Structural analysis of a Vibrio phospholipase reveals an unusual Ser–His–chloride catalytic triad

Phospholipases can disrupt host membranes and are important virulence factors in many pathogens. VvPlpA is a phospholipase A2 secreted by Vibrio vulnificus and essential for virulence. Its homologs, termed thermolabile hemolysins (TLHs), are widely distributed in Vibrio bacteria, but no structural information for this virulence factor class is available. Herein, we report the crystal structure of VvPlpA to 1.4-Å resolution, revealing that VvPlpA contains an N-terminal domain of unknown function and a C-terminal phospholipase domain and that these two domains are packed closely together. The phospholipase domain adopts a typical SGNH hydrolase fold, containing the four conserved catalytic residues Ser, Gly, Asn, and His. Interestingly, the structure also disclosed that the phospholipase domain accommodates a chloride ion near the catalytic His residue. The chloride is five-coordinated in a distorted bipyramid geometry, accepting hydrogen bonds from a water molecule and the amino groups of surrounding residues. This chloride substitutes for the most common Asp/Glu residue and forms an unusual Ser–His–chloride catalytic triad in VvPlpA. The chloride may orient the catalytic His and stabilize the charge on its imidazole ring during catalysis. Indeed, VvPlpA activity depended on chloride concentration, confirming the important role of chloride in catalysis. The VvPlpA structure also revealed a large hydrophobic substrate-binding pocket that is capable of accommodating a long-chain acyl group. Our results provide the first structure of the TLH family and uncover an unusual Ser–His–chloride catalytic triad, expanding our knowledge on the biological role of chloride.

Phospholipases are lipolytic enzymes that hydrolyze glycerophospholipids, the major component of eukaryotic membranes. It is known that many pathogens produce phospho-lipases (1), such as ExoU of Pseudomonas aeruginosa (2) and SlaA of group A Streptococcus (3). During infection, these phospholipases can directly lyse host cells by disrupting the phospholipid membrane, and the products from cell lysis, such as lysophosphatidylcholine, can act as signaling molecules to further induce apoptosis and inflammation (4). Therefore, phospholipases are important virulence factors of these pathogens and can be promising targets for antivirulence therapy.
Vibrio vulnificus is an opportunistic pathogen for human and marine animals, and it causes gastroenteritis and even severe septicemia with a high mortality rate of Ͼ 50% (5). The bacterium secretes a phospholipase, VvPlpA, 3 by a type II secretion system. VvPlpA shows phospholipase A 2 activity, with a substrate preference toward phosphatidylcholine (a major component of lecithin). VvPlpA is significantly induced in a host environment. It can disrupt the phospholipid membrane and thus lead to host cell lysis, contributing to systemic infection (6). Therefore, VvPlpA is an essential virulence factor of V. vulnificus and may serve as a promising antivirulence drug target.
Sequence analysis has revealed that mature VvPlpA is composed of an N-terminal domain with unknown function and a C-terminal phospholipase A 2 domain, which belongs to the SGNH hydrolase superfamily (Fig. 1a). The SGNH hydrolase superfamily is characterized by containing four invariant catalytic residues (Ser, Gly, Asn, and His, which indeed give the name SGNH) in four conserved blocks (I, II, III, and V) (7). Ser is in Block I and acts as the nucleophile. The Ser residue, Gly residue (in Block II), and Asn residue (in Block III) function together as proton donors to the oxyanion hole. Block V contains a His that acts as a base to polarize the Ser nucleophile by deprotonating the hydroxyl group. In addition, Block V can contain an acid, such as Asp or Glu, at the third residue preceding the catalytic His, thus forming a Ser-His-Asp/Glu catalytic triad. However, the Asp/Glu at this position is not conserved and thus it is not included in the name. Residues other than Asp/Glu often make the catalytic triad degrade to a catalytic dyad (8). It is worth noting that the residue of VvPlpA at this position is a Gly. The influence of this kind of residue variation on catalytic site composition has not been reported previously.
Based on sequence analysis, VvPlpA homologs are widely distributed in Vibrio (see Fig. 1a), such as VHH in Vibrio harveyi (9,10), VaPlpA in Vibrio anguillarum (11), Lec in Vibrio cholera (12), and PhlA in Vibrio mimicus (13). These proteins are also annotated as thermolabile hemolysins (TLHs) or lecithin-dependent hemolysins in terms of their ability to hydrolyze erythrocyte membranes. Some of these proteins have been proposed to be important virulence factors. Structural information of this protein family would be greatly helpful for understanding their functions and for developing inhibitors targeting them.
Herein, we report the crystal structure of VvPlpA, which represents the first characterized structure of the TLH family. The results reveal an unusual catalytic triad as well as substratebinding pocket information, shedding light on the catalytic mechanism of VvPlpA and other TLH family members.

Overall structure of VvPlpA
The structure of VvPlpA was solved by the single-wavelength anomalous dispersion method, with one monomer per asymmetric unit. Although the signal peptide of VvPlpA was included to facilitate soluble expression of the protein, this part is disordered and shows no electron density. Based on electron density, residues 24 -83 and 87-417 of VvPlpA and five His residues from the fusion tag were built into the final structural model. Three residues (Gly-157, Leu-175, and Val-211) are marked as outliers on the Ramachandran plot; however, they were correctly modeled, as confirmed by electron density inspection.
Consistent with sequence analysis, the VvPlpA structure can be divided into two domains (Fig. 1b). The N-terminal domain (residues 24 -133) is a functionally unknown domain and mainly comprises ␤ strands. In the middle of this domain is ␤1, which is a rather long ␤ strand. At one side of ␤1, two short ␤ strands (␤2 and ␤5) together with ␤1 form a small threestranded antiparallel sheet in the order of ␤2-␤1-␤5, whereas at the other side three short ␤ strands (␤3, ␤4, and ␤6) along with ␤1 form a four-stranded antiparallel sheet in the order of ␤1-␤3-␤4 -␤6. This domain also contains helices, including the 3 10 helix 1, ␣1 preceding ␤1, and ␣2 between ␤4 and ␤5. The C-terminal domain (residues 134 -417) is a phospholipase domain belonging to the SGNH hydrolase superfamily and adopts an ␣/␤/␣ architecture. In the middle of the C-terminal domain lies a parallel ␤ sheet composed of five strands in the order of ␤8 -␤7-␤9 -␤10 -␤11. One side of this ␤ sheet is flanked by helices ␣5 and ␣7-8, and the other side is flanked by helices ␣4 and ␣11. Remarkably, between these secondary structures are unusually long curly loops in which a couple of small secondary structural elements, including 3 10 and ␣ helices and ␤ strands, are sparsely distributed. Notably, the N-terminal domain and C-terminal domain of VvPlpA are closely packed against each other. The interface is estimated to be ϳ1700 Å 2 for the buried solvent-accessible surface, with extensive hydrogen bonding, salt bridge, and hydrophobic interactions between the two domains.
Generally, structurally and functionally important residues are evolutionarily conserved among homologs. Hence, we marked the conserved residues of the TLH family on the VvPlpA structure for the sake of uncovering functionally important regions (Fig. 1, a and c). The conserved residues are mainly distributed in the interior part of the phospholipase domain of VvPlpA. This conserved part contains residues of Blocks I, II, III, and V and represents the putative active site. In addition, ␤1 of the N-terminal domain is also rather conserved, which implies a potential role of ␤1 in protein function.

Identifying a chloride ion at the Block V region
At the region where Block V residues His-392 and Gly-389 are located, an autoassigned water molecule drew our attention. We observed significant positive peaks on the difference electron-density map (Fig. S1a). Due to the excellent quality of the diffraction data, we were able to calculate an anomalous map, which showed weak anomalous peaks for the sulfur atoms of some cysteine and methionine residues. Remarkably, a clear anomalous peak occurs at this position as well. Therefore, a heavier ion instead of a water molecule should be modeled here.
Then we intended to model this peak with a metal ion, which is commonly seen in protein structures (14). However, modeling a metal is not sufficient for interpreting the electron density. Modeling a Mg 2ϩ showed a positive peak, whereas modeling a Ca 2ϩ or K ϩ showed a significant negative peak on the Fourier difference map (Fig. S1). Furthermore, the coordinating distance to this ion is longer than 3.17 Å (Fig. S1), deviating significantly from metal-protein interaction geometry (15,16). More importantly, the chemical groups around this ion are mainly NH groups, which are unlikely to donate lone electron pairs to coordinate a metal ion at the current geometric setting. Therefore, we excluded the possibility of modeling a metal ion at this position. Indeed, supplementing a metal ion, such as Mg 2ϩ , Ca 2ϩ , or K ϩ , did not enhance the enzymatic activity of VvPlpA (Fig. S2). Moreover, adding EDTA to remove possible bound metal ions from the enzyme only slightly affected enzymatic activity (Fig. S2). These results suggest that VvPlpA activity is independent of metal ions, supporting our structural analysis.
Subsequently, we modeled this peak as a chloride ion (Cl Ϫ ) based on three aspects. First, Cl Ϫ is able to show a weak anomalous signal even at the wavelength that was used for our data collection (f ٞ ϭ ϳ0.3 electron at 0.978-Å wavelength) (Fig. 2a). Second, Cl Ϫ can well interpret the electron density with good refinement behavior (Figs. 2b and S3a). Third, the coordination chemistry is reasonable (Fig. 2b). Cl Ϫ is five-coordinated with a distorted trigonal-bipyramid geometry, accepting hydrogen bonds from the main-chain NH groups of Gly-389 and His-392, the NH groups of the imidazole ring of His-392 and the indole ring of Trp-173, and a water molecule. The hydrogen-bonding distances are in a range of 3.19 -3.41 Å, which is in good agreement with chloride chemistry identified from known structures (17).
Finding a Cl Ϫ in an enzyme is unusual, particularly at a position that is close to catalytically relevant residues. To confirm the correctness of our modeling, we performed a bromide (Br Ϫ ) substitution experiment (18). Generally, Cl Ϫ should be substituted by Br Ϫ , which has similar chemistry, with the advantage EDITORS' PICK: The Ser-His-chloride catalytic triad of being much heavier and thus easily distinguishable on the electron-density map. Furthermore, Br Ϫ can show a strong anomalous signal at its absorption edge (0.920 Å) that is easily tuned on a synchrotron. We expressed, purified, and crystal-lized VvPlpA in solutions containing Br Ϫ in place of Cl Ϫ . The diffraction data were collected at 0.916 Å (the higher-energy side of the absorption edge of bromine). As expected, a very strong anomalous signal peak with a height of ϳ44 was found EDITORS' PICK: The Ser-His-chloride catalytic triad at the position where Cl Ϫ was located (Fig. 2c). Additionally, a significant peak with a height of ϳ15 at the same position showed up on the isomorphous difference Fourier map calculated with the diffraction data of Cl Ϫ -bound and Br Ϫ -bound VvPlpA (Fig. 2c). Moreover, Br Ϫ fits the electron density well, and the coordinating geometry is similar to that of Cl Ϫ (Figs. 2d and S3b). Therefore, the Cl Ϫ was successfully replaced by a Br Ϫ , confirming the correctness of modeling a Cl Ϫ in the VvPlpA structure.

The unusual Ser-His-chloride catalytic triad
Because VvPlpA belongs to the SGNH hydrolase superfamily, the active-site residues are easily identified from sequence analysis (Figs. 1, a and b, and 3). The nucleophile Ser-152 is located at the loop region following ␤7; the base His-392 is located at the loop between ␣11 and ␤13, with the N⑀2 atom of its imidazole ring forming a short hydrogen bond with the hydroxyl group of Ser-152. The third acid component, Asp/ Glu, commonly seen in the SGNH hydrolase superfamily is missing in VvPlpA, and it is substituted with a Gly residue (Fig.  1a). Instead, a Cl Ϫ sits at the position where the carboxyl group of conventional Asp is located (Figs. 2b and 3). This Cl Ϫ accepts a hydrogen bond from the N⑀2 atom of catalytic His-392, with the role of maintaining the proper orientation and tautomeric state of the imidazole ring of His-392. In addition, the negative charge of Cl Ϫ can potentially stabilize the positive charge of the imidazole ring of His-392 that develops during catalysis. Therefore, this Cl Ϫ may functionally subrogate Asp to be the third component of the catalytic triad of VvPlpA. Tyr-367 is also close to His-392 but only forms a very weak hydrogen bond with His-392, which implies a trivial role of Tyr-367 in orientating His-392. Residues Ser-152, Gly-203, and Asn-247 are supposed to constitute the oxyanion hole in VvPlpA. Interestingly, although the main-chain NH groups of Ser-152 and Gly-

EDITORS' PICK:
The Ser-His-chloride catalytic triad 203 are in the positions for donating hydrogen bonds to the oxyanion hole, the carboxamide group of Asn-247 seems to occupy the position of the oxyanion hole, forming hydrogen bonds with Ser-152 and Gly-203 (Fig. 3). Therefore, the oxyanion hole observed here is in an inactive state, and a conformational change at this site, probably induced by substrate binding, is expected for constructing a functional oxyanion hole.
We tested the roles of individual residues at the catalytic site by mutagenesis and enzymatic experiments. The mutants of the nucleophile Ser-152 (S152G), the base His-392 (H392N), and the oxyanion hole residue Asn-247 (N247D) all exhibited negligible enzymatic activity, which confirmed their key roles in catalysis, as expected for an SGNH enzyme. To understand the contribution of Tyr-367 to catalysis, we mutated it to a Phe. Compared with the wildtype (WT) enzyme, Y367F exhibited a significantly higher K m and a slightly decreased k cat ( Table 1), indicating that Tyr-367 primarily contributes to substrate binding and that the hydrogen bonding to His-392 is not essential for the catalytic power.
The importance of Cl Ϫ was tested by monitoring the chloride dependence of enzymatic activity. The Cl Ϫ -depleted VvPlpA was prepared by the gel-filtration method (19), and subsequently a certain amount of NaCl was added to the sample for the enzymatic activity assay. VvPlpA activity increased with increasing Cl Ϫ concentration in a specific manner (Fig. 4). Depletion of Cl Ϫ led to a loss of more than 90% of enzymatic activity. In contrast, the activity of G389D, in which the Cl Ϫ was replaced by the Asp-389 side chain (Fig. S4a), was not dependent on Cl Ϫ . Therefore, specific binding of a Cl Ϫ at this position is required for the full catalytic power of VvPlpA. By fitting the activity curve, the binding affinity for Cl Ϫ was estimated with an apparent K d of 6.6 Ϯ 1.2 mM. As expected, Br Ϫ could also enhance enzymatic activity of the Cl Ϫ -depleted VvPlpA and was specifically bound to the enzyme with an apparent K d of 5.9 Ϯ 1.3 mM (Fig. S5). Combining the structural and enzymatic evidence, we propose that this Cl Ϫ acts as an essential component of the catalytic site.
To probe the catalytic roles of Cl Ϫ , we compared the activity of WT with that of the G389D and G389N mutants, which were supposed to replace Cl Ϫ to form a catalytic triad and a catalytic dyad, respectively. In the G389D structure, the mutated Asp-389, which is generally seen as the third component in the classic catalytic triad, successfully substituted for the Cl Ϫ (Fig. S4a), in agreement with the fact that activity of G389D is independent of Cl Ϫ (Fig. 4). Except for the mutated residue, the local structure at the active site of G389D is essentially similar to that of WT. In the G389D mutant, Asp-389 forms a hydrogen bond with catalytic His-392 and seemingly pulls His-392 away from the catalytic Ser-152, with the hydrogen bond between Ser-152 and His-392 being longer than that in WT (Fig. S4a). G389D showed a lower k cat and similar K m compared with those of WT ( Table 1), suggesting that Cl Ϫ can functionally subrogate Asp with an even better catalytic power. Concerning the G389N mutant, the mutated Asn-389 was supposed to occupy a position similar to that of Asp-389 and thus replace the Cl Ϫ . Because Asn lacks the negative charge that can be provided by Cl Ϫ or Asp, the catalytic activity of the G389N mutant was expected to decrease dramatically, as reported for the D102N mutant of trypsin where the catalytic power was decreased by 10,000-fold (20). Interestingly, G389N was crystallized in a different space group, even under a crystallization condition similar to that of G389D. Asn-389 indeed replaces the Cl Ϫ . However, unlike the mutated Asp-389, the mutated Asn-389 has no interaction with His-392 because the carboxamide group of Asn-389 points away from the coordination position (Fig. S4b). Instead, His-392 is hydrogen-bonded to Tyr-367, with the imidazole ring rotated about 30°compared with that of WT (Fig.  S4b). Nevertheless, the N␦1 atom of His-392 and the O␥ atom of Ser-152 stay in positions as seen in WT and thus form a hydrogen bond with a length almost identical to that of WT. Therefore, His-392 in the G389N mutant is well-orientated in an alternative but still catalytically competent conformation. Enzymatic kinetics showed a decreased k cat for the G389N mutant (Table 1), indicating a weakened catalytic rate. This may reflect the loss of the charge-stabilization effect on the protonated His-392 in this mutant. Surprisingly, the G389N mutant maintained a relatively high activity, showing a stronger substrate binding affinity than that of WT (reflected by a decreased K m ), which may relate to the structural changes in the mutant, including alternative conformations of residues 211-215, 288 -299, 360 -367, and 389 -392 and the flipping of the side chain of Trp-173 (Fig. S6). Taken together, we conclude that Cl Ϫ can functionally subrogate Asp as the third component to form an unusual Ser-His-chloride catalytic triad, with pos- Table 1 Kinetics data of VvPlpA and its mutants Every data point was measured three times (n ϭ 3), and standard errors were calculated by the fitting software. Relative activity (%) is defined as catalytic efficiency (k cat /K m ) of the enzyme divided by that of WT. The kinetics of S152G, H392N, and N247D were not determined due to their very weak activity in the assay.

The substrate-binding pocket
The crystal structure of G389N contains three monomers in one asymmetric unit, and each monomer contains extra nonprotein electron density. Considering the shape of the molecule and the crystallization composition, we modeled this electron density with a hexaethylene glycol (P6G) molecule. P6G is located in an elongated pocket of the phospholipase domain of VvPlpA, with an extended "S"-like conformation (Figs. 5, a and  b, and S7). Based on the structural knowledge of SGNH hydrolases, this pocket is supposed to be the substrate-binding pocket, and P6G would mimic the aliphatic chain of the fatty acid at the sn-2 position of a phospholipid substrate. The bottom and wall of this pocket are mainly formed by hydrophobic residues, whereas its entrance part contains both hydrophobic and hydrophilic residues (Fig. 5b). The catalytic Ser-152 is at a position close to the pocket entrance. Remarkably, the residues forming the substrate-binding pocket are highly conserved among TLH proteins (Figs. 1, a and c, and 5b), indicating that TLHs may share significant similarity in the sn-2 acyl moiety of substrates. Estimated from the size of P6G, the aliphatic chain up to ϳ16 carbons can be accommodated in the pocket (Fig.  5b). In contrast, we also noticed that the P6G molecule is not close to the catalytic Ser-152, and the oxyanion hole is still occupied by the side chain of Asn-247 and is thus in an inactive state. Therefore, the binding mode of P6G could be different from that of the substrate. An enzyme-substrate complex structure is still required to understand the true substrate-binding mode.

Comparison with homologous proteins
A Dali search identified a couple of SGNH hydrolase superfamily proteins that share a similar fold with the C-terminal domain of VvPlpA, whereas no significant homologs of the N-terminal domain were found. The top two hits, the lipase EstJ15 from Photobacterium (PDB code 5XTU) and the esterase EstA from P. aeruginosa (PDB code 3KVN), were superimposed to VvPlpA with a root-mean-square deviation of 1.51 and 1.96 Å 2 for the aligned C␣ atoms, respectively. VvPlpA shares a conserved core SGNH hydrolase fold (corresponding to ␣4 -9, ␣11, ␤7, and ␤9 -11 in VvPlpA) with EstJ15 and EstA, whereas it exhibits obvious variations in the parts outside the core fold, including the long loops and additional secondary structural elements (Fig. 6a). Alignment of their catalytic sites showed that the catalytic residues His and Ser are well-superimposed. The catalytic Cl Ϫ of VvPlpA occupies the position that corresponds to the carboxyl moiety of Asp residues in the other two enzymes. The oxyanion hole-forming residues, Asn-247 and Gly-203, of VvPlpA sit in positions different from those of corresponding residues of EstA and EstJ15 in which the oxyanion holes are well-formed (Fig. 6b). Although VvPlpA G389N contains a P6G molecule in the substrate-binding pocket, EstA also contains a detergent molecule, tetraethylene mono-octylether, in its substrate-binding pocket. When G389N and EstA structures were superimposed, P6G and tetraethylene mono-octylether are located in similar positions but not well-overlapped (Fig. S8), supporting that the substrate-binding pockets of VvPlpA and EstA share similarities in their hydrophobic nature but show significant variations in terms of shape and the property to accommodate specific substrates. Apparently, VvPlpA is evolutionally close to other known SGNH esterases/lipases, but it also exhibits clear structural variations, such as containing a structurally unique N-terminal domain, which could be important for its specific function.

Discussion
We report the crystal structure of VvPlpA, a virulence factor of V. vulnificus. VvPlpA is composed of an N-terminal functionally unknown domain and a C-terminal phospholipase domain with an SGNH hydrolase fold, and these domains are closely packed together. VvPlpA contains an unusual Ser-Hischloride catalytic triad and a hydrophobic substrate-binding pocket that is highly conserved in TLH proteins. This is the first report regarding structural information of TLH proteins.

EDITORS' PICK: The Ser-His-chloride catalytic triad
The discovery of an unusual catalytic triad variant with chloride as the third component is a remarkable result of this study. A catalytic triad refers to a set of three amino acid residues that are accurately arranged in space to form a nucleophile-baseacid triad that facilitates covalent catalysis. The prototypic catalytic triad, Ser-His-Asp, was first demonstrated in the structures of chymotrypsin (21) and trypsin (22,23). The roles of each catalytic residue have been extensively studied. Ser is the nucleophile, which attacks substrate to initiate a covalent catalysis, whereas His acts as the base to deprotonate Ser for a higher nucleophilicity. The acid Asp orients the catalytic His, stabilizes the charge on its imidazole ring developed during catalysis, and likely deprotonates the catalytic His as well (24); mutating Asp-102 to Asn in trypsin decreased its activity by 10,000-fold (20). Such a triad is so powerful for covalent catalysis that it has been found in numerous hydrolases and transferases that exhibit diverse structures and functions as a result of convergent evolution. Nevertheless, variations to this catalytic triad have been reported, such as Ser-His-Glu (25)(26)(27), Ser-His-His (28 -31), Ser-Asp-Glu (32,33), and Ser-Ser-Lys (34,35). In some cases, the triad can also degrade to a dyad, such as Ser-His (36 -40) and Ser-Lys (41,42). These variants result from alternative combinations of amino acid residues, which finely tune enzymatic activity for their individual biological functions (43)(44)(45). In VvPlpA, a chloride anion structurally and functionally subrogates the Asp/Glu acid as the third component of the catalytic triad. The importance of the Cl Ϫ for VvPlpA was confirmed by the dependence of enzymatic activity on Cl Ϫ concentration. This Cl Ϫ could be substituted by a Br Ϫ or other monovalent anions. In the present study, we report the first catalytic triad variant that employs an inorganic anion as the third component. Like the Asp in the classic triad, the roles of the chloride ion in VvPlpA could include the following: 1) forming a hydrogen bond with the catalytic His for maintaining the correct orientation and tautomeric state of the imidazole ring and 2) electrostatically stabilizing the positive charge on the imidazole ring of the catalytic His developed during the catalytic process. However, due to the extremely low pK a of HCl (about Ϫ3), Cl Ϫ might not be able to deprotonate the catalytic His using the charge-relay mechanism that has been proposed for the classic Ser-His-Asp triad (24). Based on these analyses, we supposed that the catalytic power of this Ser-His-chloride would be lower than that of the classic triad but higher than that of a dyad in which only the orientating effect on the catalytic His would be retained. We compared activity of the WT with that of G389D and G389N, which employ a classic triad and a dyad, respectively. Remarkably, we observed that WT VvPlpA showed a catalytic power that was even higher than that of G389D. We may not conclude that Cl Ϫ is a better third component than Asp because the higher activity of the WT could be a result of local structural optimization for preferential use of Cl Ϫ in VvPlpA. Nevertheless, the comparison shows that Cl Ϫ is an effective substitute of Asp as the third component. When compared with G389N, the WT enzyme exhibited a higher catalytic rate. Because His-392 is well-orientated by Tyr-367 in G389N, the decreased catalytic power of this mutant suggests that the electrostatic stabilization of the protonated His-152 by Cl Ϫ during the catalytic process also makes a significant contribution to catalysis (46). We also noticed that the extent of decrease of k cat was not as dramatic as expected, suggesting that the Tyr-367, which is hydrogen-bonded to His-392, may act as a weak acid to stabilize the charged imidazole ring, albeit in a weaker manner compared with Cl Ϫ or Asp. In general, the comparison results shed light on the roles of Cl Ϫ . In the future, a theoretical study on this unusual triad would be desired to understand the roles of Cl Ϫ more precisely.
Cl Ϫ is the most abundant anion in life and plays important biological functions, such as cell homeostasis and neural signal transduction (47)(48)(49). As NaCl is the most used buffer component, Cl Ϫ is also widely found in protein structures as attached solvent molecules (Protein Data Bank). However, compared with numerous studies on cations, there have been very few reports on Cl Ϫ regarding its specific roles in protein function (50). For years, it has been well-known that photosystem II requires chloride for maximal O 2 -evolving activity (51), but the exact role of Cl Ϫ is not fully understood yet (50). Structural  3KVN) and EstJ15 (PDB code 5XTU). The well-superimposed SGNH hydrolase fold is colored gray with the corresponding secondary structures in VvPlpA indicated. The varied parts are colored orange for VvPlpA, teal for EstA, and pink for EstJ15. b, superimposition of the catalytic site residues for VvPlpA, EstA, and EstJ15. The residues are indicated in the order of VvPlpA, EstA, and EstJ15 using the same color code as in a.

EDITORS' PICK:
The Ser-His-chloride catalytic triad studies of photosystem II have shown that the chloride ion is located close to the oxygen-evolving complex (OEC), but it is not a ligand of the OEC (18,52,53). Cl Ϫ may facilitate the OEC to access and deprotonate the substrate water (54). It is also assumed that Cl Ϫ maintains the structure of the OEC as well as the proton channel (18). In angiotensin-converting enzyme, Cl Ϫ takes a structural role to maintain a proper local structure at the active site to stabilize the enzyme-substrate complex (55). In chloride-dependent amylases, Cl Ϫ contributes to catalysis by tuning pK a of the catalytic acid residue and preventing undesired salt-bridge formation in the active site (19,56). In the present study, VvPlpA provides another precious example of using the chloride ion as an enzyme cofactor and exhibits novel features that are different from previous examples. The three aforementioned examples all include Arg/Lys as a crucial residue for Cl Ϫ binding. Mutating Arg/Lys to Gln can remove Cl Ϫ dependence of the enzymatic activity of angiotensin-converting enzyme (57), amylase (19,58), and possibly photosystem II (50). The positively charged Arg/Lys residue is absent in VvPlpA; instead, the main-chain NH groups of surrounding residues and the side-chain NH groups of Trp and His coordinate Cl Ϫ . Interestingly, substituting Gly-389, the main-chain NH group of which coordinates Cl Ϫ , with Asp is also able to convert VvPlpA into a chloride-independent enzyme. Moreover, unlike the aforementioned examples where Cl Ϫ seems to function as a peripheral element to regulate catalytic activity by maintaining local structure and/or tuning pK a of residues at the active site, Cl Ϫ is a core component of the catalytic machinery of VvPlpA. As such, Cl Ϫ functionally subrogates the catalytic acid residue Asp in the classic catalytic triads. Taken together, our results expand the knowledge of the chemistry and function of Cl Ϫ as an enzyme cofactor.
Presenting the first structure of the TLH family, our results provide invaluable information to understand structures and functions of the TLH family. The high similarity in the TLH sequences indicates that all TLHs would share similar structural folds. A similarity in the active site would be expected as well, except for the difference of one single residue, which could be Asp, Glu, or Gly at the third position preceding the catalytic His. Based on the type of this specific residue, we assume that TLHs can be divided into two groups: chloride-dependent and chloride-independent. Homologs with a Gly at this position, such as VaPlpA, are likely chloride-dependent enzymes and use a Ser-His-chloride catalytic triad as seen in VvPlpA. In contrast, TLH proteins with Asp/Glu at this position may use a Ser-His-Asp/Glu triad, functioning in a chloride-independent manner. As shown, VvPlpA can be converted into a chlorideindependent enzyme by mutating Gly to Asp. It is likely these two groups of TLH proteins can be exchangeable in the context of chloride dependence by substituting Gly with Asp/Glu or vice versa. Our analysis also revealed that the substrate-binding pockets of TLHs are formed by highly conserved residues, indicating that the substrates of TLH proteins share similarity in the acyl chain at the sn-2 site of phospholipids. Nevertheless, we cannot yet conclude that the substrate spectrums of TLHs are similar as the head group and the acyl moiety at the sn-1 position of phospholipids may also play a role in substrate preference. Lastly, as VvPlpA and other TLH proteins are virulence factors, our structure provides invaluable information for antivirulence drug design. It is noteworthy that VvPlpA contains a large portion of coils that confer conformational flexibility. This structural flexibility might be a common feature of TLH proteins, accounting for their "thermolabile" feature as indicated by their name "thermolabile hemolysins." Indeed, there are other experimental indicators for this structural flexibility. It is likely that the structural variations observed in the G389N mutant reflected the conformational flexibility of VvPlpA. Furthermore, the oxyanion hole in VvPlpA is malformed, indicating that VvPlpA undergoes conformational changes upon substrate binding. The structural flexibility of TLHs might be of importance to their functions and should be carefully considered during design of inhibitors.

Gene cloning and protein production
The DNA sequence encoding VvPlpA (residues 1-417) was amplified by a standard PCR method using primers (forward, 5Ј-AGTCCATGGGCAAGAAGATAACTATTCTGTTGG-3Ј; reverse, 5Ј-ATGAATTCTTAGTGATGGTGATGGTGA-TGAAAATTAAAGCGTTGCATGCC-3Ј) with the genomic DNA of V. vulnificus ATCC 27562 as template. The PCR product was ligated into the pETM13 vector (European Molecular Biology Laboratory) using the NcoI and EcoRI restriction sites. The sequence-verified gene construct was transformed into the Escherichia coli strain C43 (DE3). Then the strain was cultivated in Luria-Bertani medium supplemented with 50 g/ml kanamycin until the OD 600 value reached about 0.8. Overexpression was induced by adding isopropyl ␤-D-thiogalactopyranoside to a final concentration of 60 M, with further incubation at 16°C for 14 h with vigorous shaking. Subsequently, the induced cells were harvested by centrifugation at 4000 ϫ g for 15 min, resuspended in the binding buffer (containing 50 mM Tris-HCl, 150 mM NaCl, 10 mM imidazole, pH 8.0), lysed with a high-pressure cell disruptor, and centrifuged at 16,000 ϫ g for 1 h at 4°C. The supernatant was applied onto a nickel-chelating Sepharose affinity chromatography column (GE Healthcare), and the purified protein was finally eluted with a buffer containing 50 mM Tris-HCl, 150 mM NaCl, 250 mM imidazole, pH 8.0. The elution was collected, concentrated, and finally purified with size-exclusion chromatography using a HiLoad 16/600 Superdex 75 column (GE Healthcare) equilibrated in a buffer of 10 mM HEPES, 150 mM NaCl, pH 7.5. The purified protein was concentrated to about 10 mg/ml, determined by the absorbance at 280 nm, and was then freshly frozen at Ϫ80°C for later use. A typical expression and purification profile of the recombinant protein was analyzed by SDS-PAGE (Fig. S9).
The Se-Met-derived VvPlpA (Se-Met-VvPlpA) was produced using the metabolic inhibition method as described previously (59). The purification was carried out as that for the WT, but the buffer for size-exclusion chromatography was supplemented with 3 mM DTT. The Br Ϫ -bound VvPlpA (VvPlpA-Br Ϫ ) was overexpressed in a medium in which the chloride ion was totally replaced by bromide ion. From the cell lysis onward, purification of VvPlpA-Br Ϫ was done with HEPES buffer instead of Tris-HCl buffer, and NaCl was completely replaced EDITORS' PICK: The Ser-His-chloride catalytic triad by NaBr. Mutations were performed by overlap-extension PCR with appropriate mutagenesis primers. The mutant proteins were prepared in a procedure similar to that of WT except for the S152G mutant. In purification of S152G, gel-filtration buffer containing 10 mM Tris-HCl, 150 mM NaCl, pH 7.5, was used.
The diffraction data were collected on beamlines BL17U1 and BL19U1 of the Shanghai Synchrotron Radiation Facility (SSRF) (60). The data were processed by autoPROC (Global Phasing Ltd.) (61,62). The structure of VvPlpA was solved by the single-wavelength anomalous dispersion method using Se-Met-derived crystals. Seven data sets of the Se-Met crystals were merged together to obtain sufficient anomalous signal. Then the substructure of selenium was successfully solved by SHELXD (63,64). Phasing and initial model building were done with SHELXE (65). The partial model was further extended by AutoBuild (66). The structure was refined with autoBUSTER (67) or phenix.refine (68). Alternately, model building on Coot (69) was performed. Structures of VvPlpA-Br Ϫ and mutants were solved by molecular replacement with the WT VvPlpA structure as the search model using Phaser (70). All of the crystallographic data are summarized in Table S1.
Sequence alignment was done on the ESPript webserver (71). Structural superimpositions were performed using Align (72). Domain interface was analyzed with the PISA program (73). The MolProbity server (74) and other programs in CCP4 (75) and Phenix (76) were also used for structural analyses. Structural graphics were prepared using PyMOL (Schrödinger).

Phospholipase A 2 activity assay
The phospholipase A 2 activities of VvPlpA and the mutants were measured following the manual of the EnzChek Phospholipase A2 Assay kit (Invitrogen). The determination of kinetic parameters was carried out in a 100-l reaction system containing 50 mM Tris-HCl, 100 mM NaCl, 1 mM CaCl 2 , pH 7.4, supplemented with 2.5-25 M (for Y367F, 2.5-50 M) fluoro-genic phospholipid substrates (Red/Green BODIPY PC-A2) and 423 nM enzyme. The fluorescence intensity (excitation wavelength, 480 nm; emission wavelength, 515 nm) was measured continuously in a microplate (Costar 96-well, black) by an M1000 Pro Microplate Detector (TECAN). The fluorescence inner filter effect was corrected based on previous literature (77). The initial velocity was calculated with Origin, and the kinetic parameters were calculated by the Michaelis-Menten equation (v ϭ V max [S]/(K m ϩ [S]) using a nonlinear fitting method.
The effects of metal ions on the biological function of VvPlpA were determined by supplementing 5 mM EDTA, 100 mM MgCl 2 , 100 mM CaCl 2 , or 100 mM KCl in the reaction buffer containing 50 mM Tris-HCl, 100 mM NaCl, 2 M substrate, 423 nM enzyme, pH 7.4. To test the Cl Ϫ dependence of enzymatic activity, we depleted Cl Ϫ in the protein samples by running the samples twice through the Superdex 75 (16/600 HiLoad) column equilibrated with the buffer 50 mM Na 2 HPO 4 /NaH 2 PO 4 , pH 7.4 (19). The enzymatic activity was measured in a 100-l reaction system containing 423 nM Cl Ϫ -depleted sample, 2 M substrate, and a buffer (50 mM Na 2 HPO 4 /NaH 2 PO 4 , pH 7.4) supplemented with varying concentrations of NaCl. The apparent K d value was estimated by fitting the chloride concentrationdependent activity to the formula v ϭ V max [S]/(K d(apparent) ϩ [S]) with Origin. The dependence of the VvPlpA on the Br Ϫ was determined by the same method as described above. The correct folding of the proteins, including Cl Ϫ -depleted VvPlpA, Y367F, S152G, H392N, and N247D, the structures of which were not measured, was confirmed by size-exclusion chromatography and circular dichroism (CD) spectra (Fig. S10).

Accession numbers
The coordinates and diffraction data of VvPlpA, Br Ϫ -bound VvPlpA, G389D, and G389N have been deposited in the Protein Data Bank with accession numbers 6JKZ, 6JL0, 6JL1, and 6JL2, respectively.