Mechanical competition triggered by innate immune signaling drives the collective extrusion of bacterially-infected epithelial cells

SUMMARY

Intracellular pathogens alter their host cell mechanics to promote dissemination through tissues. Conversely, host cells may respond to the presence of pathogens by altering their mechanics to limit infection. Here, we monitored epithelial cell monolayers infected with intracellular bacterial pathogens Listeria monocytogenes or Rickettsia parkeri over days. Under conditions where these pathogens trigger innate immune signaling through NF-κB and use actin-based motility to spread non-lytically intercellularly, we found that infected cell domains formed three-dimensional mounds. These mounds resulted from uninfected cells moving toward the infection site, collectively squeezing the softer and less contractile infected cells upward and ejecting them from the monolayer. Bacteria in mounds were less able to spread laterally in the monolayer, limiting the focus’ growth, while extruded infected cells underwent cell death. Thus, coordinated forceful action by uninfected cells actively eliminates large domains of infected cells, consistent with this collective cell response representing an innate immunity-driven process.

eTOC blurb

Bastounis et al. show that, during infection with the intracellular bacterial pathogen L. monocytogenes, large infected domains in epithelial cell monolayers extrude to form three-dimensional mounds due to collective onslaught by their uninfected neighbors. This mechanical competition between uninfected and bacterially-infected cells limits pathogen dissemination through the epithelium.

Graphical Abstract

INTRODUCTION

Mammalian cells communicate through both biochemical and biomechanical signals. Examples of the latter include forces that cells transduce to each other and to their extracellular matrix (ECM), critical for maintaining tissue integrity and barrier function. Intracellular bacterial pathogens can manipulate host cell functions, including mechanotransduction, to disseminate. For example, two intracellular bacterial pathogens that generate actin comet tails for propulsion through the cytoplasm of infected host epithelial cells, Listeria monocytogenes (L.m.) and Rickettsia parkeri (R.p.), both secrete virulence factors that reduce intercellular tension, making it easier for bacteria to spread (Lamason et al., 2016). Conversely, host cells may exhibit biomechanical alterations after infection that could function as innate immune responses limiting bacterial spread. In intestinal epithelial cells infected with Salmonella enterica serovar Typhimurium, (S.Tm.) innate immune activation of the inflammasome triggers apical extrusion of single cells, which are shed into the intestinal lumen, without disrupting barrier function (Knodler et al., 2014; Sellin et al., 2014). Similar single cell extrusion occurs in human intestinal enteroids infected with S.Tm. or L.m. (Co et al., 2019). The mechanism resembles that which normal epithelia use to extrude apoptotic cells (Gudipaty and Rosenblatt, 2017). A dying cell contracts and signals its neighbors to assemble a multicellular contractile “purse-string” that moves basally, squeezing out the extruding cell and promoting formation of cell-cell junctions among the new neighbors (Kuipers et al., 2014; Rosenblatt et al., 2001; Tamada et al., 2007). At low cell density, crawling of neighboring cells also contributes to apoptotic cell extrusion (Kocgozlu et al., 2016). Some bacterial pathogens resist this innate immune extrusion reaction by secreting virulence factors that stabilize cell-ECM adhesions (Kim et al., 2009).

A particular challenge is faced by epithelia that are infected by pathogens capable of direct, non-lytic intercellular spread. In addition to L.m. and R.p., other intracellular bacterial pathogens have developed biochemically distinct mechanisms to trigger the rapid assembly of host cell actin filaments at one pole of the bacterium (Stevens et al., 2006). The pushing forces generated by actin network assembly propel the bacteria through the host cell cytoplasm and into membrane-bound protrusions at the host cell surface (Tilney and Portnoy, 1989). In an epithelial monolayer, these protrusions facilitate bacterial intercellular spread without causing lysis of the originally infected cell (Lamason and Welch, 2017). Actin-driven spread of L.m. in epithelia gives rise to infected foci comprising hundreds of cells within < 24 h after the initial invasion of a single cell (Ortega et al., 2019). In this situation, extrusion of single infected cells using a “purse-string” might be insufficient to limit infection.

Infection of epithelial cells by many pathogens triggers activation of the transcription factor NF-κB (Dev et al., 2011). In cells at rest, NF-κB is inactive and retained in the cytoplasm. During infection, pathogen-associated molecular patterns initiate signal transduction cascades that lead to NF-κB activation and translocation into the nucleus. There it induces expression of genes encoding cytokines and adhesion molecules, promoting the recruitment of immune cells to the infection site (Jiang et al., 2011). For several intracellular bacterial pathogens, including L.m., infection of epithelial cells in a monolayer triggers NF-κB activation and cytokine production not only in the infected cells but also in nearby uninfected bystander cells (Dolowschiak et al., 2010; Kasper et al., 2010). The bystanders may play a role in the innate immune response to infection by amplifying alarm signals produced by infected cells.

To determine whether epithelial monolayers alter their biomechanics to limit pathogen spread, we used time-lapse microscopy to monitor individual infected cell foci for several days after exposure to L.m. or R.p.. For both pathogens, uninfected neighbor cells undergo a dramatic behavioral change reminiscent of the epithelial to mesenchymal transition (EMT) and act collectively to squeeze and force the extrusion of large infected cell domains, forming “mounds”. Mounds arise due to changes in the mechanics of two battling populations, the uninfected “surrounders” and the infected “mounders”. Innate immune signals, and specifically NF-κΒ activation, are major drivers in this mechanical competition. Importantly, cells infected with wild-type (WT) R.p. do not activate NF-κB or form mounds whereas R.p. mutants lacking the outer membrane protein B (OmpB) activate NF-κB and do form mounds. Our results demonstrate that the mechanical changes in epithelial cells leading to mounds are not simply a result of cytoskeletal remodeling associated with infection but are triggered by innate immune signals. Moreover, our findings connect EMT, a conserved cell behavioral transformation critical in development and pathologies (e.g. fibrosis, cancer) with innate immune mechanisms. They also underline the dynamic remodeling capability of epithelia and lead us to propose that coordinated mechanical forces may be a mechanism employed by host epithelial cells to eliminate bacterial infection.

RESULTS

Infection with Listeria monocytogenes (L.m.) alters the organization and kinematics of host epithelial cells to form large multicellular mounds

To study the effects of infection on the kinematics of host cells, we used Madin-Darby Canine Kidney (MDCK) epithelial cells. They form polarized monolayers in culture and are often used to study L.m. infection (Ortega et al., 2019; Pentecost et al., 2006; Robbins et al., 1999). We grew MDCK to confluence and infected with L.m. such that fewer than 1 in 10³ host cells were invaded. Cells were maintained with gentamicin so L.m. could only spread from cell to cell. Infected cell foci grew to include hundreds of host cells over a period of 24–48 h. MDCK cells in uninfected monolayers exhibited regularly spaced nuclei confined to a single layer (Figure 1A). In contrast, L.m.-infected foci formed mounds of infected host cells piling on top of one another to form structures ~50 μm tall and with a volume of ~1.7 × 10⁵ μm³ (Figures 1B–1C and S1A–S1C). We also observed formation of mounds in monolayers of human A431D epithelial cells expressing E-cadherin and untransformed human enteroid-derived cells (Figures 1C and S1D–S1E). Thus infection mound formation may be a widely conserved phenomenon for mammalian epithelial monolayers, including both cell lines and untransformed cells.

Figure 1. Changes in the kinematics of L.m.-infected epithelial cell monolayers.

(A-B) Orthogonal views of uninfected or L.m.-infected MDCK 24 h p.i. (host nuclei: yellow, L.m.: black). In (B) a mound is shown. (C) Barplots of L.m. mound volume for MDCK (N=77), A431D (N=5), and human ileal enteroid-derived cells (N=9) 24 h p.i. (mean+/−SD). (D-E) 3D displacements of MDCK nuclei from uninfected (D) or L.m.-infected well (E, field of view centered in mound). Nuclei tracked 24–44 h p.i.. Trajectories colored by time. (F) Violin plot of mean nuclear speed of MDCK from uninfected or L.m.-infected wells centered in mounds (N=5, n=271+/−65 tracks/experiment). (G-H) Plots of weighted average for each time delay of all MSD curves of MDCK nuclei shown in examples in (D-E) (gray: weighted SD over all MSD curves; error bars: SEM; red line: weighted mean MSD curve from which diffusion coefficient D is calculated; see STAR methods). Fit made in first 500 min. (G) shows zoomed inset. (I-J) Same as (F) for the average D (I) and scaling exponent α (J). (F, I-J) dashed line: median; dotted lines: 25% and 75% quartiles of all tracks. For each experiment (N=5) the average value over the field of view and time shown as triangle (mean+/−SD, USTT: * p<0.01). See also Figure S1 and Movie S1.

Mounds arose due to directional movement of the epithelial cells, as evidenced by monitoring the movement of host nuclei via time-lapse microscopy from 24 h post-infection (p.i.) onwards. Cells from uninfected wells moved randomly at low speed, appearing confined by their neighbors (Figure 1D). In contrast, L.m.-infected cells within foci and their uninfected neighbors moved rapidly and persistently toward the center of the focus, and infected cells also moved upward to form the 3D mound (Figure 1E; Movie S1). Net speeds inside and near L.m.-infected foci increased by 3-fold relative to uninfected monolayers, and cell tracks became straighter (Figures 1F and S1F). Over 16 h of observation the average mean squared displacement (MSD) for cells from an uninfected well remained under 50 μm² (~7 μm net displacement). Movement was constrained, with the MSD reaching a plateau, rather than increasing linearly with time as in unconstrained random movement (Figure 1G). In contrast, the average MSD for cells in and near L.m. foci grew to over 1000 μm² over this same time frame and increased superdiffusively indicative of directed motion (Figures 1H–1J). Thus, cells undergo a behavioral phase transition, switching from a caged to a highly motile state at late stages of infection.

Changes in kinematics were accompanied by alterations in cell morphology. In uninfected monolayers, MDCK formed randomly oriented regularly packed polygons (Figures 2A and S2A). In infected monolayers, infected cells at the mounds’ base (mounders) had smaller z-projected areas than uninfected cells (surrounders) up to 200 μm away from the mounds or cells from uninfected wells (Figures 2A and S2C). Strikingly, the orientation of the long axis of surrounders pointed toward the mound center, and this persisted over 150 μm away from the mound (Figures 2A–2C). Surrounders showed alignment of their long axes parallel to that of their immediate neighbors, indicative of local nematic order in the monolayer (Figure S2B). The aspect ratio and shape factor q (perimeter divided by square root of area), dimensionless numbers used to characterize cell shapes in EMT (Mitchel et al., 2019), were also increased in cells from infected relative to uninfected wells (Figures 2D and S2D). To determine if these changes are due to cytoskeletal rearrangements associated with infection, we performed immunostaining (Figure 2E). While F-actin localized pericellularly for cells in uninfected wells, in infected wells cells contained less pericellular actin. Actin tails were visible in infected cells, and surrounders showed prominent stress fibers. Microtubules were disrupted in infected cells at the mounds’ base, as were intermediate filaments vimentin and cytokeratin (Figures 2E and S2E–S2F). The cytoskeletal changes of mounders versus surrounders are consistent with these cells displaying distinct mechanical properties that could contribute to mounding. Alternatively, these changes could be the result of mounders being squeezed and extruded and not the cause of mounding.

Figure 2. MDCK cells in L.m.-infected wells alter their morphology.

(A) MDCK from uninfected and L.m.-infected wells 24 h p.i. Columns: brightfield image, L.m. fluorescence, cells colorcoded by area and radial alignment θ (see STAR methods). Purple circle surrounds mound. 4^th column shows zoomed view. (Β) Sketch showing that radial alignment is the angle θ between the direction of the cell’s major axis and the axis between the cell centroid and the mound center. (C) Mean θ versus distance from the center of the field of view of cells from uninfected or L.m.-infected wells for the example shown in (A) (solid line: median, shaded areas: 40, 60 percentiles). (D) Violin plot of aspect ratio of cells from uninfected wells (n=5817), uninfected surrounders (n=5163, considered up to ~200 μm away from mound) and infected mounders (n=1248). Dashed line: median, dotted lines: 25, 75% quartiles. For each experiment (N=3) the average value over the field of view is shown as triangle (mean+/−SD, USTT: * p<0.05). (E) Orthogonal views of MDCK from uninfected or L.m.-infected wells 24 h p.i. (host nuclei: yellow, L.m.: black) and corresponding F-actin or vimentin localization. See also Figure S2.

Mechanical cellular competition drives infected cell extrusion en masse

To test whether mounding requires competition between two cell populations, we compared epithelial cells at low (~20% infected) and high (~100% infected) multiplicity of infection (MOI). Although mounds were visible at low MOI, they were absent in wells infected at high MOI (Figures 3A–3B). This suggests that mounding may not be a consequence of loss of cell mechanical integrity by infected cells but instead may require the involvement of uninfected surrounders in a form of mechanical competition. To test this, we used traction force microscopy (TFM) to measure the forces exerted by cells on a soft 3 kPa gel. We found that, as the infection focus grew, infected cells imparted reduced deformations and stresses onto their ECM (Figure 3C; Movie S2). In contrast, the radial deformations u_r of surrounders just at the edge of the mound were large, indicative of them grabbing the ECM and pulling it away from the mound as they move directionally towards it. This traction stress orientation is not consistent with extrusion generated by a “purse-string” but is consistent with lamellipodial protrusion and directed cell migration (Kocgozlu et al., 2016). Thus, mounds are not caused by contraction of infected cells but rather by active crawling of uninfected surrounders that migrate toward the focus, squeezing and extruding the infected cells. Kymographs of u_r and L.m. fluorescence as a function of radial distance from the mound’s center further confirmed that the largest u_r coincided with the edge of the infection focus (Figures 3D and S3A–S3B). This behavior persisted up to 54 h p.i., with the peak of forces moving outwards as the infection focus grew. After 54 h p.i. the peak L.m. fluorescence started decaying. This could be due to the extrusion of infected cells and the inability of the bacteria they carry to spread to new uninfected cells in the basal cell layer. To explore this further, we infected cells with fluorescent L.m. and measured the effect of repetitive washes (to get rid of extruded cells) on the recovery of infected cells from the monolayer. We found that late in infection, half of the infected cells had been extruded into mounds (Figure 3E).

Figure 3. Mechanical competition between uninfected and L.m.-infected cells.

(A-B) Phase contrast images 24 h p.i. of MDCK and corresponding L.m. fluorescence for wells that are 20% or 100% infected. (C) TFM on MDCK adherent on 3 kPa gel and infected at low MOI with L.m. Columns: phase contrast image, L.m. fluorescence, deformations (μm), traction stresses (Pa), radial deformations (u_r: positive values indicate deformations pointing away from the focus center) and overlay of u_r and L.m. fluorescence. Rows: p.i. time. (D) Kymograph of mean u_r and radial L.m. fluorescence as a function of time. The infection focus center is considered the center of the polar coordinate system. (E) Boxplots of fold increase relative to 8 h p.i. in % of L.m.-infected MDCK (flow cytometry). Cells were washed (or not) to get rid of extruded cells (N=2, n=6 samples/experiment, mean+/−SD, WRST: * p<0.01). (F) Violin plots of cell stiffness (kPa) for MDCK from uninfected wells (n=68), surrounders (n=126) and mounders (n=130). Cells were indented 2–8 times. Dashed line: median, dotted lines: 25 and 75% quartiles. For each cell its average stiffness is shown as triangle (mean+/−SD, WRST: * p<0.01). See also Figure S3 and Movie S2.

Infected cells might generate reduced traction because of secreted L.m. virulence factors that weaken cellular tension, such as the internalins C (InlC) (Rajabian et al., 2009) and P (InlP) (Faralla et al., 2018). To test this, we infected MDCK with ΔinlC⁻ or ΔinlP⁻ L.m. but found no differences in resulting mound volumes (Figures S3C–S3E). Since the listeriolysin O toxin (LLO) could also modulate mechanotransduction, we infected cells with LLO^G486D L.m. (Rengarajan et al., 2016). This led to formation of smaller mounds; however, the total L.m. load was also decreased comparably to the degree of mound volume reduction. These results suggest that the functions of known secreted bacterial virulence factors do not drive cellular changes leading to mechanical competition and mounds.

The weakening in mounders’ tractions could be due to changes in their cytoskeletal integrity, resulting in reduced ability to transmit forces. To determine whether cell stiffness changes in infection, we used atomic force microscopy (AFM). We found that surrounders’ stiffness was 4-fold higher than that of mounders (Figure 3F). As compared to cells from uninfected wells, surrounders were 2.5-fold stiffer and mounders 1.6-fold softer, consistent with the cytoskeletal disruption of infected cells at the mounds’ base noted in Figure 2D. In sum, we found that infected cells become softer and less contractile than cells in uninfected monolayers. Concurrently, nearby uninfected cells become stiffer and directionally polarized, grip the substrate, and actively pull on it to move themselves directionally toward the focus. This competition results in infected cells getting squeezed and ejected from the monolayer plane.

Cell-cell adhesions and actomyosin contractility contribute to infection mounds

As mounders and surrounders differ in their biomechanics, we hypothesized that perturbations of cellular forces should inhibit mounds. When L.m.-infected MDCK were treated with ROCK inhibitors Y-27632 and H1152 or the myosin II inhibitor blebbistatin to decrease actomyosin contractility and cellular tractions, we found a reduction in mound volume (Figures 4A–4B and S4A). We then examined the role of intercellular forces by infecting MDCK deleted for αE–catenin (αEcat KO) (Ortega et al., 2017) or knocked down for E-cadherin with L.m. and found dramatic reduction in mound volume (Figures 4C–4D and S4B). Compared to WT MDCK, the speed of αEcat KO cells was increased, but directional persistence and neighbor coordination were decreased (Figure S4C), suggesting that coordination of movement of neighboring cells though robust contacts is important for mounding. Moreover, we found that, as the infection focus grew, differences in tractions of infected versus uninfected cells were abrogated in αEcat KO MDCK (Figure 4F; Movie S3). These findings confirm that cell-cell and cell-ECM forces are tightly coupled and that a difference in force transduction between the two populations is important for mounds to emerge.

Figure 4. Actomyosin contractility and cell-cell adhesions contribute to mound formation.

(A) Barplots of relative L.m. mound volume 24 h p.i. for MDCK treated with 30 μΜ of Y27632, 1 μΜ of H1152, or 50 μΜ of blebbistatin. For each experiment, values are normalized relative to the control mound volume (mean+/−SD, WRST: * p<0.01). (B) Orthogonal views of L.m.-infected MDCK treated with vehicle control or 30 μΜ of Y27632 24 p.i. (host nuclei: yellow, L.m.: black). (C) Barplots of relative L.m. mound volume same as (A) but for WT MDCK, αEcat KO MDCK or MDCK treated with siRNA against E-cadherin. (D) Orthogonal views of WT or αEcat KO MDCK infected with L.m. 24 h p.i.. (E) Boxplots same as in Figure 3E but for L.m.-infected αEcat KO MDCK (N=2, n=6 samples/experiment). (F) Same as Figure 3C but for L.m.-infected αEcat KO MDCK. See also Figure S4 and Movie S3.

Interestingly, L.m. infection foci for αEcat KO (compared to WT MDCK) were larger and had increased total L.m. fluorescence but were less dense in fluorescence intensity (Figures S4D–S4F). Consistently, we found that the fold increase in L.m.-infected αEcat KO cells 24 and 48 relative to 8 h p.i. was higher as compared to infected WT MDCK, and vigorous washing of the L.m.-infected αEcat KO monolayers did not reduce the percentage of infected cells (compare Figures 3E and 4E), further supporting the conclusion that bacterial spread through the monolayer is more efficient when mounds cannot form. To confirm the role of adherens junctions in L.m. spread and mounds, we infected with L.m. human A431D cells mostly not expressing E-cadherin and found a similar inhibition of mounds (Figures S4G–S4H).

To examine the organization of cell-cell adhesions in mounders and surrounders, we performed immunostaining for E-cadherin and ZO-1 to localize adherens and tight junctions. We found more fluorescent puncta of E-cadherin and moderate disruption of ZO-1 for infected cells at the mounds’ base (Figure S4I). To explore the relative contributions of cell-cell junctions in mounders versus surrounders, we then conducted infections in monolayers of αEcat KO MDCK mixed with WT MDCK expressing E-cadherin-RFP (to distinguish populations) in varying ratios. When 75% of host cells were WT, mounds resembled 100% WT monolayers. In contrast, when 75% of cells were αEcat KO, mounds were reduced and resembled 100% αEcat KO monolayers (data not shown). Intriguingly, the higher the percentage of αEcat KO in the infected monolayer the lower the correlation length of movement between neighboring cells, suggesting that lack of coordinated cell movement may inhibit mounding (Figure S4J). Interestingly, when cells were mixed 1:1, we only observed mounds in foci where infected mounders and immediate surrounders were predominately WT, while cells further away from the mound could be of either lineage (Figure S4K). This result, further explored below through simulations, suggests that the presence of functional E-cadherin-based junctions, at least in the infected mounders and the uninfected cells immediately surrounding the mound, is necessary for mounds to emerge.

Simulations suggest that cell stiffness, contractility and intercellular adhesions are critical for mound formation

To explore the mechanism whereby changes in cell mechanics leads to collective extrusion of infected cells, we followed a computational approach formulating the problem in 3D. In our simplified model, infected versus uninfected cells can have distinct mechanical properties, but infection is fixed (bacteria do not spread/replicate). Cells are hexagonal, can deform in all directions, are divided in 3 domains (contractile, adhesive, protruding), and have both a passive stiffness arising from the cytoskeleton and an active contractility associated with the actomyosin contractile apparatus (Figures 5A and S5A–S5C) (Sunyer et al., 2016). Cell-ECM interactions are considered through contact interactions and cell-cell adhesions as a continuous material, following the mechanism depicted in Figure S5A. If cell-ECM displacements are large when cells contract, new cell-ECM adhesions are formed. If cell-ECM displacements are small and there is tensional asymmetry, new protrusions form at the edge of the cells experiencing minimum stress, followed by another round of cell contraction. If there is no tensional asymmetry, there is no protrusion and the given cell just contracts.

Figure 5. Simulations reveal that changes in cell stiffness, contractility and intercellular forces drive mound formation, and laser wounding that tension is reduced in infection.

(A) Cases: (1) all cells uninfected, (2–3) infected cells (denoted with asterisk) are softer and uninfected cells cannot (2) or can (3) create new cell-ECM adhesions, (4) infected cells are softer but cell-cell adhesions are disrupted, (5–6) infected cells reduce only their contractility (5) or passive stiffness (6) and uninfected cells form new ECM adhesions. Plots show cell displacements in the vertical (z) direction in two views. (B) Laser wounding response for MDCK from uninfected (top) or L.m.-infected wells (bottom) 24 h p.i.. Columns: brightfield image superimposed with L.m. fluorescence (green); corresponding E-cadherin localization prior to wounding (t=0 min); E-cadherin localization immediately after 1 min of laser illumination (t=1 min); cell displacement vectors u (50-fold larger); average u_y at t=1 min (positive/negative values point towards/away from wound); average u_y over the time period 30–40 min. (C) Kymographs of the average, with respect to the wound, u_y as a function of time, t and vertical position, y for the examples in (B) (see single and double crosses). (C) Barplots of the mean u_y (along a distance of 100 μm away from the wound, calculated immediately after ablation) for cells originating from uninfected wells, surrounders and infected mounders (N= 3 experiments, mean+/−SD, USTT: * p<0.01). See also Figure S5 and Movie S4.

We assumed that infected cells decrease their passive stiffness and active contractility, based on our experimental results, and examined 4 different cases (Figures 5A and S5D–S5F). In case 1 all cells are uninfected, so during the contraction phase cell displacements and principal stresses are symmetrical and thus cells do not create protrusions. However, in case 2 where a cluster of 7 cells is infected, uninfected cells in close proximity develop tensional asymmetry but since they are unable to form new cell-ECM adhesions they get displaced away from the focus’ center, opposite of our experiments. In contrast, in case 3 where we allow uninfected cells to form new cell-ECM adhesions during the protrusion phase, we find that uninfected cells in close proximity to infected cells develop tensional asymmetry and are displaced towards the infection focus. This leads to squeezing of the infected cells and an increase in their height. Finally, in case 4 where intercellular adhesions are absent, there is no tensional asymmetry and thus no protrusions. Cases 3 and 4 are consistent with our experiments and suggest that surrounders need to form protrusions and cell-ECM adhesions towards the infection focus for mounds to form, and that lack of intercellular adhesions prevents mounding.

We also examined what would happen if only one of the mechanical parameters were altered in infected cells and found that for mounds to arise it is necessary for infected cells to decrease their passive stiffness and/or active contractility while forming robust cell-cell adhesions (Figures 5A and S5G, cases 5 and 6). However, if infected mounders and surrounders immediately adjacent to them (immediate surrounders) form cell-cell adhesions but uninfected cells further away from mounds do not, that is also sufficient to drive mounding, but not the opposite case where infected mounders and immediate surrounders do not form adhesions but uninfected cells further away do (Figure S5H). In both cases cells unable to form intercellular adhesions create stronger cell-ECM adhesions and tensional asymmetry arises when cells contract. However, the levels of stress and asymmetry are different, thus when protrusion takes place mounds occur only in the first scenario. Thus cell-cell adhesions specifically for mounders and immediate surrounders are critical since lack of those inhibits mounding, consistent with our experiments.

Laser ablation demonstrates alterations in tension at cell-cell junctions for infected epithelial monolayers

Because simulations indicated the importance of tensional asymmetries at junctions between infected cells and their uninfected neighbors, we sought to confirm these predictions by performing laser ablation (Fernandez-Gonzalez et al., 2009; Joshi et al., 2010; Kiehart et al., 2000; Kong et al., 2019; Smutny et al., 2015). When we ablated rows of cells in uninfected monolayers, we observed a rapid, symmetric recoil of cells away from the wound on both sides over the first few minutes after wounding (Figures 5B–5D). This response is indicative of preexisting cell-cell tension throughout the monolayer and was followed by active crawling of cells to close the wound. However, the magnitude of the immediate recoil in response to ablation for cells at the margin of L.m.-mounds was substantially less than for uninfected cells, and markedly asymmetric (Figures 5B–5D). Surrounders performed less recoil than infected mounders on the opposite side of the wound. Active crawling to close the wound was observed in both infected and uninfected cells. We also used our computational model to examine predicted responses, which were consistent with our experiments (Figure S5I). These findings show first, that the infected cells are capable of mounting a wound-healing response and move away from the focus center, toward the wound margin. This reinforces that cell kinematics associated with mounding are specifically due to mechanical competition between surrounders and mounders, where surrounders are stronger, even though mounders retain their normal ability to mount a directional wound-healing response despite their infected status. Second, surrounders exhibit reduced intercellular tension relative to cells from uninfected monolayers.

RNA sequencing reveals distinct transcriptional profiles among uninfected cells, infected mounders and surrounders

To explore how signaling regulates the mechanical changes associated with mounding, we followed a transcriptomics approach. L.m.-infected MDCK were flow sorted 24 h p.i. into mounders (L.m.-positive) and surrounders (L.m.-negative) (Figure 6A). RNA sequencing revealed significant numbers of differentially expressed genes (DEGs) when comparing these two groups to each other and to cells originating from uninfected wells (Figures 6B and 6C; Table S1). Compared to cells not exposed to infection, both mounders and surrounders downregulate genes related to intercellular and focal adhesions and to the regulation of the actin cytoskeleton (Figure 6D; Table S1). Genes whose changes in expression are associated with EMT were differentially regulated for both populations originating from infected wells compared to cells not exposed to bacteria (e.g. MMP1, VIM, SNAI1, SNAI2). Surprisingly, mounders showed upregulation in genes associated with DNA replication and ferroptosis, hinting at a stress response or death-related process that infected cells might undergo, which could be linked to mounding. Indeed using several approaches (see STAR methods) we confirmed that cells in mounds undergo death, reinforcing the idea that this process is beneficial for the host, since infected cells are being cleared out of the monolayer (Figures S6A–S6E) and raising the question of whether cell death precedes or follows extrusion.

Figure 6. Infected mounders, surrounders and cells from uninfected wells display distinct transcriptional profiles.

(A) Sketch of the populations RNA-sequenced (N=4 replicates): cells from uninfected well, L.m.-positive and L.m.-negative cells from infected well. (B) Volcano plots of DEGs. The -log₁₀ p-values are plotted against the average log₂ fold changes in expression. For each pair of conditions compared upregulated genes of each group are shown in the corresponding color. (C) PCA on top genes that have ANOVA p value ≤ 0.05 on FPKM abundance estimations. PC1 versus PC3. (D) Pathway enrichment analysis. Infected mounders or surrounders are compared to uninfected cells based on their enrichment score (-log₁₀p). (E) Barplots of cytokine concentration in the supernatant of cells originating from an uninfected or L.m.-infected well 4 or 24 h p.i. (n=8, N=2 experiments, mean+/−SD, WRST: * p<0.01). (F) Barplots of gene expression of CSF2, CXCL8 and CCL2 for infected mounders or surrounders compared to cells from uninfected wells (n=4, mean+/−SD, WRST: * p<0.01). See also Figure S6.

Surrounders and to a lesser degree mounders showed upregulation of genes associated with the IL-17, NF-κB and TNF innate immune signaling pathways, and NF-κB activation in infection was further confirmed through Western blotting (Figure S6F). This indicates that, although surrounders are not themselves infected with L.m., their distinct mechanical behavior might arise from cytokines released in infection which influence their phenotype, gene expression and mechanics. Indeed, we found that the supernatant of cells infected for 24 h with L.m. had enriched GM-CSF, IL-8 and MCP-1 compared to uninfected cells’ supernatant (Figure 6E). Interestingly, gene expression for those cytokines was upregulated to a greater extent for surrounders as compared to mounders (Figure 6F and Table S1). All three cytokines are associated with NF-κΒ signaling, either being NF-κΒ activators, or expressed and secreted in response to it, or both (Hoesel and Schmid, 2013; Kang et al., 2007; Manna and Ramesh, 2005). To assess whether cytokine signaling alone, without actual infection, can induce changes in cell polarization and nematic order as observed in surrounders, we exposed MDCK to conditioned media coming from L.m.-infected wells for 24 h and found that the area and aspect ratio of MDCK were both greater than control cells (Figure S6G–H). Thus at least part of the response and collective behavioral change of the surrounders is due to paracrine signaling.

Apoptosis inhibition does not attenuate mounds, but perturbations in NF-κΒ signaling do

Given that infected cells undergo cell death, we tested whether this was required for mounding by treating infected monolayers with either Z-VAD-FMK or GSK’872 3, which inhibit apoptosis or necroptosis respectively associated with L.m. infection (Zhang and Balachandran, 2019). We found that mounds were similar in volume to control mounds (Figure 7A–B), suggesting that cell death occurs after infected cell extrusion, not before.

Figure 7. NF-κΒ activation contributes to mounding for two unrelated bacterial pathogens.

(A) Barplots of L.m. mound volume 24 h p.i. for MDCK treated with 5 μM GSK’872, or 100 μM Z-VAD-FMK relative to control mound volumes (mean+/−SD, WRST: * p<0.01). (B) Orthogonal views of L.m.-infected MDCK treated with vehicle control, 5 μM GSK’872, or 100 μM Z-VAD-FMK (host nuclei: yellow, L.m.: black). (C) Barplots similar to (A) of relative L.m. mound volume for cells treated with 50 μM of NEMO-binding peptide. (D) Barplots similar to (A) of relative L.m. mound volume for MDCK cells treated with siRNA against SNAI1 compared to control non-targeting siRNA. (E) Orthogonal views of L.m.-infected MDCK treated with 50 μM of NEMO-binding peptide (left) or siRNA against SNAI1 (right). (F) Orthogonal views of WT R.p.- or ompB⁻ R.p.-infected MDCK 72 h p.i.. (G) Barplots of mound volume 72 h p.i. for MDCK infected with WT R.p. or ompB^- R.p. (N=2 experiments) treated (or not) with the NEMO-binding peptide relative to control mound volume (N=1 experiment) (mean+/−SD, WRST: * p<0.01). (H) Boxplots of relative levels of NF-κΒ target genes obtained by RT-qPCR for MDCK originating from uninfected or infected wells with L.m., WT R.p. or ompB^- R.p. For infection with L.m. or R.p, expression levels are normalized relative to expression of control uninfected cells ( N=4, mean+/−SD, t- test: * p<0.01). See also Figure S7.

Since NF-κB-related pathways were upregulated for cells originating from infected wells, we tested whether inhibiting NF-κΒ with the NEMO-binding peptide would affect mound volume, and found a ~25% reduction compared to control mounds (Figures 7C and 7E). Inhibition of NF-κΒ was confirmed through Western blotting and immunostaining (Figures S6F and S7A–S7B). Moreover, treatment of infected cells with the NEMO-binding peptide suppressed the loss of vimentin from the cells at the mound’s base, to a comparable extent as loss of cell-cell junctions (Figures S7C–S7E). This is consistent with the idea that cytoskeletal changes associated with the decrease in passive stiffness of infected cells may be downstream consequences of NF-κΒ activation. Since both MCP-1 and NF-κΒ have been implicated in the regulation of EMT genes such as SNAI1 (Li et al., 2017; Pires et al., 2017; Strippoli et al., 2008; Tian et al., 2018; Weichhart et al., 2015) and given the 6- and 9-fold increase in SNAI1 expression in surrounders and mounders respectively as compared to cells from uninfected cultures, we also knocked down SNAI1 in MDCK (Figure S7F), and found a ~50% decrease in mound volume (Figures 7D–7E and S7F). Since reactive oxygen (ROS) and nitric oxide species (NOS) could be generated in cells in L.m.-infected wells and trigger NF-κΒ activation (Gouin et al., 2019; McFarland et al., 2018; Morgan and Liu, 2011), we treated cells with diphenyleneiodonium (DPI) and L-NIL that inhibit ROS and NOS generation respectively and found a significant reduction in mound volume only for cells where ROS production was inhibited (Figures S7G–S7H). We also used the DCFHA assay to assess ROS generation and found a small increase in DCF fluorescence, indicative of ROS generation, for cells originating from L.m.-infected as compared to uninfected wells (Figure S7I). Overall, these results suggest that innate immune signaling through NF-κΒ activation driven possibly due to ROS production, and subsequent cytokine signaling play a role in promoting mounds, contributing to loss of mechanical stiffness in mounders and triggering cell shape changes and motility responses similar to EMT in surrounders.

If our conclusion that NF-κB activation contributes to mounds is correct, then we would also observe mounds if we were to infect MDCK with a different intracellular bacterial pathogen that activates NF-κB. To test this, we infected MDCK with WT R.p. or the ompB⁻ mutant (Figures 7F–7G). OmpB protects the rickettsial surface from ubiquitination in the cytoplasm of infected host cells (Engstrom et al., 2019), and when S.Tm. become ubiquitinated, NF-κB is activated (Noad et al., 2017). We thus hypothesized that ompB⁻ R.p. would trigger NF-κB activation whereas WT R.p. would not. Observation of NF-κB nuclear translocation and RT-PCR on NF-κB target genes confirmed this hypothesis (Figures 7H and S7A–S7B). Importantly, MDCK infected with ompB⁻ R.p. formed mounds, while those infected with WT R.p. did not, and ompB⁻ R.p. mounds were inhibited by treatment with the NEMO-binding peptide (Figure 7F–G). Similar to L.m. mounds, we found that ZO-1 localization was perturbed for ompB⁻ R.p. mounds but not for WT R.p.-infected MDCK which resembled cells from uninfected wells (Figure S7J). Also, similar to L.m.-infected MDCK, the supernatant coming from ompB⁻ R.p.-infected MDCK was 2-fold richer in MCP-1 compared to supernatant from WT R.p.-infected MDCK and uninfected cells (Figure 7K). Together, these results support that cellular changes in mechanics associated with mounds are not simply a result of cytoskeletal alterations driven by pathogens such as L.m. and R.p. that use actin comet tails for intercellular spread. Rather, NF-κΒ signaling is key to eliciting the mechanical competition between infected and uninfected cells that leads to mounds and collective clearance of infected cells in epithelial monolayers.

DISCUSSION

Mechanical forces drive the remodeling of tissues during morphogenesis and homeostasis (Ohsawa et al., 2018). In the epithelium, these forces promote cell extrusion to prevent accumulation of excess or unfit cells (Grieve and Rabouille, 2014; Gudipaty and Rosenblatt, 2017). Infection of epithelial monolayers with intracellular bacterial pathogens also triggers extrusion of single infected cells, which represents an innate immune protective mechanism of the host (Knodler et al., 2014; Sellin et al., 2014), although in some cases it contributes to bacterial spread (Knodler et al., 2010). Here we describe a collective mechanical response of epithelial monolayers, actively forming mounds that can contribute to clearance of domains comprising hundreds of infected cells. We argue that this is beneficial for the host, since bacteria carried by extruded cells cannot spread along the basal monolayer, limiting the infection focus’ growth. Additionally, extruded cells undergo death, probably due to their forcible separation from survival signals communicated to normal epithelia by direct contact with their basement membrane. In the geometry associated with the primary site of L.m. infection in mammalian hosts, the small intestine, the extrusion of infected cells would drive their shedding into the intestinal lumen, where they could be eliminated from the body by the fecal route. Intriguingly, bacterial infections in the intestine of Drosophila melanogaster, model for human intestinal infection (Apidianakis and Rahme, 2011), trigger massive remodeling of the intestinal epithelium, featuring elimination of infected cells and replacement of damaged tissue by remodeling (Buchon et al., 2010). Likewise, infection of epithelia with intracellular viral pathogens increases polarization and motility of uninfected cells, leading to mounds of infected cells in vitro and in vivo (Abaitua et al., 2013; Beerli et al., 2019).

In our system, mound extrusion depends on active crawling of uninfected cells surrounding the infection focus, with those becoming morphologically polarized and nematically ordered. Both the softer and less contractile infected cells of the mound and their uninfected neighbors undergo a biomechanical transition reminiscent of EMT (compare Yang et al., 2020), suggesting that EMT can be specifically harnessed in the context of bacterial infection to contribute to clearance of infected cells and re-epithelization. Mechanical competition between two cell populations, where one of them loses and gets eliminated, have been studied during oncogenesis and various processes are thought to lead to losers’ elimination (Matamoro-Vidal and Levayer, 2019)Hogan et al., 2009; Wagstaff et al., 2016). Here, we find that paracrine signaling arising due to infection is a major driver, consistent with previous studies on viral infection of epithelia (Abaitua et al., 2013; Beerli et al., 2019). Differential sensitivity to mechanical stress is also key since both cell stiffness and tractions are dramatically different between mounders and surrounders. Notably though, mounders remain capable of initiating a migratory wound-healing response, demonstrating that they are not mechanically passive, but rather cannot resist the coordinated onslaught by surrounders. Other signals may also contribute to coordinated behavior of surrounders. For example, the MEK-ERK axis has been implicated in the collective extrusion of UV-treated cells (Aikin et al., 2019) and of virally-infected cells (Beerli et al., 2019).

In this work, we are intrigued by the central role of NF-κB in mound formation for both L.m. and R.p., because of its involvement in the host’s innate immune response to infection by many pathogens (Dev et al., 2011). Studies have shown that L.m. infection of certain cell types leads to NF-κB activation (Rahman and McFadden, 2011), although the specific trigger(s) from the bacterial side may vary (Drolia et al., 2018; Gouin et al., 2010; Mansell et al., 2000). ROS promotes intercellular communication in infection (Dolowschiak et al., 2010) and signaling that leads to NF-κΒ activation in L.m.-infected cells (Herb et al., 2019). We found that ROS are involved in L.m.-mounding, but whether they act upstream of NF-κΒ is to be determined. In our experiments, we did not observe NF-κB activation in MDCK infected with WT R.p., which correlated with lack of mounds, in contrast to L.m.- and ompB⁻ R.p.-infected cells. This provides one rationale as to why many pathogens, including viruses, that are highly adapted to an intracellular lifestyle, and specifically those that are capable of non-lytic intercellular spread, may have evolved mechanisms to suppress activation of immune signaling pathways that otherwise would lead to mounds (Rahman and McFadden, 2011; Reddick and Alto, 2014).

Our discovery underscores the importance of mechanics in regulating infection and the relationship between innate immune signals and cellular biomechanics, specifically EMT. Intriguingly, NF-κB activation is linked to EMT in contexts not involving infection (Strippoli et al., 2008; Tian et al., 2018). Studying the dynamics of these signals using live-cell biosensors will further reveal the spatiotemporal crosstalk between innate immunity and cell biomechanics. In this work, we propose a general cooperative epithelial extrusion mechanism, that could quickly help limit the local spread of infection in epithelia, particularly in the vulnerable surfaces of the lung and intestine, common sites of pathogen invasion where the outside is separated from the inside of the body is separated by a single epithelial monolayer. Whether these results can be recapitulated in vivo in the presence of additional cells, varying ECM topography and mechanical cues has not yet been explored. Continued investigation using more complex assays will uncover the precise biomechanical signals that regulate infection and how host cells modulate those by orchestrating an innate immune response to clear infection.

STAR METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for reagents may be directed to and will be fulfilled by the Lead Contact Julie Theriot jtheriot@uw.edu (J.A.T.).

Materials availability

Materials developed in this study are available on request to the corresponding author.

Data and code availability

Data collected and computer codes are available on request to the corresponding authors. The RNA sequencing data (FASTq files) generated during this study and subsequent analysis are available at the Gene Expression Omnibus (GEO) database (weblink: https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE140626, series record number: GSE140626.) All differential expression analysis results of this study are included as supplementary tables in this article.

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Cell culture

Type II MDCK cells (generous gift from the Bakardjiev lab, University of California, San Francisco) were cultured in high glucose DMEM medium (Thermofisher; 11965092) containing 4.5 g/L glucose and supplemented with 10% fetal bovine serum (GemBio; 900–108) (Ortega et al., 2017). Passages were between P10-P40. MDCK cells expressing E-cadherin-RFP were a generous gift from the Nelson lab, Stanford University (Perez et al., 2008). α-catenin knockout (αEcat KO) MDCK cells were also a generous gift from the Nelson lab, Stanford University (Ortega et al., 2017). A431D human epithelial carcinoma cells were a generous gift from Cara Gottardi, Northwestern University and were grown in DMEM with high glucose (4.5 g/L) in the presence of 10% FBS and 1% penicillin-streptomycin. Transduced cell lines that express full length E-cadherin were cultured in media supplemented with 800 μg/mL geneticin and were generated as described previously (Ortega et al., 2017).

Human ileal enteroid cells

Human ileal enteroids (HIE5) were originally cultured from normal healthy ileal tissue obtained from surgical resection (Holly and Smith, 2018). For continuous culture, human ileal enteroids were maintained in Matrigel domes (Corning, Growth Factor Reduced) in human complete crypt culture medium (hCCCM). hCCCM medium is composed of 50% Wnt3a conditioned media (CM), 10% R-spondin-1 CM, 10% Noggin CM, 1X B-27 supplement, 10 mM HEPES, 1X Glutamax, 1X antibiotic-antimycotic, 1X Non-Essential Amino Acids Solution (Caisson Labs), 1 mM N-acetyl-L-cysteine (Sigma), 10 μM Y-27632 (Abcam), 50 ng/ml mouse epidermal growth factor (EGF, Peprotech), 10 nM gastrin (Sigma), 50 ng/ml fibroblast growth factor-2/basic (Peprotech), 100 ng/ml insulin like growth factor-1 (BioLegend), 500 nM A 83–01 (Tocris) and 10 mM nicotinamide in DMEM. CM production and quality control were performed as described elsewhere (Holly and Smith, 2018). Medium components are from Thermo Fisher Scientific unless otherwise noted.

Monolayers of human ileal enteroid cells derived from human ileal enteroids were generated as described (Holly and Smith, 2018). Briefly, human ileal enteroids were removed from Matrigel using cell recovery solution (Thermo Fisher Scientific), dissociated in 0.05% trypsin at 37°C for at least 6 min, quenched with DMEM containing 10% FBS, 10 mM HEPES, and 1X Glutamax, and mechanically dissociated with a pipette. Cells were then passed through a 40 μm cell strainer, centrifuged at 400 x g for 5 min and then resuspended in hCCCM lacking antibiotic-antimycotic and nicotinamide. 3.5 × 10⁵ cells per well were plated onto 96-well thin (~ 200 μm) polystyrene plates. The wells were first coated overnight at 37°C with 1 mg/mL human placental collagen, type IV (Sigma-Aldrich) diluted 1:30 in sterile dH₂0 prior to adding the cells. The cultures were incubated for 7 days prior to L.m. infection, with media changes at least every 2 days.

Bacterial strains used in this study

The Listeria monocytogenes (L.m.) strains used in this study are: JAT607 (Species: L.m. 1043S, Genotype/Description: ActAp::mTagRFP) (Ortega et al., 2017), JAT605 (Species: L.m. 1043S, Genotype/Description: Constitutive GFP) (Ortega et al., 2017), ΔinlC (Species: L.m. 1043S, Genotype/Description: ΔinlC ActAp::mTagRFP), ΔinlP (Species: L.m. 1043S, Genotype/Description: ΔinlP, obtained from the Bakardjiev lab) (Faralla et al., 2018), JAT983 (Species: L.m. 1043S, Genotype/Description: LLOG486D ActAp::mTagRFP) (Rengarajan et al., 2016), Δhly (Species: L.m. 1043S, Genotype/Description: Δhly) (Rengarajan et al., 2016). Note that we used the LLO^G486D L.m. to test the role of LLO in induction of mounds and in activating signaling pathways that affect mechanotransduction., LLO^G486D L.m. is a strain that carries a point mutation in the gene encoding LLO and thus exhibits 100x less hemolytic activity than the WT protein, permitting a moderate degree of bacterial proliferation and intercellular spread (unlike strains completely lacking LLO which are unable to proliferate in host cells) (Rengarajan et al., 2016).

WT Rickettia parkeri (R.p.) strain Portsmouth was originally obtained from Dr. Chris Paddock (Center for Disease Control and Prevention, NCBI accession no. NC_017044.1), and the ompB⁻ mutant (ompB^STOP::tn) was isolated and validated in a recent study (Engström et al., 2019). R.p. propagation and purifications were from five T175 flasks of Vero cells growing in DMEM (Gibco, 11965) with high glucose (4.5 g l⁻¹) and 2% FBS (Benchmark) that after 5–6 days of infection were harvested in the media using a cell scraper. Infected cells were then centrifuged 12000g for 15 min at 4 °C, resuspended in cold K-36 buffer (0.05 M KH₂PO₄, 0.05 M K₂HPO₄, pH 7, 100 mM KCl and 15 mM NaCl), and bacteria were then released using a dounce-homogenizer. The homogenate was centrifuged at 200g for 5 min at 4 °C and the supernatant was then overlaid onto cold 30% v/v MD-76R (Mallinckrodt Inc., 1317–07) diluted in K-36, and centrifuged at 58300g for 20 min at 4 °C in an SW-28 swinging-bucket rotor. The bacterial pellet was resuspended in cold 1x BHI broth, aliquoted and frozen at −80 °C.

METHOD DETAILS

Bacterial growth conditions and infections

MDCK or A431D cells seeded on glass or polyacrylamide collagen I-coated substrates were infected with L.m. as previously described with the following modifications (Bastounis et al., 2018; Ortega et al., 2019). Briefly, three days before performing the infection assay, L.m. were streaked out from a frozen glycerol stock onto brain heart infusion (BHI, BD; 211059)) agar plates containing 200 μg/mL streptomycin and 7.5 μg/mL chloramphenicol and were placed at 37°C for 2 days. L.m. were then inoculated from plate colonies and grown in 2 mL BHI liquid cultures containing 200 μg/mL streptomycin and 7.5 μg/mL chloramphenicol at room temperature, in the dark for ~ 16 h. The day prior to infection, host cells were seeded at a density of 2×10⁵ cells/well for cells residing on wells of 24-well plates and grown for 24 h. Flagellated overnight cultures of L.m. (OD₆₀₀ approximately 0.8) were washed twice by centrifugation at 2000 x g in PBS to remove any soluble factors. Infections were then performed in normal MDCK growth media by adding 1 mL of L.m. to 15 mL of media and then adding 1 mL of the mixture per well. After 30 min of incubation at 37 °C, samples were washed 3 times in PBS and then media was replaced with media supplemented with 20 μg/mL gentamicin. Multiplicity of infection (MOI) was determined by plating bacteria at different dilutions, on BHI agar plates with 200 μg/mL streptomycin and 7.5 μg/mL chloramphenicol and measuring the number of colonies formed two days post-infection. The resulting MOI was ~200 bacteria/cell. Similar approach was followed when A431D cells were infected with L.m.

Flow cytometry experiments of infected host cells were performed 24 h after infection, unless otherwise stated. For drug exposure experiments, unless otherwise indicated, media was removed from the cells and replaced with media containing either the drug or vehicle control 4 h p.i. Infection mounds were examined at 24 h p.i. Pharmacological inhibitors used were: para-nitroblebbistatin (target: Myosin II inhibitor, concentration: 50 μM, source: Fisher, NC1706059); Y-27632 (target: ROCK inhibitor, concentration: 30 μM, source: Sigma, 129830–38-2); Z-VAD-FMK (target: pancaspace inhibitor, concentration: 100 μM, source: MilliporeSigma, 5.30389.0001); GSK’872 (target: RIPK1 and RIPK3 inhibitor, concentration: 5 μM, source: Fisher, 64–921-0); H1152 (target: ROCK inhibitor, concentration: 1 μM, source: Tocris, 2414); Diphenyleneiodonium chloride (DPI) (target: NADPH oxidase, nitric oxide synthase, ROS, concentration: 0.5 μM, source: Sigma, D2926); L-NIL dihydrochloride (target: inducible nitric oxide synthase, concentration: 1 μM, source: SCBT, 159190–45-1); NEMO binding peptide (target: NF-κΒ inhibitor, concentration: 50 μM, source: Fisher, 480025).

For L.m. infection of human ileal enteroid-derived cells, prior to infection cells were incubated with 100 μL of 4 μg/mL Hoechst for 45 min to stain the host cell nuclei. L.m. infection was then performed similar to L.m. infection of epithelial cells described above. At 24 h p.i. confocal images of L.m. infection mounds of human ileal enteroid cells were obtained without fixation since fixation steps detach extruded cells of the infection mounds. We acquired confocal images of the Hoechst-stained host cell nuclei and of the fluorescent bacteria with 0.2 μm spacing using a spinning disk confocal with a 60× 1.4NA Plan Apo oil objective.

R.p. infections of epithelial cells were carried as previously described with the following modifications (Engström et al., 2019). Briefly, 3.5×10⁵ MDCK cells were seeded in 24-well plates with sterile circle 12-mm coverslips (Thermo Fisher Scientific, 12–545-80) ~24 h prior to infection. The following day, 3×10³ purified WT or ompB^STOP::tn infectious R.p. were used to infect each well in a 24-well plate. Diluted bacterial suspensions were centrifuged onto cells at 300 x g for 5 min at room temperature and subsequently incubated at 33°C for 24, 48, 72 or 96 h as indicated, fixed for 10 min at room temperature in pre-warmed (37°C) 4% paraformaldehyde (Ted Pella Inc., 18505) diluted in PBS, pH 7.4, then washed 2 x with PBS and stored at 4°C prior to immunofluorescence staining.

Flow cytometry of MDCK epithelial cells infected with L.m.

8, 24 or 48 h p.i., MDCK cells infected with L.m. (JAT607, ActAp::mTagRFP) were detached from the substrate with 200 μL of 0.25% trypsin-EDTA for 10 min. For removing extruded infected cells wells were washed three times in PBS prior to addition of trypsin-EDTA. Trypsin-EDTA solutions in each well were then pipetted up and down 6 times to ensure single cell suspensions and 200 μL of complete media was added to inactivate trypsin in each well. Suspensions were transferred into 35-μm cell strainers, (Falcon, 352235) and spun through at 500 x g followed by fixation in 1% paraformaldehyde. Samples were then washed once in PBS and stored in PBS with 1% BSA. Flow cytometry analysis was performed on a BD FACS Canto RUO analyzer (University of Washington Cell Analysis Facility). 10,000–20,000 cells were analyzed per each replicate. To ensure analysis of single cells, the bulk of the distribution of cell counts was gated using the forward versus side scatter plot. This gating strategy ensures that single cells are analyzed and debris or cell doublets or triplets are eliminated from the analysis. A second gating step was then performed to exclude cells that autofluorescence by measuring the fluorescence of control-uninfected cells and gating the population of infected cells to exclude autofluorescence.

MDCK cell transfection with siRNA

For each well of a 24-well plate, 2×10⁴ MDCK cells suspended in serum free media were reverse-transfected with siRNAs at 20 nM final concentration using 0.25 μL lipofectamine RNAiMAX (Invitrogen 13778075). The transfection mix was replaced by full media 8 h later. Synthetic siRNA pools (including 2 distinct siRNA sequences) to target CDH1 and SNAI1 were purchased from Dharmacon (ON-TARGETplus Non-Targeting Pool, catalog number: D-001810–10; Custom CDH1 duplex, catalog number: CTM-521910 and CTM-521911; Custom SNAI1 duplex, catalog number: CTM-544153 and CTM-544153). MDCK cells were treated with control (non-targeting) or experimental siRNA in accordance to the manufacturer’s recommendations. Specifically, to demonstrate that transfection performed was sufficient to get siRNAs into the cells, we transfected cells with synthetic siRNA, siGLO, that makes transfected cells fluorescent 24 h post-transfection To track cell cycle phenotype and to verify that knockdown has occurred with our protocol, we transfected cells with siKif11 which results in substantial cell death of transfected cells approximately 24–48 h post-transfection and can be verified by microscopy using phase optics. Bacterial infections were performed approximately 72 h after transfection.

RT-PCR

To confirm the knockdowns, MDCK cells treated with control (non-targeting) or experimental siRNA 72 h after transfection (approximate cell concentration was 1.6×10⁵ cells/ well) were harvested and lyzed using the QIAshredder Kit (Qiagen, 79656). mRNA was harvested using the RNeasy Plus Micro Kit (Qiagen, 74004) and eluted in 30 μL RNAase free water RNA concentrations were measured spectrophotometrically (NanoDrop) and were comparable between conditions. cDNA was prepared using the Superscript III First-strand Synthesis SuperMix (Invitrogen, 18080085). RT-qPCR was performed using the SYBR qPCR Master mix by Arraystar Inc. Genes of interest were amplified using the appropriate primers: for CDH1 forward:5’ AGACCCAGTAACTAACGACGG 3’ and reverse:5’ ACACCAAAGTCTTCAGGGATT 3’; for SNAI1 forward:5’ CCCCCATTTGGCTGTGTTG 3’ and reverse:5’ ATCAGTCTGTCGGCTTTTATCCT 3’; for β-actin forward:5’ CCCAGCACAATGAAGATCAAGAT 3’ and reverse:5’ CAAGAAAGGGTGTAACGCAACT 3’. Briefly the steps followed were: (1) perform RT-qPCR for each target gene and the housekeeping gene GAPDH; (2) determine gene concentration using the standard curve with Rotor-Gene Real-Time Analysis Software 6.0; (3) calculate relative amount of the target gene relative to GAPDH.

Similar procedure was followed to confirm expression of NF-κΒ target genes. MDCK cells were infected or not for 24 h with L.m.. MDCK cells were also infected or not or for 72 h with WT R.p. or the ompB⁻ mutant. For each condition 4 replicates were prepared. Cells were harvested and lyzed, and their mRNA was harvested as described above. RT-qPCR was performed as above. NF-kB target genes tested were NFKBIA, CXCL8, ICAM1, SNAI1 and SNAI2. Genes of interest were amplified using the appropriate primers: for NFKBIA forward:5’ CACATTCCCTAGCCAGAAACATT 3’ and reverse:5’ TACACCTGGTTGTCACGCATC 3’; for CXCL8 forward:5’ GAAAACTCAGAAATCATTGTAAAGC 3’ and reverse:5’ ATCTTGTTTCTCAGCCTTCTTTAG 3’; for ICAM1 forward:5’ ACCGAGGGTTGGGATTGTT 3’ and reverse:5’ GTCTCTGGCAGGACAAAGGTT 3’; for SNAI1 forward:5’ CCCCCATTTGGCTGTGTTG 3’ and reverse:5’ ATCAGTCTGTCGGCTTTTATCCT 3’; for SNAI2 forward:5’ GAAGCATTTCAACGCCTCC 3’ and reverse:5’ ACTCACTCGCCCCAAGGAT 3’; for β-actin forward:5’ CCCAGCACAATGAAGATCAAGAT 3’ and reverse:5’ CAAGAAAGGGTGTAACGCAACT 3’.

Fabrication of polyacrylamide hydrogels and traction force microscopy (TFM)

Polyacrylamide hydrogel fabrication was done as previously described (Bastounis et al., 2018). Briefly, glass-bottom plates with 24 wells (MatTek, P24G-1.5–13-F) were incubated for 1 h with 500 μL of 1 M NaOH, then rinsed with distilled water, and incubated with 500 μL of 2% 3-aminopropyltriethoxysilane (Sigma, 919–30-2) in 95% ethanol for 5 min. Following rinsing with water 500 μL of 0.5% glutaraldehyde were added to each well for 30 min. Wells were rinsed with water again and dried at 60°C. To prepare polyacrylamide hydrogels of 3 kPa, mixtures containing 5% acrylamide (Sigma, A4058) and 0.1% bis-acrylamide (Fisher, BP1404–250) were prepared (Bastounis et al., 2018). Two mixtures were prepared, the second of which contained 0.2 μm fluorescent beads at 0.03% (Invitrogen, F8811) for TFM experiments. 0.06% ammonium persulfate and 0.43% TEMED were then added to the first solution to initiate polymerization. First, 3.6 μL of the first mixture without the beads was added at the center of each well, capped with 12-mm untreated circular glass coverslips, and allowed to polymerize for 20 min. After coverslip removal 2.4 μL of the mixture containing tracer beads was added and sandwiched again with a 12-mm untreated circular glass coverslip and allowed to polymerize for 20 min. Next, 50 mM HEPES at pH 7.5 was added to the wells, and coverslips were removed. Hydrogels were UV-sterilized for 1 h and then activated by adding 200 μL of 0.5% weight/volume heterobifunctional cross-linker Sulfo-SANPAH (ProteoChem; c1111) in 1% dimethyl sulfoxide (DMSO) and 50 mM HEPES, pH 7.5, on the upper surface of the hydrogels and exposing them to UV light for 10 min. Hydrogels were washed with 50 mM HEPES at pH 7.5 and were coated with 200 μL of 0.25 mg/ml rat tail collagen I (Sigma-Aldrich; C3867) in 50 mM HEPES at pH 7.5 overnight at room temperature. Next morning, the collagen coated surfaces were washed with HEPES and gels were stored in HEPES.

TFM was performed as previously described (del Álamo et al., 2007; Lamason et al., 2016). Briefly, in TFM, cells actively pull on their ECM depending on how well their focal adhesions are organized and connected to the underlying cytoskeleton, and cellular force generation can be inferred from displacement of fluorescent tracer particles embedded in the deformable ECM (Bastounis et al., 2014; del Álamo et al., 2007). Prior to seeding cells hydrogels were equilibrated with cell media for 30 min at 37°C. Cells were then seeded to a concentration of 2×10⁵ cells per well directly onto the hydrogels 24 h prior to infection. Cells are then either infected or not with low dosage of L.m. and imaging started 4 h post-infection. Multi-channel time-lapse sequences of fluorescence (to image the beads within the upper portion of the hydrogels) and phase contrast images (to image the cells) were acquired using an inverted Nikon Eclipse Ti2 with an EMCCD camera (Andor Technologies) using a 40X 0.60NA Plan Fluor air objective or a 20× 0.75 NA Plan Apo air objective and MicroManager software package (Edelstein et al., 2014). The microscope was surrounded by a box type incubator (Haison) maintained at 37°C and 5% CO₂. Images were acquired every 10 min for approximately 8 h. Subsequently, at each time interval we measured the 2D deformation of the substrate at each point using an image correlation technique similar to particle image velocimetry. We calculated the local deformation vector by performing image correlation between each image and an undeformed reference image which we acquired by adding 10% SDS at the end of each recording to detach the cells from the hydrogels. We used interrogation windows of 32 × 8 pixels (window size x overlap). Calculations of the two-dimensional traction stresses that cell monolayers exert on the hydrogel are described elsewhere (Bastounis et al., 2014; Lamason et al., 2016).

Atomic Force Microscopy (AFM) for determination of cell stiffness

A Bioscope Resolve AFM (Bruker, Santa Barbara) was used for MDCK cell stiffness measurements. The AFM operates with the Nanoscope 9.4 software and the data was analyzed using Nanoscope Analysis 1.9 software. The AFM was integrated with an inverted optical microscope (Zeiss AXIO Observer Z1) to allow for correlation with fluorescence and light microscopy. Fluorescence microscopy was used to position the AFM tip over the MDCK cells that resided on glass substrates and were immersed in PBS and was critical for assessing which cells have internalized fluorescent bacteria.

First, we measured cell stiffness using the PFQNM-LC-A-CAL probe (Bruker) that has a conical shape (15 degrees half-cone angle) terminated with 130 nm diameter sphere, and a spring constant 0.096 nN nm⁻¹ (pre-calibrated by manufacturer using Laser Doppler Vibrometer). We performed force-volume mapping using a ramp size of 4 μm, windows of 30 μm x 30 μm, 64 samples per line x 64 lines, ramp rate of 10 Hz, and a trigger force threshold of 600 pN. Since the indentation depth of >0.5 μm far exceeded the diameter of the spherical tip apex (130 nm), data were analyzed using the Sneddon Model on the extend curves (no adhesion). A total of 5 different windows were recorded for each condition.

We also probed cellular stiffness using colloidal probes with 5 μm diameter spheres on a pre-calibrated silicon nitride cantilever purchased from Novascan (PT.Si02.SN.5.CAL). Spring constant was 0.16 Nm⁻¹ (pre-calibrated by manufacturer). Force-distance spectroscopy measurements were conducted with a ramp rate 0.2 Hz, trigger force threshold 1 nN, and ramp size 6 μm (to accommodate for increased adhesion due to larger contact area). At each field of view, we selected 4 cells and collected 10 measurements per cell. A total of 4 different fields of view were selected for each condition, leading to 16 cells examined per condition. Data were analyzed using the Hertz Model and the extend curves (no adhesion). Each AFM tip was used only for one experiment and then discarded.

RNA isolation and RNA sequencing

Regarding sample preparation MDCK cells were cultured in high glucose DMEM medium (Thermofisher, 11965092) containing 4.5 g/L glucose and supplemented with 10% FBS (GemBio, 900–108). To generate confluent cell monolayers, 24-well plates glass-bottom for microscopy were coated with 50 μg/ml rat-tail collagen-I (diluted in 0.2 N acetic acid) for 1 h at 37°C, air-dried for 15 min, and UV-sterilized for 30 min in a BSC. MDCK cells were seeded at a density of 2 × 10⁵ cells/well for 24 h. 24 h post-seeding MDCK cells were exposed to L.m. (MOI = 200) for 30 min. After washing out the bacteria three times with PBS, media containing 20 μg/mL gentamycin was added to cells to kill extracellular bacteria. 24 h post-infection cells were trypsinized and either left in tubes or sorted into infected and uninfected populations. All tubes containing cells for RNA extraction were treated in parallel (4 replicates per condition). Cells in solution were spun down and lyzed using the QIAshredder Kit (Qiagen; 79656). mRNA was harvested using the RNeasy Plus Micro Kit (Qiagen; 74004) and eluted in 30 μL RNAase free water. A NanoDrop ND-1000 spectrophotometer was used to determine concentration (abs 260) and purity (abs260/abs230) of total RNA samples. Agarose gel electrophoresis was used to check the integrality of total RNA samples (performed by Arraystar Inc.).

For library preparation and RNA sequencing the procedure described below was followed. The total RNA was depleted of rRNAs by Arraystar rRNA Removal Kit. We used Illumina kits for the RNA-seq library preparation, which included procedures of RNA fragmentation, random hexamer primed first strand cDNA synthesis, dUTP based second strand cDNA synthesis, end-repairing, A-tailing, adaptor ligation and library PCR amplification. Finally, the prepared RNA-seq libraries were qualified using Agilent 2100 Bioanalyzer and quantified by qPCR absolute quantification method. The sequencing was performed using Illumina NovaSeq 6000 using the High-Output Kit with 2×150 read length. We had an average of 40 million reads per sample. The DNA fragments in well mixed libraries were denatured with 0.1M NaOH to generate single-stranded DNA molecules, loaded onto channels of the flow cell at 8 pM concentration, and amplified in situ using TruSeq SR Cluster Kit v3-cBot-HS (#GD-401–3001, Illumina). Sequencing was carried out using the Illumina X- ten/NovaSeq according to the manufacturer’s instructions. Sequencing was carried out by running 150 cycles.

Raw sequencing data generated from Illumina X-ten/NovaSeq that pass the Illumina chastity filter were used for following analysis. Trimmed reads (trimmed 5’, 3’-adaptor bases) were aligned to the reference genome (CanFam3). Based on alignment statistical analysis (mapp.i.ng ratio, rRNA/mtRNA content, fragment sequence bias), we determined whether the results can be used for subsequent data analysis. To examine the sequencing quality, the quality score plot of each sample was plotted. Quality score Q is logarithmically related to the base calling error probability (P): Q = −10log10(P). For example, Q30 means the incorrect base calling probability to be 0.001 or 99.9% base calling accuracy. After quality control, the fragments were 5’, 3’-adaptor trimmed and filtered ≤ 20 bp reads with Cutadapt software. The trimmed reads were aligned to the reference genome with Hisat 2 software. In a typical experiment, it is possible to align 40 ∼ 90% of the fragments to the reference genome. However, this percentage depends on multiple factors, including sample quality, library quality and sequencing quality. Sequencing reads are classified into the following classes: (1) Mapped : reads aligned to the reference genome (including mRNA, pre-mRNA, poly-A tailed lncRNA and pri-miRNA); (2) mtRNA and rRNA: fragments aligned to rRNA, mtRNA; and (3) Unmapped: Reads that are not aligned.

Differentially expressed genes and differentially expressed transcripts are calculated. The novel genes and transcripts are also predicted. The expression level (FPKM value) of known genes and transcripts were calculated using ballgown through the transcript abundances estimated with StringTie. The number of identified genes and transcripts per group was calculated based on the mean of FPKM in group ≥ 0.5. Fragments per kilobase of transcript per million mapped reads (FPKM) is calculated with the formula:$\mathit{F}\mathit{P}\mathit{M}\mathit{K}=\frac{\mathit{C}\times {\mathbf{10}}^{\mathbf{6}}}{\mathit{L}\times \mathit{N}}$, where C is the number of fragments that map to a certain gene/ transcript, L is the length of the gene/transcript in Kb and N is the fragments number that map to all genes/transcripts. Differentially expressed gene and transcript analyses were performed with R package ballgown. Fold change (cutoff 1.5), p-value (≤ 0.05) and FPKM (≥ 0.5 mean in one group) were used for filtering differentially expressed genes and transcripts.

Principal component analysis (PCA) and mRNA function enrichment analysis

Principal Component Analysis (PCA) and Hierarchical Clustering scatter plots and volcano plots were calculated for the differentially expressed genes in R or Python environment for statistical computing and graphics. PCA was performed using the plotPCA function in R with genes that have the ANOVA p value ≤ 0.05 on FPKM abundance estimations (Not available for samples with no replicates). Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analyses of the whole data set of DEG were performed using the R package GAGE “Generally Acceptable Gene set Enrichment” (GAGE v.2.22.0) package implemented in R. The analysis allowed us to determine whether the differentially expressed mRNAs are enriched in certain biological pathways. The p-values calculated by Fisher’s exact test are used to estimate the statistical significance of the enrichment of the pathways between the two groups. The R package “Pathview” v.1.12.0 and KEGGGraph v1.30.0 were used to visualize gene set expression data in the context of functional pathways.

Immunofluorescence microscopy

Uninfected or L.m.-infected MDCK cells residing on glass substrates were incubated with 1 μg/mL Hoechst to stain the cells’ nuclei for 10 min at 37 degrees prior to fixation. The following steps were carried out at room temperature. Cells were washed with PBS and fixed with 4% EM grade formaldehyde in PBS for 10 min. Following a wash with PBS, samples were permeabilized for 5 min in 0.2% Triton X-100 in PBS and washed again with PBS. Samples were then blocked for 30 min with 5% BSA in PBS and then incubated with primary antibodies (anti-vimentin: Abcam, ab8069; anti-pan cytokeratin: ThermoFisher, 53–9003-80; anti-tubulin: Abcam, ab52866; anti-NF-κΒ: Abcam, ab190589; anti-Ε-cadherin: Cell Signaling, 3195; anti-ZO-1: Thermofisher, ZO1–1A12; anti-R.p., I7205 (gift from Dr Ted Hackstadt, NIH Rocky Mountain Laboratories) diluted 1:100 in PBS containing 2% BSA for 1 h for all cases other than anti-R.p. where dilution was 1:1000. Samples were washed in PBS three times and then incubated with the appropriate secondary fluorescent antibodies (Invitrogen) diluted 1:250 in PBS containing 2% BSA for 1 h. For actin staining we used 0.2 μM AlexaFluor488 phalloidin (Thermo Fisher, A12379). Samples were washed three times in PBS and stored in 1 mL PBS for imaging. N > 8 mounds were analyzed per condition (N> 1200 cells). For epifluorescence imaging, we used an inverted Nikon Eclipse Ti2 with an EMCCD camera (Andor Technologies) and a 40× 0.60NA Plan Fluor air or a 20× 0.75NA Plan Apo air objective and MicroManager software. For confocal imaging we used a Yokogawa W1 Spinning Disk Confocal with Borealis upgrade on a Leica DMi6 inverted microscope with a 50um Disk pattern, a 60× 1.4NA Plan Apo oil objective and MicroManager software.

Multiplexed magnetic Luminex cytokine immunoassay and data analysis

Samples were centrifuged at 8000 x g to remove debris and assayed immediately. The remaining samples were stored at −80°C. All steps were performed at room temperature. 25 μL of each standard or control was added into the appropriate wells. 25 μL of assay buffer was added for standard 0 (background). 25 μL of assay buffer was added to the sample wells. 25 μL of tissue culture media solution was added to the background, standards, and control wells. 25 μL of sample supernatant was added into the appropriate wells. The beads were vortexed briefly, and 25 μL of the mixed beads was added to each well. We used the Canine Cytokine Magnetic Bead Panel 96-Well Plate Assay (Millipore Sigma, # CCYTMG-90K-PX13). The plate was sealed, wrapped with foil and incubated with agitation on a plate shaker over for 2 h. Following that, the well contents were gently removed, and the plate was washed two times by manually adding and removing 200 μL wash buffer per well. Then, 25 μL of detection antibodies were added into each well. The plate was again sealed and covered with foil and incubated with agitation for 1 h. After incubation, 25 μL Streptavidin-Phycoerythrin was added to each well containing the detection antibodies. (Plate was not washed after incubation with detection antibodies as per manufacturer’s protocol). The plate was sealed and covered with foil and incubated with agitation for 30 min. Then, well contents were gently aspirated out, and the plate was washed two times. Finally, 150 μL of sheath fluid was added to all wells, and the beads were resuspended on a plate shaker for 5 min at room temperature. The plate was run on Bio-Plex 200 flow cytometer. Data analysis was performed using the Bio-Plex Manager 5.0. We used a 5-parameter log fit curve against the signal intensities for the analytes present in the Canine Cytokine Magnetic bead Panel (GM-CSF, IFN-γ, IL-2, IL-6, IL-7, IL-8, IL-15, IP-10, KC-like, IL-10, IL-18, MCP-1, TNF-A) to generate a standard curve for the assay. Sample signal intensities were then back interpolated to the standard curve to determine the analyte concentrations. The upper limit of quantitation (ULOQ) was determined as the highest point at which the interpolated data of the standard curve are within 25% (±) of the expected recovery. ULOQ represents the highest concentration of an analyte that can be accurately quantified. Similarly, the lower limit of quantitation (LLOQ) was determined as the lowest point at which the interpolated data of the standard curve are within 25% (±) of the expected recovery. LLOQ represents the lowest concentration of an analyte that can be accurately quantified.

Western blotting

To assess phosphorylation of the p65 subunit of NF-κΒ and expression levels of ΝF-κΒ, MDCK cells were seeded at a concentration of 2 × 10⁵ cells/well (24-well plates) for 24 h and then infected or not for 24 h with intracellular L.m. at a MOI~200 bacteria/cell. Cells in infected wells were treated with vehicle control or the NEMO binding peptide (50 μM) starting 4 h p.i.. Cells were then lysed with a buffer containing 0.5M Tris-HCl, pH 7.4, 1.5M NaCl, 2.5% deoxycholic acid, 10% NP-40, 10mM EDTA and a protease inhibitor mixture (phenylmethylsulfonyl fluoride [PMSF], leupeptin, aprotinin, and sodium orthovanadate). The total cell lysate was separated by SDS–PAGE (10% running, 4% stacking) and transferred onto a nitrocellulose membrane (Immobilon P, 0.45-μm pore size). The membrane was then incubated with the designated antibodies. Immunodetection was performed using the Western-Light chemiluminescent detection system (Applied Biosystems). Membrane blots from 76 kDa to 52 kDa were imaged for phospho-p65 and p65 (65 kDa) and 52 kDa to 33 kDa were imaged for the corresponding β-actin (42 kDa) (Figure S6F).

Measurement of reactive oxygen species (ROS) production

The DCFDA cellular ROS detection assay kit (Abcam; ab113851) was used to measure ROS production. DCFDA is a cell permeable fluorogenic dye that measures hydroxyl, peroxyl and other ROS activity within cells. Briefly, MDCK cells in monolayer were infected or not with L.m. (JAT607, ActAp::mTagRFP) at low MOI. At 24 h p.i. cells were collected in tubes and then stained with 20 μM of DCFDA for 30 min at 37°C, as per manufacturer’s instructions. The dye was then washed out and the fluorescence signal per cell was measured via flow cytometry. Positive controls were treated with 50 μM Tert-Butyl Hydrogen Peroxide (TBHP) for 1 h prior to conduction of measurements.

Nuclear segmentation, tracking and characterization of dynamics of motion

We acquired time-lapse fluorescence confocal images of Hoechst-stained host epithelial cell nuclei and of the L.m. fluorescence using a spinning disk confocal with a 63× 1.2NA Plan Apo water objective at 30 min intervals (N=5 independent experiments for each condition). For each time point, z stacks including the total height of the cell layer was also acquired with a z spacing of 0.7 μm. To segment and track the host cell nuclei in 3D we used using IMARIS software (Bitplane) by first using the spot detection (segmentation) and then the tracking modules. The x, y and z positions of all identified objects and tracks were exported for import into MATLAB (MathWorks) for further analysis was performed to calculate cell displacements and speeds of migration. For each experimental recording originating from an uninfected well or an infected well (centered around an infection mound) we followed the tracks of n=271+/−65 tracks/experiment. The violin plots shown in Figure 1 show the median and the dotted lines the 25 and 75% quartiles of all tracks considered. For each experiment (N=5) the average value over the entire field of view and over time is shown as a triangle. For mean square displacement analysis (MSD) we used the @msdanalyzer MATLAB class (Tarantino et al., 2014). We first computed for each cell its MSD as a function of time interval, Δt: MSD(Δt)= 〈|r_i(t + Δt)−r_i(t)|²〉, where r_i (t) indicates the position of cell i at time t and 〈…〉 denotes the average over all time t. We then calculated the weighted average of all MSD curves, where weights are taken to be the number of averaged delay in individual curves. In the average MSD plots the grayed area represents the weighted standard deviation over all MSD curves and the errorbars the standard error of the mean. We estimated the self-diffusion coefficient D_s = lim_Δt→∞MSD(Δt)/(4Δt) by linear weighted fir of the mean MSD curve on the first 500 min. In the plots shown in Figure 1F–G the grayed area represents the weighted standard deviation over all MSD curves. The error bars show the standard error of the mean which is approximated as the weighted standard deviation divided by the square root of the number of degrees of freedom in the weighted mean. The red line represents the weighted mean MSD curve from which the diffusion coefficient D is calculated. Finally, for analysis of the motion type of the objects (i.e. diffusive, subdiffusive or superdiffusive) we performed log-log fitting and modeled the MSD curve by the following power law MSD(Δt) = Γ × t^α. For purely diffusive motion α=1, for subdiffusive α<1 and for superdiffusive α>1. To determine the power law coefficient we take the logarithm of the power law so that it turns linear: log(MSD) = α × log(t) + log(Γ ) and by fitting log(MSD) versus log(t) we retrieve α. For more details please refer to @msdanalyzer MATLAB class (Tarantino et al., 2014).

2D tracking of L.m.-infected wild-type or αEcat KO MDCK cells was performed using epifluorescence microscopy following the movement of Hoechst-stained host cell nuclei between 6 to 16 h p.i. 2D nuclear segmentation and tracking was performed using IMARIS software (Bitplane). The x, y and z positions of all identified objects and tracks were exported for import into MATLAB (MathWorks) for further analysis of cellular speed, net/total distance travelled and coordination scores as described previously (Hayer et al., 2016). Regarding coordination score calculation we considered a radius of 65 μm around each given cell for the calculation of the pairwise velocity correlations (Hayer et al., 2016).

The correlation length shown in Figure S4J is calculated as previously documented (Tambe et al., 2011) and it is defined as in the next equation:

$$C(R)={\langle \frac{\sum _{j}d\left(r_j\right)\cdot d\left(r_j+R\right)}{\sum _{j}d\left(r_j\right)\cdot d\left(r_j\right)}\rangle }_{t,\phi }$$

where C is the correlation length for all vectors separated by a distance R, d(r_j) is the modified displacement vector for a given position r_j and the angle brackets signify an average over all directions and time. To avoid the dependence of the correlation value with the absolute value of the displacement, we substrate the mean displacement of each frame $(\overline{x})$ to the displacement vector $\left(x\left(r_j\right)\right),d\left(r_j\right)=x\left(r_j\right)-\overline{x}$. We fit the correlation over the radius using the function C(R) = exp(−R/R₀) and determine the correlation length as R₀.

Cell segmentation and extraction of cell shape characteristics

Epithelial cells were segmented based on pericellular E-cadherin localization using the Tissue Analyzer ImageJ plugin (Aigouy et al., 2016). Masks of individual cells were exported and imported in MATLAB for further analysis. We used custom-made code for calculating the cell area, radial alignment angle θ, aspect ratio and shape parameter q $(\mathit{perimeter}/\sqrt{\mathit{area}})$. Radial alignment angle θ is defined as the angle between the radial direction (from the center of the field of view for uninfected cells or the center of the mound for L.m.-infected wells) and the direction of the major axis of the cell (Figure 2A). If the angle is close to zero, cells are oriented radially towards the center of the image (or the mound). If the orientation is close to 90° the orientation is circumferential. To quantify whether there is any ordering in the orientation of cells, we also calculated for each cell how many of its 100 nearest neighbors share similar orientation angles (Δθ=+/−10°).

Characterization of mound volume and infection focus area

We acquired confocal images of Hoechst-stained host epithelial cell nuclei and of the fluorescent bacteria in fixed samples using a spinning disk confocal with a 63× 1.2NA Plan Apo water objective. The z spacing used was 0.2 μm for MDCK cells and 0.7 μm for A431D cells, since the field of view imaged for the latter was approximately 9-fold larger. The fields of view were selected so that the infection mounds are approximately at the center of field of view. To quantify the volume of the cells extruded from the monolayer, we first cropped out the lower z slices of each confocal image to exclude cells within the monolayer. We then segmented the channel containing host cell nuclei to form binary images of each mound. An alpha radius of 50 was used to draw a tight boundary around the segmented nuclei using the MATLAB (MathWorks) function alphaShape (Figure S1A–C). The area enclosed within this alpha shape was calculated for each z slice. Lastly, we added up the area of each z slice, converted the area into μm² and multiplied by the increment between each z slice (0.2 μm for MDCK cells and 0.7 μm for A431D cells) to obtain the volume of the mound. An additional step was included in the analysis of αEcat KO MDCK cells and all mounds resulting from infection with R.p. As we observed extrusion of uninfected cells in these cases, we binarized only infected nuclei using input from the bacterial channel. To do so, bacteria were segmented and then enlarged using the imdilate function in MATLAB (MathWorks). This image was applied as a mask to the nuclei channel and only fluorescence within the mask was considered in the analysis. For each different experiment all values were normalized relative to the corresponding values for infected WT MDCK cells treated with vehicle control. Unless indicated three independent experiments were considered and N=13–19 infection mound volumes were calculated in total.

There is no standard metric in the field to measure the efficiency of L.m. cell-cell spread through a monolayer of host cells. To characterize the area of the infection focus we chose to draw a convex hull, the smallest convex polygon that encompasses a set of points, around the bacteria (Bastounis et al., 2018). This is a computationally inexpensive and consistent way to measure the efficiency of L.m. spread.

Live/dead staining, TUNEL and BrdU assays

We used the NucRed Dead 647 ReadyProbes Reagent (Thermofisher, R37113), a cell impermeant stain that emits bright far-red fluorescence when bound to DNA, for staining dead cells. This reagent stains the cells without plasma membrane integrity and was used to measure cell viability. Briefly, one drop of this reagent was added in each well containing live cells and cells were incubated with this reagent for 20 min before samples were imaged. The live-dead stain (Rincón et al., 2018) confirmed that many of the extruded cells retain the stain (Figure S6E).

In addition, we used the TUNEL assay for DNA fragmentation (Abcam, ab66108) as a means to detect number of apoptotic cells. For staining cells for DNA fragmentation, we followed the manufacturer’s instructions and quantified fluorescein-labeled DNA by flow cytometry for cells at 24 h post-infection. The TUNEL assay performed to cells from uninfected and L.m.-infected wells showed that the latter exhibited a 2-fold increase in TUNEL-positive cells (Figure S6D). We also used the BrdU staining kit for flow cytometry APC (eBioscience; 8817–6600-42) to identify proliferating cells (Crane and Bhattacharya, 2013). However, BrdU can also stain apoptotic cells due to the break-up of their genomic DNA by cellular nucleases (Darzynkiewicz and Zhao, 2011). For the BrdU assay we followed the manufacturer’s instructions and incubated cells at 24 h post-infection with 10 μm BrdU for 4 h at 37°C before harvesting the cells and proceeding with the above protocol. We found that only 1% of cells from uninfected wells were BrdU-positive, as compared to 0.5% of surrounders and 8% of mounders (Figure S6B). Interestingly, mounders that appeared BrdU-positive coincided with the infected cells having the highest bacterial load (Figure S6C), suggesting that labeling under these conditions reflects apoptosis rather than proliferation. Indeed, when we measured the number of cells collectively in uninfected versus infected wells, we found no differences in the number of cells between the two conditions, suggesting that hyperproliferation in infection is probably not occurring (Figure S6A).

Laser ablation experiments

Laser ablation experiments were performed on L.m.-infected MDCK cells at 24 h p.i. using the Firefly system (UGA-42 Firefly; Rapp OptoElectronic) as per manufacturer’s instructions. The region of ablation was approximately 816 pixels in length and a 405 nm laser traversed this region 270 times for a time period of 1 min passing through the entire height of the mound. During this time (immediately before and after), we acquired confocal images of RFP-E-cadherin and the fluorescent bacteria using a spinning disk confocal with a 60× 1.4NA Plan Apo oil objective to determine relaxation of the system and subsequent wound healing. Images were every 1 min immediately before and after laser ablation and with a z spacing of 0.7 μm. The E-cadherin channel was imaged using the 561 nm laser at full power with an exposure of 300 ms, the bacteria were imaged using the 488 nm laser at full power with an exposure of 20 ms, and the brightfield image was taken with an exposure of 50 ms. The cells were imaged every minute for up to 1.5 h post-ablation.

To analyze the displacements of cells upon laser ablation (recoil or relaxation) and their corresponding wound healing response, we used the maximum intensity projections of E-cadherin images of cells and compared subsequent frames starting before laser ablation and following cells every 1 min up to 1.5 h post-laser ablation using particle image velocimetry (PIV)-like technique (Bastounis et al., 2014). For PIV, we used windows of 36 pixels with a 50% overlap. Using the component of the velocity that is perpendicular to the laser ablated segment (u_y) we then calculated for each instant of time the mean u_y along the x-axis. Lining these means along the x-axis over time, we obtained kymographs of mean displacements above and below the laser ablated line segment. In these kymographs time, t (min) corresponds to the horizontal axis and position in the y-axis corresponds to the vertical axis (Figure 5C). This representation allows for a detailed quantitative analysis of the recoil behavior of cells upon laser ablation (magnitude of recoil, spatial location and temporal evolution). To quantitatively characterize recoil displacement based on the behavior of cells in response to laser ablation for data originating from independent experiments, for each experiment we calculated the mean u_y along the x-axis but also along the y-axis considering in this case just the first 100 μm away from the ablated region.

Computational modeling of cell monolayers during infection

We assume the cells are arranged in the monolayer as regular hexagons with side length of 7 μm, the thickness of the monolayer is set to 7 μm in the undeformed state following experimental observations and previous computational works (Escribano et al., 2019; O’Dea and King, 2012; Schmedt et al., 2012). We model the mechanical behaviour of the cell distinguishing two main parts: active and passive. The active part represents the actomyosin contractility motors and the passive one the rest of the cytoskeleton (Borau et al., 2011; Moreo et al., 2008). In this way, both parts are assumed linear elastic material, where the total stiffness of each cell is the sum of both (E_cell = E_active + E_passive. Following the experimental AFM measurements, the elastic modulus of uninfected cells is set in 1000 Pa and 250 Pa for the infected cells (E_cell). As a first approach, we assume both parts of the cells are working in parallel, therefore, the active (E_active) and passive (E_passive;) elastic modulus are 500 Pa for uninfected cells and 125 Pa for infected cells. We assume the passive part of the cell nearly incompressible, thus the Poisson ratio is set to 0.48 (Moreo et al., 2008). The active part of the cell is mainly actomyosin contraction and we assume this contraction is not isotropic, it just occurs in the plane of the monolayer. Thus, the Poisson ratio of the active part is assumed zero to uncouple effects in the monolayer plane and in the vertical direction for the active part of the cell.

We model the substrate as a linear elastic material with elastic modulus of 3 kPa corresponding to collagen I-coated polyacrylamide hydrogels used in the TFM experiments. We consider in the model both cell-cell and cell-matrix interactions. We model cell-cell interactions as a continuum and adopt a linear elastic model with elastic modulus of 1000 Pa and shear modulus of 500 Pa. When we simulate the inhibition of cell-cell adhesions we set these values close to zero, thus cells do not interact anymore mechanically with each other. Meanwhile, the cell-ECM interactions are simulated as cohesive contacts. We assume that cells adhere in different ways to the substrate (Escribano et al., 2018; Sunyer et al., 2016) depending on whether there are cell-cell interactions or not. If the cell-cell adhesions are active (cell monolayer behaves collectively), its adhesion to the substrate is weaker than if the cell-cell adhesions are inhibited (cells forming the monolayer behave individually). The stiffness of the stiffer adhesion is assumed to be 1000 nN/μm³ per area of contact both in the normal and shear direction, and the weak one 10 nN/μm³ in the normal direction and negligible in the shear direction. The total area of contact of each cell is 31.61 μm² (6 zones of 5.268 μm²), thus, the total stiffness of the total adhesion of each cell is 31608 nN/μm and 31.608 nN/μm for the high or low adhesion, respectively. We analyse a short period of time, in which just one contraction and, if it occurs, one protrusion event are simulated. These events could be repeated in time and as a result infection mounding would be more prominent.

To simulate cell ablation, we assume the cell loses its active part just considering the passive part of the cell (Nikolaev et al., 2014; Vileno et al., 2007). Therefore, cell stiffness is reduced and, moreover, ablated cells cannot contract anymore. Cell-cell junctions weaken due to ablation, thus the stiffness of cell-cell junctions in the ablated area is set close to zero during cell contraction and the cell-ECM adhesions are removed (stretching-based interactions). However, during cell protrusion cells are in contact with the ablated area (compression-based contact).

To simulate how host cells behave when certain cells are able to form cell-cell adhesions while others not, and to better understand in which scenarios an infection mound will be formed, we consider two cases. In the first one, all the infected cells and their immediate surrounders are able to form cell-cell adhesions, but all the rest uninfected cells are not. In the second one, the infected cells and their immediate surrounders are not able to form cell-cell adhesions, but uninfected cells far from the infection mound are able to. As in previous simulations, when one edge of the cell cannot adhere to another cell’s edge, the cell-ECM interaction forces increase in these edges of the cell. Thus, a cell without cell-cell adhesions in any direction would be an isolated cell with high cell-ECM traction forces in all directions.

We implement the model into a commercial finite element model (FE-based) ABAQUS⁷⁰. To simulate cells two meshes are overlapping sharing the nodes to represent the active and passive components of the cell. We discretise the model with tetrahedral and hexahedral linear elements of average size 2 μm.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical parameters and significance are reported in the Figures and the Figure Legends. Data are determined to be statistically significant when p < 0.05 or p<0.01 by an unpaired Student’s T-Test (USTT), or Wilcoxon Rank Sum Test (WRST), where indicated. As such, asterisk denotes statistical significance as compared to indicated controls. For statistical analysis of the kinematics and shape morphometrics of large number of cells characterized in each experiment, we used violin plots considering all nuclei tracked (Figures 1F, 1I, 1J, S4C and S4J) or all cells characterized (Figures 2D and S2C–D). In the violin plots dashed line represents the median of the distribution and dotted lines the 25 and 75% quartiles considering data from all experiments. For each independent experiment (replicate) the mean value was also calculated and indicated as a colored inverted triangle. Those means where then used to calculate their average (horizontal bar) and standard deviation (vertical bars) and their p-value using an USTT (Lord et al., 2020). For the rest of the data represented using barplots (e.g. relative infection mound volumes) midlines denote mean value and whiskers the standard deviation (vertical bars). P-value was calculated using the non-parametric WRST and asterisk denotes p < 0.05. Statistical analysis was performed in GraphPad PRISM 8.

Supplementary Material

1

Movie S1. Time-lapse movie showing orthogonal views of MDCK cell nuclei originating from an uninfected or a L.m.-infected well over the course of 20 h. Related to Figure 1.

On the left orthogonal views of Hoechst-stained uninfected MDCK nuclei in a confluent monolayer over the course of 20 h are shown, whereas on the right the corresponding views of Hoechst- stained L.m.-infected MDCK host cell nuclei over the same time frame are shown (host nuclei: yellow, L.m.: black). Acquisition started 24 h p.i. (frame interval: 30 min). Scale bar and corresponding time p.i. are indicated. One image from each different plane (i.e. x-y, x-z, and y-z planes) is shown.

2

Movie S2. Traction force microscopy on WT MDCK cells infected with low multiplicity of L.m. Related to Figure 3.

Upper left panel shows the phase contrast image of host MDCK cells. Scale bar and corresponding time p.i. are indicated. Upper middle panel shows the corresponding L.m. fluorescence. The infection focus is centered in the middle of the field of view chosen. Upper right panel shows the cell-matrix deformations that cells are producing onto their matrix (color indicates deformation magnitude in μm). Bottom left panel shows the radial deformation, u_r maps (positive deformations values in μm indicate deformations pointing away from the center of the focus, outwards) and bottom middle image the overlap of u_r maps and L.m. fluorescence. Bottom right panel shows the traction stresses (color indicates stress magnitude in Pa) exerted by the MDCK host cells that are adhering onto soft 3 kPa hydrogels.

3

Movie S3. Traction force microscopy on αEcat KO MDCK cells infected with low multiplicity of L.m. Related to Figure 4.

Similar to Movie S2 but for αEcat KO MDCK cells.

4

Movie S4. Stress distribution during cell contraction and mound formation while uninfected surrounder cells protrude. Related to Figure 5.

Simulations showing the stress distribution (mPa) during cell contraction and mound formation (μm) when uninfected cells (surrounders) protrude. Upper left image shows the case where all cells are healthy. Upper right image shows the case where seven cells are infected, and their cell-matrix contractility decreases while uninfected cells in contact with infected ones can create new cell-ECM adhesions. Bottom left image shows the case where seven cells are infected, and their passive stiffness decreases while uninfected cells in contact with infected ones create new cell-ECM adhesions. Finally, the bottom right image shows the case where seven cells are infected, and their passive stiffness decreases but all cell-cell adhesions of both infected and uninfected cells are disrupted.

5

Table S1. Differentially expressed genes and KEGG pathways between uninfected cells, L.m.-infected mounders and uninfected surrounders. Related to Figure 6.

Sheets 1-3 of the table present the DEGs identified when comparing the transcriptome of uninfected cells (UU) to surrounders (UI) (1^st sheet), uninfected cells (UU) to infected mounders (II) (2^nd sheet) and uninfected surrounders (UI) to infected mounders (II) (3^rd sheet). Rows correspond to the genes identified and columns A-K correspond to different parameters explained within the table. Columns L-S provide the FPMK of genes in samples for each sample from the corresponding groups compared. Sheets 4-6 of the table show KEGG pathways significantly perturbed when comparing uninfected cells, infected mounders and uninfected surrounders. The gage package was used for pathway analysis which has precompiled databases for mapping genes to KEGG pathways so to perform pathway enrichment analysis of DEGs between all three populations. Table shows comparisons performed between uninfected cells (UU) to surrounders (UI) (4^th sheet), uninfected cells (UU) to infected mounders (II) (5^th sheet) and uninfected surrounders (UI) to infected mounders (II) (5th sheet). KEGG pathways upregulated for each group are indicated. Rows show the KEGG pathways and columns different parameters explained in the table including upregulated genes of each category.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Rabbit anti-Rickettsia I7205	Ted Hackstadt (NIH, RML)	N/A
Mouse anti-vimentin	Abcam	Cat# ab8069; RRID: AB_306239
Mouse anti-pan cytokeratin	Thermo Fisher Scientific	Cat# 53-9003-80; RRID: AB_1834351
Rabbit anti-tubulin	Abcam	Cat# ab52866; RRID:
Rabbit anti-E-cadherin	Cell Signaling	Cat# 3195; RRID: AB_869989
Mouse anti-ZO-1	Thermo Fisher Scientific	Cat# ZO1-1A12; RRID: AB_2533147
Rabbit anti-NF-κB	Abcam	Cat# ab190589; RRID: AB_2728800
Rabbit NF-κB p65 (D14E12)	Cell Signaling	Cat# 8242; RRID: AB_10859369
Rabbit phospho-NF-κB p65 (Ser536) (93H1)	Cell Signaling	Cat# 3033; RRID: AB_331284
Rabbit IκBα	Cell Signaling	Cat# 9242; RRID: N/A
Rabbit phospho-IκBα (Ser32) (14D4)	Cell Signaling	Cat# 2859; RRID: AB_AB_561111
Rabbit GAPDH (14C10) Rabbit mAb	Cell Signaling	Cat# 2118; RRID: AB_561053
Biological Samples
Human ileal enteroids (HIE5)	J. Smith (Holly and Smith, 2018)	N/A
Chemicals, Peptides, and Recombinant Proteins
AlexaFluor488 phalloidin	Thermo Fisher Scientific	Cat# A12379
SulfoSanpah	Thermo Fisher Scientific	Cat# 22589
Collagen I	Sigma	Cat# C3867
Para-nitroblebbistatin	Fisher	Cat# NC1706059
Y-27632	Sigma	Cat# 129830-38-2
Z-VAD-FMK	MilliporeSigma	Cat# 5.30389.0001
GSK’872	Fisher	Cat# 64-921-0
H1152	Tocris	Cat# 2414
Diphenyleneiodonium chloride (DPI)	Sigma	Cat# D2926
L-NIL dihydrochloride (L-NIL)	SCBT	Cat# 159190-45-1
NEMO binding peptide	Fisher	Cat# 480025
Critical Commercial Assays
QIAshredder Kit	Qiagen	Cat# 79656
RNeasy Plus Micro Kit	Qiagen	Cat# 74004
Canine Cytokine Magnetic Bead Panel 96-Well Plate Assay	Millipore Sigma	Cat# CCYTMG-90K-PX13
DCFDA cellular ROS detection assay kit	Abcam	Cat# ab113851
Deposited Data
RNA-seq data	This paper	GSE140626
Experimental Models: Cell Lines
MDCK II	A. Bakardjiev (Faralla et al., 2018)	N/A
α-catenin knockout MDCK II	W. Nelson (Ortega et al., 2017)	N/A
MDCK II E-cadherin-RFP	W. Nelson (Perez et al., 2008)	N/A
A431D	C. Gottardi (McEwen et al., 2014)	N/A
A431D expressing full length E-cadherin	C. Gottardi (McEwen et al., 2014)	N/A
Experimental Models: Organisms/Strains
L. monocytogenes: L.m.-ActAp::mTagRFP	F. Ortega (Ortega et al., 2017)	N/A
L. monocytogenes: L.m.-GFP	F. Ortega (Ortega et al., 2017)	N/A
L. monocytogenes: L.m.-ΔinlC ActAp::mTagRFP	F. Ortega (Ortega et al., 2017)	N/A
L. monocytogenes: L.m.-ΔinlP	A. Bakardjiev (Faralla et al., 2018)	N/A
L. monocytogenes: L.m.-LLOG486D ActAp::mTagRFP	M. Rengarajan (Rengarajan et al., 2016)	N/A
R. parkeri Portsmouth strain	Chris Paddock	N/A
R. parkeri: R.p.- ompBSTOP::tn	P. Engstrom (Engstrom et al., 2019)	N/A
L. monocytogenes: L.m.-Δhly	M. Rengarajan (Rengarajan et al., 2016)	N/A
Oligonucleotides
siRNA targeting sequence for CDH1: Custom CDH1 duplex	GE Healthcare Dharmacon, Inc.	CTM-521910 and CTM-521911
siRNA targeting sequence for SNAi1: Custom SNAI1 duplex	GE Healthcare Dharmacon, Inc.	CTM-544153 and CTM-544153
Forward Primer for CDH1 (forward:5’ AGACCCAGTAACTAACGACGG 3’)	This paper	N/A
Reverse Primer for CDH1 (reverse:5’ ACACCAAAGTCTTCAGGGATT 3’)	This paper	N/A
Forward Primer for SNAI1 (forward:5’ CCCCCATTTGGCTGTGTTG 3’)	This paper	N/A
Reverse Primer for SNAI1 (reverse:5’ ATCAGTCTGTCGGCTTTTATCCT 3’)	This paper	N/A
Forward Primer for β-actin (forward:5’ CCCAGCACAATGAAGATCAAGAT 3’)	This paper	N/A
Reverse Primer for β-actin (reverse:5’ CAAGAAAGGGTGTAACGCAACT 3’)	This paper	N/A
Forward Primer for SNAI2 (forward:5’ GAAGCATTTCAACGCCTCC 3’)	This paper	N/A
Reverse Primer for SNAI2 (reverse:5’ ACTCACTCGCCCCAAGGAT 3’)	This paper	N/A
Forward Primer for NFKBIA (forward:5’ CACATTCCCTAGCCAGAAACATT 3’)	This paper	N/A
Reverse Primer for NFKBIA (reverse:5’ TACACCTGGTTGTCACGCATC 3’)	This paper	N/A
Forward Primer for ICAM1 (forward:5’ ACCGAGGGTTGGGATTGTT 3’)	This paper	N/A
Reverse Primer for ICAM1 (reverse:5’ GTCTCTGGCAGGACAAAGGTT 3’)	This paper	N/A
Forward Primer for CXCL8 (forward:5’ GAAAACTCAGAAATCATTGTAAAGC 3’)	This paper	N/A
Reverse Primer for CXCL8 (reverse:5’ ATCTTGTTTCTCAGCCTTCTTTAG 3’)	This paper	N/A
Software and Algorithms
ImageJ	Schneider et al., 2012	https://imagej.nih.gov/ij/
MicroManager	Open Imaging	https://www.micro-manager.org/
MATLAB	MathWorks	http://www.mathworks.com/products/matlab/?requestedDomain=www.mathworks.com
GraphPad Prism v6	GraphPad	http://www.graphpad.com/scientific-software/prism/
Hisat 2	(Kim et al., 2015)	https://ccb.jhu.edu/software/hisat2/index.shtml)
Cutadapt	(Martin, 2011)	https://cutadapt.readthedocs.io/en/stable/
R package GAGE	(Luo et al., 2009)	https://bioconductor.org/packages/release/bioc/html/gage.html
R package “Pathview”	(Luo et al., 2017)	https://www.bioconductor.org/packages/release/bioc/html/pathview.html
ABAQUS	Dassault systèmes	https://www.3ds.com/products-services/simulia/products/abaqus/
Imaris	Bitplane	https://imaris.oxinst.com/

Highlights.

Epithelial host cells form extruded mounds after infection with L. monocytogenes

3D mounds result from mechanical competition between infected and uninfected cells

Inhibition of cell-matrix and/or cell-cell forces inhibits infection mound formation

Innate immune signals control this mechanical competition that limits pathogen spread