Age-Dependent Enterocyte Invasion and Microcolony Formation by Salmonella

The coordinated action of a variety of virulence factors allows Salmonella enterica to invade epithelial cells and penetrate the mucosal barrier. The influence of the age-dependent maturation of the mucosal barrier for microbial pathogenesis has not been investigated. Here, we analyzed Salmonella infection of neonate mice after oral administration. In contrast to the situation in adult animals, we observed spontaneous colonization, massive invasion of enteroabsorptive cells, intraepithelial proliferation and the formation of large intraepithelial microcolonies. Mucosal translocation was dependent on enterocyte invasion in neonates in the absence of microfold (M) cells. It further resulted in potent innate immune stimulation in the absence of pronounced neutrophil-dominated pathology. Our results identify factors of age-dependent host susceptibility and provide important insight in the early steps of Salmonella infection in vivo. We also present a new small animal model amenable to genetic manipulation of the host for the analysis of the Salmonella enterocyte interaction in vivo.


Introduction
Non-typhoidal Salmonella are one of the major causal agents of bacterial gastrointestinal infections in all age groups worldwide. In addition, they also significantly contribute to systemic infection particularly in the pediatric population. Non-typhoidal Salmonella together with pneumococci and group B streptococci are among the most frequent causal agents of sepsis and meningitis in neonates and young infants in Africa [1][2][3][4][5]. Transmission of this highly endemic bacterium occurs perinatally from infected mothers or after birth from infected family members via the fecal-oral route. The underlying mechanisms of the enhanced susceptibility to systemic disease after infection during the postnatal period have not been systematically investigated.
Salmonella pathogenicity is conferred by horizontally acquired chromosomal regions so called Salmonella pathogenicity islands (SPI) that encode sets of virulence factors required for the various steps in microbial virulence. SPI1 encodes a type III secretion systems (T3SS) and a set of effector molecules that are translocated into the target cell. It mediates pathogen-induced internalization in non-phagocytic cells [6][7][8]. The critical importance of cellular invasion by Salmonella and intracellular proliferation in nonphagocytic cells is supported by epidemiological data and has been extensively studied in vitro [6,[9][10]. Although intracellular Salmonella have been observed in epithelial cells and cells of the lamina propria also in vivo, this host cell interaction and the functional importance have remained less defined [7][8][11][12][13][14][15]. Instead, translocation through microfold (M) cells has been demonstrated in adult animals. Such cells overlay Peyer's patches and forward luminal antigens and bacteria to the underlying immune cells [16][17][18]. Similarly, uptake via lamina propria resident myeloid cells that generate membrane extrusions extending into the intestinal lumen has been proposed to facilitate Salmonella barrier penetration [19]. More recently, rapid translocation via the colonic epithelium in the absence of intracellular proliferation was observed [15]. The underlying host and bacterial factors facilitating enterocyte invasion, intraepithelial proliferation and mucosal translocation have, however, remained largely undefined.
Here we comparatively analyzed oral Salmonella infection of neonate and adult mice. We demonstrate marked age-dependent differences in intestinal colonization, mucosal translocation and systemic spread. We observed Salmonella enterocyte invasion in neonate mice in vivo and characterize intraepithelial microcolony formation, epithelial translocation and epithelial innate immune stimulation after oral infection. We characterize developmental peculiarities of the newborn's intestinal epithelium that allow enterocyte invasion and characterize the contribution of Salmonella virulence mechanisms. Our results identify age-dependent factors of infection susceptibility and provide a novel in vivo model that allows the use of genetically modified hosts to investigate the intimate host microbial interaction at the intestinal epithelium.

Mucosal translocation in neonates depends on Salmonella-induced internalization by non-phagocytic cells
One-day-old C57BL/6 neonates were orally infected with various inocula (10 2 -10 5 ) of mid-log grown Salmonella enterica subsp. enterica sv. Typhimurium (S. Typhimurium). Colony counts and visualization of bioluminescent S. Typhimurium in the intestine using optical imaging technology demonstrated rapid colonization of the small intestine and colon ( Fig. 1A-C). Even low dose infection (10 2 CFU) resulted in systemic spread with maximal bacterial counts in spleen and liver at day 4 post infection (p.i.) (Fig. 1D-E). The absence of bacteria in lung tissue excluded accidental primary pulmonary infection (Fig. S1A). Comparative analysis of 1-day-old, 6-day-old and streptomycin pretreated adult animals at day 4 after oral administration of 10 2 , 5610 2 , and 5610 8 CFU S. Typhimurium, respectively, demonstrated the efficient colonization of the neonate intestine and the marked spread to liver and spleen tissue in 1-day-old animals ( Fig. 1F-G). We therefore focused on 1-day-old animals and applied low dose infection and analyzed animals at 4 days p.i. for all subsequent experiments, if not stated otherwise.
As SPI1 is involved in Salmonella internalization in nonphagocytic cells, we next analyzed the requirement of Salmonella SPI1 effector translocation for oral infection of neonatal mice. Spread to liver (p,0.001), spleen (p,0.001) and mesenteric lymph nodes (MLN, p,0.001) was practically abolished in the absence of a functional SPI1 system (DinvC) ( Fig. 2A-C). In fact, spleen tissue of all mice and liver and MLN tissue of the majority of neonates (8/14 and 18/23, respectively) remained sterile at day 2 and 4 after infection by SPI1 deficient Salmonella. Similarly, systemic dissemination of DinvC Salmonella was also highly significantly reduced after high dose (10 5 CFU) infection (Fig. S2D). In accordance with these results, immunostaining (Fig. 2D), plating of isolated epithelial cells after gentamicin-treatment to remove extracellular Salmonella (Fig. 2E-F) and flow cytometry (Fig. 2G) detected enterocytes infected by WT but not SPI1-deficient Salmonella. Both, WT and DinvC Salmonella after high and low dose infection spontaneously colonized the intestinal tract ( Fig.  S2A-C). The reduced total small intestinal organ counts at later points after low dose infection may result from the lack of intraepithelial bacteria ( Fig. 2E-F).

Requirement for enterocyte invasion in the absence of M cells in the neonate intestine
In contrast to the critical role of SPI1 in neonate mice, intestinal colonization but also spread to spleen, liver and MLN in adult streptomycin-pretreated animals was largely SPI1-independent ( Fig. 2H and Fig. S2E). To identify differences that might account for the requirement of SPI1 in neonate but not adult animals, the gene expression profile of isolated primary neonate and adult epithelial cells was compared. Unexpectedly, genes of differentiated M cells such as Spi-B and Ccl9 were found to be markedly reduced in neonate epithelial cells (Fig. 3A-B). In adult hosts, M cell-mediated bacterial translocation represents the major entry pathway of Salmonella [16,17]. To confirm the age-dependent differentiation of M cells, mRNA expression of the M cell specific transcription factor Spi-B was examined [18]. As expected, Spi-B expression was found to be strongly diminished in neonates and an increase was only observed starting at day 8 after birth (Fig. 3C). Also, immunostaining of intestinal tissue for the established M cell markers glycoprotein 2 (gp2), Ccl9 and Ulex europaeus agglutinin (UEA)-1 confirmed the appearance of M cells only after the neonatal period (Fig. 3D). Finally, wildtype (WT) and fimD mutant Salmonella disseminated to a similar degree to the spleen and liver tissue after 4 days infection of 1-day-old neonates ( Fig.  S3A). fimD mutant Salmonella are unable to express type I pili and cannot attach to the M cell surface protein gp2. This significantly reduces their ability to invade the adult host via M cells [20]. In contrast to a recent report on Salmonella-induced Spi-B expression in adult animals, Spi-B expression in neonate mice was reduced rather than enhanced following Salmonella infection (Fig.  S3B) [21]. Hence, Salmonella in neonate mice spontaneously colonize the intestine, invade enterocytes, penetrate the mucosal barrier and spread to systemic organs in a SPI1-dependent but M cell-independent fashion.

Epithelial invasion leads to intraepithelial proliferation and microcolony formation
Salmonella infection of neonatal enterocytes in vivo was subsequently studied in more detail. Immunostaining revealed Salmonella-positive enterocytes (Fig. 4A). Unexpectedly, Salmonella generated multi-bacterial intraepithelial colonies of variable size consistent with the formation of Salmonella-containing vacuoles (SCV) previously observed in epithelial cell lines in vitro ( Fig. 4A-B). Quantification of Salmonella-positive cells was achieved by flow cytometric analysis on isolated enterocytes.

Author Summary
Non-typhoidal Salmonella are among of the most prevalent causative agents of infectious diarrheal disease worldwide but also very significantly contribute to infant sepsis and meningitis particularly in developing countries. The underlying mechanisms of the elevated susceptibility of the infant host to systemic Salmonella infection have not been investigated. Here we analyzed age-dependent differences in the colonization, mucosal translocation and systemic spread in a murine oral infection model. We observed efficient entry of Salmonella in intestinal epithelial cells of newborn mice. Enterocyte invasion was followed by massive bacterial proliferation and the formation of large intraepithelial bacterial colonies. Intraepithelial, but not non-invasive, extracellular Salmonella induced a potent immune stimulation. Also, enterocyte invasion was required for translocation through the mucosal barrier and spread of Salmonella to systemic organs. This requirement was due to the absence of M cells, specialized epithelial cells that forward luminal antigen to the underlying immune cells, in the neonate host. Our results identify age-dependent factors of host susceptibility and illustrate the initial phase of Salmonella infection. They further present a new small animal model amenable to genetic manipulation to investigate the interaction of this pathogen with epithelial cells and characterize the early steps in Salmonella pathogenesis.  0.2460.12% and 0.8060.61% of all murine epithelial cell adhesion molecule (EpCAM) + cells at 2 and 4 days p.i., respectively were found to be infected (uninfected controls: 0.0760.03%, p,0.05). Co-infection using a 1:1 mixed inoculum of genetically green-and red-labeled bacteria revealed solely single-colored intracellular colonies (Fig. 5A). This suggests that individual microcolonies originated from a single event of bacterial invasion.
In sharp contrast, only minute numbers of (questionably) intraepithelial Salmonella without any indication of bacterial proliferation were detected in tissue sections of adult mice after high dose (1-2610 8 CFU) infection despite numerous attempts both with and without streptomycin pretreatment (data not shown). Consistently, flow cytometric analysis did not detect Salmonella-infected intestinal epithelial cells (Fig. S2F and). To identify molecular differences that might account for the agedependent susceptibility of the epithelium to Salmonella invasion, the gene expression profile of neonate and adult enterocytes was compared (Fig. 3A). Known SCVs constituents (with the notable exception of iNOS) and autophagosomal markers were equally expressed in neonate and mature adult epithelial cells. However, the neonate intestine exhibited a markedly reduced mucus layer thickness and mucin glycoprotein expression (Fig. S3C-D). Also, reduced cyclin expression was noted in accordance with minimal epithelial cell turn-over and reduced crypt-villus migration in neonates [22,23]. Finally, a severely reduced spectrum of antimicrobial peptides was observed consistent with previous reports [24]. All of these factors might facilitate epithelial invasion, prolong the lifetime of infected epithelial cells and thereby ultimately allow intraepithelial proliferation and microcolony formation.

Invasion-dependent innate immune stimulation of epithelial cells
We next investigated the neonate's host response to Salmonella infection. Highly enriched primary enterocytes (Fig. S4) were subjected to quantitative RT-PCR. A significant time-dependent increase in Cxcl2 and Cxcl5 mRNA expression was measured after infection by WT or the complemented DinvC pinvC but not SPI1-deficient non-invasive Salmonella (Fig. 6A-B). Global gene array analysis confirmed the absence of any detectable increase in epithelial gene expression upon administration of invC deficient, non-invasive Salmonella (Fig. 6C). It further identified a large number of additional genes involved in metabolism, cellular responses and intercellular communication induced after Salmonella WT infection (Fig. 6D). In accordance with the requirement of enterocyte invasion for innate immune stimulation (Fig. 6C) and the observed age-dependent susceptibility to enterocyte invasion ( Fig. 2D-G and S2F), Cxcl2 and Cxcl5 mRNA expression in enterocytes isolated at day 4 p.i. from infected 6-day-old or adult mice was severely reduced (Fig. 6E-F).
We next examined the innate immune receptors and signaling adaptors involved in Salmonella-induced enterocyte stimulation. Expression of Cxcl2 and Reg3c mRNA by epithelial cells was severely reduced in the absence of toll-like receptor (Tlr)4, MyD88 as well as Unc93B1 and Tlr9. Although some variability in the expression levels was noted, significantly decreased mRNA expression was also noted in epithelial cells devoid of Tlr2, Tlr5, or Nod2 (Fig. 7A-B). Of note, intestinal colonization (Fig. 7C), enterocyte invasion ( Fig. 7E-G) and dissemination to systemic organs ( Fig. 7C-D) was observed also in the absence of the most potently stimulated innate immune receptor, Tlr4. Although the production of proinflammatory mediators significantly enhanced the recruitment of polymorphonuclear cells and blood macrophages to the site of infection, no gross loss of epithelial barrier integrity or histopathological signs of tissue destruction were observed at day 4 p.i. in contrast to the situation in the adult host ( Fig. 8) [25]. Thus, Salmonella in the neonate efficiently invades small intestinal enterocytes, activates innate immune responses mainly via Tlr stimulation and induces the formation of intraepithelial microcolonies, a hallmark of the Salmonellaenterocyte interaction.

Discussion
Previous work had demonstrated enterocyte invasion in oral and intestinal loop infection models and ex vivo tissue explants of calfs, rabbits, swine and guinea pigs [26][27][28][29][30][31][32]. However, these host animals are not amenable to genetic manipulation and studies are restricted to the early course of infection. Enterocyte invasion, rapid egress at the basolateral side of the epithelium and the presence of Salmonella in lamina propria cells has also been described in the mouse large intestine [7,[11][12][13][14][15]. The present study now provides a new small animal model that allows the use of genetically modified hosts to analyse intracellular proliferation in small intestinal epithelial cells and the formation of microcolonies by Salmonella in vivo. It might thereby facilitate a better understanding of a hallmark of Salmonella pathogenesis in vitro, the internalization by non-phagocytic epithelial cells and the formation of intraepithelial microcolonies, so called Salmonella containing vacuoles (SCVs).
Invasion of non-phagocytic epithelial cells is facilitated by virulence factors encoded by the pathogenicity island SPI1 [6]. The critical role of SPI1 for enteric disease is illustrated by its presence in all subspecies of Salmonella enterica [33] and its high prevalence in clinical Salmonella isolates [9]. In accordance, SPI1 effector proteins are required to cause diarrhea and mucosal inflammation in several in vivo models although also SPI2 effector molecules contribute to intraepithelial survival and proliferation [25,[34][35][36]. The analysis of the interaction of S. Typhimurium with polarized epithelial cells within their anatomical environment represents a prerequisite to understand the functional role of individual bacterial virulence factors.
Whereas previous reports using streptomycin pretreated adult animals failed to demonstrate infection of the small intestinal epithelium, we observe efficient Salmonella invasion of enterocytes in neonate mice. Known host constituents of mature SCVs and autophagosome factors are similarly expressed by neonatal and adult epithelial cells [37][38][39]. However, a number of fundamental differences exist between the neonate and adult gut epithelium in mice. For example, we demonstrate that the synthesis of various mucin glycoproteins and thus the thickness of the mucus layer are significantly enhanced in adult animals. The mucus layer was shown to significantly impair mucosal translocation of Salmonella in adult mice [40]. Also, we and others have shown that the antimicrobial peptide repertoire is severely reduced in the neonate intestine in the absence of mature Paneth cells. In adult mice, Paneth cell-derived antimicrobial peptides cells significantly influence the course of enteropathogen infection [41,42] and cooperate with the mucus layer to generate an antibacterial and antiinflammatory shield [43]. Finally, the well-established constant renewal of the epithelium associated with enterocyte migration and exfoliation at the villus tip is not yet established in newborn mice [22,23]. The reduced epithelial cell turn-over is the consequence of the lack of crypts and the reduced pool of rapidly proliferating cells during this early developmental stage. Thus, enterocytes during the neonatal period remain longer at the same anatomical position. This might allow Salmonella to proliferate intracellularly and form microcolonies. Shedding of Salmonellainfected enterocytes at the villus tip was previously observed in bovine, pig and rabbit intestinal loop models and might represent an important mechanism to remove infected enterocytes [28,30,44]. Finally, the prolonged presence of the more diverse enteric microbiota in the adult host may contribute to the enhanced resistance to Salmonella enterocyte invasion and this issue requires future investigations.
In the neonate host both Salmonella enterocyte invasion and mucosal translocation were dependent on SPI1. This suggests that penetration of the neonate's intestinal barrier occurs secondary to the observed enterocyte invasion. This is in contrast to the situation in adult mice but also other species. Here, M cells were shown to represent the major port of entry [16][17][18]29]. M cells as part of the follicle-associated epithelium overlaying Peyer's patches facilitate uptake of particulate antigen [17,18,45]. Thus, translocation takes place independent of bacteria-induced internalization  [16]. In contrast, the absence of M cells in the neonate mouse intestine shifts the major entry pathway to enterocyte invasion. It thereby pronounces the requirement for SPI1 for mucosal translocation. This may also explain the recent finding that Yersinia enterocolitica translocation is highly reduced in neonate mice [46]. Mucosal translocation of enteropathogenic Yersinia is also mostly mediated through M cell transport [47]. However, we cannot formally exclude that additional host factors also contribute to the enhanced requirement for a functional SPI1 system in the neonate host.
Strong innate immune receptor-mediated stimulation of the epithelium was noted upon infection in accordance with previous analyses of infected total mucosal tissue [48,49]. The present study provides a global gene expression analysis of isolated primary enterocytes after Salmonella infection in vivo. The restriction of innate immune stimulation to invasion-competent Salmonella is consistent with reports on the requirement of SPI1 for mucosal inflammation and clinical disease [25,[34][35]49]. Epithelial stimulation was mediated by recognition of Salmonella lipopolysaccharide (LPS) through Tlr4 and signaling via MyD88 consistent with previous reports on the involvement of Tlr4 in the adult host defense against Salmonella [50,51]. Similarly, deficiency in Tlr9 and the processing molecule Unc93B1 significantly reduced the epithelial cell response to Salmonella infection. The strong effect observed in 3d UNC93B1 mutant mice indicates the possible involvement of additional UNC93B1 dependent innate immune receptors such as Tlr3, 7 or 13 [52]. Our results indicate the synergistic action of Tlr4 and Tlr9 in the epithelial response to Salmonella in accordance with previous reports [53]. A minor but significant role was also found for Tlr2, Tlr5 and Nod2 previously implicated in the antimicrobial response of the adult host to Salmonella [54]. The synergistic stimulation of Tlr2, 4 and 9 was shown to be required for endosomal acidification and SPI2 effector protein translocation in bone marrow-derived macrophages [53]. A similar scenario may apply to intestinal epithelial cells. In accordance, lack of Tlr4 alone did not significantly influence enterocyte invasion or intraepithelial proliferation in vivo. Further studies are required to investigate the role of cooperative immune signaling by different receptors for the process of enterocyte infection and microcolony formation.
Colonization of the neonate intestine occurred largely independent of the invasion-dependent innate immune stimulation consistent with a low degree of colonization resistance in the neonate host. Reduced organ counts in the total small intestine after low dose infection of SPI1 mutant Salmonella might result from the lack of proliferating intraepithelial Salmonella. Alternatively, invasion-induced immune stimulation might promote metabolic changes that favor pathogen colonization of the neonate intestine as recently described in adult animals [55]. In accordance, a number of metabolic genes were significantly influenced during bacterial challenge supporting the idea of a heavily altered intestinal host metabolism during neonatal infection [56]. In addition, several innate immune response genes such as the chemokines, serum amyloid proteins, iNOS, calprotectin, mucins and antimicrobial lectins were strongly upregulated in infected epithelial cells in accordance with previous reports [57]. As a consequence, large numbers of polymorphonuclear cells (PMNs) but also macrophages were recruited to the site of infection. The cellular immune response at this time point, however, did not lead to significant tissue alteration. The antiapoptotic effect of NF-kB stimulation might have prevented the induction of epithelial apoptosis previously reported in both Salmonella-infected and non-infected enterocytes [44].
In conclusion, we analyzed age-dependent differences in the interaction of Salmonella with the host epithelium. We identify host and bacterial factors responsible for the infection process in the neonate host and characterize the epithelial innate immune response. Our results demonstrate enterocyte invasion and intracellular proliferation of Salmonella in differentiated and polarized epithelial cells and established a new animal model amenable to genetic manipulation to study the enterocytepathogen interaction in vivo. However, neonate mice exhibit an immature mucosal immune system, which might contribute to the observed phenotype and requires further characterization. Also, limitations such as the small animal size and lack of a suitable anesthesia required for intravital microscopy as well as the possible exchange of bacteria between newborn mice and the dam exist. Nevertheless, we believe that the described model opens new avenues of research to unravel the functional role of individual effector proteins in the Salmonella-enterocyte interaction and to better understand the cellular and immunological events of the Salmonella pathogenesis.

Bacterial strains
The isogenic strain MvP818 harboring a deletion of invC has been described before [58]. For complementation of the invC mutation, plasmids p3545 was generated as follows: The promoter of invF was amplified from S. Typhimurium genomic DNA using PinvF-For-EcoRI (CCGGAATTCTCCATCCAG ATGACAA-TATC) and PinvF-Rev-SmaI (ATATCTAGATCCATCCA-GATG ACAATATCTG). The resulting product was digested by EcoRI and SmaI and subcloned in pWSK29 to obtain p3537. invC was amplified using invC-For-SmaI (atacccgggtttagtcg gtcgctaatgag) and invC-Rev-XbaI (GTATCTAGATTAAT TCTGGTCAGCGA ATGC), the resulting fragment was subcloned as SmaI/XbaI fragment in p3537 to generate p3545. Correct clones were confirmed by DNA sequencing and functional complementation of the invasion defect of MvP818 was verified in non-phagocytic intestinal epithelial m-IC cl2 cells (Fig. S5A-C). S. Typhimurium ATCC 14028 and SPI-1 mutant S. Typhimurium MvP813 DinvC (Kana R ) carrying a green fluorescent protein (GFP) expression plasmid (pGFP, Amp R , kindly provided by Brendan Cormack, Stanford, USA) and S. Typhimurium ATCC 14028 carrying a mCherry expression plasmid (pFPV-mCherry, Amp R , kindly provided by Leigh Knodler, NIH, Hamilton, USA) were used for in vivo infection experiments. Maintenance of the plasmid pGFP in S. Typhimurium under in vivo conditions was confirmed by comparative plating of spleen and liver tissue simultaneously on selective (100 mg/mL ampicillin) and nonselective LB agar plates (Fig. S6C-D). For oral infection with two different Salmonella strains, a 1:1 mixture of 10 2 pGFP and 10 2 pFPV-mCherry carrying bacteria were administered orally in 2 ml PBS. Life imaging (IVIS) was performed employing an isogenic S. Typhimurium strain with a chromosomal insertion of the luxoperon from Photorhabdus luminescens under the control of the constitutive ß-lactamase promoter.
Salmonella enterica subsp. enterica serovar Typhimurium ATCC14028 (NCTC12023) WT and isogenic mutant strains were cultured in Luria Bertani (LB) broth overnight at 37uC, diluted 1:10 and incubated at 37uC until reaching the logarithmic phase (OD 600 approximately 0.5). Bacteria were washed and adjusted to OD 600 0.55-0.60 containing approximately 1.5-2.0610 8 CFU/mL and diluted to obtain the appropriate infection dose. All experiments with neonatal mice were performed using 1day-old C57BL/6 animals. Neonates were infected orally with the indicated number of S. Typhimurium in a volume of 1 ml PBS. The administered inoculum was confirmed by serial dilution and plating. In vivo imaging was performed using the IVIS200 imaging system (PerkinElmer). For image analysis and visualization the LivingImage 4.3.1 software was used. Oral infection of 6 week-old adult female mice was performed as previously described [25]. Bacterial counts were obtained after homogenization of spleen, liver and mesenteric lymph nodes by serial dilution and plating.

Ethics statement
All animal experiments were performed in compliance with the German animal protection law (TierSchG) and approved by the local animal welfare committee (approval 12/0697, 12/0693, 13/ 1097 and 14/1385 of the Niedersachsische Landesamt für Verbraucherschutz und Lebensmittelsicherheit Oldenburg, Germany).

Primary cell isolation and flow cytometry
Primary intestinal epithelial cells were isolated from small intestinal tissue after incubation in 30 mM EDTA PBS at 37uC for 10 min as described previously [61]. Cells were passed through a 100 mm nylon cell strainer (BD Falcon), washed with 10% FCS/ PBS, and harvested by centrifugation. The purity was verified by flow cytometry (Fig. S4). In order to determine the number of cellassociated versus intracellular bacteria, one fraction of epithelial cells was plated directly (total number of epithelium-associated bacteria); another fraction was incubated in 100 mg/mL gentamicin for 1 h prior to plating (number of gentamicin-protected, intracellular bacteria). For flow cytometry analysis of intracellular Salmonella, cells were fixed with 4% PFA for 15 min. on ice. The cells were washed, stained with APC-conjugated anti-EpCAM (anti-CD326 clone 48.8, diluted 1:500, from eBioscience) and finally resuspended in 5% FCS/PBS. Prior to analysis, cells were filtered through a 35 mm pore size BD Falcon Polystyrene tube (BD, cat. no. 352235). 300,000 events were acquired and GFP expression was detected using a FACS Calibur apparatus (BD). For purity confirmation of isolated primary enterocytes, cells were fixed and incubated on ice with APC-conjugated anti-EpCAM

Immunostaining
Immunostaining using a chicken anti-GFP (dilution 1:500), a mouse anti-mCherry (diluted 1:500, both from Abcam), a mouse anti Salmonella O4-antigen (diluted 1:500, from Abcam), a rat anti-Lamp1 (diluted 1:500, from the Developmental Studies Hybridoma Bank (DSHB), University of Iowa, US), a rabbit anti-active Caspase 3 (diluted 1:200, from Cell Signaling), a rabbit anti-Muc2 (diluted 1:100, generous gift from Gunnar Hansson, Göteborg, Sweden) and a mouse anti-E-cadherin (diluted 1:100, from BD Transduction Laboratories), was performed on 3 mm paraformaldehyde-fixed paraffin-embedded tissue sections in combination with the appropriate fluorophore conjugated secondary antibody (Jackson ImmunoResearch). Sections were deparaffinized in xylene and rehydrated in ethanol followed by antigen retrieval in 10 mM sodium citrate and blocking with PBS 10% serum 5% BSA. Staining of constitutively GFP expressing bacteria was performed to enhance the sensitivity of detection. In addition, this approach ensured bacterial detection also under low oxygen conditions, which might be present within the intestinal lumen and have been reported to alter GFP emission. The accuracy of the anti-GFP staining to detect plasmid-bearing S. Typhimurium was verified by staining of two consecutive sections cut from the same tissue block with anti-GFP and a biotinylated 1:100 diluted rat monoclonal anti-O4/O5 antibody generously provided by M. Kim (Kim Laboratories, Champaign, IL). Both staining methods revealed a very similar staining pattern suggesting that the anti-GFP staining method used was both sensitive and specific (Fig.  S6A-B). Staining with the rhodamine-conjugated lectin UEA-1 (Vector), a rat anti-GP2 (diluted 1:100, from MBL) or a goat anti-Ccl9 (diluted 1:100, from R&D) was performed on 8 mm sections obtained from freshly frozen OCT (Tissue-Tek) embedded tissue. Sections were dried and fixed using methanol at 220uC for 10 min. followed by rehydration in PBS for 15 min. Blocking in 0.2% BSA, 0.2% saponin, 10% serum in PBS was performed prior to immunostaining. Peyer's patches were visualized using a FITCconjugated anti-CD45 antibody (clone 30-F11, BioLegend). Hematoxylin and eosin staining of paraformaldehyde-fixed tissue sections was performed according to Mayer's protocol using reagents from Roth. Slides were mounted in Vectashield (Vector) supplemented with DAPI and pictures were taken with an Apo-Tome microscope connected to a digital camera (Zeiss).

Electron microscopy
For electron microscopical analysis of infected intestinal tissue, samples were fixed in 150 mM HEPES, pH 7.35, containing 4% formaldehyde and 0.1% glutaraldehyde at room temperature for 1 hour and then stored over night in fixative at 4uC. Samples were dehydrated in acetone and embedded in EPON. 60 nm sections were mounted onto formvar-coated copper grids, stained with 4% uranyl acetate and lead citrate as previously described by Reynolds and visualized in a Morgagni TEM (FEI), operated at 80 kV [61].

Gene expression analysis
Total RNA was extracted from isolated enterocytes using TRIzol (Ambion) and the RNA concentration was determined on a NanoDrop 1000 spectrophotometer (Thermo Scientific). Firststrand complementary DNA (cDNA) for quantitative RT-PCR was synthesized from 5 mg of RNA with Oligo-dT primers and RevertAid reverse transcriptase (Fermentas). Taqman technology based RT PCR was performed using absolute QPCR ROX mix (Thermo Scientific), sample cDNA and the Taqman probes hprt (Mm00446368_m1), Spi-B (Mm03048233-m1) Cxcl2 (Mm00436450_m1), Cxcl5 (Mm00436451_g1) and Reg3c (Mm01181783_g1) from Life Technologies. SYBR green-based real-time PCRs were performed as described in the Extended Experimental Procedures. Microarray analysis was performed using Whole Mouse Genome Oligo Microarray v2 (4644k) (Agilent Technologies) following the SC_AgilentV5.7 protocol provided by the manufacturer. Heat map analysis was performed using the MeV 4_5_1 software. Clusters of orthologous groups (COG) analysis was performed using the online bioinformatic tool PANTHER (http://www.pantherdb.org/). The expression array data are accessible through GEO Series accession numbers GSE51160. Reviewer can access the raw data files using the following link: http://www.ncbi.nlm.nih.gov/geo/query/acc. cgi?token = ehufucgejpwtjwf&acc = GSE51160.

In vitro cell culture
Intestinal epithelial m-IC cl2 cells were cultured for 6 days with medium change every other day to obtain a confluent cell monolayer as previously described [62]. m-IC cl2 cells were incubated with wildtype S. Typhimurium, the SPI1 mutant strain (DinvC) or the complemented SPI1 mutant strain (DinvC pinvC) at a multiplicity of infection of 1:10 for 1 h at 37uC, washed and gentamicin (Sigma Aldrich) was added at a concentration of 100 mg/mL for 1 h at 37uC. Cells were washed 3 times in PBS and lysed directly or maintained in culture in the presence of 20 mg/ mL gentamicin. At the indicated incubation period cells were washed with PBS and lysed in 500 ml 0.1% Triton X-100 for 2 min. at room temperature. Viable counts were determined by serial dilution and plating on LB agar plates. Flow cytometric analysis was performed on trypsinized m-IC cl2 cells following fixation using a FACS Calibur apparatus (BD).

Statistical analysis
Results for bacterial growth in organ tissues or quantitative RT-PCR show counts for individual animals plus the median. The One-way ANOVA Kruskal-Wallis test (with Dunn's posttest) and the Mann-Whitney test were employed for statistical analysis of bacterial growth in organ tissue at different time points, comparative analysis of quantitative RT PCR results or bacterial counts after infection with WT versus SPI1 mutant Salmonella respectively. The GraphPad Prism Software 5.00 was used for statistical evaluation. p values are indicated as follows: ***p, 0.001; **p,0.01, and *p,0.05. (C and D) 1-day-old mice (n = 6-7) were orally infected with 10 2 CFU S. Typhimurium carrying the GFP expression plasmid (pGFP) and the viable organ count in liver and spleen tissue was determined at (C) 2 and (D) 4 days p.i. by dilution and plating on LB agar plates without supplement (2) or supplemented with 100 mg/mL ampicillin (+). (TIF)