Identification, Structure, and Function of a Novel Type VI Secretion Peptidoglycan Glycoside Hydrolase Effector-Immunity Pair*

Background: The bacterial type VI secretion system (T6SS) translocates toxic effector proteins into target cells. Results: Novel T6S peptidoglycan glycoside hydrolase effector-immunity families are identified and a representative pair is structurally and functionally characterized. Conclusion: Peptidoglycan glycoside hydrolase effectors are an important component of the T6S effector arsenal. Significance: This work expands the repertoire of known T6S substrates and reports the first structure of a peptidoglycan glycoside hydrolase effector. Bacteria employ type VI secretion systems (T6SSs) to facilitate interactions with prokaryotic and eukaryotic cells. Despite the widespread identification of T6SSs among Gram-negative bacteria, the number of experimentally validated substrate effector proteins mediating these interactions remains small. Here, employing an informatics approach, we define novel families of T6S peptidoglycan glycoside hydrolase effectors. Consistent with the known intercellular self-intoxication exhibited by the T6S pathway, we observe that each effector gene is located adjacent to a hypothetical open reading frame encoding a putative periplasmically localized immunity determinant. To validate our sequence-based approach, we functionally investigate a representative family member from the soil-dwelling bacterium Pseudomonas protegens. We demonstrate that this protein is secreted in a T6SS-dependent manner and that it confers a fitness advantage in growth competition assays with Pseudomonas putida. In addition, we determined the 1.4 Å x-ray crystal structure of this effector in complex with its cognate immunity protein. The structure reveals the effector shares highest overall structural similarity to a glycoside hydrolase family associated with peptidoglycan N-acetylglucosaminidase activity, suggesting that T6S peptidoglycan glycoside hydrolase effector families may comprise significant enzymatic diversity. Our structural analyses also demonstrate that self-intoxication is prevented by the immunity protein through direct occlusion of the effector active site. This work significantly expands our current understanding of T6S effector diversity.

groups: those that act as amidases, cleaving the peptidoglycan molecule within its peptide stems and cross-links and those that act as glycoside hydrolases, cleaving the glycan backbone of the molecule. T6S amidase effectors have been studied extensively (3,9,10,(12)(13)(14)(15). The enzymes are broadly distributed among Proteobacteria and form four phylogenetically distinct families that constitute the Tae (type IV secretion amidase effector) superfamily. Interestingly, the preferred cleavage site within peptidoglycan can vary between Tae families, suggesting the possibility that optimal effector specificity is dependent on the organism(s) targeted and/or the precise structure of the peptidoglycan found in those organisms.
A commonality among all identified effectors with antibacterial activity is that their corresponding open reading frames are found adjacent to genes encoding cognate immunity proteins. These proteins confer immunity by specifically binding to and inactivating their associated effector (3,(13)(14)(15). As effectors do not access the donor cell periplasm in transit to the recipient cell, the immunity proteins of cell wall-targeting effectors, which reside in the periplasm, serve the exclusive purpose of preventing intercellular self-intoxication.
The sequence divergence and sporadic distribution of T6S effectors present a challenge for the identification of these important mediators of interbacterial interactions. Owing to their frequent horizontal inheritance, bacteria related at the genus, or even the species level, often contain a completely unique repertoire of these proteins. For example, the plant commensal bacterium Pseudomonas protegens does not contain homologs of the three established effectors of the P. aeruginosa Hcp secretion island I-encoded T6SS (H1-T6SS), Tse1-3; however, this organism possesses a T6SS orthologous to the H1-T6SS (17).
One way in which the challenge of identifying T6SS effectors has been overcome is by exploiting the tendency of their corresponding genes to reside within or in close proximity to T6SSencoding gene clusters. This approach was used for the identification of Tae4 family members from Serratia marcescens (12). Alternatively, mass spectrometry-based methodologies have been successful in the identification of T6S effectors from P. aeruginosa, Burkholderia thailandensis, and S. marcescens (2,9,18). Finally, our group utilized a sequence homology-independent informatic search based on common properties found within effector-immunity (E-I) pairs to identify the Tae superfamily (9). These properties, applied independently to the candidate effector and immunity protein, included size, isoelectric point, predicted subcellular localization, and the presence of a cysteine-histidine catalytic dyad.
In this study, we performed an informatic search for T6SS substrates and found previously unidentified families of peptidoglycan glycoside hydrolase effectors, herein named Tge proteins (type VI secretion glycoside hydrolase effectors). Characterization of a representative Tge from P. protegens showed that the protein displays periplasmic toxicity, is secreted in a T6-dependent manner, and confers a fitness advantage when P. protegens is grown in competition against P. putida, a co-occurring soil bacterium. Additionally, we solved the 1.4 Å crystal structure of P. protegens Tge in complex with its cognate immunity protein. Together, our findings show a broader distribution of T6S glycoside hydrolase effectors than was previously appreciated and offer insights into the molecular basis for glycoside hydrolase activity and inhibition.

EXPERIMENTAL PROCEDURES
Bioinformatic Screen-Putative effector-immunity candidates were identified using a similar informatic search protocol as described previously (9). Briefly, a custom Perl script was used to search 115 T6SS ϩ genomes for bicistronic genes with the following criteria for the encoded effector protein: 1) no predicted signal sequence, 2) a predicted pI greater than 8.0, and 3) fewer than 200 amino acids. The criteria for the immunity protein included the presence of a predicted signal sequence and fewer than 200 amino acids. Protein sequences obtained from this screen were submitted in batch mode to the Phyre 2 server and examined manually for the presence of lysozyme-like folds (19). Candidate peptidoglycan glycoside hydrolases and associated immunity proteins were then used as Blastp search queries to identify all unique family members in the NCBI database.
Bacterial Strains and Growth Conditions-All P. protegens strains generated in this study were derived from the sequenced strain Pf-5 (20). P. protegens strains were grown in Luria-Bertani (LB) media at 30°C supplemented with 15 g ml Ϫ1 gentamycin and 25 g ml Ϫ1 irgasan where appropriate. The pEXG2 suicide vector was used for in-frame chromosomal deletions in P. protegens as described previously for P. aeruginosa (21). Similar to P. aeruginosa, deletion of retS is required for activation of T6S in P. protegens (22,23). Locus tags for retS, tge2 PP , tgi2 PP , and clpV are PFL_0664, PFL_3037, PFL_3036, and PFL_6093, respectively. The P. putida strain used for competition assays was derived from the sequenced strain KT2440 (24) and grown in LB media at 30°C. E. coli strains used included DH5␣ for cloning, SM10 for conjugal transfer of plasmids into P. protegens, BL21 pLysS for toxicity assays, growth curves, and phase contrast microscopy, and Shuffle T7 Express lysY (New England Biolabs) for expression of proteins for purification. E. coli strains were either grown in LB or LB low salt (LB-LS) at 37°C supplemented with 50 g ml Ϫ1 kanamycin, 150 g ml Ϫ1 carbenicillin, 30 g ml Ϫ1 chloramphenicol, 200 g ml Ϫ1 trimethoprim, 0.1% (w/v) L-rhamnose and the indicated concentrations of IPTG as required.
Interbacterial Competitions-Competition experiments were performed in a similar manner as described previously for B. thailandensis (9). Briefly, P. protegens and P. putida strains were mixed in a 1:1 ratio, and 10 l of the mixture was spotted on 0.2-m nitrocellulose membrane overlaid on LB-LS 3% agar plates. Plate counts were taken of the initial innoculum and after 18 h of competition at 30°C. The recipient P. putida strain contained a GFP expression construct integrated into the attTn7 site allowing for differentiation of the donor and recipient cells using fluorescence imaging as described previously (1,25,26). Statistical analyses were carried out using a two-tailed Student's t test.
P. protegens Secretion Assay and Western Blot Analysis-Tge2 PP , including its native ribosome binding site, were cloned into pPSV35-CV using the SacI/XbaI restriction sites. The resulting construct encodes Tge2 PP fused to a C-terminal vesicular stomatitis virus G (VSV-G) epitope tag to facilitate detection by immunoblotting. Overnight cultures of P. protegens ⌬retS and ⌬retS ⌬clpV harboring pPSV35-CV::tge2 PP were subcultured 1:500 into LB containing 100 M IPTG and grown to mid-log phase. Cell and supernatant fractions were separated and analyzed for the presence of Tge2 PP by Western blot using an anti-VSV-G antibody as described previously (21).
Growth Curves-Overnight cultures of E. coli BL21 pLysS containing pET-22b(ϩ), pET-22b(ϩ)::tge2 PP and pET-22b(ϩ)::tge2 PP E69Q were subinoculated to an optical density at 600 nm (A 600 ) of 0.01 in LB-LS and grown at 37°C. A 600 measurements were taken every 30 min until the cultures reached an A 600 of ϳ0.2, at which point protein expression was induced with 100 M IPTG. Following IPTG induction, A 600 measurements were taken every 15 min for the remainder of the growth curve.
Protein Expression and Purification-For the isothermal titration calorimetry (ITC) experiment, full-length tge2 PP and tgi2 PP lacking its encoded signal sequence (Tgi2 PP (24 -156) were cloned into pET-28b(ϩ) and pET-15b using the NdeI/ XhoI and NdeI/BamHI restriction sites, respectively, to generate N-terminal His 6 tags on both proteins. For crystallization, full-length tge2 PP was cloned into pET-24a(ϩ) using the NdeI/ XhoI restriction sites to generate an untagged construct suitable for coexpression with the Tgi2 PP (24 -156) construct described above. All protein constructs were expressed in Shuffle T7 Express lysY cells, which allow for cytoplasmic disulfide bond formation, and grown in LB broth supplemented with the appropriate antibiotics. For each construct, cells were grown at 37°C to an A 600 of 0.6 before protein expression was induced with 1 mM IPTG at room temperature for 18 h. For the protein samples used in the ITC experiment, cells were resuspended in 50 mM sodium phosphate, pH 7.2, 300 mM NaCl, 5 mM imidazole, and lysed by sonication. Following centrifugation, the cleared cell lysates were purified by Ni 2ϩ -nitrilotriacetic acid affinity chromatography using a linear gradient of 5-400 mM imidazole. The purified proteins were then dialyzed against 20 mM sodium phosphate, pH 7.2, 150 mM NaCl prior to use in downstream experiments. For crystallization, cells containing coexpressed Tge2 PP /Tgi2 PP (24 -156) were resuspended in 50 mM Hepes, pH 7.5, 300 mM NaCl, 5 mM imidazole, and lysed by sonication. The complex was then purified using Ni 2ϩ -nitrilotriacetic acid affinity chromatography in 20 mM Hepes, pH 7.5, 300 mM NaCl, 5-400 mM imidazole followed by size exclusion chromatography in 20 mM Hepes, pH 7.5, 150 mM NaCl.
Isothermal Titration Calorimetry-Purified Tge2 PP and Tgi2 PP (24 -156) were degassed before experimentation. ITC measurements were performed with a VP-ITC microcalorimeter (MicroCal Inc., Northampton, MA). Titrations were carried out with 250 M Tgi2 PP (24 -156) in the syringe and 16 M Tge2 PP in the cell. The titration experiment consisted of one 2-l injection followed by 29 5-l injections with 300-s intervals between each injection. The ITC data were analyzed using the Origin software (version 5.0, MicroCal, Inc.) and fit using a single-site binding model.
Phase Contrast Microscopy-Phase contrast microscopy images were acquired with a Nikon Ti-E inverted microscope fitted with a 60ϫ oil objective, a xenon light source (Sutter Instruments), and a CCD camera (Clara series, Andor). Overnight cultures of E. coli BL21 pLysS pET-22b(ϩ)::tge2 PP and pET-22b(ϩ)::tge2 PP E69Q were subcultured 1:300 into LB broth containing the appropriate antibiotics and grown to an A 600 of 0.2. IPTG was then added to a concentration of 200 M, and after a 45-min incubation, the bacterial suspension (1 l) was spotted onto growth pads made with LB-LS medium and 1% (w/v) agarose, and cells were imaged immediately.
Crystallization and Structure Determination-Selenomethionyl-incorporated Tge2 PP ⅐Tgi2 PP (24 -156) complex was concentrated to 10 mg ml Ϫ1 by spin ultrafiltration (10 kDa molecular mass cut-off, Millipore) and screened against commercially available sparse matrix crystal screens (Crystal Screens 1 and 2, Hampton Research). Crystal trials were setup in 48-well plates using the sitting drop vapor diffusion technique. Protein and crystallization solutions were mixed in a 1:1 ratio with a final drop size of 2 l suspended over 250 l of crystallization solution and stored at room temperature.
After 3-4 days, diffraction quality crystals grew in 0.1 M Tris-HCl, pH 8.5, 0.2 M CaCl 2 , and 30% (w/v) PEG 4000 and were flash frozen without any added cryoprotectant. X-ray diffraction data were collected on beamline 5.0.3 at the Lawrence Berkeley National Laboratory Advanced Light Source (27). The data were merged, integrated, and scaled using the xia2 system (28). Experimental phases were obtained using the Phenix AutoSol Wizard (29) resulting in density modified selenium single-wavelength anomalous diffraction phased maps of excellent quality, which allowed for automated model building with the Phenix AutoBuild Wizard (30). Subsequent model adjustments were made manually in COOT (31) between iterative rounds of refinement carried out using PHENIX.REFINE (32). The final model was refined to an R work /R free of 14.9% and 17.5% (Table 1).

Identification of Peptidoglycan Glycoside Hydrolase Effector/
Immunity Families-Previously, we reported the development of a heuristic search method that was successfully employed for the identification of a widespread type VI peptidoglycan amidase effector superfamily (9). Using a customized Perl script, T6SS ϩ genomes were searched for bicistrons that encode protein products with a defined set of properties commonly associated with T6S effector-immunity pairs. For the effector protein, this included the absence of a signal peptide, an isoelectric point greater than 8.0, fewer than 200 amino acids, and the presence of conserved catalytic histidine and cysteine residues. Parameters used for the immunity protein included the presence of a signal peptide and primary sequence length limitation (Ͻ200 amino acids). These selection criteria yielded 418 total candidate E-I pairs, within which amidase effectors were identified by structure prediction algorithms and subsequent direct experimentation (19). To adapt this pipeline for the identification of glycoside hydrolase effectors, we excluded the histidine and cysteine constraints, generating a list of 831 candidate E-I pairs from 115 T6SS ϩ genomes. These were then examined by structure prediction servers to identify lysozyme-like folds and manually curated to remove systematic false-positives (see "Experimental Procedures"). This approach yielded two phylogenetically distinct families, which combine with P. aeruginosa Tse3 to form the basis for three distinct type VI peptidoglycan glycoside hydrolase effector and immunity groups (Tge1-3 and Tgi1-3) ( Fig. 1 and supplemental Figs. S1 and S2). Although it likely functions as a muramidase, we have opted not to include V. cholerae VgrG-3 in this reclassification of T6S glycoside hydrolase effectors. This emphasizes the additional postulated structural role of VgrG-3 in the T6S apparatus, which has not been observed for non-VgrG-fused effectors similar to the Tge1-3 proteins (11,16).
Many peptidoglycan glycoside hydrolases share a lysozymelike fold; however, they can be divided into two enzymatic subgroups based on their cleavage mechanism. Lysozymes and lytic transglycosylases cleave the ␤-1,4-glycosidic bond between MurNAc and GlcNAc, the former hydrolyzing the bond forming a reducing MurNAc end, and the latter catalyzing the formation of 1,6-anhydroMurNAc (33). In contrast, N-acetylglucosaminidases cleave the ␤-1,4-glycosidic bond between GlcNAc and MurNAc, leaving GlcNAc as the reducing sugar. Irrespective of their substrate specificity, peptidoglycan glycoside hydrolases universally employ a conserved glutamate, acting as a catalytic acid, to initiate cleavage of the glycosidic bond (33). Consistent with our hypothesis that Tge2 and Tge3 proteins such as Tse3 function as peptidoglycan glycoside hydrolase enzymes, we identified a strictly conserved Glu within a putative active motif in both enzyme families (supplemental Figs. S1 and S2). Additionally, our sequence analyses indicated that Tge2 proteins likely hydrolyze peptidoglycan using a standard goose-type (G-type) single displacement mechanism (34), whereas the Tge3 family contains a Glu-Asp-Thr catalytic triad utilized by phage-type lysozymes (35,36).
Tge2 PP -Tgi2 PP Are an Effector-immunity Pair Secreted by the T6SS-Tge1 from P. aeruginosa (Tge1 PA , formerly Tse3) has been demonstrated previously to function as a peptidoglycan glycoside hydrolase effector. To validate a new family identified by our informatics search, we chose to characterize a Tge2-Tgi2 E-I pair from the soil-dwelling bacterium P. protegens (Tge2 PP -Tgi2 PP ). To confirm that Tge2 PP functions as a periplasmic effector, we performed toxicity assays in E. coli. As expected for a peptidoglycan-degrading enzyme, we observed a significant decrease in E. coli viability and optical density when Tge2 PP was artificially targeted to the periplasm via a sec-dependent leader sequence (peri-Tge2 PP , Fig. 2, A and B). Furthermore, these effects of Tge2 PP periplasmic toxicity were specific to Tge2 PP catalytic activity, as a conservative mutation of its predicted catalytic glutamate (E69Q) abolished toxicity. Consistent with its proposed role as a peptidoglycan glycoside hydrolase effector, peri-Tge2 PP induced cell rounding and lysis compared with the catalytic point mutant (Fig. 2C). Importantly, Western blot analysis confirmed that the observed reduction in toxicity and cell rounding was not due to decreased expression levels of the E69Q protein relative to wild-type (Fig. 2D). Despite our ability to observe Tge2 PP activity in a cellular context, we were unable to detect significant degradation of purified peptidoglycan sacculi by Tge2 PP over a range of conditions, raising the intriguing  SEPTEMBER 13, 2013 • VOLUME 288 • NUMBER 37 possibility that the effector may require a yet unidentified periplasmic cofactor or binding partner.

Structure of a Glycoside Hydrolase Effector-Immunity Pair
Next, we examined whether Tgi2 PP could inhibit the toxicity exhibited by peri-Tge2 PP . Tgi2 PP co-expressed with peri-Tge2 PP significantly rescued Tge2 PP -dependent toxicity, indicating that Tgi2 PP functions as an immunity protein (Fig. 2E). As predicted by its lack of sequence homology to Tgi2 PP , co-expression of P. aeruginosa Tgi1 (Tgi1 PA , formerly Tsi3) was not able to rescue E. coli viability. Next, we investigated the mechanism of inhibition of Tge2 PP activity by Tgi2 PP . To probe for a direct binding mechanism, we conducted ITC measurements on purified Tge2 PP and Tgi2 PP . The binding isotherm obtained by titration of the effector with Tgi2 PP fit a single-site binding model with a stoichiometry of ϳ1:1 (n ϭ 0.93) and a dissociation constant of 0.26 Ϯ 0.04 nM (Fig. 2F). Titration of Tgi2 PP into buffer and buffer into Tge2 PP confirmed that the observed saturation of the heats of injection was due to the Tge2 PP -Tgi2 PP interaction and not between either of the proteins and a component of the buffer solution (supplemental Fig. S3). The tight binding observed between Tge2 PP and Tgi2 PP is consistent with the affinities measured for other T6SS E-I pairs, including P. aeruginosa Tae1-Tai1, Enterobacter cloacae Tae4-Tai4, and two Tae4-Tai4 family members from S. marcescens, which all bind with low nanomolar to high picomolar dissociation constants (12,13,15). P. aeruginosa possesses three functionally non-redundant T6SSs, encoded by HSI-I-III. Of these, the H1-encoded system is post-transcriptionally activated by the Gac/Rsm pathway (23). P. protegens possesses only a single T6SS. This system is orthologous to the H1-T6SS, and is also regulated by the Gac/ Rsm pathway (22). The RetS hybrid sensor kinase functions within the Gac/Rsm pathway as a negative regulator of T6S (2,22,23). Given that T6SSs are largely quiescent under in vitro cultivation conditions, we used a P. protegens ⌬retS strain to test the hypothesis that Tge2 PP is a T6SS substrate. Robust secretion of Tge2 PP was observed in the supernatant fraction of the parental strain; however, this was abrogated in a P. protegens strain containing the additional deletion of the gene encoding the essential type VI component, ClpV (Fig. 2G). Given that Tge2 PP is secreted by the T6SS, we next sought to ascertain its contribution to interbacterial fitness. When P. protegens strains were competed under cell contact-promoting conditions against P. putida, a bacterium that inhabits similar environments (37,38), those lacking tge2 PP were ϳ6-fold less fit compared with the wild-type (Fig. 2H). Although this difference is highly significant and confirms the role of Tge2 PP as a T6 effector, it is also interesting to note that P. protegens strains with a non-functional T6SS display a ϳ1000-fold decrease in competitive fitness against P. putida. This finding suggests that there are other currently uncharacterized effectors produced by P. protegens that mediate antagonistic interactions between these organisms. In total, these data indicate that Tge2 PP is an antibacterial T6S effector.
Structure of the Tge2 PP ⅐Tgi2 PP Complex-To gain further insight into Tge2 PP function and its mode of inhibition by Tgi2 PP , we determined the 1.4 Å crystal structure of the Tge2 PP ⅐Tgi2 PP complex using the single-wavelength anoma-lous diffraction technique with selenomethionyl-incorporated protein ( Table 1). Crystals of the Tge2 PP ⅐Tgi2 PP complex grew in the orthorhombic space group P2 1 2 1 2 1 and contained a single 1:1 complex in the asymmetric unit. Disordered residues not built in the final model include 1-29, 101-104, and 152-155 of Tge2 PP , and residues 24 -26, 155-156, and the N-terminal His 6 tag of Tgi2 PP . The final model was refined to an R work and R free of 14.9% and 17.5%, respectively.
Tge2 PP Resembles Family 73 Glycoside Hydrolases-Tge2 PP adopts a lysozyme-like fold consisting of large and small lobes that are positioned such that a substrate-binding groove is formed between them (Fig. 3A). The large lobe is made up of a short 3 10 helix, ␣1-␣2 and ␣4-␣6, whereas the small lobe consists of ␣3 and a ␤-hairpin formed by ␤1-␤2. The termini are covalently linked by a Cys 32 -Cys 177 disulfide bond, which connects the N-terminal 3 10 helix to the C-terminal loop. As determined by DALI (39), the overall structure resembles closely that of glycoside hydrolase family 73 members, which includes the  peptidoglycan hydrolase FlgJ from Sphingomonas sp. strain A1 (Z-score, 16.4; C ␣ root mean square deviation of 1.6 Å over 124 equivalent positions) and Auto, a peptidoglycan N-acetylglucosaminidase from Listeria monocytogenes (Z-score, 13.9; C ␣ root mean square deviation of 2.6 Å over 126 equivalent positions) (Fig. 3B) (40,41). It is proposed that members of this family catalyze the hydrolysis of peptidoglycan using a single displacement mechanism similar to that originally demonstrated for G-type egg white lysozyme (GEWL). In this mechanism, a catalytic glutamate donates a proton to the scissile glycosidic bond, resulting in its cleavage and formation of an oxocarbenium intermediate. This short-lived intermediate then undergoes nucleophilic attack by a water molecule, resulting in the hydrolyzed product with inverted stereochemistry at the anomeric position (34). In agreement with our functional studies, the catalytic glutamate in Tge2 PP corresponds to Glu 69 , inferred from its nearly identical position to the corresponding glutamate residue in both FlgJ and Auto (Fig. 3B). Glu 69 protrudes from the end of ␣2 and lies deep within the substrate-binding groove. In addition, the hydroxyl group of Tyr 147 forms a hydrogen bond with Glu 69 , in a manner similar to the aforementioned peptidoglycan glucosaminidase and G-type lysozyme enzymes, perhaps to orient it for catalysis. Tgi2 PP Inhibits Tge2 PP by Protruding into the Substrate-binding Groove-The structure of Tgi2 PP consists of a short N-terminal 3 10 helix and a central five-stranded ␤-sheet flanked by three ␣-helices (Fig. 3C). A Cys 117 -Cys 152 disulfide bond connects ␤5 to the C-terminal loop, anchoring ␣3 to the ␤-sheet core. Tgi2 PP does not share strong structural similarity to other proteins of known structure; however, it does contain a similar topology to the periplasmic E. coli colicin M immunity protein (Z-score, 3.5; C ␣ root mean square deviation of 3.7 Å over 62 equivalent positions) (42). Colicin M is a bacteriocin with phosphatase activity that targets the undecaprenyl phosphatelinked peptidoglycan precursors lipid I and lipid II in the periplasm of recipient bacteria (43). The observation that the colicin M immunity protein resembles Tgi2 PP suggests that these immunity proteins may have undergone diversifying selection to acquire effector specificity.
The interaction between Tge2 PP and Tgi2 PP involves the insertion of the ␤-sheet core of Tgi2 PP into the substrate-binding groove of Tge2 PP . Most notably, the elongated loop connecting ␤2 and ␤3 protrudes into the active site in an orientation predicted to prevent the catalytic glutamate from accessing its peptidoglycan substrate. Overall, the Tge2 PP ⅐Tgi2 PP complex consists of a 1237 Å 2 interface stabilized predominantly by hydrogen bonding and hydrophobic interactions. Conservation mapping of Tgi2 PP homologs suggests that invariant residues Gly 94 and Ala 95 , which are located in the ␤2-␤3 loop, are important for direct interaction with the active site of Tge2 PP (Fig.  3D). Binding site analysis using PDBePISA indicates that the amide nitrogen and oxygen atoms of Gly 94 form hydrogen bonding interactions with the main chain of Tyr 147 and His 88 of Tge2 PP , respectively, whereas the side chain of Ala 95 becomes desolvated upon complex formation (44). Additional conserved interfacing residues include Tyr 56 , Asp 66 , Leu 69 , His 78 , Tyr 88 , Pro 91 , Trp 99 , Gly 100 , Leu 101 , Glu 113 , Cys 117 , Gly 149 , Leu 150 , Gly 151 , Cys 152 , and Asp 154 . The inhibition mode observed in the Tge2 PP ⅐Tgi2 PP complex is reminiscent of bacterial proteina-ceous inhibitors of eukaryotic lysozymes. The E. coli inhibitor of vertebrate lysozyme (Ivy), P. aeruginosa membrane-bound lysozyme inhibitor of C-type lysozyme (MliC), and E. coli PliG (periplasmic lysozyme inhibitor of G-type lysozyme) all use active site occlusion as a mechanism of inhibition ( Fig. 4) (45)(46)(47). Moreover, in each case, the region involved in inhibition consists of a loop region connecting two ␤-strands. However, it is interesting to note that there is little fold similarity between these lysozyme inhibitors, suggesting they arose through convergent evolution.

DISCUSSION
In this work, we identified two families of T6S peptidoglycan glycoside hydrolase effectors using informatic methods. Previously, we used homology-independent search criteria to identify a T6S peptidoglycan amidase effector superfamily. The identification of the peptidoglycan glycosyl hydrolase families described herein further exemplifies the utility of this method as a means to identify novel effector-immunity pairs. Prior characterization of Tge1 PA has shown that this enzyme possesses ␤-(1,4)-N-acetylmuramidase (lysozyme) activity. Furthermore, Tge1 PA contains a GLXQ motif found in both GEWL and E. coli soluble lytic transglycosylase (Slt70) in addition to its catalytic glutamate, suggesting it is likely structurally similar to these enzymes (3,48). Interestingly, Tge3 family members contain a conserved Glu-Asp-Thr catalytic triad reminiscent of phage T4 lysozyme. Thus, similar to Tge1, we expect this family to exhibit ␤-(1,4)-N-acetylmuramidase activity. In contrast, we find that Tge2 PP most closely resembles members of a peptidoglycan hydrolase family with ascribed N-acetylglucosaminidase activity. Members of this family contain a catalytic glutamate residue embedded in an AAXE(S/T) motif. . Structurally characterized lysozyme inhibitors display distinct folds but act via a common mechanism. Comparison of the Tge2 PP ⅐Tgi2 PP complex to vertebrate lysozymes in complex with proteinaceous bacterial lysozyme inhibitors. Shown are the Tge2 PP ⅐Tgi2 PP complex (A), E. coli periplasmic lysozyme inhibitor of G-type lysozyme (PliG EC ) in complex with salmon G-type lysozyme (SalG) (B), E. coli proteinaceous inhibitor of vertebrate lysozyme (Ivy EC ) in complex with hen egg white lysozyme (HEWL) (C), and membrane-bound lysozyme inhibitor of C-type lysozyme (MliC) in complex with hen egg white lysozyme (D) (45)(46)(47). For each structural model, the lysozyme-like protein and its associated inhibitor are shown as surface and ribbon representations, respectively. This observation suggests that Tge2 enzymes could target a different chemical bond in the peptidoglycan backbone to facilitate breakdown of the cell wall in recipient bacteria.
Bacteria are known to chemically modify the glycan moieties of peptidoglycan. As this can affect their susceptibility to cell wall-targeting enzymes, it is intriguing to speculate that the different peptidoglycan glycoside hydrolase effectors delivered by the T6SS could harbor specificity for the presence or absence of these modifications. Modifications influencing muramidase activity include MurNAc and GlcNAc de-N-acetylation, as well as MurNAc O-acetylation (49). These generally result in lysozyme resistance, whereas N-acetylglucosaminidase enzymes can be unaffected by the acetylation state of peptidoglycan (50). Future studies will be needed to address the specificity of the Tge families, as well as the extent to which particular cell wall modifications influence their activity.
In light of the peptidoglycan glycoside hydrolases identified in this study, it is now apparent P. aeruginosa is not the only bacterium that secretes multiple T6 effectors targeting peptidoglycan. For example, in addition to Tge2 peptidoglycan glycoside hydrolases, Salmonella Typhi and P. protegens also possess Tae2 and Tae3 peptidoglycan amidases, respectively (9). The reason for secreting a multitude of peptidoglycan-targeting T6S effectors may be myriad. It is conceivable that the T6 donor bacterium requires complete digestion of recipient peptidoglycan sacculi to utilize it as a carbon source after lysis occurs. A non-mutually exclusive explanation may be the existence of enzymatic synergy between amidase and muramidase effectors, wherein the breakdown product of one enzyme may make the substrate of the second more accessible. The importance of this type of enzymatic synergy was recently demonstrated for cellulose-degrading enzymes (51). Alternatively, the various E-I loci within a given organism may be subject to differential regulation, such that the effector most beneficial in a particular environmental condition is expressed. The magnitude of this disproportionate benefit could be dependent upon the target organism(s) present, the physiological state of the target(s) (e.g. growing versus sessile), and the osmolarity of the surrounding milieu.