S.Tm Infection Triggers Prompt Focal Contractions in Murine IEC Monolayers.

To explore epithelial tissue dynamics upon infection, we developed enteroid-derived epithelial 2D-monolayers placed atop a flexible hydrogel in live-imaging chambers [informed by (26); see SI Appendix, Materials and Methods]. C57BL/6J mouse jejunal enteroids were first kept in 3D-culture containing CHIR99021 (CHIR, a glycogen synthase kinase-3 inhibitor) and valproic acid (VPA) to maximize proliferative potential (33, 34). Enteroids were subsequently fragmented into single cells, seeded out onto collagen-1 hydrogels, and the CHIR/VPA eliminated 24 h postseeding. IECs quickly attached to the hydrogel and extensive proliferation was noted during the first 24 to 48 h (Fig. 1 A–D). By 72 h, a postmitotic confluent epithelial monolayer was established (Fig. 1 A–D), which maintained at 96 h (Fig. 1C), and thereafter gradually deteriorated. The mature monolayer expressed low transcript levels of the stem cell marker Lgr5 (Fig. 1E), elevated levels of the differentiated enterocyte markers alkaline phosphatase and Ezrin (Fig. 1E), and featured a standing IEC arrangement with adherence junctions (E-cadherin) and an apical actin brush border (Fig. 1F).

Fig. 1.

S.Tm elicits prompt focal contractions in nonconstrained murine IEC monolayers. (A–F) Characterization of murine IEC monolayers atop hydrogels. (A) Murine enteroids were disrupted into single cells and seeded over a collagen-1 hydrogel. Depicted are representative DIC images at the indicated time points (hours postseeding; h p.s.) at low (Top) and high (Bottom) magnification. Red lines indicate borders of IEC islands. (Scale bars, 100 µm.) (B) Representative micrographs of monolayers (as in A) incubated with EdU (magenta). Nuclei visualized by DAPI (blue). (Scale bars, 10 µm.) (C) Quantification of monolayer confluence (as in A) and D) percentage EdU-positive IECs (as in B) over time. Data in C–D presented as mean ± SD of at least three independent experiments up to 72h p.s. Kruskal Wallis with Dunn´s post hoc test and Benjamini-Hochberg correction, comparing against the 24h sample (*P < 0.05; **P < 0.01). (E) Transcriptional profiles of 3D-enteroids and mature 2D-monolayers (72h p.s.). Data presented as mean ± SD of six biological replicates per condition, pooled from two separate experiments. Data points shown as filled (3D) or empty (2D) circles, two-tailed Mann–Whitney U test (**P < 0.01). (F) Representative micrographs of monolayers (72h p.s.) stained for Ezrin (Left; white), Actin (Middle; white), E-cadherin (Right; white), and DAPI (nuclei; blue). Depicted are x-y (Left) and z projection (Right) micrographs of confocal Z-stacks. (Scale bars, 10 µm.) (G) Example of an S.Tm-elicited contraction focus and the framework for flow analysis (schematic in margin). Depicted are DIC images (Top) from the indicated time frames (min p.i.), with visualization of vector fields (middle row panels) and speed–color maps (bottom row panels) derived from Cellocity. (Scale bars, 100 µm.) (H) Representative time-lapse series of a monolayer infected with S.Tm^wt at MOI 2. Depicted are DIC images (upper row panels) and speed–color maps (bottom row panels) at the indicated time points. Data representative for >10 independent experiments. (Scale bars, 100 µm.) (I) Quantification of contraction focus (as in G and H) cell numbers, and (J) maximum speeds over time. Each black dot represents one contraction focus. Means depicted by red lines. Gray spotted area represents contraction patchwork (∞).

To assess the impact of bacterial infection, we challenged mature IEC monolayers atop hydrogels with wild-type (WT) S.Tm SL1344 (“S.Tm^wt”) at a multiplicity-of-infection (MOI) of 0.2-2 and imaged them by time-lapse differential interference contrast (DIC) microscopy. Strikingly, within 10 to 20 min postinfection (p.i.), large contraction foci became visible (Movie S1). We employed an optical flow image analysis pipeline to quantify the speed and direction of these epithelial movements [Movies S2 and S3 and Fig. 1G) (35); see SI Appendix for details]. Speed maps calculated from vector fields showed that each focus involved a 0.21 ± 0.14 mm² patch of epithelium (i.e., ∼200 to 1700 IECs; Fig. 1 H and I), contracting toward a single epicentre with inward movements typically lasting 1.2 ± 0.3 min and maximal pixel speeds ranging between ∼2 and 8 µm/min (Fig. 1 G and J). Early upon infection, isolated foci were readily detectable (Fig. 1H, 13 min p.i.), which subsequently coalesced into a patchwork (Fig. 1H, 23 min p.i.). Signs of epithelial cell death (rounding and dislodging of IECs) became evident thereafter, and contractile movements eventually ceased (Fig. 1H, 83 min p.i.). Infection-elicited contractions were also observed upon inoculation with another S.Tm^wt isolate (ATCC 14028s) (SI Appendix, Fig. S1 A and B) and in monolayers from a donor mouse housed in another facility (SI Appendix, Fig. S1 C and D). Exceedingly rare and small contraction foci could in fact occasionally be observed also prior to infection (Fig. 1 I and J). From this we conclude that murine IEC monolayers are wired to elicit prompt focal contractions upon S.Tm infection.

S.Tm TTSS-1 Docking and Invasion Elicits Focal Epithelial Contractions.

S.Tm employs flagella and TTSS-1 to reach and invade the epithelium (3637–38). Using fluorescence-labeled bacteria, we noted that contraction foci emerged in areas with S.Tm bound to the epithelial surface (Fig. 2A). This led us to postulate that S.Tm docking or invasion elicits focal contractions. To test this hypothesis, we infected murine monolayers with S.Tm^wt or S.Tm^ΔinvG (lacks a structural component of TTSS-1). As expected, S.Tm^wt invaded IECs, while S.Tm^ΔinvG failed to do so (SI Appendix, Fig. S1 E and F). Live DIC imaging over a range of MOIs (0.2 to 20) showed that epithelial contractions increased in frequency and intensity with S.Tm^wt dose (Fig. 2B). By sharp contrast, S.Tm^ΔinvG infection did not elicit epithelial contractions, not even at the highest MOI (Fig. 2B). This points to a central role for TTSS-1 in triggering S.Tm-induced epithelial contractions.

Fig. 2.

S.Tm elicits focal epithelial contractions dependent on TTSS-1. (A) Representative time-lapse series of a murine IEC monolayer (as in Fig. 1A, 72h p.s.) infected with mCherry-S.Tm at MOI 0.2. Depicted are processed fluorescence images and speed–color maps at the indicated time points. Fluorescence channel background was corrected by addition of an inverse Gaussian blurred version of the image, and shown are means of five frames to visualize only nonmoving S.Tm (adherent; white). Yellow arrow indicates an S.Tm located at the contraction focus epicenter. (Scale bars, 100 µm.) (B) Speed–color maps of monolayers left uninfected or infected with S.Tm^wt or S.Tm^ΔinvG at the indicated MOIs and time points. Graphs to the right show raw speeds for the whole field of view (f.o.v.) (gray) and local regression (black line) over time, stopping when monolayer integrity was compromised. Red dots atop graphs represent contraction foci over time. Red thick stripe represents contraction patchwork (∞). One representative experiment out of three is shown for each condition. (Scale bars, 100 µm.) (C) Speed–color maps of monolayers infected with S.Tm mutants deficient for TTSS-1 effectors SipA (ΔsipA), SopB, SopE, and SopE₂ (ΔsopBEE2), all four effectors combined (Δ4), or the structural component InvG (ΔinvG). Graphs to the right show raw speeds for the whole f.o.v. upon infection with S.Tm^wt (black) and S.Tm^ΔinvG (gray) (same data in all three graphs), plus the indicated mutant (red), over time. One representative experiment out of three is shown. (Scale bars, 1 mm.)

TTSS-1 docks to the host cell membrane via a SipBC translocon, resulting in the transfer of TTSS-1-effectors into the host cell (39). Among the effectors, SipA stabilizes actin filaments, while SopB, SopE, and SopE2 drive actin filament nucleation through Rho and Arf GTPases (40). These four effectors largely explain TTSS-1–dependent host cell invasion both in vivo and in tissue culture models (38, 41). We used single- and multiple-effector mutant bacteria to assess the relationship between S.Tm IEC docking, invasion, and epithelial contraction. As evident in Fig. 2C, deletion of SipA (S.Tm^ΔsipA), or the combined deletion of SopB, SopE, and SopE2 (S.Tm^ΔsopBEE2), delayed the contraction response. Importantly however, even a strain lacking all four effectors (i.e., ΔsipAsopBEE2; “S.Tm^Δ4”) elicited contractions which exceeded those of S.Tm^ΔinvG-infected monolayers (Fig. 2C). Hence, TTSS-1 docking to the epithelium suffices to elicit focal epithelial contractions, and TTSS-1 invasion effectors augment this effect.

NAIP/NLRC4 Sensing of the Infection at Contraction Focus Epicenters.

An epithelial NAIP/NLRC4 inflammasome can sense bacterial ligands (TTSS rod and needle proteins, flagellin), trigger IEC death, and restrict bacterial spread to deeper tissues in vivo (12, 16, 42). To examine if the epithelial contraction response might be linked to NAIP/NLRC4, we generated enteroids from Nlrc4^−/− (43) and Villin^rfpNaip1-6^Δ/Δ mice (lacking all NAIP receptors and expressing a red fluorescent protein (RFP) reporter; see SI Appendix, Materials and Methods) and established monolayers atop hydrogels for infection. In side-by-side experiments, WT control monolayers again reacted to S.Tm^wt by vigorous contractions (Fig. 3 A–D). This response was, however, completely abolished in both Nlrc4^−/− (Fig. 3 A and B) and Villin^rfpNaip1-6^Δ/Δ (Fig. 3 C and D) monolayers across all MOIs tested. In fact, NLRC4- or NAIP-deficient monolayers appeared as refractory as WT monolayers infected by S.Tm^ΔinvG (compare Figs. 3 A and C and 2B).

Fig. 3.

Epithelial NAIP/NLRC4 sensing of bacterial ligands triggers focal contractions. (A–D) WT and either Nlrc4^−/− (A and B) or Villin^rfpNaip1-6^Δ/Δ (C and D) murine IEC monolayers (as in Fig. 1A, 72h p.s.) were infected in parallel with S.Tm^wt at the indicated MOIs. (A) Speed–color maps of infected WT and Nlrc4^−/− monolayers (time points on top). Graphs to the right show raw speeds for the whole f.o.v. (gray) and local regression (black line) over time, stopping when monolayer integrity was compromised. Red dots atop graphs represent single contraction foci over time. Red thick stripe represents contraction patchwork (∞). One representative experiment out of three is shown. (Scale bars, 100 µm.) (B) Quantification of contraction foci numbers in infected WT and Nlrc4^−/− monolayers (as in A) in the 0 to 10, 10 to 20, 20 to 30, and 30 to 40 min time periods p.i. at the indicated MOIs (uninf; uninfected). Each dot represents one monolayer. Bars indicate means. The symbol “∞” represents contraction patchwork. Data pooled from two to three independent experiments. (C) Speed–color maps and graphs of infected WT and Villin^rfpNaip1-6^Δ/Δ monolayers, depicted as in A. One representative experiment out of three is shown. (D) Quantification of contraction foci numbers in infected WT and Villin^rfpNaip1-6^Δ/Δ monolayers (C), depicted as in B. Data pooled from two independent experiments. (E and F) Chimeric murine IEC monolayers derived from WT (noncolored) and Villin^rfpNaip1-6^Δ/Δ enteroids (red) were infected with S.Tm^wt at MOI 2 through 20. (E) Representative fluorescence plus DIC overlay (Left) and fluorescence-only (Right) image from a time-lapse series of an infected (MOI 2) chimeric monolayer. Stars represent contraction focus epicenters. (Scale bar, 100 µm.) Graph to the right shows the localization of epicenters to WT (black) or Villin^rfpNaip1-6^Δ/Δ (red) regions. Bars represent mean ± SD. Data pooled from four to five monolayers per MOI from two independent experiments, χ² test (*P < 0.05; ***P < 0.001). (F) Representative speed–color map (Left) and fluorescence image (Right) of a single contraction focus. Star represents the epicenter. (Scale bar, 100 µm.) Graph to the right shows the propagation of contraction foci over WT (black) or Villin^rfpNaip1-6^Δ/Δ (red) regions, normalized to the relative area in the f.o.v. Bars represent mean ± SD. Data pooled from in total 53 contraction foci, analyzed from two independent experiments. Paired Mann–Whitney U test (n.s., nonsignificant). (G) Speed–color maps of WT monolayers left uninfected or infected with S.Tm^ΔmotA or S.Tm^ΔfljB/fliC at MOI 30 (time points on top). (H) Speed–color maps of WT monolayers treated with 10 µg/mL flagellin (Fla), 10 µg/mL FlaTox, or 10 µM nigericin (time points on top). In G and H, one representative experiment out of two is shown. (I) Quantification of contraction foci numbers in infected monolayers (as in G), depicted as in B. (J) Contraction foci numbers in treated monolayers (as in H) in the 0 to 30, 30 to 60, 60 to 90 and 90 to 120 min time periods p.t. at the indicated concentrations. Data in I and J pooled from two independent experiments. ND, not determined.

The involvement of NAIP/NLRC4 in epithelium contraction (Fig. 3 A–D) could be explained by several models. In the simplest one, NAIP/NLRC4 activation in an infected IEC initiates the contraction response with that cell at the epicenter. Alternatively, NAIP/NLRC4 might fuel propagation of contraction across the tissue following its initiation by other means. To resolve these possibilities, we established genetically chimeric IEC monolayers. 3D-enteroids were generated from WT and Villin^rfpNaip1-6^Δ/Δ mice, the enteroids fragmented, and the fragments mixed in equal proportions before seeding atop hydrogels. This resulted in the outgrowth of confluent monolayers with alternating red Naip1-6^Δ/Δ and noncolored WT regions (Fig. 3E). The chimeric monolayers were infected with S.Tm^wt, contraction foci tracked by time-lapse DIC imaging, and the epicenter locations overlaid with the RFP-fluorescence channel (Fig. 3E and Movie S4). Strikingly, contraction epicenters mapped virtually exclusively onto noncolored WT regions (>95% of epicenters; Fig. 3E and Movie S4). Following their initiation, however, contractions propagated equally across Villin^rfpNaip1-6^Δ/Δ and WT regions (Fig. 3F). These data support that NAIP/NLRC4 bacterial sensing at invasion sites triggers the focal contraction response.

We next assessed the impact of specific NAIP ligand(s). Epithelial contractions were elicited also by a S.Tm^ΔfljBfliC strain that lacks flagellin (Fig. 3 G and I; nonmotile but flagellated S.Tm^ΔmotA used as control). Combined with the previous results (Fig. 2 B and C), this suggests that TTSS-1 is sufficient to elicit contractions. Nevertheless, delivery of Legionella pneumophila flagellin into the IEC cytosol by the N-terminal domain of Bacillus anthracis lethal factor [“FlaTox” (44, 45)] also elicited contractions, while extracellular Fla or the NLRP3 inducer nigericin failed to do so (Fig. 3 H and J). These findings are consistent with the fact that naïve murine IECs express little NLRP3 but significant levels of all NAIP receptors [i.e., NAIP1, 2, 5, and 6 (9)] and are therefore poised to elicit focal contractions upon sensing of either bacterial TTSS or flagellin ligands.

Gasdermin D Sublytic Pore Formation Sparks Myosin-Dependent Focal Contractions that Precede and Can Be Uncoupled from IEC Death and Expulsion.

Activated NAIP/NLRC4 can in turn activate Caspase-1 or Caspase-8, leading to either Gasdermin D pore formation and subsequent lytic cell death [“pyroptosis” (46)] or apoptotic cell death, depending on the cellular context (12, 47). In the intestinal epithelium, this results in the expulsion of infected IECs into the lumen (12, 15, 16) and the release of proinflammatory mediators (12, 15). We used genetic and pharmacological approaches to probe the relationship between the epithelial contraction response and potential pathways downstream of NAIP/NLRC4.

The adaptor protein Apoptosis-Associated Speck-Like Protein Containing CARD (ASC) associates with activated NAIP/NLRC4 and links this inflammasome to Caspase-8 activation in IECs (12, 47). However, S.Tm^wt infection elicited focal contractions in both Asc^−/− (43) monolayers and matched controls (SI Appendix, Fig. S2 A and B), hence excluding a primary involvement of this pathway. By sharp contrast, monolayers established from mice lacking Caspase-1 [Casp1/11^−/− (48)] or Gasdermin D [GsdmD^−/− (49)] did not exhibit epithelial contractions (Fig. 4 A–D) (i.e., in line with findings in NAIP/NLRC4–deficient monolayers). To examine if lytic IEC death downstream of Caspase-1/Gasdermin D might explain the focal contractions, infections were repeated in WT monolayers in the presence of the dye Draq7, which enters lysed cells. Tracing of fluorescence within contraction foci showed that Draq7-positive expelled IECs only became visible in some cases and always later than the focal contraction phase (Fig. 4 E and F). Moreover, pretreatment with the broad-spectrum Caspase-inhibitor Z-VAD-FMK drastically reduced IEC death upon S.Tm infection but did not impact early contractions (Fig. 4 G–J).

Fig. 4.

Gasdermin D sublytic pore formation sparks myosin-dependent contractions. (A) WT and either Casp1/11^−/− (A and B) or GsdmD^−/− (C and D) murine IEC monolayers were infected with S.Tm^wt at the indicated MOIs. (A) Speed–color maps of infected WT and Casp1/11^−/− monolayers infected at MOI 2 (time points on top). Graphs to the right show raw speeds for the whole f.o.v. (gray) and local regression (black line) over time, stopping when monolayer integrity was compromised. Red dots atop graphs represent single contraction foci over time. Red thick stripe represents contraction patchwork (∞). One representative experiment out of ≥3 shown. (Scale bars, 100 µm.) (B) Quantification of contraction foci numbers in infected monolayers as in A, in the 0 to 10, 10 to 20, 20 to 30, and 30 to 40 min time periods p.i. Each dot represents one monolayer. Bars indicate means. The symbol “∞” represents contraction patchwork. Data pooled from ≥3 independent experiments. (C) Speed–color maps and graphs of infected WT and GsdmD^−/− monolayers at MOI 2, depicted as in A. One representative experiment out of ≥3 shown. (D) Quantification of contraction foci numbers in infected WT and GsmdD^−/− monolayers (C), depicted as in B. Data pooled from ≥3 independent experiments. (E and F) Murine IEC monolayers were infected with S.Tm^wt at MOI 2 in the presence of Draq7. (E) Example of contraction focus at max speed (speed–color map, left) and image series (time relative to contraction) of the Draq7 channel (magenta, top row) and Draq7 channel merged with DIC (bottom row). Black arrows indicate an expelling IEC. (Scale bars, 100 µm, 10 µm in the time-lapse series.) (F) Quantification of Draq7 pixels within focal epicenters. Gray shading represents contraction focus duration. In total, 14 epicenters pooled from two independent experiments. (G) Representative time-lapse series of murine IEC monolayers infected with S.Tm^wt at MOI 2 in the presence of vehicle (DMSO) or 50µM Z-VAD-FMK and Draq7. Depicted are DIC plus fluorescence image overlays (upper row panels; Draq7 in magenta) and speed–color maps (bottom row panels; montage from 0 to 120 min p.i.). (Scale bars, 100 µm [Top] or 1 mm [speed–color maps].) (H) Quantification of the raw speeds for the whole f.o.v. (thin lines) and the total Draq7 signal area (thick lines) of monolayers as in G. (I) Quantification of Draq7-positive cells at 2 h p.i. in infected monolayers as in G. Each dot represents one monolayer. Bars indicate mean ± SD. Data pooled from two independent experiments, one-tailed Mann–Whitney U test (**P < 0.01). (J) Quantification of contraction foci numbers in monolayers as in G, depicted as in B. (K and L) Quantification of contraction foci numbers in infected monolayers treated with the indicated concentrations of K Gd³⁺ or L blebbistatin (bleb), depicted as in B. Data pooled from three (K) or five (L) independent experiments.

Neighboring epithelial cells form an actin purse string to help expel a dying epithelial cell from a cell layer [a process referred to as “extrusion” (50)]. It remained plausible that this phenomenon could be linked to the large focal contractions noted here. However, in the live cell tracing experiments, we only observed IECs expelling from the monolayer subsequent to the contraction phase (exemplified in Fig. 4E). Homeostatic removal of IECs in uninfected monolayers (known to involve a typical extrusion process) were also not accompanied by any contraction foci propagating across the surrounding tissue (SI Appendix, Fig. S3A). Most importantly, in Caspase-1(-11)–deficient monolayers, S.Tm still elicited expulsion of infected IECs, but not focal epithelial contractions, hence formally uncoupling these two phenomena from each other (SI Appendix, Fig. S3B). Finally, typical cell extrusion relies on Sphingosine-1-phosphate (S1P) signaling between neighboring IECs (51), but we found that S.Tm-induced focal contractions proceeded normally also in the presence of S1P biosynthesis inhibitors (SI Appendix, Fig. S3C).

Proinflammatory signaling occurs concomitant with the death of S.Tm-infected IECs and involves interleukin-18 and eicosanoid release (12, 15). Neutralization of IL-18 by saturation with IL-18–binding protein (IL-18BP) had no impact on the epithelial contraction response (SI Appendix, Fig. S4A), neither did preincubation of monolayers with eicosanoid synthesis inhibitors (SI Appendix, Fig. S4 B and C). Transfer of culture supernatants from infected monolayers to naïve monolayers failed to elicit any contraction foci (SI Appendix, Fig. S4D). We also assessed proinflammatory gene transcription and found that this occurred in both S.Tm^wt- and S.Tm^ΔinvG-infected monolayers (SI Appendix, Fig. S4E), whereas exclusively S.Tm^wt-infected monolayers produced contraction foci (Fig. 2). Taken together, these results suggest that NAIP/NLRC4–dependent focal epithelial contractions require minute quantities of active Caspase-1 and the pore-forming protein Gasdermin D (46), but precede and can be uncoupled from IEC death and lysis, the IEC extrusion process itself, and generalized proinflammatory signaling.

Work in macrophages has shown that early pore formation by a small number of Gasdermin D molecules causes immediate ion fluxes, which can be blocked by Gd³⁺ (in the form of Gadolinium chloride) (52). Strikingly, preincubation of WT monolayers with Gd³⁺ revealed a dose-dependent reduction in the number of S.Tm^wt-elicited contraction foci (Fig. 4K). A brief incubation with the Calcium chelator EGTA also markedly attenuated the contraction response (SI Appendix, Fig. S5 A and B). In epithelial cells, nonmuscle myosin II (hereafter simply “myosin”) can respond to ion fluxes and is fundamental for force generation, force transmission at adherence junctions, and apical constriction (53). To inhibit myosin activity, blebbistatin was used, as previously reported by others (54). Short-term blebbistatin incubation did not affect the epithelial monolayer morphology nor the numbers of tissue-associated S.Tm or late-stage IEC death (SI Appendix, Fig. S5 C–F). However, at low dose (10µM), blebbistatin dramatically reduced the number of S.Tm-induced contraction foci, and at high concentration (50µM), contractile movements ceased entirely (Fig. 4L and SI Appendix, Fig. S5G). From these data, we conclude that early sublytic Gasdermin D pore formation results in ion fluxes and the swift propagation of focal myosin contraction pulse(s) across the epithelium.

Early Focal Contractions Densify Infected Epithelial Regions and May Prevent Subsequent Tissue Disintegration.

To determine if infection-induced contractions are relevant also for human gut epithelia, human jejunal enteroids were established and 2D-monolayers cultured atop hydrogels (see SI Appendix, Materials and Methods for details). Imaging of S.Tm^wt-infected human monolayers kept in expansion medium revealed some subtle epithelial movements (Fig. 5A, Upper; compare with Fig. 2B). When the medium was exchanged for Wnt3a-deficient medium at the time of confluence, human monolayers adopted a more differentiated appearance (SI Appendix, Fig. S6 A and B). In response to infection, these human monolayers displayed more pronounced focal contractions (Fig. 5A, Bottom and Movie S5). It should be noted that contraction foci were still fewer, and the contraction events less vigorous, than in the murine monolayers. Notably, neither the human transformed cell line Caco-2 C2Bb1 nor the immortalized mouse cell lines m-ICc₁₂ produced contraction foci when placed atop hydrogels and infected with S.Tm (SI Appendix, Fig. S6 C and D). These results indicate that infection-elicited gut epithelial contraction is a response shared between species but only observed in untransformed epithelial models, and most potent in murine epithelia.

Fig. 5.

Focal contractions compact the epithelial layer and may prevent disintegration. (A) Human enteroid-derived IEC monolayers were generated atop Collagen-1 hydrogels and cultured in expansion medium or differentiation medium (see SI Appendix, Materials and Methods for details). Depicted are representative speed–color maps from time-lapse imaging of monolayers infected with S.Tm^wt at MOI 2 (time points on top). (Scale bars, 100 µm.) (B and C) Murine WT IEC monolayers (as in Fig. 1A, 72h p.s.) were detached from the culture chamber wall and infected with S.Tm^wt or S.Tm^ΔinvG. (B) DIC images at the edge of the monolayer (time points on top). Dashed lines represent the monolayer edge location at 0 (yellow), 30, and 60 min p.i. (red). (Scale bars, 100 µm.) (C) Quantification of the percentage of the monolayer area remaining in the f.o.v. (as in B). Data presented as mean ± SD of five monolayers, pooled from three independent experiments. (D) Quantification as in C of WT and Nlrc4^−/− monolayers infected with S.Tm^wt. Data pooled from seven monolayers from three independent experiments. In C–D, two-way ANOVA with TukeyHSD was performed (*P < 0.05; ***P < 0.001). (E and F) Murine 3D-enteroids were microinjected with mCherry-S.Tm^wt or mCherry-S.Tm^ΔinvG. (E) DIC plus fluorescence image overlays (S.Tm in magenta) at the indicated time points of the time-lapse imaging series. (Scale bars, 100 µm.) (F) Quantification of enteroid size (cross-section area) after microinjection of S.Tm^wt, S.Tm^ΔinvG, or mock, as in E. Data presented as mean ± SD of ≥8 enteroids per condition, pooled from four independent experiments. Two-way ANOVA with TukeyHSD, comparing S.Tm^wt to mock, was performed (n.s., nonsignificant; ***P < 0.001). (G) Representative speed–color map (Left) and blow-up DIC images (Right) from an S.Tm-elicited contraction focus at the indicated time points (seconds before/after maximum speed of contraction pulse). Yellow lines indicate outlines of three IECs. (Scale bars, 100 or 10 µm in speed–color map or DIC images, respectively.) (H) Relative cell area to maximal area over time for IECs outlined in G. (I) Quantification of individual IEC area change within contraction foci in S.Tm-infected monolayers (before and after maximal speed of contraction pulse compared) and in randomly picked regions of uninfected monolayers within the same time frame. Each dot represents one IEC. Mean ± SD depicted by red lines and error bars. Data pooled from three independent experiments. Two-tailed Mann–Whitney U test was performed (***P < 0.001). (J) Representative time-lapse series of murine WT, Casp1/11^−/−, and GsdmD^−/− monolayers infected with S.Tm^wt at MOI 2 in presence of Draq7 (magenta). Yellow lines indicate gaps in the monolayer. (Scale bars, 100 µm.) (K and L) Quantification of Draq7-positive cells at 2 h p.i. in K WT and Casp1/11^−/− monolayers, or L WT and GsdmD^−/− monolayers, infected in parallel at the indicated MOIs. (M and N) Quantification of area occupied by M WT and Casp1/11^−/− monolayers, or N WT and GsdmD^−/− monolayers, at 2 h p.i. with the indicated MOIs. In K–N, each dot represents one monolayer. Bars indicate mean ± SD. Data pooled from ≥3 independent experiments in each case. One-tailed Mann–Whitney U test in K and L and two-tailed Mann–Whitney U test in M and N were performed (*P < 0.05; **P < 0.01; ***P < 0.001, exact P value given in M).

The appearance of multiple contraction foci, each affecting hundreds to thousands of IECs (Fig. 1 H and I), could have consequences for the compaction of the epithelial layer. We addressed this possibility, first by measuring the retraction of the murine monolayer edge, after manually detaching it from the chamber wall prior to infection. Infection with S.Tm^wt produced an edge retraction in WT monolayers already within the first ∼30 to 60 min p.i. (Fig. 5 B and C). By contrast, only negligible edge retraction was observed in response to noninvasive S.Tm^ΔinvG (Fig. 5 B and C) or when Nlrc4^−/− monolayers were infected with S.Tm^wt (Fig. 5D).

To explore the consequences of the 3D epithelial geometry found in the intact gut, we next microinjected S.Tm into the lumen of murine 3D-enteroids and followed the epithelial behavior by DIC imaging. In line with observations from monolayers, focal contractions were visible in infected 3D-enteroids. Strikingly, microinjection of S.Tm^wt, but not S.Tm^ΔinvG or mock, resulted in a reduction of the total enteroid size by ∼25 to 50% during the first 30 to 60 min p.i. (Fig. 5 E and F). This shows that S.Tm infection causes prompt compaction also of 3D-arranged murine epithelia.

As shown here and previously (12, 15, 16), S.Tm infection causes death and removal of infected IECs. We hypothesized that prompt and early focal contractions may prepare the epithelium to withstand the effects of subsequent IEC loss. Indeed, focal contractions led to an increased IEC packing in the affected regions, as evident from an up to 40% (mean 12.3%) reduction in the surface area of individual IECs (Fig. 5 G–I). To further test our hypothesis, we exploited the fact that Caspase-1– or Gasdermin D–deficient IECs still undergo NAIP/NLRC4–dependent cell death and expulsion (SI Appendix, Fig. S3B and refs. 12, 47) but that monolayers from these genetic backgrounds fail to elicit focal, epithelium-compacting, contractions (Fig. 4 A–D). WT, Casp1/11^−/−, and GsdmD^−/− monolayers were infected with S.Tm^wt and imaged for a period extending beyond the contraction phase (i.e., for 2 h). As expected, Casp1/11^−/− and GsdmD^−/− monolayers produced fewer Draq7-positive IECs than WT controls at the end of the imaging series (Fig. 5 J–L), but we still observed expelling (mostly Draq7-negative) IECs (Fig. 5J and SI Appendix, Fig. S3B). Remarkably, despite the attenuated IEC death response, both Casp1/11^−/− and GsdmD^−/− monolayers lost epithelium integrity to a higher degree than WT controls (Fig. 5 M and N). At the end of the imaging series, intact WT monolayer still covered ≥93% of the infected areas, whereas Casp1/11^−/− and GsdmD^−/− monolayers displayed frequent gaps and covered a mean of 62 to 83% of the surface depending on the MOI (Fig. 5 M and N).

Taken together, these data associate early NAIP/NLRC4/Caspase-1/Gasdermin D/myosin–dependent focal contractions with increased local IEC packing and an enhanced capacity of the epithelium to withstand the potentially disruptive effects of subsequent IEC loss.

Ethical Statement.

Human enteroid cultures were established from tissue resected from the proximal small bowel during bariatric surgery. Written informed consent was obtained from each subject. All personal data were pseudonymized before reaching the laboratory. The procedures were approved by the local governing body, Regionala Etikprövningsnämnden, Uppsala, Sweden, license number 2010–157 with addendum 2010–157-1 (2018-06-13). Procedures for keeping of laboratory mice and tissue harvesting were approved by the Kantonales Veterinäramt Zürich, Switzerland, approval no. 193/2016, or by Umeå Regionala Djurförsöksetiska Nämnd, Sweden, approval no. A25-2019, and were performed in accordance with the legal regulations.

A detailed description of bacterial strains and plasmids, FlaTox expression and purification, murine and human enteroid culture, 2D monolayer establishment, epithelial cell line culture, RT-qPCR analysis, infection conditions and chemicals, time-lapse microscopy, immunofluorescence, microinjection, and image analysis can be found in SI Appendix.

Statistical Analysis.

Where applicable, statistical significance was assessed by the Kruskal Wallis with Dunn´s post hoc test and Benjamini-Hochberg correction, the Mann–Whitney U test, the χ² test, or the two-way ANOVA with Tukey’s HSD (Honest Significant Difference), as indicated in the figure legends. Tests were performed in GraphPad Prism (8.4.2) or in R studio.