The 1.5-Å Structure of Chryseobacterium meningosepticum Zinc β-Lactamase in Complex with the Inhibitor, D-Captopril*

The crystal structure of the class-B β-lactamase, BlaB, from the pathogenic bacterium, Chryseobacterium meningosepticum, in complex with the inhibitor, d-captopril, has been solved at 1.5-Å resolution. The enzyme has the typical αβ/βα metallo-β-lactamase fold and the characteristic two metal binding sites of members of the subclass B1, in which two Zn2+ ions were identified. d-Captopril, a diastereoisomer of the commercial drug, captopril, acts as an inhibitor by displacing the catalytic hydroxyl ion required for antibiotic hydrolysis and intercalating its sulfhydryl group between the two Zn2+ ions. Interestingly, d-captopril is located on one side of the active site cleft. The x-ray structure of the complex of the closely related enzyme, IMP-1, with a mercaptocarboxylate inhibitor, which also contains a sulfhydryl group bound to the two Zn2+ ions, shows the ligand to be located on the opposite side of the active site cleft. A molecule generated by fusion of these two inhibitors would cover the entire cleft, suggesting an interesting approach to the design of highly specific inhibitors.

The crystal structure of the class-B ␤-lactamase, BlaB, from the pathogenic bacterium, Chryseobacterium meningosepticum, in complex with the inhibitor, D-captopril, has been solved at 1.5-Å resolution. The enzyme has the typical ␣␤/␤␣ metallo-␤-lactamase fold and the characteristic two metal binding sites of members of the subclass B1, in which two Zn 2؉ ions were identified. D-Captopril, a diastereoisomer of the commercial drug, captopril, acts as an inhibitor by displacing the catalytic hydroxyl ion required for antibiotic hydrolysis and intercalating its sulfhydryl group between the two Zn 2؉ ions. Interestingly, D-captopril is located on one side of the active site cleft. The x-ray structure of the complex of the closely related enzyme, IMP-1, with a mercaptocarboxylate inhibitor, which also contains a sulfhydryl group bound to the two Zn 2؉ ions, shows the ligand to be located on the opposite side of the active site cleft. A molecule generated by fusion of these two inhibitors would cover the entire cleft, suggesting an interesting approach to the design of highly specific inhibitors.
Antibiotic resistance of pathogenic bacteria is a major clinical concern because it has increased significantly during recent years, rendering antimicrobial therapy progressively less effective. One mechanism by which bacteria can escape the action of antibiotics is the production of metallo-␤-lactamases (MBLs) 1 or class B ␤-lactamases (1). These enzymes were initially described as a curiosity in innocuous strains of Bacillus cereus (2) but were subsequently identified in several pathogenic species responsible for human infections in which the enzymes are either encoded by resident chromo-somal genes or by horizontally acquired genes carried on mobile genetic elements (3)(4)(5)(6)(7). Class B ␤-lactamases have a broad spectrum of substrates, including penicillin, cephalosporin, and carbapenem, some of which are not substrates for active site serine ␤-lactamases (3,4). At present no inhibitors of these enzymes are used clinically, making them a potential source of antibiotic resistance.
The MBL family has been divided into three different subclasses, B1, B2, and B3, on the basis of sequence similarities (8). MBLs have two metal binding sites (the His site, or first site, and the Cys site, or second site), which bind Zn 2ϩ ions required for enzyme activity. In general, both the mono-Zn 2ϩ (Zn 2ϩ in the His site) and di-Zn 2ϩ forms are active but have different kinetic properties (9, 10); the exception is CphA from Aeromonas hydrophilia, which is inhibited by the presence of the second zinc ion (11). In the mono-Zn 2ϩ enzyme, a watermediated mechanism for the hydrolysis of the antibiotic ␤-lactam ring has been suggested in which a water molecule is activated by the presence of the Zn 2ϩ ion in the His site, then carries out nucleophilic attack against the carbonyl carbon of the ␤-lactam. Residues Asp-120 and Cys-221 from the second site would help keep the active water molecule in position. Asp-120 would also act as a general base, transferring the proton from the water molecule to the nitrogen atom of the ␤-lactam ring, causing bond cleavage and opening of the ring, leading to inactivation of the antibiotic (12,13). In the bi-Zn 2ϩ enzyme, a hydroxide ion is positioned close to the two Zn 2ϩ ions. The second Zn 2ϩ would also help to anchor Asp-120 (14) and would orient the ␤-lactam ring by binding to the ␤-lactam nitrogen of the antibiotic (15). Although both the mono-and di-Zn 2ϩ forms exist, it has been hypothesized that only the mono-Zn 2ϩ form is present under normal physiological conditions (16).
Because of the great interest in MBL inhibition, several classes of inhibitors have been described (17,18); these include phenazines (19), ketone derivatives of L-and D-alanine and trifluoromethyl alcohol (20), thioesters (21,22), biphenyl tetrazoles (23,24), amino acid-derived hydroxamates (25), thiols (26 -30), and tricyclic natural products (31). However, despite many of these having good inhibitory properties, only a few thiols have a broad spectrum of inhibition for MBLs. In this context, the analysis of the three-dimensional structures of MBLs in complex with potential inhibitors becomes a powerful tool for rational drug design. At present, the structures of Bacteroides fragilis CcrA complexed with Mes, biphenyl tetrazole, or a tricyclic natural product (32) and of (23, 31) Pseudomonas aeruginosa IMP-1 complexed with mercaptocarboxylate (30) or with succinic acid derivatives (33) have been reported.
BlaB is a chromosome-encoded MBL produced by Chryseobacterium meningosepticum. It is a member of subclass B1 has a broad substrate profile (34 -36) and is produced constitutively (37). C. meningosepticum is an ubiquitous Gramnegative rod bacterium of clinical relevance because it can cause neonatal meningitis, adult septicemia, and nosocomial infections and is resistant to most ␤-lactams, including carbapenems (38,39).
, used clinically to control high blood pressure. It acts as an angiotensin-converting enzyme inhibitor and increases plasma level of the vasodilator, bradykinin (40). D-Captopril, which was synthesized to test its inhibitory effect on MBLs, is a good competitive inhibitor of BlaB (K i ϭ 70 -100 M) at neutral pH. 2 In this work, we describe the first high resolution (1.5 Å) three-dimensional structure of BlaB complexed with D-captopril.

EXPERIMENTAL PROCEDURES
Synthesis of D-Captopril-D-Captopril was synthesized from commercially available compounds, starting with D-proline (41,42). The final hydrolysis reaction was carried out using 1 N NaOH under an argon atmosphere.
Expression, Purification, and Crystallization of the D-Captopril Complex-The overexpression, purification, and characterization of the protein have been described (34). For crystallization, the hanging drop method was used. Crystals were obtained by mixing 2 l of protein (5 mg/ml in 10 mM sodium cacodylate, 100 M zinc acetate, 100 M dithiothreitol, pH 6.5) and 1 l of the reservoir solution (28% polyethylene glycol 4000, 0.2 M sodium acetate, and 0.1 M Tris-HCl, pH 8.4) at 8°C; crystals appeared after approximately 3 weeks. A crystal was soaked for ϳ5 weeks in previously equilibrated drops of the reservoir solution containing 2 mM D-captopril, then the crystal was flash-cooled in liquid nitrogen using precipitant containing 15% glycerol as cryoprotectant and stored in liquid nitrogen for subsequent data collection at a synchrotron source.
Data Collection and Processing-Diffraction data sets were collected at the ID14 -4 beamline of the European Synchrotron Radiation Facility in Grenoble, France. A data set of 360 frames at ϭ 0.9792 Å was collected at 1.5-Å resolution with 0.5°oscillation steps. Data integration performed using DENZO/SCALEPACK (43) gave an initial ambiguity in the space group determination, since according to the automated indexing procedure, the crystal could be cubic or trigonal, with similar distortion indexes. The data were re-indexed using DENZO/SCALE-PACK and scaled using SCALA (44), trying different space groups to solve this ambiguity. After this stage, the crystal was considered to be cubic I23 or I2 1 Table I).
Structural Determination and Refinement-The structure of the MBL, BcII, from B. cereus at 1.8-Å resolution (45) with no ions or water molecules was used as a model for molecular replacement using the program AMoRe (46). Because the x-ray data could be indexed in three space groups (I2 1 3, I23, and R3), various tests were performed. In brief, data indexed in R3 were used successfully in AMoRe with a BcII model performing a four molecule/asymmetric unit search (c-factor ϭ 36.3%, R-factor ϭ 48.0%). With these solutions and using a noncrystallographic symmetry-averaging map calculated using the CNS program (47), a partial model of BlaB was built using TURBO (48). A full model of BlaB at 1.5 Å was made using ARP/WARP (49). From the initial (2F o Ϫ F c ) and (F o Ϫ F c ) maps, the inhibitor, D-captopril, was localized close to the Zn 2ϩ ions in the active site. Further cycles of energy minimization, B-factor refinement, and water picking were performed using CNS programs (47). Models for D-captopril were not included until their conformations were well defined by the unbiased ͉F o Ϫ F c ͉, calc electron density maps. D-Captopril molecules in each of the four monomers were modeled. The model was refined without any non-crystallographic-symmetry restrains, yielding R working ϭ 18.95% and R free ϭ 20.76%.

RESULTS
The BlaB monomer shares a structural feature with other members of the MBL family, an ␣␤/␤␣ fold, in which a compact core of two ␤-sheets is surrounded by ␣-helices (Fig. 1). The crystal, which shows a cubic pseudo-symmetry, belongs to the space group R3 and has four monomers per asymmetric unit (A, B, C, and D).
The four monomers are practically identical, with a few local differences due to slightly different environments or to effects of crystal packing (e.g. monomers A, B, and D are missing the last four C-terminal residues, whereas monomer C is missing the last six). The root mean square deviation between the conserved C␣ atoms of the four molecules is about 0.1 Å. The active site of BlaB shares a feature with other subclass B1 structures, namely that two Zn 2ϩ ions (Zn1 and Zn2) occupy the two metal binding sites, the His site (His-116, His-118, His-196) and the Cys site (Asp-120, Cys-221, His-263) (Fig. 2, a  and c). A clear electron density corresponding to a molecule of D-captopril was found close to the active site. The inhibitor intercalates its sulfhydryl group between the two Zn 2ϩ ions (distance S-Zn1, 2.32 Å; S-Zn2, 2.30 Å in monomer A) (Fig. 2). The sulfur atom is also close to the NE2 of His-196 (3.34 Å) ( Table II). The inhibitor is located in the positively charged groove leading to the active site (Fig. 3a) and is present in the four molecules/asymmetric unit with full occupancy.
Other interactions enhance the binding of the inhibitor molecule. The carboxylic group of D-captopril is stabilized by hydrogen bonds between its O 2 and the NZ of Lys-167 and two water molecules; the other carboxylic oxygen, O3, is bound to the amide main chain of Asp-119. The O1 carbonyl oxygen of 2 H. W. Adolph, personal communication. the inhibitor also makes a hydrogen bond with another water molecule (Fig. 2b). In addition, the inhibitor is stabilized by several hydrophobic contacts between its carbons and His-118, Asp-120, and Tyr-233 ( Fig. 2b and Table II).

DISCUSSION
Structural Comparisons-The sequence of BlaB shows 25-35% identity with those of other subclass B1 ␤-lactamases. The secondary structure topology of the four available x-ray structures (BcII, CcrA, BlaB, and IMP-1) is very similar, the only difference being that the first short ␤-strand seen in BcII and CcrA is absent in BlaB and IMP-1. Pairwise superposition of the C ␣ atoms of BlaB and BcII (1BMC), Ccra (1ZNB), or IMP-1 (1JJE) gives a root mean difference density of 1.5, 1.4, and 1.9 Å, respectively.
In subclass B1, there is a loop between ␤2 and ␤3 (residues 61-65), which is disordered in the structure of native BcII (12). However, in the presence of an inhibitor, the loop is stabilized (30), transforming the active site groove into a tunnel-shaped cavity. In the BlaB structure, the loop is also well defined, probably due to the presence of D-captopril in the active site (Fig. 2c). Interestingly, D-captopril has no direct interaction with any residues in the loop (Fig. 2, b and c; Table II). The flexibility in the loop may help to accommodate the inhibitor, which itself may help to stabilize the loop. This stabilization takes place indirectly by Trp-87, which makes a double hydrogen bond between its NE1 and the OD1 of Asp-119 and the water molecule that hydrogen bonds to the O1 carbonyl oxygen of D-captopril (Table II). Trp-87 also makes hydrophobic contacts with the active site residue, His-263. All these interactions determine the position of Trp-87, which establishes hydrophobic interactions between its ring and two residues, Phe-61 and Asn-59. The loop is strongly stabilized by hydrogen bonds between the carbonyl oxygens and the amide nitrogens of the backbone of this region (amino acids 57-69). In the apical part of the loop facing the solvent, a water molecule forms a bridge between the ND2 of Asn-62 and the carbonyl oxygen of Phe-61, which also makes a hydrogen bond with the amide backbone of Thr-64. In addition, the amide backbone of Phe-61 makes hydrogen bonds with the carbonyl oxygen of the main chain of Thr-64. A second water molecule makes hydrogen bonds with this bridging water molecule and the hydroxyl group of Tyr-67. In BlaB, the loop appears to be closer to the were not included in the phase calculation. The residues labeled in black belong to the His site, and those in red belong to the Cys site. Zinc ions are represented as brown spheres. The figure was produced using BOBSCRIPT (54). b, protein-ligand interactions between BlaB and D-captopril depicted in monomer A using LIGPLOT (55). In the schematic drawing, strong interactions are shown as dashed green lines. Ligand and protein hydrophobic contacts are represented as curved red combs. c, active site cleft of BlaB. Zinc ions and water molecules are represented in brown and red spheres, respectively. D-Captopril is displayed in sticks (carbon, nitrogen, and sulfur atoms colored in violet, blue, and green, respectively). Amino acid residues cited in the text under "Structural Comparisons" are labeled, and Zn 2ϩ ligands are colored in gray.
rest of the protein than in the other three structures, with the hydrophobic residues, Phe-61, Tyr-67, and Trp-87, pointing toward the inside of the protein and creating a hydrophobic environment even though the loop does not interact directly with the inhibitor. In addition, the loops in BlaB and BcII are in different positions, e.g. the distance between equivalent residues in BlaB and BcII is 5.44 Å for Phe-61 and 5.77 Å for Asn-62. Interestingly, in subclass B1 MBLs, bulky hydrophobic residues are often found at positions 61, 67, and 87. In all MBL structures, the Asp or Asn residue at residue 84 has an angle outside the allowed range ( ϳ 60 Ϯ 1°) and plays an important structural role, contributing to the architecture of the active site. Of the residues that are highly conserved and/or potentially involved in substrate binding in subclass B1 ␤-lactamases, three positions are particularly interesting in the BlaB structure, 224, 233, and 236 (Fig. 2c).
Lys-224 is very well defined in the electron density map (NZ, B factor of 13.5 Å 2 ). The NZ makes three hydrogen bonds with the carbonyl oxygen of the preceding amino acid (Ile-223) and two water molecules. Lys-224 is assumed to interact with the carboxyl group of the ␤-lactam antibiotic (30). However, Lys-224 is not found in subclass B3 MBLs, and no equivalent residue is observed in a similar spatial position in the L1 (15) or FEZ-1 structures (50).
Based on x-ray crystallographic structures, Asn-233 and Zn1 were thought to form an oxyanion hole to polarize the carbonyl group of the ␤-lactam antibiotic (50,51). In BlaB, Asn-233 is replaced by Tyr, and the hydroxyl of the tyrosine is 7.88 Å away from Zn1 (Fig. 2c). Interestingly, Glu-236, which is an aspartic acid residue in the BcII, CcrA, and IMP-1 enzymes, makes three hydrogen-bonds with OH Tyr-233, NEZ His-118 (a zinc ligand), and a water molecule, whereas Asp-236 in the other three x-ray structures only hydrogen-bonds to NEZ His-118 and a water molecule. Consequently, no oxyanion hole is observed in the present structure. However, the observed position of Tyr-233 could be the result of complex formation, since C 3 , C 5 , and C 6 of D-captopril make hydrophobic contacts with the tyrosine side chain. Nevertheless, rotation around 1 is sterically forbidden, and only a large conformational change of the 227-237 loop (Fig. 2c) upon substrate binding could bring OH Tyr-233 or other side chain residues into positions where they could form part of the oxyanion hole.
Binding of the Inhibitor-The main interaction between BlaB and D-captopril occurs through intercalation of the sulfur atom of the inhibitor between the two Zn 2ϩ ions, leading to the removal of the activated water (OH Ϫ group) (Fig. 2, a and b; Table II). A similar situation has been seen in the three-dimensional structure of IMP-1 in complex with a mercaptocarboxylate inhibitor (2-mercaptomethyl-4-phenyl-butyrylimino)-(5tetrazol-1-ylmethyl-thiophen-2-yl)-acetic acid) (30).
Structural superposition of the structures of the two proteininhibitor complexes shows that the inhibitors intercalate their sulfur groups almost identically between the two Zn 2ϩ ions of the active site. A striking observation is that, although they are located in different parts of the active site, together the two inhibitors cover the entire active site cavity (Fig. 3a). The fusion of these two inhibitor molecules could be a good starting point for generating a new family of more effective inhibitors (Fig. 3b). Based on the structural superposition using this new molecule, many interactions would be retained, including those between the S atom and the two Zn 2ϩ ions, the carboxylate groups and the main-chain amide nitrogens (N119 and N233), the carboxylate groups and the NZ of two lysines (Lys-167 and Lys-224), and many hydrophobic contacts.
In the crystal structure of IMP-1 complexed with a succinic acid derivative (33), one oxygen of the carboxylic group of the inhibitor is located between the two Zn 2ϩ ions, displacing the hydroxyl. Curiously, in the three-dimensional structure of the subclass B3 MBL FEZ-1 complexed with D-captopril (50), the inhibitor is not located in the active site, because a sulfate ion from the crystallization solution intercalates one of its oxygen atoms between the two Zn 2ϩ ions. In the complex of B. fragilis CcrA with a biphenyl tetrazole (23), the tetrazole ring of the inhibitor acts as an extra ligand for the second Zn 2ϩ , changing its coordination and provoking a distortion of the active site and the removal of the water molecules from the active site.
Taking together our results and all currently known threedimensional structures of MBLs complexed with inhibitors (23, 30, 32, 33), we can conclude that there is a high variability in the modes of binding of the inhibitors depending on the ␤-lactamase sequence and the size and chemical properties of the inhibitor. However, one feature is constant, which is the intercalation of a charged group between the Zn 2ϩ ions and the removal of the bridging hydroxide directly involved in watermediated hydrolysis of the substrate. Moreover, the new BlaB structure presented here highlights the structural diversity in the MBL family. The evolution of broad substrate profile enzymes, such as BlaB, poses a serious threat to antibacterial therapy, and a detailed analysis of inhibitor-enzyme complexes is vital for the design of new inhibitors.