ANTIMICROBIAL AGENTS AND CHEMOTHERAPY, Oct. 2009, p. 4464–4471 Vol. 53, No. 100066-4804/09/$08.00�0 doi:10.1128/AAC.00288-09Copyright © 2009, American Society for Microbiology. All Rights Reserved.

The Structure of the Dizinc Subclass B2 Metallo-�-Lactamase CphAReveals that the Second Inhibitory Zinc Ion Binds in the

Histidine Site�

Carine Bebrone,1,3†* Heinrich Delbruck,2† Michael B. Kupper,2 Philipp Schlomer,2Charlotte Willmann,2 Jean-Marie Frere,1 Rainer Fischer,2

Moreno Galleni,3 and Kurt M. V. Hoffmann2

Centre for Protein Engineering, University of Liege, Allee du 6 Aout B6, Sart-Tilman 4000 Liege, Belgium1; Institute ofMolecular Biotechnology, RWTH-Aachen University, c/o Fraunhofer IME, Forckenbeckstrasse 6, 52074 Aachen, Germany2;

and Biological Macromolecules, University of Liege, Allee du 6 Aout B6, Sart-Tilman, 4000 Liege, Belgium3

Received 3 March 2009/Returned for modification 10 July 2009/Accepted 27 July 2009

Bacteria can defend themselves against �-lactam antibiotics through the expression of class B �-lactamases,which cleave the �-lactam amide bond and render the molecule harmless. There are three subclasses of classB �-lactamases (B1, B2, and B3), all of which require Zn2� for activity and can bind either one or two zinc ions.Whereas the B1 and B3 metallo-�-lactamases are most active as dizinc enzymes, subclass B2 enzymes, such asAeromonas hydrophila CphA, are inhibited by the binding of a second zinc ion. We crystallized A. hydrophilaCphA in order to determine the binding site of the inhibitory zinc ion. X-ray data from zinc-saturated crystalsallowed us to solve the crystal structures of the dizinc forms of the wild-type enzyme and N220G mutant. Thefirst zinc ion binds in the cysteine site, as previously determined for the monozinc form of the enzyme. Thesecond zinc ion occupies a slightly modified histidine site, where the conserved His118 and His196 residues actas metal ligands. This atypical coordination sphere probably explains the rather high dissociation constant forthe second zinc ion compared to those observed with enzymes of subclasses B1 and B3. Inhibition by the secondzinc ion results from immobilization of the catalytically important His118 and His196 residues, as well as thefolding of the Gly232-Asn233 loop into a position that covers the active site.

Class B �-lactamases (also called zinc �-lactamases or me-tallo-�-lactamases) play a key role in bacterial resistance to�-lactam antibiotics by efficiently catalyzing the hydrolysis ofthe �-lactam amide bond. The existence of such enzymes is aparticular concern because they are effective against most�-lactam antibiotics (including the carbapenems), the corre-sponding genes are easily transferred between bacteria, andthere are no clinically useful inhibitors. On the basis of theknown sequences, three different lineages, identified as sub-classes B1, B2, and B3, can be characterized (13, 14). All classB �-lactamases possess two potential zinc-binding sites andshare a small number of conserved motifs bearing some of theresidues that coordinate the zinc ion(s), notably His/Asn/Gln116-Xaa-His118-Xaa-Asp120 and Gly/Ala195-His196-Ser/Thr197 (14). Structural analysis of subclass B1 enzymes showsthat one zinc ion has a tetrahedral coordination sphere involv-ing His116, His118, His196, and a water molecule or OH� ion(histidine site), whereas the other has a trigonal-pyramidalcoordination sphere involving Asp120, Cys221, His263, andtwo water molecules (cysteine site) (Table 1). One water mol-ecule/hydroxide serves as a ligand for both metal ions (3). Inthe mononuclear structures of B1 enzymes (BcII, VIM-2,SPM-1, and VIM-4), the sole metal ion was found to be located

in the histidine site (5, 16, 26; P. Lassaux, unpublished data). Insome monozinc B1 structures, the Cys221 residue is foundunder an oxidized form. In subclass B3, the histidine site issimilar to that found in subclass B1, but Cys221 is replaced bya serine residue and the second zinc ion is coordinated byAsp120, His121, His263, and the nucleophilic water molecule,which again forms a bridge between the two metal ions (Table1) (33).

Aeromonas hydrophila CphA is a subclass B2 metallo-�-lactamase characterized by a uniquely narrow specificity. CphAefficiently hydrolyzes only carbapenems and shows very pooractivity against penicillins and cephalosporins, in contrast tosubclass B1 and B3 enzymes, which hydrolyze nearly all �-lac-tam compounds, with the exception of monobactams (11, 30).In contrast to subclass B1 and B3 enzymes, which are moreactive as dizinc species, subclass B2 is inhibited in a noncom-petitive manner by the binding of a second zinc ion. For CphA,the dissociation constant of the second zinc ion (Kd2) is 46 �Mat pH 6.5 (17). In agreement with extended X-ray absorptionfine-structure studies (18) and site-directed mutagenesis (34),the crystallographic structure of CphA shows that the first zincion is in the cysteine site (15). However, the second binding sitein subclass B2 enzymes remains to be determined becauseGarau and colleagues could not produce crystals of the dizincform, despite the presence of zinc concentration well aboveKd2 (15). The structure of Sfh-1, a subclass B2 enzyme fromSerratia fonticola, has been solved recently and once againinvolves only one zinc ion in the active site (12). In subclass B2,the His116 residue found throughout the metallo-�-lactamasesuperfamily (with the exception of the B3 GOB enzymes in

* Corresponding author. Mailing address: Centre for Protein Engi-neering, University of Liege, Allee du 6 Aout B6, Sart-Tilman, 4000Liege, Belgium. Phone: 3243663315. Fax: 3243663364. E-mail: carine.bebrone@ulg.ac.be.

† Both authors contributed equally to this work.� Published ahead of print on 3 August 2009.

4464

which Gln is present) is replaced by an asparagine (9, 13, 14).It has previously been shown that the Asn116 residue has norole in the binding of the zinc ions in CphA (34). On the basisof spectroscopic studies with the Aeromonas veronii cobalt-substituted ImiS enzyme, reports from the laboratory of Craw-ford and coworkers postulated that the second metal bindingsite was not the traditional histidine site but a site, remote fromthe active site, which involves both His118 and Met146 as zincligands (7, 8). However, a recent study of potential zinc ligandmutants contradicts this hypothesis and indicates that the po-sition of the second zinc ion in CphA is probably equivalent tothe histidine site observed with subclass B1 and B3 enzymes,with His118 and His 196 involved in the binding of this secondzinc, with perhaps Cys221 (or Asn116 or Asp120) as the thirdligand (4).

Since crystals of the wild-type CphA protein could not beobtained directly, single-site mutants were engineered by site-directed mutagenesis and overproduced. These mutants wereselected in order to introduce residues that are conserved ineither subclass B1 or B3, or both (2). Among these mutants,the N220G mutant (2) easily yielded crystals which were usedas seeds to grow wild-type CphA crystals (15). The kineticparameters of the wild-type and N220G mutant CphA enzymeswere similar, indicating strong conservation of enzymatic prop-erties in the mutant (2). However, the N220G mutation resultsin a slightly higher Kd2 value (86 �M) (2). We therefore set outto solve the crystallographic structures of the dizinc forms ofthe wild-type CphA enzyme and its N220G mutant. We con-firmed that the second zinc ion occupies a slightly modifiedhistidine site, in which the conserved His118 and His196 resi-dues act as metal ligands. The implication of this discovery forwhat is known about the coordination of zinc ions by metallo-�-lactamases is discussed. We propose an explanation for theinhibition by the second zinc ion. Moreover, the role of theAsn233 residue in this inhibition phenomenon is also appre-hended based on a site-directed mutagenesis study. In thiswork, a structural role for the zinc ions is also investigated.

MATERIALS AND METHODS

Protein purification and crystallization. Site-directed mutagenesis, proteinexpression, and purification of the proteins were carried out as described previ-ously (2), with the addition of a third purification step consisting of a size-exclusion chromatography (Hiload 16/60 Superdex 75 prep-grade column; Phar-macia Biotech). The purified enzyme solution was dialyzed against 15 mMsodium cacodylate (pH 6.5). Monodispersity of the protein solutions and thehydrodynamic radii of the dissolved protein molecules were checked by dynamiclight scattering. The crystallization conditions used previously (15) were not easyto reproduce because the crystallization drop comprised two phases. Therefore,a screen was carried out to find new conditions. Initial screens were carried outat 8°C using commercial Hampton crystal screens 1 and 2, grid screen ammo-

nium sulfate, and grid screen PEG 6000 (Hampton Research, CA). N220GCphA (10 mg/ml) was crystallized from 2.2 to 2.4 M ammonium sulfate and 0.1M morpholineethanesulfonic acid (pH 6.0 to 6.5), using the sitting-drop methodwith new crystallization plates designed by Taorad GmbH (Aachen, Germany).The reservoir solution (1 �l) was mixed with the protein solution (1 �l), and themixture was left to equilibrate against the solution reservoir. Typically, ortho-rhombic crystals of the N220G mutant grew within a few days to dimensions of80 by 100 by 100 �m and belonged to space group C2221 (a � 42.68 Å, b �101.06 Å, and c � 116.82 Å). Crystals of the wild-type CphA enzyme wereobtained under similar conditions using mutant microcrystals as starting seedsand showed similar orthorhombic symmetry in space group C2221 (a � 42.80 Å,b � 101.50 Å, and c � 116.49 Å). Before data collection, 1 �l of 100 mM ZnCl2was added to the drops containing the wild-type and the mutant crystals, and thecrystals were soaked for 1 day.

Data collection and processing. For data collection, wild-type and mutantcrystals were transferred to a cryoprotectant solution containing reservoir solu-tion supplemented with 30% (vol/vol) glycerol. The mounted crystals were flash-frozen in a liquid nitrogen stream. Near-complete X-ray data sets were collectedusing a Bruker FR591 rotating anode X-ray generator and a Mar345dtb detector.Diffraction data were processed with XDS (21) and scaled with SCALA from theCCP4 suite (6).

Structure determination and refinement. The structures of dizinc variantsfrom the wild-type and N220G mutant enzymes were solved using a molecularreplacement approach with Molrep (6) and the monozinc structure of CphA(Protein Data Bank [PDB] number 1X8G) as a starting model. The resultingmodel was rebuilt with ARP/wARP (29). The structures were refined in a cyclicprocess, including manual inspection of the electron density with Coot (10) andrefinement with Refmac (27). The refinement to convergence was carried outwith isotropic B values and using a translation/libration/screw parameter (28).The positions of the zinc, sulfate, and chloride ions were verified by calculatinganomalous maps, and the occupancies of the ions were determined by thedisappearing of the difference electron density. Alternative conformations weremodeled for a number of side chains, and occupancies were adjusted to yieldsimilar B values for the disordered conformations.

CD spectroscopy. Circular dichroism (CD) spectra for the apo-, mono-, anddizinc forms of the wild-type and N220G enzymes (0.3 mg/ml) were obtainedusing a JASCO J-810 spectropolarimeter. The apoenzymes were obtained in thepresence of 10 mM EDTA. The dizinc forms were obtained in the presence of500 �M Zn2�. The spectra were scanned at 25°C with 1-nm steps from 200 to 250nm (far UV). Thermal stability was assessed for the apo-, mono-, and dizincforms of the wild-type and N220G proteins using CD spectroscopy at 220 nmwith increasing temperature (25°C to 85°C). Urea denaturation studies of thewild-type and N220G enzymes were also carried out in the presence and absenceof 500 �M Zn2�. The mono- and dizinc enzymes were incubated for 18 h inincreasing concentrations of urea (0 to 8 M) at 4°C. Stability was assessed for themono- and dizinc forms of both proteins using CD spectroscopy at 220 nm, withincreasing concentrations of urea.

Thermal shift assay. Thermal shift assays were carried out using a LightCyclerreal-time PCR instrument (Roche Diagnostics). Briefly, 10 �l of wild-type CphAenzyme or the N220G mutant (0.3 mg/ml) was mixed with 10 �l of 5,000� SyproOrange (Molecular Probes) diluted 1:500 in 15 mM cacodylate (pH 6.5). Sampleswere heat denatured from 25 to 90°C at a rate of 0.5°C per minute. Proteinthermal unfolding curves were monitored by detecting changes in Sypro Orangefluorescence. The inflection point of the fluorescence-versus-temperature curveswas identified by plotting the first derivative over the temperature, and theminima were referred to as the melting temperature (Tm). Buffer fluorescencewas used as the control.

Differential scanning calorimetry. Thermal denaturation was investigated us-ing an NDSC 6100 differential scanning calorimeter (Calorimetry Sciences Corp.,Lindon, UT) after dialysis of the N220G enzyme sample (1 mg/ml) against 15mM cacodylate (pH 6.5). The dialysis buffer was used as a reference. The datawere collected and analyzed with an MCS Observer software package, assuminga two-state folding mechanism.

Analysis of the N233A mutant. The QuikChange site-directed mutagenesis kit(Stratagene Inc., La Jolla, CA) was used to introduce the N233A mutation intoCphA, using pET9a-CphA wild type as the template, forward primer (5�-GGAGAAGCTGGGCGCCCTGAGCTTTGCCG-3�), and reverse primer (5�-CGGCAAAGCTCAGGGCGCCCAGCTTCTCC-3�). The vector was then introducedinto Escherichia coli strain BL21(DE3) pLysS Star. Overexpression and purifi-cation of the mutant protein; CD spectroscopy; electrospray ionization-massspectrometry; and the determination of metal content, kinetic parameters, andresidual activity in the presence of increasing concentrations of zinc were carriedout as previously described (2, 4, 34).

TABLE 1. Composition of the two metal-binding sites in the threemetallo-�-lactamase subclasses

Enzyme Histidine site Cysteine site

B1 M�L His116-His118-His196 Asp120-Cys221-His263B2 M�L His118-His196 Asp120-Cys221-His263B3 M�L His116-His118-His196 Asp120-His121-His263B3 M�L GOB (Gln116)-His118-His196a Asp120-His121-His263

a This histidine site is still putative, since the three-dimensional structure ofthe GOB enzyme is not available.

VOL. 53, 2009 INHIBITORY ZINC ION BINDS IN HISTIDINE SITE OF CphA 4465

Protein structure accession numbers. Coordinates and structure factors weredeposited in the PDB using accession numbers 3F90 (dizinc form of wild-typeCphA) and 3FAI (dizinc form of the N220G mutant).

RESULTS

Structure determination. The crystal structures of the zinc-saturated wild-type and N220G CphA enzymes were solved.Stereochemical parameters were calculated by PROCHECK(22) and WHAT_CHECK (19) and fell within the range ex-pected for a structure with similar resolution. The crystallo-graphic and model statistics for the dizinc form of the wild-typeand N220G mutant structures are shown in Table 2.

Dizinc form of the wild-type CphA metallo-�-lactamase.The crystal structure of the dizinc form of the wild-type CphAenzyme was solved by molecular replacement using the struc-ture of the monozinc form (PDB number 1X8G) as the startingmodel. The structure was refined to a resolution of 2.03 Å. TheRwork and Rfree values for the refined structure were 0.1589 and0.2020, respectively. The crystals adopted a C2221 space group,with one molecule in the asymmetric unit. Only one residue

(Ala195) was found in a disallowed region of the Ramachan-dran plot. Ala195 (� � 105°; � 155°) is located on the loopbetween strands �8 and �9, with His196 at its apex.

The model of the wild-type dizinc structure includes 226amino acid residues (41 to 312, according to the BBL number-ing [14]). The last serine residue at the C terminus is missing.The electron density was good enough to identify three zincions (with two in the active site and one on the surface) and asulfate ion in the active site. The composition of the visiblesolvent sphere is mentioned in Table 2. The Gly232-Phe236mobile loop region had a low electron density, and the mainchain for the residues was modeled in two conformations. ThePhe236 side chain was not visible in the electron density map,and other side chains were modeled with alternative confor-mations.

CphA shows the typical �� fold of the metallo-�-lacta-mase superfamily. The overall fold is not altered by the pres-ence of the second zinc. The dizinc form has an average rootmean square deviation (RMSD) value of 0.272 Šfor backboneatoms with respect to the monozinc form.

The positions of the zinc ions were verified at the level of theelectron density and by using anomalous signals. Both zinc ionswere refined to a 100% occupancy and show an increasedmobility. This is indicated by the B factors (17.68 and 20.63 forthe first and second zinc ions, respectively) lying over the meanB factor (9.81). The first zinc ion is found in the Asp120-Cys221-His263 site or cysteine site as previously observed withthe monozinc form (15) (Fig. 1). The distances between thezinc ion and the Asp120 carboxyl oxygen atom, the Cys221sulfur atom, and the His263 side chain nitrogen atom are 2.00,2.26, and 2.09 Å, respectively. In the monozinc form, thesedistances were 1.96, 2.20, and 2.05 Å, respectively (15). Thetetrahedral coordination sphere is completed by a sulfate ionfrom the crystallization solution at a distance of 2.04 Å.

The second zinc ion is coordinated by the His118 and His196residues. Both histidine residues belong to the conserved his-tidine site in metallo-�-lactamases. The distances between thezinc ion and the His118 ND1 and the His196 NE2 are 2.01 and

FIG. 1. Structure of the wild-type CphA metal-binding sites. Zincions and water molecule are shown as gray and red spheres, respec-tively. ZnHis is the zinc ion that binds in the histidine site, while ZnCysis the zinc ion that binds in the cysteine site. The anomalous map at acontour level of 4 � is shown in orange. Relevant distances betweenzinc ions and ligands are indicated.

TABLE 2. X-ray data collection and structure refinement

Parameter

Value(s)

CphA WTdizinca

CphA N220Gdizinc

Data statisticsResolution range (Å) 19.7–2.03 19.6-1.70Wavelength (Å) 1.5418 1.5418Unit cell parameters

Space group C2221 C2221a (Å) 42.80 42.68b (Å) 101.05 101.06c (Å) 116.49 116.83

Rsym 0.062 0.032No. of observed reflections 235,065 206,898No. of unique reflections 16,754 28,046Completeness (%) 99.7 99.7I/�(I) 34.4 42.1Multiplicity 14 7.4B (Wilson)b 15.69 11.96

Refinement statisticsNo. of reflections used 16,754 28,046No. of molecules/asymmetric unit 1 1Rworking 0.1568 0.1388Rfree 0.2020 0.16024No. of residues in disallowed

region1 1

RMSD from idealBonds (Å) 0.011 0.011Angles (°) 1.217 1.550

Mean B factor 9.81 7.84No. of atoms (non-H) 2,090 2,198No. of atoms (protein) 1,850 1,882No. of H20 molecules 202 252

Other molecules/ions (no.)Zn2� 3 3SO4 4 7CO3 2Cl� 7 7Glycerol 3

a WT, wild type.b Value resulting from the Wilson plot.

4466 BEBRONE ET AL. ANTIMICROB. AGENTS CHEMOTHER.

2.03 Å, respectively. In CphA, the third ligand His116 is notconserved and is replaced by an asparagine residue, which isnot involved in the binding of the second zinc (the distance ismore than 4 Å). A sulfate ion (1.95 Å) and a water molecule(2.01Å) complete the vacant coordination positions of the zincion in the histidine site, forming a tetrahedral coordinationsphere. The sulfate ion acts as a bridging agent between thezinc ions (Fig. 1). The distances between the sulfate ion andzinc in the cysteine site and zinc in the histidine site are 2.04and 1.95 Å, respectively. The distance between both zinc ionsis 4.08 Å.

The third observed zinc ion was found on the surface, awayfrom the active site (about 17 Å). It is coordinated by His289(2.09 Å from His NE), and the coordination sphere is com-pleted by three chloride ions (2.19, 2.25, and 3.17 Å).

Dizinc form of the N220G mutant. The crystal structure ofthe dizinc form of N220G was solved by molecular replace-ment based on the monozinc structure (PDB number 1X8G).The structure was refined to 1.70 Å with Rwork and Rfree valuesof 0.1308 and 0.1602, respectively.

The model of the N220G dizinc structure included all the227 residues of the protein (41 to 313 according to the BBLnumbering [14]), three zinc ions (two in the active site and oneon the surface), and one sulfate ion located in the active site.The complete content of the solvent sphere is described inTable 2. The C-terminal residues were highly flexible. Ala195was found in a disallowed region of the Ramachandran plot.The electron density for the Gly232-Phe236 mobile loop re-gion allowed a well-defined model to be constructed.

The active site of the N220G mutant with two bonded zincions had a conformation nearly identical to that of the wild-type dizinc enzyme. The RMSD for all main chain atoms was0.216 Å. In contrast to the zinc ion that occupied two sites 1.5Å apart in the monozinc form of the N220G mutant (15), in thedizinc form of the enzyme described here, the zinc ion was inthe classical cysteine site, with distances to ligands identical tothose in the wild-type structure. Also, a sulfate ion and a watermolecule (both with 50% occupancy) were included in thecoordination sphere at distances of 2.04 and 2.27 Å, respec-tively. Under the new crystallization conditions describedabove, a monozinc form of the N220G mutant, in which a fullyoccupied zinc ion was present in the canonical cysteine site,was also obtained (data not shown).

The second zinc ion could not be modeled with an occu-pancy higher than 0.8. This second zinc ion shows a B factor of9.85, nearly the same as the mean B factor, 7.84, and in thesame range as the B factor of the zinc ion in the cysteine site,10.50. This zinc ion was coordinated by His118 and His196.The coordination sphere was completed by a water molecule(2.02 Å) and the sulfate ion (2.15 Å). An additional watermolecule was visible at a distance of 2.54 Å.

The third zinc ion was found in nearly the same position asthat in the wild-type structure. Here, the zinc ion and thecoordinating chloride ions exhibited 50% occupancy. The co-ordination sphere was affected by a water molecule and adisordered side chain from Glu292.

Comparison between the mono- and dizinc structures ofCphA. The binding of the second zinc ion in the active site hasno influence on the global structure of CphA. This is confirmedby the low RMSD values obtained by superposition of the

mono- and dizinc forms. The differences between the mono-and dizinc structures are limited to the Gly232-Asn233 loopregion (Fig. 2). This loop, located at the entrance to the activesite, is known to be stabilized upon the binding of the substrate(15) or inhibitors (20, 23). For the wild-type dizinc structure,this loop can exist in two conformations (both half occupied),corresponding, respectively, to the open form (as in themonozinc structures 1X8G and 1X8H) and the closed form (asin structures with substrate or inhibitor in the active site, 1X8I,2QDS, and 2GKL). In the dizinc form of the N220G mutant,this loop is found in the closed form (Fig. 2). The closedconformation is stabilized by additional H bonds between theAsn233 side chain and the Ser235 side-chain oxygen and back-bone nitrogen.

The active site of the dizinc forms can be superimposednearly perfectly over all the previously available CphA struc-tures (PDB numbers 1X8G, 1X8H, 1X8I, 2QDS, and 2GKL).For example, the RMSD is 0.56 Å for the superposition of thecysteine and histidine sites over their counterparts in 1X8G.The positions of the zinc in the cysteine sites are nearly iden-tical, with the exception of that of the monozinc N220G mu-tant, in which a disordered zinc ion was observed (15). Thecoordinating His118 residue in the histidine site is not af-fected by the binding of the second zinc. In contrast, theHis196 residue displays a slight flexibility. It is noticeablethat the orientation of the imidazole ring of His196 in allX-ray structures is selected to achieve an optimal pattern ofcontacts in the coordination sphere (H bonds and zinc bind-ing). Therefore, in dizinc structures, the His196 imidazolering is rotated 180° compared to its position in the monozincstructures (Fig. 2).

FIG. 2. Superposition of the mono- (gray) and dizinc (yellow)forms of the N220G CphA enzyme. The Gly232-Asn233 loop region ofthe dizinc form is represented in red to underline the conformationalchange. The positions of the His196 residue in the monoform(His196mono) and in the dizinc form (His196di) are represented. Thezinc ions in the histidine site (ZnHis) and in the cysteine site (ZnCys) areshown as gray spheres. The distance between His196di and the zinc ionlocated in the histidine site is indicated.

VOL. 53, 2009 INHIBITORY ZINC ION BINDS IN HISTIDINE SITE OF CphA 4467

Comparison with the active sites in B1 and B3 metallo-�-lactamases. The active sites of all three metallo-�-lactamasesubclasses show a high degree of correlation (RMSD valuesbetween 0.5 Å and 1.6 Å). In subclass B2, the coordinationsphere of the zinc in the histidine site is altered by the substi-tution of His116 by an asparagine residue. In subclasses B1 andB3, the His116 residue is involved in the binding of the zinc ionin the histidine site (Fig. 3). Removing one coordinating resi-due causes this zinc ion to shift by 1.12, 1.03, 1.03, and 0.92 Åcompared to the BcII (PDB number 1BVT) (Fig. 3), VIM-2(PDB number 1KO3), VIM-2 (PDB number 1KO2), and L1(PDB number 2FM6) structures, respectively. The position ofthe zinc in the cysteine site is nearly unchanged compared tothe available structures of subclass B1 enzymes.

Stability study. As already reported (17), the helical con-tents of the apo-, mono-, and dizinc forms of the wild-typeCphA enzyme are similar. The apparent Tm values determinedby CD spectroscopy were 46.9, 58.7, and 61.6°C for the apo-,mono-, and dizinc forms, respectively. Monozinc wild-typeCphA is also less stable than the dizinc form with urea as adenaturing reagent. Transitions between native and denaturedstates occurred at 3.4 and 4.4 M urea for the mono- and dizincforms, respectively (data not shown).

The far-UV CD spectra of the apo-, mono-, and dizinc formsof the N220G mutant exhibited small but significant differences(Fig. 4a). However, as observed for the wild-type enzyme, thehelical content did not vary significantly. Thermal denaturationwas irreversible in all cases. The apparent Tm values deter-mined by CD spectroscopy were 49.2, 61.7, and 62.9°C for theapo-, mono-, and dizinc forms, respectively (Fig. 4b). Also, theapparent Tm value determined for the monozinc form ofN220G using a thermal shift assay (Fig. 4c) and by differentialscanning calorimetry was 62°C but decreased to 50°C in thepresence of 10 mM EDTA (Fig. 4c). The monozinc form of theN220G mutant was slightly less stable than the dizinc form withurea as the denaturing reagent. Transitions between native and

denatured states occurred at 4.1 and 4.7 M urea for the mono-and dizinc forms, respectively (data not shown).

Is the closed position of the Gly232-Asn233 loop importantfor inhibition by the second zinc ion? Analysis of the N233Amutant. An Asn233 residue is present in all metallo-�-lacta-mases, with the exception of the L1, BlaB, and IND-1 enzymes(14). In CphA, the binding of a carbapenem substrate or in-hibitor modifies the Asn233 � angle so that its side chain closesthe entrance of the active site in a conformation stabilized bythe formation of an H bond between the oxygen of the Asn233side chain and the hydroxyl of Ser235 (15, 20, 23). In thestructure of the dizinc form of CphA, the loop adopts theclosed position, even in the absence of substrate, thus hinder-ing access to the active site (Fig. 2). To explore the role of theAsn233 residue in the inhibition phenomenon, the N233Amutant was expressed in Escherichia coli BL21(DE3) pLysSStar, purified to homogeneity, and characterized. The mass ofthe protein was verified by electrospray ionization mass spec-trometry. Within experimental error, the mutant was foundto exhibit the expected mass (25,144 Da measured versus25,146 Da expected). The far-UV CD spectrum of the mu-tant enzyme showed the same /� ratio as that of the wildtype. The enzyme was stored in 15 mM sodium cacodylate(pH 6.5) with �0.4 �M free zinc. Under these conditions,the N233A protein contained one zinc ion per molecule, asreported for the wild-type �-lactamase. The activity of themutant was measured in the absence of added Zn2� (�0.4�M) with imipenem, nitrocefin, and CENTA (Table 3). TheN233A mutant showed an increased Km value for imipenem.In contrast to the wild-type enzyme, the N233A mutant wasable to hydrolyze (albeit rather poorly) both cephalosporins,with quite low Km values. The hydrolysis of imipenem byN233A was inhibited by the binding of the second zinc ion,but the Kd2 value decreased to 11 2 �M compared to 46�M for the wild-type enzyme.

FIG. 3. The histidine site in CphA (a) and in BcII (PDB number 1BVT) (b). Zinc ions and water molecules are represented as gray and redspheres, respectively.

4468 BEBRONE ET AL. ANTIMICROB. AGENTS CHEMOTHER.

DISCUSSION

There are three subclasses of metallo-�-lactamases (B1, B2,and B3), all of which require Zn2� for activity and can bindeither one or two zinc ions (13, 14). Subclass B2 metallo-�-lactamases are active only in the monozinc form, because thebinding of a second zinc ion inhibits the enzyme in a noncom-petitive manner (17). This behavior contrasts with that of theB1 and B3 subclasses, in which the presence of a second zincion in the active site has either little effect or facilitates enzymeactivity. Since it has thus far proven difficult to determine thestructure of the dizinc form of a subclass B2 �-lactamase toinvestigate the basis of this inhibitory effect, we set out tocrystallize the dizinc form of Aeromonas hydrophila CphA andsolve its structure.

Previous attempts to determine the crystal structure of diz-inc A. hydrophila CphA have not been successful, even with ahigh concentration of zinc in the mother liquor (15). It wasproposed that this might reflect the conditions used to growcrystals and/or the presence of a carbonate ion in the activesite. Also, it was suggested that the binding of the second zincion required a change of active site conformation impossible toobtain in the crystal. Indeed, nuclear magnetic resonance mea-surements indicate major conformation changes upon bindingof the second zinc ion (C. Damblon, personal communication).However, CD spectra showed that the secondary structures ofthe monozinc forms of the wild-type (17) and N220G CphAenzymes are only slightly different from those of the dizincforms. By modifying the crystallization conditions, we suc-ceeded in obtaining CphA crystals with two zinc ions in theactive site. Moreover, the overall �� conformation of theenzyme was not affected by the presence of a second zinc ion.The dizinc form of CphA was obtained by soaking the crystalin a solution containing a zinc concentration well above the Kd2

value. In the crystal, major movements leading to the signifi-cant unfolding of the protein as observed by nuclear magneticresonance are impossible.

In the monozinc form, a carbonate ion was present in theactive site (15). In our structures, a sulfate ion from the crys-tallization solution was bound to both zinc ions.

As previously postulated (4, 34), the second inhibitory zincion is located in a slightly modified histidine binding site, withthe two conserved His118 and His196 residues as ligands. TheMet146 residue, proposed as one of the inhibitory zinc ionligands together with His118 in the ImiS metallo-�-lactamase(7), was 12.54 Å from this ion in CphA, and the sulfur atompoints in the opposite direction. Also, the space between the

FIG. 4. Structural studies of the N220G mutant. (a) Circular di-chroism spectra of the apo-enzyme (dotted line), monozinc (dashedline), and dizinc forms (solid line) of the N220G enzyme (0.3 mg/ml inboth cases). (b) Thermal denaturation of the N220G mutant (0.5mg/ml); fraction of native protein versus temperature (°C). Apo-N220G (dotted line), monozinc N220G (dashed line), dizinc N220G(solid line). (c) Thermal shift assay for N220G (0.3 mg/ml). N220Gplus 10 mM EDTA (dotted line) and monozinc N220G (dashed line).

TABLE 3. Kinetic parameters of the wild-type and N233ACphA enzymesa

�-Lactamantibiotic

WT N233A

kcat (s�1) Km

(�M)kcat/Km

(M�1 s�1) kcat (s�1) Km

(�M)kcat/Km

(M�1 s�1)

Imipenem 1,200 340 3,500,000 �700 �1,000 700,000Nitrocefin 0.008 1,300 6 0.03 55 520CENTA NH NH NH 0.0035 14 250

a All the experiments were performed at 30°C in 15 mM cacodylate (pH 6.5)with �0.4 �M Zn2�. Standard errors were below 10%. NH, no observed hydro-lysis; WT, wild type.

VOL. 53, 2009 INHIBITORY ZINC ION BINDS IN HISTIDINE SITE OF CphA 4469

His118 and Met146 residues is occupied by the backbone ofthe Asn114-His118 loop, connecting strand �6 and helix 2,and thus does not allow the accommodation of a zinc ion.The sequence of ImiS is 96% identical to that of CphA, soit would be remarkable if these few substitutions producedsuch a radically different mechanism for the binding of theinhibitory zinc ion.

This atypical coordination sphere with only two of the threeconserved histidine residues (Fig. 3) could probably explain therather high value of the dissociation constant for the histidinesite in CphA (46 �M) compared to that in subclasses B1 andB3. The value of the dissociation constant for the histidine siteis 1.8 nM for BcII (a subclass B1 enzyme), and subclass B3enzymes have high affinity constants for both binding sites(Kd � 6 nM) (35).

Vanhove et al. proposed an explanation for the inhibition ofsubclass B2 enzymes by the second zinc ion. The side chain ofthe Cys221, which is essential for binding the first zinc ion,would be displaced by the binding of the second zinc ion (34).However, a comparison between mono- and dizinc structuresshowed that this residue was not displaced. On the basis of theresults obtained here, we can confirm that the binding of theinhibitory zinc ion to the catalytically important His118 andHis196 residues prevents them from playing their catalyticroles in the hydrolysis of carbapenems (4, 15), in which His118is the general base which activates the hydrolytic water mole-cule and His196 contributes to the oxyanion hole by forming anH bond with the carbonyl oxygen of the �-lactam bond (15).This model is supported by the very weak activities of theH118A and H196A mutants (4). This conclusion reinforces thehypothesis that the nucleophile is not adequately metal acti-vated in B2 �-lactamases (31, 36). Superimposition of thestructures of the dizinc forms of BcII (B1) and CphA showsthat in the latter, the water molecule does not bridge the twozinc ions but is in interaction only with that in the histidine site(Fig. 5). It is possible that an interaction with this sole zinc iondoes not sufficiently decrease the water pKa, so the concentra-tion of OH� ions remains very low. Moreover, His118, whichnow serves as a zinc ligand, can no longer play the role of ageneral base. The closure of the Gly232-Asn233 loop in thedizinc form could also help to inhibit the enzyme, but Asn233probably plays only a minor role since the N233A mutant is stillinhibited by the binding of the second zinc ion. Finally, a thirdzinc ion is bound to a superficial histidine residue, but this isunlikely to be involved in the inhibition phenomenon. Its bind-ing is probably due to the high zinc concentration utilized.

Whereas the N233S mutant of ImiS exhibits kinetic param-eters similar to those of the wild-type enzyme (31), the N233Amutant of CphA shows an increased Km value for imipenemand thus seems to play a role in the binding of this substrate.Two-thirds of all sequenced metallo-�-lactamases have an Asnat position 233 (13, 14), and this residue was shown to beinvolved in substrate binding and activation by interacting elec-trostatically with the substrate �-lactam carbonyl (37). Thesame role for the Asn233 residue in substrate binding by CphAwas already predicted by computational modeling (36) and isin agreement with our results.

In contrast to the zinc ion located in the cysteine site, thesecond zinc ion does not play an important structural role.Indeed, the binding of the first zinc ion stabilizes the wild-type

and N220G CphA enzymes significantly, whereas the secondzinc ion has a much more limited impact on stability. Both themono- and dizinc forms of the N220G mutant are somewhatmore stable than their wild-type counterparts, perhaps explain-ing why crystals of the N220G mutant are obtained more easilythan those of the wild-type enzyme.

In conclusion, the catalytic metal ion is located in the cys-teine site of CphA, and the second inhibitory zinc ion binds toa slightly modified histidine site. The histidine site had firstbeen considered the catalytic site in the mononuclear metallo-�-lactamases. Indeed, in the first reported monozinc structures(B1 enzymes, such as BcII, VIM-2, VIM-4, and SPM-1), thesingle metal ion was shown to be located in the histidine site (5,16, 26). In contrast, a B3 enzyme called GOB was shown to beactive as a monozinc enzyme with the metal ion located in thecysteine site (25). Recently, reports from the laboratory of Vilaand coworkers proposed that B1 enzymes can be active in theirmono- and dizinc forms and that in monocobalt BcII, the metalion is localized in the cysteine site (24, 32). Two mononuclearmutants of BcII in which each of the metal binding sites wasselectively removed produced inactive variants. They con-cluded that the mononuclear form can be active only if assistedby residues at the other binding site (1). According to themodel proposed by Tioni et al. (32), the metal ion in thehistidine site would activate the hydrolytic water molecule inthe dizinc enzyme. Reports from the laboratory of Vila andcoworkers proposed that this is facilitated by a net of H-bondinteractions in the mononuclear enzyme (24, 32), includingprobably His118 or Asp120, as already proposed for CphA (4,

FIG. 5. Superimposition of the structures of the dizinc forms ofBcII (PDB number 2BFK) and CphA. Zinc ligands of BcII and CphAare represented as gray and green sticks, respectively. Zinc ions of BcIIand CphA are shown as gray and green spheres, respectively. ZnHis isthe zinc ion that binds in the histidine site, while ZnCys is the zinc ionthat binds in the cysteine site. Water molecules of BcII and CphA areshown as purple and red spheres, respectively. The sulfate ion thatbridges the two zinc ions in the CphA structure (SO4CphA) is alsorepresented, as well as glycerol, the fifth ligand of ZnCys in BcII(glycerolBcII).

4470 BEBRONE ET AL. ANTIMICROB. AGENTS CHEMOTHER.

15, 36). However, Asp120 appears unlikely because it alreadycoordinates a zinc ion. In subclass B2, as previously postulatedby Vanhove et al. (34), Asn116 is not involved in the binding ofthe second metal ion, but the N116H (34) and N116H-N220G(2) mutants could have a reconstituted His116-His118-His196site and behave similarly to B1 enzymes. Indeed, the N116H-N220G mutant has an extended substrate spectrum, and itsdizinc form is active (2).

The crystal structure of the subclass B2 CphA �-lactamase inits metal-inhibited form thus provides an answer to the struc-tural characterization of the inhibitory site that has been elu-sive and controversial so far.

ACKNOWLEDGMENTS

The work in Liege was supported by the Belgian Federal Govern-ment (PAI P5/33) and grants from the FNRS (Brussels, Belgium;FRFC grant no. 2.4511.06 and Lot. Nat. 9.4538.03). C.B. is an FRS/FNRS postdoctoral researcher.

REFERENCES

1. Abriata, L. A., L. J. Gonzalez, L. I. Llarrull, P. E. Tomatis, W. K. Myers,A. L. Costello, D. L Tierney, and A. J. Vila. 2008. Engineered mononuclearvariants in Bacillus cereus metallo-�-lactamase are inactive. Biochemistry47:8590–8599.

2. Bebrone, C., C. Anne, K. De Vriendt, B. Devresse, J. Van Beeumen, J. M.Frere, and M. Galleni. 2005. Dramatic broadening of the substrate profile ofthe Aeromonas hydrophila CphA metallo-�-lactamase by site-directed mu-tagenesis. J. Biol. Chem. 17:180–188.

3. Bebrone, C. 2007. Metallo-�-lactamases (classification, activity, genetic or-ganization, structure, zinc coordination) and their superfamily. Biochem.Pharmacol. 74:1686–1701.

4. Bebrone, C., C. Anne, F. Kerff, G. Garau, K. De Vriendt, R. Lantin, B.Devreese, J. Van Beeumen, O. Dideberg, J. M. Frere, and M. Galleni. 2008.Mutational analysis of the zinc and substrate binding sites in the CphAmetallo-�-lactamase from Aeromonas hydrophila. Biochem. J. 114:151–159.

5. Carfi, A., S. Pares, E. Duee, M. Galleni, C. Duez, J. M. Frere, and O.Dideberg. 1995. The 3-D structure of a zinc metallo-�-lactamase from Ba-cillus cereus reveals a new type of protein fold. EMBO J. 14:4914–4921.

6. Collaborative Computational Project, Number 4. 1994. CCP4 suite: pro-grams for protein crystallography. Acta Crystallogr. D 50:760–763.

7. Costello, A. L., N. P. Sharma, K. W. Yang, M. W. Crowder, and D. L.Tierney. 2006. X-ray absorption spectroscopy of the zinc-binding sites in theclass B2 metallo-�-lactamase ImiS from Aeromonas veronii bv. sobria. Bio-chemistry 45:13650–13658.

8. Crawford, P. A., K. W. Yang, N. Sharma, B. Bennett, and M. W. Crowder.2005. Spectroscopic studies on cobalt(II)-substituted metallo-�-lactamaseImiS from Aeromonas veronii bv. sobria. Biochemistry 44:5168–5176.

9. Daiyasu, H., K. Osaka, Y. Ishino, and H. Toh. 2001. Expansion of the zincmetallo-hydrolase family of the �-lactamase fold. FEBS Lett. 503:1–6.

10. Emsley, P., and K. Cowtan. 2004. Coot: model-building tools for moleculargraphics. Acta Crystallogr. D 60:2126–2132.

11. Felici, A., G. Amicosante, A. Oratore, R. Strom, P. Ledent, B. Joris, L.Fanuel, and J. M. Frere. 1993. An overview of the kinetic parameters of classB �-lactamases. Biochem. J. 291:151–155.

12. Fonseca, F., A. Correia, and J. Spencer. 2008. Structural and kinetic char-acterisation of Sfh-1, the B2 metallo-�-lactamase of Serratia fonticola, p. 46.In Proceedings of the 10th �-Lactamase Meeting, Eretria, Greece.

13. Galleni, M., J. Lamotte-Brasseur, G. M. Rossolini, J. Spencer, O. Dideberg,J.-M. Frere, and the Metallo-�-Lactamases Working Group. 2001. Standardnumbering scheme for class B �-lactamases. Antimicrob. Agents Chemother.45:660–663.

14. Garau, G., I. Garcia-Saez, C. Bebrone, C. Anne, P. S. Mercuri, M. Galleni,J. M. Frere, and O. Dideberg. 2004. Update of the standard numberingscheme for class B �-lactamases. Antimicrob. Agents Chemother. 48:2347–2349.

15. Garau, G., C. Bebrone, C. Anne, M. Galleni, J. M. Frere, and O. Dideberg.2005. A metallo-�-lactamase enzyme in action: crystal structures of themonozinc carbapenemase CphA and its complex with biapenem. J. Mol.Biol. 345:785–795.

16. Garcia-Saez, I., J. D. Docquier, G. M. Rossolini, and O. Dideberg. 2008. Thethree-dimensional structure of VIM-2, a Zn-�-lactamase from Pseudomonasaeruginosa in its reduced and oxidised form. J. Mol. Biol. 375:604–611.

17. Hernandez-Valladares, M., A. Felici., G. Weber, H. W. Adolph, M. Zeppe-zauer, G. M. Rossolini, G. Amicosante, J. M. Frere, and M. Galleni. 1997.Zn(II) dependence of the Aeromonas hydrophila AE036 metallo-�-lactamaseactivity and stability. Biochemistry 36:11534–11541.

18. Hernandez-Valladares, M., M. Kiefer, U. Heinz, R. Paul Soto, W. Meyer-Klaucke, H. Friederich Nolting, M. Zeppezauer, M. Galleni, J. M. Frere,G. M. Rossolini, G. Amicosante, and H. W. Adolph. 2000. Kinetic andspectroscopic characterization of native and metal-substituted �-lactamasefrom Aeromonas hydrophila AE036. FEBS Lett. 467:221–225.

19. Hooft, R. W. W., G. Vriend, C. Sander, and E. E. Abola. 1996. Errors inprotein structures. Nature 381:272.

20. Horsfall, L. E., G. Garau, B. M. Lienard, O. Dideberg, C. J. Schofield, J. M.Frere, and M. Galleni. 2007. Competitive inhibitors of the CphA metallo-�-lactamase from Aeromonas hydrophila. Antimicrob. Agents Chemother.51:2136–2142.

21. Kabsch, W. 1993. Automatic processing of rotation diffraction data fromcrystals of initially unknown symmetry and cell constants. J. Appl. Crystal-logr. 26:795–800.

22. Laskowski, R. A., M. W. MacArthur, D. S. Moss, and J. M. Thornton. 1993.PROCHECK: a program to check the stereochemical quality of proteinstructures. J. Appl. Crystallogr. 26:283–291.

23. Lienard, B. M. R., G. Garau, L. Horsfall, A. I. Karsisiotis, C. Damblon, P.Lassaux, C. Papamicael, G. C. K. Roberts, M. Galleni, O. Dideberg, J. M.Frere, and C. J. Schofield. 2008. Structural basis for the broad-spectruminhibition of metallo-�-lactamases by thiols. Org. Biomol. Chem. 6:2282–2294.

24. Llarull, L. I., M. F. Tioni, and A. J. Vila. 2008. Metal content and localizationduring turnover in Bacillus cereus metallo-�-lactamase. J. Am. Chem. Soc.130:15842–15851.

25. Moran-Barrio, J., J. M. Gonzalez, M. N. Lisa, A. L. Costello, M. Dal Peraro,P. Carloni, B. Bennett, D. L. Tierney, A. S. Limansky, A. M. Viale, and A. J.Vila. 2007. The metallo-�-lactamase GOB is a mono-Zn(II) enzyme with anovel active site. J. Biol. Chem. 282:18286–18293.

26. Murphy, T. A., L. E. Catto, S. E. Halford, A. T. Hadfield, W. Minor, T. R.Walsh, and J. Spencer. 2006. Crystal structure of Pseudomonas aeruginosaSPM-1 provides insights into variable zinc affinity of metallo-�-lactamases. J.Mol. Biol. 357:890–903.

27. Murshudov, G. N., A. A. Vagin, and E. J. Dodson. 1997. Refinement ofmacromolecular structures by the maximum-likelihood method. Acta Crys-tallogr. D 53:240–255.

28. Painter, J., and E. A. Merritt. 2006. Optimal description of a protein struc-ture in terms of multiple groups undergoing TLS motion. Acta Crystallogr.D 62:439–450.

29. Perrakis, A., R. Morris, and V. S. Lamzin. 1999. Automated protein modelbuilding combined with iterative structure refinement. Nat. Struct. Biol.6:458–463.

30. Segatore, B., O. Massida, G. Satta, D. Setacci, and G. Amicosante. 1993.High specificity of cphA-encoded metallo-�-lactamase from Aeromonas hy-drophila AE036 for carbapenems and its contribution to �-lactam resistance.Antimicrob. Agents Chemother. 37:1324–1328.

31. Sharma, N. P., C. Hajdin, S. Chandrasekar, B. Bennett, K. W. Yang, andM. W. Crowder. 2006. Mechanistic studies on the mononuclear Zn(II)-containing metallo-�-lactamase ImiS from Aeromonas sobria. Biochemistry45:10729–10738.

32. Tioni, M. F., L. I. Llarull, A. A. Poeylaut-Palena, M. A. Marti, M. Saggu,G. R. Periyannan, E. G. Mata, B. Bennett, D. H. Murgida, and A. J. Vila.2008. Trapping and characterization of a reaction intermediate in carbap-enem hydrolysis by Bacillus cereus metallo-�-lactamase. J. Am. Chem. Soc.130:15852–15863.

33. Ullah, J. H., T. R. Walsh, I. A. Taylor, D. C. Emery, C. S. Verma, S. J.Gamblin, and J. Spencer. 1998. The crystal structure of the L1 metallo-�-lactamase from Stenotrophomonas maltophilia at 1.7 Å resolution. J. Mol.Biol. 284:125–136.

34. Vanhove, M., M. Zakhem, B. Devreese, N. Franceschini, C. Anne, C.Bebrone, G. Amicosante, G. M. Rossolini, J. Van Beeumen., J. M. Frere, andM. Galleni. 2003. Role of Cys221 and Asn116 in the zinc-binding sites of theAeromonas hydrophila metallo-�-lactamase. Cell. Mol. Life Sci. 60:2501–2509.

35. Wommer, S., S. Rival, U. Heinz, M. Galleni, J. M. Frere, N. Franceschini, G.Amicosante, B. Rasmussen, R. Bauer, and H. W. Adolph. 2002. Substrate-activated zinc binding of metallo-�-lactamases: physiological importance ofmononuclear enzymes. J. Biol. Chem. 277:24142–24147.

36. Xu, D., D. Xie, and H. Guo. 2006. Catalytic mechanism of class B2 metallo-�-lactamase. J. Biol. Chem. 281:8740–8747.

37. Yanchak, M. P., R. A. Taylor, and M. W. Crowder. 2000. Mutational analysisof metallo-�-lactamase CcrA from Bacteroides fragilis. Biochemistry 39:11330–11339.

VOL. 53, 2009 INHIBITORY ZINC ION BINDS IN HISTIDINE SITE OF CphA 4471