Stress‐induced host membrane remodeling protects from infection by non‐motile bacterial pathogens

Abstract While mucosal inflammation is a major source of stress during enteropathogen infection, it remains to be fully elucidated how the host benefits from this environment to clear the pathogen. Here, we show that host stress induced by different stimuli mimicking inflammatory conditions strongly reduces the binding of Shigella flexneri to epithelial cells. Mechanistically, stress activates acid sphingomyelinase leading to host membrane remodeling. Consequently, knockdown or pharmacological inhibition of the acid sphingomyelinase blunts the stress‐dependent inhibition of Shigella binding to host cells. Interestingly, stress caused by intracellular Shigella replication also results in remodeling of the host cell membrane, in vitro and in vivo, which precludes re‐infection by this and other non‐motile pathogens. In contrast, Salmonella Typhimurium overcomes the shortage of permissive entry sites by gathering effectively at the remaining platforms through its flagellar motility. Overall, our findings reveal host membrane remodeling as a novel stress‐responsive cell‐autonomous defense mechanism that protects epithelial cells from infection by non‐motile bacterial pathogens.


Introduction
Intestinal epithelial cells (IECs) constitute a physical and biochemical barrier between host and microorganisms. While not considered professional immune cells, IECs are crucial contributors to immune surveillance and the initial inflammatory responses against infection (Sansonetti, 2004;Peterson & Artis, 2014). For example, IECs secrete pro-inflammatory cytokines and chemokines in response to infection by certain enteroinvasive bacterial pathogens such as Shigella flexneri and Salmonella Typhimurium. This, in turn, leads to a massive infiltration of professional immune cells into the sites of inflammation, from which ensues a local increase in reactive oxygen species and a profound hypoxia (Colgan & Taylor, 2010;Zeitouni et al, 2016;Arena et al, 2017). Such change in environment not only constrains the pathogen but also dramatically affects the intestinal epithelium. Interestingly, while this inflammationrelated stress used to be considered as posing additional harm to the affected tissues, there is recent evidence to suggest that it makes crucial contributions to counteract infection processes at the cellular level (Chovatiya & Medzhitov, 2014).
Notwithstanding the diverse active mechanisms used by bacteria to mediate binding and invasion of host cells [e.g., pili, fimbriae, adhesins/invasins, T3SS (Pizarro-Cerda & Cossart, 2006;Stones & Krachler, 2016)], it has also become clear that efficient bacterial entry requires permissive sites in the host membrane. Membrane rafts, which are highly dynamic membrane domains enriched in sphingolipids and cholesterol that mediate the compartmentalization of signaling proteins and receptors (Lingwood & Simons, 2010;Sezgin et al, 2017), have been shown to be utilized by numerous bacterial pathogens (reviewed in Refs: Lafont & van der Goot, 2005;Bagam et al, 2017). For example, Shigella uses its IpaB effector protein to bind the host raft-associated CD44 transmembrane receptor (Lafont et al, 2002); entry of Listeria monocytogenes into host cells requires the localization of the host receptors E-cadherin and HGF-R/Met in specific lipid domains (Seveau et al, 2004). In addition to receptors, plasma membrane composition itself, specifically cholesterol and sphingolipid membrane content, impacts the binding and internalization of various bacterial pathogens, including Shigella and Salmonella species (Garner et al, 2002;Lafont et al, 2002;Misselwitz et al, 2011a;Santos et al, 2013).
Here, we investigated the impact of the general stress response of epithelial cells on the infection by S. flexneri and Salmonella Typhimurium. We found that induction of stress in epithelial cells by inflammatory cues and oxidative insults prevents the binding of Shigella, a non-motile pathogen, to host cells. We demonstrate that this inhibition results from extensive remodeling of the host plasma membrane following a stress-induced activation of the acid sphingomyelinase (ASM). By contrast, the related motile pathogen Salmonella can overcome this barrier, using flagellar motility to reach and accumulate at the remaining permissive entry sites. Moreover, we show that intracellular replication of Shigella activates ASM and subsequent membrane remodeling, thus suppressing re-infection by non-motile pathogens. Collectively, our findings demonstrate a role for the host stress response in protecting cells against Shigella infection and demonstrate the involvement of ASM and membrane remodeling in this process.

Host cell response to stress inhibits Shigella infection
To investigate whether host cell stress has a deleterious effect on the outcome of Shigella infection, we treated HeLa cells, an epithelial cell line commonly used to study Shigella infection, with sub-lethal concentrations of sodium arsenite ( Fig 1A). Arsenite is widely used to induce oxidative stress (Bernstam & Nriagu, 2000;Liu et al, 2001). Following arsenite removal, cells were extensively washed and then infected with Shigella; infection efficiency was monitored at early, intermediate, and late stages of infection (0.5, 2, and 6 hpi, respectively; Fig 1A) by: (i) fluorescence microscopy, (ii) colonyforming unit (cfu) assays, and (iii) qRT-PCR. Interestingly, pretreatment of cells with arsenite strongly reduced Shigella infection, at all time points tested (4.7-to 8.8-fold compared to control,cfu;Figs 1B and D,and EV1A and B). Validating these observations, Shigella infection was also inhibited by arsenite in all tested colon epithelial cells, namely HCT-8, .
Response to environmental stress in eukaryotic cells generally dampens bulk protein synthesis due to impaired mRNA translation initiation (Holcik & Sonenberg, 2005). However, inhibition of translation by puromycin and cycloheximide did not affect Shigella infection ( Fig EV1E-G), demonstrating that the effect of arsenite on Shigella infection is unrelated to translation shutdown.
To understand whether inhibition of Shigella infection by cellular stress is a broad phenomenon, we tested other stress inducers, namely anisomycin, hydrogen peroxide (H 2 O 2 ), hypoxia, or the inflammatory cytokine TNF-a. All the stress inducers have been widely used to mimic conditions encountered by cells during inflammation (arsenite, anisomycin) or are per se stimuli present during inflammation (hypoxia, TNF-a, and H 2 O 2 ). These stimuli converge in the production of reactive oxygen species (ROS), key signaling molecules during inflammation. Consistently, pre-treatment with the various stimuli strongly inhibited Shigella infection, already at early times post-infection (0.5 hpi;Figs 1E and F,and EV1H and I). The various compounds did not affect host cell viability, at the concentrations and incubation periods tested ( Fig EV1J). It should be noted that host cells were extensively washed prior to infection to remove any remaining stressors. Moreover, arsenite or anisomycin treatment did not impair Shigella growth (Fig EV1K), thus excluding direct effects on the bacteria.
Importantly, treatment with the antioxidant N-acetyl-L-cysteine (NAC) reverted the inhibitory effect of arsenite or anisomycin on Shigella infection (Figs 1G and H,and EV1L), further confirming that the effect of these stressors on Shigella infection is mediated by oxidative stress. Overall, these results demonstrate that the response of epithelial cells to oxidative stress limits Shigella infection.

Shigella binding to host cells is inhibited upon cellular stress
The evident inhibition of Shigella infection observed at 0.5 hpi strongly indicates that host cell stress affects the early steps of Shigella interaction with host cells. Accordingly, pre-treatment of cells with arsenite or anisomycin strongly inhibited Shigella binding to HeLa cells (ca. 10.0-and 2.0-fold compared to control, cfu, respectively; Fig Fig S1F). A possible effect of the stress inducers on actin cytoskeleton integrity and dynamics was excluded, since normal induction of actin-rich membrane ruffles by wild-type (WT) Shigella was observed in cells pre-treated with arsenite or anisomycin (white arrowheads in Fig 2A and Appendix Fig S1C, respectively). Accordingly, the efficiency of ruffle formation, i.e., the percentage of ruffles induced upon bacterial contact, was similar in cells treated with stressors and control cells (Appendix Fig S1G). To uncouple Shigella binding to host cells from subsequent steps of invasion, we performed parallel experiments with the Shigella DipaB mutant strain, which binds efficiently to host cells but is unable to invade (Menard et al, 1993). Treatment of cells with arsenite or anisomycin inhibited binding of Shigella DipaB, similar to the WT bacteria (Fig 2A-E, and Appendix Fig S1B, C, E and F). Binding of Shigella DicsA mutant strain to host cells was also inhibited by arsenite or anisomycin treatment (Appendix Fig  S1H and I), demonstrating that the effect of the stressors is independent of the role of IcsA in Shigella adhesion (Brotcke Zumsteg et al, 2014). Overall, these results show that host cell stress has a strong inhibitory effect on Shigella binding.
Analysis of Shigella cell-to-cell spreading in control and arsenite or anisomycin-treated cells, by quantifying the area of Shigella infection foci using fluorescence microscopy and automated image analysis (Sunkavalli et al, 2017), excluded an effect of stress in the actin-based spreading of Shigella to neighboring cells (Appendix Fig  S1J). In these experiments, infections were performed at different MOIs (MOI 10 for control, MOI 50 for anisomycin, and MOI 100 for arsenite), to achieve comparable levels of bacterial invasion.

ASM-dependent membrane remodeling upon stress inhibits Shigella binding
Considering that binding of Shigella to host cells is inhibited by stress and that this occurs in a relatively short timeframe (e.g., 15 min pre-treatment with TNF-a; Figs 1E and F, and EV1H and I), we reasoned that the decreased binding of Shigella could result from modifications of the cellular membrane. In line with this possibility, membrane composition, specifically the presence of sphingolipids and cholesterol at the cell surface, has been shown to be required  A Schematic representation of the experimental design. B, C Representative images of HeLa (B) or HCT-8 (C) cells infected with Shigella WT pre-treated or not with arsenite, analyzed at the indicated times post-infection. D Cfu quantification of intracellular bacteria in HeLa and HCT-8 cells pre-treated or not with arsenite and infected with Shigella. E, F Representative images (E) and cfu quantification (F) of intracellular bacteria in HeLa cells infected with Shigella WT after pre-treatment with TNF-a, H 2 O 2 , anisomycin, hypoxia, and corresponding controls, analyzed at 0.5 hpi. G, H Representative images (G) and cfu quantification (H) of intracellular Shigella in HeLa cells pre-treated with arsenite, anisomycin, stressors plus NAC, and corresponding controls.
Data information: Shigella infection was performed at MOI 10. Results are shown as mean AE s.e.m. of five independent experiments; *P < 0.05, **P < 0.01, ***P < 0.001 (t-test adjusted for multiple comparison for D-HeLa; paired t-test for D-HCT-8 and F; one-way ANOVA for H). Scale bars, 50 lm.
To investigate whether the activation of sphingomyelinases and consequently the disruption of sphingolipid-rich membrane domains could account for the inhibition of Shigella binding to host cells observed upon stress, we used specific inhibitors of these enzymes. Inhibition of NSM by GW4869 did not rescue the impairment of Shigella infection prompted by arsenite or anisomycin (Fig EV2A-F). However, treatment with the ASM inhibitor amitriptyline partially reverted the inhibitory effect of arsenite on Shigella infection (Fig 3A-C) and fully reverted that caused by anisomycin (Fig 3D-F). Corroborating these results, knockdown of ASM blunted the effect of arsenite on Shigella infection (Fig 3G-I). Of note, in the absence of stress, the inhibition of ASM activity by amitriptyline or ASM knockdown did not affect Shigella infection (Fig EV2G-J). The effect of amitriptyline in blunting ASM enzymatic activity and ceramide production upon arsenite or anisomycin treatment was confirmed (Fig EV2K-M). In addition, the decreased ASM activity, RNA, and protein levels upon knockdown were confirmed (Fig EV2N-P). Taken together, these results show that, in conditions of cellular stress, ASM activity is responsible for the inhibition of Shigella infection.
Acid sphingomyelinase is usually associated with the lysosomal compartment, but, upon activation, it is redistributed to the outer leaflet of the plasma membrane, where it hydrolyzes sphingomyelin giving rise to ceramide-rich platforms (Grassme et al, 2001). To validate whether ASM is relocalized as part of the cell response to stress, we used confocal microscopy followed by 3D reconstruction to visualize the accumulation of ASM and ceramide at the membrane. Indeed, treatment of cells with arsenite or anisomycin induced a marked accumulation of ASM at the cell surface, in both HeLa and HCT-8 cells ( Fig  S2D). The ASM staining specificity was confirmed in cells transfected with ASM siRNA, in which a considerable decrease in ASM foci was observed (Appendix Fig S2B). To further reinforce these observations, we quantified ceramide in live cells, by flow cytometry. In live cells, the ceramide antibody can only recognize extracellular/ Data information: Shigella infection was performed at MOI 10 for Shigella WT or at MOI 50 for the DipaB mutant strain; cells were incubated with the bacteria for 25 min. Results are shown as mean AE s.e.m. of 5 (C) or 6 (B, E) independent experiments; **P > 0.01, ***P < 0.001 (paired t-test). Scale bars, 25 lm. exposed epitopes and thus exclusively labels ceramide present at the cellular surface. This analysis confirmed a significant increase of ceramide in cells treated with arsenite and anisomycin, and a reversion of the increase upon amitriptyline treatment (Fig EV2L and M).
Overall, these results show that the binding of Shigella to host cells in conditions of cellular stress is impaired as a consequence of the membrane remodeling induced by ASM activation and translocation to the plasma membrane.

p38 MAPK is required for the inhibition of Shigella infection upon stress
Previous studies have shown a strong relation between sphingomyelinases and p38 mitogen-activated protein kinase (MAPK) activity. Oxidative stress leads to the activation of sphingomyelinases and ceramide production, which activates p38 MAPK pathway (Chen et al, 2008). In turn, p38 MAPK was shown to activate ASM in glial cells (Bianco et al, 2009). Thus, we hypothesized that the p38 MAPK pathway could be an important player in the activation of the sphingomyelinases and consequent inhibition of Shigella infection, in a context of cellular stress. Firstly, we confirmed that the various stressors (arsenite, anisomycin, TNF-a, and H 2 O 2 ) activate the p38 MAPK pathway, by Western blotting for the active phosphorylated form of p38 ( Fig 4A). Supporting our initial hypothesis, inhibition of p38 MAPK pathway by siRNA knockdown (Mapk14) or treatment with SB203580 reverted the inhibitory effect of arsenite (Figs 4B-D and EV3A) or anisomycin (Figs 4E and F, and EV3B and C) on Shigella infection. Importantly, we verified that inhibition of p38 MAPK pathway, by treatment with SB203580, inhibited the production of ceramide induced by arsenite and anisomycin (Figs 4G and H,and EV3D and E), demonstrating that p38 MAPK pathway is required for ASM activation and membrane remodeling upon stress. The effective inhibition of the p38 MAPK pathway by SB203580 upon arsenite or anisomycin treatment, and the knockdown of p38 were confirmed by Western blotting (Fig EV3F-H). p38 knockdown or SB203580 treatment had no effect on Shigella infection in the absence of stress (Fig EV3I and J). Taken together, these results demonstrate that the inhibition of Shigella infection during stress relies on the activation of the p38 MAPK pathway. 6 of 17 The EMBO Journal 37: e98529 | 2018 ª 2018 The Authors Bacterial motility overcomes the restriction posed by stressinduced host cell membrane remodeling Membrane composition has been shown to play critical roles for the adhesion and invasion of various bacterial pathogens in addition to Shigella, including the closely related pathogen Salmonella Typhimurium (Garner et al, 2002;Santos et al, 2013). We thus reasoned that the stress-induced membrane remodeling could likewise affect Salmonella infection. Surprisingly, however, Salmonella infection at 0.5 hpi was normal in cells pre-treated with arsenite or anisomycin ( Fig 5A-C), suggesting that Salmonella is able to cope with the remodeled host cell membrane. Interestingly, the pattern of infection was altered in cells treated with the stress inducers, showing fewer infected cells, but more bacteria per infected cell ( Fig 5A). This observation was further strengthened by the results of Salmonella binding assays, in which a significant increase in the number of bacteria per infected cell was observed in cells pre-exposed to arsenite or anisomycin, compared to control cells (Fig 5D and E, and Appendix Fig S3A). To confirm that this phenotype was related to bacterial binding, we used an invasion-deficient but binding proficient Salmonella mutant strain (Salmonella D4), which lacks four effector proteins (SopE, SopE2, SopA, and SipB) essential for invasion (Schlumberger et al, 2005;Lara-Tejero & Galan, 2009;Misselwitz et al, 2011b). Similar to WT Salmonella, in conditions of cellular stress the D4 mutant strain infected a lower number of cells than in the absence of stress, but also accumulated in specific sites, resulting in a higher number of bacteria per infected cell (Fig 5D and F, and Appendix Fig S3A). Moreover, a direct effect of stress on Salmonella intracellular replication was excluded, given that arsenite treatment post-invasion did not affect infection (Appendix Fig  S3B and C). These results indicate that, although the total bacterial load is similar, in cells subjected to stress there are fewer Salmonella-infected cells, though exhibiting a higher number of bacteria bound per cell. The patchy pattern of Salmonella binding to cells exposed to stress suggests that the bacteria are able to accumulate in the remaining permissive binding sites after stress-induced membrane remodeling has occurred. In contrast to Shigella, Salmonella is a motile pathogen, and therefore, a possible explanation for this phenomenon is that Salmonella can compensate for the lack of permissive sites by scanning the cellular surface through a process of near cellular surface swimming, enabled by its flagellar motility (Misselwitz et al, 2012). Most Salmonella serovars, including Typhimurium, have two flagellin genes, fliC and fljB, allowing antigen switches between two alternative forms of its flagellin filament protein (Andrewes, 1922). To test the relevance of bacterial motility in conditions of host cell stress, we used the Salmonella motility-deficient mutant strain DfliC, which lacks the fliC gene but retains expression of FljB. However, the frequency of flagellin switching is low [approximately 1/5,000 per cell per generation, biased toward the fljB OFF orientation (Gillen & Hughes, 1991)], and thus, the Salmonella DfliC mutant strain is essentially non-motile. Additionally, we used a completely non-motile Salmonella mutant (DflhC), due to an insertion of the MudJ transposon into the flagella master regulator flhC gene. Remarkably, arsenite and anisomycin pre-treatment inhibited very efficiently the binding of the non-motile Salmonella DfliC and DflhC mutants (Fig 5G and H,and Appendix Fig S3D and E). Complementation of the DfliC mutant, by ectopic expression of FliC from a native promoter, restored the ability of Salmonella to bind to cells subjected to stress, at levels comparable to that of control cells (Fig 5G and H, and Appendix Fig S3D and E), demonstrating causality between bacterial motility and the ability to overcome membrane remodeling. Based on these observations, we next addressed the effect of host stress on the infection by two additional pathogens, L. monocytogenes and Yersinia pseudotuberculosis, which repress flagella at 37°C and are thus non-motile when grown in these conditions (Peel et al, 1988;Kapatral & Minnich, 1995). In agreement with the results reported above for Shigella and Salmonella, arsenite treatment inhibited the infection by Listeria and Yersinia (Fig 5I-L), demonstrating that membrane remodeling generally impairs infection by non-motile bacterial pathogens.
Overall, these results show that motility is a strong determinant of binding of bacteria to cells exposed to stress. Under these conditions, while binding of non-motile bacteria to host cells is strongly compromised, motile bacteria are able to efficiently sample the host cellular surface and accumulate at the remaining permissive sites for efficient bacterial invasion. Shigella intracellular replication inhibits re-infection by nonmotile bacteria Given that Shigella intracellular replication causes stress, entailing oxidative stress, cytokine secretion, and MAPK activation (Pedron et al, 2003;Carneiro et al, 2009;Kasper et al, 2010), and that the host cell response to stress ultimately inhibits Shigella binding, we hypothesized that cells infected with Shigella could become refractory to re-infection. To test this possibility, we designed a reinfection assay in which mock-treated or cells infected with Shigella-mCherry for 3 h (primary infection) were re-infected with Shigella-GFP (secondary infection; Fig 6A). Interestingly, the secondary infection with Shigella-GFP was significantly decreased in cells previously infected with Shigella-mCherry, when compared to mock-treated cells (Figs 6B and C,and EV4A). To test whether the internalization of a replication-deficient Shigella mutant strain is sufficient to block re-infection, we performed parallel experiments using the Shigella-mCherry DipaB/Inv strain for the primary infection. The Shigella DipaB/Inv strain has a compromised T3SS secretion, though it is internalized due to the expression of the Yersinia invasin protein (Isberg et al, 1987). Once inside cells, this mutant is unable to escape the internalization vacuole and does not replicate (Suzuki et al, 2006). Interestingly, we observed that primary infection with this mutant strain did not compromise reinfection, which was comparable to that observed in mock-treated cells (Figs 6B and C, and EV4A). Taken together, these observations demonstrate that Shigella intracellular replication and/or presence of the bacteria in the host cytoplasm is required for inhibition of re-infection. In agreement with the results described above showing that host response to stress does not restrict infection by motile bacteria, infection with Salmonella WT-GFP (secondary infection) was not inhibited in cells previously infected with Shigella-mCherry (Figs 6D and E, and EV4B). Similar results were obtained in HCT-8 cells; specifically, we observed that primary infection with Shigella-mCherry inhibits re-infection by Shigella-GFP (Fig EV4E-G), but not by Salmonella-GFP (Fig EV4H-J). Of note, the primary infection with Shigella-mCherry WT or DipaB/Inv mutant strain was comparable in the different experimental conditions, both in HeLa and in HCT-8 cells (Fig EV4C, D, K and L). In agreement with the lack of membrane remodeling observed upon Salmonella infection (Appendix Fig S3F), the primary infection with Salmonella-mCherry did not affect re-infection by Shigella-GFP (Fig EV4M and N).
To evaluate whether the motility of Salmonella is related to its ability to circumvent the restriction to infection caused by Shigella intracellular replication, we performed re-infection assays using the Salmonella motility mutants described above. Infection with Salmonella DfliC and DflhC was inhibited in cells containing replicating Shigella-mCherry (Fig 6F and G), whereas infection with the DfliC/ pFliC complemented strain was not affected (Fig 6F and G). This clearly reinforces the notion that the motility of Salmonella is the decisive feature to overcome the inhibition posed by the cellular response to Shigella infection. Of note, we confirmed that the levels of the primary Shigella-mCherry infection were similar in the experiments with the different Salmonella strains, excluding any indirect effect on the primary infection ( Fig EV4O).
Collectively, these results demonstrate that Shigella intracellular replication in epithelial cells inhibits subsequent binding and reinfection by Shigella and, presumably, of other non-motile bacteria.

Shigella intracellular replication induces host cell membrane remodeling in vitro and in vivo
Next, we pondered whether Shigella infection could cause membrane remodeling, and thus explain the inhibition of re-infection. To address this question, we either performed the reinfection experiment in ASM knockdown cells or used amitriptyline to inhibit ASM activity in cells at 2 hpi with Shigella, proceeding with re-infection with Shigella-GFP after 1 h of amitriptyline treatment ( Fig 6H). Under these conditions, secondary infection by Shigella-GFP was not inhibited by the primary infection (Figs 6I-K and EV5A-C). Similarly, inhibition of the p38 MAPK pathway achieved by p38 knockdown or treatment with SB203580 (Figs 6L and M, and EV5D and E), as well as oxidative stress mitigation by NAC treatment (Fig EV5F and G), dampened the inhibitory effect of the primary infection. Of note, treatment with amitriptyline, SB203580, NAC, as well as ASM or p38 knockdown, had no effect on the primary infection by Shigella-mCherry (Fig EV5H-N). Overall, these results imply that ASM-mediated membrane remodeling, induced by oxidative stress-activated p38 MAPK pathway, is responsible for the inhibition observed in the re-infection experiments reported above.
To provide direct evidence that membrane remodeling occurs as a consequence of the host cell response to Shigella infection, we measured ASM activity and analyzed the accumulation of ASM and ceramide at the membrane of Shigella-infected cells. In agreement with an involvement of ASM in this process, we observed a clear accumulation of ASM in the membrane of Shigella WT-infected cells (Figs 7A and EV5O), alongside a clear increase of ceramide (Figs 7B and EV5P); this phenotype was not observed in Shigella DipaB/Inv (Fig EV5R and S), which is in perfect agreement with the observation that primary infection with this mutant strain does not compromise re-infection. Interestingly, both ASM and ceramide accumulated to a higher extent in bystander cells, compared to cells with intracellular bacteria (Fig 7A and B, and EV5O-Q). Enzymatic assays on membrane fractions from Shigella-infected cells (3 hpi) ▸ Figure 6. Re-infection by non-motile bacteria is inhibited by Shigella intracellular replication.

A
Schematic representation of the experimental design for re-infection experiments. Cells were infected with Shigella-mCherry (primary infection) or mock-treated. At 3 hpi, cells were re-infected with Shigella or Salmonella expressing GFP (secondary infection). Secondary infection was analyzed at 0.5 hpi (i.e., 1 h after addition of bacteria to cells).

B-E Representative images (B, D) and cfu quantification (C, E) of re-infection assays with Shigella WT (B, C) or Salmonella WT (D, E) of HeLa cells primarily infected with
Shigella WT or DipaB/Inv mutant strain, or mock-treated. Secondary infection is shown in bottom images. showed a modest but significant increase in the activity of ASM (Fig 7C), correlating with a clear increase in ceramide levels ( Fig 7D). Collectively, our results demonstrate that Shigella intracellular replication in epithelial cells leads to the activation of ASM, resulting in formation of ceramide domains, which impairs re-infection by Shigella.
Having shown that Shigella infection induces major changes in membrane composition in vitro, we wondered whether this process could be observed in vivo. For this purpose, we used a guinea pig model, a well-established animal model for the study of Shigella infection (Arena et al, 2015), and compared the ceramide accumulation in colon tissue samples from non-infected and Shigella-infected animals. In complete agreement with the in vitro data, we observed  Shigella-infected animals (4 hpi of challenge with Shigella WT) compared to non-infected controls (Fig 7E-I). Interestingly, the enhanced ceramide signal in infected samples was clearly more noticeable in bystander cells than in cells with intracellular bacteria. Overall, these results strongly corroborate the notion that Shigella intracellular replication, both in vitro and in an animal model of Shigella infection, leads to the activation of ASM and consequent membrane remodeling.

Discussion
Although intestinal epithelial cells are now considered crucial players in the defense against infection by bacterial pathogens, exactly how their general stress response helps to fend off potential bacterial invaders is incompletely understood. In this study, we reveal that epithelial cell stress generated by inflammatory cues and oxidative insults, namely arsenite, anisomycin, hypoxia, hydrogen peroxide, and cytokines, causes a strong inhibition of Shigella binding to host cells. Mechanistically, we determined that this inhibition results from an extensive remodeling of the host plasma membrane due to stress-induced activation of ASM that is enforced by p38 MAPK activation. This depletes permissive binding sites on the plasma membrane drastically and so hinders Shigella's attempts to bind and, ultimately, invade host cells (Fig 7J).
The relevance of ASM activity and ensuing membrane remodeling in the context of infection by diverse bacterial pathogens has been studied before. ASM is required for efficient infection by Neisseria gonorrhoeae and Neisseria meningitidis (Grassme et al, 1997;Simonis et al, 2014;Faulstich et al, 2015), whereas for other pathogens, ASM is involved in the host defense against infection. For example, ASM KO mice are much more susceptible to L. monocytogenes infection than wild-type animals (Utermohlen et al, 2003), a phenotype that was linked with lysosomal dysfunction in macrophages (Utermohlen et al, 2003Schramm et al, 2008). Based on the results described herein, we suggest that the higher susceptibility of ASM KO mice to Listeria infection can also be linked with the role of ASM in the remodeling of epithelial cell membrane to restrict adhesion and entry of non-motile pathogens. It is conceivable that the absence of ASM precludes membrane remodeling events that would normally occur, effectively increasing the number of permissive sites on the cell surface. In agreement with this hypothesis, the entry of Listeria into host cells was shown to require the localization of the host receptors E-cadherin and HGF-R/Met in specific membrane domains (Seveau et al, 2004). Of note, Listeria should be considered a non-motile pathogen in the human host due to the temperature-dependent expression of flagella, which is strongly reduced at mammalian physiological temperature [37°C (Peel et al, 1988)] and therefore likely to be strongly affected by host plasma membrane remodeling (see below).
Noticeably, ASM KO mice are also more susceptible to Salmonella infection than wild-type animals, albeit less so than to Listeria (Utermohlen et al, 2003). The apparent discrepancy with our results (showing that Salmonella binding to epithelial cells is not affected by ASM activation) can likely be explained by the fact that the reported increased susceptibility of ASM KO mice to Salmonella derives from an impaired ability of ASM KO macrophages to kill Salmonella (McCollister et al, 2007). Nonetheless, we cannot exclude the possibility that in vivo the inflammatory conditions associated with Salmonella infection can induce a stronger effect at the level of membrane remodeling leading to inhibition of Salmonella invasion of epithelial cells.
Our results reinforce the relevance of the flagellar motility as a crucial mechanism used by bacteria to reach preferred sites of infection, particularly when these are scarcely available, such as under stress conditions. This is well illustrated by the ability of Salmonella-a motile pathogen-to overcome host defenses based on host membrane remodeling, which are thus particularly detrimental to non-motile pathogens. Here, we show that global binding and invasion of Salmonella are unaffected by stress, since flagellar motility enables this bacterium to scan the host cellular surface and accumulate at the remaining permissive entry sites. Conversely, non-motile Salmonella mutants show impaired binding to host cells in stress conditions. These in vitro observations are in agreement with previous reports showing that mutations affecting Salmonella motility impair bacterial colonization of the inflamed mouse intestine, but not of the normal gut (Stecher et al, 2004(Stecher et al, , 2008. This phenotype is, at least in part, attributed to flagella-facilitated chemotactic movement, which facilitates access to the nutrient-rich molecules released as part of the mucosal defense during inflammation (Stecher et al, 2008). However, our findings showing that host membrane remodeling in response to ▸ Figure 7. Intracellular Shigella replication induces remodeling of the host cell plasma membrane in vitro and in vivo.
A, B 3D reconstruction of representative images of ASM (A) or ceramide (B) staining in HeLa cells infected with Shigella WT or mock-treated, analyzed at 3 hpi. ASM, ceramide, Hoechst, and Shigella staining were surface-converted by voxel distance. C, D ASM enzymatic activity (C) and ceramide (D) measurement in HeLa cells infected with Shigella WT or mock-treated, analyzed at 3 hpi. The ASM enzymatic activity was determined in the membrane fraction corresponding to 3.0 × 10 5 cells per condition. Ceramide is shown normalized to mock-treated cells. E-H Representative spinning-disk confocal images of colon tissue sections of non-infected (E and F) and Shigella-infected (G and H) guinea pigs, collected at 4 hpi.
Panels (E and G) are montages of multiple image fields; dashed boxes in (E and G) are shown enlarged in (F and H), respectively. I Mean fluorescence intensity of ceramide signal in surface epithelial cells (outermost tissue layer) of non-infected or Shigella-infected guinea pig colon tissue sections. J Model depicting the effect of stress-induced host plasma membrane remodeling on infection by bacterial pathogens. Different stressors, including TNF-a stimulation and Shigella intracellular replication, induce the relocalization and activation of acid sphingomyelinase (ASM) to the host plasma membrane, leading to the conversion of sphingomyelin into ceramide. As a consequence of membrane remodeling, the binding of non-motile bacteria, such as Shigella, to host cells is inhibited. Motile bacteria, such as Salmonella, are able to scan the host membrane for the remaining permissive binding sites, and thus circumvent this defense mechanism.
Data information: Shigella infection was performed at MOI 100 (A, B). Results are shown as mean AE s.e.m. of 9 (C) or 6 (D) independent experiments; whiskers in the box-plot (I) correspond to min/max from 10 (non-infected) or 13 (Shigella) tissue sections. *P < 0.05, **P < 0.01, ***P < 0.001 (one-way ANOVA for C, D; unpaired t-test for I). stress protects epithelial cells from infection by non-motile pathogens might help to explain the reported impaired in vivo fitness of the non-motile mutant Salmonella strains. The relevance of motility is further reinforced by the demonstration that L. monocytogenes and Y. pseudotuberculosis, which repress flagella at 37°C and are thus non-motile when grown at this temperature (Peel et al, 1988;Kapatral & Minnich, 1995), also exhibit impaired binding/ invasion in stress conditions. Unlike Salmonella, Shigella has lost flagellar motility (Ewing, 1949) despite motility being an important trait for pathogenicity in the intestinal environment at various stages of infection, including reaching the sites of infection, adhesion to and invasion of host cells, and biofilm formation (reviewed in Refs: Rossez et al, 2015;Erhardt, 2016). However, flagellin, the major component of the bacterial flagellum, is a potent inducer of the innate immune response, and it is therefore deleterious for bacterial survival within the host, with the ensuing greater clearance of flagellated versus non-flagellated strains (Lockman & Curtiss, 1990;Dons et al, 2004;Lai et al, 2013;Olsen et al, 2013). This trade-off between higher motility/higher immunogenicity and lower motility/lower immunogenicity is particularly interesting in light of the results presented here.
Importantly, ASM activation and subsequent membrane remodeling under stress have propagative properties linked to the inducing stimuli. For instance, most stress conditions occurring during colonic inflammation, such as TNF-a stimulation, hypoxia, or oxidative stress, elicit responses at the level of cell populations rather than just individual cells. Moreover, a common denominator of the response to stress is the ability of cells to produce signals or activate pathways to alert neighboring cells to the presence of the stress elicitor, and orchestrate a common response (Chovatiya & Medzhitov, 2014). Along this line, our results show that, even in the context of Shigella infection, the remodeling of the host membrane is not restricted to infected cells, but is in fact more evident in the yet non-infected adjacent cells. Specifically, we show that Shigella infection induces ASM activation and strong ceramide accumulation in the bystander cells. Paracrine signaling, for example, TNF-a release upon Shigella infection, can explain the host membrane remodeling in the bystander cells. In addition, direct cell-cell communication can be involved, given that the p38 MAPK pathway activation has been shown to be propagated from Shigella infected to bystander uninfected cells via gap junctions (Kasper et al, 2010). These mechanisms are further supported by our results showing that stress conditions, namely TNF-a, arsenite, or anisomycin treatment, activate the p38 MAPK pathway and that this stimulation is required for the inhibition of Shigella infection upon stress. P38-dependent ASM activation upon infection is also in agreement with the observed stronger accumulation of ceramide in bystander cells, when compared with cells with replicating bacteria. Indeed, p38 activation in cells with internalized bacteria is modest due to the phosphothreonine-lyase activity of the Shigella OspF effector protein (Arbibe et al, 2007;Li et al, 2007).
We show that intracellular replication of Shigella induces host membrane remodeling and inhibition of re-infection by extracellular non-motile bacteria. Importantly, we have also observed membrane remodeling, visualized by a strong ceramide accumulation, in colon samples of guinea pigs infected with Shigella. Accordingly, we propose that membrane remodeling constitutes part of the host response to the stress elicited by Shigella infection, acting as a protective mechanism against further pathogen invasion. In addition to bacterial pathogen infection, this mechanism might also be relevant to restrict commensal bacteria translocation through the intestinal epithelium. Given that the mucus layer is reduced during inflammation (Swidsinski et al, 2007) and thus more bacteria are probing the epithelial surface, it is conceivable that the depletion of membrane rafts limits the docking of such bacteria and their potential translocation through the intestinal barrier. Indeed, commensal bacteria have been shown to exploit membrane raft-mediated transcellular pathways to cross the intestinal epithelium, in mechanisms dependent (Kalischuk et al, 2009) or independent (Clark et al, 2005) of co-infection with bacterial pathogens.
In summary, we describe a novel mechanism by which host epithelial cell response to stress limits Shigella binding and invasion via reinforcement of the physical barrier formed by epithelial cells, through the activation of ASM and resulting depletion of permissive bacterial binding sites. Moreover, we provide evidence that global binding of motile pathogens to host cells is not affected in stress conditions, indicating that motility is an important feature to overcome host defenses based on host membrane remodeling.

Bacterial infections
For bacterial infections, overnight cultures were diluted 1:100 in LB and grown aerobically at 37°C with shaking until OD 600 0.4 (Shigella) or OD 600 2.0 (Salmonella). Bacteria were harvested by centrifugation (12,000 g, 2 min) and resuspended in complete cell culture medium. Bacterial infections were performed at the multiplicity of infection (MOI) indicated in the Figure legends. Following addition of bacteria, cells were centrifuged at 2,000 g for 15 min (Shigella) or 250 g for 10 min (Salmonella) at room temperature. Cells were then incubated at 37°C in a 5% CO 2 humidified atmosphere for 15 or 20 min, respectively, followed by replacement with fresh medium supplemented with 50 lg/ml gentamycin (defined as time point 0 of infection) and incubated for 30 min to kill extracellular bacteria. The medium was then replaced, and cells were maintained in medium supplemented with 10 lg/ml gentamycin, until analysis.
For binding experiments, bacteria were added to the host cells as above, and processed immediately after the centrifugation step (Salmonella) or after an additional 10-min incubation at 37°C (Shigella). Cells were then extensively washed with PBS to remove non-bound bacteria and processed for microscopy, colony-forming units (cfu) assays, or RNA extraction. For Listeria infections, overnight cultures were diluted 1:50 in BHI medium and grown at 37°C with shaking until OD 600 0.7. Listeria was added to the cells after dilution of bacterial cultures in complete medium. Cells were then centrifuged at 250 g for 10 min at room temperature and incubated for 10 min at 37°C. For Yersinia infection, overnight culture was performed at 28°C. The culture was then diluted to OD 600 0.05 and grown with shaking at 28°C for 1 h, followed by 2 h growth at 37°C reaching approx. OD 600 2.0. Dilution of bacteria and infections was performed as described for Shigella, except that following addition of bacteria, cells were centrifuged at 400 g for 10 min, followed by 1-h incubation at 37°C in a 5% CO 2 humidified atmosphere.
To quantify intracellular bacterial replication by cfu assays, cells were washed three times with PBS and lysed in PBS containing ª 2018 The Authors The EMBO Journal 37: e98529 | 2018 0.1% Triton X-100. The lysates were then serially diluted in PBS and plated on LB agar plates. For re-infection assays, the cells were washed with PBS at 3 hpi (primary infection) and re-infected as described above. For treatment with compounds, these were added directly to the medium containing 10 lg/ml of gentamycin at the appropriate time prior to the secondary infection. Cells were then washed with PBS and reinfected. Discrimination of the colonies derived from primary and secondary infections was possible since the strains used for the reinfection experiments have different antibiotic resistance cassettes as follows: Shigella-mCherry, kanamycin resistance, and Salmonella-mCherry, ampicillin resistance (primary infection); Shigella-GFP and Salmonella-GFP (secondary infections), chloramphenicol resistance.

Animals, infections, and sample preparation
Shigella infection of guinea pigs (120-250 g; Charles River Laboratories) and sample preparation were performed as described previously (Arena et al, 2015). Infections were performed in accordance with Institut Pasteur animal protocols (no. 2013-0113). Two animals per condition (non-infected and Shigella infected) were sacrificed at 4 hpi, and the distal 10 cm of colon was harvested and processed for immunofluorescence analysis.
Confocal microscopy images were acquired with a Leica SP5 laser scanning confocal microscope (Leica Microsystems). For the analysis of the number of bacteria per cell, at least 50 infected cells per condition and independent experiment (total of at least 250 cells) were counted manually from maximum-projected Z-stack confocal images.
The 3D reconstruction of Z-stack confocal images was performed using the Imaris software (Bitplane). An average of 20-25 Z-stacks were acquired for each image; the various channels were surfacerendered by voxel distance.
For quantification of Shigella infection and spreading, image acquisition was performed using Operetta automated high-content screening fluorescence microscope (PerkinElmer) at a 20× and 10× magnification, respectively; a total of 9 (infection) or 1 (spreading) images were acquired per coverslip/well, corresponding to approximately 2,500 cells analyzed. Image analysis was performed using Columbus image analysis software (PerkinElmer) as described previously (Sunkavalli et al, 2017).
For the guinea pig colon tissue sections (10-lm-thick transversal sections), samples were permeabilized with 0.5% Triton X-100 in PBS for 2 h and blocked in 1% BSA in PBS for 1 h. Sections were then incubated overnight at 4°C with the primary antibody against ceramide (1:50; Enzo Life Sciences, ALX-804-196-T050), a rabbit polyclonal serum specific for Shigella-LPS (1:300; P.S. Lab), and Alexa Fluor 647 Phalloidin (1:200; Life Technologies, A22287), in 0.1% Triton X-100/1% BSA solution. After two washes with PBS, samples were incubated for 1 h at room temperature with the goat anti-mouse and goat anti-rabbit secondary antibodies conjugated with Alexa Fluor 568 and Alexa Fluor 488 (Life Technologies), followed by DAPI staining (1:1,000; Life Technologies, 62247) for 5 min at room temperature. The sections were then washed with PBS and mounted with ProLong Gold (Life Technologies, P36930).
Images were acquired in an Opterra swept-field confocal microscope (Bruker) with a 100× objective. Montages of multiple image fields were used to reconstruct the tissue sections. The quantification of the mean fluorescent intensity (MFI) of ceramide staining in the sections was performed using Fiji (Schindelin et al, 2012). Briefly, maximal intensity projections of 20 z-stacks were generated and specific region of interests (ROIs) were manually drawn on the surface epithelial layer using the phalloidin staining as reference, and MFI was calculated.

Plasma membrane acid sphingomyelinase activity assays
For determining membrane sphingomyelinase activity, HeLa cells (3.5 × 10 5 cells/well) were seeded in 6-well plates 2 days prior to the assay. ASM activity was measured in the membrane fractions, as described previously (Tonnetti et al, 1999), using the Amplex Red Sphingomyelinase Assay Kit (Invitrogen, A12220).
To obtain the membrane fractions, following treatment with the various compounds or after infection, the cells were washed with PBS, and harvested in PBS with 5 mM EDTA and washed three times in ice-cold PBS with 100 lM sodium orthovanadate at 4°C. Cells were then resuspended in 100 ll of lysis buffer (20 mM Tris-HCl pH 7.5, 2 mM EDTA, 5 mM EGTA, 1 mM sodium orthovanadate, 10 mM b-glycerol phosphate, 1 mM PMSF, 1× protease inhibitors cocktail) and disrupted by five cycles of freezing and thawing in a methanol/ dry-ice bath. The lysate was centrifuged for 10 min at 1,000 g at 4°C, and the supernatant (post-nuclear homogenate) was transferred to a new tube. The supernatant was centrifuged for 1 h at 100,000 g at 4°C. The resulting pellet (membrane fraction) was resuspended in 40 ll 1× reaction buffer provided with the kit. For each treatment/ infection, a replicate well was used for cell counting; 1.2 × 10 6 cells were used on average as input for the extraction of the membrane fractions. 10 ll of the membrane fractions (i.e., corresponding to 3.0 × 10 5 cells per condition) was used for the measurement of ASM activity, which was performed according to the manufacturer's instructions. A standard curve was established for each experiment,

of 17
The EMBO Journal 37: e98529 | 2018 ª 2018 The Authors based on 1:4 serial dilutions of the provided commercial sphingomyelinase (10 U/ml; sphingomyelinase from Bacillus cereus; 1 U corresponds to the amount of enzyme required to hydrolyze 1 lmole substrate in 1 min at 37°C), starting from 4 × 10 À3 U to 0.1 × 10 À5 U in a total volume of 100 ll of 1× reaction buffer. The reactions were incubated at 37°C for 1 h, and fluorescence was measured with a microplate reader using excitation at 530 nm and emission at 590 nm. For each experiment, the fluorescence of the samples was converted to enzyme units using a standard curve.

Ceramide quantification by flow cytometry
After treatment with stressors or infection, cells were collected in PBS + 5 mM EDTA and washed two times in PBS. Cells were then resuspended in 1% BSA in PBS containing the ceramide antibody (1:30, 1 h room temperature, with agitation; Enzo Life Sciences, ALX-804-196-T050). Cells were washed three times in PBS and again resuspended in 1% BSA in PBS containing the anti-mouse secondary antibody conjugated with Alexa Fluor 488 (1:300, 1 h room temperature, with agitation; Life Technologies). Finally, cells were washed and resuspended in cold PBS before measurement. Cells incubated with only the secondary antibody were used as a control. Flow cytometry was performed on a BD Accuri C6 or a FACSAria III flow cytometer (BD Biosciences). FlowJo software (Tree Star Inc) was used for data analysis.

Cell viability
Analysis of cell viability was performed using 7-amino-actinomycin (7-AAD; BD Biosciences, 51-68981E), a viability dye excluded from cells with intact membranes (viable cells), as described previously (Maudet et al, 2014). Flow cytometry was performed on a BD Accuri C6 flow cytometer (BD Biosciences); a minimum of 10,000 cells were analyzed per sample. FlowJo software (Tree Star Inc) was used for data analysis.

Statistical analysis
All data are presented as mean AE standard error of the mean (s.e.m.), with the exact number of experiments performed indicated in the respective Figure legend. Statistical analysis was performed using Prism software (GraphPad). For statistical comparison of datasets from two groups/conditions, two-tailed Student's t-test was used; for data from three or more groups/conditions, one-way ANOVA with Tukey's or Dunnett's post hoc test or two-way ANOVA with Sidak's multiple comparisons test was used. Mann-Whitney test was used for the comparison of medians from two groups. A Pvalue < 0.05 was considered significant. Violin plots were generated using the BoxPlotR web tool.
Expanded View for this article is available online.