The C-terminal part of ExlA, but not the five RGD motifs, confers on the protein its cytolytic activity.

ExlA is a 172-kDa protein belonging to the family of polymorphic toxins (24). In addition to the conserved secretion motifs based on an analysis in the Pfam database (http://pfam.xfam.org), ExlA shares five β-sheet domains (Fil HAEM) and one hemagglutinin domain (HAEMA) (Fig. 1C) with FHA (25). The ExlA protein also carries five arginine-glycine-aspartic acid (RGD) motifs that are often involved in cell-to-cell recognition with proteins of the integrin family (26). As well, in the majority of polymorphic toxins, as well as in the filamentous hemagglutinin adhesion protein FHA (27), the C-terminal portions specify various activity functions and share conserved features (24), However, the ExlA C-terminal part of approximately 300 amino acids shares no homology with any other protein or domain of known function. Therefore, we undertook to determine the roles of the RGD motifs and the C-terminal domain in cytotoxicity. To that aim, in ExlA we replaced the five RGD motifs with RGA, a mutation known to prevent integrin binding, and also engineered a truncated ExlA (ExlA_ΔCter) which lacks the last 296 amino acids. Both mutant proteins were produced and secreted at the same level as the wild-type protein (Fig. 1D). Cytotoxicity assays on epithelial A549 cells showed that while the ExlA_RGA pentamutant exhibited high cytotoxicity (90%) similar to that of the wild-type ExlA, ExlA_ΔCter completely lost the ability to lyse epithelial cells (Fig. 1D), showing that the cytolytic activity of the ExlA protein requires its C-terminal domain but does not depend on the RGD motifs.

ExlA forms pores in RBCs.

Strains expressing ExlA induce plasma membrane rupture of cultured eukaryotic cells, leading to rapid cell death (within 3 h postinfection) (6). To further decipher the mechanism of ExlA action, we examined its capacity to form pores within membranes, as is the case for several bacterial membrane-damaging toxins, including ShlA of Serratia marcescens (28). The classical model to study the ability of a toxin to form pores in biological membranes utilizes red blood cells (RBCs), where pore formation is evaluated by measuring hemoglobin release (29). RBCs were incubated with P. aeruginosa strains expressing different ExlA variants, and hemoglobin release was measured after 90 min of coincubation. The well-characterized pore-forming toxin α-hemolysin from Staphylococcus aureus, S. marcescens expressing ShlA, and P. aeruginosa PP34, which produces the T3SS-exported phospholipase ExoU, all induced more than 90% hemolysis within 90 min of incubation (Fig. 2A). Negative controls included S. marcescens strain 21c4, unable to secrete ShlA, and P. aeruginosa lacking ExoU (PP34ΔexoU), which showed weak (ca. 20%) hemolysis (Fig. 2A). The T2SS/T3SS-deficient P. aeruginosa PAO1ΔxcpRΔpscC (“Parental” in Fig. 2A) was poorly hemolytic with approximately 25% hemolysis, whereas expression of exlB-exlA in the same strain resulted in >80% hemolysis. As already observed for the cytolysis of A549 cells, P. aeruginosa expressing ExlA_ΔCter induced reduced hemolysis; however, this was higher (by about 50%) than that of the parental ExlA⁻ strain (Fig. 2A). For the demonstration that ExlA is indeed a pore-forming toxin, we approximated the pore size using a variety of osmoprotectants in the cytolytic assays. In these experiments, osmoprotectants (sugars and polyethylene glycols [PEGs]) of different sizes allowed us to evaluate the diameter of the pore (Fig. 2B). As expected based on a previous report (30), the α-hemolysin forms a pore of about 1 nm and the hemolysis is inhibited by raffinose, while ShlA forms a pore of about 2 nm, with hemolysis being inhibited by PEG 2000 (31). Hemolysis induced by ExoU was not inhibited by any sugars or PEGs, an observation in agreement with the fact that ExoU is not a pore-forming toxin but a potent phospholipase (32). The smallest molecule inhibiting ExlA-induced hemolysis was maltopentaose, which indicated that the diameter of the ExlA pore is approximately 1.6 nm, based on the hydrodynamic radius of the osmoprotectant molecules, as described previously (29). In contrast, the 50% hemolysis generated by ExlA_ΔCter could not be inhibited by any of the osmoprotectants. This response is similar to that observed with ExoU, suggesting that this hemolysis was not due to pore formation but rather was caused by some other membrane-disruptive activity detectable on RBCs. Initially, we observed that natural clinical isolates synthetizing ExlA are in general poorly hemolytic (9). We hypothesized that the difference between the clinical isolates and the PAO1 strain ectopically overexpressing ExlA, which was used in the hemolytic assays, is due to the difference in the quantities of ExlA secreted by individual strains. Since PA7 is not amenable to genetic manipulation, we selected strain IHMA, a clinical isolate from our laboratory collection that secretes ExlA and is lytic on epithelial cells (see Table S1 in the supplemental material). In this strain, we introduced pBAD-exlB-exlA into the chromosome and performed a hemolysis assay using variable arabinose concentrations to induce different levels of exlA expression. Indeed, raising the concentrations of arabinose significantly increased the capacity of a clinical strain to lyse RBCs in a dose-dependent manner (Fig. 2C), and this lytic activity correlates with the increased quantity of secreted ExlA (Fig. 2D).

Open in a separate window

FIG 2 

Exolysin is a pore-forming toxin. (A) Lysis of RBCs incubated with α-hemolysin from S. aureus or infected (at an MOI of 10) with S. marcescens secreting ShlA (Db11), an S. marcescens ShlA⁻ mutant (21c4), P. aeruginosa strain PP34 expressing T3S toxin ExoU (ExoU), P. aeruginosa PP34ΔExoU, the P. aeruginosa PAO1ΔxcpRΔpscC strain (Parental), and the P. aeruginosa PAO1ΔxcpRΔpscC::pSW196exlBA strain secreting ExlA (ExlA) or the P. aeruginosa PAO1ΔxcpRΔpscC::pSW196exlBA_ΔCter strain secreting ExlA_ΔCter (ExlAΔCter). (B) Hemolysis assays performed with α-hemolysin (100 nM) and the indicated strains as described in the legend to Fig. 3A in the presence of 30 mM sugars and PEG osmoprotectants. (C) Hemolysis following incubation with P. aeruginosa IHMA or IHMA::pSW196exlBA. For induction of exlBA, bacteria were grown in the presence of 0.025%, 0.075%, and 0.1% arabinose for 3 h and used at an MOI of 10. In a control, the hemolysis assay was done with IHMA in the presence of arabinose (0.1%) in EBM-2. In panels A to C, hemolysis was determined by measuring the release of hemoglobin into the supernatants (OD₅₄₉). Complete (100%) lysis of RBCs was accomplished using 1% SDS. Statistical analysis was performed by analysis of variance (*, P ≤ 0.05; n.s., not significant). (D) Western blot analysis of ExlA secreted by IHMA or IHMA::pSW196exlBA in the presence of arabinose. FliC (Sup) probed with anti-FliC antibody was used as a loading control.

Both the recombinant ExlA and its C-terminal domain induce liposome leakage.

The observed phenotypes of ExlA-expressing P. aeruginosa strains on nucleated cells and on RBCs strongly suggest that ExlA may directly interact with membranes to induce pore formation. To further gain insight into the mode of action and to evaluate the association of the protein with membranes, we purified recombinant ExlA variants and tested their ability to induce membrane damage on the phospholipid bilayers of lipid vesicles. Three ExlA variants were designed for the experiment: (i) the full-length ExlA protein lacking the N-terminal export signal peptide and starting at Gly³⁵ (ExlA_noSP), (ii) a truncated version of ExlA lacking the last C-terminal 296 amino acids (ExlA_ΔCter), and (iii) a polypeptide comprising the C-terminal 296-amino-acid region (Cter_ExlA). After optimization of the overexpression conditions and purification steps (see Materials and Methods for further details), the three variants were obtained in soluble forms and in quantities required for biochemical characterization (1, 6, and 28 mg of protein/liter of culture for ExlA_noSP, ExlA_ΔCter, and Cter_ExlA, respectively). Proteins were stable except for ExlA_noSP, which over time showed some degradation (Fig. 3A). Size exclusion chromatography with detection by multiangle laser light scattering analysis (SEC-MALLS) of the three variants showed that all of them are produced as soluble monomers (Fig. 3A). The various proteins were tested in a liposome leakage assay as described previously (33, 34, 35) (Materials and Methods), where the large unilamellar vesicles (LUVs) incorporating sulforhodamine B (SRB) were prepared by freeze-thaw cycles followed by extrusion and subsequently checked by dynamic light scattering. As a control for vesicle preparation, we used the T3SS translocator PopB, which was previously shown to promote membrane permeabilization at pH 5 (33, 35). We tested the three ExlA variants at various pHs and on LUVs consisting of phosphatidylcholine (PC) alone or a mixture of PC and phosphatidylserine (PS) or other lipids. ExlA_noSP was highly efficient in inducing SRB leakage from LUVs composed of PC/PS at neutral pH (Fig. 3B), whereas other types of LUVs were resistant to permeabilization even at higher protein concentration (up to 5 times the initial concentration) (Fig. S1). For comparison, ShlA similarly requires PS and PC for pore formation, and the addition of phosphatidylethanolamine (PE) furthermore increases the susceptibility of LUVs to ShlA (34). ExlA_noSP lacking the C-terminal region (ExlA_ΔCter) was unable to induce significant liposome leakage at any tested pH, confirming the experiments on nucleated cells, which showed that the C-terminal domain is required for cytolysis (Fig. 3B). Moreover, the 35-kDa polypeptide comprising only the C-terminal domain (Cter_ExlA) possesses on its own a significant membrane permeabilization activity, though exclusively at pH 4 (Fig. 3B). This result shows that the Cter_ExlA protein’s activity requires conformational changes induced by low pH and these are not required for the full-length protein activity, suggesting that the N-terminal domain contributes to conformational changes required for membrane binding and the induction of liposome leakage.

Open in a separate window

FIG 3 

ExlA induces dye leakage from PC/PS liposome. (A) SEC-MALLS analysis of three ExlA variants, ExlA lacking the signal peptide (ExlA_noSP), ExlA_noSP lacking the C-terminal domain (ExlA_ΔCter), and the C-terminal fragment of ExlA (Cter_ExlA), showing their soluble and monomeric form. The insets show SDS-PAGE analyses of the corresponding proteins at different stages of purification (lane 1, total lysate; lane 2, soluble lysate; lane 3, eluted fraction after the second step of purification). The values to the left are molecular masses in kilodaltons. (B) Permeabilization of PC/PS liposomes by ExlA proteins. Sulforhodamine B-containing liposomes composed of PC and PS were incubated with ExlA proteins, and the release of SRB was measured over 500 s. The initial rate of dye release (V0) was plotted for the three ExlA proteins, at pH 4 and pH 7. Statistical analysis was done by analysis of variance (*, P ≤ 0.05).

Kinetics of pore formation and membrane disruption on nucleated cells seen by video microscopy.

Having established that ExlA displays a pore-forming activity on RBCs and liposomes, we sought to determine the kinetics of the various events leading to the death of eukaryotic cells induced by clinical strains of P. aeruginosa expressing ExlA. To that aim, we observed epithelial A549 cells infected with IHMA or its genetic variants by time-lapse microscopy. To follow the plasma membrane permeability, the infection was performed in the presence of YoPro-1, a membrane-impermeant dye (629 Da) that fluoresces upon interaction with RNA in the cytoplasm and DNA in the nucleus and can be used as a marker of small pores occurring during early apoptosis (Invitrogen). In A549 cells infected with ExlA-expressing strains, we observed and measured an increase of YoPro-1 fluorescence at early time points in the cytoplasm and later in the nucleus (as illustrated in Fig. 4A). The data show an initial phase of slow fluorescence increase from 30 to 60 min corresponding to the complete entry of the dye in the cytoplasm. This is followed by a phase of rapid increase of fluorescence until it reaches a plateau corresponding to the entry into the nucleus. Infection with IHMAΔexlA and IHMAexlA_ΔCter did not show any fluorescence in the cytoplasm or in the nuclei at any time points (Fig. 4A). We quantified the first phase of YoPro-1 incorporation on 100 cells for each condition and found significant differences in calculated slope values for IHMA versus IHMAΔexlA and IHMA versus IHMAΔexlA::pSW196exlBA_ΔCter (Fig. 4B). We conclude that the ExlA-induced pore formation, as visualized by YoPro-1 incorporation, starts at 30 min postinfection. To further characterize early steps of ExlA-induced morphological changes in epithelial cells, we performed time-lapse microscopy with a confocal microscope to generate three-dimensional (3D) images. A549-EGFP (enhanced green fluorescent protein) cells, with membranes labeled with wheat germ agglutinin (WGA)-Alexa 647, were infected with various strains secreting ExlA. Under these conditions, we observed that IHMA-infected cells initially became round (Fig. 4C; time 140 min), followed by a loss of their internal EGFP content, leaving shells of empty plasma membranes (Fig. 4C; time 220 min). Therefore, ExlA rapidly forms pores in the membranes of A549 cells, allowing the entry of small molecules (such as YoPro-1) into the cytoplasm. At later time points, the cell morphology changes and cells become round. ExlA eventually leads to plasma membrane rupture, an event detected by GFP release and visualized by discontinuous WGA labeling.

Open in a separate window

FIG 4 

ExlA-induced pore formation precedes plasma membrane disruption. (A) Incorporation of YoPro-1 (green) into infected A549 cells was visualized and measured by time-lapse microscopy. The representative images (inset figures) taken with an ArrayScan wide-field microscope are shown for the wild-type infection. Bars, 10 μm. A549 cells were infected with P. aeruginosa IHMA (WT), IHMAΔexlA (ΔexlA), and IHMAexlA_ΔCter (exlA_ΔCter). (B) YoPro-1 incorporation was analyzed for P. aeruginosa IHMA (WT)-, IHMAΔexlA (ΔexlA)-, IHMAΔexlA::pSW196exlBA (ΔexlA/exlBA)-, and IHMAΔexlA::pSW196exlBA_ΔCter (ΔexlA/exlBA_ΔCter)-infected A549 cells and compared to a noninfected (NI) control. Fluorescence intensities were measured using HCS studio analysis software. The box plot represents slopes of phase I YoPro-1 intensities (from 30 to 60 min postinfection) (n = 100 cells per condition). Asterisks indicate statistically significant differences (Kruskal-Wallis, P < 0.05). (C) A549-EGFP cells, whose plasma membranes were labeled with WGA-Alexa 647, were infected at an MOI of 10 with P. aeruginosa IHMA (WT) or IHMAΔexlA (ΔexlA) and observed by confocal spinning-disk microscopy. Two inset images represent the same cell infected by the wild-type strain at 200-min and 220-min time points. Note the cell rounding (at 140 min postinfection), absence of EGFP labeling, and rupture of cell membrane (at 220 min postinfection) for infection with the wild-type strain (WT). 3D images were constructed from z-planes. Bars, 10 μm. The film with the entire sequence is provided in Movie S1 in the supplemental material.

Genetic manipulations.

All deletion mutants were obtained by gene splice extension overlap (SOE) PCRs. The exlA_RGA gene was synthesized by GenScript and cloned in BglII-AvrII sites of pUC57. exlB-exlA_RGA was cloned into pSW196 between the EcoRI and SacI restriction sites, to obtain the plasmid pSW196exlB-exlA_RGA. For complementation of mutants, wild-type genes were amplified by PCR and cloned into the integrative plasmid pSW196 containing the arabinose-inducible promoter pBAD (52). The constructs in the pSW196 plasmid were transferred into P. aeruginosa strains by triparental mating. Primers used in PCR are listed in Table S3.

Cell culture and cytotoxicity assays.

The epithelial cell line A549 (ATCC CCL-185) was grown in 1× RPMI (Gibco Life Technologies) supplemented with 10% fetal bovine serum (FBS; Sigma) at 37°C and 5% CO₂. Cells were plated in 96-well plates (50,000 cells/well) and incubated for 1 night. Two hours before infection, RPMI medium was replaced by endothelial growth basal medium (EBM-2; Lonza Clonetics). The infection was done at a multiplicity of infection (MOI) of 10 (bacteria per eukaryotic cell). The level of cytotoxicity was determined by measuring the release of lactate dehydrogenase (LDH) using a cytotoxicity detection kit (Roche). Infection supernatants were sampled at 4 h postinfection. Negative controls were noninfected cells; positive controls were cells lysed by the addition of 200 µl of 2% Triton X-100. The optical density (OD) was measured at 492 nm. The extent of cytotoxicity was calculated by determining the percent cytotoxicity.

Time-lapse microscopy.

A549 cells were plated in 96-well plates (50,000 cells/well) in RPMI medium (1×) for a night. Before infection, the RPMI medium was replaced with EBM-2 and the cells were labeled with vital Hoechst stain (1 μg/ml) and YoPro-1 (1 μM) followed by infection with the bacteria at an MOI of 10. Cell acquisitions were made using the ArrayScan high-content system (Thermo Fisher) every 5 min. Analyses were performed by measuring the fluorescence intensity of each cell using the HCS studio analysis software. Slopes were calculated between the 5 last time points before a sharp increase of fluorescence. The box plot was made using SigmaPlot software (Systat Software Inc., San Jose, CA), and statistical analysis was done by the Kruskal-Wallis test.

For confocal spinning-disk microscopy, the A549-EGFP cells were grown in RPMI medium supplemented with 10% fetal calf serum (all from Lonza). Cells were seeded at 25,000 cells per well on Lab-Tek II 8-chambered (Dutscher Scientific) coverslips and used 24 h later. Cell membranes were labeled after incubation with WGA-Alexa 647 (Life Technologies) for 10 min at 37°C and then infected with bacteria at an MOI of 10 in EBM-2. Cell infections were observed by confocal spinning-disk microscopy for 4 h, with images recorded every 20 min. Successive planes in 3D stacks were taken every 0.5 μm. The 3D reconstruction was performed using the 3D viewer plug-in of ImageJ software.

Analysis of ExlA secretion.

Secretion of ExlA was monitored either in liquid LB cultures or under infection conditions. A 30-ml bacterial culture, in LB at an optical density at 600 nm (OD₆₀₀) of 1, was centrifuged at 6,000 rpm for 15 min. The supernatants (9 ml) were filtered, followed by addition of 90 µl of 2% sodium deoxycholate (DOC). After a 30-min incubation at 4°C, 900 µl of 100% trichloroacetic acid (TCA) was added and the proteins were allowed to precipitate overnight at 4°C followed by centrifugation at 15,000 × g for 15 min at 4°C. Pellets were resuspended in 100 µl of Laemmli loading buffer. For the analysis of proteins during infection of cultured cells, the bacteria were added to 100-mm petri dishes (at an MOI of 10) for 2 h, the supernatants were centrifuged, and the proteins were precipitated with TCA as described above. Samples were separated by SDS-PAGE (12%) and transferred onto polyvinylidene difluoride (PVDF) membranes for Western immunoblotting. Primary rabbit polyclonal antibodies were raised against either synthetic peptides of ExlA (6) or the recombinant C-terminal portion of the ExlA fragment, used at a 1:1,000 dilution. Anti-RpoA (Neoclone) and anti-FliC (39) were used at 1:2,000 and 1:1,000 dilutions, respectively. For the detection of the abundant RpoA protein, the bacterial lysates were further diluted 5-fold. Secondary antibodies were anti-rabbit-horseradish peroxidase (HRP) and anti-mouse-HRP (Sigma). The membranes were developed with Luminata Classico Western HRP (Millipore) substrate.

Hemolysis assay.

Human red blood cells (RBCs) were obtained from the Etablissement Français du Sang (EFS), Grenoble, France. RBCs were counted, and 5 × 10⁷ were incubated in 100 µl of EBM-2 with bacteria at the MOI of 10 or with 100 nM α-hemolysin from S. aureus, in 96-well plates. After centrifugation at 2,000 rpm and gentle agitation, the plates were incubated at 37°C for 90 min. The plates were centrifuged at 2,000 rpm at 4°C, allowing sampling of 100 µl of the supernatant. The optical density was measured at 560 nm. The negative controls were noninfected RBCs, and the positive control was obtained by inducing complete lysis by adding 100 µl of 1% SDS. For osmoprotection assays, RBCs were incubated with 30 mM osmoprotectants in EBM-2.

Expression and purification of ExlA variants.

ExlA genes were cloned in pET15b for Cter_ExlA and in pET28a for ExlA_noSP and ExlA_ΔCter. E. coli BL21(DE3) cells were used for the production of ExlA variants. Precultures were done in 30 ml of LB with ampicillin (Amp; 100 μg/ml) for pET15b and kanamycin (Km; 30 μg/ml) for pET28a at 25°C overnight. Expression of the recombinant proteins was induced in a 1-liter culture grown at 37°C by addition of 0.5 mM isopropyl-β-d-thiogalactopyranoside (IPTG) for 1.5 h. Bacteria were centrifuged, and the pellets were resuspended in 20 ml of IMAC10 buffer (25 mM Tris-HCl, 500 mM NaCl, 10 mM imidazole, pH 8) and lysed using a Microfluidizer M_110P2 (Microfluidics, Westwood, MA) at 15,000 lb/in² for 3 min. Soluble fractions were sampled after centrifugation at 30,000 rpm for 30 min at 4°C. Lysates were passed through a Ni²⁺ affinity column (HisTrap HP; GE Healthcare) using the Äkta purifier 10 instrument (GE Healthcare), and the bound proteins were eluted using a 20 mM to 200 mM step gradient of imidazole. Fractions were monitored by gel electrophoresis, pooled, and further purified on cation exchanger (HiTrap SP) and anion exchanger (HiTrap Q) columns (HiTrap; GE Healthcare). Protein concentrations in each fraction were quantified by bicinchoninic acid assay (BCA) using the Optima kit.

Liposome leakage assay.

L-R-phosphatidylcholine (PC), L-R-phosphatidylethanolamine (PE), L-R-phosphatidylserine (PS), and cholesterol (Ch) were from Avanti Polar. Lipids and liposomes were stored under nitrogen in chloroform at −20°C. Liposomes containing PS/PC (18%/82%), PC (100%), PC/PE (82%/18%), PC/PS/PE (72%/18%/10%), and PC/Ch (80%/20%) were prepared by evaporation of lipids in a rotatory evaporator, hydrated, and resuspended in 1 ml of 25 mM Tris-HCl, 250 mM NaCl, pH 8, containing 50 mM sulforhodamine B by six cycles of freezing-thawing. To obtain large unilamellar vesicles (LUVs) of 100 nm, liposomes were filtered through a 0.1-µm filter by extrusion and the nonencapsulated dye was removed with size exclusion chromatography on a PD10 column (Amersham) equilibrated with 25 mM Tris-HCl, 250 mM NaCl, pH 8. LUV disruption was monitored by an increase of fluorescence intensity upon dye dilution in the same buffer. Measurements were performed on a Jasco FP6500 fluorimeter with excitation set at 565 nm and emission at 586 nm in 2 ml of 10 µM LUV and recorded during 600 s after the addition of the protein. A 10 μM concentration of LUV was incubated with 10 nM protein in Tris or acetate buffer at different pHs. Fluorescence intensities were normalized considering the initial intensity before protein addition and the maximal intensity generated by the addition of Triton X-100 at the end of each kinetics measurement. The initial rates of dye release were derived from the first 10 s of the kinetics, using linear regression.

Construction of transposon library.

A transposon library was constructed in P. aeruginosa IHMA by mating with an E. coli donor carrying plasmid pBTK24 containing the mariner transposon with a Gm resistance cassette (53). P. aeruginosa IHMA was grown in LB at 42°C, and E. coli SM10λpir harboring pBTK24 was grown at 37°C with agitation for 18 h. Bacteria in 300 separate aliquots (100 μl) from each culture were combined, concentrated by centrifugation to 50 μl, and spotted on LB agar plates which were further incubated for 5 h at 37°C. Spots were scraped off, resuspended in 100 μl LB, and plated on LB agar plates containing Irgasan (25 mg/ml) and Gm (75 µg/ml) to isolate single colonies for testing in the cytotoxicity assay. The locations of the transposon insertions in candidate mutants were determined by semirandom PCR (primers are listed in Table S2) and sequencing.

Cellular screen for noncytotoxic mutants by high-content microscopy.

A549 cells were plated in 384-well plates (2,500 cells/well) in 1× RPMI with 10% FBS and incubated overnight. IHMA_tn single colonies from the transposon library were robotically transferred one by one in 96-well plates containing LB with 15 μg/ml Gm. The colonies were then transferred with 96-pin replicators into 384-well plates containing 30 µl of LB and grown statically for 18 h. The day of the screen, A549 cells were washed once with EBM-2, and EBM-2 containing vital Hoechst (1 μg/ml) and Draq7 (3 μM final concentration) stains was added. Infection of cells was initiated with a 384-pin replicator and run for 3 h at 37°C, and then the cells were fixed by adding an equal volume of 2% formaldehyde in phosphate-buffered saline (PBS) to each well. The image acquisition was done using a high-content microscope (ImageXpress Micro; Molecular Devices) at the ICCB Longwood facility, Harvard Medical School, Boston, MA. Analysis was done using MetaXpress software.

Twitching motility assay.

A determination of twitching motility by P. aeruginosa was carried out by inoculating bacteria at the interface between the plastic petri dish and LB agar (10 g/liter tryptone, 5 g/liter yeast extract, 10 g/liter NaCl, 1% agar). After 48 h of incubation at 37°C, agar was removed and the zone of twitching was revealed following Coomassie blue staining.

We thank Stewart Rudnicki from screening facilities at ICCB-Longwood, Harvard Medical School, Boston, MA; Emmanuelle Soleilhac from the Screening Center for Bioactive Molecules (CMBA), BIG, CEA-Grenoble; and Laëtitia Kurzawa, μLife microscopy facility (BIG, Grenoble), for their help in training, setting up experiments, and data analysis. P.B. thanks Xindan Wang for help in using the colony-transferring robot (Hudson) (Rudner Lab, Department of Microbiology and Immunobiology, Harvard Medical School, Boston, MA). Pseudomonas aeruginosa strain IHMA87 was kindly provided by International Health Management Association (IHMA; USA), and Serratia strains were obtained from J. Ewbank (CIML, Marseille, France). We thank Thomas Dougherty for critical reading of the manuscript.

This work was supported by grants from Laboratory of Excellence GRAL, ANR-10-LABX-49-01, and Agence national de Recherche ANR-15-CE11-0018-01. P.B. received the Ph.D. fellowship from the French Ministry of Education and Research, and E.R. received the Ph.D. fellowship from Laboratory of Excellence GRAL.

P.B., M.R., S.E., E.R., G.G., S.B., and E.F. performed experiments. P.H., G.G., E.F., and S.L. contributed reagents. P.B., S.E., P.H., S.L., E.F., and I.A. analyzed the data. P.B. and I.A. conceived the study and wrote the manuscript.