Catalytic Role of the Metal Ion in the Metallo-β-lactamase GOB*

Metallo-β-lactamases (MβLs) stand as one of the main mechanisms of bacterial resistance toward carbapenems. The rational design of an inhibitor for MβLs has been limited by an incomplete knowledge of their catalytic mechanism and by the structural diversity of their active sites. Here we show that the MβL GOB from Elizabethkingia meningoseptica is active as a monometallic enzyme by using different divalent transition metal ions as surrogates of the native Zn(II) ion. Of the metal derivatives in which Zn(II) is replaced, Co(II) and Cd(II) give rise to the most active enzymes and are shown to occupy the same binding site as the native ion. However, Zn(II) is the only metal ion capable of stabilizing an anionic intermediate that accumulates during nitrocefin hydrolysis, in which the C–N bond has already been cleaved. This finding demonstrates that the catalytic role of the metal ion in GOB is to stabilize the formation of this intermediate prior to nitrogen protonation. This role may be general to all MβLs, whereas nucleophile activation by a Zn(II) ion is not a conserved mechanistic feature.

Molecular structures of M␤Ls from the three subclasses have been solved by x-ray crystallography (11,13,14,24,28,31). Comparison of their structures reveals a common ␣␤/␤␣ sandwich fold, in which diverse insertions and deletions have resulted in different loop topologies and, ultimately, in different zinc coordination environments and metal site occupancies (Fig. 1). M␤Ls bind up to two metal ions in their active sites. In B1 and B3 enzymes, Zn1 is tetrahedrally coordinated to three histidine ligands (His 116 , His 118 , and His 196 in Fig. 1, A and C) and a water/OH Ϫ molecule (3H site), which is the attacking nucleophile (13,14,28,31). The coordination polyhedron of Zn2 in B1 enzymes is provided by Asp 120 , Cys 221 , His 263 , and one or two water molecules (DCH site) (Fig. 1A) (13,14). Notably, this site constitutes the active species in mono-Zn(II) B2 enzymes (Fig. 1B) (24). Instead, two mutations (C221S and R121H) affect the Zn2 coordination geometry in B3 M␤Ls, and the metal ion is now bound to Asp 120 , His 121 , His 263 , and one or two water molecules, whereas Ser 221 is no longer a metal ligand (DHH site) (Fig. 1C) (28,31). A remarkable exception is constituted by the deepest branching member of the M␤L B3 subclass, GOB from E. meningoseptica (4). In all reported GOB sequences, His 116 and Ser 221 are substituted by Gln and Met, respectively, suggesting the presence of an unusually perturbed metal binding site within this subclass.
We have recently reported a biochemical and biophysical characterization of GOB-18 (33). In contrast to all known M␤Ls, GOB-18 is fully active against a broad range of ␤-lactam substrates using a single Zn(II) ion. Based on spectroscopic, mutagenesis, and modeling experiments, we already proposed that the Zn(II) ion is bound to Asp 120 , His 121 , His 263 , and one or two solvent molecule(s) (i.e. in the canonical Zn2 site of dinuclear M␤Ls) (Fig. 1D).
These findings are puzzling because, although it is accepted that B2 lactamases may work without a metal-activated nucleophile (24,36), this is not the case for B3 lactamases (37). A crystal structure of mono-Zn(II) L1 has revealed that the single metal ion is located in the 3H site (i.e. the one involved in nucleophile activation) (38). A recent kinetic and spectroscopic study in L1 has also indicated that mono-Zn(II) L1 in solution is active with the metal ion localized in the 3H site (37). Thus, there is no precedent of a B3 lactamase without a metal-activated nucleophile. The absence of a crystal structure for GOB has precluded a possible description of the role of the metal site in this enzyme, which is still unclear. Here we report a series of experiments that allow us to propose that the metal site in GOB is involved in stabilization of an anionic intermediate that accumulates during catalysis, thus favoring C-N bond cleavage after the nucleophilic attack. These results are unprecedented because these intermediates have been only identified in catalysis by dinuclear lactamases (16,17,37,39) and disclose the existence of a common catalytic feature in M␤Ls from all subclasses.

EXPERIMENTAL PROCEDURES
Chemicals-All reagents were purchased from Sigma with the exception of the NAP-10 gel filtration column purchased from Amersham Biosciences. Metal-free buffers were prepared adding Chelex 100 (Sigma) to normal buffers and stirring for 0.5 h.
Metal Content Determination-Metal content determinations were performed by atomic absorption spectroscopy using a Metrolab 250 instrument operating in the flame mode.
Non-steady-state Kinetics-Nitrocefin hydrolysis catalyzed by different GOB-18 metal variants was followed, employing a Jasco V-550 spectrophotometer for slow reactions and an SX.18-MVR stopped-flow spectrometer associated with a PD.1 photodiode array (Applied Photophysics, Surrey, UK) or an absorbance photomultiplier for fast reactions. All measurements were performed in metal-free buffer (100 mM Hepes, pH 7.5, 200 mM NaCl). Reaction temperature was 30°C for all GOB-18 metal variants with the exception of Zn(II)-GOB-18, where the reaction temperature was 4°C. Enzyme and substrate concentrations were adjusted to 10 -30 and 5-15 M, respectively, to attain single turnover conditions. Nonlinear regression analysis was used to fit single-wavelength absorbance changes with the program Dynafit (40). Groups of data corresponding to hydrolysis of nitrocefin were fitted simultaneously. The molar extinction coefficients of nitrocefin used were as follows: substrate ⑀ 390 ϭ 18,400 M Ϫ1 cm Ϫ1 ; product ⑀ 390 ϭ 6,300 M Ϫ1 cm Ϫ1 , ⑀ 490 ϭ 17,400 M Ϫ1 cm Ϫ1 , ⑀ 665 ϭ 500 M Ϫ1 cm Ϫ1 . Steady-state kinetic parameters for the hydrolysis of nitrocefin were calculated as k cat ϭ k 2 k 3 /(k 2 ϩ k 3 ) and K m ϭ k 3 (k Ϫ1 ϩ k 2 )/k 1 (k 2 ϩ k 3 ) for Zn(II)-GOB-18 and k cat ϭ k 2 and K m ϭ (k Ϫ1 ϩ k 2 )/k 1 for all other GOB-18 metal variants.
Steady-state Kinetics-All reactions were performed at 30°C in buffer (15 mM Hepes, pH 7.5, 200 mM NaCl). Antibiotic hydrolysis was monitored following the absorbance variation  (24); C, dinuclear B3 M␤L L1 from S. maltophilia (Protein Data Bank code 1sml) (28); D, molecular model for the mononuclear B3 M␤L GOB-18 from E. meningoseptica (33). Zinc atoms are shown as large gray spheres, and water molecules (W) are shown as small gray spheres. Coordination bonds are shown as dashed lines.
resulting from the hydrolysis of the ␤-lactam ring, using the following extinction coefficients: imipenem, ⌬⑀ 300 ϭ Ϫ9,000 M Ϫ1 cm Ϫ1 ; penicillin G, ⌬⑀ 235 ϭ Ϫ800 M Ϫ1 cm Ϫ1 ; cefotaxime, ⌬⑀ 260 ϭ Ϫ7,500 M Ϫ1 cm Ϫ1 . The kinetic parameters k cat and K m were derived from nonlinear fit of the Michaelis-Menten equation to initial rate measurements recorded on a Jasco V-550 spectrophotometer. In each case, k cat values have been calculated taking into account the concentration of metallated mononuclear enzyme.
Circular Dichroism-Measurements were performed at 25°C, using a Jasco J-715 spectropolarimeter flushed with N 2 . Samples were prepared dialyzing the corresponding protein solution against 300 volumes of buffer 10 mM Tris-HCl, pH 7, 50 mM NaCl, twice for 8 -12 h, at 4°C.
NMR Spectroscopy-NMR spectra were recorded at 298 K on a Bruker Avance II 600 spectrometer operating at 600.13 MHz. 1 H NMR paramagnetic spectra were acquired under conditions to optimize detection of the fast relaxing isotropically shifted resonances, either using the superWEFT pulse sequence (41) or water presaturation. These spectra were recorded over large spectral widths with acquisition times ranging from 16 to 80 ms and intermediate delays from 2 to 35 ms. 1 H NMR spectra in the diamagnetic envelope were obtained employing a selective shaped pulse for solvent saturation (42). Hydrodynamic radius determination by pulsed field gradient 1 H NMR was performed according to Ref. 43. A 113 Cd NMR experiment was performed at 133 MHz, with a 5-mm reverse broad band probe. About 260,000 free induction decays were recorded, acquiring 32,768 complex points in 100 ms, with a 30°observe pulse (4 s), a spectral width of 160,000 Hz (1,200 ppm), and a relaxation delay of 0.5 s. The external reference was Cd(ClO 4 ) 2 .
111m Cd(II) PAC Spectroscopy-The experiment was performed employing a setup with six BaF 2 detectors. Radioactive 111m Cd was produced on the day of the experiments by the Cyclotron Department at the University Hospital in Copenhagen. Preparation and purification of 111m Cd were as previously described (44). A 25-l 111m Cd-containing solution was mixed with non-radioactive cadmium acetate and buffer (Hepes and NaCl; see final concentrations below). pH was adjusted to 7.5 at room temperature. Apo-GOB-18 in 15 mM Hepes, pH 7.5, 200 mM NaCl was added at a final enzyme concentration of 40 M, and pH was adjusted to 7.5 at room temperature. The total volume of the sample was 750 l. Sample was then left 2 h at room temperature to allow binding of Cd(II) to the protein and subsequently run through a gel filtration column (NAP10), using 15 mM Hepes, pH 7.5, 200 mM NaCl as eluent. 111m Cdcontaining protein fractions were pooled (four fractions with a volume of 250 l each). The sample was frozen in liquid nitrogen and placed in the PAC instrument, where the temperature was set to Ϫ20°C and controlled by a Peltier element. Fits to PAC data were carried out with 300 data points, disregarding the first five points due to systematic errors in these. The analytical expression for the perturbation function is known, and five parameters are fitted for each nuclear quadrupole interac-tion (45). In radiotracer ( 109 Cd) experiment condition were the same as for PAC experiments.

Nitrocefin Hydrolysis by Zn(II)-GOB-18 Proceeds by Means of an Anionic
Intermediate-Nitrocefin, a synthetic cephalosporin, has been extensively employed as a chromogenic probe of the mechanism of M␤Ls. Benkovic and co-workers (16,17) reported for the first time the accumulation of a reaction intermediate in the hydrolysis of nitrocefin by the B1 enzyme CcrA from B. fragilis. The same intermediate was also reported for the B3 enzyme L1 (37,46) and for an evolved mutant of the B1 lactamase BcII (39). In all cases, this intermediate was stabilized exclusively by the dinuclear forms of these enzymes.
We studied the hydrolysis of nitrocefin catalyzed by mono Zn(II)-GOB-18 employing a stopped-flow equipment coupled to a photodiode array. Besides the decay of the substrate absorption at 390 nm and the rise of the product absorption at 490 nm, the sequence of electronic absorption spectra obtained clearly shows the accumulation and the decay of a species with an intense absorption band centered at 665 nm, identical to the one reported for the anionic intermediate previously observed for dinuclear enzymes ( Fig. 2A).
In order to obtain a minimal kinetic mechanism for the UVvisible data, we performed similar experiments with different amounts of nitrocefin and enzyme measuring the evolution of the absorbance at 390, 490, and 665 nm (Fig. 2B). Data were then subjected to simultaneous global fit. The kinetic model that fits the experimental data best is as follows.
The individual kinetic constants obtained were employed to calculate k cat ϭ 4.4 Ϯ 0.1 s Ϫ1 and K m ϭ 18 Ϯ 3 M as described under "Experimental Procedures." These values compare very well with the steady-state kinetic parameters determined under the same experimental conditions: k cat ϭ 4.0 Ϯ 0.3 s Ϫ1 and K m ϭ 25 Ϯ 5 M, strongly supporting the proposed model.
The finding of a reaction intermediate in mononuclear Zn(II)-GOB-18 similar to that reported for dinuclear enzymes suggests that the Zn2 ion is involved in its stabilization in all cases. To further explore the role of this metal ion, we decided to probe the mechanism of nitrocefin hydrolysis by GOB-18 substituted with a variety of transition divalent metal ions.
Nitrocefin Hydrolysis by M(II)-GOB-18 Derivatives-Cytoplasmic overexpression of GOB-18 in E. coli gives rise to a mixture of the Fe(III) and Zn(II) variants (33). We have already optimized a procedure for metal depletion (33). The apoprotein can then be fully loaded with metal ions by dialysis (see "Experimental Procedures"). We have now prepared different metal derivatives (referred to as M(II)-GOB-18 hereafter) using Co(II), Ni(II), and Cd(II) as surrogates of the native Zn(II) ion. The metal content of each M(II)-GOB-18 variant was determined by atomic absorption spectroscopy. Similar values were obtained for several protein preparations. The metal content of the different M(II)-GOB-18 derivatives never exceeded 1 eq per protein molecule in agreement with the metal content of the native enzyme (Table 1).
In order to determine the essentiality of the Zn(II) ion in the stabilization of the anionic intermediate in the hydrolysis of nitrocefin by GOB-18 and to explore the mechanistic changes induced by the metal replacements, reactions catalyzed by the different M(II)-GOB-18 derivatives were studied similarly to those above described for the native Zn(II) enzyme. Two major facts result from these experiments: 1) the intermediate is not accumulated in the reaction catalyzed by M(II)-GOB-18 derivatives as detected by Zn(II)-GOB-18 (direct conversion of substrate to product was observed in all cases; see supplemental Fig. S1), and 2) all M(II)-GOB-18 derivatives are less active than Zn(II)-GOB-18, and among these the Co(II) variant displays higher levels of nitrocefinase activity than the Cd(II) and Ni(II) derivatives.
In each case, single wavelength traces corresponding to substrate depletion and product formation were globally fit to the model presented in Scheme 2 (supplemental Fig. S1). Kinetic constants obtained are summarized in supplemental Table S1.
The fact that there is no accumulation of the anionic intermediate in the hydrolysis of nitrocefin catalyzed by GOB-18 metal variants other than the Zn(II) native one may be interpreted as an inherent inability of the metal ions to stabilize the intermediate or as a different binding of these ions to the protein. To explore these hypotheses, we employed different spectroscopic techniques to ascertain the coordination sphere of the M(II)-GOB-18 derivatives.

Spectroscopic and Functional Study of M(II)-GOB-18
Derivatives-We exploited the features of the employed metal ions as spectroscopic probes of their coordination spheres. High spin Co(II) and high spin Ni(II) can be employed as paramagnetic probes to identify the metal ligands by following their effect on the 1 H NMR spectrum of the protein. Cd(II) can be directly interrogated by NMR and PAC spectroscopy.
We tested the activity of Co(II)-GOB-18, Ni(II)-GOB-18, and Cd(II)-GOB-18 against a series of clinically relevant substrates (penicillin G, cefotaxime, and imipenem). The corresponding steady-state kinetic parameters are summarized and compared with those of Zn(II)-GOB-18 in Table 2. The general activity trend is Zn(II) Ͼ Cd(II), Co(II) Ͼ Ni(II) (in all cases, the addition of 20 M metal ion to the reaction medium did not improve the catalytic efficiency). Remarkably, the catalytic efficiencies of M(II)-GOB-18 derivatives were mainly affected in the k cat values as compared with the native enzyme. The k cat values for Cd(II), Co(II), and Ni(II)-GOB-18 decreased by ϳ3-fold, 10-fold, and ϳ100-fold, respectively, compared with those corresponding to Zn(II)-GOB-18. Instead, K m values for each substrate were within the same order of magnitude for the different assayed metal ions.
We recorded circular dichroism and 1 H NMR spectra of M(II)-GOB-18 derivatives as well as those of apo and Zn(II)-GOB-18 in order to probe their correct folding (see supplemental Fig. S2). CD spectra in the far UV range are similar, with some minor differences in the case of apo-GOB-18. However, the CD spectra in the near UV region suggest a reduced level of tertiary structure in apo-GOB-18 as compared with the metallated forms. Also, the signals in the 1 H NMR spectrum of apo-GOB-18 show less dispersion than for any of the metallated variants (see supplemental Fig. S2), suggesting that (despite showing secondary structure) apo-GOB-18 does not adopt the

TABLE 1 Metal content of GOB-18 variants as determined by atomic absorption spectroscopy
The presented values correspond to the average of at least three independent protein preparations. Deviations from the presented values never exceeded 15%.

GOB-18 variant Metal equivalents
Zn ( 3B). When the spectrum was recorded in D 2 O, signals at 60 and 47 ppm were absent, indicating the presence of two solventexchangeable resonances that can be attributed to two His ligands. In addition, the low number of signals observed in the paramagnetic region of the spectrum and their dispersion in the chemical shift range is also consistent with a pentacoordinated Co(II) ion (47,48). The present spectrum shows a larger signal dispersion than that of mononuclear Co(II)-L1 (37). Fig. 3C shows the electronic absorption spectrum of Ni(II)-GOB-18. The signals observed in the visible range correspond to ligand field electronic transitions of the Ni(II) ion bound to the protein. The spectrum resembles that of Ni(II)-carboxypeptidase A, which contains a single hexacoordinated metal ion (49). The 1 H NMR spectrum of Ni(II)-GOB-18 reveals a set of isotropically shifted resonances spanning from 70 to 50 ppm (Fig. 3D). When the spectrum was recorded in D 2 O, the signals at 70 and 67 ppm were no longer detected, indicating the presence of two exchangeable resonances that can be attributed to two His ligands. The chemical shifts of this variant resemble those reported for Ni(II)-carboxypeptidase A (50). This observation, together with the absence of signals experiencing pseudocontact shifts, supports the hypothesis of a pseudo-octahedral Ni(II) site (50).
The 113 Cd NMR spectrum of 113 Cd(II)-GOB-18 shows only one signal (Fig. 3E) at 115 ppm, revealing only one metal binding site. Besides, the chemical shift of the observed signal is consistent with a coordination sphere containing nitrogen and oxygen donor atoms. 113 Cd(II) resonances in proteins with a mixed N/O coordination sphere are usually found between 300 and 40 ppm, O being the most shielding donor atom (51). On the other hand, each sulfur donor present at a 113 Cd(II) binding site is so deshielding that it is possible to use chemical shifts to determine the number of sulfur donor atoms at an unknown site (51). Resonances of 113 Cd(II)-BcII have been reported at 262 and 142 ppm, which have been attributed to the metal ion

Steady-state kinetic parameters for the hydrolysis of different ␤-lactam substrates by GOB-18 variants
Kinetic parameters were derived from nonlinear fit of the Michaelis-Menten equation to initial rate measurements. The reaction medium was 15 mM Hepes, pH 7.5, 200 mM NaCl, at 30°C. The presented values correspond to the average of at least three independent enzyme preparations.

Substrate
Penicillin G Cefotaxime Imipenem bound to sites DCH and D3H, respectively, differing significantly from the resonance here reported (52,53). The 115 ppm resonance observed for 113 Cd(II)-GOB-18 agrees well with experimental 113 Cd NMR data giving a resonance at 120 ppm for carboxypeptidase A (54), presumably with a N 2 O 3 coordination sphere (two histidines, one water, and a bidentate carboxylate) and also with theoretical predictions (55). Fig. 3F shows the 111m Cd(II) PAC spectroscopic data recorded for Cd(II)-GOB-18. With 2 eq of Cd(II) added to the protein and an incubation time of 2 h, the PAC signal is found at a relatively low frequency. It can be fit with just one nuclear quadrupole interaction, which would agree with the single resonance observed by 113 Cd NMR. Also, only one Cd(II) ion was observed to bind per protein molecule, as demonstrated by a radiotracer experiment using 109 Cd(II), where 2 eq of 109 Cd(II) were incubated with the protein under the same conditions as applied for the PAC experiment, and a subsequent gel filtration displayed very close to 50% of the radioactivity in the same fractions as the protein. All data from 113 Cd NMR, 111m Cd PAC, and atomic absorption spectroscopy for the fully metalloaded protein are thus consistent with one binding site.
The nuclear quadrupole interaction fall in the spectral range expected for ligands coordinating with nitrogen or oxygen atoms. The PAC spectrum obtained from a mixture of 111m Cd(II) but not protein in the buffer is very different from the one described above (not shown), indicating that in that case, no free 111m Cd(II) was present. Fitted parameters are shown in supplemental Table S2.
M(II)-D120S GOB-18 Derivatives-Spectroscopic data on Co(II)-, Ni(II)-, and Cd(II)-substituted GOB-18 unequivocally point to metal binding to a single site in all M(II)-GOB-18 derivatives. However these data could be compatible with binding to a His 2 Gln or to a His 2 Asp site. In order to unequivocally determine the metal binding site in GOB-18 metal derivatives, we obtained the metal-substituted mutant D120S GOB-18, where one of the possible metal binding sites is drastically perturbed. D120S GOB-18 as isolated contained no metal bound and displayed no activity against nitrocefin. Although exhaustive dialysis of this mutant against buffer containing an excess of ZnSO 4 , CoSO 4 , NiCl 2 , or CdCl 2 resulted in some extent of metal binding (Table 3), all metal derivatives displayed nondetectable activity against nitrocefin. All D120S GOB-18 metal derivatives were interrogated by CD, showing similar features to those presented for the wild type enzyme (not shown), thus indicating comparable secondary and tertiary structure. These results allow us to unambiguously assign site DHH as the metal binding site in all GOB-18 derivatives.

DISCUSSION
M␤Ls represent the largest group of carbapenemases (5)(6)(7)(8)(9)(10). The lack of a pan-M␤L inhibitor is mostly due to the failure to identify common structural and mechanistic features in enzymes from different organisms, which show a striking diversity in terms of substrate spectrum, active site structure, and metal ion requirements. The Zn(II) ions in M␤Ls are required for substrate binding and hydrolysis, but the specific role and essentiality of each metal binding site are still subject of intense debate (5)(6)(7)(8)(9)(10).
Zn(II) is ubiquitous in nature, being the only metal ion present in enzymes from all six groups in the EC nomenclature (56). This fact highlights its amazing chemical versatility. In the case of M␤Ls, the Zn(II) ion is able to contribute to ␤-lactam hydrolysis by 1) lowering the pK a of a bound water molecule, which may act as a nucleophile, providing a high local concentration of hydroxide ions at neutral pH (37,57); 2) acting as a Lewis acid, polarizing the CϭO bond and therefore augmenting the electrophilic nature of the carbonyl carbon (57); and 3) stabilizing a negative charge in the bridging nitrogen of the lactam moiety, after C-N bond cleavage (16,17,24,36,58,59). These three roles have been invoked for the two metal binding sites in M␤Ls, but it is not clear yet which of them are essential for catalysis.
Substrate docking studies to the active site of B1 and B3 lactamases suggest that the attacking nucleophile in the dinuclear forms is the bridging water/OH Ϫ ligand (13,14,28,36,60). This moiety is asymmetrically positioned with respect to the two metal ions, lying closer (1.9 -2.1 Å) to the metal ion in the 3H site than to the Zn(II) ion in the DCH or DHH sites (2.1-3.1 Å) (13,14,28). The shorter bond length in the former case is consistent with this ligand being a hydroxide and with the idea that the 3H site is responsible for lowering the pK a of a water molecule, thus being responsible for nucleophile activation (57). The role of CϭO polarization by this same zinc ion is not supported by QM/MM calculations and by different docking studies, which reveal that the ␤-lactam bond may not directly bind to the metal ion (60,61).
The availability of crystal structures of mono-Zn(II) BcII (a B1 enzyme) (11) and L1 (B3) (38) disclosing the presence of one metal ion localized in the preserved 3H site and theoretical studies of substrate binding to mononuclear BcII (62-64) have supported the hypothesis that only this metal site would be essential, delivering the attacking nucleophile. The structure of the mono-Zn(II) B2 enzyme CphA revealed that the only metal ion in this case is bound to the DCH site because one of the His residues of the 3H binding site is replaced by an Asn (24). In this case, the attacking nucleophile has been proposed to be a water molecule that is not activated by a metal ion (24,36). However, B2 enzymes have been considered as an exception, being limited to few bacteria and performing as exclusive carbapenemases.
The recent report of GOB, a B3 enzyme, which is active as a mononuclear enzyme and is not inhibited by excess Zn(II) (33,36), provides an excellent opportunity to examine the role of the metal ion in a broad spectrum M␤L. Here we have shown that nitrocefin hydrolysis mediated by mono-Zn(II) GOB pro-  (66). The kinetic data of nitrocefin hydrolysis by Co(II), Cd(II), and Ni(II) GOB can also be fit to Scheme 1 by assuming k 3 Ͼ Ͼ k 2 . In this model, k 2 reflects the nucleophilic attack in concert with the C-N bond cleavage, the latter being triggered by a metal-nitrogen interaction. We therefore conclude that the reduced k 2 values can also be accounted for by a less efficient interaction with the nitrogen atom of the substrate. Another possibility that we cannot fully discard at this point is that the metal site is also involved in nucleophile activation. Nevertheless, QM/MM calculations on B2 enzymes do not support this hypothesis (36).
We have recently shown that second shell mutations in the B1 lactamase BcII obtained by in vitro evolution are able to fine tune the position of the metal ion in the DCH site of this enzyme, stabilizing the nitrocefin intermediate, which is not accumulated in hydrolysis mediated by wild type BcII (39).
Here we observe the same phenomenon; subtle changes (such as those that may be induced by metal substitution) determine the stability of this intermediate.
Based on these observations, we propose a ␤-lactam hydrolysis mechanism for GOB, which does not require a metal-activated nucleophile (Fig. 4). Instead, the role of the metal ion is to steer substrate binding and to provide electrostatic stabilization of the anionic intermediate. This mechanism is in agreement with the proposal that BcII could be active as a mononuclear enzyme with an empty 3H site (67) (i.e. with a requirement of a Zn(II) ion for intermediate stabilization rather than for nucleophile activation (58)), which is formed by shifting of an equilibrium between the two sites (52) upon substrate binding (67).
This conclusion suggests that the positioning of the metal ion at the DCH/DHH site is crucial to define a catalytically active lactamase. This is in excellent agreement with the finding that engineering a more buried position for this metal center seriously impairs the lactamase activity (68), whereas rendering a more substrate-accessible Zn(II) at this site enhances the catalytic efficiency (39). Although nitrocefin is not a clinically useful antibiotic, its usefulness as a probe of the catalytic mechanism of M␤Ls should not be underestimated. It has been shown recently that carbapenem hydrolysis by a B1 M␤L also proceeds by means of an anionic intermediate, after C-N bond cleavage (58). This evidence points to a general role of the Zn(II) ion in the DCH/DHH site of M␤L from all subclasses, which may be targeted as a common mechanistic element for inhibitor design.