Diverse type VI secretion phospholipases are functionally plastic antibacterial effectors.

Membranes allow the compartmentalization of biochemical processes and are therefore fundamental to life. The conservation of the cellular membrane, combined with its accessibility to secreted proteins, has made it a common target of factors mediating antagonistic interactions between diverse organisms. Here we report the discovery of a diverse superfamily of bacterial phospholipase enzymes. Within this superfamily, we defined enzymes with phospholipase A1 and A2 activity, which are common in host-cell-targeting bacterial toxins and the venoms of certain insects and reptiles. However, we find that the fundamental role of the superfamily is to mediate antagonistic bacterial interactions as effectors of the type VI secretion system (T6SS) translocation apparatus; accordingly, we name these proteins type VI lipase effectors. Our analyses indicate that PldA of Pseudomonas aeruginosa, a eukaryotic-like phospholipase D, is a member of the type VI lipase effector superfamily and the founding substrate of the haemolysin co-regulated protein secretion island II T6SS (H2-T6SS). Although previous studies have specifically implicated PldA and the H2-T6SS in pathogenesis, we uncovered a specific role for the effector and its secretory machinery in intra- and interspecies bacterial interactions. Furthermore, we find that this effector achieves its antibacterial activity by degrading phosphatidylethanolamine, the major component of bacterial membranes. The surprising finding that virulence-associated phospholipases can serve as specific antibacterial effectors suggests that interbacterial interactions are a relevant factor driving the continuing evolution of pathogenesis.

Within proteobacterial genomes, predicted lipases are often encoded adjacent to homologs of the vgrG gene[6]. The VgrG protein is strongly associated with, and functionally important for, the cell contact-dependent T6S protein delivery pathway[7]. This pathway, which is distributed throughout all classes of Proteobacteria, can target both eukaryotic and bacterial cells; however, it is the specificity of its effectors that dictates the consequences of intoxication by the system. Known T6 effectors are few and include enzymes that either modify actin or degrade peptidoglycan – both domain-restricted molecules[8,9]. Thus, one would speculate that a barrier to the expansion or alteration of domain targeting would be the acquisition of a new effector or the evolution of one that is preexisting.

To understand the significance of the T6S-associated lipases we undertook an informatic approach to examine their genetic context, sequence, and phylogenetic distribution. This analysis uncovered 377 putative lipases comprising five divergent families (type VI lipase effector 1–5, Tle1-5) that share no detectable overall sequence homology (Fig. 1a and Supplementary Fig. 1–5). However, the families are united by a broad sporadic distribution pattern within Gram-negative bacteria and conserved putative catalytic motifs. Four of the families (Tle1-4) exhibit the GxSxG motif common in esterases and many lipases, while the fifth (Tle5) possesses dual HxKxxxxD motifs found in PLD enzymes (Fig. 1b)[1]. Outside of catalytic motifs, Tle1-4 members lack significant homology with known lipase enzymes, suggesting these proteins could represent previously uncharacterized diversity in the lipase superfamily.

Figure 1

Overview of the Tle superfamily

a, Evolutionary trees, genetic organization, and phylogenetic distribution of select Tle family members. Genes are coloured by their predicted protein product (blue, Tle proteins with a GxSxG catalytic motif; purple, Tle proteins with dual HxKxxxxD catalytic motifs; grey, VgrG proteins; yellow, putative periplasmic immunity proteins). Branch lengths are not proportional to evolutionary distance. Asterisks denote tle genes without an apparent adjacent vgrG gene. b, Domain organization of a single member of the GxSxG and dual HxKxxxxD catalytic classes of Tle proteins. Regions comprising these catalytic motifs are labeled in grey, and positions of all putative catalytic residues are denoted. Sequence logos were generated from alignments of the catalytic motif from Tle1-4 (GxSxG) and catalytic motifs from Tle5 (HxKxxxxD).

Our prior work has shown that antibacterial T6S effectors are encoded adjacent to cognate immunity genes, which are essential due to the self-targeting activity of the T6S apparatus[9,10]. Moreover, due to a direct inactivation mechanism, the localization of the immunity protein indicates the cellular compartment targeted by the effector. Examination of the genomic context of the putative lipase-encoding genes revealed each is found adjacent to an open reading frame encoding a predicted periplasmic protein (Fig. 1a). Thus, we hypothesize that contrary to prevailing views of bacterial lipase function, vgrG-associated lipase families could universally serve roles in interbacterial competition, possibly targeting phospholipids accessible from the periplasm. Consistent with our hypothesis, one of the putative lipase enzymes that we identified, V. cholerae VC1418 (Tle2VC, Supplementary Fig. 2), was recently found to act as an effector in amoeba defense and intraspecies bacterial competition[11]. Though the biochemical activity of Tle2VC was not elucidated, this suggests a capacity for Tle proteins to target a structure conserved in eukaryotes and bacteria.

To determine whether Tle2VC participates in interspecies bacterial antagonism, we tested its ability to provide fitness to V. cholerae in competition with E. coli. We observed V. cholerae strains lacking tle2 display a striking impairment in their capacity to kill E. coli, approaching that of a strain lacking T6S function (Fig. 2a and Supplementary Fig. 6). It is of note that a prior study probing the function of Tle2VC did not observe a contribution of the protein to fitness in an interspecies setting[11]. This study was performed with strain V52, in which T6S-associated genes exhibit constitutively high expression[12]. Therefore, a potential explanation for the apparent discrepancy is that in a hyperactive state the absence of one effector is not sufficient to diminish antibacterial activity to a measurable level.

Figure 2

Tle GxSxG-type proteins are antibacterial phospholipase effectors delivered by the T6SS

Error bars for all panels ± s.d. a, Outcome of growth competitions between the indicated V. cholera strains and E. coli. The ΔtssM strain is inactive for T6S. Asterisks denote competitive outcomes significantly different than those obtained with wild-type (P<0.05, n=3). b, Growth competition assays between the indicated B. thailandensis donor and recipient strains. The ΔclpV1 strain is inactivated for T6SS-1, required for Tle1BT export[10]. The parental strain for all experiments in this panel is ΔI2701-2703. Asterisks denote competition outcomes significantly different between indicated recipient strains (left) or indicated donor strains (right) (P<0.05, n=3). c, d, Enzymatic activity of the designated proteins against vesicles containing phospholipid derivatives with fluorescent moieties at the sn1 or sn2 positions (n=4). e, Enzymatic activity of MH–Tle1BT on sn2-labeled phospholipids as measured in (c) upon the addition of the indicated immunoprecipiate (arrow) (n=5). f, Representative cropped micrograph series displaying three propidium iodide (P.I.) uptake and subsequent lysis events in a growth competition experiment between B. thailandensis wild-type and a Tle1BT-sensitive recipient, ΔI2698-I2703. Each event spans two frames and is highlighted by arrowheads. The mask frames depict cell assignments made by gating cells based on fluorescence (black, donor; green, recipient; red, P.I.-positive). g, Quantitation of P.I. staining events from B. thailandensis growth competitions using automated custom software. The recipient strain was the same as used in (f). Shading indicates counting error.

Tle2 represents only one of four divergent GxSxG families within the broader superfamily. As a first step toward understanding the functional significance of other GxSxG families, we examined B. thailandensis BTH_I2698 (Tle1BT, Fig. 1 and Supplementary Fig. 1), which we previously demonstrated to be a substrate of an antibacterial T6SS[10]. The tle1 gene is found adjacent to genes encoding two homologous periplasmic lipoproteins, I2699 and I2700, which we posited could serve as Tle1BT immunity proteins. Additionally, tle1, I2699, and I2700 appear to have been subject to a duplication event, with homologs of all three genes present immediately upstream (Supplementary Fig. 7). To simplify our analysis, we generated a mutant strain lacking one copy of this duplicated region. Using labeled derivatives of this strain co-cultured under T6S-conducive conditions, we found that recipient strains lacking tle1 and its putative immunity determinants exhibit significantly decreased fitness in competition with donor strains possessing tle1 and a functional T6SS, and that expression of I2699 in the recipient strain was necessary and sufficient to restore competitive fitness (Fig. 2b). These data show that Tle1BT is an antibacterial effector delivered between cells by T6S, and that I2699, henceforth referred to as Tli1BT (type VI secretion lipase immunity 1), protects against Tle1BT.

Having demonstrated that members of two GxSxG Tle families function as antibacterial T6S effectors, we next sought to investigate their biochemical activity. To characterize Tle1BT and Tle2VC, we purified the proteins and catalytic nucleophile substitution mutant derivatives (Tle1BT(S267A) and Tle2VC(S371A)) as N-terminal fusions to hexahistidine-tagged maltose binding protein (MH–), which we found necessary to generate and maintain soluble protein (Supplementary Fig. 8 and 9). Importantly, Tle1-4 possess a Ser-Asp-His catalytic triad utilized by a diversity of esterase enzymes, including thioesterases, acetylesterases, and assorted lipase and phospholipases[1]. Given this wide range of potential activities, we initially confirmed general esterase activity of MH–Tle1BT and MH–Tle2VC by demonstrating that these effectors, but not their catalytic substitution mutants, hydrolyze a model substrate, polysorbate 20 (Supplementary Fig. 10). Next we asked whether MH–Tle1BT or MH–Tle2VC possess phospholipase activity. Using vesicle substrates doped with fluorescent phospholipid derivatives, we determined that MH–Tle1BT acts specifically as a PLA2, and MH–Tle2VC as a PLA1 (Fig. 2c,d). Linking these activities to the antibacterial phenotypes we observed associated with the proteins in vivo, neither Tle1BT(S267A) nor Tle2VC(S371A), both catalytically inactive, serve as antibacterial effectors (Fig. 2a,b). Moreover, we found that the PLA2 activity of MH–Tle1BT is robustly inhibited by the addition of its immunity protein, Tli1BT (Fig. 2e).

If GxSxG family Tle proteins serve as antibacterial T6SS phospholipases, we reasoned that their activity against sensitive recipients should correlate with an increase in cellular permeability. To test these predictions, we performed single-cell measurements of propidium iodide (P.I.) uptake within interbacterial competitions of B. thailandensis. Consistent with our hypotheses, the lack of Tle1 immunity within cells corresponded to significantly increased P.I. uptake (Fig. 2f and Supplementary Videos 1–3). Using automated cell identity and tracking algorithms[13], we further demonstrated that the increase in P.I. uptake depended upon direct contact with donor cells possessing a functional T6SS (Fig. 2g and Supplementary Fig. 11).

With our data validating members of two GxSxG families as antibacterial phospholipase effectors, we explored whether these findings could be extended to the HxKxxxxD family (Tle5). This catalytic motif is strongly indicative of PLD activity[1], which has heretofore not been associated with an antibacterial enzyme. We choose P. aeruginosa PldA, henceforth referred to as Tle5PA (Fig. 1 and Supplementary Fig. 5), as a representative Tle5 family member. We began our study by confirming the enzymatic activity of the protein, as its function was previously studied in the context of cellular extracts[3]. Consistent with prior observations, Tle5PA catalyzes the release of choline from phosphatidylcholine (PC), in a manner dependent upon a predicted catalytic histidine residue (His855) (Fig. 3a and Supplementary Fig. 12). Under similar conditions neither Tle1BT nor Tle2VC showed appreciable activity in this assay, underscoring the diverse substrate specificity within the Tle superfamily.

Figure 3

Tle5PA is an HxKxxxxD-type interspecies antibacterial phospholipase effector delivered by the H2-T6SS of P. aeruginosa

Error bars for all panels ± s.d. a, PC-specific PLD activity of the indicated proteins against mixed lipid vesicles (n=3). b, Lysis of recipient strains grown in co-culture with the indicated donor strains. Asterisks mark experiments wherein recipient lysis is significantly different between indicated recipients (left), or between indicated donors (right) (P<0.05, n=3). c, Competitive growth of indicated P. aeruginosa strains against P. putida under T6SS-conducive conditions. Asterisks denote competition outcomes significantly different than those obtained with wild-type P. aeruginosa (P<0.05, n=3). d, Summary of phospholipid profiles of the indicated P. aeruginosa strains. Statistical significance noted (n=4, * P<0.01, ** P <0.001). e, Generalized schematic of a phospholipid indicating the activities defined in this study.

A candidate Tle5PA periplasmic immunity protein is not readily apparent, as the adjacent gene, PA3488, is predicted to encode a cytoplasmic protein. However, expression of PA3488 from a second, upstream, predicted start site yields a periplasmically-localized protein, henceforth referred to as Tli5PA, that binds specifically to Tle5PA (Supplementary Fig. 13). To probe the role of Tle5PA and Tli5PA in interbacterial interactions, we generated a lysis reporter strain bearing a deletion of the tle5PA tli5PA bicistron. Lysis of this strain was highly elevated when co-cultured with a wild-type, but not a Δtle5PA donor strain (Fig. 3b). Additionally, expression of tli5 in the recipient was sufficient to protect from Tle5PA-dependent lysis. Together, these data demonstrate Tle5PA acts as an antibacterial toxin and that Tli5PA is its cognate immunity determinant.

The P. aeruginosa genome encodes three T6SSs, the H1-3-T6SSs. The H1-T6SS is the only system with known substrates and a demonstrated role in interbacterial interactions[14]. To define the T6SS involved in Tle5PA transport, we constructed strains bearing individual in-frame deletions of the critical ATPase genes, clpV1-3, associated with the H1-3 systems, respectively. Specific inactivation of the H2-T6SS in a donor strain abrogated Tle5PA-dependent toxicity, indicating that this system is responsible for Tle5PA delivery (Fig. 3b).

The finding that Tle5PA transits the H2-T6S pathway is interesting in light of data that implicate this T6SS as a virulence factor in plant, mammalian cell culture, worm, and mouse models of infection[4,5]. To more thoroughly explore the role of Tle5PA and the H2-T6SS in interbacterial interactions, we measured their influence on competition outcomes between P. aeruginosa and a model T6S target, P. putida[9]. Our results showed that both Tle5PA and the H2-T6SS significantly contribute to the fitness of P. aeruginosa in interspecies competition under T6S-conducive conditions (Fig. 3c and Supplementary Fig. 14). These findings show that Tle5PA is a potent antibacterial effector delivered by the H2-T6SS.

While our data thus far show that Tle1BT, Tle2VC, and Tle5PA possess phospholipase activity in vitro, this did not allow us to definitively assign the toxic consequences of these effectors to membrane destruction. The phospholipase activity of the effectors could be accessory to a second toxicity mechanism found in these large, multidomain proteins. To resolve this remaining ambiguity concerning Tle function, we focused our studies on Tle5PA. Since a mixture of healthy and intoxicated cells could complicate our measurements, we decided to assay Tle5PA effects in self-intoxicating monocultures of Δtli5, wherein each cell serves both as a donor and a sensitive recipient. As expected, this strain exhibited increased membrane permeability in a manner dependent on an active H2-T6SS and Tle5PA (Supplementary Fig. 15).

Under conditions promoting intercellular delivery of Tle5PA,we harvested lipids of both non-intoxicated (wild-type) and intoxicated (Δtli5) cells and quantified their phospholipid composition using mass spectrometry. This analysis revealed that the unchecked action of Tle5PA leads to a severely perturbed membrane phospholipid composition. Strikingly, phosphatidic acid (PA), a product of PLD activity and a minor constituent of wild-type membranes (0.17%), was present at 8.1% in Δtli5PA – a 48-fold enrichment (Fig. 3d and Supplementary Table 1). The increased PA appeared to derive primarily from PE, as it underwent a concomitant decrease of similar magnitude. Finally, we noted that phosphatidylglycerol (PG) increased slightly in Δtli5 relative to the wild-type. We speculate this latter result either derives from a compensatory effect or from Tle5PA activity against cardiolipin, a minor component of P. aeruginosa membranes not detectable by the analysis method we used. Taken together, these data strongly suggest that Tle5PA-imposed cell death occurs through PA accumulation via PLD activity, primarily directed against PE. The precise physiological consequences of massive PA accumulation in bacterial cells are not known, however the strong negatively charged character of the molecule is likely to have a detrimental impact on both integral and peripheral membrane-associated proteins. It is known that PA induces membrane curvature that can promote fusion and fission events[15]; therefore, Tle5PA activity might also lead to generalized membrane destabilization, membrane blebbing and depolarization. Interestingly, the in vivo specificity of Tle5PA for PE, the major phospholipid constituent of most bacterial membranes, affords P. aeruginosa the capacity to use this enzyme against a vast array of competitors.

The discovery of T6SS-delivered phospholipase effectors has many implications. Critically, their biochemical activity does not intrinsically limit their toxicity to bacterial cells (Fig. 3e). Indeed, two specificities now ascribed to Tle superfamily members, PLD and PLA2, are both highly represented in host cell-targeting bacterial toxins[2]. As these effectors are found in numerous established and emerging opportunistic pathogens, our work highlights the need to understand the biochemical, genetic, and evolutionary basis of inter-domain targeting by the T6SS. Such knowledge may ultimately become a component of a larger strategy to develop predictive algorithms for the evolution of bacterial pathogens. In addition, our findings add a new dimension to our understanding of the mechanisms employed during bacterial competition. Based on our data it appears that membrane targeting evolved independently on multiple occasions as an antibacterial strategy. This convergent evolution underscores the susceptibility of the bacterial membrane to attack, a theme mirrored by the prior observation that bacteriolytic T6S effectors likewise degrade an essential, conserved bacterial structure[9]. The continued discovery of antibacterial effectors promises to illuminate additional vulnerabilities of the bacterial cell, and thus may aid our efforts to define promising therapeutic targets.

Online-only Methods

Bacterial strains and growth conditions

B. thailandensis strains used in this study were derived from the sequenced strain E264[16]. B. thailandensis strains were grown on either Luria-Bertani media (LB), or the equivalent lacking additional NaCl (LB low salt (LB-LS): 10 g bactopeptone and 5 g yeast extract per liter) at 37°C supplemented with 200 μg ml−1 trimethoprim and 25 μg ml−1 irgasan where necessary. For introducing in-frame deletions, B. thailandensis was grown on M9 minimal medium agar plates with 0.4% glucose as a carbon source and 0.1% (w/v) p-chlorophenylalanine for counter-selection[21]. V. cholerae strains used in this study were derived from the O1 El Tor strain A1552[18]. V. cholerae was grown on LB or LB with 340 nM NaCl at 37°C or 30°C supplemented with 100 μg/ml rifampin, 100μg/ml carbenicillin and stated concentrations of arabinose as needed. In order to introduce in-frame deletions V. cholerae was grown on LB supplemented with 10% (w/v) sucrose at 30 °C for counter-selection[22,23]. P. aeruginosa strains used in this study were derived from the sequenced strain PAO1[17]. P. aeruginosa strains were grown on LB at 37 °C supplemented with 25 μg ml−1 irgasan, 30 μg ml− gentamycin, and stated concentrations of IPTG as required. To generate in-frame deletions P. aeruginosa was grown on LB-LS supplemented with 5% (w/v) sucrose at 30 °C for counter-selection[24]. For intra- and inter-species competition P. aerguinosa was grown on synthetic cystic fibrosis sputum media (SCFM) at 23°C[19]. P. putida used in this study was the sequenced strain, KT2440[25]. P. putida was grown on LB at 30°C or on SCFM at 23°C. E. coli strains included in this study included DH5α for plasmid maintenance and production of Tli1BT immunoprecipitate, SM10 λpir for conjugal transfer of plasmids into B. thailandensis, V. cholerae, and P. aeruginosa, MC4100 for competition assays with V. cholerae, BL21(DE3) plysS for Tle5PA immunoprecipitation studies, and Shuffle T7 plysY Express(New England Biolabs), for purification of Tle proteins. All E. coli strains were grown on LB or 2xYT at 37 °C supplemented with 150 μg ml−1 carbenicillin, 50 μg ml−1 kanamycin, 30 μg ml−1 chloramphenicol, 200 μg ml−1 trimethoprim, 50μg/ml streptomycin, 15μg/ml gentamycin, 0.1% rhamnose, and 100 mM IPTG as needed.

DNA manipulations

The creation, maintenance, and transformation of plasmid constructs followed standard molecular cloning procedures. All primers used in this study were obtained from Integrated DNA Technologies. DNA amplification was carried out using either Phusion (New England Biolabs) or Mangomix (Bioline). DNA sequencing was performed by Genewiz Incorporated. Restriction enzymes were obtained from New England Biolabs. SOE PCR was performed as previously described[26].

Plasmid construction

Plasmids used for expression in this study were pET28b:His6-MBP-TEV-His6[27], pET22b+ (Novagen), and pSCrhaB2[28] for E. coli, pPSV35CV[29] for P. aeruginosa, and pBAD24[30] for V. cholerae. Complementation in B. thailandensis was performed using the Tn7-based integration vector pUC18T-miniTn7T-Tp::PS12[31]. In-frame deletions were generated utilizing the suicide vectors pJRC115 for B. thailandensis[21], pVCD442 for V. cholerae[32], and pEXG2 for P. aeruginosa[24]. For the production of deletion constructs either 600 bp (B. thailandensis and P. aeruginosa) or 500 bp (V. cholerae) regions flanking the deletion were amplified, ligated together using SOE PCR, and subsequently cloned into pJRC115, pEXG2, or pVCD442 respectively. To generate the tle1 mutation construct, 600 bp regions flanking the mutation with an additional overlapping extension consisting of the desired mutation were amplified and ligated together using SOE PCR and subsequently cloned into pJRC115. For B. thailandensis complementation constructs genes were amplified along with predicted ribosomal-binding sites and cloned into pUC18T-miniTn7T-Tp::P12. For P. aeruginosa complementation and expression constructs, genes were amplified with their native ribosomal binding sites into pPSV35CV with a 3′ fusion to the VSV–G (vesicular stomatitis virus glycoprotein) epitope tag. To further generate the pPSV35CV::tle5 and –V constructs, the entire tle5 gene was amplified from pPSV35CV::tle5 and SOE PCR was used to introduce the desired base pair mutations. For pSCrhaB2 E. coli expression constructs and pBAD24 V. cholerae expression and complementation constructs, genes were cloned downstream of the optimized ribosomal binding site already present in these vectors with a fusion to a 3′ VSV–G-linker. To further generate the pBAD24::tle2–V construct, the entire tle2 gene was amplified from pBAD24::tle2–V and SOE PCR was used to introduce the desired base pair mutations. This product was subsequently cloned into pBAD24. For the Tle purification constructs tle genes were amplified and cloned into pET28b:His6-MBP-TEV-His6 to generate an N-terminal fusion to an MBP protein and a hexahistadine purification tag. SOE PCR was then used to generate the desired catalytic nucleophile substitution mutants. For the Tle5PA periplasmic expression construct, tle5 was amplified and cloned into pET22b+ to generate an N-terminal fusion to the PelB leader peptide and a C-terminal fusion to a hexahistidine epitope tag.

Informatic identification of Tle proteins

All sequences were obtained from NCBI (http://www.ncbi.nlm.nih.gov), and Genbank accession numbers for all Tle proteins identified in this study are found in Supplementary Fig. 1–5. BTH_I2698 from B. thailandensis E264, PA0260, PA1510, PA3487, and PA5089 from P. aeruginosa PAO1, and VC1418 from V. cholerae V52, all encoded adjacent to vgrG genes, were identified as putative lipases utilizing the PHYRE 2 structural prediction server[33]. Using the amino acid sequences of these predicted lipases, blastp analyses were performed against the non-redundant protein database (ftp://ftp.ncbi.nih.gov/blast/db/) to identify unique instances of their homologs. Homology identified by the blast server was used to distribute these proteins into five distinct Tle families. Each family was aligned using the MUSCLE algorithm and phylogenetic trees were generated using the PHYML 3.0 method with bootstrap analysis of 1000 replicates[34,35]. Proteins encoded by the genes shown in Figure 1a were analyzed for subcellular localization utilizing the SignalP 3.0 and TMHMM 2.0 servers, and VgrG proteins were identified utilizing blastp[36,37]. Regions depicted in Figure 1a were extracted based on boundaries defined by the presence of a tle, tli, or vgrG gene. Fig. 1b catalytic residues were determined both by PHYRE 2 structural alignment with known lipase enzymes and conservation of those residues within the Tle family alignments. Sequence logos were generated from a manual alignment of conserved catalytic motifs utilizing Geneious software.

Western blot analyses

Whole cell fractions were prepared as described previously[29]. Anti-RNA polymerase, anti-VSV–G, anti-beta-lactamase, anti-His5, and anti-CRP Western blot analyses were performed utilizing previously-defined methods[9,23,38]. To analyze the expression of epitope-tagged Tle2VC and Tle2VCS371A in V. cholerae, cells were grown in LB medium at 37 °C to an optical density at 600 nm (OD600) of 0.5, induced with 0.0002% (w/v) arabinose, and then harvested at a final OD600 of 2.0. To analyze the expression of epitope-tagged Tle5PA, Tle5PA H167R, and Tle5PA H855R, in P. aeruginosa, cells were grown in LB medium supplemented with 1mM IPTG at 37 °C and harvested at an OD600 of 1.0. Subcellular localization of epitope-tagged Tle1PA and Tli1PA in P. aeruginosa was performed identically to previous localization studies of Tsi1 and Tsi3[9,39]. For immunoprecipitation experiments, BL21(DE3) plysS cells co-expressing periplasmic hexahistidine-tagged Tle5PA from a pet22b+ vector and VSV–G tagged immunity proteins from pSCrhaB2 vectors were pelleted and resuspended in lysis buffer (20 mM Tris-Cl pH 7.5, 50 mM KCl, 8.0% (v/v) glycerol, 1.0% (v/v) triton, supplemented with DNase I (Roche), lysozyme (Roche), and 200μM PMSF). Cells were disrupted by sonication and the solution clarified by centrifugation. A sample of supernatant was then taken for analysis of total protein. The remainder of the supernatant was incubated with anti-VSV–G agarose beads (Sigma) for 1 h at 4°C. Beads were washed four times with IP-wash buffer (100 mM NaCl, 25 mM KCl, 0.1% (v/v) triton, 20 mM Tris-Cl pH 7.5, and 2% (v/v) glycerol). Proteins were removed from beads with SDS loading buffer (125 mM Tris, pH6.8, 2% (w/v) 2-mercaptoethanol, 20% (v/v) glycerol, 0.001% (w/v) bromophenol blue and 4% (w/v) SDS) and analyzed by Western blot.

Bacterial competition experiments

Burkholderia competition experiments were performed as described previously[10]. Recipient strains (Fig. 2b, left), or donor strains (Fig. 2b, right) were labeled with a GFP-expression constructed integrated into the attTn7 site, allowing the disambiguation of donor and recipient colonies through fluorescence imaging[40]. For V. cholerae competition experiments with E. coli, both strains were grown to an OD600 of 0.5 in LB before being mixed 1:1 by volume. This mixture was then spotted on a nitrocellulose membrane on a 1.5% (w/v) agar LB plate containing 300 mM NaCl and 0.002% (w/v) arabinose. Competitions were incubated for 5 h at 37°C. Cells were then harvested and competitions analyzed. Initial and final colony-forming units (CFUs) of V. cholerae and E. coli were enumerated on LB plates supplemented with rifampin and streptomycin respectively. For P. aeruginosa competitions with P. putida, strains were grown overnight on solid LB media at 37 °C (P. aeruginosa) or 30 °C (P. putida) and resuspended in water to an OD600 of 0.3. Cells were mixed 1:1 and spotted on 1.5% (w/v) agar SCFM media plates, or inoculated into liquid media of the same. After 23 h of incubation at 23°C, a temperature previously demonstrated conducive to H2-T6SS and Tle5PA expression under in vitro conditions[41], cells were harvested and relative numbers of bacteria determined. Both initial and final counts of P. aeruginosa and P. putida were determined by plate counts. P. aeruginosa self-intoxication assays were performed under identical conditions to solid media competition assays, save for the addition of 1 mM IPTG. After 23 h of growth, cells were stained with 5 μg ml−1 propidium iodide in PBS pH 7.0 for 10 minutes and washed prior to fluorescence measurements at an excitation/emission of 535/617 nm. Values shown were corrected for cellular density as measured by OD600. Competition results for B. thailandensis and P. aeruginosa experiments are the change in ratio of donor cells to recipient cells, competition results from V. cholerae represent the final ratio alone. Data from all competitions were analyzed by a two-tailed Student’s T-test, and data from monoculture experiments were analyzed by a one-tailed Student’s T-test for a significant increase in P.I. staining.

Enzymatic assays of lipase activity

The hydrolysis of polysorbate 20 was measured as described by Tigerstrom and Stelmaschuk[42]. These experiments were performed at 28°C at a final enzyme concentration of 60 nM in a buffer consisting of 20 mM Tris-Cl pH 7.2, 100 mM NaCl, 3 mM CaCl2 and 2% (v/v) polysorbate 20. Fluorescence assays for phospholipase A activity were performed utilizing PED-A1 (sn1-labeled) and PED6 (sn2-labeled) fluorescent substrates according to manufacturer’s directions (Invitrogen). Activity of Tle1PA and Tle2VC on these substrates was measured at an enzyme concentration of 300 nM (Tle1PA) or 30 nM (Tle2VC) at 28°C. For Tle1BT-inhibition assays, immunoprecipitate was obtained as detailed under Western blot analyses from E. coli DH5α bearing a pSCrhaB2::tli1–V expression construct or the equivalent empty vector control, with the modification that proteins were eluted from anti-VSV–G agarose beads by the addition of VSV–G peptide at a concentration of 100 μg ml−1 and no PMSF was used. After the addition of immunoprecipitate to Tle1BT enzymatic reactions, samples were incubated for four minutes after which the first reading was normalized to the measurement immediately prior to treatment. Fluorescent assays for phospholipase D activity were performed by measuring the production of peroxide by choline oxidase through the generation of the fluorescent molecule resorufin from Amplex red reagent (Invitrogen) according the manufacturer’s directions with the following modifications: reactions were performed in a buffer consisting of 50 mM Tris-Cl pH 7.2, 100 mM NaCl, 5 mM CaCl2, and 2 mM MgCl2, and vesicles consisting of equal amounts dioleoylphosphatidylcholine and dioleoylphosphatidylglycerol were used as a substrate at a final reaction concentration of 16.7 μM for each lipid species. Activity was measured at an enzyme concentration of 130 nM at 28°C. In all assays fluorescent values were corrected for fluorescence as measured in a buffer-only control.

Competitive lysis assays

The lysis of P. aeruginosa reporter strains was determined by the relative partitioning of LacZ to the supernatant. Lysis reporter strains were generated by the chromosomal integration of a previously-described miniCTX vector containing lacZ under the expression of a constitutive promoter[43]. Lysis reporter strains and unmarked donor strains were grown overnight on solid LB media at 37°C and then resuspended in water to an OD600 of 0.3. Donor and recipient strains were mixed 1:1 and spotted on 1.5% (w/v) agar SCFM plates supplemented with 1 mM IPTG and incubated at 23°C for 23 h. Relative levels of supernatant LacZ activity as compared to total LacZ activity were then determined as previously described[20]. Data were analyzed using a two-tailed Student’s T-test.

Microscopic analyses of interbacterial competitions

Time-lapse fluorescence microscopy sequences were acquired with a Nikon Ti-E inverted microscope fitted with a 60X oil objective, automated focusing (Perfect Focus System, Nikon), a Xenon light source (Sutter Instruments), a CCD camera (Clara series, Andor), and a custom environmental chamber. NIS Elements (Nikon) was used for automated image acquisition. Overnight cultures of recipient (B. thailandensis ΔBTH_I2698-I2703 attTn7::gfp) and donor (either B. thailandensis wild-type, ΔBTH_I2698 ΔBTH_I2701-3, or ΔBTH_I2598) strains were mixed 1:1 and diluted 2-fold with LB. The resulting bacterial suspension (~2 μL) was spotted onto growth pads made with LB broth, 2.5% (w/v) agarose, 0.2% (w/v) sodium nitrate, and 2.5 μg mL−1 propidium iodide. Automated image acquisition was performed at 5-min intervals for 6–8 h at 30°C. Cell identification, cell linking, and donor-contact analyses were performed using customized Matlab-based software (2012a, Mathworks) as described previously[13]. Donor (unlabeled) and recipient (GFP-labeled) populations were identified using an empirically determined green fluorescence gate. A P.I. uptake event was defined as the first frame in which a cell achieved an empirically determined mean red fluorescence intensity threshold. Counting error was calculated as the square root of measurable events. Results represent two fields of view from a single experiment; each experiment was independently repeated at least three times. Videos generated from cropped regions of the three growth competition experiments depicted are provided (Supplementary Videos 1–3).

Protein purification

For purification, Tle proteins were expressed from pET28b:His6-MBP-TEV-His6 in Shuffle T7 pLysY Express cells (New England Biolabs). Proteins were purified to homogeneity using nickel chromatography followed by size-exclusion chromatography using previously-reported methods, with the exception that reducing agents were excluded[44].

Lipidomic analyses

Wild-type and tli5 mutant P. aerguinosa strains were grown as 20 individual 10 μl spots on 1.5 % (w/v) agar SCFM plates for 23 h at 23°C. These spots were then resuspended in PBS and lipids were extracted using the Bligh-Dyer method[45]. Purified lipid samples were analyzed for PE, PC, PG, and PA content by the Kansas State Lipidomics Research Center. An automated electrospray ionization-tandem mass spectrometry approach was used, and data acquisition and analysis were carried out as described previously[46,47] with modifications. The lipid samples were dissolved in 1 ml chloroform. An aliquot of 50 μl of extract in chloroform was used. Precise amounts of internal standards, obtained and quantified as previously described[48], were added in the following quantities (with some small variation in amounts in different batches of internal standards): 0.6 nmol di12:0-PC, 0.6 nmol di24:1-PC, 0.6 nmol 13:0-lysoPC, 0.6 nmol 19:0-lysoPC, 0.3 nmol di12:0-PE, 0.3 nmol di23:0-PE, 0.3 nmol 14:0-lysoPE, 0.3 nmol 18:0-lysoPE, 0.3 nmol di14:0-PG, 0.3 nmol di20:0(phytanoyl)-PG, 0.3 nmol di14:0-PA, and 0.3 nmol di20:0(phytanoyl)-PA. The sample and internal standard mixture was combined with solvents, such that the ratio of chloroform/methanol/300 mM ammonium acetate in water was 300/665/35, and the final volume was 1.4 ml. Unfractionated lipid extracts were introduced by continuous infusion into the ESI source on a triple quadrupole MS/MS (4000QTrap), Applied Biosystems, Foster City, CA). Samples were introduced using an autosampler (LC Mini PAL, CTC Analytics AG, Zwingen, Switzerland) fitted with the required injection loop for the acquisition time and presented to the ESI needle at 30 μl/min. Sequential precursor and neutral loss scans of the extracts produce a series of spectra with each spectrum revealing a set of lipid species containing a common head group fragment. Lipid species were detected with the following scans: PC and lysoPC, [M + H]+ ions in positive ion mode with Precursor of 184.1 (Pre 184.1); PE and lysoPE, [M + H]+ ions in positive ion mode with Neutral Loss of 141.0 (NL 141.0); PG, [M + NH4]+ in positive ion mode with NL 189.0 for PG; and PA, [M + NH4]+ in positive ion mode with NL 115.0. The collision gas pressure was set at 2 (arbitrary units). The collision energies, with nitrogen in the collision cell, were +28 V for PE, +40 V for PC, +25 V for PA, and +20 V for PG. . Declustering potentials were +100 V for all lipids. Entrance potentials were +15 V for PE and +14 V for PC, PA, and PG. Exit potentials were +11 V for PE and +14 V for PC, PA, and PG. The scan speed was 50 or 100 u per sec. The mass analyzers were adjusted to a resolution of 0.7 u full width at half height. For each spectrum, 9 to 150 continuum scans were averaged in multiple channel analyzer (MCA) mode. The source temperature (heated nebulizer) was 100 °C, the interface heater was on, +5.5 kV were applied to the electrospray capillary, the curtain gas was set at 20 (arbitrary units), and the two ion source gases were set at 45 (arbitrary units). The background of each spectrum was subtracted, the data were smoothed, and peak areas integrated using a custom script and Applied Biosystems Analyst software, and the data were isotopically deconvoluted. The first set of mass spectra were acquired on the internal standard mixture only. Peaks corresponding to the target lipids in these spectra were identified and molar amounts calculated in comparison to the two internal standards on the same lipid class. To correct for chemical or instrumental noise in the samples, the molar amount of each lipid metabolite detected in the “internal standards only” spectra was subtracted from the molar amount of each metabolite calculated in each set of sample spectra. The data from each “internal standards only” set of spectra was used to correct the data. Values expressed are the percentage of the total polar lipid signal detected. Statistical significance analyzed by a two-tailed Student’s T-test.