorptive epithelial cells (S1G and S1H Fig). Hence, by microscopy of the middle part of the cecum, we can sensitively quantify S.Tm invasion of gut absorptive epithelial cells in vivo. As a baseline, we infected wild-type mice for 8-12h with S. Tm wt or isogenic strains deficient in TTSS-1 (ΔinvG) or the Sii adhesin system (Δspi-4), all harboring the pssaG-GFP reporter. S.Tm wt were as expected abundant in the cecal epithelium, whereas the S.Tm ΔinvG strain exhibited virtually undetectable numbers of invasion events (~200-fold attenuated at 8h; >1000-fold attenuated at 12h p.i.; S1J-S1L Fig). Deletion of spi-4 resulted in~20-fold lower levels of tissue-residing bacteria at 8h, but only a minor difference at 12h p.i. (S1J-S1L Fig). Consequently, early invasion of the naïve murine gut absorptive epithelium critically relies on TTSS-1 and is further enhanced by the SPI-4-encoded adhesin system.
Next, we infected wild-type mice (orally as above) with S.Tm wt , S.Tm ΔsipA , S.Tm ΔsopBEE2 and S.Tm Δ4 strains to analyze the dependence on TTSS-1 effectors. All strains equally colonized the gut lumen at 12h p.i. (Fig 1B). Remarkably, and in sharp contrast to results from the cell lines (Fig 1A), the S.Tm ΔsipA strain reached dramatically lower pathogen densities in the gut epithelium than S.Tm wt (~200-fold lower; Fig 1C and 1D). S.Tm ΔsopBEE2 was also attenuated, but still~10-20-fold more invasive than S.Tm ΔsipA (Fig 1D). In fact, the epithelial loads of S.Tm ΔsipA were similar to those of the strain lacking all four effectors (S.Tm Δ4 ; Fig 1D), or the TTSS-1 apparatus itself (S.Tm ΔinvG ; S1L Fig). Plasmid complementation restored SipA protein expression and epithelial invasion (Fig 1D and S1M Fig). Finally, similar observations were also made in a strain background lacking a functional TTSS-2 apparatus, which is important for intracellular S.Tm life (Fig 1D, right panels). Taken together, our data reveal a fundamentally different impact of TTSS-1 effectors on S.Tm invasion into epithelial cell lines-primarily driven by the ruffle-inducers SopBEE2 (Fig 1A), vs the murine gut absorptive epithelium in vivo-primarily driven by SipA (Fig 1B-1D).
SipA also affects S.Tm-induced gut inflammation (S1N Fig). Based on prior proposals [48], it remained conceivable that SipA could indirectly facilitate bacterial access to the epithelium in vivo, through inflammation-induced thwarting of protective barriers (i.e. affecting immune cell influx, mucus structure, or antimicrobial peptide production). If that was the case, then

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium co-infection with a separate SipA-expressing strain should help S.Tm ΔsipA achieve wild-type invasion efficiency. However, while a co-administered SipA-proficient helper strain (S.Tm Δsop-

BEE2
) augmented the early signs of inflammation (median pathoscore increased from 1 to 5; S1O Fig), it did not increase the invasion efficiency of a S.Tm ΔsipA reporter strain (Fig 1E and  1F; compare with SipA plasmid complementation in 1D). We conclude that SipA needs to be expressed by the invading bacterium itself to drive epithelial cell entry in vivo.
One feature that distinguishes the gut absorptive epithelium from common cell lines is its confluent, columnar, polarized cell arrangement. Epithelial cell polarization may be influenced by and/or affect the impact of S.Tm effectors [49,50]. We investigated if this property could account for the exquisite dependence on SipA for S.Tm invasion of the gut epithelium. MDCK cells were cultured in parallel as subconfluent non-polarized vs confluent polarized cell layers  (Fig 1C and 1D).
In summary, S.Tm invasion of the murine gut absorptive epithelium is facilitated by the SPI-4 adhesin system and critically depends on TTSS-1 delivery of the primordial effector SipA, with a surprisingly small contribution of the three ruffle-inducers SopBEE2.

Experiments in inflammasome-deficient mice verify the importance of SipA for gut absorptive epithelium invasion
Several TTSS-1 effectors, and in particular SipA, have been shown to shape the intracellular S. Tm niche and/or promote the bacterium's replicative potential subsequent to host cell entry [45,[52][53][54]. This could potentially skew the results above that rely on the pssaG-GFP reporter to quantify invasion efficiency in vivo. Moreover, infected epithelial cells in wild-type mice are frequently and quickly expelled into the lumen by an inflammasome response [30,55,56]. This reduces pathogen loads in the mucosal tissue and might hamper precise quantification of epithelial cell invasion efficiencies. To substantiate the observations above, we conducted S.Tm infections in inflammasome-deficient (Nlrc4 -/-) mice [57]. This allowed infections for a longer time period (18h) without overt epithelium destruction and resulted in~50-100-fold higher total bacterial numbers in the cecal absorptive epithelium (Fig 2A-2D and S1I Fig). At these high loads, robust tissue plating experiments could be performed without problems arising from carry-over contamination by the dense luminal S.Tm population (approximately 10 9 cfu/ g content; Figs 1B and 2A).
We first infected Nlrc4 -/mice for 18h with S.Tm wt or S.Tm ΔsipA harboring the pssaG-GFP reporter. The strains colonized the gut lumen equally (Fig 2A), but S.Tm ΔsipA again exhibited 100-fold lower total numbers of epithelium-lodged GFP+ bacteria (Fig 2B and 2C

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium mucosal tissue followed by plating of total intracellular bacteria ( Fig 2D). Finally, staining of permeabilized cecal sections with anti-Salmonella-LPS antibodies revealed plenty of tissueresiding S.Tm wt (Fig 2E), whereas S.Tm ΔsipA invasion foci were virtually absent from the epithelial tissue ( Fig 2E). Instead, S.Tm ΔsipA were found enriched on the epithelial surface (Fig 2E,  rightmost panel and insert). These results exclude that the effect of SipA on epithelial S.Tm loads in vivo can be explained by altered reporter maturation, or by a SipA effect on replication (although we do not refute that such effects could also exist). Of further note, intraepithelial S. Tm foci on average contain only a low number (mean~2) of bacteria during the first 12-18h of infection ( [30]; see also Fig 5E below). This means that the frequency of invasion events, rather than intraepithelial replication, predominantly governs the intraepithelial S.Tm load during early gut infection. Taken together, our results support that SipA deletion results in normal S.Tm gut lumen colonization, normal approach of and binding to the gut epithelium, but a profound epithelial cell invasion defect.
In vivo infections are subject to large animal-to-animal variations. Furthermore, in infections with one individual mutant per mouse, mutants attenuated at an early step of the infection process may face delayed onset of host defense and thereby grow or survive differently than S.Tm wt . This could complicate scoring of attenuation phenotypes in vivo. To substantiate the contribution of S.Tm effectors to epithelial cell invasion under internally controlled conditions, we employed our recently developed method for barcoded consortium infections [19]. Unique inert 40-nucleotide tags (informed by [58]) were placed on the bacterial chromosome of each strain of interest and infections conducted with a mixed inoculum comprising equal amounts of each strain. Quantitative PCR of genomic DNA was employed to quantify the relative abundance of each tag in enrichment cultures of the input (inoculum or luminal content) and output (intracellular) bacterial populations (see materials and methods for details).
For the consortium infections, we employed seven isogenic barcoded strains: S.Tm wt -tag A, S.Tm ΔsipA -tag B, S.Tm ΔsopBEE2 -tag C, S.Tm Δ4 -tag D, S.Tm ΔinvG -tag E, and the additional control strains S.Tm Δspi-4 -tag F, and S.Tm ΔinvGΔspi-4 -tag G (S1 Table). These strains were mixed in Next, we infected Nlrc4 -/mice with the same seven-strain consortium and extracted the luminal and epithelial tissue-residing bacterial populations. This yielded the expected total bacterial loads in both compartments (~10 9 S.Tm/g cecum content, S3F Fig; and~10 6 S.Tm in tissue/cecum, S3G Fig). S.Tm wt -tag A outperformed the other strains with respect to epithelial invasion (~55% of the intracellular population; Fig 2F right panel). S.Tm Δspi-4 -tag F exhibited a modest attenuation (~23% of intracellular population) compared to the wild-type. Importantly, deletion of SipA (S.Tm ΔsipA -tag B) again resulted in a dramatic loss in invasiveness (�1% of total intracellular population), whereas the strain lacking the ruffle-inducers (S. Tm ΔsopBEE2 -tag C) performed markedly better (~16% of total intracellular population) ( Fig  2F right panel). Taken together, multiple experimental approaches demonstrate that S.Tm invasion of the murine gut absorptive epithelium critically depends on TTSS-1 translocation of the primordial effector SipA. This contrasts starkly to observations of the invasion process in cultured epithelial cell lines.

S.Tm invades gut absorptive epithelial cells through discreet entry structures
The results above (Figs 1 and 2) indicate that SipA is a key driver of epithelial cell invasion in vivo, in the absence or presence of SopBEE2. Importantly, previous work has suggested that SipA on its own is incapable of inducing large ruffles in cultured cell lines (e.g. [26,59]). The strong SipA-dependence for gut epithelium entry raises the question whether the model for S. Tm invasion through large SopBEE2-dependent ruffles applies to the in vivo scenario. We hypothesized that the primary, differentiated, polarized, and neighbor-connected nature of absorptive epithelial cells in vivo steer S.Tm invasion towards a SipA-dependent, and away from a ruffle-dependent, entry mechanism. We therefore examined the presence and morphology of entry structures around invading S.Tm in cell lines exhibiting different degrees of polarization, and in the mouse gut.
We began by characterizing S.Tm entry structures in MDCK cells, cultured in parallel either as flat-growing or polarized cell layers on plastic. The cells were infected with S.Tm wt expressing constitutive GFP (pM965; S1 Table) and infections terminated by fixation. We consistently noted induction of~3.5-8μm high actin ruffles in non-polarized MDCK cells (Fig 3A  and 3B). Polarized MDCK cells produced significantly smaller entry structures (~2-4.5 μm height) in response to S.Tm wt /pGFP (Fig 3A and 3B). In agreement with previous work [27], these ruffles were also shaped differently, often with a circular appearance when viewed from the top. Similar results were obtained with an S.Tm strain expressing a SopE-M45 reporter protein (S1 Table), which allowed focusing the analysis on host cells that recently experienced delivery of TTSS-1 effectors ( Fig 3B). Moreover, when m-ICc12 and Caco-2 C2Bbe1 cells were grown as semi-polarized/polarized monolayers atop Transwell inserts (see growth conditions in S2D and S2E Fig), they produced smaller ruffles upon S.Tm infection, as compared to their subconfluent non-polarized counterparts (Fig 3C-3E). These data show that polarization of epithelial cells reduces their propensity to generate large ruffles in response to S.Tm docking and effector translocation.
High loads of luminal bacteria complicate the use of constitutive reporters to visualize epithelial cell invasion in vivo. Using S.Tm/pssaG-GFP in mice, we did not observe overt perturbation of the apical actin brush border of infected epithelial cells at the resolution of lightmicroscopy ( Fig 3F and 3G, S4A Fig). This could imply that i) no ruffles are forming, or that ii) in vivo entry structures are exquisitely short-lived and the apical actin returns to normal before the ssaG-GFP-reporter produces visible fluorescence. To resolve this uncertainty, we again utilized the S.Tm wt /psopE-M45 reporter strain. Of note, translocated effector proteins can by this approach be detected within less than a minute of TTSS-1 secretion [60]. Indeed, SopE-M45 could be detected as a deposit in occasional infected epithelial cells in vivo ( Fig 3H  and S4B Fig). Since this effector exhibits a short half-life within host cells (half-life << 1h; [61]), the approach allowed us to focus the analysis of entry structures to epithelial cells that had recently experienced TTSS-1 effector translocation. Notably, we again did not detect any pronounced perturbations of the actin brush border of infected cells compared to uninfected neighbors ( Fig 3H and 3I, S4B Fig).
Next, we used scanning electron microscopy (SEM) to investigate the ultrastructural appearance of S.Tm entry sites. As expected, S.Tm inoculums contained rod-shaped~0. . Importantly, no large lamellipodial ruffles were observed (scrutiny of the cecal mucosa in N = 5 mice). From these data, we conclude that S.Tm enters the intact gut absorptive epithelium through discreet entry structures, distinct from the expansive SopBEE2-dependent ruffles observed in epithelial cell lines. This agrees with an entry mechanism dominated by TTSS-1 delivery of the primordial effector SipA (Figs 1 and 2). We term this mode of S.Tm entry into the murine gut epithelium "discreet-invasion".

SipA drives ruffle-independent invasion into a tight vacuolar compartment
We next aimed to resolve the dynamic features of SipA-driven epithelial cell invasion at high temporal resolution. We began by infecting non-polarized epithelial cell lines, i.e. sub-confluent MDCK and HeLa cells expressing LifeAct (for actin visualization), with fluorescent S.Tm strains. At the end of the time-series, fixation and anti-S.Tm-LPS staining without prior permeabilization allowed us to determine which pathogen-host cell encounters had led to successful invasion.

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium varied greatly, however. The strain invading through SipA (i.e. S.Tm ΔsopBEE2 ) has a poor capacity to enter cultured cell lines (Fig 1A-1C), which prompted us to carry out the corresponding live experiments at a high MOI over a longer time period (MOI 500, 40min). Nevertheless, all S.Tm ΔsopBEE2 invasion events detected in both cell types occurred in the complete absence of actin ruffles (Fig 4C-4E  Cryo-electron tomography of the thin edges of plunge-frozen HeLa cells was used to determine the ultrastructural underpinnings of the wild-type and "SipA-only"-driven invasion processes in a near-native state. The cryo-tomograms illustrated abundant actin bundles underneath S.Tm wt captured at early stages of host cell binding (S6 Movie and Fig 4F). For events captured at a later stage of invasion, these bundles were replaced by a multidirectional meshwork of actin filament-rich protrusions around the bacterium (S7 Movie and Fig 4G; red pseudo-coloring delineates the bacterial membrane and white pseudo-coloring the host cell membrane). Again by sharp contrast, invading S.Tm ΔsopBEE2 bacteria were found within a tightly wrapped and smooth membrane compartment (S8 Movie and Fig 4H). Hence, the S. Tm ΔsopBEE2 strain, which invades specifically through SipA, enters non-polarized epithelial host cells by sinking into a tight vacuole without inducing complex higher-order actin meshworks. This is especially notable since non-polarized epithelial cells are highly permissive for ruffling if exposed to bacteria expressing SopB, SopE and/or SopE2. Taken together, our timelapse and 3D-reconstruction data provide a direct link between the primordial TTSS-1 effector SipA and a discreet-invasion mode for epithelial cell entry.

Non-cooperative S.Tm invasion into the murine gut absorptive epithelium
Cooperative invasion, i.e. that an actively invading bacterium promotes the entry also of bystander bacteria, is prevalent in cultured cell lines infected with S.Tm [5,18,19]. Expansive

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium membrane ruffles elicited by the primary invader generates physical obstacles where secondary motile bacteria get entangled and taken up [5]. Cooperative invasion depends on the likelihood of secondary bacteria finding a ruffle and will therefore increase in frequency with higher MOI, larger ruffle size, and/or longer ruffle duration [5,18,19]. As such, cooperative invasion provides a functional readout for S.Tm-elicited host cell ruffling responses.
In the mouse cecum, the mucus barrier covers the crypts, whereas the top~50% of the epithelial layer is in contact with motile luminal S.Tm [62]. Based on this prior knowledge, we used confocal microscopy to estimate the effective MOI in the cecum at 12h p.i. (in Nlrc4 -/mice to prevent epithelium distortion). We quantified the total luminal S.Tm population and the number of accessible epithelial cells per section. S.Tm were evenly spaced over the gut lumen cross-section with a modest enrichment at the epithelial border ( Fig 5A). As expected, only few bacteria localized within crypts, while S.Tm were frequently found in contact with the differentiated part of the gut epithelium (Fig 5B). Repeated analysis resulted in an MOI estimation of 91+/-20 (mean +/-SD; Fig 5C). It should here be noted that luminal S.Tm loads in wild-type and Nlrc4-deleted mice are equal during early infection (Figs 1B and 2A) [30]. Moreover, the gut luminal pathogen population reaches a stable plateau of colonization (~10 8 −10 9 CFUs/gram content) already at 6-8h p.i. (S1J Fig) [30]. Hence, the naïve mouse cecal epithelium experiences close contact with a dense and motile luminal S.Tm population for several hours during acute infection.
We infected Nlrc4 -/mice with a 1:1 mix of two differentially labelled wild-type S.Tm strains (S.Tm wt /pssaG-mCherry and S.Tm wt /pssaG-GFP) to begin assessing cooperative invasion frequency. At 12h p.i.~55% of all epithelial invasion foci carried only one bacterium,~30% carried two bacteria, and the remaining foci carried three bacteria or more (Fig 5D and 5E). Most notably, only~5% of all invasion foci carried a mix of green and red bacteria (Fig 5D-5F; N tot = 1055 foci in 4 mice analyzed). This subfraction of invasion foci could either have resulted from co-invasion (i.e. two active invasion events into the same host cell, occurring in parallel or in sequence), or alternatively from true cooperative (helped) invasion. In either case, the results show that at an estimated MOI of 91+/-20 in the mouse gut, cooperative epithelial cell invasion is at best rare.
To estimate the frequency of true cooperative invasion events, we adapted the dual-colored mixed inoculum to include one actively invading strain (S.Tm wt /pssaG-mCherry) and one strain incapable of TTSS-1-mediated active entry (S.Tm ΔinvG /pssaG-GFP; >1000-fold reduced invasion capacity in single strain infections; S1L Fig). A mixed invasion focus could with this setup only arise if the S.Tm wt strain promoted cooperative entry of S.Tm ΔinvG . As reference, the mixed inoculum was used to infect cultured HeLa and m-ICc12 cells at MOIs spanning across the range noted in vivo (MOI 20-320, 1h infection). HeLa cells produce expansive S. Tm-induced ruffles (Fig 3J), and as expected cooperative invasion was highly prevalent at all MOIs tested (Fig 5G-5J). At an MOI close to the in vivo estimate (MOI 80) >80% of all invaded cells carried S.Tm of both colors (Fig 5J). Mouse m-ICc12 cells produce somewhat smaller ruffles (Fig 3K), and consequently cooperative invasion was less common than in HeLa cells, but still comprised~40% of all invasion events at MOI 80 (Fig 5H-5J). By stark contrast, we did not observe a single case of cooperative epithelial cell invasion in the mouse cecum (Fig 5I and 5J; N tot = 578 foci in 5 mice analyzed). This means that even when the gut luminal S.Tm population supports a high MOI in close contact with the mucosa for several hours (Fig 5A-5C), cooperative S.Tm entry does not occur (Fig 5J). These data provide functional support for S.Tm discreet-invasion into the murine gut absorptive epithelium.

S.Tm discreet-invasion preferentially occurs proximal to cellular junctions
While studying the impact of apical S.Tm wt binding to the gut epithelium in vivo, we noted that bacteria frequently localized to the cell-cell junctional zones, separating individual

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium epithelial cells (e.g. Fig 3H and 3L panel i, S4B Fig). These circumstantial observations prompted us to investigate if S.Tm discreet-invasion exhibits preference for specific apical locations. To study the surface-binding S.Tm population in vivo in isolation, luminal bacteria were removed by repeated gentle washing of infected tissue, prior to fixation and staining of the remaining adherent S.Tm wt (Fig 6A). Again, we noted only modest perturbations of the local actin brush border proximal to attached S.Tm. Moreover, S.Tm surface binding exhibited a highly non-random pattern;~80% of all bacteria could be found within a 2μm distance from the closest cell-cell junction (Fig 6B and 6C).
We adapted the procedure for SEM imaging of the gut epithelial surface from the luminal side ( Fig 6D). To quantify the distribution of bound S.Tm, the surface of each epithelial cell was subdivided into three zones of equal area. Zone 1 covered the junction-proximal part, zone 2 the intermediate part, and zone 3 the mid part of the cell surface (Fig 6E). The center of each bound S.Tm was subsequently mapped onto these zones. To increase the number of bacteria that could be observed in this transient pre-invasion state, we infected mice with S.Tm Δ4 , which lacks all the major TTSS-1 effectors, but remains competent for host cell binding. A marked enrichment of S.Tm binding was noted in zone 1, which carried a higher fraction of all bound bacteria than zone 2+3 combined (Fig 6F).
To test if not only binding, but also S.Tm invasion, exhibited preference for the apicolateral zone in polarized epithelial cell layers, we next performed a similar analysis in live polarized MDCK cells expressing LifeAct. Fixation and staining for S.Tm-LPS without permeabilization was used at the end point of the infection to focus the analysis on successful invasion events (Fig 6G). Each event was traced back to the moment of entry in the live series, and the apical host cell membrane subdivided into zones as above. In full agreement with results from the binding experiments in vivo, a majority of S.Tm invasion events mapped to zone 1 (Fig 6H  and 6I). Based on these data, we conclude that an apicolateral surface region represents a hotspot for S.Tm binding and invasion of the polarized gut absorptive epithelium.
By contrast to homogeneous epithelial cell lines, the intact gut epithelium comprises multiple cell types in addition to absorptive epithelial cells. Of these, mucus-producing goblet cells make up~12% of the total cell-count in the murine cecum. In the SEM analysis, we frequently observed that epithelium-adherent S.Tm localized close to the junctional zones between an absorptive epithelial cell and its goblet cell neighbor (Fig 6J). The S.Tm/pssaG-GFP reporter

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium strain was therefore again used to examine the neighborhoods of epithelial cells targeted by S. Tm invasion. The results revealed a~5-fold enrichment of S.Tm invasion events into gobletcell-neighboring epithelial cells, as compared to goblet cell non-neighbors (Fig 6K and 6L). This points to a non-random targeting of S.Tm epithelial cell discreet-invasion both at the cellular (i.e. goblet-cell neighbors) and subcellular (junctional zone-proximal) level.

Discussion
Tissue culture studies have established that S.Tm invades epithelial cells through TTSS-1 and SopBEE2-driven large membrane ruffles [3,4]. We have confirmed these findings across cell culture models from diverse species and found that ruffle-invasion accounts for 100% of the S. Tm entry events. Such ruffles are characterized by a mix of actin meshwork-containing lamellipodia and spike-like filopodial protrusions, induced by parallel activation of several actin regulatory Rho and Arf GTPases (e.g. Rac1, Cdc42, Arf1; [4]). The large size and dynamic nature of S.Tm-induced ruffles, combined with bacterial near-surface swimming, also accounts for the prevalent cooperative uptake of bystander bacteria [5,18] (Fig 7A). Importantly however, we here argue that S.Tm invasion of absorptive epithelial cells in the mouse gut proceeds by discreet-invasion, a process with distinct molecular and morphological properties (Fig 7B). Specifically, discreet-invasion of the murine gut absorptive epithelium i) is facilitated by the SPI-4 adhesin system, ii) depends strongly on the primordial TTSS-1 effector SipA, iii) exhibits only a moderate dependence on the ruffle-inducers SopBEE2, iv) does not promote cooperative entry, v) drives formation of discreet and transient entry structures distinct from prototypical ruffles, vi) preferably targets the apicolateral membrane at cellular junctions, and vii) results in preferential invasion of goblet-cell neighboring epithelial cells. All of these features contrast to observations of the ruffle-invasion process in epithelial cell lines (Fig 7A and 7B). Moreover, discreet-invasion also appears distinct from the TTSS-1-independent entry mechanism(s) that have been described e.g. in cultured fibroblasts [63,64], since discreet-invasion of the gut epithelium requires both TTSS-1 and SipA.
Differences in epithelial cell polarity may contribute to the observed differences in S.Tm invasion mechanisms. When epithelial cell lines were grown as a semi-polarized/polarized cell layers, the size of S.Tm-induced ruffles decreased in comparison to non-polarized counterparts. By time-lapse imaging of polarized MDCK cells we also detected a~5% fraction of S. Tm wt invasion events that proceeded in the absence of visible actin ruffles. Still, bacterial entry through ruffling appears commonplace in this host cell context (this study and [27,65,66]). Previous work also showed that S.Tm invasion of polarized epithelial cell lines involves cooperative entry [65], similar to in non-polarized cell lines [5,18,19,26], while S.Tm invasion of the mouse gut absorptive epithelium occurs in the absence of cooperativity (this study). Furthermore, we observed only a modest SipA-dependent invasion phenotype in polarized epithelial

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium cell lines cells (~1.5-3-fold attenuation), compared to the �100-fold invasion defect in the gut epithelium in vivo. Based on this, we propose that polarized cell lines morphologically resemble columnar in vivo epithelia, but may retain (or acquire in culture) some immature properties that still overemphasize ruffling responses elicited by SopBEE2. It is here noteworthy that cellular transformation often results in overexpression of cytoskeletal regulators, including Rho-GTPases [67], which are common targets for these S.Tm effectors.
Early electron microscopy studies of S.Tm infection in starved and opium-treated guinea pigs [40], or in Peyer's patches of calves, mice, and pigs [41][42][43][44], revealed S.Tm invasion of both M-cells, absorptive epithelial cells, and goblet cells. Peyer's patch M-cells exhibited membrane ruffling with lamellipodial features in response to S.Tm [41,42]. In pig absorptive epithelial cells, elongated and distorted microvilli could be found at entry sites, i.e. in line with our results herein [44]. However, some examples of more pronounced cell surface perturbations were also noted in calves [41,43]. Hence, it appears plausible that the mechanism of S.Tm epithelial cell invasion in vivo may vary along the spectrum from "ruffle-invasion" to "discreetinvasion", as a consequence of the epithelial cell type afflicted, the host species, the developmental stage of the epithelium and/or the effector repertoire of the S.Tm strain. In gut absorptive epithelial cells of adult mice infected with S.Tm SL1344, discreet-invasion appears to constitute the norm (Fig 7).
The prominent role of SipA during epithelial cell invasion in mice was unexpected, and contrasts to findings in cell lines ( [25,27]; this study). On the molecular level, SipA has been shown to bind directly to actin, stabilize and bundle actin filaments, and in combination with SipC stimulate actin nucleation [22,25,68]. In cellular extracts, addition of SipA also prevents actin filament disassembly by ADF/Cofillin or Gelsolin [24]. Villin, another actin-binding and severing protein, is highly enriched in the brush border of intestinal epithelial cells. Knockdown of Villin in a polarized epithelial cell line attenuated S.Tm invasion and Villin -/mice showed a blunted mucosal tissue response to S.Tm infection [66]. It appears plausible that one or several of the actin-supporting functions of SipA can explain the discreet outgrowth of

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium microvilli around invading S.Tm in vivo, as noted in our SEM analysis. Furthermore, a screen for host cell factors involved in SipA-driven HeLa cell invasion identified members of the SPIRE family of actin nucleation factors [26]. SPIRE2 is also highly expressed in gut epithelial cells [69]. It, however, seems less likely that SPIRE(s) act directly downstream of SipA, since SPIRE1/2 ablation resulted in a general decrease in invasiveness also for S.Tm strains invading by other means [26]. Nevertheless, the profound impact of SipA on epithelial cell invasion in adult (this study), as well as neonate mice [45], warrants further studies of how the biochemical activities of this effector are integrated in the in vivo context. Previous work has demonstrated important role(s) for SipA in shaping the S.Tm niche in the vacuole or cytosol after host cell entry [45,[52][53][54]. Our data do not refute that SipA can act in this fashion also in the intact epithelium of adult mice. However, several lines of evidence still support a predominant effect of SipA during the epithelial cell invasion step in vivo. First, our approach with multiple methods confirmed that our results do not depend on the use of specific fluorescent reporters. Second, microscopy of washed tissue showed S.Tm ΔsipA to be enriched on the epithelial surface similar to S.Tm wt , but to be incapable of accumulating inside the mucosal epithelium. Third, S.Tm replicative foci are on average small (mean~2 bacteria) and cytosolic replication rare in the epithelium of adult wild-type mice [30,70], where we noted �100-fold reduced bacterial loads upon SipA deletion (this study). Consequently, total S.Tm loads in the murine gut epithelium depend strongly on the de novo S.Tm invasion rate, which in turn depends on the TTSS-1 effector SipA.
Moreover, we found that S.Tm targets absorptive epithelial cells preferentially proximal to cell-cell junctions in vivo. Near-surface swimming allows S.Tm to scan along a plastic dish or host cellular surface, resulting in preferential trapping and docking at the base of physical obstacles (e.g. preexisting membrane ruffles) [5]. The junctions between individual gut epithelial cells, and between goblet cells and their neighboring enterocytes, constitute interruptions of an otherwise homogenous surface. Hence, the observed S.Tm targeting preference may stem from near-surface swimming along the epithelium surface and subsequent trapping at cell-cell junctional sites of unevenness. Direct testing of this hypothesis will require further development of in vivo imaging technology. We have by such technology recently shown that S.Tm exhibits "near-mucus surface swimming" in the murine gut [62], which can explain the high likelihood of goblet cell neighboring epithelial cells being targeted for S.Tm discreetinvasion.
From an evolutionary standpoint, SPI-1 acquisition represented a significant leap in the evolution of a harmless commensal E. coli-like bacterium towards modern pathogenic Salmonella spp. It is here relevant to note that TTSS-1 and SipA, but not SopBEE2, are encoded directly within SPI-1. Hence, TTSS-1 and SipA can be considered the primordial virulence arsenal that enabled Salmonella spp. to invade the host's gut epithelium. The genes for the ruffle-inducers were acquired as separate elements, SopB located within SPI-5 [71], SopE on a bacteriophage found only in some serovars [72], and SopE2 on a phage remnant [9]. Despite their invasion-promoting potential in cultured cell lines, the functions of SopB, SopE, and SopE2 during gut infection remain a matter of debate. These effectors may generally enhance invasive behavior, drive ruffle-invasion into specific host cell types, or promote absorptive epithelial cell invasion in particular host species or developmental stages. Alternatively, acquisition of SopBEE2 has primarily served the purpose of driving gut inflammation through activation of pro-inflammatory signaling in the mucosa [35,38,39,[73][74][75]. In either case, also without SopBEE2, TTSS-1 combined with SipA constitutes a remarkably efficient minimal system for S.Tm discreet-invasion of absorptive gut epithelial cells in their native niche. The present study motivates continued efforts towards uncovering the molecular facets of this in vivo invasion mechanism.

Ethics statement
All animal experiments were performed in accordance to the Swiss Federal Government guidelines in the animal experimentation law (SR 455.163 TVV). The protocols used were approved by the Cantonal Veterinary Office of the canton Zürich, Switzerland (Kantonales Veterinäramt ZH licenses 223/2010, 222/2013, and 193/2016).

Bacterial strains, plasmids and culture conditions
All S.Tm strains used in this study were isogenic derivatives of SL1344 [76]. Strains and bacterial plasmids are detailed in S1 Table. For infections, S.Tm strains were grown in LB broth/ 0.3M NaCl, supplemented with 50μg/mL streptomycin (AppliChem), 50μg/mL ampicillin (AppliChem), or 12.5μg/mL chloramphenicol (Sigma Aldrich) for 12h at 37˚C in a rotating wheel incubator. Cultures were diluted 1:20 in the same broth w/o antibiotics and cultured for an additional 4h at 37˚C. For epithelial cell line infections, the inoculum was diluted in tissue culture medium to achieve the indicated MOI. For mouse infections, the inoculum was washed in sterile phosphate-buffered saline pH 7.4 (PBS; Gibco or Amimed) and reconstituted in PBS to a concentration of~10 9 CFUs/ml.

Mice and in vivo infections
Mice were bred and kept in individually ventilated cages under specific pathogen free conditions (RCHCI and EPIC facilities, ETH Zürich). Wild-type C57BL/6 mice were originally from Charles River, Rag1 -/-(B6.129S7-Rag1tm1Mom/J) from Jackson Laboratory, and Nlrc4 -/mice have been described elsewhere [57]. S.Tm infections were performed as detailed in [33]. In brief,~8-12 week old mice were treated with 25mg streptomycin sulphate per oral gavage. This step is required to suppress gut microbiota colonization resistance and thereby permit luminal expansion of S.Tm. 24h later, mice were infected per oral gavage with 5x10 7 CFUs of the indicated S.Tm strain. Bacterial loads in gut lumen content and organs were monitored by plating homogenized samples on MacConkey agar (Oxoid) with 50μg/ml streptomycin. Plating of cecum tissue was done subsequent to a 30min incubation in PBS/400μg/ml gentamicin, six rigorous washes in excess PBS, and tissue homogenization. For histopathology scoring, cecum tissue was frozen in optimum cutting temperature medium (OCT; Tissue-Tek). 5μm cross-sections were air dried and stained with hematoxylin and eosin. Histopathology scoring was done blindly as described [33]. Briefly, a score was assigned based on the degree of submucosal edema, polymorphonuclear leukocyte infiltration, goblet cell numbers, and epithelial damage. The possible scores range from 0 (uninflamed) to 13 (maximum inflammation). 15.000 LifeAct-expressing HeLa (LifeAct-mCherry) or MDCK (LifeAct-GFP) cells were seeded into glass-bottomed culture dishes (ibidi) in culture medium 24 hours prior to the experiment. To induce polarization of MDCK cells, 25.000 cells were seeded into culture dishes and incubated for 7 days, with medium exchange after 3, 6, and 7 days. Short-term live microscopy was performed in HBSS (Gibco)/10% FCS/20mM Hepes. Cells were infected with S.Tm wt /pmCherry or S.Tm wt /pGFP (MOI~50) or with S.Tm ΔsopBEE2 /pmCherry or S.Tm Δsop-BEE2 /pGFP (MOI~500). Movies were acquired on a Nikon Eclipse T1 inverse microscope equipped with a Yokogawa CSU-W1-T2 spinning-disk confocal unit and two EMCCD ixon888 cameras, using a 20x objective (non-polarized cells; PLAN Apochromat, NA 0.75) or a 60x oil objective (polarized MDCK; PLAN Apochromat, NA 1.4). Movies were acquired for 14 min-40 min and time-lapse data were reconstructed and analyzed in Image J ×64. Subsequent to live cell imaging, infected cells were washed with HBSS/10% FCS, fixed with 4% PFA, washed with 4% sucrose, saturated in 20% sucrose, and incubated in a PBS/3% bovine serum albumin/3% sucrose blocking buffer. Extracellular bacteria were labeled using α-S.Tm LPS and α-rabbit-Cy5. Cells were subsequently permeabilized in 0.1% Triton X-100 and stained with DAPI. Image acquisition of the same areas previously recorded by live microscopy was performed using a 100X oil objective (PLAN Apochromat, NA 1.49; non-polarized cells) or a 60X objective (PLAN Apochromat, NA 1.4; polarized MDCK cells). Image z-stacks were collected using the VisiVIEW software and further analyzed in Fiji. For consistency, the LifeAct signal is presented as red and S.Tm as green in all figure panels.

Cryo-electron microscopy and cryo-electron tomography
EM finder grids (gold NH2 R2/2, Quantifoil) were sterilized under UV light and then glow discharged. Grids were placed on the bottom of wells in a 12-well plate (Nunc, Thermo Fisher) and equilibrated in DMEM. Subsequently, 30.000 HeLa cells were seeded into each well and incubated overnight. Cells were infected with S.Tm wt and S.Tm ΔsopBEE2 as indicated. Plunge freezing was performed according to [78]. Briefly, grids were removed from the wells using tweezers, these subsequently mounted in a Vitrobot (Thermo Fisher) and the grids blotted from the backside by installing a Teflon sheet on one of the blotting pads. Grids were plungefrozen in liquid ethane-propane (37%/63%) and stored in liquid nitrogen as described [79]. Infected cells were examined by cryo-electron microscopy (cryoEM) and cryo-electron tomography (cryoET) as detailed in [78]. Images were recorded on a Tecnai Polara TEM (Thermo Fisher) equipped with post-column GIF 2002 imaging filter and a K2 Summit direct electron detector (Gatan), or on a Titan Krios TEM (Thermo Fisher) equipped with a Quantum LS imaging filter and a K2 Summit. Both microscopes were operated at 300kV and the imaging filters were set to 20 eV slit width. The pixel size at the specimen level ranged between 4.29-5.95Å. Tilt series covered an angular range from -60˚to +60˚with 1˚increments and -8 μm defocus. The total dose of a tilt series was 120 e -/Å 2 . Tilt series and 2D projection images were acquired automatically using UCSF Tomo [80] or SerialEM [81]. Three-dimensional reconstructions and segmentations were generated using the IMOD program suite [82].

Field emission scanning electron microscopy
Infected mouse cecal tissue or HeLa cells were fixed in 2.5% glutaraldehyde (Polyscience). After washing in Krebs-Ringer buffer, 1% OsO 4 (Polyscience) was used for post-fixation, or samples were washed in TE buffer (TRIS 10 mM, EDTA 1 mM, pH 7.0) without further osmification. HeLa cells were additionally incubated in 0.5% carbohydrazide and treated with 1% OsO 4 for a second time. All samples were washed before dehydration in a graded series of acetone. Critical-point-drying was performed with liquid CO 2 using an Autosamdri-931 (Tousimis or Bal-Tec CPD030). Thereafter, samples were mounted on aluminum SEM stubs, sputter coated with 5nm platinum/palladium or palladium/gold (Safematic CCU-010 or Bal-Tec SCD500). Samples were examined using a Zeiss Merlin Gemini II ultra-high resolution field emission scanning electron microscope at an acceleration voltage of 5 kV. Images were captured and analyzed with Zeiss SmartSEM and Image J ×64.

Barcoded consortium infections
Barcoded S.Tm strains are detailed in S1 Table. Strains were grown in LB/0.3M NaCl/12.5μg/ ml chloramphenicol for 12h at 37˚C, diluted 1:20 in LB/0.3M NaCl and cultured for another 4h at 37˚C. A 1:1:1:1:1:1:1 mix of all seven strains (or a 1:1:1:1:1:1 mix of a six strain consortium for S3E Fig) was used as inoculum. Cell lines were seeded in wells or atop Transwell inserts as indicated and infected with the mixed consortium inoculum for 7-20min at low MOI (i.e. 0.2-2; to limit the impact of cooperative invasion). Extracellular S.Tm were killed by 200μg/ml gentamicin for 1h and cells were washed and lysed in 0,1% sodium deoxycholate. The diluted inoculum and the retrieved intracellular bacterial population were enriched for 16h in tubes with LB (+/-12.5μg/ml chloramphenicol) at 37˚C. For mouse infections, the mixed inoculum was resuspended in PBS and used to infect sm-pretreated Nlrc4 -/mice with 5×10 7 total CFUs per oral gavage. Animals were sacrificed~18h p.i. The cecum was opened longitudinally, washed extensively in PBS, and treated with 400μg/ml gentamicin for 30min. The tissue was washed an additional nine times in PBS and homogenized in a TissueLyzer (Qiagen). S.Tm populations in cecum content and cecal tissue lysate were enriched in parallel in LB/12.5μg/ml chloramphenicol at 37˚C. Total bacterial loads in cecum content and tissue was evaluated by plating a subfraction of the homogenates on MacConkey agar with 50μg/ml streptomycin. Genomic DNA from enrichment cultures was extracted using the GenElute™ Bacterial Genomic DNA Kit (Sigma Aldrich) and quantitative PCR analysis performed on 9ng of total genomic DNA, using the Maxima SYBR Green/ROX qPCR Master Mix (2X) (Thermo Scientific). Primers are detailed in S2 Table. The relative abundance of each barcoded strain was normalized to the corresponding mixed inoculum for HeLa cell infections, or to the corresponding cecum content for animal infections.

Statistical analysis
Where applicable, statistical significance was assessed by the Mann-Whitney U-test, the Kruskal-Wallis with Dunn's post-test, or the one-way ANOVA with Dunnett´s multiple comparisons test, as indicated in the figure legends.

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium Supporting information S1 Fig. (supporting data for Fig 1). (A-C) Invasion efficiency of S.Tm wt and S.Tm Δspi-4 pssaG-GFP reporter strains in the indicated epithelial cell lines, infected for 20min, and analyzed at 4h p.i. by automated microscopy. Data points represent mean +/-range of two to three replicate infections. (D) m-ICc12 cells were infected with the indicated strains at MOI 62.5 for 20min. Quantification of intracellular bacteria after gentamycin treatment. Bars represent mean +/-SD of six replicate infections. One-way ANOVA with Dunnett´s test (n.s., not significant; �� p<0.01). (E) S.Tm CFU counts in cecum content of three C57BL/6 wild-type mice infected with S.Tm wt for 12h. (F) Quantification of intraepithelial S.Tm per 20μm section across the entire length of the cecum in the three mice described in E. Note that S.Tm invasion events distribute evenly across the cecum length.

PLOS PATHOGENS
Salmonella Typhimurium discreet-invasion of the murine gut absorptive epithelium into a moderate TEER (semi-polarized) monolayer, while Caco-2 C2Bbe1 cells form a tight monolayer with high TEER. Longer incubation times then those presented did not significantly increase the TEER values further for either of the cell lines. (F-G) Invasion efficiency of the indicated S.Tm strains in (F) semi-polarized m-ICc12 cells and (G) polarized Caco-2 C2Bbe1 cells grown atop Transwell inserts (conditions at 72/96h end-point of confluent growth, indicated by the inverted triangle symbols in D-E). Cells were infected at MOI 62.5 for 7 or 20min, and analyzed by selective plating of intracellular bacteria. Shown are CFU data expressed as the percentage of the inoculum retrieved in the intracellular population. Each circle symbol corresponds to one replicate infection. Bars represent mean+/-SD. One-way ANOVA with Dunnett´s test (n.s., not significant; �� p<0.01). (TIF) S3 Fig. (supporting data for Fig 2). (A) Dependence on SipA for S.Tm absorptive epithelial cell invasion in vivo. Inflammasome-deficient (Nlrc4 -/-) mice were orally infected with the indicated S.Tm/pssaG-GFP strains for 18h. Graph shows quantification of the number of intraepithelial S.Tm foci per 20μm section (mice also analyzed for total intraepithelial S.Tm loads in Fig 2C). Each data point corresponds to one animal. Line at median. (B-G) Barcoded consortium infections of epithelial cell lines and in vivo in mice. (B) Relative abundance of the individual strains in the barcoded consortium inoculum. The pie chart depicts the average from seven replicate experiments, where the relative abundance of each strain was assessed by quantitative PCR after enrichment culture. Note that none of the strains in the consortium is significantly over/underrepresented in the inoculum. (C-D) Barcoded consortium infections of (C) m-ICc12 cells on plastic, and (D) polarized Caco-2 C2Bbe1 cells grown atop Transwell inserts. The cells were infected for 20min at a total MOI of 2, using the same seven strain barcoded consortium as in Fig 2F and S3B Fig. Bars correspond to mean +/-SD of six (C) or three (D) replicate infections (circle symbols). (E) Barcoded consortium infection of polarized Caco-2 C2Bbe1 cells grown atop Transwell inserts with a less complex consortium. The cells were infected for 7, 10 or 20min at a total MOI of 2, using a barcoded consortium containing six tagged strains; two S.Tm wt (tag C and tag D), two S.Tm ΔsipA (tag B and tag F), and two S. Tm Δ4 strains (tag E and tag G) (see S1 Table). The relative abundance for S.Tm wt , S.Tm ΔsipA , and S.Tm Δ4 was calculated based on the summed abundance of the two internal technical replicates for each strain. Data points correspond to mean +/-SD of three replicate infections with separately prepared consortia. In C-E, One-way ANOVA with Dunnett´s test (n.s., not significant; � p<0.05, ��� p<0.001). (F-G) Total S.Tm CFU counts in cecum content (F) and washed cecal tissue (G) of Nlrc4 -/mice infected with the seven strain barcoded consortium for 18h (barcode quantification data presented in Fig 2F). Each data point corresponds to one animal. Line at median. (TIF) S4 Fig. (supporting data for Fig 3).  Fig. (supporting data for Fig 3). (A-C) Additional SEM micrographs of S.Tm invasion into epithelial cell lines and the absorptive gut epithelium in vivo in mice. (A) SEM micrographs of the S.Tm wt inoculum used in Fig 3J-3L. (B) Additional SEM micrographs of HeLa cells infected with S.Tm wt for 6-10min at MOI 400, as in Fig 3J. (C) Additional SEM micrographs of the cecal epithelium in mice, either uninfected, or upon infection with S.Tm wt , as in Fig 3L. Scale bars indicated separately for each panel. Arrow heads point to S.Tm. For each micrograph containing a dashed white box, the panel directly to the right of it represents the same area at higher magnification. (TIF) S6 Fig. (supporting data for Fig 4).