Minimal SPI1-T3SS effector requirement for Salmonella enterocyte invasion and intracellular proliferation in vivo

Effector molecules translocated by the Salmonella pathogenicity island (SPI)1-encoded type 3 secretion system (T3SS) critically contribute to the pathogenesis of human Salmonella infection. They facilitate internalization by non-phagocytic enterocytes rendering the intestinal epithelium an entry site for infection. Their function in vivo has remained ill-defined due to the lack of a suitable animal model that allows visualization of intraepithelial Salmonella. Here, we took advantage of our novel neonatal mouse model and analyzed various bacterial mutants and reporter strains as well as gene deficient mice. Our results demonstrate the critical but redundant role of SopE2 and SipA for enterocyte invasion, prerequisite for transcriptional stimulation and mucosal translocation in vivo. In contrast, the generation of a replicative intraepithelial endosomal compartment required the cooperative action of SipA and SopE2 or SipA and SopB but was independent of SopA or host MyD88 signaling. Intraepithelial growth had no critical influence on systemic spread. Our results define the role of SPI1-T3SS effector molecules during enterocyte invasion and intraepithelial proliferation in vivo providing novel insight in the early course of Salmonella infection.

Introduction Non-typhoidal Salmonella like Salmonella enterica subsp. enterica sv. Typhimurium (S. Typhimurium) represent a major causative agent of gastroenteritis in humans worldwide [1,2]. Infection usually occurs through the ingestion of contaminated food or drinking water. Salmonella colonizes the distal small intestine, penetrates the intestinal epithelium and induces a strong inflammatory tissue response provoking the main clinical symptoms such as abdominal pain and diarrhea. In the healthy human host, non-typhoidal Salmonella remain restricted to the intestinal tissue. In contrast, spread to systemic organs associated with high mortality is observed in immunosuppressed individuals and newborns. This renders Salmonella one of the most important causative agents of neonatal sepsis and meningitis in parts of Asia and sub-Saharan Africa [3,4] Salmonella employs a multitude of virulence factors to overcome the mucosal barrier and evade the cellular and humoral host defence [5]. Effector molecules secreted by the Salmonella pathogenicity island (SPI)1-type 3 secretion system (T3SS) act during the early phase of infection and enable Salmonella to penetrate the intact intestinal epithelial barrier and reach the subepithelial tissue [6,7]. SPI1-T3SS effector molecules such as SipA, SopA, SopB, and SopE 2 intimately interact with host cell processes and manipulate cellular functions such as F-actin dynamics, signal transduction, chemokine secretion, fluid homeostasis, membrane trafficking and tight junction formation [6,[8][9][10][11][12][13][14][15][16][17][18][19][20][21][22][23][24]. Thereby, they facilitate enterocyte invasion followed by intraepithelial proliferation, histological hallmark of Salmonella pathogenesis [7,[25][26][27][28][29]]. An intact SPI1-T3SS has been strongly associated with mucosal inflammation and diarrhea and thus the clinical symptoms of Salmonella gastrointestinal infection in different models [30][31][32][33][34][35][36]. The presence in all Salmonella subspecies and clinical isolates indicates a critical and nonredundant function of SPI1 also during human infection [37,38].
The functional relevance of the SPI1-T3SS for tissue infiltration, mucosal inflammation and enhanced fluid secretion in vivo has first been characterized using the bovine ileal loop or oral calf infection model [13,[32][33][34][39][40][41][42]. Infection of calves mimics the disease in humans, characterized by small intestinal mucosal inflammation with chemokine secretion and leukocyte infiltration as well as enhanced fluid secretion [43][44][45]. Epithelial invasion has also been observed in guinea pig, swine and rabbit intestinal tissue [25-27, 46, 47]. In the most widely used animal model of adult mice, however, oral administration of Salmonella does not lead to detectable epithelial invasion and mucosal inflammation but causes a typhoid fever-like systemic infection following a largely SPI1-independent M cell-mediated mucosal translocation [48][49][50]. Streptomycin administration prior to infection facilitates bacterial expansion and leads to mucosal inflammation [35]. Salmonella-induced tissue pathology, however, is largely restricted to the caecum and colon and SPI1-dependent epithelial invasion is observed at low frequency due to rapid enterocyte exfoliation [28,29,35,51,52]. Although enterocyte invasion and intracellular proliferation has been observed in epithelial cells of the caecum of S. Typhimurium infected adult animals and mechanisms of host defence have been investigated [29,53], the role of individual SPI1-T3SS mediated effector molecules during the early steps of Salmonella infection in vivo has remained ill-defined.
We have recently introduced a novel oral Salmonella infection model using neonate mice [50]. In this model, Salmonella invades the neonatal small intestinal epithelium, forms intraepithelial microcolonies and spreads to systemic organs in a strongly SPI1-T3SS dependent manner.
Here, we used this model and employed wild type Salmonella, sopABE 2 sipA quadruple mutants complemented with individual effector molecules, the respective triple mutants, sopE 2 sipA, sopAE 2 , and sopBE 2 double mutants as well as sipA, sopB, and sopE 2 single mutants in combination with wild type and myeloid differentiation primary response gene 88 (MyD88) -/mice to investigate the contribution of individual SPI1-T3SS effector molecules and host factors during epithelial cell invasion and intraepithelial microcolony formation in vivo.

Comparative analysis of individual SPI1-T3SS effector molecules by complementation
We initially employed a previously established approach using a sopABE 2 sipA quadruple mutant S. Typhimurium strain complemented in trans by an expression plasmid encoding the respective SPI1 virulence factors SipA, SopA, SopB or SopE 2 [42]. Consistent with our previous results obtained using the SPI1-T3SS defective S. Typhimurium invC mutant, the sopABE 2-sipA quadruple mutant was unable to invade murine enterocytes, failed to induce a significant transcriptional epithelial response and remained restricted to the gut lumen (S1A-S1F Fig) [50]. Complementation with SipA alone significantly enhanced S. Typhimurium viable counts in isolated enterocytes, spleen, liver and mesenteric lymph node (MLN) tissue and induced transcriptional epithelial activation (S1A-S1F Fig . Importantly, albeit able to invade the epithelium, sipA, sopE 2 and sopE complemented ΔsopABE 2 sipA Salmonella were unable to proliferate and generate intraepithelial microcolonies (S1G Fig, S2G Fig).

Analysis of enterocyte invasion using triple mutants
To overcome the technical obstacles associated with trans-complementation such as plasmid loss, multiple effector gene copies or an impaired access of regulatory elements we next employed triple mutants leaving the fourth SPI1-T3SS effector gene under the native regulatory control. In accordance with our previous results, sopABE 2 and sopABsipA mutant Salmonella expressing solely a functional SipA and SopE 2 effector, respectively, displayed the ability to invade primary enterocytes and enhance Cxcl2 and Cxcl5 mRNA expression in the intestinal epithelium ( Fig 1A and 1C, S3C Fig). Both mutants also penetrated the mucosal barrier reaching liver, MLN, and spleen tissue at numbers comparable to wild type bacteria (Fig 1B, S3A  and S3B Fig). In contrast, sopBE 2 sipA and sopAE 2 sipA mutant bacteria expressing sopA or sopB alone failed to significantly invade and penetrate the epithelium, stimulate a transcriptional response and spread to systemic organs (Fig 1A-1C). Consistently, immunostaining identified intraepithelial Salmonella in the presence of SipA or SopE 2 but not SopA or SopB (Fig 1D). Again, despite the intraepithelial localization, SipA expressing ΔsopABE 2 or SopE 2 expressing ΔsopABsipA Samonella failed to proliferate intracellularly.
Following internalization, Salmonella manipulates maturation of the endosomal compartment. This recruits the lysosomal-associated membrane protein (LAMP)1 from Golgi-derived vesicles [58] and provides the environmental signals that coordinate the expression of SPI2-T3SS effector genes and the development of a replicative compartment [59,60]. LAMP1 recruitment to intraepithelial bacteria was observed following infection with wild type but not ΔsopABE 2 or ΔsopABsipA Salmonella (Fig 1E and 1F). Additionally, we used a previously described SPI2 reporter construct expressing gfp under the control of the ssaG promoter to analyze the induction of SPI2-encoded genes [52]. SPI2 reporter activity was detected in intraepithelial wild type but not SipA expressing ΔsopABE 2 or SopE 2 expressing ΔsopABsipA Salmonella (Fig 1G and 1H).
Together, these results identifed the critical but redundant role of SopE 2 and SipA for enterocyte invasion in vivo and highlighted the requirement of enterocyte invasion for transcriptional stimulation. Notably, invasion per se appeared not to be sufficient to generate an appropriate intracellular niche allowing intraepithelial bacterial proliferation. On the other hand, the formation of intraepithelial microcolonies did not significantly promote systemic spread. The ability and degree of individual SPI1-T3SS effector molecules to confer an enterocyte-invasive phenotype differed markedly between the situation in vivo and a classical in vitro cell culture-based invasion assay (S1H Fig, S2H Fig, S3D Fig) illustrating the need for a detailed investigation of the bacteria-epithelial cell interaction in vivo [61].

Infection-induced innate stimulation and intraepithelial proliferation
Host innate immune recognition through Toll-like receptor (TLR)2, 4 and 9 and signaling through the common adapter molecule MyD88 has previously been shown to provide the stimulatory signal for SPI2 effector gene expression and represent a prerequisite for the expression of SPI2-encoded effector genes, intracellular growth and microcolony formation in myeloid cells [60]. Interestingly, innate immune stimulation in Salmonella-infected neonate mice was also induced by TLR stimulation and mediated by the common adapter molecule MyD88 [50]. We therefore tested the requirement of intact MyD88-dependent signaling for intraepithelial Salmonella proliferation in vivo. As expected, MyD88-deficient mice exhibited significantly reduced epithelial Cxcl2 and Cxcl5 mRNA expression (S4A and S4B Fig). However, in contrast to the situation in myeloid cells, Salmonella maintained its ability to generate intraepithelial microcolonies with similar numbers of microcolonies per villus also in the absence of MyD88 signaling as illustrated by immunostaining and transmission electron microscopy (Fig  2A-2C). Also, LAMP1 was recruited to the majority of Salmonella-containing vacuole (SCV) in enterocytes and the SPI2 reporter was induced in the absence of MyD88-dependent signal transduction (Fig 2D-2G). The disease course was significantly accelerated rather than delayed, most probably as a result of an impaired MyD88-mediated antimicrobial host defence

Requirement of SPI1-T3SS effector molecules for intraepithelial microcolony formation
Although SipA or SopE 2 were sufficient to facilitate enterocyte invasion, Salmonella failed to induce intraepithelial microcolony formation. This suggested that intracellular proliferation required the contribution of SPI1-T3SS effector proteins beyond their invasion-promoting function. We first investigated the requirement of SipA and/or SopE 2 for intraepithelial proliferation. As expected, ΔsopE 2 sipA Salmonella exhibited severely impaired enterocyte invasion, transcriptional stimulation and organ spread (Fig 3A-3C, S5A-S5C Fig). In contrast, sipA or sopE 2 mutant bacteria displayed only a minor phenotype with intact enterocyte invasion, transcriptional stimulation and spread to systemic organs (Fig 3D-3F, S5D-S5F Fig). Strikingly, however, sopE 2 mutant bacteria similar to wild type Salmonella generated LAMP1-positive intraepithelial microcolonies, whereas ΔsipA Salmonella failed to do so ( Fig 3G, 3H and 3I). In contrast, both ΔsopE 2 and ΔsipA Salmonella exhibited detectable intraepithelial SPI2 reporter activity (Fig 3J and 3K). Consistent with intracellular proliferation of ΔsopE 2 Salmonella, the presence of SipA as compared to SopE 2 resulted in significantly higher intraepithelial viable counts ( Fig 3D). The enhanced number of intraepithelial bacteria was in turn associated with increased Cxcl2 mRNA expression as well as augmented spread to spleen, MLN and liver tissue ( Interestingly, sopBE 2 mutant Salmonella failed to generate microcolonies, whereas the majority of sopAE 2 mutant-infected epithelial cells exhibited intraepithelial growth (Fig 4D and 4E). Consistently, significantly lower intraepithelial bacterial numbers were found for ΔsopBE 2 but not ΔsopAE 2 Salmonella as compared to wild type bacteria ( Fig 4A). Also, sopAE 2 mutant but not sopBE 2 mutant Salmonella recruited LAMP1 (Fig 4F and 4G) and upregulated SPI2 gene expression (Fig 4H and 4I). These results highlighted the role of SipA and SopB for intraepithelial proliferation and excluded SopA as critical SPI1 component for intraepithelial microcolony formation.

The role of SopB during the interaction of Salmonella with the epithelium in vivo
To further evaluate SopB as a potential essential effector for intraepithelial proliferation, we next tested single sopB mutant Salmonella in vivo. Unexpectedly, sopB mutant Salmonella   (Fig 4), these results suggested that SopB by a hitherto unidentified mechanism counteracts the proapoptotic activity induced by SopE 2 [62,63].
Strikingly, immunostaining revealed that Salmonella deficient solely for sopB were still able to proliferate intraepithelially demonstrating that SopB is not essential for intraepithelial microcolony formation ( Fig 5G). Consistently, ΔsopB Salmonella generated LAMP1-positive intraepithelial endosomal compartments (Fig 5H and 5I) and showed induction of SPI2 reporter gene activity (Fig 5J and 5K). The fact that both sopE 2 and sopB single mutant Salmonella were able to grow intraepithelially (Figs 3G and 5G), whereas sopBE 2 double mutant Salmonella failed to form intraepithelial microcolonies (Fig 4D and 4E) suggests that SopE 2 and SopB exert a critical but redundant role during intraepithelial proliferation and microcolony formation.

The contribution of SipA to mucosal inflammation and intraepithelial proliferation
Neonate mice infected with the SipA-expressing sopABE 2 triple mutant Salmonella exhibited a more severe disease phenotype. Whereas mice infected with wild type Salmonella showed low but still significant weight gain during the first days after infection, mice infected with sopABE 2 mutant Salmonella failed to increase their body weight and exhibited an aggravated course of the disease (S8A and S8B Fig). We therefore infected 4-day-old mice in some experiments to facilitate the analysis. The accelerated disease course of ΔsopABE 2 Salmonella was associated with an increased tissue inflammation and epithelial barrier impairment illustrated by an enhanced translocation of labelled 4 kDa dextran at day 3 p.i. and a reduced length of the small intestine ( [64,65]. However, our results fail to confirm a significant importance of the caspase 3 cleavage site in SipA for this activity [12,66]. The proinflammatory activity of SipA was reported to occur independently of enterocyte invasion, following secretion and binding of SipA to the epithelial surface molecule p53-effector related to PMP-22 (PERP) [12,65,67,68]. Consistently, the aggravated disease phenotype was also observed after infection with Salmonella that lack SopA, SopB and SopE 2  To test whether the SipA-mediated immune stimulation contributes to the intraepithelial growth and microcolony formation, we next infected mice with a 1:1 mixture of SipA-sufficient wild type Salmonella and SipA-deficient ΔsopABsipA Salmonella. As shown in previous experiments (Fig 1A and 1D), ΔsopABsipA Salmonella were able to invade epithelial cells through the activity of SopE 2 but failed to proliferate intraepithelially. Following double infection, wild type Salmonella readily formed microcolonies in intestinal epithelial cells (Fig 6D, double stained appearing in yellow). In contrast, ΔsopABsipA Salmonella invaded but failed to proliferate intracellularly (Fig 6D, in red). Thus, the induction of a proinflammatory signal by SipA-expressing wild type Salmonella was not sufficient to induce a replicative endosomal environment in distant enterocytes and rescue the intraepithelial proliferation defect of SipAdeficient ΔsopABsipA Salmonella.
Next, we investigated the intrinsic activity of SipA in vivo. We therefore analyzed ΔsipA Salmonella complemented in trans with the gene for (i) wild type SipA (ΔsipA psipA) (ii) SipA with two point mutations at position 635 and 637 previously shown to impair its actin stabilization and reduce the invasion-promoting activity (ΔsipA psipA K635A E637W ) [67] (S8I Fig), (iii) SipA that lacks the caspase 3 cleavage site reported to be critical for the proinflammatory activity [66] (ΔsipA psipA D434A ) and (iv) SipA that lacks both, the actin-stabilizing and proinflammatory activity (ΔsipA psipA D434A K635A E637W ). Expression of these constructs was tested in the presence of SopE 2 to ensure enterocyte invasion and allow the comparative analysis of the different SipA variants in respect to intraepithelial growth and microcolony formation. As evaluation of the percentage of intraepithelial ΔsopBE 2 and ΔsopAE 2 S. Typhimurium associated with LAMP1 staining. All microcolonies from three tissue sections per infected neonate were analyzed (n = 3-4) at day 4 p.i.. Results represent the mean ± SD. (H) Co-immunostaining for ΔsopBE 2 and ΔsopAE 2 Salmonella (red) and the GFP-expressing SPI2 reporter (pM973; green) in small intestinal tissue sections at day 4 p.i.. Counterstaining with E-cadherin (white) and DAPI (blue). Bar, 5 μm. (I) Quantitative analysis of the percentage of intraepithelial S. Typhimurium expressing the SPI2 reporter. All microcolonies from three tissue sections per infected neonate were analyzed (n = 3-4) at day 4 p.i.. Results represent the mean ± SD. https://doi.org/10.1371/journal.ppat.1006925.g004 Salmonella-enterocyte interaction in vivo  (Fig 6G, S9C Fig). As observed before, ΔsipA Salmonella exhibited significantly reduced intraepithelial bacterial numbers and Cxcl2 mRNA expression, and failed to proliferate and form intraepithelial microcolonies (Fig 6E, 6G and 6H). Infection with ΔsipA Salmonella was also associated with a moderately reduced mortality (S9D Fig). In contrast, ΔsipA Salmonella complemented with psipA K635A E637W , psipA D434A , psipA D434A K635A E637W behaved indistinguishably from wild type Salmonella and readily formed LAMP1-positive intraepithelial microcolonies (Fig 6I and 6J). Together, these results demonstrate the critical importance of SipA. However, in the presence of SopE 2 , neither the actin-stabilizing activity nor the reported caspase 3-dependent proinflammatory activity were required for intraepithelial proliferation and microcolony formation.

Discussion
Salmonella expressing SopE, SopE 2 or SipA but not ΔsopE 2 sipA Salmonella were able to invade the epithelium indicating that the production of SopE, SopE 2 or SipA is necessary and sufficient to facilitate enterocyte invasion in vivo. SopE or SopE 2 activate the Rho GTPases Rac-1 and Cdc42 or only Cdc42, respectively, leading to actin assembly, membrane ruffling and bacterial internalization [22-24, 54, 56]. SipA with its C-terminal domain stabilizes actin filaments, promoting bacterial invasion in the absence of prominent membrane ruffling [9-12, 42, 69]. Their potent activity was illustrated by the finding that the presence of only one of the two effector molecules facilitated enterocyte invasion at levels indistinguishable from wild type bacteria. Notably, also the ubiquitin E3 ligase SopA was previously shown to contribute to invasion of polarized epithelial cells in vitro [15,16,42]. Also SopB was shown to promote membrane fission and bacterial invasion [18,19,70,71]. It was suggested that SopB supports SopE-mediated actin filament polymerization by recruiting ARNO (Cytohesin2) to the site of invasion [72]. Our in vivo results do not support a significant role for SopA or SopB in enterocyte entry. However, although unlikely, we cannot exclude that enterocyte invasion by strains expressing effectors other than sipA or sopE 2 occurred but remained undetected due to rapid enterocyte apoptosis or exfoliation as observed in adult mice [29].
Consistent with previous in vitro results, enterocyte invasion was not sufficient to allow proliferation and microcolony formation despite the presence of a fully functional SPI2 locus [73]. ΔsopB, ΔsopE 2 and ΔsopAE 2 but not ΔsipA Salmonella were able to generate intraepithelial microcolonies assigning SipA a critical non-redundant role for intracellular growth. ΔsipA Salmonella failed to recruit LAMP1 to the endosomal epithelial compartment, confirming a recent report on the contribution of the NH 2 -terminal domain of SipA (aa1-458) to endosomal maturation and intracellular replication [74]. SipA was recently also shown to promote intracellular proliferation via interaction with the actin nucleator family member Spire2 [69]. In our study, the SipA activity required for intraepithelial microcolony formation was independent of its actin stabilizing function or proinflammatory activity reported to depend on caspase 3 processing [66,67]. This might be explained by the recently described cooperative action of SipA with the SPI2 effector SifA to promote phagosome maturation and the generation of a replicative intraepithelial compartment [74]. Interestingly, intraepithelial ΔsipA Salmonella exhibited expression of the SPI2 reporter indicating that SPI2 effector expression can occur independently of SipA.
Whereas both ΔsopE 2 and ΔsopB Salmonella readily proliferated in enterocytes, sopBE 2 mutant Salmonella were unable to form intraepithelial microcolonies. Consistently, ΔsopBE 2 and ΔsopABE 2 (in contrast to ΔsopE 2 or ΔsopB) Salmonella failed to recruit LAMP1 to the endosomal compartment and to turn on SPI2 gene expression. Thus, SopB and SopE 2 contributed in a redundant fashion to the generation of a replicative endosomal compartment in enterocytes. Both, SopE 2 and SopB activate Rho GTPases and mitogen activated protein kinases (MAPK), which might contribute to endosomal modification [22][23][24]72]. Alternatively, SopB was described to directly or indirectly lead to the accumulation of phosphatidylinositol 3-phosphate on the outer leaflet of the SCV, altering the recruitment of host cell endocytic trafficking molecules and thereby preventing SCV-lysosomal fusion [17,18,75,76]. Although SopE 2 was shown to influence endosomal maturation and intracellular proliferation, it is currently unknown how it could compensate for the lack of SopB [77].
Enterocyte invasion was consistently associated with a transcriptional stimulation of the intestinal epithelium. In contrast, non-invasive ΔsopABE 2 sipA Salmonella failed to evoke a significant epithelial response despite the presence of high numbers of bacteria in the intestinal lumen. Also, the presence of a functional T3SS system in the absence of invasion in sopAE 2-sipA, sopBE 2 sipA or sopE 2 sipA mutant bacteria was unable to induce a significant epithelial transcriptional response. Whereas invasion led to a more than 100-fold increase in Cxcl2 and Cxcl5 mRNA expression, intraepithelial proliferation only moderately contributed to this response. Our results therefore suggest that the presence of intraepithelial Salmonella per se or, alternatively, downstream events such as stimulation of lamina propria immune cells as a consequence of penetration of the epithelial barrier drive epithelial transcriptional stimulation [67]. This is consistent with the previously reported requirement of a functional SPI1 system to evoke PMN transmigration and fluid secretion in calves and the bovine ligated loop model [32][33][34]. It is also in accordance with the presence of a functional SPI1 locus among all isolates from symptomatic human patients [37,38].
In addition to the stimulation of pattern recognition receptors the activation of host cell signaling pathways by Salmonella virulence factors has been described. For example, SopA was reported to contribute to tissue inflammation by targeting two host E3 ubiquitin ligases, TRIM56 and TRIM65 [13,15,78]. Also SopB was shown to indirectly stimulate Rho family GTPases and nuclear responses [56]. However, no significant influence of SopA or SopB on chemokine expression was observed in our study. Instead, our results suggest that SopA together with SopB and/or SopE 2 may contribute to balance the adverse effects of the inflammatory activity of SipA. Additionally, T3SS-mediated translocation of SipA and SopE 2 was reported to directly induce host cell activation [23,79]. SopE-mediated activation of the Rho family of small molecular weight GTPase Cdc42 and activation of p21-activated kinase was reported to induce mitogen-activated protein (MAP) kinase and NF-κB stimulation [23, 80,81]. However, invasive ΔsopABsipA and ΔsopABE 2 Salmonella (in the absence of SipA and SopE 2 , respectively) still provoked potent transcriptional enterocyte stimulation in vivo.
SipA (SipA 294-424 ) was additionally shown to induce PMN transmigration and tissue inflammation via PKCα-dependent secretion of the chemotactic eicosanoid hepoxillin A 3 (HXA 3 ) [12,65,82]. Notably, this pro-inflammatory activity was demonstrated to be independent of actin filament stabilization and enterocyte invasion and thus differed from the invasion-mediated immune stimulation discussed above [52,65,67,82]. Binding of SipA to the epithelial surface molecule p53-effector related to PMP-22 (PERP) was shown to inhibit the phospholipid glutathione peroxidase (GPX4) [12,65,67,68]. This effect was also observed in vivo with major influence on epithelial barrier integrity, immune cell infiltration and the outcome of the disease. Consistent with previous reports, immune stimulation was still observed in sipA K635A E637W ΔsopABE 2 Salmonella despite a reduced invasiveness, supporting the idea that SipA acts extracellularly to stimulate the mucosal immune system [65,68]. However, no influence of this SipA-mediated immune stimulation was observed on intraepithelial proliferation. The fact that ΔsopABE 2 Salmonella exhibited a more severe proinflammatory effect as compared to SipA-competent wild type bacteria suggests that SopA, SopB or SopE 2 might counteract this SipA effect in vivo.
Enhanced epithelial apoptosis and increased mortality were noted in the absence of SopB, consistent with a recent report that demonstrated protection from Nlrc4/ASC-mediated apoptosis by SopB in vitro [20,83]. SopB might additionally prevent apoptosis by other mechanisms [21,84]. Epithelial barrier damage after infection with ΔsopB Salmonella explains the higher bacterial load in the mesenteric lymph node. Intriguingly, enhanced disease progression following infection with ΔsopB Salmonella was absent using ΔsipAsopBE 2 and ΔsopBE 2 Salmonella. This suggests that the apoptosis-promoting effect is SopE 2 -mediated. Indeed, SopE 2 has long been described to activate the c-jun NH 2 -terminal kinase (JNK) [23, 80] and JNK signaling is known to drive ROS-induced caspase 3-dependent apoptosis [85]. Alternatively, SopE 2 was recently shown to activate caspase 1 [63]. Thus, our results identify a previously undescribed role for SopB: to protect from SopE 2 -mediated epithelial cell damage.
Despite the lack of intraepithelial microcolonies, ΔsopABsipA, ΔsopABE 2, ΔsopBE 2 or ΔsipA Salmonella efficiently penetrated the epithelial barrier and spread to systemic organs, reaching levels in spleen and liver tissue comparable to wild type Salmonella. This suggests that intraepithelial replication, although considered a hallmark of Salmonella pathogenesis, does not represent a prerequisite for efficient penetration of the epithelial barrier. This is consistent with the observation by Müller et al., who described SPI1-T3SS dependent penetration of the colon epithelium in the absence of detectable intraepithelial growth [28]. We even cannot exclude that two different bacterial populations simultaneously drive mucosal translocation and systemic spread. These results indicate that the functional relevance of intraepithelial replication and SCV formation has not been fully established and warrants further investigation.
Together, we validate the new neonatal infection model and present the first detailed in vivo analysis of the host and bacterial factors required for enterocyte invasion and intraepithelial microcolony formation, hallmark of Salmonella pathogenesis (Fig 7, S3 Table). We define the redundant and cooperative function of the SPI1-T3SS effectors SopA, SopB, SopE 2 and SipA for invasion of the intestinal epithelium and intraepithelial growth and characterize their influence on innate immune stimulation, mucosal translocation and spread to systemic organs. Our results thereby significantly extend our insight in the early steps of the microbialhost interplay in vivo and might reveal new targets for future preventive and therapeutic measures.

Bacterial strains and plasmids
All bacterial mutants and plasmids used in this study are listed in the S1 Table. Salmonella enterica subsp. enterica serovar Typhimurium ATCC14028 were used as wild type bacteria. The sopABE 2 sipA quadruple mutant S. Typhimurium carrying the low copy number pWSK29 vector encoding sopA, sopB, sipA, SopE or sopE 2 , or the empty pWSK29 vector as control, as well as the respective isogenic S. Typhimurium triple, double or single mutants were used to analyze individual SPI1-T3SS effector molecules. Deletions of genes encoding effector proteins were generated by Red-mediated recombination basically as described before [86]. Strains with multiple deletions of effector genes were generated by P22-mediated transduction of mutant alleles containing the aph or CAT cassette. If required to generate multiple mutations  [68,89]. This effect appears to be invasion-independent and strongly enhanced in the absence of SopA, SopB and SopE 2 suggesting that these effectors exert regulatory functions. (2) Among the studied effector molecules, expression of SipA (red dots), SopE 2 (green dots), or SopE (not shown here) alone is sufficient to facilitate enterocyte invasion. Intraepithelial Salmonella then reside within a LAMP1 negative endosomal compartment and fail to proliferate or express SPI2 encoded genes. Salmonella-enterocyte interaction in vivo by Red-mediated mutagenesis, antibiotic resistance genes were removed by FLP-mediated recombination between FRT sites. For the generation of a strain with point mutations in chromosomal sipA a two-step scarless mutagenesis approach according to Hoffmann et al. was applied [86]. A targeting cassette from pWRG717 was amplified using sipA633 In717 For and sipA939 In717 Rev (S2 Table). Salmonella WT harboring pWRG730 was induced for expression of redαβγ and transformed to kanamycin resistance. The resulting sipA mutant allele was transferred to other mutant strains using P22 transduction. The mutant allele sipA K635A E637W was amplified by PCR from plasmid p4758 target strains and the resulted DNA was used to transform MvP2511 haboring pWRG730. Resistance to I-SceI-mediated double strand breaks was used for selection of homologous recombination basically as described by Hoffmann et al. [86]. The point mutations in resulting strain MvP2520 were confirmed by DNA sequencing. Finally, the sopE 2 ::aph mutation was introduced by P22 transduction to yield MvP2521.
For complementation of sipA, sopA, sopB, sopE or sopE 2 deletions, the corresponding genes were amplified from S. enterica sv. Typhimurium genomic DNA introducing 3' HA tag sequences for detection of the encoded proteins by immunoblotting. Oligonucleotides for amplification are listed in S2 Table. PCR products were cloned in low copy number vector pWSK29 and E. coli DH5α was used to propagate plasmids p4041, p4042, p4043 and p4044, SopA::HA, SopB::HA, SopE::HA, and SopE 2 ::HA, respectively. Since sipA is the terminal gene of the sicAsipBCDA operon, the P sicA promoter was amplified and cloned upstream of sipA:: HA to generate p4040. Mutant strains were transformed with complementation plasmids listed in S1 Table. Synthesis and translocation of the effector proteins by S. Typhimurium was confirmed by immunolabelling of the HA tag. Furthermore, the plasmids gradually complemented the invasion defect of a multi-effector mutant strain. The function of SipA was additionally analyzed using ΔsipA Salmonella carrying the sipA gene with point mutations at specific functional positions. Vector p4758 carrying two point mutations in the sipA gene at the actin binding site (amino acid position 635 and 637), vector p4890 carrying a point mutation in the sipA gene at the caspase 3 motif (amino acid position 434), and vector p4892 carrying all three point mutations in the sipA gene were generated by site-directed mutagenesis. Site-directed mutagenesis was performed using the Q5 SDM kit according to the manufacturers' protocol (NEB). Multiple rounds of mutagenesis were performed if required for the generation of double or triple mutations using primers listed in the S2 Table. The plasmid with constitutive green fluorescent protein (GFP) expression (pGFP, AmpR) was kindly provided by Brendan Cormack, Stanford, USA. The reporter construct pM973 expressing gfp under the control of the SPI2 promoter pssaG (AmpR) kindly provided by Wolf D. Hardt, ETH Zürich, Switzerland was used to analyze the expression of SPI2 effector molecules by intraepithelial Salmonella.

In vitro infection experiments
Intestinal epithelial m-IC cl2 cells were cultured as previously described [87]. Bacteria were cultured in Luria Bertani (LB) broth at 37˚C in the presence of 100 μg/mL ampicillin, 50 μg/mL kanamycin, or 100 μg/mL carbenicillin. Overnight cultures were diluted 1:10 and incubated at 37˚C for approximately 80 min until reaching the logarithmic phase (OD 600 approximately 0.5). Bacteria were then washed three times in PBS and the OD 600 was adjusted to 0.55-0.60 corresponding to approximately 2.0×10 8 CFU bacteria per mL. The bacterial suspension was subsequently diluted to obtain the appropriate infection dose. Bacteria were added to the m-IC cl2 cells (2×10 5 cells per well) at a multiplicity of infection (MOI) of 1:10. The cell culture plate was centrifuged at 300xg for 5 min and incubated at 37˚C for 1 h. Cells were washed three times with PBS and cell culture medium supplemented with 100 μg/mL gentamicin (Sigma) was added. After incubation at 37˚C for 1 h, cells were washed again three times in PBS and lyzed for 2 min at room temperature in 500 μL 0.1% Triton X-100 (Roth) in aqua dest. Viable bacteria were quantified 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 (

In vivo infection experiments
Adult C57BL/6J wild type and B6.129P2(SJL)-MyD88 tm1.1Defr /J (MyD88 -/-, stock no. 009088) were obtained from the Jackson Laboratory (Bar Harbour, USA) and bred locally under SPF conditions. Bacteria were cultured as described above. 1-and 4-day-old animals were orally infected with 100 CFU of wild type S. Typhimurium or the indicated isogenic mutants in a volume of 1 μL PBS. At various time points post infection (p.i.), liver, mesenteric lymph mode (MLN) and spleen were collected and homogenized in sterile PBS. Viable counts were obtained by serial dilution and plating on LB agar plates supplemented with the appropriate antibiotics. Alternatively, small intestinal tissues were collected at the indicated time point p.i. and prepared for immunostaining or electron microscopy. For co-infection, 100 CFU of Salmonella wild type carrying pGFP and sopABsipA mutant Salmonella were administered orally to 1-day-old mice at a ratio of 1:1 in 2 μl.

Primary epithelial cell isolation
Primary small intestinal epithelial cells (IEC) from neonate mice were isolated as previously described [87]. Briefly, the small intestinal tissue was cut in small pieces and incubated in 30 mM EDTA/PBS at 37˚C for 10 min. IEC were detached from the underlying tissue by shaking. Cells were then filtered through a 100 μm nylon cell strainer (BD Falcon) and harvested by centrifugation at 300xg for 10 min. The cell pellet was resuspended in 10% FCS/PBS and harvested by centrifugation at 300xg for 10 min. To obtain the intraepithelial bacterial count, IEC were treated with 100 μg/mL gentamicin for 1 h at room temperature as previously described and subsequently lyzed and plated in serial dilutions on selective LB agar plates [50].

Gene expression analysis
Total RNA was extracted from isolated IEC using TRIzol (Invitrogen) and the RNA concentration was determined on a NanoDrop 1000 spectrophotometer (Thermo Scientific). Firststrand complementary DNA (cDNA) was synthesized from 5 μg total RNA using Oligo-dT primers, RevertAid reverse transcriptase and RiboLock RNase inhibitor (ThermoFisher Scientific). RT-PCR was performed using the Taqman technology with an absolute QPCR ROX mix (Thermo Scientific), sample cDNA and the Taqman probes Hprt (Mm00446968-m1), Cxcl2 (Mm00436450_m1) and Cxcl5 (Mm00436451_g1) (Life Technologies). Results were calculated by the Δ2-CT method. For data analysis, values were normalized to the hprt housekeeping gene and were presented as fold induction over age-matched controls.

FITC dextran-mediated analysis of epithelial barrier function
Infected mice were orally administered 2 μL of a 0.6 mg/μL 4kDa FITC dextran solution (TdB Consultancy) at the indicated time point p.i.. After four hours, serum was collected and the serum concentration of FITC dextran was measured by fluorometry using a SpectraMax i3 at an excitation of 492nm (9 nm bandwidth) and an emission of 518nm (15 nm bandwidth) using a serially diluted FITC dextran solution as standard.

Transmission electron microscopy
1-day-old MyD88 +/+ and MyD88 -/mice were infected with wild type S. Typhimurium and sacrificed at day 4 p.i.. Tissue samples were prepared for ultrastructural analysis as previously described [88]. After embedding, samples were post-fixed with 1% osmium tetroxide and contrasted with 2% uranyl acetate, both for 2 h. Samples were dehydrated with a graded ethanol series, followed by infiltration with epoxy resin and overnight heat polymerization. Thin, 70 nm sections were prepared using an ultramicrotome Ultracut UCT (Leica Microsystems) and contrasted with 0.2% lead citrate. Sections were analyzed with a JEM-1400 TEM microscope (Jeol) and images were recorded with TemCam-F216 camera using EM MENU software (both Tvips).

Statistical analysis
The One-way ANOVA Kruskal-Wallis test (with Dunn's posttest) and the Mann-Whitney test were employed for statistical analysis of bacterial counts in organ tissue and the comparative transcriptional analysis. Mortality was analyzed by log-rank (Mantel-Cox) test. The GraphPad Prism Software 5.0 was used for statistical evaluation. p values are indicated as follows: ÃÃÃ p<0.001; ÃÃ p<0.01, and Ã p<0.05. . Individual values and the mean from at least two independent experiments are shown (n = 4-7 animals per group). The data for uninfected control animals and Salmonella WT infected mice are identical to S3A-S3C Fig. (D) Survival following S. Typhimurium infection. 1-day-old C57BL/6 mice were orally infected with 100 CFU WT (solid line), ΔsipA (broken line), or ΔsipA complemented with sipA K635A E637W (broken line) S. Typhimurium. Animals that had to be euthanized due to a rise in the clinical score were included in the analysis (see Material and Methods). (TIF) S1