Type VI secretion delivers bacteriolytic effectors to target cells.

Peptidoglycan is the major structural constituent of the bacterial cell wall, forming a meshwork outside the cytoplasmic membrane that maintains cell shape and prevents lysis. In Gram-negative bacteria, peptidoglycan is located in the periplasm, where it is protected from exogenous lytic enzymes by the outer membrane. Here we show that the type VI secretion system of Pseudomonas aeruginosa breaches this barrier to deliver two effector proteins, Tse1 and Tse3, to the periplasm of recipient cells. In this compartment, the effectors hydrolyse peptidoglycan, thereby providing a fitness advantage for P. aeruginosa cells in competition with other bacteria. To protect itself from lysis by Tse1 and Tse3, P. aeruginosa uses specific periplasmically localized immunity proteins. The requirement for these immunity proteins depends on intercellular self-intoxication through an active type VI secretion system, indicating a mechanism for export whereby effectors do not access donor cell periplasm in transit.

Competition among bacteria for niches is widespread, fierce and deliberate. These organisms elaborate factors ranging in complexity from small diffusible molecules, to exported proteins, to multicomponent machines, in order to inhibit the proliferation of rival cells[1,2]. A common target of such factors is the peptidoglycan cell wall[3-6]. The conserved, essential, and accessible nature of this molecule makes it an Achilles heel of bacteria.

The T6SS is a complex and widely distributed protein export machine capable of cell contact-dependent targeting of effector proteins between Gram-negative bacterial cells[7-10]. However, the mechanism by which effectors are delivered via the secretory apparatus, and the function(s) of the effectors within recipient cells, have remained elusive. Current models of the T6SS derive from the observation that several of its components share structural homology to bacteriophage proteins[11-13]; it has been proposed that target cell recognition and effector delivery occur in a process analogous to bacteriophage entry[14].

The observation that T6S can target bacteria was originally made through studies of the hemolysin co-regulated protein secretion island I (HSI-I)-encoded T6SS (H1-T6SS) of P. aeruginosa, which exports at least three proteins, Tse1-3[7,13]. These proteins are unrelated to each other and lack significant primary sequence homology to characterized proteins. One substrate, Tse2, is toxic by an unknown mechanism in the cytoplasm of recipient cells lacking Tsi2, a Tse2-specific immunity protein. Here we show that Tse1 and Tse3 are lytic enzymes that degrade peptidoglycan via amidase and muramidase activity, respectively. Unlike related enzymes associated with other secretion systems[15], these proteins are not required for the assembly of a functional secretory apparatus. Instead, Tse1 and Tse3 function as lytic antibacterial effectors that depend upon T6S to breach the barrier imposed by the Gram-negative outer membrane.

Contacting P. aeruginosa cells actively intoxicate each other with Tse1 and Tse3. However, the peptidoglycan of P. aeruginosa is not inherently resistant to the activities of these enzymes. To protect itself, the bacterium synthesizes immunity proteins – type VI secretion immunity 1 and 3 (Tsi1 and Tsi3) – that specifically interact with and inactivate cognate toxins in the periplasm. Orthologs of tsi1 and tsi3 appear restricted to P. aeruginosa, therefore the species is able to exploit the H1-T6SS to target closely related organisms that are likely to compete for overlapping niches, while minimizing the fitness cost associated with self-targeting.

Tse1 and Tse3 are lytic enzymes

To identify potential functions of Tse1 and Tse3, we searched their sequences for catalytic motifs using structure prediction algorithms[16]. Interestingly, motifs present in peptidoglycan degrading enzymes were apparent in both proteins (Fig. 1a and Supplementary Fig. 1). Tse1 contains invariant catalytic amino acids present in cell wall amidases (DL-endopeptidases)[17], whereas Tse3 possesses a motif that includes a catalytic glutamic acid found in muramidases[18,19].

Figure 1

Tse1 and Tse3 are lytic proteins belonging to amidase and muramidase enzyme families

a. Genomic organization of tse1 and tse3 and their homology with characterized amidase and muramidase enzymes, respectively. Highly conserved (boxed) and catalytic (starred) residues of the respective enzyme families are indicated. SWISS-PROT entry names for the proteins shown are: Tse1 (Q9I2Q1_PSEAE), Spr (SPR_ECOLI), P60 (P60_LISIN), Tse3 (Q9HYC5_PSEAE), GEWL (LYG_ANSAN), Slt70 (SLT_ECOLI). See Supplementary Fig. 1 for full alignments.

b,d. Partial HPLC chromatograms of sodium borohydride-reduced soluble E. coli peptidoglycan products resulting from (b) digestion with Tse1 and subsequent cleavage with cellosyl or (d) digestion with Tse3 alone. Peak assignments were made based on MS; predicted structures are shown schematically with hexagons and circles corresponding to sugars and amino acid residues, respectively. Reduced sugar moieties are shown with grey fill. Full chromatograms and MS data are provided in the supplement (Supplementary Fig. 3 and Supplementary Table 1).

c. Simplified representation of Gram-negative peptidoglycan showing cleavage sites of Tse1 and Tse3 based on data summarized in b and d.

e. Growth in liquid media of E. coli producing the indicated peri-Tse proteins. Periplasmic localization was achieved by fusion to the PelB leader sequence[35]. Cultures were induced at the indicated time (arrow). Error bars ± s.d. (n=3).

f. Representative micrographs of strains shown in e acquired prior to complete lysis. The lipophilic dye TMA-DPH is used to highlight the cellular membranes. Supplementary Fig. 5 contains the full microscopic fields from which these images were derived. All images were acquired at the same magnification. Scale bar = 2 μm.

To test our predictions, we incubated purified Tse1 and Tse3 (Supplementary Fig. 2) with isolated E. coli peptidoglycan sacculi. Soluble products released by the enzymes were separated by high performance liquid chromatography (HPLC) and analyzed by mass spectrometry (MS). To generate separable fragments, Tse1-treated samples were digested with cellosyl, a muramidase, prior to HPLC. The observed absence of the major crosslinked fragment, and the formation of two Tse1-specific products, is consistent with enzymatic cleavage of an amide bond in the peptidoglycan peptide crosslink (Fig. 1b and Supplementary Fig. 3). Moreover, our MS data suggest that the enzyme possesses specificity for the γ-D-glutamyl-L-meso-diaminopimelic acid bond in the donor peptide stem (Fig. 1c and Supplementary Table 1). A variant of Tse1 containing an alanine substitution in its predicted catalytic cysteine ((C30A), Tse1*) did not degrade peptidoglycan (Fig. 1b).

Soluble peptidoglycan fragments released by Tse3 confirmed our prediction that the enzyme cleaves the glycan backbone between N-acetylmuramic acid (MurNAc) and N-acetylglucosamine (GlcNAc) residues (Fig. 1d and Supplementary Fig. 3). Enzymes that cleave this bond can do so hydrolytically (lysozymes) or non-hydrolytically (lytic transglycosylases); the latter results in the formation of 1,6-anhydroMurNAc. Our analyses showed that Tse3 possesses lysozyme-like activity and furthermore suggest that its activity is limited to a fraction of the MurNAc-GlcNAc bonds. The enzyme solubilized a significant proportion of the sacculi to release non-crosslinked peptidoglycan fragments and high molecular weight, soluble peptidoglycan fragments (Fig. 1c, Supplementary Fig. 3 and Supplementary Table 1). A Tse3 protein with glutamine substituted at the site of the predicted catalytic glutamic acid ((E250Q), Tse3*) displayed significantly diminished activity.

If Tse1 and Tse3 degrade peptidoglycan, we reasoned the enzymes might have the capacity to lyse bacterial cells. Ectopic expression of Tse1 and Tse3 in the cytoplasm of Escherichia coli resulted in no significant lysis (Supplementary Fig. 4a,b). However, periplasmically-localized forms of both proteins (peri-Tse1, peri-Tse3) abruptly lysed cells following induction (Fig. 1e and Supplementary Fig. 4c). In accordance with our in vitro studies, peri-Tse1* and peri-Tse3* did not induce lysis at expression levels equivalent to those of the native enzymes (Supplementary Fig. 4d). We also examined cells producing the periplasmically localized enzymes using fluorescence microscopy. Consistent with our biochemical data, cells producing peri-Tse1 were amorphous or spherical, while those producing peri-Tse3 were swollen and filamentous (Fig. 1f and Supplementary Fig. 5). In total, these data demonstrate that Tse1 and Tse3 are enzymes that degrade peptidoglycan in vivo, and that, unlike related enzymes involved in cell wall metabolism, they possess no inherent means of accessing their substrate in the periplasmic space.

T6S function does not require Tse1&3

Since the Tse enzymes alone are unable to reach their target cellular compartment, we hypothesized that their function must be linked to export by the T6SS. In this regard, they could: 1) remodel donor peptidoglycan to allow for the assembly of the mature T6S apparatus, 2) remodel recipient cell peptidoglycan to facilitate the passage of the T6S apparatus through the recipient cell wall, or 3) act as antibacterial effectors that compromise recipient cell wall integrity. To determine if Tse1 and Tse3 are essential for T6S apparatus assembly, we examined whether the enzymes are required for export of the third effector, Tse2. The secretion of Tse2 was not diminished in a strain lacking tse1 and tse3, suggesting that assembly of the T6S apparatus is unhindered by their absence (Fig. 2a). If Tse1 and Tse3 act as enzymes that remodel recipient cell peptidoglycan to facilitate effector translocation, Tse2 action on recipient cells should be severely impaired or nullified in the Δtse1 Δtse3 background. Instead, we found that this strain retained the ability to functionally target Tse2 to recipient cells (Fig. 2b). These findings led us to further examine the hypothesis that Tse1 and Tse3 are effector proteins rather than accessory enzymes of the T6S apparatus.

Figure 2

Tse1 and Tse3 are not required for Tse2 export or transfer to recipient cells via the T6S apparatus

a. Western blot analysis of supernatant (Sup) and cell-associated (Cell) fractions of the indicated P. aeruginosa strains. The parental background for all experiments represented in this figure is PAO1 ΔretS, a strain in which the H1-T6SS is activated constitutively[13,36].

b. Growth competition assays between the indicated donor and recipient strains under T6S-conducive conditions. Experiments were initiated with equal colony forming units (c.f.u.) of donor and recipient bacteria as denoted by the dashed line. The ΔclpV1 strain is a T6S-deficient control. Asterisks indicate significant differences in competition outcome between recipient strains against the same donor strain. **P < 0.01. Error bars ± s.d. (n=3).

Immunity proteins inhibit Tse1&3

Previous data indicate that P. aeruginosa can target itself via the T6SS[7]. If Tse1 and Tse3 act as antibacterial effectors, it follows that P. aeruginosa must be immune to their toxic effects. The tse1 and tse3 genes are each found in predicted bicistronic operons with a hypothetical gene, henceforth referred to as tsi1 and tsi3, respectively. Immunity proteins often inactivate their cognate toxin by direct interaction[20]; therefore, as a first step toward defining a functional link between cognate Tsi and Tse proteins, we asked whether they physically associate. A solution containing a mixture of purified Tse1 and Tse3 was mixed with E. coli lysates containing either Tsi1 or Tsi3. Co-immunoprecipitation studies indicated that Tsi1 and Tsi3 interact specifically with Tse1 and Tse3, respectively, and interactions between non-cognate pairs were not detected (Fig. 3a). To investigate the immunity properties of the Tsi proteins, we measured their ability to inhibit toxicity of peri-Tse1 and peri-Tse3 in E. coli. Both Tsi1 and Tsi3 significantly decreased the toxicity of cognate, but not non-cognate Tse proteins (Fig. 3b). These results show that the activity of periplasmic Tse1 and Tse3 is specifically inhibited by cognate Tsi proteins.

Figure 3

Tsi1 and Tsi3 provide immunity to cognate toxins

a. Western blot analysis of hexahistidine-tagged Tse proteins (–His6) in total and bead-associated fractions of an α-VSV–G (vesicular stomatitis virus glycoprotein) immunoprecipitation of VSV–G epitope fused Tsi proteins (–V) from E. coli.

b. Growth of E. coli harboring a vector expressing the indicated tse gene (top panels) or vectors expressing the indicated tse and tsi genes (bottom panels). Numbers at top indicate 10-fold serial dilutions.

c. Fluorescence micrographs showing colony growth of the indicated strains. The parental background for this experiment was PAO1 ΔretS attTn7::gfp. Growth of the Δtsi strains was rescued by the addition of 1.0% w/v NaCl to the underlying medium. For quantification of data and complementation analyses see Supplementary Fig. 7.

d. Replication rates of the indicated P. aeruginosa strains in liquid medium of low osmolarity formulated as in c. The parental strain used in this experiment was PAO1 ΔretS. Error bars ± s.d. (n=3).

T6S delivers Tse1&3 to the periplasm

Most genes encoding immunity functions are essential in the presence of their cognate toxins. However, mutations that inactivate tsi1 and tsi3 are readily generated in P. aeruginosa strains that constitutively express and export Tse1 and Tse3. Based on this observation, we hypothesized that under standard laboratory conditions, the Tse proteins do not efficiently access their substrate in the periplasm. This suggests that T6S occurs by a mechanism wherein effectors are denied access to donor cell periplasm and are instead released directly to the periplasm of the recipient cell. According to this mechanism, the tsi genes would only be essential when a strain is grown under conditions that permit intercellular transfer of effectors between neighboring cells by the T6SS. As predicted, deletions in tsi1 and tsi3 severely impaired the growth of P. aeruginosa on a solid substrate, a condition conducive to T6S-based effector delivery (Fig. 3c and Supplementary Fig. 6)[21,22]. In contrast, this growth inhibition did not occur in liquid media, which is not conducive to effector delivery by the T6SS (Fig. 3d). The growth inhibition phenotype required a functional T6SS and intact cognate effector genes, and consistent with the proposed functions of Tse1 and Tse3 in compromising cell wall integrity, growth of immunity deficient strains was fully rescued by increasing the osmolarity of the medium (Fig. 3c).

Bioinformatic analyses suggested that the Tsi proteins reside in the periplasm – Tsi1 as a soluble periplasmic protein and Tsi3 as an outer membrane lipoprotein. These predictions were confirmed by subcellular fractionation experiments, which indicated enrichment of the proteins in the periplasmic compartment (Fig 4a). This result, taken together with the observation that the Tsi proteins interact directly with their cognate Tse proteins (Fig. 3a), provided us with a means of addressing whether the T6SS delivers Tse proteins intercellularly to the periplasm. We reasoned that if the Tse proteins are indeed delivered to the periplasm of another bacterial cell, not only should we be able to observe intoxication between distinct donor and recipient strains of P. aeruginosa, but the production of an otherwise competent immunity protein that is mislocalized to the cytoplasm should not be able to prevent such intoxication.

Figure 4

Tse1 and Tse3 delivered to the periplasm provide a fitness advantage to donor cells

a. Western blot analyses of cytoplasmic (Cyto) and periplasmic (Peri) fractions of P. aeruginosa strains producing Tsi1–V, Tsi3–V or Tsi3–SS–V. Equivalent ratios of the Cyto and Peri samples were loaded in each panel. RNA polymerase (RNAP) and β-lactamase (β-lac) enzymes were used as cytoplasmic and periplasmic fractionation controls, respectively. The presence of Tsi3–a predicted outer membrane lipoprotein–in the periplasmic fraction is consistent with previous studies utilizing this method of fractionation[37].

b. Growth competition assays between the indicated donor and recipient strains under T6S-conducive conditions. Experiments were initiated with equal c.f.u. of donor and recipient bacteria as denoted by the dashed line. The parental strain used in this experiment was PAO1 ΔretS. All donor strains were modified at the attB site with lacZ. Asterisks indicate outcomes significantly different than parental versus Δtse3 Δtsi3 (top bar). Error bars ± s.d. (n=4). **P < 0.01.

c. Lysis of EDTA-permeabilized or intact P. aeruginosa cells with equal quantities of Tse1, Tse1*, or Lysozyme (Ly). Lysis was normalized to a buffer control. Error bars ± s.d. (n=3).

d. Competitive growth of P. aeruginosa against P. putida on solid (open circles) or in liquid (filled circles) medium. Competition outcome was defined as the c.f.u. ratio (P. aeruginosa/P. putida) divided by the initial ratio. The dotted line represents the boundary between competitions that increase in P. aeruginosa relative to P. putida (above the line) and those that increase in P. putida relative to P. aeruginosa (below the line). The parental strain used in this experiment was P. aeruginosa PAO1. Asterisks above competitions denote those where the outcome (P. aeruginosa/P. putida) was significantly less than the parental (P < 0.05). Horizontal bars denote the average value for each dataset (n=5).

In growth competition assays between distinct donor and recipient strains of P. aeruginosa, we found that recipient cells that lack Tse3 immunity and are incapable of self-intoxication (Δtse3 Δtsi3), display a growth disadvantage against donor bacteria. This phenotype depends on H1-T6SS function and Tse3 in the donor strain. In the recipient strain, ectopic expression of wild-type tsi3, but not an allele encoding a signal sequence-deficient protein (Tsi3–SS), rescues the fitness defect (Fig. 4b). Importantly, the Tsi3–SS protein used in this experiment does not reach the periplasm, and retains activity in vitro as judged by interaction with Tse3 (Fig. 4a and Supplementary Fig. 7). The Tsi3–SS protein also fails to rescue the intercellular self-intoxication growth phenotype of Δtsi3 (Supplementary Fig. 6). Analogous experiments with Tsi1 were not feasible, as the protein was unstable in the cytoplasm.

The most parsimonious explanation for T6S-mediated intercellular toxicity by Tse1 and Tse3 is that the apparatus provides a conduit for the effectors through the outer membrane of recipient cells. This led us to predict that exogenous Tse1 and Tse3 would not lyse intact P. aeruginosa. Furthermore, we posited that if the outer membrane was the relevant barrier to Tse1 and Tse3 toxicity, compromising its integrity should render P. aeruginosa susceptible to exogenous administration of the enzymes.

To test these predictions, we measured lysis of permeabilized and intact P. aeruginosa following addition of exogenous Tse1. We did not test Tse3, as the filamentous phenotype induced by this enzyme would not affect non-growing, permeabilized cells. Intact P. aeruginosa cells were not affected by the addition of exogenous Tse1; conversely, permeabilized P. aeruginosa was highly susceptible to lysis by the enzyme (Fig. 4c). Lysis induced by Tse1 is linked to its enzymatic function, as Tse1* failed to significantly lyse cells. In total, our data show that the T6SS breaches the outer membrane to deliver lytic effector proteins directly to recipient cell periplasm.

To determine whether the T6SS can target the Tse proteins to cells of another Gram-negative organism, we conducted growth competition assays between P. aeruginosa and P. putida. These bacteria can be co-isolated from the environment[23] and are likely to compete for niches[24]. While inactivation of either tse1 or tse3 only modestly affected the outcome of P. aeruginosa-P. putida competition assays, the fitness of P. aeruginosa lacking both genes or a functional T6SS was dramatically impaired (Fig. 4a). This partial redundancy is congruent with the enzymes exerting their effects through a single target–peptidoglycan–in the recipient cell. The fitness advantage provided by Tse1 and Tse3 was lost in liquid medium, consistent with cell contact-dependent delivery of the proteins to competitor cells (Fig. 4d). These data indicate that the T6SS targets its effectors to other species of bacteria and that these proteins can be key determinants in the outcome of interspecies bacterial interactions. In contrast with intraspecies intoxication, interspecies intoxication via the T6SS does not require the inactivation of a negative regulator of the system (eg. ΔretS), suggesting that T6S function is stimulated in response to rival bacteria.

Discussion

Our data lead us to propose a model for T6S-catalyzed translocation of effectors to the periplasm of recipient bacteria (Fig. 5). This model provides a mechanistic framework for understanding the form and function of this complex secretion system. Our findings strengthen the existing hypothesis that the T6SS is evolutionarily and functionally related to bacteriophage[8,14,25]. Neither the T4 bacteriophage tail spike nor other components of the puncturing device are thought to cross the inner membrane; instead, bacteriophage DNA is released to the periplasm and subsequently enters the cytoplasmic compartment using another pathway[26]. By analogy, the Tse proteins would utilize T6S components as a puncturing device to gain access to the periplasm, whereupon Tse2 may then utilize an independent route to access the cytoplasm (Fig. 5).

Figure 5

Proposed mechanism of T6S-dependent delivery of effector proteins

The schematic depicts the junction between competing bacteria, with a donor cell delivering the Tse effector proteins through the T6S apparatus (grey tube) to recipient cell periplasm. Effector and immunity proteins are shown as circles and rounded rectangles, respectively. Bonds in the peptidoglycan that are predicted targets of the effector proteins are highlighted (red). Cytoplasm (C), inner membrane (IM), periplasm (P), and outer membrane (OM) of both bacteria are shown.

Niche competition in natural environments has clearly selected for potent antibacterial processes; however, the human body is also home to a complex and competitive microbiota[27,28]. Commensal bacteria form a protective barrier, and the ability of pathogens to colonize the host is not only dependent upon suppression or subversion of host immunity, but also can depend on their ability to displace these more innocuous organisms[29-31]. In polymicrobial infections, Gram-negative bacteria, including P. aeruginosa, often vie with other Gram-negative bacteria for access to nutrient-rich host tissue[32]. Factors such as the T6SS, that influence the relative fitness of these organisms, are thus likely to impact disease outcome.

Methods Summary

P. aeruginosa strains used in this study were derived from the sequenced strain PAO1[33]. All deletions were in-frame and unmarked, and were generated by allelic exchange. E. coli growth curves were conducted using BL21 pLysS cells harboring expression plasmids for tse and tsi genes. Intercellular self-intoxication and interbacterial competition assays were performed by spotting mixed overnight cultures on a nitrocellulose membrane placed on a 3% agar growth medium. Samples were incubated at 37°C (P. aeruginosa-P. aeruginosa) or 30°C (P. aeruginosa-P. putida) for 12 or 24 hours. Tse1-catalyzed P. aeruginosa lysis was measured by placing cells in a minimal buffer ± 1.5 mM EDTA containing either Tse1, Tse1* or lysozyme. The change in optical density at 600 nm following 5 min of incubation was used to calculate lysis. For determination of Tse1 and Tse3 activity, isolated E. coli peptidoglycan sacculi were incubated with the purified enzymes (100 μg/mL). The resulting peptidoglycan and soluble fragments released by the enzymes were separated by HPLC and their identities were determined using MS as described previously[34].