Cell-type-specific hypertranslocation of effectors by the Pseudomonas aeruginosa type III secretion system

Summary

Many Gram-negative pathogens use a type III secretion system (T3SS) to promote disease by injecting effector proteins into host cells. Common to many T3SSs is that injection of effector proteins is feedback inhibited. The mechanism of feedback inhibition and its role in pathogenesis are unclear. In the case of P. aeruginosa, the effector protein ExoS is central to limiting effector injection. ExoS is bifunctional, with an amino-terminal RhoGAP and a carboxy-terminal ADP-ribosyltransferase domain. We demonstrate that both domains are required to fully feedback inhibit effector injection. The RhoGAP−, but not the ADP-ribosyltransferase domain of the related effector protein ExoT also participates. Feedback inhibition does not involve translocator insertion nor pore-formation. Instead, feedback inhibition is due, in part, to a loss of the activating trigger for effector injection, and likely also decreased translocon stability. Surprisingly, feedback inhibition is abrogated in phagocytic cells. The lack of feedback inhibition in these cells requires phagocytic uptake of the bacteria, but cannot be explained through acidification of the phagosome or calcium limitation. Given that phagocytes are crucial for controlling P. aeruginosa infections, our data suggest that feedback inhibition allows P. aeruginosa to direct its effector arsenal against the cell types most damaging to its survival.

Plain language summary

Many bacteria use specialized nano-machines to inject proteins into host cells in order to cause disease. Here we show that injection of proteins by the bacterium Pseudomonas aeruginosa is feedback inhibited by two of the injected proteins. Unlike other cells, injection into phagocytic cells, which constitute our first line of defense against bacteria, is not feedback inhibited, suggesting that feedback inhibition has evolved to allow P. aeruginosa to focus its arsenal against the very cells that are needed to control the infection.

Introduction

Pseudomonas aeruginosa, like many Gram-negative pathogens, uses a type III secretion system (T3SS) to promote disease (Hauser, 2009, Engel & Balachandran, 2009, Galan et al., 2014). The T3SS is required for systemic spread of the organism (Lee et al., 2005, Vance et al., 2005, Balachandran et al., 2007, Nicas et al., 1985) and contributes materially to morbidity and mortality associated with these infections (Roy-Burman et al., 2001, Hauser et al., 2002, El-Solh et al., 2012). A primary function of the T3SS in P. aeruginosa infections is to stave off killing by neutrophils that are rapidly mobilized to the site of infection (Diaz & Hauser, 2010, Diaz et al., 2008, Rangel et al., 2014, Sun et al., 2012, Vareechon et al., 2017). Effector injection inactivates the phagocytic machinery in these cells (Rangel et al., 2014) and results in an inability to produce a reactive oxygen species burst (Vareechon et al., 2017). These phenotypes are dependent on effector proteins that are injected into host cells via the T3SS. To date, four effector proteins have been described in P. aeruginosa isolates. ExoS and ExoT are highly homologous bifunctional effector proteins with an amino-terminal Rho-GTPase activating protein (RhoGAP) domain and a carboxy-terminal ADP-ribosyl transferase domain (ADPRT)(Barbieri & Sun, 2004). ExoU, is a potent phospholipase (Sato et al., 2003, Tyson & Hauser, 2013, Sato & Frank, 2014), and ExoY is an adenylate cyclase (Yahr et al., 1998). The distribution of effectors among strains varies. Almost all strains of P. aeruginosa produce ExoT, the distribution of ExoS and ExoU is for the most part mutually exclusive, with about 70% of strains producing ExoS (Feltman et al., 2001, Borkar et al., 2013). While ExoS may be more prevalent, the presence of ExoU has been associated with more severe disease (Finck-Barbancon et al., 1997). ExoY seems to only contribute to disease when overexpressed from a plasmid (Munder et al., 2018).

Translocation of effector proteins requires a specialized structure called the translocon (Sory & Cornelis, 1994, Persson et al., 1995, Rosqvist et al., 1995). For P. aeruginosa, the translocon consists of a pore formed in the host cell plasma membrane by two translocator proteins, PopB and PopD (Frithz-Lindsten et al., 1998), that is thought to be docked to the needle tip, which consists of a pentamer of PcrV (Broz et al., 2007). The T3SS needle-tip mediates insertion of the pore-forming translocator proteins into the host cell plasma membrane (Goure et al., 2004, Goure et al., 2005, Picking et al., 2005). Sensing of host cell contact, which triggers effector secretion, is thought to be mediated by a conformational change in the translocation pore that is transmitted to the needle-tip and from thence down the needle to the base of the apparatus (Kenjale et al., 2005, Torruellas et al., 2005, Russo et al., 2019, Armentrout & Rietsch, 2016).

Translocon assembly, stability, and perhaps also triggering of effector secretion are controlled by host cell factors. Translocator insertion by Vibrio parahaemolyticus T3SS2 requires host cell surface fucosylation (Blondel et al., 2016). Translocation by the Yersinia pseudotuberculosis T3SS requires a functional actin cytoskeleton (Viboud & Bliska, 2001). It has been proposed that C-C chemokine receptor 5 (CCR5) coordinates cytoskeletal activities required for translocation by Y. pseudotuberculosis (Sheahan & Isberg, 2015). In Shigella, translocon stability is promoted by an interaction with the intermediate filament protein vimentin (Russo et al., 2016). Activation of the T3SS in P. aeruginosa is prevented if host cells are pre-treated with pore-forming toxins, or depleted of cholesterol with methyl-ß-cyclodextrin, arguing that host cell processes are required to bring the T3SS to bear (Bridge et al., 2010, Cisz et al., 2008). The ability of P. aeruginosa to inject effector proteins appears to be linked to adherence of the target cell to a surface (Bridge et al., 2010, Verove et al., 2012), and it was proposed that cortical actin dynamics at the leading edge of cells are important for delivery of effectors into host cell (Bridge et al., 2010).

In several species, including P. aeruginosa, injection of effector proteins by the T3SS is feedback regulated. In other words, an effector protein, once injected into the host cell, limits further injection of effectors into the targeted cell. This phenomenon was first described in Yersinia spp. The secreted protein YopK controls effector translocation into host cells (Holmstrom et al., 1997, Garcia et al., 2006, Dewoody et al., 2011, Dewoody et al., 2013). YopK controls effector translocation from the host cell cytosol, and interacts with the translocation pore via the translocator protein YopD (Dewoody et al., 2013). YopK is not conserved among other type III secretion systems, arguing that this particular mechanism of regulation is specific to Yersinia spp. The injection of effector proteins by Yersinia spp. is also controlled by YopE, a GTPase-activating protein (GAP) effector (Aili et al., 2006, Aili et al., 2008). Evidence suggests that regulation depends on the GAP activity of YopE, which controls pore formation by manipulating the actin cytoskeleton of the host cell (Mejia et al., 2008). Consistently, the effect of YopE on pore formation can be overcome by producing constitutively active versions of RhoA and Rac1 (Viboud & Bliska, 2001). In P. aeruginosa, feedback inhibition of effector translocation was linked to ExoS (Cisz et al., 2008, Bridge et al., 2010, Bridge et al., 2012, Novotny et al., 2013). Delivery of ExoS into host cells prevents subsequent triggering of effector secretion in intoxicated host cells and inactivating both RhoGAP and ADPRT activity of ExoS results in hypertranslocation of ExoS into epithelial cells. Notably, the RhoGAP activity of ExoS also restores feedback inhibition in the Yersinia system; however, in contrast to P. aeruginosa, the ADPRT activity of ExoS has no effect in Yersinia, highlighting a difference between the two organisms (Aili et al., 2008). Feedback inhibition of effector translocation in enteropathogenic Escherichia coli (EPEC)(Mills et al., 2008) is linked to the activity of the effector EspZ, which interacts with the pore-forming translocator EspD (Creasey et al., 2003, Berger et al., 2012).

Here we further dissect the mechanism of feedback inhibition of T3SS effector translocation in P. aeruginosa. We demonstrate that the RhoGAP activities of ExoS and ExoT can both contribute to feedback inhibition of effector translocation. Their contribution is redundant, arguing that they share a cellular target. The ADPRT activity of ExoS, but not ExoT, is also required for full feedback inhibition of effector translocation. Feedback inhibition is only weakly suppressed by a deletion of the gene encoding the negative regulator of effector secretion, Pcr1, arguing that feedback inhibition is also controlled at the level of translocon stability. Translocator insertion and pore-formation are not affected. Surprisingly, we found that feedback inhibition differs depending on the cell type infected. P. aeruginosa hypertranslocates effectors into phagocytic cells. Feedback inhibition is restored by blocking phagocytosis of the infecting bacteria, arguing that injection of effectors by bacteria in the phagosomal compartment is no longer feedback inhibited.

Results

Feedback inhibition of translocation into epithelial cells relies on the RhoGAP activities of ExoS and ExoT, as well as the ADPRT activity of ExoS

In P. aeruginosa, triggering of effector secretion results in up-regulation of T3SS gene expression by export of the negative regulator ExsE via the T3SS (Rietsch et al., 2005, Urbanowski et al., 2005). This up-regulation can be captured using GFP reporter of exoS transcription (Rietsch & Mekalanos, 2006). Activation on cell contact requires translocon-dependent injection of ExsE into host cells (Urbanowski et al., 2007). We had previously observed feedback inhibition of effector translocation by infecting cells with strains producing only ExoS and subsequently monitoring the ability to trigger effector injection using our GFP reporter strain (Cisz et al., 2008). In these experiments, injection of ExoS completely blocked activation of translocation in the GFP reporter strain, whereas injection of ExoT and ExoY somewhat reduced activation of the reporter, but did not rise to the level of statistical significance (Cisz et al., 2008). Injection of ExoS in which either RhoGAP or ADPRT activities had been inactivated blocked activation of the GFP reporter, arguing that either activity was sufficient for feedback inhibition of effector translocation. In a similar vein, both the RhoGAP and ADPRT activity of ExoS contributed to translocation control in T24 epithelial cells and a rat mammary gland adenocarcinoma cell line, MTLn3 (Bridge et al., 2012, Novotny et al., 2013). This however, is at odds with the observation that ExoT is not involved in feedback inhibition, even though it has a RhoGAP domain that is highly homologous to that of ExoS with similar target specificity (Henriksson et al., 2002, Goehring et al., 1999, Krall et al., 2000). We therefore decided to revisit the role of ExoS and ExoT in effector translocation by directly detecting the amount of ExoS translocated into targeted host cells.

We first examined translocation of ExoS into A549 epithelial cells when delivered by a strain of P. aeruginosa in which exoT and exoY had been deleted. In these experiments, A549 cells were infected, the cells were removed from the culture dishes by proteinase K digestion, and lysed with either Triton X-100 (lyses A549 cells, but not bacteria) or sodium dodecyl sulfate (SDS, lyses both bacteria and A549 cells), which was confirmed by detecting tubulin and RpoA (bacterial RNA polymerase alpha subunit). Translocated (Triton X-100, T) and total, bacterial and epithelial cell-associated (SDS, S), ExoS were detected by Western blot. Inactivation, through point mutations, of the RhoGAP domain or ADPRT domain alone resulted in an intermediate increase in translocated ExoS, compared to the level of translocation observed if both activities were inactivated (Figure 1A). The shift in ExoS molecular weight in lysates from cells infected with a strain harboring ExoS with an intact ADP-ribosyltransferase activity is due to auto-ADP-ribosylation of translocated ExoS (Riese et al., 2002). Differences in ExoS levels in the SDS fraction (bacteria + epithelial cells), when comparing wild-type and GAP−/ADPR− bacteria, are likely due to differences in the amount of translocated protein, rather than a change in expression, since the strains are all exsE null mutants, which constitutively up-regulate expression of the type III secretion genes, and production of ExoS is equivalent in these strains (Figure 2C, S1). When we performed the experiment in a strain in which exoT and exoY were intact, we found that inactivation of the ExoS RhoGAP activity no longer affected translocation, whereas the ADPRT activity of ExoS did (Figure S2). A simple explanation for these data is that the RhoGAP activity of ExoT is able to substitute for the RhoGAP activity of ExoS. To test this hypothesis, we deleted exoT from strains producing either wild-type ExoS or the RhoGAP/ADPRT− double mutant protein. Removing exoT resulted in a slight, approximately six-fold increase in wild-type ExoS translocation (compared to the 100x increase seen when ExoS(GAP−/ADPRT−) is translocated in the absence of ExoT) (Figure 1B). The ability of ExoT to control effector translocation was more evident in the strain producing the double mutant ExoS (RhoGAP/ADPRT−). Here ExoT reduced ExoS translocation to 14% of the level seen in the absence of ExoT (Figure 1B). Notably, simply inactivating the RhoGAP activity of ExoT restored ExoS translocation to the level seen in the exoT deletion mutant, demonstrating that the ADPRT activity of ExoT is not involved in feedback inhibition. ExoY is not important for feedback inhibition, since the strains producing ExoT also produce ExoY (Figure 1B). Feedback inhibition affects effector translocation in general since translocation of ExoT is also feedback inhibited (Figure S3), and our previously published data indicates that injection of the regulatory ExsE protein is also inhibited (Cisz et al., 2008). Taken together, these data indicate that the RhoGAP and ADPRT activities of ExoS collaborate to feedback inhibit effector translocation in P. aeruginosa. Moreover, the RhoGAP activity of ExoT similarly contributes to feedback inhibition and is functionally redundant with the RhoGAP activity of ExoS.

Figure 1. Feedback inhibition of effector translocation is a function of the RhoGAP activities of ExoS and ExoT, as well as the ADPRT activity of ExoS.

A549 cells were infected with P. aeruginosa for 90 minutes, at which point cells were washed, resuspended by proteolysis, and lysed with either Triton X-100 (translocated protein fraction, “T”) or SDS (total bacterial and A549 cell protein fraction, “S”). ExoS, tubulin, and RpoA were detected by Western blot. (A) P. aeruginosa lacking exoT and exoY and producing ExoS [“S+”], ExoS(GAP−) [“G−”], ExoS(ADPRT−) [“A−”], or ExoS (GAP−/ADPRT−) [“G/A−”]. (B) P. aeruginosa producing either wild-type ExoS [“S+”] or ExoS(GAP−/ADPRT−) [“S(G/A−)”], in the presence or absence of ExoT [“ΔT”], ExoY, or ExoT(GAP−) [“T(G−)”], as indicated. Blots representative of at least three independent experiments are shown. Quantitation of translocated ExoS normalized to tubulin, is presented in the graphs to the right of the blot images and is represented as a percentage of the amount of ExoS translocated by the strain producing ExoS(GAP−/ADPRT−) in the absence of ExoT and ExoY. Mean and standard deviation values are indicated above each bar. Averages of translocated protein were compared to levels of the ExoS(GAP−/ADPRT−) protein by one-way ANOVA with Tukey multiple comparison test. * p < 0.05, ** p < 0.01, *** p < 0.0001, **** p < 0.0001, and n.s. not significant.

Figure 2. Deletion of pcr1 partially overcomes feedback inhibition in epithelial cells.

A549 cells were infected for 90 minutes with P. aeruginosa producing either wild-type ExoS [“S+”], or ExoS(GAP−/ADPRT−) [“G/A−”]. Where indicated, the strains lacked the gene encoding the negative regulator of effector secretion, Pcr1 [“Δpcr1”]. Subsequently, cells were washed, resuspended by proteolysis, and lysed with either Triton X-100 (translocated protein fraction, “T”) or SDS (total bacterial and A549 cell protein fraction, “S”). ExoS, tubulin, and RpoA were detected by Western blot. (A) representative image of the translocation experiment. (B) Quantitation of four biological replicates, expressed as percentage of translocated ExoS relative to the matched ExoS(GAP−/ADPRT−) mutant (pcr1+ and Δpcr1). Mean and standard deviation values are indicated above each bar. (C) Media samples from the A549 cell infection experiment in (A) were removed, separated into bacterial cells and supernatant protein, and separated by SDS-PAGE. ExoS and RpoA were detected by Western blot. Mean and standard deviation values are indicated above each bar. Ratios of translocated ExoS:ExoS(GAP−/ADPRT−) were compared by two-tailed Student’s T-test (B), * p < 0.05.

Feedback inhibition is partially due to the loss of the host-cell trigger for effector secretion

Translocation into host cells can be affected by any number of parameters, including translocator insertion, pore-formation, triggering of effector secretion, and translocon stability. In order to begin to dissect the step that is impacted by feedback inhibition we decided to examine whether feedback inhibition can be reversed in bacteria that do not require a specific trigger to commence effector secretion. To this end, we compared the feedback inhibition phenotype of exoS+ and exoS (RhoGAP/ADPRT−) strains to mutant derivatives lacking the secretion regulator Pcr1. Pcr1 is a component of the PopN complex which prevents effector secretion before cell contact (Yang et al., 2007, Lee et al., 2014). Consequently, a pcr1 null mutant strain no longer needs to receive a host cell signal to commence effector secretion. Notably, while a pcr1 mutant secretes effector proteins constitutively (Figure 2C), it still requires an intact translocon to inject effector proteins into host cells (Armentrout & Rietsch, 2016). If feedback inhibition is a function of deactivation of the triggering host cell signal, then deleting pcr1 should result in high-level injection of wild-type ExoS into the cell. ExoS translocation was elevated about five-fold relative to the amount of ExoS(RhoGAP−/ADPRT−) translocated, indicating that ExoS feedback inhibition, at least in part, inactivates the host-cell trigger of effector secretion (Figure 2A, B). However, while injection of wild-type ExoS was elevated when compared to the corresponding pcr1+ strains, injection of ExoS did not revert to the level seen in the exoS(RhoGAP−/ADPRT−) strain, arguing that feedback inhibition of effector secretion must also involve another step in the process such as translocon assembly and/or stability.

In order to determine whether translocator insertion or pore-formation are affected by ExoS, we assayed the ability of P. aeruginosa to insert PopB and PopD into host cell membranes, and translocon-dependent pore-formation. To assay pore-formation, we infected A549 cells for 2 hours, washed cells with PBS containing 1M KCl to remove peripherally associated proteins, and then liberated translocator proteins inserted into the plasma membrane using Triton X-100. As a control, we included a strain of P. aeruginosa that lacks the needle-tip protein PcrV. This strain can still make and secrete translocator proteins, but fails to insert them into host membranes (Armentrout & Rietsch, 2016). Both exoS+ and exoS(RhoGAP−/ADPRT−) strains were able to insert PopB and PopD, with no apparent defect in the presence of functional ExoS (Figure 3A).

Figure 3. ExoS does not interfere with translocator insertion or pore formation.

(A) A549 cells were infected with strains RP2318 (“S+”), RP2349 (“G/A−”), or RP9268 (“G/A− ΔpcrV”), cells were washed with 1M KCl to remove peripherally associated translocator proteins, then lysed with SDS (total cell associated proteins, including attached bacteria) or Triton X-100 (plasma membrane-associated proteins). Fractions were separated by SDS-PAGE and PopB, PopD, as well as E-cadherin (plasma membrane protein), and RpoA (bacterial RNA polymerase) were detected by Western blot. The experiment shown is representative of three biological replicates. (B) Pore formation was assessed by infecting A549 cells with strains RP2318 (“S+”), RP2349 (“G/A−”), RP3670 (“G/A− ΔHBD”, lacking pcrHpopBD) in the presence of propidium iodide for 1.5h. Cells were washed and mounted, and PI uptake was assessed by fluorescence microscopy. Representative phase contrast and red fluorescence images are shown for each infection condition. The graph to the right shows average numbers of PI+ cells from four biological replicates.

We next examined whether pore-formation is affected by the activity of ExoS. In order to assay translocon-dependent pore formation, we performed the infection the presence of propidium iodide (PI), which is membrane impermeant and is taken up if pores are introduced into the host cell plasma membrane (Corrotte et al., 2015, Pathak-Sharma et al., 2017). PI uptake was visualized by fluorescence microscopy. ExoS did not interfere with the ability of P. aeruginosa to introduce translocon-dependent pores into host cells (Figure 3B). We noted that PI uptake seemed to be enhanced in the strain producing wild type ExoS. In order to analyze this phenotype more carefully, we examined PI uptake over time, both by microscopy (Figure S4A) and by using a fluorescence plate reader (Figure S4B). The experiments demonstrate that the time at which pore formation is first detectable doesn’t differ between the two strains, but the rate of PI uptake is markedly enhanced by the presence of ExoS. This increased uptake of PI cannot be explained by activation of caspases and subsequent activation of gasdermins (Shi et al., 2015), since the pan-caspase inhibitor zVAD was not able to block PI uptake (Figure S4B). The data are consistent with ExoS interfering with membrane damage repair. Inactivating membrane damage repair by removing calcium from the medium results in accelerated PI uptake in cells that have been damaged by the bacterial toxin streptolysin O (Idone et al., 2008). Taken together, these data argue that feedback inhibition is imposed at the level of triggering of effector secretion and translocon stability, but not translocator insertion or pore-formation.

P. aeruginosa hypertranslocates effector proteins into phagocytic cells

While P. aeruginosa can inject effector proteins into cells from organisms as divergent as Dictyostelium discoideum, Acanthamoeba castellanii (Pukatzki et al., 2002, Matz et al., 2008) and mammals, some cell type differences have been observed. Specifically, certain cell types, such as undifferentiated HL-60 cells are resistant to T3SS-mediated effector injection (Rucks & Olson, 2005, Verove et al., 2012). We therefore decided to examine effector translocation in other cell types.

Surprisingly, we found that the relative translocation of ExoS, compared to the ExoS(RhoGAP/ADPRT−) mutant protein, was strongly elevated in J774 cells, compared to A549 cells (Figure 4A, B). As we had observed previously for A549 epithelial cells, ExoS intoxication does not affect attachment of bacteria to J774 cells ((Cisz et al., 2008), Table S1). J774 cells are a murine macrophage-like cell line that is phagocytic in vitro. Consistent with this observation, translocation into human corneal epithelial cells and murine fibroblasts is feedback inhibited, whereas translocation into the human macrophage cell line U937 mirrors the hypertranslocation result seen with J774 cells (Figure S5). We therefore examined whether phagocytosis of P. aeruginosa could be contributing to the hypertranslocation phenotype. Treatment of J774 cells with the actin polymerization inhibitor cytochalasin D strongly interfered with phagocytic uptake of P. aeruginosa (Figure 4C). Notably, this treatment also restored feedback inhibition of effector translocation in J774 cells, while having no appreciable effect on translocation into A549 cells (Figure 4A, B). As was the case for A549 cells (Cisz et al., 2008, Armentrout & Rietsch, 2016), translocation into J774 cells is translocon dependent (Figure S6A), and both ExoS and ExoT are hypertranslocated into J774 cells in a strain that produces both effector proteins (Figure S6C). Taken together, these experiments argue that P. aeruginosa hypertranslocates effector proteins into phagocytic cells, and that this hypertranslocation depends on phagocytosis of the pathogen.

Figure 4. Hypertranslocation into phagocytic cells.

(A) A549 cells or J774 cells were infected for 30 minutes with strains RP2318 and RP2349, producing wild-type ExoS [“S+”] or ExoS(GAP−/ADPRT−) [“G/A−”], respectively, washed and incubated for an additional 60 minutes before resuspending the cells by proteolysis. Where indicated, cells were pre-incubated with cytochalasin D to depolymerize the actin cytoskeleton. The inhibitor was present throughout the experiment. Cells were lysed with either Triton X-100 (translocated protein fraction, “T”) or SDS (total bacterial and A549 cell protein fraction, “S”). ExoS, tubulin, and RpoA were detected by Western blot. Representative images are shown. (B) Quantitation of translocated ExoS normalized to tubulin is graphed for five biological replicates and is represented as a percentage of the amount of ExoS translocated by the strain producing ExoS(GAP−/ADPRT−). Mean and standard deviation values are indicated above each bar. (C) Phagocytosis was assayed by infecting A549 cells or J774 cells with the indicated strain of P. aeruginosa in the presence or absence of cytochalasin D. Average, gentamicin resistant colony forming units (CFU) from at least 3 biological replicates are plotted. Ratios of translocated ExoS:ExoS(GAP−/ADPRT−) were compared by one-way ANOVA with Tukey multiple comparison test. ** p < 0.01, *** p < 0.0001, and n.s. not significant.

Hypertranslocation into phagocytes is a function of phagocytosed bacteria and cannot be explained by changes in intracellular calcium or pH.

In order to investigate the hypertranslocation phenotype further, we decided to examine the contribution of phagocytosed bacteria. To this end, we performed the translocation assay as in Figure 4 (30 min. infection, followed by a wash step and continued incubation with fresh media for 60 minutes), except that gentamicin was added to 150μg/ml during the second incubation step in order to kill extracellular bacteria. ExoS was still hyperinjected under these conditions (Figure 5A), and injection occurred after addition of gentamicin, arguing that it was carried out by intracellular bacteria (Figure S6B). In vitro, type III effector secretion can be triggered by removing calcium from the medium (Yahr et al., 1997). We had demonstrated previously that shuttling calcium into cells did not interfere with triggering of effector secretion on host cell contact, arguing that this is a non-physiological cue (Cisz et al., 2008). Shuttling calcium into J774 cells using the ionophore ionomycin similarly had no effect on the hypertranslocation phenotype, suggesting that the intracellular bacteria are not experiencing a low-calcium environment, resulting in constitutive activation of the T3SS (Figure 5A). We also examined whether a change in pH, e.g. acidification of the phagosome, could result in activation of the T3SS. A change in pH from acidic to neutral activates the SPI-2 T3SS, which is produced inside host cells by Salmonella enterica sv. Typhimurium (Yu et al., 2010). However, shifting P. aeruginosa from medium adjusted to pH5 to medium adjusted to pH 7.2 did not result in triggering of effector secretion. Moreover, pH5 interferes with effector secretion, arguing that acidification of the phagosome would, if anything, be deleterious (Figure 5B). These data indicate that phagocytosed bacteria hypertranslocate effector proteins into the engulfing host cell, and that this phenotype is not the result of low calcium levels inside the phagocyte, or acidification of the phagosome.

Figure 5. Killing extracellular bacteria or shuttling calcium into cells does not block hypertranslocation, nor do pH changes explain the hypertranslocation phenotype.

(A) J774 cells were infected for 30 minutes with strains RP2318 and RP2349, producing wild-type ExoS [“S+”] or ExoS(GAP−/ADPRT−) [“G/A−”], respectively, washed and incubated for an additional 60 minutes before resuspending the cells by proteolysis. In one set of experiments, the culture medium was supplemented with 150μg/mL gentamicin during the 60 minute incubation step to kill extracellular bacteria. In a second set of experiments, cells were pre-treated for 15 minutes with 10μM ionomycin (which was maintained throughout the experiment) to shuttle calcium from the medium into cells. Cells were lysed with either Triton X-100 (translocated protein fraction, “T”) or SDS (total bacterial and A549 cell protein fraction, “S”). ExoS, tubulin, and RpoA were detected by Western blot. Representative images are shown. The graph to the right shows quantitation of translocated ExoS normalized to tubulin for four biological replicates, represented as a percentage of the amount of ExoS translocated by the strain producing ExoS(GAP−/ADPRT−). Mean and standard deviation values are indicated above each bar. (B) To assess whether changes in pH can trigger effector secretion, bacteria were grown to early log in LB buffered to pH 5 or pH 7.2, pelleted and resuspended in media of the same or opposite pH, in the presence or absence of calcium. Bacteria were grown for an additional 30 minutes, at which point cells and supernatants were separated by centrifugation and normalized samples were separated by SDS-PAGE. ExoS, PopD, and RpoA were detected by Western blot. The image shown is representative of 3 biological replicates.

Discussion

Many pathogens feedback inhibit translocation of effector proteins into host cells. It has been proposed that this process helps limit toxicity of translocated effector proteins, or affects recognition by intracellular pattern recognition molecules (Brodsky et al., 2010, Zwack et al., 2017, Berger et al., 2012). Interestingly, the mechanism of feedback inhibition seems to vary between pathogens. YopK and EspZ seem to exert their influence by interacting directly with the translocon. However, they share no discernable homology and interact with different components of the translocon: YopD, in the case of YopK, and EspD, which is homologous to YopB), in the case of EspZ(Dewoody et al., 2013, Creasey et al., 2003, Berger et al., 2012). Similarly, even though the ExoS RhoGAP activity can restore feedback inhibition in a Yersinia yopE null mutant background, the ADPRT activity, which contributes significantly to feedback inhibition in P. aeruginosa, has no effect (Aili et al., 2008). Moreover, interfering with actin polymerization reduces effector translocation in Yersinia pseudotuberculosis, but does not interfere with translocation in the case of P. aeruginosa (Cisz et al., 2008, Viboud & Bliska, 2001)(Figure 4). Taken together, these observations suggest that imposing feedback regulation on effector injection is an important enough characteristic of T3SSs to have evolved in parallel in multiple organisms.

Here we expand our analysis of the effector functions that impose feedback inhibition in P. aeruginosa. We had reported previously that ExoS feedback inhibits effector translocation into host cells. While ExoS injection prevented triggering of effector secretion by subsequently attaching bacteria, injection of the other two effector proteins produced by P. aeruginosa PAO1, ExoT and ExoY did not result in a significant amount of inhibition (Cisz et al., 2008). Curiously, in a similar superinfection assay, the RhoGAP activity of ExoS alone was able to block effector injection. Given that the RhoGAP activities of ExoT and ExoS are thought to have similar target specificities, this suggested that perhaps the noise of the fluorescent reporter assay was obscuring a role for ExoT in feedback inhibition. This is particularly interesting since a large percentage of P. aeruginosa isolates (~30%, up to 50%, depending on the study) do not produce ExoS, but instead produce ExoU, whereas almost all isolates of P. aeruginosa produce ExoT (Feltman et al., 2001, Karthikeyan et al., 2013, Toska et al., 2014). To resolve this discrepancy, we examined translocation of ExoS directly. We first examined feedback inhibition in a strain of P. aeruginosa that only produces ExoS. Inactivation of either the RhoGAP or ADPRT activities, or both, by point mutation, led to a loss of feedback inhibition. The phenotype of knocking out either the RhoGAP or ADPRT activities alone was intermediate, when compared to the wild-type and RhoGAP/ADPRT− double mutant, arguing that both activities collaborate to impose feedback inhibition. When we performed the experiment in the context of a strain producing ExoT, only the ADPRT activity of ExoS downregulated effector translocation, suggesting that ExoT was compensating for the loss of the RhoGAP activity. Indeed, in a strain lacking the RhoGAP activity of ExoT, feedback inhibition was once again a function of the dual RhoGAP and ADPRT activities of ExoS. Since a complete deletion of exoT behaved similarly to the point mutant in which only the RhoGAP activity had been inactivated, we conclude that the RhoGAP activity, but not the ADPRT activity of ExoT can feedback inhibit effector translocation. The RhoGAP activities of ExoS and ExoT act redundantly, and in concert with the ADPRT activity of ExoS to feedback inhibit effector translocation. Notably presence or absence of ExoY had no significant effect on translocation.

We also examined feedback inhibition in different cell types. We found that translocation is feedback inhibited in fibroblasts and epithelial cells, but that feedback inhibition was greatly diminished in phagocytic cells. Hypertranslocation into phagocytic cells was tied to phagocytic uptake of the bacteria. Preventing phagocytosis resulted in the restoration of feedback inhibition. The presence of wild type ExoS or the catalytically inactive mutant did not affect attachment to phagocytic J774 cells. Moreover, in experiments where extracellular bacteria were killed through the addition of gentamicin, translocation of ExoS occurred after the initial infection, during the 60-minute incubation with gentamicin. The hypertranslocation phenotype is therefore a function of the phagocytosed bacteria. These data are consistent with recent work studying translocon assembly in Yersinia enterocolitica (Nauth et al., 2018). The authors found that bacteria associated with HeLa cells have fewer assembled translocons than bacteria associated with macrophage. Interestingly, conditions that up-regulate translocation in Yersinia, deletion of yopE or production of a constitutively active Rac1 variant, also up-regulated the number of translocon-producing bacteria on infected HeLa cells. Using a clever system of fluorescent labeling, the authors went on to demonstrate that bacteria with assembled translocons associate with a prevacuolar compartment, suggesting that translocon assembly correlates with invasion (Nauth et al., 2018).

We further characterized the parameters that govern hypertranslocation into phagocytic cells. Shuttling of calcium into cells had no effect, arguing that low-calcium triggering of effector secretion cannot explain this phenotype. Nor could hypertranslocation be explained by changes in pH as would be expected upon phagosome acidification, which serves as a trigger of effector secretion for the Salmonella Typhimurium SPI-2 T3SS (Yu et al., 2010). Finally, we examined whether the hypertranslocation phenotype indicates that the bacteria in the phagosome experience a constant trigger of effector secretion. However, removing the need for host cell contact to trigger effector secretion, by deleting the gene encoding the negative regulator pcr1, only partially reversed the feedback inhibition phenotype. These data suggest that phagocytosis also promotes hypertranslocation by affecting translocon stability, since neither translocator insertion or pore formation are negatively regulated by ExoS. Our observations are consistent with the increase in assembled translocons seen in phagocytosed Y. enterocolitica (Nauth et al., 2018). However, given that we performed our experiments in a strain in which expression of the T3SS is constant, and given that the number of needles produced by P. aeruginosa under these conditions is very low (unlike Yersinia) (Kudryashev et al., 2013, Nauth et al., 2018, Rietsch & Mekalanos, 2006, Lee et al., 2014), we favor a model whereby phagocytosis stabilizes the assembled T3SS conduit by preventing dissociation of the bacterium from the host cell.

Feedback inhibition of effector translocation is a common feature of many T3SS, however its role in infection is unclear. It has been proposed that feedback inhibition helps the pathogen limit inflammation. For example, hypertranslocation of YopB and YopD in yopK mutant Yersinia is pro-inflammatory (Brodsky et al., 2010, Zwack et al., 2015, Zwack et al., 2017). Our work raises the possibility that feedback inhibition controls how effectors are allocated to different cell types in vivo. Phagocytic cells, primarily neutrophils, are the first responders to an infection and are crucial for controlling P. aeruginosa in an infection (Diaz et al., 2008, Diaz & Hauser, 2010, Vareechon et al., 2017, Sun et al., 2010, Hazlett et al., 1992, Kooguchi et al., 1998). They are the primary cell type injected early on in the infection process (Diaz & Hauser, 2010), and eliminating neutrophil recruitment abrogates the need for a T3SS to establish an infection (Sun et al., 2012). We therefore propose that P. aeruginosa limits injection of effector proteins into epithelial cells and fibroblasts, using the concomitant up-regulation of T3SS gene expression (Cisz et al., 2008, Urbanowski et al., 2007) to arm itself for the subsequently recruited neutrophil response. Neutrophils kill P. aeruginosa using a reactive oxygen species (ROS) burst that is rapidly generated upon phagocytosis, but is efficiently blocked by ExoS and ExoT (Vareechon et al., 2017). Our data suggest a model whereby the up-regulation of the rate of effector secretion (Lee et al., 2014) and hyper-translocation of effectors by phagocytosed bacteria allow P. aeruginosa to win the race between neutrophils mounting an ROS response and the effector-mediated block of ROS production (Vareechon et al., 2017).

Our work also raises an additional possibility for explaining differential targeting of host cells in animal models of infection. Target cell selection in animal models of infection has been studied using beta-lactamase reporter fusions to effector proteins by treating cells isolated from infected tissues with the fluorescence resonance energy transfer (FRET) reporter CCF2 (Deuschle et al., 2016, Diaz & Hauser, 2010, Geddes et al., 2007, Sheahan & Isberg, 2015, Marketon et al., 2005, Charpentier & Oswald, 2004). Cleavage of CCF2 results in loss of the FRET, and changes the molecule's fluorescence emission from green to blue. Notably, the rate of conversion is a function of the amount of the beta-lactamase fusion that has been translocated (Dewoody et al., 2011). In fact, it has been used to monitor feedback inhibition in vitro (Dewoody et al., 2011, Dewoody et al., 2013). Our data raise the possibility that the initial bias for injection into phagocytic cells seen in infections with P. aeruginosa and Yersinia spp. may be a result of hyperinjection of the fusion protein into these cells. Conversely, the later targeting of epithelial cells in infection, which has been documented in a lung model of P. aeruginosa infection (Rangel et al., 2014), could be the result of invasion of these cells, which has been documented in vivo, in an animal model of corneal infection (Fleiszig et al., 1994), and in vitro, after prolonged infection of epithelial cells with P. aeruginosa (Angus et al., 2008, Heimer et al., 2013, Nieto et al., 2019).

In summary, we have presented evidence that feedback inhibition of effector translocation in P. aeruginosa infected epithelial cells is a function of the ADPRT activity of ExoS, as well as the RhoGAP activities of ExoS and ExoT, which function redundantly. Feedback inhibition did not impact translocator insertion or pore formation, and was only partially overcome by removing the need for a host-cell trigger of effector secretion, arguing that it likely also involves control of translocon stability. Notably, feedback inhibition was largely abolished in phagocytic cells. Here, hyperinjection of effectors is mediated by phagocytosed bacteria, and is not controlled by changes in pH or calcium levels. Our work suggests that P. aeruginosa modulates injection of effectors into host cells, in order to specifically target phagocytic cells, which are the primary cell type responsible for controlling the pathogen in vivo.

Experimental Procedures

Cell Lines, Bacterial Strains, and Plasmids

Strains and plasmids are listed in Table 1. Bacteria were grown in high salt LB (HS-LB), which consists of 10g/L tryptone, 5g/L yeast extract, 11.7g/L NaCl (200mM final concentration) with MgCl₂ and CaCl₂ added after autoclaving to 5mM or 0.5mM, respectively. For pH shift experiments, media were set to pH 5 or 7.6 using a sodium phosphate buffer (25mM final concentration). Where indicated, calcium was chelated by adding 5mM ethylene glycol tetraacetic acid (EGTA) to the medium.

Table 1.

Strains and plasmids

Strain #	Description	Reference
RP1910	PAO1F ΔexsE	(Cisz et al., 2008)
RP2318	PAO1F ΔexsE ΔexoT ΔexoY	(Cisz et al., 2008)
RP2347	PAO1F ΔexsE ΔexoT ΔexoY exoS(RhoGAP−)	(Cisz et al., 2008)
RP2348	PAO1F ΔexsE ΔexoT ΔexoY exoS(ADPR−)	(Cisz et al., 2008)
RP2349	PAO1F ΔexsE ΔexoT ΔexoY exoS(RhoGAP−/ADPR−)	(Cisz et al., 2008)
RP5989	PAO1F ΔexsE exoS(RhoGAP−)	This study
RP6483	PAO1F ΔexsE exoS(ADPRT−)	This study
RP6484	PAO1F ΔexsE exoS(RhoGAP−/ADPRT−)	This study
RP10551	PAO1F ΔexsE exoS(RhoGAP−/ADPRT−) exoT(RhoGAP−)	This study
RP2996	PAO1F ΔexsE ΔexoT ΔexoY Δpcr1	(Armentrout & Rietsch, 2016)
RP2998	PAO1F ΔexsE ΔexoT ΔexoY exoS(RhoGAP−/ADPR− ) Δpcr1	This study
RP9268	PAO1F ΔexsE ΔexoT ΔexoY exoS(RhoGAP−/ADPR− ) ΔpcrV2	(Armentrout & Rietsch, 2016)
RP3624	PAO1F ΔexsE ΔexoT ΔexoY ΔpcrHpopBD	(Armentrout & Rietsch, 2016)
RP3670	PAO1F ΔexsE ΔexoT ΔexoY exoS(RhoGAP−/ADPR− ) ΔpcrHpopBD	(Armentrout & Rietsch, 2016)
RP3201	PAO1F ΔexsE ΔexoT ΔexoY ΔpopB	This study
RP3203	PAO1F ΔexsE ΔexoT ΔexoY exoS(RhoGAP−/ADPR− ) ΔpopB	This study
Plasmid
pEXG2	allelic exchange vector, colE1 origin, oriT, gentamicin resistance, sacB	(Rietsch et al., 2005)
pEXG2-exoS(GAP−)	introduces R146K mutation into exoS	(Cisz et al., 2008)
pEXG2-exoS(ADRPT−)	introduces E379D/E381D mutation into exoS	(Cisz et al., 2008)
pEXG2-exoT(GAP−)	introduces R149K mutation into exoT	(Sun et al., 2012)
pEXG2-Δpcr1	deletion of codons 10-81 of pcr1	(Tomalka et al., 2012)
pEX-popB	popB deletion construct	(Sundin et al., 2002)
pP42-malE-mCherry	constitutively produced MBP-mCherry fusion, operator-less lacUV5 promoter, gentamicin resistance	This study

A549 cells (ATCC# CCL-185) and mouse embryonic fibroblast (MEF) cells (ATCC# SCRC-1008) were obtained from the American Type Culture Collection. J774 cells, U937 cells, and 10.014 pRSV-T human corneal epithelial cells were gifts from Dr. Clifford Harding, Dr. Jacek Skowronski, and Dr. Edward Medof, respectively, here at Case Western Reserve University. All cells, with the exception of the corneal epithelial cells were grown at 37°C in a CO₂ incubator (5% CO₂), in RPMI1640 media supplemented with 10%FBS (RP10), and, in the case of U937 cells, 25mM HEPES, pH7.4. U937 cells were differentiated in the presence of 10ng/ml phorbol-12-myristate-13-acetate (PMA) for 2 days before the experiment. On the day of the experiment the cells were washed 2x with PBS and the media replaced with fresh medium lacking PMA. Corneal epithelial cells were grown in media consisting of Keratinocyte-Serum Free Media (Gibco), supplemented with 5ng/ml human recombinant epidermal growth factor (Gibco), 50μg/ml bovine pituitary extract (Gibco), 5μg/ml insulin (Sigma), and 100ng/ml hydrocortisone (Sigma). Tissue culture flasks for culturing human corneal epithelial cells were precoated for 2 hours at 37°C with 10μg/ml bovine serum albumin (BSA, Sigma), 5μg/ml fibronectin (Sigma), and 30μg/ml collagen I (Vitrogen, Cohesion Tech) in phosphate buffered saline.

Bacterial strains used in this study are listed in Table 1. Chromosomal mutations were generated by allelic exchange. Plasmids harboring the mutant allele (Table 1) were mated into P. aeruginosa using E. coli strain SM10 λpir. Cointegrates were selected in LB plates with 30μg/ml gentamicin and 5μg/ml triclosan. Cointegrates were grown briefly (2-3h) in LB lacking salt, and then plated on sucrose plates (LB agar without salt but with 5% sucrose) and grown at 30°C to select for loss of the sacB counterselection marker. Colonies were tested for gentamicin sensitivity (loss of the pEXG2 vector backbone) and for the presence of the mutation by PCR. Plasmid pPSV41 was generated by replacing the lacUV5 promoter in plasmid pPSV37 with a promoter lacking the lacO operator sequence by digesting pPSV37 (Lee et al., 2010) with DraIII and EcoRI and ligating a linker containing the promoter into the vector backbone. The linker was derived by annealing two primers UV5noOp1 (5’-GTGCTTTACACTTTATGCTTCCGGCTCGTATAATGTGTGG-3’) and UV5noOp2 (5’-AATTCCACACATTATACGAGCCGGAAGCATAAAGTGTAAAGCACGTA-3’). Plasmid pPSV42 was derived from plasmid pPSV41, by deleting the lacI open reading frame. To this end, plasmid pPSV41 was digested with SspI and religated. malE lacking its signal sequence was amplified from the chromosome of E. coli MG1655 using primers malE-5R (5’-ATATAgaattcTAAGGAGGCGCCCCCATGAAAATCGAAGAAGGTAAACTGG-3’) and malE-3K (5’-ATATAggtaccTgcggccgcCTTGGTGATACGAGTCTGCGC-3’). mCherry was amplified using primers mCherryfMBP-5Not (5’-AAAAAgcggccgcaGTGAGCAAGGGCGAGGAGGAT-3’) and Cherry3H (5’-AAAAAaagcttTTACTTGTACAGCTCGTCC-3’). The malE PCR product was digested with NotI and EcoRI, the mCherry PCR product was digested with NotI and HindIII, and both were ligated into pPSV42 digested with EcoRI and HindIII to generate pP42-malE-mCherry.

Translocation assay

For cytochalasin D treatment experiments, J774 cells (est. 1*10⁶ cells/ well, seeded 7*10⁵ cells in a 6-well plate) and A549 cells (est. 5*10⁵ cells/well seeded 3*10⁵ cells/well in a 6-well plate) infected in 4ml of RP10 w/ 1mM pyruvate for 30 min with 1*10⁶ CFU of P. aeruginosa, then washed 1x with phosphate buffered saline supplemented with 5mM MgCl₂ and 0.5mM CaCl₂ (PBS-MC), and incubated with 4ml of RP10 with pyruvate for 1h. Where indicated, cells were pre-incubated with 5μM cytochalasin D for 30’ and the inhibitor was maintained throughout the experiment, the other cells were treated with an equal amount of DMSO. After the incubation, the cells were washed 2x with PBS-MC and rinsed with 250μg/ml proteinase K in PBS-MC. The protease solution was removed and the cells were incubated at room temperature for 15 minutes. Subsequently, the cells were resuspended in 1ml PBS-MC with 2mM phenylmethylsulfonyl fluoride (PMSF), pelleted by centrifugation (300 xg for 5 minutes), and resuspended in 50μl PBS-MC with 2mM PMSF and 0.25% Triton X-100 to permeabilize host cells (15’ on ice). After the incubation, 24μl of the suspension were removed and mixed with 8μl 4x SDS sample buffer (“S”, SDS sample), the remaining cells were spun down (9600 xg for 3 minutes) and 24 μl supernatant (“T”, Triton fraction) were mixed with 8μL 4x SDS sample buffer. The samples were boiled for 15’ at 95°C. 10μl of each sample were separated on a 10% polyacrylamide gel (BioRad), transferred to a PVDF membrane, and blocked with 5% non-fat milk in PBS-0.05% Tween20 (PBS-T). The blots probed for ExoS using an affinity purified rabbit antibody in blocking buffer, washed 3x with PBS-T, and probed with an anti-rabbit-horse radish peroxidase secondary antibody. Horse-radish peroxidase activity was detected chemiluminescently using WesternBright Sirius reagent (Advansta) and recorded using an ImageQuant LAS 4000 scanner. Blots were subsequently stripped by incubation with stripping buffer (50 mM Tris (pH 6.8), 2% SDS, with 8 μL beta-mercaptoethanol added to 10 mL just before use) at room temperature for 20 minutes, washed 2x with tris-buffered saline-Tween 20 (TBS-T) and re-blocked with 5% milk in TBS-T. Stripped blots were probed for tubulin (Cell Signaling Technologies) and RpoA (BioLegend), and detected using an anti-mouse IgG horseradish peroxidase antibody.

Experiments in which extracellular bacteria were killed using gentamicin were carried out analogously, however the infection was carried out in RP10 with pyruvate and the media was replaced with media containing 150μg/ml gentamicin after the initial 30 minute infection period. In the case of the ionomycin experiments, the ionophore was added to a final concentration of 10μM 15 minutes before the infection and maintained throughout the experiment.

In experiments comparing translocation of ExoS active site mutants into A549 cells, as well as when analyzing the function of the pcr1 mutant derivatives, cells were infected continuously for 90 minutes with 5*10⁶ CFU. Subsequently, cells were processed as indicated above.

Phagocytosis assay

For each strain and condition assayed, two wells of a 24-well plate were seeded with 7.5*10⁴ A549 cells or 1.5*10⁵ J774 cells/well (assuming 1*10⁵ A549 cells/well and 2*10⁵ J774 cells/well on the day of the experiment). 30 minutes before infection, the cells were washed 1x with PBS, and the media was replaced with 1ml pre-warmed RP10/pyruvate with DMSO (0.1%) or cytochalasin D (5μM, from a 5mM stock in DMSO). After the pre-incubation period, the cells were infected for 30 minutes with 2.5*10⁵ cfu/ml P. aeruginosa. After 30 minutes of infection, the media were removed, the cells were washed 1x with PBS, and pre-warmed RP10 with 150μg/ml gentamicin was added back. The cells were incubated for an additional 30 minutes to kill extracellular bacteria, then washed 2x with PBS, and lysed with 500μl of PBS-MC/0.1% Triton X-100. Dilutions of the lysate were plated to quantitate surviving CFU.

Secretion assay

For examining the effect of pH on secretion, bacteria were grown overnight in high salt LB medium set to pH 5 or pH 7.2 by adding 25mM monobasic (pH5) or dibasic (pH 7.2) sodium phosphate and adjusting the pH using the converse sodium phosphate solution. Cultures were diluted 1:300 into two culture tubes with 5ml fresh medium of the same pH and grown for ~2.5h, at which point bacteria from 2ml aliquots were pelleted and resuspended in 2ml of LB pH 5 or pH 7.2, with or without 5mM EGTA to chelate calcium. Cultures were grown for an additional 30 minutes, before bacteria from 1ml of culture were pelleted and resuspended in 1x SDS sample buffer. Secreted proteins from 500μL of culture supernatant were precipitated with 10% trichloroacetic acid, washed 1x with acetone, dried and resuspended in 1x SDS sample buffer. Pellet and supernatant fractions were normalized by the OD₆₀₀ of the culture. 10μL aliquots were separated by SDS polyacrylamide gel electrophoreses and analyzed by Western blot using antibodies directed against ExoS, PopD, and RpoA.

For infection supernatant samples at the end of the experiment, before washing the cells, 500μl of culture supernatant were removed from the tissue culture flask. Bacteria were pelleted by centrifugation (3 minutes 21100 xg). 75 μL of supernatant were combined with 25μL of 4x SDS sample buffer to generate the supernatant samples. Bacteria were resuspended in 500μL of PBS and 75μL were mixed with 25μL of 4x SDS sample buffer to generate the bacteria samples. All samples were boiled and separated by SDS-polyacrylamide gel electrophoresis. ExoS and RpoA were detected by Western blot.

To examine effector and translocator production by strains used in this study (Figure S1), bacteria were grown in high salt LB to mid-logarithmic phase, at which point the cultures were diluted 1:1 with high salt LB supplemented with 10mM EGTA (5mM final concentration) and grown for another hour. At this point, 75μL of culture were removed, mixed with 25μL 4x SDS sample buffer, and boiled for 10 minutes. 500μL of culture were removed and the bacteria were pelleted by centrifugation. 75μL of supernatant were mixed with 25μL of 4x SDS sample buffer and boiled for 10 minutes. Optical density at 600nm was determined for each culture and total culture and supernatant samples were adjusted to correspond to the same OD using 1x SDS sample buffer.

Membrane insertion assay

For each condition, one T75 flask of A549 cells (~70% confluent) was infected with P. aeruginosa at an MOI of 25 for 2h at 37°C with 5% CO₂ in RP10 (or, as a control, left uninfected). Cells were washed 1x with PBS-MC and incubated 15’ with 0.5ml of RPMI1640 (no FBS) with 4U/μl of SLO (activated for ~10’ with 10mM DTT) in the presence of 2mM PMSF. The cells were then washed again with 1x PBS-MC, 1x with PBS-MC with 1M KCl, and 1x with PBS-MC. The cells were then scraped into 1ml of PBS-MC with 2mM PMSF and pelleted in a microcentrifuge (3’ 300 xg) and the supernatant discarded. Cells were resuspended in 60μL of PBS-MC with 0.25% Triton-X100 and 2mM PMSF and incubated on ice for 15’. Each cell suspension was then vortexed and 12μL were mixed with 4μL 4xSDS sample buffer and diluted 1:1 with 16μL 1x SDS sample buffer to make the input sample. The remainder was pelleted and 45μL of the supernatant were mixed with 15μL 4x sample buffer to make the membrane (Triton X-100) fraction. Both samples were boiled 15’ at 95°C, and 10μL were separated on a 4-15% gel by SDS-PAGE and transfered to PVDF. E-cadherin, PopB, PopD, and RpoA were detected by Western blot.

Pore formation assay

A549 cells were seeded in a 24-well plate (~1.2*10⁵ cells/well) on glass cover slips one day before the experiment. On the day of the experiment, the cells were washed 1x with phosphate buffered saline (PBS), and the media was replaced with 270μl prewarmed RPMI1640 media (no phenol red, no FBS). Propidium iodide was added in 30μl of RPMI1640 to achieve a final concentration of 4μg/ml for microscopy experiments and 2μg/ml for spectrophotometric experiments in each well. Where indicated, the pan-caspase inhibitor zVAD-fmk (Selleckchem) was added to a final concentration of 50μM/well (Hodges et al., 2019), other wells received an equal volume of vehicle (DMSO). For microscopy experiments, cells were infected at a MOI of 100 for 1.5h, then washed 2x with PBS. Cover slips were removed with tweezers and inverted onto mounting medium (ProLong Gold, Invitrogen) which had been spotted onto glass slides. The mounting medium was cured overnight, at which point the cover slips were rinsed to remove buffer residue and imaged using a Nikon inverted fluorescence microscope. The same exposure conditions were used for each strain. Levels of microscopy images were adjusted using Acorn (Flying Meat software), using the same levels settings for all images being compared. In order to monitor PI uptake over time by fluorescence spectroscopy, cells were seeded in a 24-well plate (without cover slip) and infected as above. PI fluorescence was gathered using a Synergy HT plate reader (BioTek) heated to 37°C. Fluorescence for each well was measured before infection, and the baseline was set to 0% fluorescence. At the end of the 1.5h time course, Triton X-100 was added to each well to a concentration of 0.1% in order to permeabilize the cells and read the maximal fluorescence (set to 100%).

Quantitation of blots and statistical analysis

Blot images were quantified using the ImageJ software package. Graphs were generated using Microsoft Excel, and statistical analysis of data was performed using Graphpad Prism 8. Data were compared by ANOVA with Tukey post-hoc test.

Data Availability Statement

The data that supports the findings of this study are available in the supplementary material of this article.