Structural Determinants of Substrate Binding to Bacillus cereus Metallo-β-lactamase*

Binding and hydrolysis of the β-lactams cefotaxime, cephapirin, imipenem, and benzylpenicillin by the metallo-β-lactamase from Bacillus cereus were studied by presteady state kinetic measurements. In all cases, the substrate was unmodified in the most populated reaction intermediate, and no chemically modified substrate species accumulated to a detectable amount. The cephalosporins tested showed similar formation rate constants for this intermediate, and they differed mostly in their decay rates. Formation of a non-productive enzyme ·substrate complex was detected for imipenem. The substrate binding differences can be accounted for by considering the structural features of each substrate. The apoenzyme could not bind any of the substrates, but binding was restored when the apoenzyme was reconstituted with Zn(II), revealing that the metal ions are the main determinants of substrate binding. This evidence is in line with the lack of an optimized substrate recognition patch in B1 and B3 metallo-β-lactamases that provides a broad substrate spectrum.

The most prevalent mechanism of bacterial resistance to ␤-lactam antibiotics is the production of ␤-lactamases (1, 2), which inactivate these drugs by hydrolyzing the active ␤-lactam bond. Metallo-␤-lactamases (class B ␤-lactamases) constitute a distinct class within this family of enzymes (3)(4)(5). In most cases, these metalloenzymes exhibit a broad substrate spectrum, being able to hydrolyze most of the ␤-lactam antibiotics currently available ( Fig. 1) (6 -8). The spread of metallo-␤-lactamase genes among pathogenic bacterial strains is raising an increasing concern in the biomedical community because no clinically useful inhibitors have yet been developed.
Metallo-␤-lactamases are classified in three groups according to sequence homology: subclasses B1, B2, and B3 (5). The metal content of the enzymes is variable. Most of them are active as bi-Zn(II) enzymes, while the lactamases from subclass B2 are active as mono-Zn(II) enzymes (9,10). In particular, Bacillus cereus metallo-␤-lactamase (BcII) 1 differs from other homologous subclass B1 lactamases because it exhibits different binding constants for the 2 Zn(II) equivalents, and it is active in both mono-and bi-Zn(II) forms (11)(12)(13). The mono-Zn(II) forms of the CcrA enzymes from Bacteroides fragilis and IMP-1 have also been characterized as active species (10,14,15) even if results from different groups are contrasting (16). Crystal structures have been solved for subclass B1 and B3 enzymes, while no structure of subclass B2 lactamases is yet available (17)(18)(19)(20)(21). All class B lactamases share a common fold, and the residues that constitute the metal binding site are mostly conserved, although the overall sequence similarity among them is low. In all structures but one, two Zn(II) ions are found in the active site. One of the metal ions (Zn 1 ) is bound to three histidine residues of the protein and a water molecule (17), which is thought to act as a nucleophile in catalysis. The coordination geometry of the other metal ion (Zn 2 ) is more variable among the available crystal structures (18 -21).
The active site of these enzymes is a shallow cleft hosting the Zn(II) ions and the activated water nucleophile in its bottom. Subclass B1 and B3 enzymes differ in the length and amino acid composition of the loops flanking the active site (17,20,21). Several crystal structures of metallo-␤-lactamases in complex with inhibitors are available (21)(22)(23)(24)(25)(26). All characterized inhibitors are able to bind the active site metal ions. Docking studies of substrates on the three-dimensional structures of these enzymes have been performed to delineate the interactions responsible for substrate binding and catalysis (18 -20, 27-29). In subclass B1 lactamases, the conserved residue Lys-224 is thought to bind the substrate carboxylate. Aside from this residue, no single group of residues can account for binding of the widely differing, efficiently hydrolyzed substrates, leaving an open question on the main determinants of substrate binding by class B lactamases.
The hydrolysis of ␤-lactams by metallo-␤-lactamases is thought to be similar to the base-catalyzed hydrolysis (30,31). It proceeds through an initial nucleophilic attack on the carbonyl carbon of the substrate by the zinc-bound water/hydroxide moiety, leading to the formation of a tetrahedral intermediate. Presteady state kinetic studies on the subclass B1 metallo-␤-lactamases from B. fragilis (CcrA) and from Pseudomonas aeruginosa (IMP-1) and the subclass B3 enzyme from Stenotrophomonas maltophilia (L1) have demonstrated that the hydrolysis of the chromophoric cephalosporin nitrocefin proceeds through an anionic intermediate in most cases (32)(33)(34)(35)(36). However, Spencer and co-workers (37) have demonstrated for the L1 enzyme that the formation of this intermediate is not a general feature of ␤-lactam hydrolysis catalyzed by metallo-␤-lactamases but is rather due to the particular chemical na- ture of nitrocefin, which allows the stabilization of the negative charge on the nitrogen atom through an extended -delocalization. In the same study, the ␤-lactam cleavage was identified as the rate-limiting step for the hydrolysis of another cephalosporin, cefaclor, and for the carbapenem substrate meropenem. Presteady state kinetic studies on the hydrolysis of nitrocefin and benzylpenicillin by Zn-BcII have also been carried at subzero temperatures using cryosolvents (38).
Altogether, the information gathered in the above mentioned studies points to the existence of a mechanistic heterogeneity in the hydrolysis of ␤-lactam antibiotics catalyzed by metallo-␤-lactamases caused both by differences in the active sites of the enzymes and in the structures of the substrates. With the aim of generating a more complete picture of ␤-lactam binding and hydrolysis by metallo-␤-lactamases, we studied the presteady state kinetics of the reaction of BcII with clinically used antibiotics: benzylpenicillin, cefotaxime, cephapirin, and imipenem. We were not able to detect any spectroscopically distinguishable intermediate during the hydrolysis of any of the substrates. Analysis of the substrate binding kinetics showed a marked heterogeneity among these ␤-lactams. Finally the apoenzyme could not bind any of the substrates tested, indicating that the metal ions play a crucial role in substrate binding to BcII.

EXPERIMENTAL PROCEDURES
Reagents-All chemicals were of the best quality available. Benzylpenicillin, cefotaxime, and cephapirin were purchased from Sigma. 10 -20 mM stock solutions of benzylpenicillin, cefotaxime, and cephapirin and 5-10 mM stock solutions of imipenem were prepared in distilled water. Benzylpenicillin stocks were not stored; cefotaxime and cephapirin stocks were stored at Ϫ20°C for no longer than 1 week, and imipenem stocks were stored at Ϫ70°C for no longer than 1 week. The concentrations of the stock solutions were determined spectrophotometrically, measuring the ⌬A at max after complete hydrolysis catalyzed by BcII. The following values of ⌬⑀ max were used: benzylpenicillin, ⌬⑀ 235 ϭ Ϫ800 M Ϫ1 ⅐cm Ϫ1 ; cefotaxime, ⌬⑀ 262 ϭ Ϫ7500 M Ϫ1 ⅐cm Ϫ1 ; cephapirin, ⌬⑀ 259 ϭ Ϫ7000 M Ϫ1 ⅐cm Ϫ1 ; and imipenem, ⌬⑀ 300 ϭ Ϫ9000 M Ϫ1 ⅐cm Ϫ1 .
Preparations of Enzyme Samples-Expression and purification of BcII were performed as reported previously (11). Purity of the enzyme preparations was checked by SDS-PAGE. Protein samples used in all experiments reported were exchanged from the final purification buffer to 15 mM Hepes, pH 7.5, using a HiPrep® 26/10 desalting column (Amersham Biosciences) or by two dialysis steps against Ͼ100 volumes of the same buffer. Buffer solutions were treated by stirring with Chelex to remove traces of divalent metal ions. The Zn(II) content was measured using the colorimetric reagent 4-(2-pyridylazo)resorcinol in denaturing conditions (16, 39 -41). The total metal content of protein samples obtained by dialysis or column buffer exchange was 1.4 -1.5 Zn(II)/ enzyme. A 6-h treatment of the native enzyme samples with 4-(2pyridylazo)resorcinol in the same buffer used for kinetic determinations removed ϳ0.3 eq of Zn(II)/enzyme, indicating that they were loosely bound. Apoenzymes were obtained as described previously (42). Protein concentrations were measured spectrophotometrically using ⑀ 280 ϭ 30,500 M Ϫ1 ⅐cm Ϫ1 (12).
Presteady State Kinetic Measurements and Data Analysis-Experiments were carried on an Applied Photophysics SX.18-MVR stoppedflow spectrometer. Inner tubing of the apparatus, syringes, and valves are devoid of metallic surfaces. Enzyme fluorescence measurements were performed by excitation at 280 nm (slit width ϭ 0.5 mm) and detection through a cut-off filter at Ͼ305 nm. The photomultiplier voltage was set to obtain a 4.0-or 0.8-V signal with the free enzyme. The background signal obtained after washing the observation cell with buffer ranged within 2-5% of the enzyme signal. Absorbance changes were measured at the corresponding max for each substrate using an absorption photomultiplier. The absorbance reference was set to the final absorbance value measured after completion of the reaction. For both absorbance and fluorescence measurements, the excitation/absorption pathlength was 0.2 cm. All reactions were performed in 15 mM Hepes, pH 7.5, without extra Zn(II) added except when indicated. The whole sample head was kept at 15°C using a Lauda RC6 thermostatted circulator.
Kinetic runs under pseudo-first order conditions to follow substrate binding were performed by mixing 2 or 4 M enzyme solutions with 10 -160 M substrate (syringe concentrations) and measuring the protein intrinsic fluorescence on a split time base. 20 runs were averaged for all tested conditions due to the low amplitude of the signal obtained and the high observed rate constants even at low substrate concentration. For the cephalosporin substrates, the enzyme fluorescence quenching was satisfactorily fitted to simple exponential functions. A simple reaction scheme was assumed for these substrates (Scheme 1). A linear fit of the data points was therefore performed to calculate k Ϫ1 ϩ k cat and k ϩ1 . Subtraction of the k cat value obtained from steady state experiments allowed us to calculate an estimate for k Ϫ1 . The reaction rates for imipenem hydrolysis were obtained from a global analysis of fluorescence and absorption reaction traces at different substrate concentration using the software DynaFit (43).
In all cases, the enzyme fluorescence was followed until the substrate was fully consumed. The final part of the reactions (when the substrate concentration is well below the K m value) were fitted to single exponential functions, and the whole reaction curves were normalized by subtraction of the estimated final fluorescence value. This allowed us to evaluate the amplitude of the quenching process for cefotaxime, cephapirin, and imipenem where inner filter effects due to absorption of the substrate at the excitation wavelength were estimated to be negligible.
The quenching amplitudes can be correlated to the maximum fraction of enzyme that binds substrate at a given substrate concentration (44). Therefore, the quenching amplitudes were fitted to the following function, where Q is the observed quenching amplitude and Q max is the expected fluorescence quenching at saturating substrate concentration, to obtain values of the fraction of fluorescence quenched in the E⅐C* complex and of the apparent equilibrium constant K S . Single turnover experiments were performed at enzyme concentrations of 20 or 60 M and at substrate concentrations of 10 or 30 M, respectively. For all ␤-lactams tested, both substrate consumption and enzyme fluorescence were recorded. Fluorescence recovery and the whole absorbance curves were fitted to single exponential functions.
Steady state kinetic parameters in the same conditions used for the presteady state experiments were obtained by recording and analyzing the whole reaction time courses of substrate consumption or by fitting the initial rates of hydrolysis to the Michaelis-Menten equation. The steady state rate equations were obtained with the computer program Albass (45).

RESULTS
Substrate Binding-Spencer and co-workers (37) have recently shown that the intrinsic fluorescence of the metallo-␤lactamase L1 from S. maltophilia is substantially quenched  during substrate binding, thus providing an efficient way to follow this process. We used the same strategy with BcII, which shows an identical behavior upon binding of all substrates tested by us. The initial fast quenching process could be followed for the two cephalosporins, but in the case of penicillin and imipenem this event was too fast to be monitored since it occurred during the dead time of the instrument (Fig. 2).
To evaluate the rate constants for the binding of cephalosporins, fluorescence quenching was monitored under pseudofirst order conditions, and the experimental curves were fitted to single exponential functions. The observed pseudo-first order rate constants for both cefotaxime and cephapirin binding increased linearly with increasing substrate concentration in the working substrate range (Fig. 3). Determinations at higher substrate concentrations were attempted, but the reaction rates were too high to allow a reliable fit. The linear increase of the observed rate constants with substrate concentration suggests a simple one-step binding process (Scheme 1) for which both k ϩ1 and k Ϫ1 could be estimated (Table I).
The fluorescence quenching amplitude measured at different substrate concentrations can be exploited to estimate the fraction of protein fluorescence quenched in the E⅐C* complex (Q max ) and the apparent reversible substrate binding equilibrium constant K S (44). The normalized amplitudes were fitted as described under "Experimental Procedures," yielding Q max and the constant K S (Supplemental Fig. S1 and Table I).
Imipenem binding to BcII followed by enzyme fluorescence quenching revealed a biexponential behavior. First, a very fast (close to the detection limit) initial quenching process occurred followed by a further decrease in the protein fluorescence emission at a slow rate (ϳ8 s Ϫ1 ) before fluorescence was recovered due to substrate depletion (Fig. 4). The finding of a fast initial binding step is consistent with the high K m and k cat values of this substrate. The second process, however, was too slow to be included in a linear reaction pathway since the turnover number for imipenem hydrolysis is 122 s Ϫ1 . The simplest reaction schemes that can account for these observations are those shown in Scheme 2 that include formation of a non-productive enzyme⅐substrate complex (E⅐Im 2 ). The whole reaction time courses for imipenem hydrolysis and protein intrinsic fluorescence quenching were globally fitted to this reaction scheme. It was assumed that the absorptivity of the ␤-lactam chromophore is unaltered in both the non-productive (E⅐Im 2 ) and the productive (E⅐Im 1 ) complexes compared with free imipenem. Initial guesses for the relative quenching amplitudes were obtained from the biexponential fits of the first 0.3 s of the fluorescence signal. The value of k ϩ1 was fixed as diffusion-limited (10 8 M Ϫ1 ⅐s Ϫ1 ), and all other kinetic constants were fitted simultaneously to all data sets. The resulting fit and the obtained rate constants are displayed in Fig. 4.
Apo and Zn(II)-reconstituted BcII-To evaluate the influence of the metal ions in substrate binding, the same experiments were carried with apoenzymes using cefotaxime, benzylpenicillin, and imipenem as substrates. No quenching of the apoenzyme intrinsic fluorescence was detected after mixing with any of the mentioned substrates. In all cases, a residual substrate hydrolysis was detected that was attributable to the presence of a small fraction of Zn-enzyme in the apoenzyme preparations. The same apoenzyme preparations were reconstituted with excess Zn(II) (10 M Zn(II), ϳ5 eq). Fig. 5 shows that substrate binding was restored in the reconstituted apoenzyme. Quenching of the enzyme fluorescence occurs before the substrate undergoes any chemical modifications (see below). These observations indicate that apoBcII cannot bind any of the substrates tested.
␤-Lactam Hydrolysis-All substrates were assayed under single turnover conditions following both substrate consumption and enzyme fluorescence. The rates of fluorescence recovery after the initial quenching and of ␤-lactam hydrolysis (followed by absorbance) were almost identical in all cases (Table II).
The first step in the hydrolysis reaction, i.e. formation of a tetrahedral gem-diol intermediate after nucleophilic attack by the enzyme, is expected to result in a modification of the substrate spectrum. The fact that both the fluorescence recovery and changes in substrate absorbance proceeded at the same rates indicates that quenching of the enzyme intrinsic fluorescence occurs before the substrate has undergone any chemical reaction and that the fluorescence decay is mainly due to substrate binding. Although it is evident that the rate-limiting step should happen after substrate binding, the present data do not allow us to make any inference on the nature of the rate-  limiting step as none of the putative reaction intermediates seemed to reach a concentration high enough to be detected. Hydrolysis of imipenem measured under single turnover conditions did not show biphasic fluorescence curves (Supplemental Fig. S2). In the presence of excess enzyme, the substrate is expected to be hydrolyzed through the fast, direct reaction path and be depleted before any significant amount of the nonproductive complex is formed. To verify this assertion, the ex-pected fluorescence and absorbance traces were calculated using the rate constants obtained under pseudo-first order conditions for this reaction. Only the quantum yields of the E⅐Im 1 and E⅐Im 2 complexes were allowed to vary to fit the experimental data.
The steady state kinetic parameters for the reaction of BcII with the assayed substrates were measured in the same reaction conditions. The estimated values are in good agreement with the parameters calculated with the steady state rate equations using the values obtained for the individual rate constants (Table III). DISCUSSION Metallo-␤-lactamases are well known for their ability to inactivate most clinically useful ␤-lactams regardless of their structural diversity. As a member of this group of enzymes, BcII can hydrolyze several distinct ␤-lactam antibiotics albeit with varying effectiveness (6,7). Several studies have been performed on the reactivity of BcII against different antibiotics (31, 46 -48). However, there is little direct evidence on the structural reasons for their broad substrate specificity. We attempted to shed some light on these issues through the analysis of the presteady state kinetics of the reaction of BcII against penicillins, cephalosporins, and carbapenems.
Substrate binding to BcII was followed by Trp fluorescence. Binding of substrates from the three ␤-lactam classes used in this work resulted in quenching of BcII intrinsic fluorescence as already reported for S. maltophilia L1 (37). Substrate consumption was followed by the absorption change resulting from the loss of the amide group in the antibiotic. Analysis of the different processes revealed that fluorescence recovery and substrate consumption proceed essentially with the same rate, i.e. chemical modifications of the substrates cannot be distinguished from SCHEME 2. Proposed kinetic schemes for imipenem hydrolysis by BcII. The non-productive E⅐Im 2 complex was formed by direct binding to the free enzyme (A) or by reorganization of the productive E⅐Im 1 complex (B).  product release. From this evidence we conclude the following. 1) In the most populated reaction intermediates detected in this work (i.e. the quenched enzyme forms), all substrates tested are still intact but tightly bound to the enzyme.
2) The rate-limiting step is a chemical step, and no chemically modified substrate species accumulates to a detectable amount. Studies of the basecatalyzed hydrolysis of ␤-lactams (30) and quantum chemical calculations on the hydrolysis of ␤-lactams by BcII (27) suggest that the highest activation energy barrier of the reaction is the formation of a tetrahedral gem-diol intermediate after nucleophilic attack on the carbonyl carbon atom. In the quantum chemical study, it was specifically pointed out that a tetrahedral intermediate would not accumulate to an experimentally detectable amount. The results presented here give experimental support to this assertion. For cefotaxime and cephapirin, the quenching process proceeds with a rate constant low enough to allow analysis of the binding rate constants. The linear dependence of these rate constants with substrate concentration (Fig. 2) indicates a simple binding mechanism to yield the quenched complex. It is noteworthy that the second order rate constants for formation of the quenched E⅐C* complex were very similar for cefotaxime and cephapirin (31 and 32 s Ϫ1 , respectively). If one considers that substrate binding depends mainly on the interaction of the ␤-lactam core with the Zn(II) site (see below), similar binding constants would be expected for substrates sharing the same core structure. The binding behavior of the cephalosporin substrates used in this work differs mostly in the rate constant for decay of the quenched complexes, i.e. in different stabilities of the E⅐C* complexes. These differences should be due to a differential interaction of the cephalosporin substituents with residues at the active site.
Benzylpenicillin and imipenem binding to BcII occurred within the instrument dead time. We are able to propose that the high K m values of these substrates arise from a high k off rate, which, in concert with a relatively high k cat value, makes the expected relaxation process too fast to be detected. Binding of the carbapenem substrate meropenem to L1 could be detected in agreement with lower K m and k cat values (13 M and 77 s Ϫ1 , respectively) (37). In the reaction of BcII with imipenem, two substrate-bound enzyme forms with different quantum yields were clearly distinguishable (E⅐Im 1 and E⅐Im 2 , in Scheme 2). Formation of the E⅐Im 1 complex induced an initial drop in fluorescence emission, while the slow formation of complex E⅐Im 2 resulted in a further decay of the enzyme fluorescence. Two mechanisms can account for the observed behavior: a slow binding of the substrate to the enzyme giving rise to this complex (Scheme 2A) or a rearrangement of the initial enzyme⅐substrate complex E⅐Im 1 to yield the non-productive complex (Scheme 2B). The experimental data available do not allow us to distinguish between these two mechanisms, although the very low second order rate constant obtained assuming the direct formation of a non-productive form favors the mechanism sketched in Scheme 2B in which the substrate could undergo a conformational rearrangement within the active site.
In both proposed schemes, the rate constants for formation and decomposition of the non-productive complex are low. This observation indicates that this complex is relatively stable with the substrate being possibly held in the active site by stronger interactions than in the productive complex E⅐Im 1 . An important difference between carbapenems and the other ␤-lactams used is that the C-6 substituent is in an ␣ orientation instead of ␤ (as found in penicillins and cephalosporins, see Fig. 1). If we assume that the imipenem bicyclic core is oriented as proposed for other ␤-lactam substrates (positioning the ␤-lactam carbonyl for optimum nucleophilic attack by Zn(II)-bound water), the C-6 hydroxyethyl substituent points toward the active site groove "floor" (Asp-120 C␣ and C␤). An alternative substrate orientation in which the C-6 side chain is better accommodated within the active site could probably place the ␤-lactam ring away from the attacking nucleophile. Also, the placement of the C-2 substituent in the same channel occupied by the cephalosporin C-3 substituent should result in a different position of the lactam carbonyl relative to the protein frame.
The presteady state study from Spencer and co-workers (37) reveals that in L1 the E⅐S complexes for meropenem and cefaclor are longer lived than in BcII, and a substrate rearrangement occurs during the residence time. In contrast to this, hydrolysis of benzylpenicillin and cephalosporins by BcII proceeds without such a rearrangement. The protein structure around the active site in subclass B1 and B3 metallo-␤-lactamases is very different, and therefore variations in the substrate binding mode are to be expected between BcII and L1. In particular, the conserved residue Lys-224, which is proposed to interact with the carboxylate moiety of ␤-lactams in subclass B1 enzymes (19), has no counterpart in L1 (20). Substitution of Lys-224 results in markedly increased K m values for cephalosporins and imipenem in IMP-1 and for benzylpenicillin in both IMP-1 and CcrA (50,51). In contrast, the corresponding S224K L1 mutant does not show significant changes in the kinetic parameters for any substrate (52). This suggests that Lys-224 could direct electrostatically the substrate in an optimal orientation. These data point to different substrate binding mechanisms for subclass B1 and B3 metallo-␤-lactamases.
When Zn(II) was removed from the active site, the intrinsic fluorescence of the apoenzyme was not quenched by any of the substrates assayed. Chemical modification of the ␤-lactam moiety (directly monitored through the amide absorbance) proceeded at the same rate as that observed for protein fluorescence recovery, confirming that the substrate is unaltered in the quenched complex. Hence, the absence of fluorescence quenching directly proves that apoBcII does not bind any of the substrates used.
The involvement of the active site metal in substrate binding has been proposed since the earliest works on metallo-␤lactamases, but none of these proposals pointed to the essentiality of Zn(II) (17,53). Several crystal structures of metallo-␤lactamases in complex with inhibitors have been solved (22)(23)(24)(25)(26). The binding modes of the inhibitors are highly variable as are their chemical structures, but in all cases a charged group of the inhibitor is located between the two metal ions (21,22,24). However, the finding that no substrate binds to the apoenzyme is not trivial. The crystal structure of apoBcII has been solved at 2.5-Å resolution (54). The overall structure of the apoenzyme does not differ from that of the holoenzyme: only the active site is slightly more open than that of the Zn 2 form. Therefore, the structural determinants located in the protein (if any) for binding of the substrates should be present in the apoenzyme. In other thoroughly characterized Zn(II)-hydrolases, the substrate binding determinants are shared between the metal site and the protein scaffold, the contribution of each part being variable among the different enzymes and substrates. For example, apocarboxypeptidase A tightly binds peptide substrates mainly through hydrophobic interactions with the terminal residue side chain (55,56). Indeed the substrate binding pocket in carboxypeptidases confers to these enzymes their substrate specificity (57). For the Zn(II) enzyme epoxyalkane:CoM transferase, the binding determinants have been dissected with the zinc-cofactor interaction contributing to ϳ60% of the binding free energy release (58). Our data are also in agreement with the observation of an enhancement of Zn(II) affinity in metallo-␤-lactamases in the presence of substrates (59). The finding of an essential role for zinc in the substrate binding process can be related to the broad substrate profile of most metallo-␤-lactamases. Due to the diversity in the chemical structures of substrates hydrolyzed by metallo-␤lactamases, recognition of the ␤-lactam moiety (including the bicyclic carboxylate substituent) should somehow be coded into the ␤-lactamase active site. Transition metal ions in solution can catalyze nucleophilic substitution on ␤-lactams. Mechanisms involving one or two metal ions have been implicated in the hydrolysis, aminolysis, and methanolysis of ␤-lactams (30,60). Binuclear Zn(II) complexes have been synthesized that catalyze the hydrolysis of nitrocefin through the same anionic intermediate detected with the enzymes CcrA and L1 (49). Depending on the reaction conditions, the metal ion(s) can act by enhancing the nucleophilicity, by favoring the cleavage of the amide bond, and/or by stabilizing the leaving amine group. We propose that the catalytic power of metallo-␤-lactamases resides in the correct placement of one or two metal ions within an active site that merely helps to orient the substrate toward the nucleophilic machinery. This proposal is in line with the lack of an optimized substrate recognition patch in B1 and B3 metallo-␤-lactamases that provides a broad substrate spectrum.