A Secreted NlpC/P60 Endopeptidase from Photobacterium damselae subsp. piscicida Cleaves the Peptidoglycan of Potentially Competing Bacteria

Peptidoglycan (PG) is a major component of the bacterial cell wall formed by long chains of two alternating sugars interconnected by short peptides, generating a mesh-like structure that enwraps the bacterial cell. Although PG provides structural integrity and support for anchoring other components of the cell envelope, it is constantly being remodeled through the action of specific enzymes that cleave or join its components.

The present work reports the structural and functional characterization of a novel NlpC/P60-containing peptidase from Phdp (PnpA). The results show that PnpA is a PG hydrolase with a four-domain structure similar to that of Desulfovibrio vulgaris lysin (DvLysin) and specificity for the g-D-glutamyl-meso-diaminopimelic acid bond (26), but with a more hydrophobic and narrower access to the catalytic center. It is also shown that PnpA is secreted into the extracellular medium by the Phdp type II secretion system and acts on the PG of Vibrio anguillarum and Vibrio vulnificus, suggesting that it may provide Phdp an advantage over bacteria competing for the same resources or a way of obtaining nutrients in nutrient-scarce environments, either inside or outside the host. Comparison of the muropeptide compositions of PG, susceptible and resistant to PnpA activity, allowed development of a model suggesting that the susceptibility to PnpA is determined by three-dimensional structural features of the PG and not by their chemical compositions.

RESULTS
Photobacterium damselae subsp. piscicida secretes an NlpC/P60 family protein. Photobacterium damselae subsp. piscicida (Phdp) virulent strains have a relatively simple profile of secreted proteins in mid-exponential-growth-phase cultures (45). Apart from AIP56 toxin, no other proteins have been identified and characterized. Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) analysis of proteins from Phdp extracellular products (ECPs) precipitated with trichloroacetic acid (TCA) revealed a band of approximately 55 kDa that was excised from the gel and subjected to matrix-assisted laser desorption ionization2time of flight mass spectrometry (MALDI-TOF MS). The obtained MS data were used in a Mascot search against the NCBI database resulting in the identification of a hypothetical protein from Photobacterium damselae subsp. damselae (Phdd) CIP 102761 (VDA_000779; NCBI accession number EEZ39759). The 1,479-nucleotide homologous sequence in the Phdp MT1415 strain (accession number TJZ86030.1) was then amplified using primers designed based on the VDA_000779 sequence. In silico analysis (SignalP 5.0 and NCBI conserved domain search) of its 499-amino-acid translation product predicted a Sec signal peptide (M 1 to A 19 ), followed by an N_NLPC_P60 putative stabilizing domain (Pfam PF12912), an SH3b1 (Pfam PF12913/12914), and an NlpC_P60 domain (Pfam PF00877), classifying it as a protein belonging to the NlpC/P60 family, hereafter referred to as PnpA (Photobacterium NlpC-like protein A).
PnpA is encoded in a genetically unstable chromosomal region, and its expression levels are similar at exponential and stationary phases of growth. To investigate the genetic context of pnpA in Phdp MT1415 strain, the draft genome sequence of MT1415 was obtained in this study. Then, homologous DNA sequences of a number of Phdp and Phdd isolates were additionally retrieved from the GenBank database and subjected to comparative sequence analysis (Fig. 1). This revealed that the PnpA-encoding gene is invariably linked to a downstream gene encoding an RNase T and to an upstream gene encoding an a-galactosidase, the latter being a pseudogene in some Phdp isolates. As a whole, the DNA flanking pnpA underwent a massive insertion of transposase genes (IS elements of the IS1 and IS91 families) likely followed by accumulation of inactivating mutations, resulting in a collection of pseudogenes. This process of gene decay not only affected the transposase genes themselves but also flanking genes encoding enzymes putatively involved in sugar metabolism, as a-galactosidases, a-amylases, and pullulanases ( Fig. 1). Proliferation of insertion sequences that cause a high frequency of pseudogenes and gene loss is indeed a hallmark of all Phdp genomes studied thus far (48)(49)(50). The observation that PnpA-and the RNase T-encoding genes have escaped the inactivation by IS insertions suggests that these two genes may fulfill an important role in Phdp.
Expression levels of pnpA were determined by reverse transcription-PCR (RT-PCR), showing that under the culture conditions used (growth in tryptic soy broth supplemented with NaCl to a final concentration of 1% [wt/vol] [TSB-1] at 25°C), there are no differences in the level of gene transcription between exponential-and stationaryphase cultures (see Fig. S1 in the supplemental material).
Overall description of PnpA structure. For better understanding of the structurefunction relationship of PnpA, its three-dimensional structure was solved. The crystal structure of PnpA was determined at 1.4-Å resolution by molecular replacement with DvLysin (PDB entry 3M1U, 26% sequence identity), an endopeptidase from Desulfovibrio vulgaris Hildenborough (26). The crystal asymmetric unit contains two PnpA molecules, which are essentially identical (root mean square deviation [RMSD], of 0.5 Å for 457 aligned Ca atoms). Table S1 in the supplemental material summarizes the data collection, processing, and refinement statistics.
Analysis of the intermolecular packing interfaces within the crystal lattice suggests that the molecule behaves as a monomer in solution, which is in agreement with the molecular mass estimated by size exclusion chromatography. The PnpA monomer has an overall structure similar to that of DvLysin (26), namely, one N-terminal "c-clip" or "N_NLPC_P60" stabilizing domain (residues N 20 -N 133 ), two SH3b domains (SH3b1, residues I 134 -V 218 ; SH3b2, residues D 219 -T 295 ), and the C-terminal NlpC/P60 catalytic domain (residues P 296 -K 499 ) ( Fig. 2A). The three-dimensional models of DvLysin and PnpA display an RMSD of 2.2 Å (for 405 aligned Ca atoms), suggesting that both proteins may be functionally equivalent. A significant number of structures sharing at least one of the PnpA domains have been identified (Table S2), although so far, PnpA and DvLysin are the only four-domain NlpC/P60-containing peptidases whose structure has been reported.
As in DvLysin (26), the PnpA c-clip domain has an extended helical conformation which surrounds and stabilizes the SH3b1 and NlpC/P60 domains, forming a planar assembly from which the SH3b2 domain protrudes (Fig. S2). Compared to DvLysin, the cclip domain of PnpA harbors an extension between helices a1 and a2, thereby forming an additional two-stranded antiparallel b-sheet (b2 and b3) and a 3 10 helix (h 4), which protrude into the catalytic groove and close one of its sides (Fig. 2B).
The presence of SH3b domains in prokaryotes has long been documented. These domains have been described as targeting domains, involved in cell wall recognition and binding (1,24,35). Despite the lack of amino acid sequence conservation (8% sequence identity), the two SH3b domains in PnpA have a conserved overall fold (RMSD of 3.9 Å for 55 aligned Ca atoms) (Fig. S3). As in DvLysin (26), both PnpA SH3b domains consist of seven conserved strands (bA-bA1-bA2-bB-bC-bD-bE), with the bA-bE strands structurally equivalent to their eukaryotic counterparts (31, 32), while bA1 and bA2 form a b-hairpin that corresponds to the RT loops of eukaryotic SH3b domains ( Fig. 2A).
As in other NlpC/P60-containing peptidases, the 204-residue-long C-terminal NlpC/ P60 catalytic domain of PnpA displays a fold resembling a primitive papain-like cysteine peptidase (24). Its secondary structure elements adopt the topology described for DvLysin, i.e., a six-stranded central b-sheet and five a-helices with aA-aB-aC-bA-aD-bB-bC-bD-bE-aE-bF topology, where aA-aB-aC and aD-aE protect either side of the central b-sheet ( Fig. 2A) (26).
PnpA has a narrow and hydrophobic access to the catalytic site. The active site of NlpC/P60 cysteine peptidases consists of a conserved cysteine-histidine dyad and a third polar residue (H, N, or Q) that orients and polarizes the catalytic histidine (24)(25)(26)(27)(28)(29). In PnpA, the residues that make up the active site are C324, H395, and N415, the latter similar to the equivalent residue found in the active site of the prototypical papain (51), but differing from the histidine (H408) at the active site of DvLysin (26) (Fig. 3A). As described for other NlpC/P60-containing peptidases (24)(25)(26)(27)(28)(29), the catalytic C324 is [orange]). The N and C termini (Nter and Cter, respectively) and secondary structure elements are labeled. (B) Cartoon representation of superposed PnpA (color code as in panel A) and DvLysin (gray). N and C termini are indicated. A close-up of the insertion between a1 and a2, forming an additional antiparallel b-sheet (b2 and b3) and a 3 10 helix (h 4) in the c-clip domain, is shown in the insert (dashed oval). located at the amino terminus of a helix packing against the central b-sheet that harbors H395 in its second strand and N415 in the third. In the PnpA structure, the thiol group of the catalytic cysteine is oxidized, resulting in the disruption of the characteristic C324 SD-H395 ND1 hydrogen bond and suggesting that the enzyme is in an inactive state (Fig. S4). As advanced for Bacteroides thetaiotamicron YkfC (BtYkfC) (26), oxidation of the catalytic cysteine most likely occurred during crystallization or exposure to X-rays (52), since recombinant PnpA from the same purification batch was used in biochemical assays and was catalytically active.
In DvLysin, access to the catalytic cysteine occurs through a groove between the NlpC/P60 domain on one side and the c-clip helices aD and aE plus the SH3b1 domain on the other, with the RT loop from the SH3b1 domain closing one end of the groove (Fig. 3B) (26). While this topology is generally maintained in PnpA, the end of the groove opposite to the RT loop is also closed by strands b2 and b3 and the 3 10 helix h 4, creating a narrower access to the catalytic site (Fig. 3B). A minor difference is observed on the "wall" formed by the NlpC/P60 domain, wider in PnpA and closed by R414 and R452 in DvLysin (Fig. 3B). Besides the narrower entrance, two clusters of amino acids confer to the active site cavity of PnpA a more polar and hydrophobic nature than observed for DvLysin (Fig. 3C). However, extensive conservation of substrateinteracting residues between PnpA and DvLysin (Fig. 3C) suggests a similar interaction with meso-diaminopimelic acid (mDAP)-D-Ala from the stem peptide.
PnpA is secreted by Phdp type II secretion system. PnpA possesses a typical Sec signal peptide and was identified in the culture supernatants of exponentially growing Phdp cultures, suggesting that it could be actively secreted by the bacteria. Many proteins that are transported via the Sec system into the periplasm are secreted across the outer membrane through a type II secretion system (T2SS) (53,54). Recently, it was shown that Phdp contains a functional T2SS (44) and that deletion of epsL, which encodes an inner membrane-spanning protein that establishes a critical link between the cytoplasmic and periplasmic parts of that system (55), abolishes the secretion of AIP56 (44). To test the involvement of the T2SS of Phdp in PnpA secretion, the presence of PnpA in total cell lysates and extracellular products of wild-type (WT), DepsL, and DepsL 1 pEpsL Phdp was analyzed by Western blotting (Fig. 4). PnpA was detected in ECPs, but not in total cell lysates of the WT strain, confirming that it is a secreted protein (Fig. 4A). In contrast, in the DepsL strain, PnpA was retained in the cell, likely in the periplasm (Fig. 4B), confirming the involvement of T2SS in PnpA secretion.
PnpA has specificity for the c-D-glutamyl-meso-diaminopimelic acid bond. To investigate the PnpA enzymatic activity toward PG muropeptides and define its substrate specificity, recombinant PnpA was incubated with monomeric trimuropeptides (M3; GlcNAcMurNAc-L-Ala-D-Glu-mDap), tetramuropeptides (M4; GlcNAc-MurNAc-L-Ala-D-Glu-mDap-D-Ala), and pentamuropeptides (M5; GlcNAc-MurNAc-L-Ala-D-Glu-mDap-D-Ala-D-Ala) and the cleavage product(s) analyzed by high-performance liquid chromatography (HPLC) (Fig. 5). PnpA converted all tested muropeptides to dipeptides (M2; GlcNAc-MurNAc-L-Ala-D-Glu), suggesting it cleaves specifically g-D-glutamyl-meso-diaminopimelic acid bond of monomeric muropeptides. PnpA does not hydrolyze Phdp peptidoglycan. In order to evaluate the involvement of PnpA in Phdp cell wall biogenesis, a Phdp DpnpA strain was generated, and the absence of PnpA expression in the mutant strain was confirmed by SDS-PAGE and Western blotting ( Fig. 6A and B). Bacterial growth was not affected in the DpnpA strain (Fig. 6C). In addition, no differences were detected in the composition of the peptidoglycan from the WT and DpnpA strains ( Fig. 6D; Table 1). In agreement with this, both WT and DpnpA strains showed similar morphology (Fig. 6E). Moreover, PnpA did not display in vitro enzymatic activity against Phdp whole sacculus, since no differences in the muropeptide composition were detected after incubating the PG with active PnpA or inactive PnpA ( Fig. 6F and Fig. S5A; Table 2). Altogether, these results suggest that PnpA is not enzymatically active toward intact Phdp PG.
PnpA has hydrolytic activity toward Vibrio anguillarum and Vibrio vulnificus PG. The facts that PnpA is actively secreted into the extracellular medium and has no enzymatic activity for Phdp PG raised the possibility that it could cleave PG from other bacteria, functioning as a weapon against competing bacteria or as part of a mechanism to acquire nutrients, e.g., muropeptides from dead bacteria. To address this issue, whole sacculi from several Gram-positive or Gram-negative bacteria were isolated and incubated in vitro with recombinant PnpA or catalytically inactive PnpA (PnpA C324A ) ( Fig. 7 and Fig. S5B to J). Interestingly, only sacculi from V. anguillarum and V. vulnificus were sensitive to the action of PnpA ( Fig. 7 and Fig. S5B and C and S6). Analysis of the insoluble sacculi resulting from digestion with PnpA showed the appearance of novel   Table 2). V. anguillarum and V. vulnificus PG present a very simple muropeptide composition with three major muropeptides, the monomer GM-tetrapeptide (GM4), the dimer GM4-GM4, and the anhydro-dimer (GM4-GanhM4 and GanhM4-GM4). The high proportion of anhydro-muropeptides indicates that V. vulnificus has a PG with short glycan chains ( Table 2). PnpA treatment led to the appearance of four new muropeptides, GM2, GanhM2, GM4-mDapA, and GanhM4-mDapA. GM2 and GanhM2 products are consistent with the hydrolysis of the g-D-glutamyl-meso-diaminopimelic acid bond. The presence of GM4-mDapA and GanhM4-mDapA are also consistent with the hydrolysis of a dimer or higher oligomers such as the major dimers GM4-GM4 and GM4-GanhM4 and the trimers GM3-GM4-GM4 and GM4-GM4-GM4 (Table 2) at the g-D-glutamyl-mesodiaminopimelic acid bond at one of the 4-amino-acid stem peptides.
Analysis of the products released from the V. vulnificus PG identified two main tetrasaccharides substituted with the L-alanine-D-glutamate dipeptide (GM2-GanhM2) and/ or a remain of the dimer cross-link (GM4-GanhM4-mDapA; Fig. 7 and Fig. S6; Table 3). Additionally, the GanhM2 monomer, the remains of the monomer stem peptide mDapA and of dimer cross-link mDapA-mDapA were also released, confirming that PnpA is indeed a g-D-glutamyl-meso-diaminopimeate endopeptidase ( Fig. 7 and Fig. S6; Table 3).
In order to assess whether PnpA could inhibit the growth of competitor bacteria, the growth of V. vulnificus was monitored in the presence of PnpA (5 mg ml 21 ), and no growth inhibition was observed (Fig. S7A). To test the hypothesis that an additional factor secreted by Phdp could assist PnpA in reaching the PG, the growth of V. vulnificus was monitored in the presence of ECPs from wild-type or DpnpA Phdp (Fig. S7B) and in coculture experiments (Fig. S7C). No growth inhibition was observed in any of these experiments. Finally, it was tested whether PnpA was able to inhibit the growth of V. vulnificus in the presence of EDTA, an external membrane-permeabilizing agent used to mimic conditions that may be encountered in the host, and no effect on growth was observed (Fig. S7D).

DISCUSSION
In this work, the structural and functional characterization of PnpA, an NlpC/P60 family peptidase secreted by Photobacterium damselae subsp. piscicida (Phdp) is reported. PnpA is not essential for Phdp cell wall biogenesis and does not cleave Phdp PG, but it degrades the PG of V. anguillarum and V. vulnificus, two bacterial species that share the same hosts and/or environment as Phdp. On the basis of these observations, it is proposed that PnpA may allow Phdp to fight competitors or to acquire nutrients from dead coinhabitants.
Many cysteine peptidases containing the NlpC/P60 domain were characterized thus far (1,2,24,26,30,(33)(34)(35), several of which display a four-domain organization similar to PnpA. However, until now, only the three-dimensional structure of DvLysin from Desulfovibrio vulgaris was reported, with a N-terminal "c-clip" or "N_NLPC_P60" stabilizing domain, two SH3b domains, and a C-terminal NlpC/P60 cysteine peptidase domain (26). Furthermore, among the known DvLysin and PnpA orthologs, only EcgA from    Salmonella enterica serovar Typhimurium was functionally characterized (56). Although the three molecules are very similar (25 to 27% amino acid sequence identity) (see Fig. S8A in the supplemental material), DvLysin does not have the insertion found in PnpA and EcgA and that in PnpA closes the side of the catalytic groove opposed to the RT loop ( Fig. 2 and 3 and Fig. S8B). Despite these differences, residues involved in substrate binding in DvLysin (26) are conserved in PnpA and EcgA ( Fig. 3C and Fig. S8B), in agreement with their specificity for the g-D-glutamyl-meso-diaminopimelic acid bond (Fig. 5) (26, 56). However, unlike DvLysin (26) and EcgA (56), which were more active toward tetra-and trimuropeptides, respectively, PnpA showed activity toward penta-, tetra-, and tripeptides (Fig. 5). So far, the cellular localization of DvLysin and its function in D. vulgaris cell wall biogenesis remain unknown (26). Regarding EcgA, its expression is induced when S. Typhimurium is inside eukaryotic cells, localizing in the inner and outer membranes where it plays a role in PG remodeling and contributes to S. Typhimurium virulence (56). In contrast, PnpA is secreted by the T2SS into the extracellular medium (Fig. 4), and deletion of pnpA does not affect Phdp growth, PG composition, and morphology ( Fig. 6C to E). Accordingly, PnpA has no in vitro hydrolytic activity toward Phdp sacculi ( Fig. 6F and Fig. S5A). Altogether, these results suggest that PnpA is not involved in Phdp cell wall biogenesis.
The resistance of Phdp PG to the activity of PnpA is in sharp contrast with the ability of PnpA to hydrolyze penta-, tetra-, and trimuropeptides, since the chemical composition of Phdp PG suggested that it would be a target of PnpA. This unexpected resistance to PnpA was not exclusively observed with PG from Phdp, as it also occurred when using sacculi from multiple bacterial species (Fig. S5). In fact, PGs from V. anguillarum and V. vulnificus were sensitive to the activity of PnpA, despite having a PG composition characteristic of Gram-negative bacteria and similar to the composition of some PG shown to be resistant to PnpA hydrolysis. Hence, PnpA specificity for V. anguillarum and V. vulnificus PG cannot be explained by their muropeptide composition and may be related to specific three-dimensional features of the PG mesh. Accordingly, the analysis of the V. anguillarum and V. vulnificus PG composition shows that these two species have a high proportion of anhydro-muropeptides, a trademark of the end of glycans, indicating that their glycan chains are rather short compared to other Gram-negative bacteria. Consequently, structural analysis of the products released upon incubation of the sacculi of V. anguillarum and V. vulnificus with PnpA identified a high proportion of the tetrasaccharide GM2-GanhM2. This suggests that the PG of V. anguillarum and V. vulnificus is enriched in tetrasaccharides. The simultaneous release of mDapA-mDapA suggests that these tetrasaccharides are linked to the rest of the PG by one or even two cross-links. These results combined with the rather simple muropeptide composition of V. anguillarum and V. vulnificus suggest that the vulnerability of V. anguillarum and V. vulnificus to PnpA might arise from the fact that their PGs rely on very short, highly cross-linked glycans. Hence, hydrolysis of the stem peptides by PnpA leads to a rapid destruction of the PG layer while in other Gram-negative species, because they have much longer glycans, PG integrity can be maintained by multiple dimers along the same glycan chain (Fig. 8).
Expression levels of pnpA in standard culture conditions do not vary between the logarithmic and stationary growth phases (Fig. S1) but increase under iron-limited conditions or in response to oxidative stress (57). However, in vivo, no changes in pnpA expression were detected after intraperitoneal infection of sole (Solea senegalensis) with Phdp (57), and deletion of pnpA did not affect Phdp virulence in a sea bass (Dicentrarchus labrax) intraperitoneal infection model (Fig. S7E). This suggests that PnpA is likely dispensable at late systemic phases of Phdp infection but does not exclude a role of PnpA in earlier stages of the infection. It is known that, during the systemic phase of Phdp-induced disease, the exotoxin AIP56 plays a major role by neutralizing host phagocytic defenses (43-45, 47, 58). However, little is known about the early stages of the infection. Here, it is shown that PnpA specifically hydrolyzes the sacculi of V. anguillarum and V. vulnificus (Fig. 7, Fig. S5B and C, and Fig. S6), two other enterobacteria present in the marine environment (59)(60)(61) and, at least in the case of V. anguillarum, reported as infecting the same hosts as Phdp (37,61). This suggests that before reaching the systemic phase, Phdp may secrete PnpA to gain competitive growth advantage over bacteria sharing a complex community environment, such as the gastrointestinal tract, or to obtain nutrients in an environment where nutrient scarcity can compromise its survival, either inside the host or in water or sediment (40). These strategies have been first described for Gram-positive bacteria (21,23), which have their PG exposed on the cell surface, accessible to secreted PG hydrolases (10,11,22). Gram-negative bacteria, despite having their PG protected by the outer membrane, can inject PG hydrolases, including NlpC/P60 family peptidases, into the periplasm of neighboring bacteria through type VI secretion systems (10,17,19,20). The examples of using bacterial exohydrolases to target Gram-negative competitors are restricted to predatory bacteria such as myxobacteria (11) and Bdellovibrio bacteriovorus (18). Another example where PG hydrolases are secreted to eliminate competing bacteria is that reported for the urogenital pathogenic protozoan Trichomonas vaginalis (16), which has acquired by lateral genetic transfer two genes of bacterial origin encoding NlpC/P60 endopeptidases that the parasite secretes to degrade bacterial PG and thus outcompete bacteria from mixed cultures (16). However, it remains unclear how these exohydrolases reach the PG of the Gram-negative targets. Here, it was also not clarified how PnpA reaches the PG in V. vulnificus and V. anguillarum cell wall, since no growth inhibition was detected in several in vitro tests with V. vulnificus (Fig. S7), suggesting that the access of PnpA to the periplasm of competing bacteria may depend on conditions present at specific stages of the Phdp life cycle, when Phdp and competitors meet.

MATERIALS AND METHODS
Bacterial strains and culture conditions. Photobacterium damselae subsp. piscicida (Phdp) virulent strain MT1415 isolated from sea bass in Italy (45) was cultured at 25°C in tryptic soy broth (TSB) or tryptic soy agar (TSA) supplemented with NaCl to a final concentration of 1% (wt/vol) (TSB-1 and TSA-1, respectively). The DepsL and DpnpA strains were cultured under the same conditions as the wild type. DepsL 1 pEpsL and DpnpA 1 pPnpA complemented strains were cultured in TSB-1 or TSA-1 supplemented with 10 mg ml 21 of gentamicin (TSB-1 Gm and TSA-1 Gm , respectively). Stocks of bacteria were maintained at -80°C in TSB-1 supplemented with 15% (vol/vol) glycerol. To obtain growth curves, bacteria grown on agar plates for 48 h were suspended in TSB-1 or TSB-1 Gm at an optical density at 600 nm (OD 600 ) of 0.5 to 0.6. These suspensions were inoculated in 20 ml TSB-1 (1:100 dilution). One-milliliter aliquots were removed (in triplicate) and transferred to 24-well culture plate, and the OD 600 was determined kinetically (1 point/h) using a BioTek Synergy 2 spectrofluorometer (BioTeK U.S., Winooski, VT, USA) at 25°C with continuous slow agitation, for 60 to 70 h. Growth curves were constructed using GraphPad Prism software (La Jolla, CA, USA).
Construction of DpnpA strain. An in-frame (nonpolar) deletion of the almost complete pnpA coding sequence was constructed following an allelic exchange procedure as previously described (62). In brief, the 39 and 59 flanking sequences were PCR amplified using suitable primers (Mut_NlpC_1Eco . The PCR products were ligated to obtain an inframe deletion of ca. 90% of the PnpA coding sequence. The deleted allele construction was cloned into the suicide vector pNidKan containing the sacB gene, which confers sucrose sensitivity, and R6K ori, which requires the pir gene product for replication. The plasmid containing the deleted allele was transferred from Escherichia coli S17-1-lpir into the rifampin-resistant derivative of Phdp MT1415 by drop mating for 24 h on TSA plates prepared with seawater. Cells were then scrapped off the plate and selected on TSA supplemented with kanamycin (Kan) (50mg ml 21 ) for plasmid integration. A selected Kan r clone was further selected for sucrose resistance (15% [wt/vol]) for a second recombination event. This led to Phdp DpnpA mutant strain, which was tested by PCR to verify the correct allelic exchange.
Bacterial cell extracts and extracellular products. Phdp was grown in TSB-1 at 25°C with shaking (160 rpm) and centrifuged (6,000 Â g, 5 min, 4°C), and the pellets (total cell extracts) and culture supernatants were collected. Supernatants were filtered (0.22 mm) to obtain extracellular products (ECPs). For SDS-PAGE, proteins in the ECPs were precipitated with trichloroacetic acid (TCA) as previously described (45).
PnpA identification. ECPs from Phdp strain MT1415 were subjected to SDS-PAGE followed by Coomassie blue staining. A protein band of approximately 55 kDa was analyzed by MALDI-TOF MS in a 4800 Proteomics Analyzer (Applied Biosystems) at TOPLAB GmbH. The MS data were used for a Mascot search against the NCBInr sequence database.
Draft genome sequence of Phdp MT1415 and genomic context of pnpA locus. To delete the PnpA-encoding gene in Phdp MT1415, it was necessary to obtain at least 2 kb of upstream and downstream sequences free of repetitive insertion sequence elements that would compromise the specific recombination steps during allelic exchange. Therefore, the draft genome sequence of strain MT1415 was obtained, using an Illumina platform as previously described (48) and deposited in the GenBank database under accession number SUMH00000000. A comparative analysis was conducted by retrieving the genomic contexts of pnpA genes in different Phdp and Photobacterium damselae subsp. damselae (Phdd) isolates whose draft or complete genomes are available in the GenBank database. The GenBank locus tag numbers of the pnpA homologues used in this analysis are VDA_000779 (Phdd type strain CIP 102761), PDPUS_2_00834 (Phdp 91-197), PDPJ_2_00460 (Phdp OT-51443), BEI67_17705 (Phdp L091106-03H), and BDMQ01000002 (Phdp DI21). For the pnpA negative Phdd strain RM-71, the draft genome sequence as a source of homologous flanking DNA sequences was used (accession number NZ_LYBT00000000.1). The DNA sequences were handled with Vector NTI 10.3.0 sequence editor (Invitrogen).
Recombinant PnpA. The pnpA open reading frame (ORF) (GenBank accession number TJZ86030.1) was amplified from Phdp MT1415 genomic DNA using Pfu DNA polymerase (Thermo Scientific) and primers 59-cgcccATGGATATAAATAAACATTTAATGC-39 and 59-gcgctcgagTTTTTCAAATAGATATTTTTC-39 (target sequences are in uppercase letters) and cloned into pET28a(1) using the NcoI and XhoI restriction sites, in frame with a C-terminal 6ÂHis tag. Mutation of C 324 to alanine was achieved by site-directed mutagenesis by inverse PCR using Q5 high fidelity DNA polymerase (New England BioLabs), pET28-PnpA as the template, and primers (59-GCCTCTGGTTTATTAAAAAGGTTATTCAGC-39 and 59-ATCATTATTGAAATC-CATTCCCCC-39). Proteins were expressed in E. coli BL21(DE3) CodonPlus-RIL (Stratagene). Four liters of LB medium with 50 mg ml 21 kanamycin and 25mg ml 21 chloramphenicol were inoculated with pET28-PnpAor pET28-PnpA C324A -transformed bacteria and incubated at 37°C until an OD 600 of 0.6 to 0.8 was reached. Cultures were cooled at 17°C for 30 min, followed by the addition of 0.5 mM isopropyl-b-D-thiogalactopyranoside (IPTG) to induce protein expression. After 20 h, cells were harvested by centrifugation, resuspended in 50 mM Bis-Tris (pH 6.5) and 500 mM NaCl, and sonicated. Lysates were centrifuged (34,957 Â g, 30 min, 4°C), and the soluble fraction was applied to a nickel-nitrilotriacetic acid (Ni-NTA) column (ABT), followed by anion-exchange chromatography (Bio-Scale Mini Macro-Prep High Q; Bio-Rad). Fractions containing the recombinant proteins were pooled and injected into a size exclusion chromatography column (Superose12 10/300 GL; GE Healthcare) equilibrated with 50 mM Bis-Tris (pH 6.5) and 500 mM NaCl. Fractions containing the desired protein were pooled, concentrated to 6 to 7 mg ml 21 , frozen in liquid nitrogen, and stored at -80°C. Protein concentration was determined in a NanoDrop ND-1000 UV-visible (UV-Vis) spectrophotometer (Thermo Fisher Scientific) considering the extinction coefficient and the molecular weight calculated with the ProtParam tool (https://web.expasy.org/protparam/).
Crystallization. Initial crystallization hits for PnpA were identified by high-throughput screening performed at the HTX Lab of the EMBL Grenoble Outstation (Grenoble, France). Crystallization experiments for refinement of the initial conditions were carried out using the hanging drop vapor diffusion method at 20°C. Crystals were obtained by mixing protein solution (6.7 mg ml 21 in 50 mM Bis-Tris [pH 6.5] and 500 mM NaCl) with an equal volume of crystallization solution (100 mM imidazole [pH 8.0], 15% [wt/vol] polyethylene 8000 [PEG 8K]). Crystals appeared after 24 to 48 h. The crystals were cryo-protected by sequential transfer into their crystallization condition with increasing concentrations of ethylene glycol (up to 30% [vol/vol]) and then flash-frozen in liquid nitrogen prior to data collection.
Data collection, structure solution, and refinement. Diffraction data were collected at beamline Proxima-1 of Synchrotron SOLEIL (Saint-Aubin, France) (64) on a Dectris Pilatus 6M detector (750 images, 0.2°rotation, 0.2-s exposure) and indexed and integrated with XDS (65). Space group determination, data scaling, and merging were performed with POINTLESS and AIMLESS from the CCP4 program suite (66). The structure of PnpA was solved by molecular replacement with Phaser MR as implemented in the CCP4 program suite (66,67) using the coordinates of a putative gamma-D-glutamyl-L-diamino acid endopeptidase from Desulfovibrio vulgaris Hildenborough (DvLysin, PDB entry 3M1U, 26% sequence identity) as the search model. Phase refinement and initial model building were performed using ARP/wARP (68). Model completion and refinement were done iteratively with COOT (69) and Phenix.refine (70,71), respectively. Refinement and structure validation statistics are summarized in Table S1 in the supplemental material. All illustrations of macromolecular models were produced with PyMOL (72). The experimental data were deposited with the Structural Biology Data Grid (73)  products of the reaction were analyzed by reverse-phase HPLC (Waters 1525 system) as previously described (56).
Peptidoglycan (PG) purification. Bacteria were grown in TSB-1 at 25°C with shaking (160 rpm) to exponential (OD 600 of 0.4 to 0.5) or stationary (OD 600 of 1.2 to 1.4) phases. Bacterial cells (;10 11 ) were centrifuged (4,200 Â g, 10 min, room temperature [rt]), washed twice and resuspended in phosphatebuffered saline (PBS), and immediately mixed 1:1 (vol/vol) with a boiling solution of 8% SDS, drop by drop. Boiling was maintained for 8 h with stirring, followed by overnight incubation at rt. Samples were centrifuged (150,000 Â g, 40 min, 4°C), the pellets were washed three times with ultrapure water (150,000 Â g, 40 min, 4°C), resuspended in 10 mM Tris (pH 7.6) and 0.06% (wt/vol) NaCl with or without 100 mg ml 21 a-amylase, and incubated at 37°C for 90 min. Samples were treated for 2 h at 60°C with 100 mg ml 21 pronase E preactivated by incubation in the same buffer for 60 min at 60°C. Pronase E digestion was stopped by adding SDS (5.3% [wt/vol] final concentration) and heating at 100°C for 20 min. PG was recovered by centrifugation (300,000 Â g, 10 min) and washed with ultrapure water.
Analysis of Phdp PG composition and PG cleavage assays. To analyze the PG composition of the Phdp MT1415 and MT1415DpnpA strains, PGs were purified as described above, digested overnight at 37°C in sodium phosphate buffer supplemented with 100 IU of mutanolysin from Streptomyces globisporus (ATCC 21553; Sigma), and reduced with NaH 4 B. After 30 min at rt and centrifugation, the reduced muropeptides were diluted in acidified water with formic acid (FA) and analyzed by high-performance liquid chromatography (HPLC) or HPLC/high-resolution mass spectrometry (HRMS). HPLC/HRMS was performed on an Ultimate 3000 UHLPC system coupled to a quadrupole orbitrap mass spectrometer (qExactive Focus; Thermo Fisher Scientific). Reduced muropeptides were eluted on an C 18 analytical column (Hypersil gold aQ; 1.9 mm, 2.1 Â 150 mm) held at 50°C under a 200 ml min 21 flow rate. A binary solvent system composed of acidified water (H 2 O 1 0.1% FA; mobile phase A) and acidified acetonitrile (CH 3 CN 1 0.1% FA, mobile phase B) was used for chromatographic separation. The composition was linearly increased to 12.5% mobile phase B over 25 min, increased to 20% mobile phase B for 5 min, and held for an additional 5 min. It was then stepped down to 0% over and held for 10 min to return initial conditions.
Exactive Focus was operated under electrospray ionization in positive mode and data-dependent acquisition mode (ddMS2) control by Xcalibur 4.0. For structural confirmation of muropeptides, higherenergy collisional dissociation (HCD) fragmentation was set up with a normalized collision energy at 20%. Data were processed both with the software TraceFinder 3.3 (Thermo Fisher Scientific) and Xcalibur 4.0 for peak area determination.
For testing PnpA activity against macromolecular PG, PGs from Phdp and several bacterial species, purified as described above, were incubated with 100 mg PnpA or inactive PnpA C324A at 37°C overnight in 50 mM Tris (pH 8.0) and 300 mM NaCl. PGs incubated with vehicle were used as controls. After digestion, PGs were analyzed by HPLC or HPLC/HRMS as described above.
Accession number(s). The draft genome sequence of strain MT1415 was obtained and deposited in the GenBank database under accession number SUMH00000000. The experimental data were deposited with the Structural Biology Data Grid (73)

ACKNOWLEDGMENTS
We are grateful for access to the HTX crystallization facility (Proposal ID: BIOSTRUCTX_8167). The support of the X-ray Crystallography Scientific Platform of i3S (Porto, Portugal) is also acknowledged.