SPI2 T3SS effectors facilitate enterocyte apical to basolateral transmigration of Salmonella-containing vacuoles in vivo

ABSTRACT Salmonella pathogenicity island (SPI) 2 type three secretion system (T3SS)-mediated effector molecules facilitate bacterial survival in phagocytes but their role in the intestinal epithelium in vivo remains ill-defined. Using our neonatal murine infection model in combination with SPI2 reporter technology and RNA-Seq of sorted primary enterocytes, we demonstrate expression of SPI2 effector molecules by intraepithelial Salmonella Typhimurium (S. Typhimurium). Contrary to expectation, immunostaining revealed that infection with SPI2 T3SS-mutants resulted in significantly enlarged intraepithelial Salmonella-containing vacuoles (SCV) with altered cellular positioning, suggesting impaired apical to basolateral transmigration. Also, infection with isogenic tagged S. Typhimurium strains revealed a reduced spread of intraepithelial SPI2 T3SS mutant S. Typhimurium to systemic body sites. These results suggest that SPI2 T3SS effector molecules contribute to enterocyte apical to basolateral transmigration of the SCV during the early stage of the infection.


Introduction
Enteric and typhoid Salmonella colonize the gut lumen and subsequently penetrate the intestinal epithelial barrier to gain entrance to the sterile subepithelial and lymphoid tissue. The critical and nonredundant role of this early step in the course of the infection in vivo is illustrated by the ability of all Salmonella subspecies including human clinical isolates to invade non-phagocytic intestinal epithelial cells 1 . Internalization by enterocytes is facilitated by Salmonella pathogenicity island (SPI)1 type three secretion system (T3SS)-translocated effector molecules. Once internalized, Salmonella resides in a membrane-enclosed cell compartment, the Salmonella-containing vacuole (SCV). Cell-derived signals induce expression of a second, SPI2-encoded T3SS and set of effector molecules. 2,3 In phagocytes, SPI2 T3SS effectors critically contribute to intracellular survival, facilitating spread to systemic organs and persistent infection. 4,5 In non-polarized epithelial cells, SPI2 T3SS effector molecules localize the SCV to the peri-Golgi region, manipulate the endosomal compartment and ensure nutrient supply and intracellular proliferation. [6][7][8][9] Several reports have linked an intact SPI2 T3SS with invasiveness and penetration of the mucosal tissue. [10][11][12][13] However, few studies have analyzed the molecular mechanisms that facilitate penetration of the intestinal epithelial barrier, i.e. apical to basolateral transmigration and enterocyte egress.
We have recently shown that infection of neonate mice allows the visualization and quantitative analysis of infected enterocytes. 14 Neonate mice exhibit reduced exfoliation of infected enterocytes. 15,16 Using this neonate infection model, we previously characterized the requirement of individual SPI1 T3SS effector molecules for enterocyte invasion and intracellular proliferation. 17 Here, we extended this work and analyzed the role of the SPI2 T3SS during this early step in the course of the infection.

Results
Using transcriptomic analysis of primary S. Typhimurium-infected intestinal epithelial cells isolated and sorted by flow cytometry on day 4 post infection (p.i.), we observed expression of mRNAs for SPI2 T3SS components and effector molecules by intraepithelial wildtype (wt) S. Typhimurium in vivo (Figure 1(a)). SPI2 induction in intraepithelial wt S. Typhimurium in vivo was confirmed using a pssaG::gfp reporter construct (Figure 1(b)). Interestingly, SPI2 T3SS effector gene expression and SCV formation by wt S. Typhimurium in intestinal epithelial cells did not require innate immune stimulation previously reported to mediate endosomal acidification and drive SPI2 T3SS effector gene expression in myeloid cells ( Fig. S3A-D). 2,3 SPI2 T3SS effector expression in primary enterocytes in our neonatal infection model prompted us to analyze the functional role of the SPI2 T3SS during enterocyte infection. Initially, we compared the clinical course and organ load between neonate mice infected with wt S. Typhimurium or isogenic mutants lacking the SPI2 T3SS translocon component SseB and unable to translocate effector molecules via the SPI2 T3SS (ΔsseB). 18 Consistent with previous reports on the critical role of SPI2 T3SS effector molecules for survival in phagocytes and the intracellular localization of a major fraction of S. Typhimurium in systemic organs, ΔsseB S. Typhimurium exhibited a strongly attenuated phenotype compared to wt bacteria illustrated by decreased bacterial numbers in spleen and liver tissue and loss of virulence ( Fig.  S1A-C). Surprisingly, the number of S. Typhimurium in isolated primary small intestinal epithelial cells was not reduced in the absence of a functional SPI2 T3SS (Figure 1(c)). In fact, the size of the intraepithelial SCV after ΔsseB SPI2 T3SS mutant S. Typhimurium infection in vivo was not reduced but rather enlarged (Figure 1(d-e)) with an unchanged number of infected cells per field of view (Figure 1(f)). Ultrastructural analysis of SPI2 T3SS mutant S. Typhimurium infected enterocytes revealed large SCVs covering a major part of the epithelial cell cytosol, enclosed by an intact membrane and filled with actively dividing bacteria (Figure 1 (g)). Despite extensive electron microscopy analysis, no endosomal rupture or cytosolic bacteria and maximal one SCV per cell were observed for ΔsseB S. Typhimurium similar to what we have previously reported for wt S. Typhimurium. 14 Also, the SCV membrane of ΔsseB S. Typhimurium was positive for the established SCV marker proteins LAMP-1 and LAMP-2, as shown by immunostaining and immunoelectron microscopy ( Fig. S1D-F). Infection with a 1:1 ratio of GFP-and DsRed-labeled SPI2 T3SS mutant bacteria resulted in only single colored SCVs, demonstrating that SCVs emerged by clonal expansion of intraepithelial bacteria similar to what had been observed for wt S. Typhimurium (Fig. S1G). 14 Further in-depth analysis of the number and size of ΔsseB SPI2 T3SS mutant SCVs during the course of the infection revealed a stable number but steady increase in size during disease progression (Figure 2 (a-c)). This increase was found after infection with ΔsseB SPI2 T3SS mutant but not wt S. Typhimurium (Figure 2(d-e)). Even SPI2 T3SS mutant S. Typhimurium additionally lacking uptake systems for important nutrients such as glucose, amino acids, fatty acids or peptides formed large SCVs (Fig. S3G-I). Nutrient availability of intraepithelial SPI2 T3SS mutant S. Typhimurium could be explained by access to luminal energy-rich material ( Fig. S3E and F).
Several SPI2 effector molecules interact with the host cell cytoskeleton and motor proteins and could account for the observed phenotype. 19,20 For example, SifA was reported to bind via the kinesininteracting protein (SKIP) the microtubule-based motor kinesin-1 and PipB2 to promote kinesin-1 accumulation on the SCV membrane. 21,22 Similarly, SseF and SseG was shown to contribute to the positioning of the SCV possibly also by regulating microtubule motors on the vacuole. 23 We therefore compared the frequency and size of    intraepithelial SCVs between wt, ΔsifA, ΔpipB2 and ΔsseFG mutant S. Typhimurium. Indeed, ΔsifA, ΔpipB2 and ΔsseFG S. Typhimurium mutants at 2 days p.i. exhibited an unaltered SCV frequency (Figure 2(f)) but significantly enhanced vacuolar size ( Figure 2(g)). 24 Thus, the SPI2 effector molecules SifA, PipB2, SseF and SseG may contribute to the described phenotype of SPI2 T3SS mutant S. Typhimurium.

SPI2
Importantly, in addition to the size, also the intraepithelial positioning of SCVs differed between wt and SPI2 T3SS mutant S. Typhimurium (Figure 3(a-f)). The number of wt SCVs positioned apical to the nucleus decreased during the course of the infection concomitant with an increase in the number of basolateral SCVs. In sharp contrast, the vast majority of SPI2 T3SS mutant S. Typhimurium SCVs remained within the apical part of the enterocyte (Figure 3 (e-f)). A less pronounced but significant effect on the intraepithelial positioning was also observed for ΔsifA mutant S. Typhimurium (Figure 3(g-h)).
To further substantiate our hypothesis of SPI2dependent enterocyte apical to basolateral transmigration as part of the epithelial barrier penetration process, we generated libraries of 22 isogenic, genome-tagged strains of each our wt and our ΔsseB S. Typhimurium strain. This neutral barcoding method allows the quantitative evaluation of population dynamics and bottlenecks during spread between different anatomical sites. 25 Since it does not rely on total bacterial counts, it allows the detection of impaired mucosal translocation and thus a reduced number of translocation events, independent of the reduced survival i.e. the absolute number of SPI2 T3SS mutant versus wt S. Typhimurium in tissue resident and systemic phagocytes. 4,5 Following oral administration of the bacterial library, i.e. an equal mixture of all 22 isogenic strains of either wt or ΔsseB S. Typhimurium, the presence and relative abundance of individual tagged strains was analyzed in the infection inoculum as well as the total spleen tissue, the total liver tissue and the intestinal epithelial cells isolated at day 4 p.i. by plating and subsequent DNA sequencing. Population similarities between the intraepithelial bacterial population and the population in liver or spleen tissue indicated by the calculated Bray-Curtis Index were markedly reduced in the absence of a functional SPI2 T3SS compared to wt S. Typhimurium (Figure 4(a-c)). These findings are consistent with a role of SPI2 T3SS translocated effector molecules in enterocyte apical to basolateral transmigration with subsequent cell egress and systemic dissemination of S. Typhimurium in the neonate host.
In order to more specifically address the role of the SPI2 T3SS during enterocyte egress and exclude an influence of the surrounding tissue on the observed phenotype, we employed a reductionist approach using highly polarized murine small intestinal epithelial m-IC cl2 cells grown on transwell filter inserts. Enterocyte invasion by S. Typhimurium in this model is SPI1 T3SSdependent as illustrated by significantly reduced numbers of intraepithelial SPI1 T3SS mutant ΔinvC S. Typhimurium in the classical gentamicin kill assay ( Fig. S1H and I). Consistent with the in vivo phenotype (Figure 1(c)), deletion of the SPI2 T3SS translocon component SseB or the ATPase SsaN required for SPI2 T3SS-mediated effector molecule translocation did not alter intraepithelial survival and growth ( Fig. S1J and K). Notably, the absence of a functional SPI2 T3SS significantly reduced the bacterial egress at the basolateral plasma membrane (Figure 4(d-e)) and this difference was not explained by an altered cell viability (Fig. S1L). 18,26 Thus, a functional SPI2 T3SS was dispensable for intraepithelial survival but required for epithelial m-IC cl2 cell egress. Consistently, wt S. Typhimurium were identified in the lamina propria close to the epithelial layer in vivo (Fig. S2).

Discussion
The results on S. Typhimurium in polarized intestinal epithelial cells obtained using our neonatal murine infection model suggest significant and previously ill-defined differences to the reported situation in phagocytes with respect to intracellular survival, SCV formation, and cellular positioning. In phagocytes, SPI2 T3SS effector molecules are mainly induced upon innate immune stimulationmediated SCV acidification and exert a critical and non-redundant role in vacuolar remodeling, intracellular bacterial survival 2,3,12 and macrophage e1973836-6 polarization. 27 In contrast, intracellular bacterial survival in polarized neonatal intestinal epithelial cells in vivo does not require an intact SPI2 T3SS. Instead, SPI1 T3SS virulence factors, in particular SipA, SopE 2 , and SopB, contribute to SCV formation in polarized epithelial cells. 17 Endosomal rupture, release of S. Typhimurium into the cytosol and rapid cytosolic proliferation associated with SPI1 effector gene expression has been described. 28,29 In our model, intraepithelial S. Typhimurium exhibited low SPI1 effector gene expression and we have been unable to detect endosomal rupture or cytosolic bacteria by electron microscopy. Although we have no evidence for endosomal rupture of intraepithelial SCVs, however, we cannot exclude the occurrence in a minority of cells possibly associated with rapid exfoliation into the intestinal lumen.
In non-polarized cells, SPI2 T3SS effectors were shown to induce the formation and extension of Salmonella-induced filaments (Sifs), longitudinal membrane tubules that extend from the SCV and promote intravacuolar nutrition. 9,24,30 In contrast, the intestinal epithelium represents an absorptive surface for dietary nutritional substances and intraepithelial S. Typhimurium in vivo have access to luminal energy-rich material even in the absence of a functional SPI2 T3SS, rendering their intraepithelial growth independent of important nutrient-uptake systems. The observed decrease in the number of S. Typhimuriumpositive enterocytes at later time points is most likely due to the start of enterocyte proliferation, migration along the crypt-villus axis and exfoliation known to start at approximately two weeks after birth as part of the postnatal mucosal tissue development. 31 Also, the role of the SPI2 T3SS for the positioning of intracellular S. Typhimurium seems to differ between non-polarized cells and polarized enterocytes. In non-polarized cells, SPI2 T3SS effectors were reported to mediate paranuclear localization. 6,24,30 In contrast, our findings in highly polarized intestinal epithelial cells in vivo suggest that the interaction of SPI2 T3SS effector molecules with the microtubule network and trafficking machinery facilitates apical to basolateral transmigration. Importantly, this is fully consistent with the different organization of the microtubule network in polarized and non-polarized cells. In nonpolarized cells, microtubules extend radially from the microtubule organizing center (MTOC) localized close to the paranuclear Golgi apparatus, whereas in polarized enterocytes they extend from the apical membrane to the basolateral side to facilitate vectoral vesicle transport. 32 The continued presence of SPI2 T3SS mutant microcolonies mainly at the apical pole is also reminiscent of the appearance of subapical vesicles in genetic disorders of the intracellular trafficking machinery. 33 Thus, the enhanced size and block in apical to basolateral transmigration of SPI2 T3SS mutant and somewhat less pronounced also for SifA mutant SCVs as well as the enhanced size of SseFG and PipB2 deficient S. Typhimurium SCVs suggest a critical role of the SPI2 T3SS and a contribution of the translocated effector molecules SifA, SseFG and PipB2 in the microtubulemediated transport of the SCV through the epithelial cell in vivo.
Our results thus explain the recent finding that the SPI2 T3SS exerts a critical role during the initial course of an oral S. Typhimurium infection. 12 They are also in accordance with the recent report on SPI2 T3SS-dependent transport of Salmonella flagellin across the epithelium, 34 and live-imaging studies of the translocation of S. Typhimurium. 11 However, in the latter study, bacterial transcytosis was observed in the cecal epithelium, occurred as SCV containing a single bacterium, and was completed within a few hours. Notably, additional alternative SPI2 T3SS-dependent and independent pathways that facilitate epithelial barrier penetration may exist, explaining the low but detectable spread of SPI2 T3SS mutant S. Typhimurium to systemic sites in our neonatal infection model.
Finally, the enlarged SCVs following infection with SPI2 T3SS mutant S. Typhimurium are reminiscent of the phenotype described for ΔssaV and ΔsseB Salmonella in hepatic leukocytes. 35 Although impaired progression from the apical to the basolateral side of polarized epithelial cells can explain the observed continuous intravacuolar growth, the detection of giant SPI2 T3SS mutant Salmonella SCVs in myeloid cells indicates an additional role of SPI2 in cell egress, both from immune cells and polarized enterocytes. Consistently, a reduced number of infectious hepatic foci was observed in the above-mentioned study after systemic SPI2 mutant Salmonella infection. 35 Similarly, a recent publication reported that SPI2-deficient Salmonella tissue persisters were impaired in their ability to reach the gut lumen and mediate spread of a conjugative plasmid to luminal recipient bacteria. 13 Thus, the reduced enterocyte apical to basolateral transmigration by SPI2 deficient S. Typhimurium could explain the critical contribution of SPI2 T3SS effector molecules to epithelial intestinal mucosal inflammation observed in murine models of Salmonella enterocolitis. 36,37 Our results assign the SPI2 T3SS a new function in vivo. Future studies are warranted to reveal the mechanistic details of the described process and to further investigate S. Typhimurium enterocyte egress at the basolateral side to complete the process of epithelial transcytosis.

Bacterial strains and plasmids
Salmonella enterica subsp. enterica serovar Typhimurium (S. Typhimurium) ATCC14028 (NCTC12023) and isogenic SPI1 T3SS (ΔinvC), SPI2 T3SS (ΔsseB, ΔssaN) and SPI-2 T3SS effector (ΔsifA, ΔsseFG, ΔpipB2) mutants as well the respective complemented strains were used for in vivo and in vitro infection experiments. For visualization in tissue sections and flow cytometric sorting, bacterial strains harbored plasmids encoding constitutively expressed mCherry, DsRed or green fluorescent protein (GFP) (pGFP, AmpR, kindly provided by Brendan Cormack, Stanford, USA). The generation of directed deletion mutants of SPI2 T3SS effector genes and other mutations of S. Typhimurium was performed as described previously. 38 The strains and plasmids used in this study are listed in the supplemental table 1.

In vitro transcytosis assay in polarized epithelial cells
Small intestinal epithelial m-IC cl2 cells were cultured in a defined medium and passaged every 2-3 days as previously described. 39 Cells were seeded on 3 µm pore size PET ThinCert transwell inserts (Greiner Bio-one, Kremsmünster, Austria) and cultured for 10-12 days with medium change every other day to obtain a covalent layer of polarized epithelial cells. Integrity of the cellular monolayer was verified by monitoring of the transepithelial electrical resistance (TEER). wt or SPI2 T3SS ΔsseB and ΔssaN mutant S. Typhimurium as well as their complemented strains ΔsseB (psseB) and ΔssaN (pssaN) were grown overnight at 37°C, diluted 1:10 and sub-cultivated with mild agitation at 37°C, until mid-logarithmic growth was reached (OD 600 : 0.5). Bacteria were adjusted by dilution, added at a multiplicity of infection (MOI) of 10:1 to the apical medium compartment and centrifuged at 1,500 rpm for 5 min. Following incubation for one hour, the medium in the apical and basolateral compartment was replaced with fresh medium supplemented with 100 µg/mL gentamicin and incubated for another hour. The number of intracellular bacteria at 2 h was determined by washing three times in warm PBS followed cell lysis in 0.

Ethics
All animals were handled in accordance with regulations defined by FELASA and the national animal welfare body GV-SOLAS (http://gv-solas. de). In vivo experiments were performed in compliance with the German animal protection law (TierSchG) and approved by the local animal welfare committee (Niedersächsisches Landesamt für Verbraucherschutz und Lebensmittelsicherheit Oldenburg, Germany; Landesamt für Natur, Umwelt und Verbraucherschutz, North Rhine Westfalia, approval 33.14.42502-04-12/0693, 81-02.04.2017. A397 and Landesamt für Gesundheit und Soziales Berlin (LaGeSo), approval 15/0293 including all approved changes. The number of animals per group was defined prior to the start of the experiments based on previous results and the expected variation of the determined read out in the application for ethical approval (see above).

Infection experiments
Adult C57BL/6 N wild type were obtained from Charles River Laboratory (Sulzfeld, Germany) and B6.129P2(SJL)-MyD88tm1.1Defr/J (MyD88-/-, stock no. 009088) mice from the Jackson Laboratory (Bar Harbor, USA), respectively. Animals were bred locally under SPF conditions. Bacteria were cultured in Luria Bertani (LB) broth containing 100 µg/mL ampicillin overnight at 37°C and 200 rpm. The next day, cultures were diluted 1:10 in fresh LB medium at identical incubation conditions until an OD 600nm of approximately 0.5 was reached. The bacterial pellet was washed in PBS twice and cultures were adjusted to the indicated infection dose and orally applied to 1-day-old mice in a volume of 1 μL PBS containing 10 2 CFU if not stated otherwise. Infection with ΔsifA, ΔsseFG and ΔpipB2 mutant strains (plus the comparative wt S. Typhimurium strain) in Figure 2(f-g) were administered at 10 4 CFU to allow an earlier analysis of the SCV size already at day 2 p.i. At later time points, SCVs generated by ΔsifA, ΔsseFG and ΔpipB2 mutant S. Typhimurium become unstable as previously reported. 40,41 Infections with the 22 WITS strains were performed with an inoculum of 10 3 CFU to allow the generation of an equal mixture of all WITS strains (see below). Litters were randomly assigned to infection with wildtype or SPI2 deficient S. Typhimurium. At the indicated time points p.i., animals were sacrificed and liver and spleen were harvested and homogenized in sterile PBS. Viable counts were obtained by serial dilution and plating on LB agar plates supplemented with the appropriate antibiotic. For the 1:1 coinfection with S. Typhimurium carrying the GFP or DsRed reporter plasmid, cultures were grown separately, adjusted and mixed 1:1 prior to oral challenge.

Isolation of primary intestinal enterocytes
Primary intestinal epithelial cells were isolated from small intestinal tissues of infected neonates as described previously. 14 Briefly, intestinal tissues of neonate mice were cut in small pieces and incubated in 30 mM EDTA PBS at 37°C for 10 min. Epithelial cells were detached from the underlying tissue by vigorous shaking and the cell suspensions were passed through a 100 µm nylon cell strainer (VWR). Cells were washed in 10% FCS/PBS, collected by centrifugation and the purity was verified by flow cytometry (E-cadherin + CD45 − ). For the cultural quantification of intraepithelial S. Typhimurium in vivo, isolated epithelial cells were incubated in gentamycin (100 µg/mL) to remove extracellular bacteria, repeatedly washed and lysed followed by serial dilution and plating as recently described. 14 To determine the bacterial organ count, organ homogenate or cell lysate was plated on LB agar supplemented with 100 µg/mL ampicillin by replica-plating.

Immunofluorescence staining
Tissues were fixed in paraformaldehyde, embedded in paraffin and cut in 3 µm slices. Sections were deparaffinized in xylene and rehydrated in ethanol. Antigen retrieval was conducted in 10 mM sodium citrate for 10 min and blocking with 10% serum in PBS/BSA solution prior to immunofluorescence staining. For immunostaining, a chicken anti-GFP (dilution 1:500, Abcam), a rabbit anti-S. Typhimurium (dilution 1:1,000, Abcam), a mouse anti-E-cadherin (dilution 1:200, BD Transduction Laboratories), a rat anti-LAMP-1 (dilution 1:500, Developmental Studies Hybridoma Bank), and a rabbit anti-mCherry (dilution 1:500, Abcam) were used in the appropriate combination with a fluorophore conjugated secondary antibody (Jackson ImmunoResearch) as indicated. Slides were mounted in Vectashield (Vector) containing DAPI and pictures were taken with an ApoTome equipped Observer Z.1 fluorescence microscope equipped with axiovision software from Zeiss (Jena, Germany) or a DMI6000B fluorescence microscope equipped with LAS X software from Leica (Wetzlar, Germany). No blinding of tissue sections prior to the visual analysis was performed.

e1973836-10
Since the visualization of the SCV membrane markers LAMP-1 and LAMP-2 by immunostaining and electron microscopy as well as the ultrastructural studies and previous work (Zhang et al., 2014 and unequivocally demonstrated an intact SCV membrane around the intraepithelial bacteria in neonatal enterocytes, we measured the size of the bacterial microcolonies as a surrogate marker for the size of the SCVs. Images for the visual quantitative analysis of the SCV number/field of view were carefully selected to ensure a similar number of epithelial cells on each field of view prior to counting. In most figure panels, data points from individual animals are indicated by colored symbols. This presentation clearly shows that the observed variation represents variation between individual SCVs and not between individual animals.

Transmission electron microscopy
Neonate mice were sacrificed 4 days p.i. The small intestinal tissue was dissected and fixed in 1% glutaraldehyde (for the ultrastructural analysis) or 4% paraformaldehyde (for immunolabeling) in 200 mM HEPES, pH 7.4 for several days. The distal end was cut into approximately 2 mm long gut sections that were subsequently longitudinally opened at one side. For the ultrastructural analysis ( Fig. 1 and 2, Fig. S2 and S3), samples were stained with 1% osmium tetroxide prepared in 1.5% potassium ferricyanide for 1 h on ice, followed by 2% aqueous uranyl acetate for 2 h, and then dehydrated with a graded ethanol series. The tissues were progressively infiltrated with epon over 2 days and then polymerized in the oven overnight. Thin, 60-80 nm sections were prepared using an Ultracut UCT ultramicrotome (Leica Microsystems) and mounted onto formvar-coated copper grids. Sections were stained with lead citrate, and imaged in a JEM-1400 TEM microscope (Jeol) equipped with a TemCam-F216 camera and the EM MENU software (both Tvips). For on-section immunolabeling ( Fig. S1E and F), tissue pieces were embedded in 12% bovine gelatine and infiltrated with 2.3 M sucrose overnight. Small blocks were mounted on aluminum pins and snap frozen in liquid nitrogen. A UC7 cryo-ultramicrotome (Leica Microsystems) was used to prepare Tokuyasu sections. 80 nm thin sections were cut at −110°C, and 250 nm semi-thin sections were cut at −90°C. Using a 1 + 1 mixture of 2% methyl cellulose and 2.3 M sucrose, thin sections were transferred onto formvar-coated copper grids, while semi-thin sections were transferred onto glass coverslips. For labeling, 1% fish skin gelatine was used for blocking and for dilution of labeling probes. Grids were incubated with 5x diluted LAMP-1 (rat monoclonal Ab, clone 1D4B) and LAMP-2 (rat monoclonal Ab, clone ABL-93) hybridoma supernatant (DSHB, University of Iowa) for 30 min, followed by 250x diluted rabbit anti-rat bridging antibody (Rockland) for 15 min and 50x diluted protein A gold (CMC, University Medical Centrum Utrecht) for 30 min. Sections were embedded in 0.2% uranyl acetate in 1.8% methyl cellulose. Grids were imaged in a JEM-1400 TEM microscope equipped with a TemCam-F216 camera and the EM MENU software. Glass coverslips were incubated with 10x diluted LAMP-1 antibody for 30 min, followed by goat anti-rat Cy3conjugated secondary Ab for 30 min, and 1 µg/mL DAPI for 5 min. Coverslips were mounted on a glass slide with Mowiol 4-88, and imaged in an inverted epifluorescence Leica DMIRBE microscope, using a 60x objective lens and a Leica application Suite (LAC) software v 3.8 (both Leica Microsystems).

Transcriptomic analysis of intraepithelial S. Typhimurium
One-day-old mice were orally infected with 10 2 CFU wt S. Typhimurium carrying a mCherry reporter plasmid. An aliquot of the bacteria used for the infection was stored for subsequent analysis (inoculum). At 4 days p.i., mice were sacrificed and primary intestinal epithelial cells were isolated from small intestinal tissues as described. E-cadherin+ CD56-cells were sorted by flow cytometry in mCherry positive (S. Typhimurium-infected) and mCherry negative (uninfected bystander) epithelial cells. Total RNA was isolated from the bacterial inoculum as well as S. Typhimurium-positive enterocytes and transcriptomic analysis was performed by RNA-seq as described in Westermann et al., 2016. 42 Notably, due to the fragility and low number of infected enterocytes (approximately 1%) 14 in each animal and the relatively low number of intraepithelial wildtype bacteria, the FACS sorted material of seven infected litters (with approximately 49 neonates in total) had to be pooled to obtain sufficient material for sequencing of bacterial RNA. This explains why only the most abundant S. Typhimurium virulence factors were detected in the intraepithelial fraction and no biological replicates could be generated. Briefly, RNA samples were treated for 45 min with DNase I (NEB) to remove genomic DNA remnants and sheared via ultrasound sonication (4 pulses of 30 s at 4°C each) to generate ~200-400 bp average fragments. Fragments <20 nt were removed using the Agencourt RNAClean XP kit (Beckman Coulter Genomics). The samples were poly(A)-tailed using poly(A) polymerase and the 5ʹ triphosphate structures were removed using tobacco acid pyrophosphatase (TAP). An RNA adapter was ligated to the 5ʹ monophosphate of the RNA fragments. First-strand cDNA synthesis was performed using an oligo(dT)adapter primer and the M-MLV reverse transcriptase (NEB). The resulting cDNA was PCR-amplified to about 10-20 ng/µL using a high fidelity DNA polymerase (barcode sequences for multiplexing were part of the 3ʹ primers). cDNA libraries were purified using the Agencourt AMPure XP kit (Beckman Coulter Genomics) and analyzed by capillary electrophoresis (Shimadzu MultiNA microchip electrophoresis system). For sequencing, cDNA samples were pooled in approximately equimolar amounts. The cDNA pool was sizefractionated in the size range of 150-600 bp using a differential clean-up with the Agencourt AMPure kit. Single-end sequencing (100 cycles) was performed on a HiSeq2000 platform. Illumina reads in FASTQ format were trimmed with a Phred quality score cutoff of 20 using fastq_quality_trimmer from FASTX toolkit version 0.0.13 (http://hannon lab.cshl.edu/fastx_toolkit/). Reads shorter than 20 nt after adapter-and poly(A)-trimming were discarded prior to the mapping. The mapping was performed using the READemption pipeline (version 0.3.5) 43 and the short read mapper segemehl and its remapper lack (version 0.2.0) 44 against the published genome sequence of S. Typhimurium ATCC 14028 (NCBI RefSeq accession number NC_016856) and mouse genome (GENCODE M2, GRCm38.p2) in parallel. Mapped reads with an alignment accuracy <90%, murine-mapped reads as well as crossmapped reads (i.e. reads which could be aligned equally well to the bacterial and mouse genome) were discarded and only S. Typhimurium-specific reads were considered for differential expression analysis. For the data presentation, the SPI1 and −2 associated genes were selected based on the detection of sequence reads in either the inoculum or intraepithelial fraction by RNA-seq and their established role and assignment to SPI1 or SPI2. For comparison of the two fractions, the genespecific sequence reads were normalized to the total number of aligned S. Typhimurium reads in that library. The resulting normalized sequence read counts for SPI1 and SPI2 genes are displayed in Figure 1(a) and the respective counts for the two established house keeping genes rpoD and gyrA are shown as control. 22 individual strains were barcoded with a 4 nucleotide sequence tag, introducing an artificial stop codon into the endogenous proV gene as described previously. 23 Equal mixtures of each barcoded strain grown separately were orally applied to 1-day-old C57BL/6 N mice in 1 µl PBS at a final concentration of 10 3 CFU (i.e. approximately 5 × 10 1 per strain). At 4 days p.i., the animals were sacrificed and small intestine, liver and spleen tissue were harvested. IEC isolation was performed as described above. Total spleen and liver homogenates and total IEC lysate were plated on ampicillin containing LB agar plates. After incubation at 37°C overnight, bacteria were harvested in 4 mL PBS. Bacteria were harvested by centrifugation for 10 min at 300 rpm, followed by gDNA isolation and amplification of proV as recently described. 25 Samples were sequenced on an Illumina MiSeq system using MiSeq Reagent Kits (73 cycles/323 cycles) at the Center for Biotechnology (CeBiTec) of the University of Bielefeld, Germany. From a total of 22 generated sequence tags, 21 tags were recovered from the inoculum and tissue samples. For calculation of sequence identities, only barcoded strains contributing to >1% to the overall population were considered. The calculated Bray-