EPEC‐induced activation of the Ca2+ transporter TRPV2 leads to pyroptotic cell death

The enteropathogenic Escherichia coli (EPEC) type III secretion system effector Tir, which mediates intimate bacterial attachment to epithelial cells, also triggers Ca2+ influx followed by LPS entry and caspase‐4‐dependent pyroptosis, which could be antagonized by the effector NleF. Here we reveal the mechanism by which EPEC induces Ca2+ influx. We show that in the intestinal epithelial cell line SNU‐C5, Tir activates the mechano/osmosensitive cation channel TRPV2 which triggers extracellular Ca2+ influx. Tir‐induced Ca2+ influx could be blocked by siRNA silencing of TRPV2, pre‐treatment with the TRPV2 inhibitor SET2 or by growing cells in low osmolality medium. Pharmacological activation of TRPV2 in the absence of Tir failed to initiate caspase‐4‐dependent cell death, confirming the necessity of Tir. Consistent with the model implicating activation on translocation of TRPV2 from the ER to plasma membrane, inhibition of protein trafficking by either brefeldin A or the effector NleA prevented TRPV2 activation and cell death. While infection with EPECΔnleA triggered pyroptotic cell death, this could be prevented by NleF. Taken together this study shows that while integration of Tir into the plasma membrane activates TRPV2, EPEC uses NleA to inhibit TRPV2 trafficking and NleF to inhibit caspase‐4 and pyroptosis.


Intimin-mediated Tir clustering and actin polymerization in
EPEC-infected macrophage lead to activation of caspase-4 and processing of GSDMD, which subsequently activates the canonical NLRP3-Asc-caspase-1 inflammasome (Goddard et al., 2019). Using EPEC wild type (WT), the EPEC-1 derivative, which expresses intimin, the T3SS and Tir as the sole effector and EPEC-0, which expresses intimin and the T3SS but none of the effectors, we showed that intimin-induced Tir clustering in the IEC model cell line SNU-C5 induces Ca 2+ and LPS internalization which leads to caspase-4mediated pyroptosis (Zhong et al., 2020). Cell death, as measured by propidium iodide (PI) uptake, was first seen at 2 hr post-infection and plateaued before 8 hr (Zhong et al., 2020). Tir-induced cell death in SNU-C5 cells was enhanced by pre-treatment with ATP or priming with IFNγ which promotes Ca 2+ -influx or expression of caspase-4 and GSDMD, respectively. In contrast, cell death can be blocked by co-expression in EPEC-1 of the caspase-4-inhibiting effector NleF, silencing of caspase-4 or GSDMD or chelation of extracellular Ca 2+ (Zhong et al., 2020).
The route of Ca 2+ entry following EPEC-1 infection had not been elucidated. Tir-membrane insertion induces localized membrane disruptions (Lin et al., 2011), which could potentially activate mechanosensitive ion channels, such as the TRP family (Liu & Montell, 2015;Startek et al., 2019). Using whole-cell proteomics, we have recently shown that infection with EPEC-1 induces increased abundance of TRPV2 (Zhong et al., 2020). TRPV2 is a non-selective cation channel with a higher affinity to Ca 2+ conductance compared to other ions and can be opened by various stimuli, including heat over 52°C, mechanical stretching (e.g., via osmotic stress), and cannabinoids (Caterina et al., 1999;Muraki et al., 2003;Petrocellis et al., 2011;Zanou et al., 2015). It is predominantly located in intracellular membranes, particularly the ER, in unstimulated cells. However, trafficking of TRPV2 to plasma membrane, preceding TRPV2 channel activation, can be induced by mechanical stress or growth factor stimulation (Kanzaki et al., 1999;Monet et al., 2009;Nagasawa & Kojima, 2015;Nagasawa et al., 2007;Reichhart et al., 2015). In this study, we investigated the role of TRPV2 in Tir-induced Ca 2+ influx and pyroptosis.

| TRPV2 contributes to Tir-induced cell death in SNU-C5 cells
We have recently shown that EPEC triggers caspase-4-and GSDMD-dependent pyroptotic cell death in SNU-C5 cells; the magnitude of cell death was amplified by pre-treatment with IFNγ, as it increases expression of caspase-4 and GSDMD (Zhong et al., 2020). EPEC-1-, but not EPEC-0, triggers activation of caspase-4 following clustering of Tir by intimin in the plasma membrane ( Figure 1a). As this leads to membrane curvature (Lin et al., 2011), we hypothesized that Tir clustering creates membrane stress which could activate mechanosensitive Ca 2+channels, leading to Ca 2+ -influx required for LPS entry that subsequently activates caspase-4. We have recently reported that following EPEC-1 infection of SNU-C5 cells the protein abundance of the mechanosensitive cation channel TRPV2, which has a preference for Ca 2+ , has increased (Zhong et al., 2020). Accordingly, we investigated if TRPV2 plays a role in EPEC-induced pyroptosis in SNU-C5 cells.
First, we used siRNA silencing to investigate the involvement of TRPV2 in EPEC-1 induced cell death. In our previous study, we have shown that EPEC-1 infection induces Ca 2+ influx after around 20 min post-infection, followed by detectable LPS entry at 2 hr post-infection, while cell death peaks at 8 hr post-infection (Zhong et al., 2020). TRPV2 silencing by  Figure S1). Importantly, caspase-4 was not cleaved in EPEC-1-infected TRPV2-silenced SNU-C5 cells, demonstrating that TRPV2 activation lies upstream of caspase-4 activation ( Figure 1e). Moreover, pre-treatment with SET2, an inhibitor selective for TRPV2-mediated Ca 2+ conductance (Chai et al., 2019), also reduced Ca 2+ influx and EPEC-1induced cell death (Figure 1f,g), implicating the channel activity of TRPV2. A combination of SET2 and YVAD, which inhibit TRPV2 and caspase-4, respectively, did not further reduce Tir-induced cell death compared to YVAD treatment alone, confirming that TRPV2 and caspase-4 activation are involved in the same cell death pathway (Figure 1g and Figure S1). However, SET2 did not block cell death to the same magnitude as extracellular Ca 2+ chelation by EGTA, which was higher than cell death seen following infection with EPEC-0 ( Figure 1g and Figure S1), suggesting the involvement of additional Ca 2+ influx routes.

| Lowering osmolality protects cells from EPEC-1-induced cell death
Activation of TRPV2 can be achieved via membrane pressure, for example changing growth medium osmolality (Muraki et al., 2003;Zanou et al., 2015). As integration of Tir into the plasma membrane causes local distortion and membrane stress (Lin et al., 2011), we hypothesized that TRPV2 activation can be regulated by modulation of osmolality. To this end we used growth media with different osmolality, adjusted by sorbitol, to reverse or promote any potential membrane disruption induced by Tir membrane integration. Incubating SNU-C5 cells in hypertonic medium (~360 mOs/kg) did not affect EPEC-1-induced Ca 2+ influx and cell death (Figure 2a

| Tir-induced activation of TRPV2 specifically induces caspase-4 pyroptosis
We have recently shown that Ca 2+ influx leads to LPS internalization and activation of caspase-4 (Zhong et al., 2020). In order to investigate whether direct activation of TRPV2 could substitute Tir-induced activation during EPEC-1 infection, we treated SNU-C5 cells with the TRPV2 agonist cannabidiol (CBD) with or without extracellular LPS. Fluo-4 assay confirmed that CBD induced Ca 2+ influx, which could be inhibited by SET2 (Figure 3a). Moreover, CBD treatment alone induced PI-uptake, which was inhibited by SET2 or TRPV2 silencing (Figure 3b,c). However, the addition of LPS to CBD-treated cells did not enhance PI uptake ( Figure 3b). The pan-caspase inhibitor zVAD, which inhibits pyroptotic as well as apoptotic caspases, partially inhibited CBD-induced cell death ( Figure 3b). However, specific silencing of caspase-4 by siRNA could not reduce CBD-induced cell death ( Figure 3c). As caspase-4 is the only pyroptotic caspase expressed in SNU-C5 (Roumeliotis et al., 2017), CBD-induced cell death is likely dependent on the traditionally apoptotic caspases.
Therefore, while activation of TRPV2 alone induces Ca 2+ influx and cell death, Tir-induced activation of TRPV2 leads specifically to caspase-4-dependent pyroptosis.
Increase in Ca 2+ influx by extracellular ATP enhances EPEC-1induced cell death in a caspase-4-dependent manner, while ATP alone does not induce cell death (Zhong et al., 2020). We, therefore,

| ER transport is required for TRPV2 activation
In resting conditions, TRPV2 resides on the ER membrane. In response to mechanical or chemical stimuli, TRPV2 is transported to the plasma membrane where it becomes activated (Nagasawa et al., 2007). Unfortunately, we were unable to stain for TRPV2 using the commercially available TRPV2 antibody and the transfection efficiency of plasmid pCMV encoding GFP-tagged TRPV2 was too low for robust quantitative analysis (data not shown). However, in order to determine the need for an ER-to-membrane trafficking in EPEC-1-induced TRPV2 activation, we pre-treated SNU-C5 cells with brefeldin A, which inhibits vesicle trafficking between ER and Golgi. A control experiment revealed that brefeldin A treatment slightly reduced cell death following LPS transfection ( Figure 4a). In contrast, infection with EPEC-1 resulted in a significant reduction in Ca 2+ influx and pyroptosis compared to untreated EPEC-1-infected cells ( Figure 4b,c), confirming the requirement of ER-to-Golgi transport.
EPEC E2348/69 injects 21 effectors into the host cell, forming a complex signaling network (Garmendia et al., 2005;Ruano-Gallego et al., 2021). It has been shown that the activity of some effectors could be antagonized by others. For example, we have shown that Tir-induced cell death could be mitigated by NleF (Zhong et al., 2020), which inhibits caspase-4 (Blasche et al., 2013;Pallett et al., 2014Pallett et al., , 2017. NleA, whose expression and subsequent injection is coupled to Tir translocation (Katsowich et al., 2017), has been shown to bind to the COPII cargo selection protein Sec24 in HeLa cells, where it prevents ER-to-Golgi transport (Thanabalasuriar et al., 2012). We, therefore, hypothesized that NleA might interfere with TRPV2 transport from the ER and prevent EPEC-1-induced F I G U R E 1 TRPV2 inhibition and silencing reduce Tir-induced Ca 2+ influx and pyroptosis. (a) Schematics of EPEC-0 and EPEC-1 interaction with epithelial cells. (b) qRT-PCR of TRPV2 mRNA from RNA extracts of SNU-C5 cells treated by control or TRPV2 siRNA for 3 days. Data were normalized to the relative quantification of TRPV2 in control siRNA-treated cells from the first biological repeat. Means ± SEM from n = 3 independent biological repeats are shown. (c) Fluo-4 assay at 20 min after EPEC-1 infection of control or TRPV2-silenced SNU-C5 cells. UI, uninfected. Means ± SEM from n = 6 independent biological repeats are shown. (d) PI uptake at 8 hr after EPEC-1 infection of control or TRPV2-silenced SNU-C5 cells. Means ± SEM from n = 6 independent biological repeats are shown. (e) Caspase-4 Western blot of combined lysates and supernatants from EPEC-1-infected control or TRPV2-silenced SNU-C5 cells. Shown is a representative blot from n = 3 independent biological repeats are shown. (f) Fluo-4 assay at 20 min after EPEC-1 infection of SNU-C5 cells with or without SET2 pre-treatment. Means ± SEM from n = 6 independent biological repeats are shown. (g) PI uptake at 8 hr after EPEC-1 infection of SNU-C5 cells with or without SET2, YVAD, YVAD + SET2, or EGTA pre-treatment. PI uptake of EPEC-0 infection is included as a negative control. Means ± SEM from n = 10 (SET2) and 3 (EGTA, YVAD, YVAD + SET2, and EPEC-0) independent biological repeats are shown. Statistical significance was determined using 2-tailed t test (b-d, f) or 1-way ANOVA with Tukey post-test (g). *p ≤ .05; **p ≤ .01; ***p ≤ .001; ns, nonsignificant F I G U R E 2 Osmolality affects Tir-induced Ca 2+ influx and pyroptosis. (a) Fluo-4 assay at 20 min after EPEC-1 infection of SNU-C5 cells incubated in isotonic or hypertonic medium. Means ± SEM from n = 3 independent biological repeats are shown. (b) PI uptake into EPEC-1-infected SNU-C5 cells incubated in isotonic or hypertonic medium. Means ± SEM from n = 3 independent biological repeats are shown. (c) Fluo-4 assay at 20 min after EPEC-1 infection in control or TRPV2-silenced SNU-C5 cells incubated in isotonic or hypotonic medium. Means ± SEM from n = 3 independent biological repeats are shown. (d) PI uptake into EPEC-1-infected SNU-C5 cells incubated in isotonic or hypotonic medium. Means ± SEM from n = 6 independent biological repeats are shown. (e) Immunofluorescence microscopy of EPEC-1-infected SNU-C5 cells incubated in isotonic or hypotonic medium. DAPI: blue; EPEC: green; Phalloidin: red. Scale bar: 5 µm. Shown is a representative image from n = 3 independent biological repeats are shown. Statistical significance was determined using two-way ANOVA with Bonferroni post-test. *p ≤ .05; **p ≤ .01; ***p ≤ .001; ns, non-significant FIGURE 3 Replacing Tir-induced TRPV2 Ca 2+ influx with pharmacological TRPV2 activation or purinergic Ca 2+ influx. (a) Fluo-4 assay at 4 min after CBD treatment of SNU-C5 cells, with or without SET2 pre-treatment. UT, untreated. Means ± SEM from n = 5 independent biological repeats are shown. (b) PI uptake at 2 hr after LPS, CBD, CBD + LPS, CBD + zVAD and CBD + SET2 treatment of SNU-C5 cells. Means ± SEM from n = 3 independent biological repeats are shown. (c) PI uptake at 2 hr after CBD and CBD + LPS treatment of SNU-C5 cells. Shown are control, caspase-4-or TRPV2-silenced cells. Means ± SEM from n = 4 independent biological repeats are shown. (d) Fluo-4 assay at 2 min after ATP treatment of SNU-C5 cells, with or without SET2 pre-treatment, and Fluo-4 assay at 20 min post-infection of SNU-C5 cells with EPEC-1. Means ± SEM from n = 3 independent biological repeats are shown. (e) PI uptake into EPEC-1-infected SNU-C5 cells, with or without SET2 and/or ATP pretreatment. Shown are control or caspase-4-silenced cells. PI uptake of EPEC-0 infection is included as a negative control. Means ± SEM from n = 3 independent biological repeats are shown. (f) PI uptake into EPEC-1-infected control or caspase-4 silenced SNU-C5 cells, with or without SET2 pre-treatment. Means ± SEM from n = 3 independent biological repeats are shown. Statistical significance was determined using 2-tailed t test (a), 1-way ANOVA with Tukey post-test (b, d, e, f) or 2-way ANOVA with Bonferroni post-test (c). *p ≤ .05; **p ≤ .01; ***p ≤ .001; ns, non-significant pyroptosis. First, we confirmed that NleA binds Sec24 in SNU-C5 cells. To this end, we infected SNU-C5 cells with EPEC-1 containing a plasmid encoding flag-tagged-NleA (EPEC-1-NleA). Following infection and flag pull-down, we found that Sec24 was co-purified with NleA (Figure 4d). We then confirmed that NleA overexpression did not affect EPEC-1-induced pedestal formation (Figure 4e).
Measuring Fluo-4 emission and PI uptake revealed that expression of NleA in EPEC-1 inhibited Ca 2+ influx and cell death similarly to brefeldin A (Figure 4f,g). A combination of NleA and SET2 did not further reduce EPEC-1-induced Ca 2+ influx and cell death compared to EPEC-1-NleA alone (Figure 4f,g), implying that the two mechanisms converge.
To further investigate the role of NleA in EPEC-induced pyroptosis, we deleted nleA from wild-type (WT) EPEC (EPECΔnleA). EPECΔnleA-induced cell death, suggesting that NleA and NleF control the same pyroptosis pathway (Figure 4j).

Taken together our data suggest that infection of SNU-C5 cells
with EPEC leads to activation of Tir-mediated mechanical stress, trafficking of TRPV2 from the ER to the plasma membrane and Ca 2+ influx, which promotes LPS entry, followed by caspase-4 activation and pyroptosis. Moreover, EPEC translocates the effectors NleA, and NleF, which inhibit trafficking of TRPV2 and caspase-4 and cell death.

| DISCUSS ION
We have previously shown that the EPEC T3SS effector Tir triggers extracellular Ca 2+ influx followed by LPS internalization and caspase-4 activation, culminating in GSDMD-dependent pyroptotic cell death. In this study, we have characterized the mechanisms by which Tir induces Ca 2+ influx. It was known that the insertion of Tir, followed by Tir-intimin interaction, distorts the local membrane structure, likely via Tir-recruited membrane curvature-inducing proteins such as IRTKS, Eps15, and epsins (Ford et al., 2002;Lin et al., 2011;Vingadassalom et al., 2009;Zhao et al., 2011). Such an event could activate mechanosensitive ion channels. From the proteomics of EPEC-1-infected epithelial cells, we found the mechanosensitive Ca 2+ channel TRPV2 is highly expressed at resting conditions and shows an increased abundance during infection of SNU-C5 cells (Zhong et al., 2020). Here we have shown that TRPV2 has a critical role in triggering the early Ca 2+ influx, leading to caspase-4 activation in EPEC-1-infected SNU-C5 cells. Consistent with previous studies on TRPV2, ER transport process is required for TRPV2dependent Ca 2+ influx and cell death during EPEC-1 infection. As a mechanosensitive channel, TRPV2 activation can be controlled by changing medium osmolality and the resulting membrane pressure (Muraki et al., 2003;Zanou et al., 2015). In our study, low osmolality inhibited Tir-and TRPV2-induced Ca 2+ influx and partially reduced cell death. It is likely that Tir activates TRPV2 through membrane structural changes that can be antagonized by the membrane alterations from hypotonic stress. On the other hand, the inability of hypertonic medium to enhance cell death could be explained by the necessity of a localized structural distortion to activate TRPV2 rather than a global cell shape change.
Cytosolic Ca 2+ overload has been associated extensively with different cell death mechanisms. Ca 2+ influx into the cytosol can be from either an extracellular or an intracellular source. Extracellular Ca 2+ influx mediated by plasma membrane ion channels, such as TRP family, purinergic and voltage-gated Ca 2+ channels, can lead to Ca 2+ uptake by organelles including ER, mitochondria, and lysosomes (Kondratskyi et al., 2015). ER can also relay Ca 2+ directly into adjacent mitochondria or release Ca 2+ to promote mitochondrial uptake, triggering the cell death signaling (Joseph & Hajnóczky, 2007;Marchi et al., 2018;Min et al., 2012). Ca 2+ overloading of mitochondria causes mitochondrial outer membrane permeability, cytochrome C release, and mitochondrial ROS production, which participate in intrinsic apoptosis (Gogvadze et al., 2006;Tait & Green, 2013). Furthermore, Ca 2+ influx, mitochondrial ROS production, and mitochondrial DNA release have been implicated in the canonical NLRP3 inflammasome formation (Broz & Dixit, 2016;Murakami et al., 2012;Shimada et al., 2012;Zhou et al., 2011). Early studies of caspase-4 activation have also suggested the role of Ca 2+ influx from stressed ER (Matsuzaki et al., 2010). Our discovery of TRPV2 upstream of caspase-4 activation makes the first connection between the non-canonical inflammasome and a specific plasma membrane Ca 2+ channel.
It remains unclear how TRPV2-dependent Ca 2+ influx leads to LPS entry. Although pharmacological activation of TRPV2 leads to cell lysis, LPS and caspase-4 appear to be dispensable in this pathway. Therefore, Tir appears to play a key role by promoting LPS entry and changing the caspase preference downstream of TRPV2. Both the intimin-Tir cluster and TRPV2 have been shown to be recruited to lipid rafts for their respective functions (Allen-Vercoe et al., 2006;Lévêque et al., 2018).
It is interesting to note that while having an antagonistic function the injection of NleA is coupled to Tir translocations (Katsowich et al., 2017). In addition to NleA, we have previously shown that EspZ reduces Tir translocation while NleF inhibits caspase-4 (Berger et al., 2012;Pallett et al., 2014Pallett et al., , 2017, all leading to reduced pyroptosis in SNU-C5 (Zhong et al., 2020). Increased pyroptosis due to loss of NleA in WT EPEC can be complemented by NleF, showing their redundant roles in pyroptosis inhibition.
Importantly, in the Citrobacter rodentium mouse model of EPEC infection, NleA is one of the three essential effectors that could not be deleted without causing significant attenuation (Ruano-Gallego et al., 2021); it remains to be seen if the main activity of NleA in vivo is to block the side effect induced by Tir membrane integration. The relationship between NleA and NleF is reminiscent to that of NleE and NleC which block the top and bottom of the NF-kB cascade, respectively (Nadler et al., 2010;Sham et al., 2011).

Pyroptosis in infected cells can release pro-inflammatory cyto-
kines such as IL-1β and IL-18 as well as DAMPs to activate immune cells and trap pathogenic bacteria, promoting bacterial clearance (Eldridge & Shenoy, 2015;Jorgensen et al., 2016). In C. rodentium- Overall, we have characterized the involvement of the mechanosensitive Ca 2+ influx via TRPV2 in Tir-induced pyroptosis and its regulation by other effectors intracellularly.

| Infection of SNU-C5 cells
EPEC infection of SNU-C5 cells was performed as described before in (Zhong et al., 2020) with modifications. Briefly, EPEC strains were primed by diluting the overnight cultures 50× in nonsupplemented DMEM (low glucose) and growing for 3 hr static at 37°C with 5% CO 2 . Isopropyl βd-1-thiogalactopyranoside (IPTG) (Sigma) at 0.5 mM was added to the bacterial culture 30 min before infection when required. Infection was carried out at a multiplicity-of-infection (MOI) of 50:1. Spent medium was replaced with fresh serum-free RPMI 1 hr before infection to avoid serum-induced TRPV2 activation (Nagasawa et al., 2007).
Alternatively, isotonic (~290 mOs/kg) and hypotonic (~100 mOs/ kg) medium was prepared using RPMI adjusted by 1 M sorbitol (Sigma) and water supplemented with CaCl 2 (Sigma) to a final con- Medium was replaced with fresh RPMI with a second siRNA dose of identical concentration 24 hr after the first dose. Medium was replaced with IFNγ-containing RPMI one day before infection.

| qRT-PCR
5 × 10 4 cells/well were seeded in 24-well plates three days prior to RNA extraction. RNA extraction and qRT-PCR were performed as described before (Zhong et al., 2020). Briefly, RNAeasy kit (QIAGEN) and RQ1 RNase-free DNase (Qiagen) were used for RNA extraction and DNase digestion. M-MLV reverse transcriptase (Promega) and oligo-dT and random primers (Promega, Madison, Wisconsin, USA) were used to perform the RT reaction at 42°C for 60 min. PowerUp SyBr-Green Mastermix and gene-specific primers ( Figure S4) were used to perform qPCR reaction in a StepOne Real-Time PCR system (Thermo).
TRPV2 cDNA-specific primer sequences were from Caprodossi et al. (2008). GAPDH cDNA level was used as an internal control.

| Immunofluorescence staining
1.5 × 10 5 cells/well were seeded in 24-well plates on glass coverslips for imaging with Zeiss AxioImager Z1 microscope (Carl Zeiss, Jena, Germany). Infection experiments were carried out as described.
Infected cells were fixed and stained as described before in Zhong et al. (2020). Briefly, cells were fixed by 4% paraformaldehyde, permeabilized by 0.2% Triton X-100 (Sigma), blocked for 10 min with 1% bovine serum albumin (BSA) before being incubated with primary antibodies for 45 min, and re-blocked for 10 min with 1% BSA before being incubated with secondary antibodies for 30 min (Table S5).
Coverslips were mounted on glass slides with Gold-Pro-Long-Antifade (Invitrogen) before imaging.

| In vitro infection and pulldown
For pulldown experiments, 3 × 10 5 cells/well were seeded in six-well plates. After priming the EPEC-1 empty vector and EPEC-1-NleA strains, 0.5 mM IPTG was used to induce NleA expression 30 min prior to infection. After 2 hr, cells were washed thrice in PBS followed by cell lysis (50 mM Tris-HCl, pH 7.4, with 150 mM NaCl, 1 mM EDTA, and 1% Triton X-100, protease inhibitor cocktail). The lysate was subjected to binding to Anti-FLAG M2 affinity resin as per the manufacturer's instructions.
Briefly, bead binding was performed for 4 hr at 4°C. For western blotting, the beads were boiled with non-reduced Laemmli buffer to elute the bound proteins and subjected to western blot analysis.
Membranes were blocked with 3% BSA or 5% milk in TBST or PBST for 1 hr at room temperature and incubated with primary antibodies overnight at 4°C, followed by secondary antibodies for 1 hr at room temperature (Table S6). Membranes were developed using the ECL Western Blotting Reagents (GE, Amersham, UK) and imaged using the ChemiDoc MP imaging system (Bio-Rad).

| PI-uptake assay
5 × 10 4 cells/well were seeded in black clear-bottom 96-well plates one day prior to infection and IFNγ and inhibitor treatment were applied as described above. Alternatively, 1.5 × 10 4 cells/ well were seeded three days prior to infection for siRNA transfection. PI-uptake assay was performed as described before in Zhong et al. (2020). Briefly, phenol-red-free DMEM (low glucose) and RPMI was used for T3SS priming and infection, respectively. Cells were incubated in phenol-red-free RPMI supplemented with 5 µg/ ml PI (Sigma) 1 hr prior to infection and throughout the infection.
Alternatively, phenol-red-free RPMI with osmolality adjusted by sorbitol and water as described above and supplemented with 5 µg/ ml PI was used. Cell-free medium-only wells were prepared as blank.
Positive control wells were prepared by incubating cells in RPMI with 5 µg/ml PI and 0.05% Triton X-100 (Sigma) 10 min prior to infection. Infections were carried out as described above. Measurement was taken with 620-nm emission and 520-nm excitation using the FLUOstar Omega Microplate Reader (BMG Labtech, Aylesbury, UK).

| Fluo-4 assay
5 × 10 4 cells/well were seeded in black clear-bottom 96-well plates one day prior to infection and IFNγ was applied as described above. Fluo-4 assay was performed as described before in Zhong et al. (2020). Briefly, cells were incubated in Fluo-4 Direct reagent (Molecular Probes, Eugene, Oregon, USA) diluted two-fold in phenol-red-free RPMI for 30 min before the first measurement.
Alternatively, 25% Fluo-4 Direct reagent was used to prepare isotonic and hypotonic medium adjusted by sorbitol and water. Cell-free medium-only wells were prepared as blank. Fluorescence readings were performed in the FLUOstar Omega Microplate Reader (BMG Labtech, Aylesbury, UK) measuring 520-nm emission with 485-nm excitation.

| Statistical analysis
All experiments were independently repeated at least three times as indicated in the figure legends. Methods of data transformation are described in their corresponding method sections.
Statistical analysis of all biochemical/biological experimental data was performed using GraphPad Prism 5.1. Student's t-test, oneway or two-way analysis of variance (ANOVA) followed by Tukey post-test or Bonferroni post-test, respectively, were performed on the means as listed in the figure legends. Significant result was defined as having a p-value <.05.

ACK N OWLED G M ENTS
We thank Luis Ángel Fernández of CNB-CSIC, Madrid, for sharing EPEC-0 and EPEC-1. QZ is supported by an Imperial College President's PhD Scholarship. The project was supported by a grant from the Wellcome Trust (107057/Z/15/Z).

CO N FLI C T O F I NTE R E S T
The authors declare no conflict.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data sharing not applicable to this article as no datasets were generated or analysed during the current study.