Intestinal restriction of Salmonella Typhimurium requires caspase-1 and caspase-11 epithelial intrinsic inflammasomes

We investigated the role of the inflammasome effector caspases-1 and -11 during Salmonella enterica serovar Typhimurium infection of murine intestinal epithelial cells (IECs). Salmonella burdens were significantly greater in the intestines of caspase-1/11 deficient (Casp1/11−/−), Casp1−/− and Casp11−/− mice, as compared to wildtype mice. To determine if this reflected IEC-intrinsic inflammasomes, enteroid monolayers were derived and infected with Salmonella. Casp11−/− and wildtype monolayers responded similarly, whereas Casp1−/− and Casp1/11−/− monolayers carried significantly increased intracellular burdens, concomitant with marked decreases in IEC shedding and death. Pretreatment with IFN-γ to mimic inflammation increased caspase-11 levels and IEC death, and reduced Salmonella burdens in Casp1−/− monolayers, while high intracellular burdens and limited cell shedding persisted in Casp1/11−/− monolayers. Thus caspase-1 regulates inflammasome responses in IECs at baseline, while proinflammatory activation of IECs reveals a compensatory role for caspase-11. These results demonstrate the importance of IEC-intrinsic canonical and non-canonical inflammasomes in host defense against Salmonella.


Introduction
Within the mammalian gastrointestinal (GI) tract, intestinal epithelial cells (IECs) provide the primary interface between the microbial-rich gut lumen and the underlying mucosal immune system. Here they play a central role in the coordination of mucosal homeostasis, tempering pro-inflammatory responses while remaining rapidly reactive to noxious stimuli such as enteric pathogens. One recently described mechanism by which IECs engage in immune defense is through the activation of cell-intrinsic inflammasomes that require inflammatory caspases, namely caspase-1 and caspase-11 in mice, or caspase-1 and caspase-4 in humans [1,2].
During the initial stages of an enteric infection, Salmonella enterica serovar Typhimurium (S. Typhimurium) migrates from the gut lumen towards the intestinal epithelium, subsequently invading IECs. The invasion and intracellular proliferation of this pathogen triggers the activation of IEC-intrinsic inflammasomes, resulting in the expulsion of infected IECs into the intestinal lumen. The more rapidly these cells are shed, the less time is available for intracellular Salmonella to proliferate and invade surrounding IECs or translocate into the underlying lamina propria. In 2014, Sellin and colleagues showed this process requires the Nod-like receptors (NLRs) Naip1-6 and Nlrc4 [2], which form an inflammasome platform that activates caspase-1. During the early stages of a S. Typhimurium infection (12 h post-infection (p.i.)), the IECs lining the ceca of Naip1-6 −/− and Nlrc4 −/− mice were found to be heavily infected, containing densely packed microcolonies of intracellular Salmonella (up to 20 bacteria per cell), which were only rarely observed in the IECs of wildtype mice [2]. Through bone marrow transplantation studies, as well as the use of Naip1-6 ΔIEC−/− mice, the authors demonstrated this microcolony phenotype was caused by the loss of Naip-Nlrc4 inflammasome activation in IECs. Notably, the protective role for Naip1-6 was extremely acute as these mice were comparable to wildtype mice in Salmonella colonization of the cecum or histopathology at later time points (36 h p.i.) In the study by Sellin et al., Salmonella loads in the mucosa of Casp1/11 −/− mice were between that of Naip1-6 −/− or Nlrc4 −/− mice and wildtype mice, whereas Casp11 −/− mice phenocopied wildtype mice at 18 h p.i. [2]. In an independent study, we demonstrated that a noncanonical inflammasome involving caspase-11 is activated at later time points during enteric S. Typhimurium infection in mice [1]. Specifically, Casp11 −/− mice carried higher Salmonella loads in the cecum and cecal lumen at 7 days p.i. and displayed an intracellular IEC microcolony phenotype similar to that described by Sellin et al. at 24 h p.i. [1]. Importantly, this phenotype was also observed in the gallbladder of Casp11 −/− mice, indicating that caspase-11 dependent control of epithelial cell shedding is not restricted to the intestine.
Thus the reports by Sellin et al. and Knodler et al. both detailed an important connection between epithelial-intrinsic inflammasome activation, cell shedding and intracellular bacterial burdens [1,2]. However, the individual contributions and potential functional overlap of caspase-1 and caspase-11 to host protection against Salmonella in the gut has yet to be determined, primarily because mice deficient only in caspase-1 were not available. Recently this has changed, as Casp1 −/− mice have been generated by a handful of groups [3][4][5]. To define the exact involvement of caspase-1 and caspase-11 in antimicrobial defenses within the gut, we directly compared S. Typhimurium colonization in Casp1 −/− , Casp11 −/− and Casp1/11 −/− mice as well as in enteroids. Our results demonstrate that caspase-1 primarily regulates inflammasome responses in IECs at baseline whereas caspase-11 plays a compensatory role upon extrinsic stimulation of inflammatory signaling pathways in IECs. Therefore, canonical and noncanonical IEC-intrinsic inflammasomes cooperate to provide an important innate immune defense against pathogen infections.

Inflammasome-deficient mice carry higher intestinal tissue and luminal Salmonella burdens
To define the exact contributions of caspase-1 and caspase-11 to enteric host defense, we infected C57BL/6 (wildtype; WT), Casp1 −/− , Casp11 −/− and double-deficient Casp1/11 −/− mice with S. Typhimurium via the orogastric route. The Casp1 −/− and Casp1/11 −/− mice proved highly susceptible to infection, carrying heavy cecal, colonic and luminal pathogen burdens at 18 h p.i. (Fig 1A). Although their cecal tissue burdens were not as high as those carried by the Casp1 −/− and Casp1/11 −/− mice, the Casp11 −/− mice also displayed significantly higher intestinal and luminal burdens than WT mice at 18 h p.i. ( � , P < 0.05, Fig 1A) and their intestinal burdens remained high at 72 h p.i. (Fig 1A). Interestingly, WT cecal burdens displayed a marked seven-fold decrease between 18 h and 72 h p.i. whereas only a minor decrease was observed in the Casp1 −/− and Casp1/11 −/− mice, while Casp11 −/− intestinal burdens remained comparable to those at 18 h p.i. This suggests the inflammatory caspase-deficient mice were unable to clear the infection from their tissues as efficiently as WT mice, a finding corroborated by their higher fecal shedding burdens (S1 Fig). Expression profiles of Casp1 and Casp11 in the cecal tissues of WT mice revealed that Casp11 transcripts increased over the course of infection, while Casp1 levels decreased (Fig 1B), which is consistent with other reports [2,6,7].

Inflammasome-deficient mice display increased numbers of infected IECs and higher intracellular Salmonella burdens
To investigate if the increased intestinal burdens recovered from the caspase-deficient mice indicated potential differences in tissue localization, we used immunofluorescence staining of infected cecal tissues (18 h p.i.) to identify S. Typhimurium via its O-antigen and epithelial cells via epithelial cadherin (E-cadherin). In all mouse backgrounds, the majority of the Salmonella were confined to the cecal lumen, however a small intraepithelial (and intracellular) subset was also observed (Fig 1C). Focusing on this intracellular subset, we noted that the cecal crypts of WT mice remained relatively sterile, with only an occasional infected cell identified per crypt (fewer than 10% of crypts carried infected IECs at 18 h p.i.) (Fig 1D). Of the infected IECs, they contained only 1-2 Salmonella per cell on average ( Fig 1E). Moreover, these infected WT IECs were largely confined to the tips of crypts or in the process of being actively shed from the epithelial surface (Fig 1C), similar to that described previously [1,2]. In contrast, all the inflammatory caspase-deficient mice showed increased numbers of Salmonella-infected IECs (Fig 1C; 22%, 38% and 42% of the cecal crypts of Casp11 −/− , Casp1 −/− and Casp1/11 −/− mice, respectively, showed infected IECs). Casp1 −/− and Casp1/11 −/− mice carried both the highest number of infected IECs/crypt ( Fig 1D) (mean of 2-3 infected IECs/crypt), but also the highest number of Salmonella per IEC, with numerous IECs containing microcolonies comprising �10 bacteria (Fig 1E).

Wildtype mice display high levels of IEC shedding largely localized to infected crypts
To address the connection between IEC shedding and S. Typhimurium invasion in vivo, 18 h p.i. cecal tissues were scored for epithelial damage. Whereas the ceca of WT mice demonstrated widespread signs of crypt and IEC deterioration, the cecal epithelium of

Casp1 −/− and Casp1/11 −/− enteroid-derived monolayers exhibit increased numbers of infected IECs and higher intracellular Salmonella burdens
To define whether the ability of inflammatory caspases to restrict S. Typhimurium infection in murine ceca reflected an IEC intrinsic role, or alternatively, confounding factors such as the intestinal microbiota or infiltrating immune cells that may alter IEC function, we generated cecal enteroids from uninfected mice. After generating 2D monolayers from these enteroids, they were infected with mCherry-S. Typhimurium (wildtype bacteria constitutively expressing the fluorescent protein, mCherry) and intracellular bacteria were enumerated using a gentamicin protection assay in combination with fluorescence microscopy. Casp1 −/− and Casp1/11 −/− monolayers proved highly susceptible to Salmonella infection and the majority of infected IECs remained intact within the monolayer (Fig 2A and 2B). Moreover, many of these infected cells contained large microcolonies of intracellular Salmonella, with some IECs containing over 100 bacteria at 10 h p.i. (Fig 2C). In contrast, WT and Casp11 −/− monolayers (Fig 2A-2C) showed stronger responses to infection, with significantly fewer infected adherent IECs ( Fig  2B) as well as relatively low numbers of intracellular Salmonella (Fig 2C). This result is consistent with a previous report that caspase-11 is not required for S. Typhimurium restriction within IECs in vivo [2]. With the exception of the Casp11 −/− monolayers, intracellular burdens in enteroids largely mirrored the in vivo findings that inflammatory caspase-deficient mice carried higher intracellular Salmonella levels than WT mice (Fig 2C). Bacterial colony forming

Wildtype and Casp11 −/− enteroid-derived monolayers display increased cell shedding and death
We have previously observed the shedding of Salmonella-infected IECs in vivo and in vitro [1,8]. Shedding IECs were also evident in our murine enteroid infection model, in agreement with a recent study of S. Typhimurium infection of human enteroids [9]. Compared to adherent IECs, shedding IECs presented a markedly different cell morphology; characterized by small, condensed DNA (leading to a comparatively stronger DAPI/nuclei signal), a 'ruffled' cytoskeletal actin signal and a slightly higher z-axis location in the monolayer, whereas adherent IECs displayed larger nuclei and clear cell-to-cell junction 'lattice-like' actin staining (Fig  2A and 2D) [8]. To quantify cell shedding, the size and intensity of DAPI signals were evaluated. IEC shedding was found to be significantly enhanced upon bacterial infection in WT and Casp11 −/− monolayers as compared to modest shedding in Casp1 −/− and Casp1/11 −/− cells ( Fig  2D and 2E). The proportion of infected, shed IECs compared to infected, adherent IECs was also comparatively higher in WT and Casp11 −/− monolayers (S4 Fig). Similarly, infected WT and Casp11 −/− monolayers exhibited higher levels of cytotoxicity compared to Casp1 −/− and Casp1/11 −/− monolayers, as measured by the release of the cytosolic enzyme, lactate dehydrogenase, into the growth media ( Fig 2F). These results support the concept that inflammatory caspases promote the expulsion of infected, dying IECs into the gut lumen [1,2,8].

Inflammatory caspase activity is present in Salmonella-infected shedding cells and not detectable in Casp1/11 −/− monolayers
We previously showed that Salmonella-infected human IECs undergoing extrusion have active inflammatory caspases [8]. To detect inflammatory caspase activity in infected murine enteroids we employed a cell permeable fluorescent caspase activity dye (660-YVAD-FMK FLICA), which covalently couples to active caspase-1 and/or caspase-11. By fluorescence microscopy, WT monolayers exhibited a strong fluorescence signal specifically in infected cells undergoing shedding, indicating active caspase-1/11 (Fig 3). This phenotype was also seen in Casp1 −/− and Casp11 −/− monolayers, despite overall lower numbers of shedding IECs in the Casp1 −/− monolayers. In contrast, active caspase-1/11 was not detected in the Casp1/11 −/− monolayers (S5 Fig), reinforcing the specificity of the FLICA probe for caspase-1 and -11. The FLICA signal in WT, Casp1 −/− and Casp11 −/− monolayers appeared diffuse throughout the cell cytoplasm, while small high-intensity puncta were also observed in WT and Casp11 −/− monolayers ( Fig   Fig 2. Epithelial intrinsic inflammasomes restrict the intracellular proliferation of Salmonella predominantly through caspase-1 induced cell shedding and death. Cecal enteroid monolayers derived from WT, Casp11 −/− , Casp1 −/− and Casp1/11 −/− mice were infected with mCherry-S. Typhimurium (MOI of 50) and bacterial colonization was assessed at 10 h p.i. by fluorescence microscopy. Representative fluorescence images depicting Salmonella (red), actin (green) and DNA (blue) (A). Original magnification ×400; scale bars 50μm. Arrows denote actively shedding or shed IECs and asterisks denote large foci of intracellular Salmonella. The severity of infection was also determined by the percentage of adherent infected IECs (B) and the number of intracellular Salmonella in each infected IEC (C).
Results are from at least 400 IECs from two independent experiments. Representative fluorescence images of Salmonella-induced cell shedding; Salmonella (red), actin (green) and DNA (blue) (D). Original magnification ×200; scale bars 50μm. Percentage of shed/ shedding IECs from the monolayer (from at least four blinded fields of view from two independent experiments) (E). IEC cytotoxicity as measured by lactate dehydrogenase activity released into the growth media at 10 h p.i. (F). Mean and SEM indicated from duplicate wells in two independent experiments. Statistical significance for 2B-C and 2E calculated using Mann-Whitney U-test and student t-test applied to 2F (n.s. p>0.05; � p<0.05; �� p<0.01; ��� p<0.001). https://doi.org/10.1371/journal.ppat.1008498.g002

Epithelial-intrinsic inflammasomes restrict Salmonella infection
3), appearing similar to the FLICA-positive signals described for Nlrc4-canonical inflammasome formation in macrophages [10]. These results suggest that caspase-1 activity is dominant in IECs, but a caspase-11 inflammasome is also present, albeit more evident in cells lacking caspase-1 function.

IFN-γ priming functionally differentiates enteroid monolayers derived from Casp1 −/− and Casp1/11 −/− mice in terms of Salmonella-induced cell shedding
We hypothesized that the limited role for caspase-11 in enteroid-derived monolayers could be due to a lack of extrinsic factors present during in vivo infection of mice, such as inflammatory cytokines and chemokines that directly modulate IEC expression of innate defense proteins [11]. Interferon (IFN)-γ is a potent cytokine released by immune cells which induces hundreds of genes promoting host defense [7,12]. It helps drive mucosal inflammation during the late

PLOS PATHOGENS
Epithelial-intrinsic inflammasomes restrict Salmonella infection stages of Salmonella-induced colitis and has been described as an early stage effector cytokine with high systemically circulating levels during the first day of oral infection [13,14]. When we analyzed cecal tissues collected from streptomycin-pretreated uninfected and S. Typhimurium-infected mice at 18 h and 72 h p.i. we noted that IFN-γ protein levels were significantly elevated in all infected genotypes at 18 h p.i., as compared to uninfected control tissues (S6 Fig). Interestingly, IFN-γ levels remained high in the infected ceca of WT mice at 72 h p.i., but significantly lower levels were produced by the inflammatory caspase-deficient mice, particularly the Casp1 −/− and Casp1/11 −/− mice. To test whether IFN-γ could alter inflammatory caspase production in IECs, pro-caspase-1 and pro-caspase-11 levels in naïve and IFN-γ treated WT cecal enteroids were compared by immunoblotting. High levels of caspase-1 were present irrespective of IFN-γ treatment, whereas caspase-11 levels increased upon IFN-γ treatment ( Fig 4A). This suggests that the activity of the non-canonical inflammasome might be potentiated in murine IECs as part of the host inflammatory response to infection.
Next, we assessed the effect of IFN-γ-priming on the enteroid monolayer response to infection. IFN-γ pretreatment alone had no overt effect on cell shedding under basal conditions for all genotypes (S7 Fig). However, IFN-γ pretreatment followed by S. Typhimurium infection significantly increased cell shedding in WT and Casp1 −/− enteroid monolayers (Fig 4B and  4C). In contrast, shedding of infected Casp11 −/− and Casp1/11 −/− monolayers was not affected by IFN-γ priming, with Casp11 −/− monolayers maintaining high levels of cell shedding that were not significantly different from its untreated, infected control, while Casp1/11 −/− monolayers exhibited little IEC shedding even after pretreatment with IFN-γ (< 2% of total IECs) (Fig 4B and 4C).

Discussion
Through the use of in vivo mouse infections as well as enteroid-derived monolayers, we have clarified the contributions of caspase-1 and caspase-11 to the IEC-intrinsic inflammasome and its ability to restrict enteric Salmonella infections. While previous work compared responses in Casp11 −/− and Casp1/11 −/− mice [1,2], here using Casp1 −/− and Casp11 −/− mice and enteroids, we unequivocally demonstrate that IECs utilize both inflammatory caspases to launch an intrinsic multilayered innate defense. In this study, we found that both caspase-1 and caspase-11 are required to effectively control an enteric S. Typhimurium infection. Caspase-1 dominates the antimicrobial response to Salmonella at early time points, while caspase-11 mediated defense plays a larger role later in the course of infection. This timeline agrees with our previous observations that Casp11 −/− mice carried significantly higher cecal burdens than WT mice at 7d p.i. with the S. Typhimurium ΔaroA strain [1].
Enteroids offer an attractive alternative to traditional cell culture, providing a physiologically relevant model system derived from the genotypic tissue of interest [15]. The use of enteroid-derived monolayers enables the study of bacterial-IEC interactions on various genetic backgrounds without the requirement for expression knockdowns prior to infection [16]. In

PLOS PATHOGENS
Epithelial-intrinsic inflammasomes restrict Salmonella infection murine cecal-derived enteroids, we showed that under baseline conditions IEC-intrinsic caspase-1 plays the major role in restricting intracellular Salmonella proliferation (Fig 2). In contrast, after pretreatment with IFN-γ to mimic the inflammatory environment that develops during an in vivo infection, high pathogen loads only developed in Casp1/11 −/− monolayers (Fig 4), indicating that both caspase-1 and caspase-11 exert potent antimicrobial responses within "inflamed" IECs.
Studies have previously shown that murine caspase-1 is highly expressed in naïve tissues whereas expression of caspase-11 requires pro-inflammatory induction, potentially in response to activation of the NF-κB, IRF3 or STAT pathways [2,6,7]. At the beginning of a S. Typhimurium infection in vivo, baseline tissue expression of Casp11 is low but is upregulated over the course of infection, whereas Casp1 transcript levels are high at baseline and decline slightly as the infection progresses [2]. This expression profile is corroborated by our finding that levels of pro-caspase-11, but not pro-caspase-1, increased in IECs in response to IFN-γ treatment. In a recent publication by Hausmann et al., it was reported that TNF-α treatment of murine small intestine-derived enteroids also increased Casp11 expression [17]. Together with our findings, this suggests that although caspase-1 is sufficient to protect naïve IECs, as the infection proceeds, host inflammatory responses upregulate caspase-11 expression in IECs, leading to increased caspase-11 activity; the combined efforts of these two caspases form a multilayered innate defense that controls intracellular Salmonella burdens and protects the host from pathogen attack.
Notably, our in vivo mouse studies show that both caspase-1 and caspase-11 are necessary for optimal host defense against S. Typhimurium (Fig 1). While our enteroid studies also show a clear role for both caspases in vitro, expression of either caspase was sufficient to promote inflammasome-mediated control over Salmonella expansion in murine IECs, including the expulsion of infected IECs from monolayers. These differences likely reflect a role for the inflammatory milieu in fine-tuning IEC inflammasome function as well as the contribution of inflammatory caspase-mediated activity from other cell types. Caspase-11 expression by IECs and macrophages needs to be induced by proinflammatory signaling before it can provide protection against S. Typhimurium [2,4,6,7]. Thus, the impaired IFN-γ expression displayed by infected inflammatory caspase-deficient mice may have been insufficient to induce caspase-11 expression to levels sufficient to protect Casp1 −/− mice.
A vital innate defense phenotype mediated by the IEC intrinsic inflammasomes is cell shedding accompanied by pyroptosis [8]. Correspondingly, we found that loss of inflammatory caspase signaling in murine enteroids led to decreased cell shedding and resulted in a heavily infected, but relatively intact, monolayers. Defective inflammasome signaling in these monolayers also led to large intracellular microcolonies of Salmonella, as compared to the relatively small intracellular burdens seen in WT monolayers. We have previously reported similar findings during S. Typhimurium infection of human IEC lines [1]. We propose that these

PLOS PATHOGENS
Epithelial-intrinsic inflammasomes restrict Salmonella infection increased bacterial burdens are due to the inability of Casp1/11 −/− monolayers to expel infected, pyroptotic IECs. However, in macrophages, there is also an inflammasome-mediated restriction of Salmonella intracellular growth that occurs prior to cell lysis, that is both caspase-1 and caspase-11 dependent, but independent of gasdermin D (Gsdmd) [18]. The identity(s) of the cytosolic antimicrobial agent(s) mediating this protective effect remains unknown, but based on our study design, we cannot rule out its contribution to overall inflammasome-dependent intracellular Salmonella restriction. Interestingly, a low level of cell shedding was observed at 10 h p.i. in Casp1/11 −/− monolayers suggesting there is a minor inflammatory caspase-independent IEC expulsion component in our enteroid infection model. Previous work found that treatment of Casp1 −/− ileal enteroids with FlaTox (a cytoplasmic delivery reagent of Naip5 ligand) induced IEC expulsion that was independent of plasma membrane disruption [3]. Through the use of murine knockout and inducible expression models, it was demonstrated that caspase-8, in the absence of caspase-1, can induce IEC cell expulsion in an Asc-dependent manner [3]. Thus, although the limited number of shed IECs we noted in the S. Typhimurium-infected Casp1/11 −/− monolayers did not exhibit caspase-1/11 activity, they may be extruded via this caspase-8 dependent mechanism.
We also observed a significant increase in the number of shedding IECs in WT and Casp1 −/ − enteroid-derived monolayers upon IFN-γ pretreatment. We hypothesize this increase is due to IFN-γ increasing caspase-11 expression, since no significant changes in cell shedding were demonstrated by Casp11 −/− or Casp1/11 −/− monolayers upon IFN-γ pretreatment. In macrophages, pro-caspase-11 is present at very low levels under resting conditions and can be induced through type I interferon (IFN-α and IFN-β) and type II interferon (IFN-γ) treatment [4,6,19,20]. Interestingly, complement signalling also amplifies caspase-11 activity in macrophages via upregulation of Casp11 transcript [21]. While transcriptional induction is one aspect of caspase-11 activity, IFN-γ signalling also potentiates non-canonical (caspase-11) inflammasome activation by inducing the expression of guanylate binding proteins (GBPs). These can control macrophage antibacterial immune responses in both inflammasome-dependent and -independent manners and are known to be expressed in IECs [22,23]. The GBPs potentiate caspase-11 based pyroptosis through the disruption of pathogen-containing vacuoles and since caspase-11 is an LPS sensor, the GBP-enhanced release of LPS into the cytosol augments non-canonical inflammasome activation [23][24][25][26].
Interestingly, the two inflammatory caspases are required for the rapid induction of IFN-γ in the gut during the early stages of S. Typhimurium infection (12 h p.i.) [27]. However, the impact of IFN-γ on pathogen burdens and mucosal inflammation is not evident until 48 h p.i. [13,14,28]. Acute sources of IFN-γ include innate lymphoid cells, NK cells, neutrophils and intestinal intraepithelial lymphocytes [13,14,[28][29][30][31]. Interestingly, Songhet et al. observed that control over S. Typhimurium burdens within IECs was IFN-γ dependent, and although bone marrow-derived IFN-γR signaling controlled systemic spread, stromal IFN-γR expression was required for this control in IECs [28]. Based on our findings, it appears that caspase-11 plays a larger role during the later stages of an acute murine infection, after its expression has been induced by proinflammatory signals such as IFN-γ. To study the contribution of IEC-intrinsic caspase-11 to overall host defense and the interplay of IFN-γ signaling, further infection studies using conditional knockout mice, cell culture, enteroids or less virulent strains of Salmonella (e.g. ΔaroA) as well as the evaluation of other cytokines induced during Salmonella infection will be required [32].
In conclusion, our study defines the importance of the IEC-intrinsic inflammasomes in early host restriction of a S. Typhimurium infection. We hypothesize that caspase-1 drives the initial inflammasome mediated antimicrobial response through Nlrc4 mediated detection of Naip ligands; Salmonella pathogenicity island-1 (SPI-1) type III secretion system (T3SS) needle components (PrgJ and PrgI) and flagella (FliC) [2,[33][34][35]. Upon inflammatory caspase activation, pyroptosis and cell extrusion is initiated, expelling the compromised IEC into the intestinal lumen. By sacrificing these infected cells, the epithelium maintains its sterility, disables the ability of Salmonella to expand their infective niche, meanwhile secreting cytokines and chemokines to recruit professional immune cells to further bolster host defenses. These proinflammatory pathways in turn induce Casp11 expression and newly primed IECs can now more effectively detect intracellular Salmonella through caspase-11 LPS recognition as well as identify those bacteria which have evaded Naip detection through downregulation of SPI-1 and/or fliC expression [36][37][38]. These actions highlight the complex and critical roles played by the intestinal epithelium in dealing with invasive bacterial pathogens, such as Salmonella enterica.

Ethics statement
All mouse experiments, maintenance and care were performed according to protocols approved by the University of British Columbia's Animal Care Committee (Permit Number: A15-0211) and in direct accordance with the Canadian Council on Animal Care (CCAC) guidelines. Animals were weighed and monitored daily to ensure animal welfare.

Tissue collection and bacterial counts
Mice were anesthetized with isoflurane and euthanized via cervical dislocation. For S. Typhimurium enumeration, the cecum, colon and combined cecal and colonic luminal contents were collected and homogenized separately in 1 mL of sterile PBS. Samples were serially diluted and plated onto streptomycin-supplemented LB agar plates and incubated at 37˚C overnight. Colonies were then enumerated and normalized to tissue weights. Tissue samples for histology and immunostaining were fixed in 10% neutral buffered formalin (Fischer Scientific) overnight then transferred to 70% ethanol. All fixed tissue was embedded in paraffin and cut into 5 μm sections.

Immunofluorescent staining of infected tissues
Immunofluorescent staining proceeded as outlined previously [1]. In brief, paraffin embedded tissues were deparaffinized by heating to 60˚C for 15 min, cleared with xylene, and rehydrated through an ethanol gradient to water. Antigen retrieval was performed in steam heated citrate buffer for 30 mins, before cooling to room temperature and washing with water. Tissues were treated in PBS, 0.1% Triton X-100 and 0.05% Tween 20 for 15 mins, then blocked with 5% donkey serum in PBS, 0.01% Triton X-100 and 0.05% Tween 20. Primary antibodies used were Salmonella O antisera Group B (Factors 1, 4, 12, 27) (1:1000, BD) and anti-E-cadherin (1:100; BD Biosciences). Tissues were then probed with Alexa Fluor 488-conjugated donkey anti-goat

Epithelial-intrinsic inflammasomes restrict Salmonella infection
IgG (1:1000; Life Technologies) and Alexa Fluor 568-conjugated donkey anti-rabbit IgG (1:2000; Life Technologies). Tissues were mounted using ProLong Gold Antifade reagent (Life Technologies) containing DAPI for DNA staining. Sections were viewed on a Zeiss AxioImager microscope and images taken using an AxioCam HRm camera operating through AxioVision software.

Intracellular Salmonella quantification in vivo
Cecal sections that had been immunostained for Salmonella, E-cadherin and DAPI, were blinded and manually studied at a magnification of ×400 to enumerate infected IECs per crypt, the number of intracellular Salmonella in IECs, and the presence of IEC(s) apically shedding from infected crypts (a score of 1 was given when shed IEC(s) were present, while 0 was awarded when no adjacent shedding IEC was present). For all enumerations, five separate cross sections from each mouse background were used, from two or more independent experiments and ten non-adjacent crypts for each cross section were selected. Epithelial integrity for the entire cross section were also evaluated as described by Barthel

Generation of cecal enteroids
Enteroids from murine ceca were isolated from each mouse background as previously described [15,16,42]. In brief, the ceca were excised, the tip and base removed, laterally opened to expose the apical surface, while luminal contents were removed and placed in Advanced DMEM/F12 (Gibco) supplemented with Pen Strep (100 U/ml, Gibco) and gentamicin (50 μg/ml, Gibco) on ice. The tissue was washed ten times in ice-cold Advanced DMEM/ F12 (Gibco) with extensive vortexing, then transferred to Cell Recovery Solution (Corning) and incubated on ice for 30 mins. Under sterile conditions, forceps were used to gently liberate cecal crypts from the underlying tissue and the remaining tissue was discarded. The solution containing the cecal crypts was then centrifuged and washed twice with base media (Advanced DMEM/F12, Gibco) supplemented with Pen Strep (100 U/ml, Gibco), GlutaMAX (1X, Gibco) and HEPES (0.01 M, Gibco)) then diluted 1:1 in Matrigel (Corning). This was pipetted into several 'domes' on a 24-well plate and incubated at 37˚C with 5% CO 2 . After the Matrigel solidified, growth media (base media supplemented with 1X condition media from L-WRN cells (CRL-3276, ATCC), N2 (Invitrogen), B27 (Invitrogen), N-acetylcystine (Sigma-Aldrich), nicotinamide (Sigma), mEGF (Invitrogen), A 83-01 (Tocris), SB 202190 (Sigma-Aldrich), and Y-27632 (Abmole)) was added to the well and incubated at 37˚C with 5% CO 2 [15]. L-WRN cells were cultured as previously described and condition media collected every 48h [15]. Media was changed every three days (growth media without Y-27632 supplementation) and the enteroids were passaged every five to seven days. For IFNγ-treated enteroids, growth media was supplemented with murine IFNγ (10 ng/mL; Peprotech) or corresponding volume of growth media for 16 h.

Enteroid monolayer seeding and Salmonella infection of monolayers
Monolayers were generated as outlined previously [16,43] with modifications. First, the growth media was removed, then four Matrigel 'domes' were pooled and disrupted through the addition of ice-cold Cell Recovery Solution and incubation on ice for 30 mins. Enteroids were then centrifuged and washed twice with base media, resuspended in Trypsin-EDTA (0.05%, Gibco) and incubated at 37˚C with 5% CO 2 for 10 mins. Enteroids were then

PLOS PATHOGENS
Epithelial-intrinsic inflammasomes restrict Salmonella infection mechanically disrupted into single cell suspensions with repeated pipetting through a p200 tip, and an equal volume of monolayer media (base media supplemented with N2 (Invitrogen), B27 (Invitrogen), and Y-27632 (Abmole) was added. Cells were centrifuged, then resuspended in monolayer media and added dropwise to Geltrex (Gibco) coated coverslips in 24-well plates. Monolayers were incubated at 37˚C with 5% CO 2 and media changed 24 h after seeding. Confluent monolayers were infected 72 h after seeding.
S. Typhimurium SL1344 WT glmS::Ptrc-mCherryST [1] was grown overnight in LB (5g/L NaCl) at 37˚C then diluted 1:300 into 10 mL of LB (Miller; 10g/L NaCl) and grown for 4 h at 37˚C with shaking [8]. The culture was then centrifuged, washed in PBS then diluted in infection media (monolayer media without Pen Strep). Salmonella was added to the monolayers at a MOI of 50:1 (bacteria:eukaryotic cell) and incubated at 37˚C with 5% CO 2 for 10 mins, then washed three times with PBS, and fresh infection media added for 20 mins. Monolayers were then treated with 50 μg/mL of gentamicin for 40 mins at 37˚C with 5% CO 2 . Media was discarded then fresh infection media supplemented with 10 μg/mL of gentamicin added and monolayers incubated at 37˚C with 5% CO 2 for a total infection period of 10 h. After infection, two 50 μl aliquots of media from each condition were transferred to a black bottom 96-well plate for LDH activity quantification through the CytoTox-ONE Homogeneous Membrane Integrity Assay (Promega) performed according to manufacturer's instructions. Monolayers were washed three times with PBS, fixed in 4% paraformaldehyde (PFA, Thermo Scientific) in the dark at RT for 30 mins, then used for immunostaining. For CFU counts, following the 10 h infection monolayers were lysed with 0.2% (w/v) sodium deoxycholate PBS solution for 10 mins with orbital shaking (400 rpm) at RT, lysates plated on LB and colonies enumerated the next day.

Cell shedding, infection and intracellular Salmonella quantification of enteroid monolayers
PFA-fixed enteroids on coverslips were treated in PBS, 0.1% Triton X-100 and 0.05% Tween 20 for 15 mins, then blocked with 2% donkey serum in PBS, 0.01% Triton X-100 and 0.05% Tween 20 overnight. Coverslips were then stained with Alexa Fluor 488-phalloidin (1:2000; Life Technologies) for 30 mins, washed and mounted using ProLong Gold Antifade reagent (Life Technologies) containing 4 0 ,6-diamidino-2-phenylindole (DAPI) for DNA staining. For determination of inflammatory caspase activity, the staining proceeded as outlined by Knodler et al. [8]. One hour prior to the end of infection, 660-YVAD-FMK (Immunochemistry Technologies) was diluted 1:30 into infection media and incubated at 37˚C with 5% CO 2 for 1 h, and further prepared according to manufacturer's instructions, before incubation with Alexa Fluor 488-phalloidin and mounted onto glass slides using ProLong Gold Antifade reagent. Sections were viewed on a Zeiss AxioImager microscope and images taken using an AxioCam HRm camera operating through AxioVision software.
Fixed enteroid monolayers were blinded and images at a magnification of ×200 (shedding) or ×400 (infection/intracellular Salmonella) were obtained then evaluated using ImageJ (version 1.52i). Shed IECs were defined as high intensity DAPI signals (signals present after gating minimal threshold >200) while intact IECs were defined as lower intensity DAPI signals (signals present after gating minimal threshold >30). Shed and total IECs were enumerated through ImageJ 'Analyzed Particles' (>125inch 2 pixel units; 0.10-1.00 circularity). Infected IECs and intracellular Salmonella were manually enumerated by eye. For all quantifications at least four images per condition were evaluated from two or more independent experiments.

Statistical analysis
All results presented in this study are expressed as the mean values ± standard errors (SEM). Mann-Whitney U-test, student t-test and one-way ANOVA were performed using GraphPad Prism software, version 7.02 for Windows. A p-value of 0.05 or less was considered significant, with asterisks denoting significance in figures.