Purification, Characterization, and Kinetic Studies of a SolubleBacteroides fragilis Metallo-β-lactamase That Provides Multiple Antibiotic Resistance*

Resistance to multiple β-lactam antibiotics traced to the expression of Zn(II) requiring metallo-β-lactamases has emerged in clinical isolates of several bacterial strains includingBacteroides fragilis, a pathogen commonly found in suppurative/surgical infections. A soluble B. fragilismetallo-β-lactamase has been purified to homogeneity from the cell growth medium after expression as a secretory protein inEscherichia coli. The enzyme requires two tightly bound Zn(II) ions for full activity, and the Zn(II) ions can be removed by EDTA from the enzyme. The apoenzyme is reactivated by stoichiometric amounts of Zn(II) and Co(II) ions. The Co(II)-substituted enzyme exhibits a UV-visible spectrum characterized by strong Co(II) d-d transitions at 510, 548, 615, and 635 nm and an EPR spectrum withg values of 5.52, 4.25, and 2.01: features that serve as useful spectroscopic handles for the mechanistic studies of the enzyme. Although steady-state and transient-state kinetic studies of the soluble Zn(II) enzyme with nitrocefin as substrate found no ionizable groups with pK a values between 5.25 and 10.0 involved in catalysis, a kinetically significant proton transfer step in turnover was implicated by studies in deuterium oxide. These studies also detected the accumulation of an enzyme-bound intermediate and provide the basis for a minimal kinetic scheme describing metallo-β-lactamase-catalyzed nitrocefin hydrolysis.

The widespread use of antibiotics has put tremendous selective pressure on bacteria to devise mechanisms to escape the lethal action of the drugs. The most common and efficient mechanism for bacteria to subvert ␤-lactam antibiotics is to produce ␤-lactamases, a class of enzymes that inactivate the antibiotics by hydrolyzing their ␤-lactam rings (1). The majority of ␤-lactamases employ an active site serine as a nucleophile to attack the ␤-lactam carbonyl to form an acyl-enzyme intermediate during the hydrolysis (2)(3)(4). Resistance caused by these enzymes can be generally overcome by carbapenem antibiotics such as imipenem (5). However, a small group of ␤-lactamases, i.e. class B (6,7) or the functional Bush group 3 (8,9) enzymes that require divalent metal ions (most often Zn(II)) for activity, cannot be inhibited by any of the available ␤-lactamase inactivators. So far, these metallo-␤-lactamases have been identified in more than 10 bacterial strains including several pathogenic genera such as Aeromonas, Pseudomonas, Serratia, and Bacteroides. Their emergence represents yet another problem in the crisis spawned by antibiotic resistance.
The amino acid sequences of the metallo-␤-lactamase from Bacillus cereus (10,11), Bacteroides fragilis (12,13), Aeromonas hydrophila (14,15), Stenotrophomonas maltophilia (16), and Serratia marcescens (17) have been determined. Sequence alignment reveals a 17-34% sequence similarity between them (18,19). Although the properties of all the metallo-␤-lactamases have been characterized to some extent, detailed studies have only been carried out on the enzymes from B. cereus, B. fragilis, and A. hydrophila. The B. cereus enzyme, which was the first metallo-␤-lactamase to be identified, binds one Zn(II) ion with high affinity and a second with much lower affinity (20). The binding of the high affinity Zn(II) ion is responsible for almost all the enzymatic activity, whereas the binding of the second Zn(II) ion to the low affinity site merely increases the enzyme efficiency (21). The B. fragilis enzyme, on the other hand, has two high affinity Zn(II)-binding sites with the occupation of both sites required for maximum enzymatic activity (22). The x-ray crystal structures of the B. cereus and of the B. fragilis enzyme have been solved to 2.5-Å and 1.85-Å resolution, respectively (18,19). The overall structures of the two enzymes are very similar, comprising an unusual protein fold of a ␤␤ sandwich with helices on each external face and the active site, i.e. the Zn(II)-binding site, at the edge of the sandwich (18,19). In the structure of the B. cereus enzyme, only the high affinity Zn(II)-binding site is occupied with the Zn(II) ion coordinated by three histidine residues (residues 86, 88, and 149) and a water molecule (18). In the structure of the B. fragilis enzyme, both Zn(II)-binding sites are occupied (19). One Zn(II) ion is bound by three histidine residues (residues 86, 88, and 149); the second Zn(II) ion is coordinated by cysteine (residue 168), aspartic acid (residue 90), histidine (residue 210), and a water molecule (19). A water molecule/hydroxide ion bridges the two Zn(II) ions (19). His-88, His-149, His-210, and Asp-90 are conserved in all five of the known metallo-␤lactamase sequences, whereas His-86 and Cys-168 are conserved in four of the five sequences (18,19). Recently, two Zn(II) ions were also found in the A. hydrophila metallo-␤lactamase, although the presence of the second Zn(II) ion results in a loss of enzymatic activity (23). It seems that metallo-␤-lactamases from different sources have differing Zn(II) requirements, which may account for the wide range of catalytic efficiencies of different metallo-␤-lactamases (24).
B. fragilis is one of the most important pathogens in polymicrobial infections in humans (25)(26)(27). Recently, a clinical isolate of B. fragilis expressing a metallo-␤-lactamase proved resistant to multiple ␤-lactam antibiotics (28), underscoring the need for clinically useful inhibitors of the enzyme. One impor-tant step to aid in this task is to elucidate the catalytic mechanism of this enzyme. However, the B. fragilis metallo-␤-lactamase currently used in biochemical and structural studies (19,22,29) was produced as inclusion bodies in E. coli. The active enzyme is obtained by renaturing the 8 M urea-solubilized inclusion bodies in metal ion-containing buffer (Zn(II) or Co(II)). The preparation is subject to aggregation at high protein concentration and is not reliably reproducible. 1 The purified Co(II) enzyme failed to show the desired spectroscopic features of a Co(II) enzyme (22). Here, we report the subcloning, overexpression, and purification of a soluble B. fragilis metallo-␤-lactamase. Using the soluble enzyme, we were able to characterize the effects of Zn(II) on activity, to prepare a Co(II)-containing enzyme whose spectroscopic features further define the metal ion binding sites, and to establish the minimum kinetic sequence for the enzyme-catalyzed nitrocefin hydrolysis.

EXPERIMENTAL PROCEDURES
Materials-E. coli strain DH5␣ was obtained from Life Technologies, Inc. E. coli strain BL21(DE3) and plasmid pET-27b(ϩ) were from Novagen (Madison, WI). Plasmid pT7CcrANDE02 harboring the B. fragilis metallo-␤-lactamase gene (13,29) was provided by Dr. Beth A. Rasmussen of the American Cyanamid Co. (Pearl River, NY). Nitrocefin was a gift from SmithKline Beecham Pharmaceuticals (Harlow, United Kingdom). Metal-free buffers were prepared by treating the buffers with Chelex-100 resin according to the manufacturer's suggestions (Bio-Rad). Metal-free dialysis tubing was prepared as described by Auld (30). Spectroscopically pure ZnSO 4 and CoCl 2 in Milli Q water stock solutions were used as Zn(II) and Co(II) sources. Restriction endonucleases and other enzymes used in cloning and sequencing were from commercial sources as indicated. The renatured B. fragilis metallo-␤-lactamase was purified from the 8 M urea-solubilized inclusion bodies according to the procedures described previously (22). All other chemicals were analytical reagents or molecular biology grade from commercial suppliers.
Construction of the Overexpression Plasmid for Soluble B. fragilis Metallo-␤-lactamase-Two oligonucleotide primers, 5Ј-CTTTTCCCT-GTCGCAGCCATGGCACAG-3Ј (having a NcoI site) and 5Ј-GGAATAC-CGGGTAGGATCCTACAATTC-3Ј (having a BamHI site), were designed based on the deoxynucleotide sequences of a truncated gene encoding the mature B. fragilis metallo-␤-lactamase (amino acids 18 -249) (13) and synthesized on an EXPEDITE nucleic acid synthesizer (PerSpective Biosystems, Inc., Framingham, MA). The primers were used to amplify the truncated B. fragilis metallo-␤-lactamase gene from pT7CcrANDE02 by polymerase chain reaction. The polymerase chain reaction product was digested with NcoI and BamHI and subsequently ligated into the NcoI/BamHI site of the pET-27b(ϩ) plasmid. The ligation product was then transformed into E. coli DH5␣ cells by electroporation, and plasmid minipreps were made using the Promega Wizard kit (Madison, WI). The sequences of the miniprep plasmids were checked using the ABI PRISM Dye Terminator Cycle Sequencing Ready Reaction system (Applied Biosystems, Inc., Foster City, CA) at the Penn State Nucleic Acid Facility. The one containing the correct truncated metallo-␤-lactamase sequence, designated as pMSZ02, was subjected to a large scale plasmid preparation using the Qiagen plasmid purification kit from Qiagen, Inc. (Santa Clarita, CA) and used as the overexpression plasmid for the soluble B. fragilis metallo-␤-lactamase.
Overexpression and Purification of the Soluble B. fragilis Metallo-␤lactamase-The overexpression plasmid, pMSZ02, was first transformed into E. coli BL21(DE3) cells by electroporation. A single colony from the transformed E. coli BL21(DE3, pMSZ02) plate was then used to inoculate 50 ml of LB medium containing 25 g/ml kanamycin. After the cell preculture grew overnight at 37°C, portions were diluted 1:100 into four 1-liter volumes of fresh LB medium containing 25 g/ml kanamycin. The cell cultures were grown at 37°C to an OD 600nm of ϳ0. 6 and were subsequently cooled down in a 25°C water bath for 15 min. The production of the metallo-␤-lactamase was induced by 1 mM isopropyl-1-thio-␤-D-galactopyranoside, and the induced cell cultures were shaken at 25°C for 18 -20 h.
The following procedures were carried out at 4°C. Cells were removed by centrifugation (7000 rpm for 20 min), and the 4 liters of supernatant were concentrated to about 15 ml using two Minitan concentrators equipped with two 10,000 NMWL membranes each (Millipore Corp., Bedford, MA) followed by an Amicon YM-10 membrane (Amicon, Inc., Beverly, MA). The resulting solution was dialyzed against 3ϫ l liter of 30 mM Tris-Cl, 100 M Zn(II), pH 7.6 buffer during a 36-h period and centrifuged at 18,000 rpm for 30 min to remove any precipitate. The supernatant was then loaded onto a 30 ϫ 200-mm Q-Sepharose column (Amersham Pharmacia Biotech, Uppsala, Sweden) that had been equilibrated with 30 mM Tris-Cl, 100 M Zn(II), pH 7.6 buffer. The column was eluted with a linear gradient formed by 500 ml of the starting buffer and 500 ml of 30 mM Tris-Cl, 100 M Zn(II), 400 mM NaCl, pH 7.6 buffer at a flow rate of 2 ml/min. Fractions containing the metallo-␤-lactamase were combined and concentrated to about 10 ml using an Amicon concentrator with an YM-10 membrane. The product from the Q-Sepharose column was subsequently loaded onto a 30 ϫ 700-mm Sephadex G-75 column (Amersham Pharmacia Biotech) that had been equilibrated with 30 mM Tris-Cl, 100 M Zn(II), pH 7.6 buffer for further purification. After loading, the column was eluted with the same buffer at a flow rate of 0.8 ml/min. The pure, active metallo-␤lactamase fractions were pooled and concentrated to a final volume of 6 ml with an Amicon YM-10 membrane. The protein solution was finally dialyzed against 4ϫ 1 liter of metal-free, 50 mM HEPES, pH 7.6 buffer in a 96-h period to remove any adventitious metal ions.
The purity of the enzyme was ascertained by SDS-polyacrylamide gel electrophoresis, and the protein concentration was determined by using the Bio-Rad protein assay kit (Bradford method) and/or ⑀ 280nm ϭ 39,000 M Ϫ1 ⅐cm Ϫ1 . The metal content of the protein was determined using a Perkin-Elmer 730 atomic absorption spectrophotometer in the flame mode at the Penn State Materials Characterization Laboratory, and the reported values are the average of readings from at least three enzyme preps. Enzyme activity was determined in 1ϫ MTEN buffer (50 mM Mes, 2 25 mM Tris, 25 mM ethanolamine, and 100 mM NaCl) at pH 7.0 and 25°C using nitrocefin as substrate; the formation of the hydrolyzed nitrocefin was monitored and quantitated using ⌬⑀ 485nm ϭ 15,600 M Ϫ1 ⅐cm Ϫ1 (31). The purified enzyme can be stored at 4°C for a week without losing any activity. It can also be stored in aliquots at Ϫ70°C for a long period of time after flash freezing in liquid nitrogen.
Preparation and Characterization of Apo-and Metal-substituted Metallo-␤-lactamase-Apo-metallo-␤-lactamase was prepared by dialyzing the Zn(II)-␤-lactamase solution containing 0.3-0.5 mM protein against three changes of 150-fold volume excess of 50 mM HEPES, 10 mM EDTA, pH 7.6 buffer at 4°C with a duration of 12 h for each dialysis. The chelating agent was then removed by extensive dialysis against eight changes of 150-fold volume excess of metal-free, 50 mM HEPES, pH 7.6 buffer at 4°C. The obtained apoenzyme was subjected to metal analysis to ensure that it was metal-free. The apo-␤-lactamase can be stored at Ϫ70°C after flash freezing in liquid nitrogen. Metal reconstitution studies were carried out by adding aliquots of the corresponding metal ion stock solution directly into the apoenzyme solution, mixing quickly, and incubating on ice for 30 min before assay. The enzymatic activities of the reconstituted enzyme were determined using nitrocefin as substrate (see above). The formation of Co(II)-substituted ␤-lactamase was also monitored spectroscopically. For the atomic absorption determination of the metal content of the Co(II)-reconstituted enzyme, the samples were made by adding a stoichiometric excess of Co(II) stock to the apoenzyme followed by dialysis against metal-free buffer to remove the unbound metal ion.
EPR spectra of the Co(II)-substituted enzyme were collected on a Bruker ESR 300E spectrometer equipped with an Oxford ESR 900 continuous-flow cryostat and an Oxford model ITC4 temperature controller. Operating temperatures were read directly from the controller, which was calibrated with a carbon glass sensor. The buffer-subtracted EPR spectra were quantitated by double-integration of signal-averaged scans using a 0.5 mM CoCl 2 standard solution as reference.
Kinetic Studies of the Soluble Zn(II)-Metallo-␤-lactamase-Steadystate kinetic studies of the Zn(II)-␤-lactamase-catalyzed hydrolysis of nitrocefin were conducted in 1ϫ MTEN buffer from pH 5.25 to 10.0 at 25°C (see above). The K m and k cat values were derived from initial rate measurements using program Curve Fit 0.7e. The reaction rates at higher pH (Ն7.5) were routinely corrected for background hydrolysis of the substrate. For solvent isotope effects, all the reactions were run in 1ϫ MTEN buffers prepared in D 2 O with their pD values adjusted with KOD or DCl. Single-turnover experiments were carried out at pH 5.5, 7.0, and pD 7.0 on a SX.17MW stopped-flow system (Applied Photo-physics, Leatherhead, United Kingdom) at 25°C. The substrate disappearance and product formation were monitored as absorbance decrease at 390 nm and increase at 485 nm, respectively (31). The absorbance changes were then corrected for an instrument dead time of 1.5 ms and converted into concentrations (32). The kinetic mechanism was simulated using the program KINSIM (Carl Frieden and Bruce Barshop, Washington University, St. Louis, MO) (33).

RESULTS
Overexpression and Purification-B. fragilis metallo-␤-lactamase is a secretory protein with its leader sequence comprising amino acid residues 1-17 (13,29). Our plan was to subclone the leaderless B. fragilis metallo-␤-lactamase gene (encoding amino acids 18 -249) into plasmid pET-27b(ϩ) so that the truncated gene is fused with a pelB leader sequence under control of a bacteriophage T7 promoter. As a result, the overexpressed metallo-␤-lactamase has the potential to be secreted into the periplasm in a soluble form. The construction of the expression plasmid, pMSZ02, is quite straightforward (Scheme 1). To overexpress the metallo-␤-lactamase, we first grew the pMSZ02transformed E. coli BL21(DE3) strain in rich medium to the midlog phase (OD 600nm ϭ 0.6) at 37°C. The cell culture was then cooled to 25°C and induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside. The induction was performed at 25°C to avoid cell lysis (34). However, very little soluble protein was found in the periplasmic space after induction; most appeared in the growth medium. The production of the soluble protein reaches a maximum after 18 -20 h of induction and accounts for more than 90% of the total protein in the medium (Fig. 1). The ␤-lactamase activity in the cell growth medium was monitored during the induction and found proportional to the level of the soluble enzyme produced (data not shown). The fact that the ␤-lactamase activity in the medium was inhibited by EDTA in a time-dependent fashion (data not shown) demonstrates that the soluble enzyme is indeed a metallo-␤-lactamase. During the expression, large amounts of insoluble metallo-␤-lactamase also accumulated in the cytoplasm.
The secretion of the enzyme into cell growth medium makes its purification relatively easy (Fig. 1). Minitan concentrators proved very efficient in concentrating the medium; four 10,000 NMWL membranes concentrated 4 liters of medium to 50 ml in about 6 h. After further concentration with an Amicon YM-10 membrane, the protein solution was dialyzed to adjust pH and to lower the salt concentration. A Q-Sepharose column was used to remove most of the positively charged components in the medium and the metallo-␤-lactamase-containing fractions eluted using a linear salt gradient between 0.15 and 0.2 M NaCl. Highly purified metallo-␤-lactamase was obtained after a final Sephadex G-75 gel filtration column (Fig. 1). We have consistently isolated 80 -100 mg of purified enzyme from 4 liters of cell culture.
Properties of the Soluble Metallo-␤-lactamase-As expected, the purified soluble metallo-␤-lactamase has a molecular mass of ϳ26 kDa (Fig. 1). Atomic absorption measurements revealed that the soluble metallo-␤-lactamase contained two Zn(II) ions per protein molecule (Table I), and no other metal ions were found in significant amounts (data not shown). With nitrocefin as the substrate, the soluble metallo-␤-lactamase exhibited a K m value similar to that of the renatured enzyme that had been purified from urea-solubilized inclusion bodies (Table I). However, the specific activity of the soluble enzyme is much higher than that of the renatured enzyme; the k cat value of the soluble enzyme is higher than 200 s Ϫ1 , whereas our best k cat value for the renatured enzyme is only 160 s Ϫ1 at pH 7.0 (Table I).
Like the renatured enzyme (22), the soluble metallo-␤-lactamase can be inactivated by Zn(II) chelating agents such as EDTA and 1,10-phenanthroline in a time-dependent manner. Attempted preparation of the apoenzyme by dialyzing the purified Zn(II)-␤-lactamase against 2 mM 1,10-phenanthrolinecontaining buffer at pH 7.6 resulted in the precipitation of the protein. However, dialysis of the Zn(II)-␤-lactamase against 10 mM EDTA-containing buffer followed by extensive dialysis against metal-free buffer to remove EDTA provided the apoenzyme with almost no activity (Fig. 2). When increasing Zn(II) concentrations were added to the apoenzyme, the initial hydrolysis rate of nitrocefin markedly increased. About 80% of the enzymatic activity was restored after 1 equivalent of Zn(II) was introduced and full enzymatic activity recovered after 2 equivalents of Zn(II) were added to the apoenzyme (Fig. 2). This result reveals an important difference between the soluble enzyme and the renatured enzyme; the renatured enzyme cannot be reactivated by any metal ions after being treated with chelating agents (22).
Reconstitution of the Apoenzyme with Other Metal Ions-Co(II) can reactivate the apoenzyme with maximal activity reached at two equivalents (Fig. 2). However, the Co(II)-substituted enzyme possesses only half of the activity of the Zn(II) enzyme ( Fig. 2 and Table I). Slight activity, approximately 4% and 8%, was restored with Cd(II) and Mn(II), respectively, but no activity was recovered by adding Ni(II) and Cu(II) to the apoprotein. Metal analysis confirmed the stoichiometric incorporation of Cd(II) and Mn(II) into the apoenzyme, whereas no bound metal ions were found in the Ni(II)-and Cu(II)-treated protein. Spectrophotometric titration of the apoenzyme with increments of Co(II), accompanied by scanning of the absorbance spectrum, confirmed that there are two Co(II) binding sites in the enzyme (Fig. 3). The addition of one equivalent of Co(II) resulted in most of the spectral changes, whereas the addition of the second equivalent of Co(II) completed the spectral changes. The UV-visible difference spectrum of the soluble Co(II) enzyme (Fig. 4, spectrum of soluble Co(II) enzyme minus the spectrum of the apoenzyme) has five features: an intense S-to-Co(II) ligand to metal charge transfer transition (LMCT) band at 340 nm (⑀ ϭ 820 M Ϫ1 ⅐cm Ϫ1 ) and four Co(II) d-d transition bands at 510, 548, 615, and 635 nm (⑀ ϭ 150, 230, 210, and 190 M Ϫ1 ⅐cm Ϫ1 , respectively) (35,36). This is very different from the UV-visible spectrum of the renatured enzyme, which failed to show the characteristic features of the Co(II) d-d transitions (22).
The low temperature EPR spectrum of the Co(II)-substituted ␤-lactamase is similar to the one obtained for the renatured enzyme (22). It is characterized by signals with apparent g values of 5.52, 4.25, and 2.01 (Fig. 5), typical of S ϭ 3/2 high spin, mononuclear Co(II) species. The EPR signal is integrated to 1.7 mol of Co(II)/mol of enzyme using a CoCl 2 sample as standard. This indicates that there are two Co(II)-binding sites in the protein and the two Co(II) ions are both high spin and do not interact with each other to form a magnetically coupled dinuclear site. The fact that the signal was temperature-dependent and disappeared at temperatures higher than 50 K (data not shown) supports the above conclusion.
Kinetic Studies of the Soluble Zn(II)-␤-Lactamase-In order to obtain accurate kinetic parameters, 1ϫ MTEN buffer, a three-buffer system known to have constant ionic strength through a wide pH range (37), was chosen for the kinetic studies. The K m and k cat values of the Zn(II)-␤-lactamasecatalyzed nitrocefin hydrolysis were determined at different pH values. The pH dependence of the logarithm of k cat and k cat /K m of the soluble Zn(II)-␤-lactamase is presented graphically in Fig. 6. Both curves are rather flat between pH 5.25 and 10.0. As the enzyme loses activity rapidly at low pH and the background hydrolysis of nitrocefin is too fast at high pH, we were unable to perform studies at pH values under 5.25 and above 10.0. It therefore appears that there are no ionizable groups with pK a values between 5.25 and 10.0 involved in the catalysis.
The pH dependence of k cat and k cat /K m for the soluble Zn(II)-␤-lactamase-catalyzed nitrocefin hydrolysis was also studied in D 2 O. Both k cat and K m values decreased when the reactions were carried out in D 2 O. The kinetic solvent isotope effect on k cat ranges from 2.35 to 2.79 and on k cat /K m is ϳ1.6 from pH/pD 6.0 to pH/pD 10.0 (Table II). These results are in accord with a proton transfer in a kinetically significant step of the reaction.  The catalytic activity of the soluble Zn(II)-␤-lactamase was also examined in detail by a stopped-flow method to determine whether any enzyme-bound intermediate could be detected. Fig. 7 shows the consumption of nitrocefin and the formation of the product during a single turnover experiment using 25 M Zn(II) enzyme and 5 M substrate at pH 5.5. The data fit well to a single exponential with no sign of a large burst or lag phase with rate constants of 900 s Ϫ1 for the consumption of the substrate and 150 s Ϫ1 for the formation of the product. The large difference between the two rates suggests that the consumption of the substrate and the formation of the product proceed through an intermediate that accumulates  Fig. 7 were simulated using the program KINSIM. The results of the best fit simulation are shown as solid lines in Fig. 7 and the kinetic constants used for the simulations are listed in Table  III. All three data sets can be well simulated by the same parameters, which in turn provide the calculated k cat and K m values (38) close to those obtained from the steady-state kinetic studies (Table IV).
The consumption of nitrocefin and the formation of the product during a single turnover experiment using 25 M Zn(II) enzyme and 5 M substrate were also studied at pH 7.0 and pD 7.0 (data not shown). The formation of an intermediate during the reaction was also evident under these conditions. Using the same kinetic mechanism (Scheme 2), the data were well simulated by the parameters listed in Table III. Therefore, Scheme 2 most likely represents the minimum kinetic sequence of the Zn(II)-␤-lactamase-catalyzed nitrocefin hydrolysis where breakdown of the intermediate is the rate-determining step.    Table  III.

DISCUSSION
By subcloning the gene encoding amino acids 18 -249 of the B. fragilis metallo-␤-lactamase in frame with the T7 pelB sequence in the pET-27b(ϩ) vector, we were able to overexpress the enzyme very efficiently in E. coli BL21(DE3). A portion of the overexpressed enzyme was subsequently processed and secreted into the cell growth medium to yield an active, soluble metallo-␤-lactamase lacking the putative signal sequence. The yield of the soluble enzyme is thus modest, but it accounts for more than 90% of the total protein in the medium (Fig. 1), simplifying purification. The purified enzyme binds 2 mol of Zn(II)/mol of protein and exhibits higher catalytic activity than the renatured enzyme, indicating that the soluble enzyme is properly folded. This is further supported by the fact that the bound Zn(II) ions can be removed to form apoenzyme and the apoenzyme reconstituted with Zn(II) to recover full enzymatic activity. The apoenzyme derived from the renatured enzyme can not be reconstituted (22). Recently, Toney and co-workers reported another system for expressing the soluble B. fragilis metallo-␤-lactamase in E. coli, but no information on whether the apoenzyme could be generated and the native enzyme then reconstituted was provided (39).
One powerful means to study Zn(II) enzymes is to substitute the Zn(II) ions with Co(II). Co(II)-substituted Zn(II) enzymes are usually active and exhibit rich spectroscopic features that are lacking in the native enzymes. However, the Co(II)-substituted enzyme made from the renatured B. fragilis metallo-␤lactamase failed to show any distinct features in the Co(II) d-d transition region in its UV-visible spectrum (22). The soluble enzyme provided Co(II)-substituted B. fragilis metallo-␤-lactamase suitable for spectroscopic studies. The enzyme reconstituted from Co(II) binds 2 mol of Co(II)/mol of enzyme, retains roughly half of the activity of the Zn(II) enzyme, and has an activity-Co(II) dependence very similar to that of the Zn(II) enzyme (Table I and Fig. 2). The UV-visible spectrum of the Co(II)-substituted enzyme has five distinct features: an intense S-to-Co(II) LMCT band at 340 nm and four Co(II) d-d transition bands at 510, 548, 615, and 635 nm (Fig. 4). The 340-nm LMCT band is due to the Cys-168 ligation to Co(II), as demonstrated in the x-ray crystal structure (19). The pattern and intensities of the d-d transition bands are in agreement with the existence of two distinct Co(II) binding sites with distorted tetrahedral and trigonal bipyramidal coordination geometries, respectively, in agreement with the crystal structure of the Zn(II) enzyme (19). The fact that all the UV-visible signals appear simultaneously upon the addition of Co(II) (Fig. 3) indicates that both sites have similar binding affinities for the metal ions. Despite the difference between their UV-visible spectra, the Co(II)-substituted renatured enzyme and soluble enzyme exhibit similar low temperature EPR spectra. The EPR spectrum suggests that the two Co(II) ions in the Co(II) proteins are both high spin and not spin-coupled.
The UV-visible spectrum of nitrocefin has an intense absorbance at 390 nm, which is shifted to 485 nm upon hydrolysis (31). The absence of spectral overlap between nitrocefin and its hydrolysis product makes it an ideal substrate for mechanistic studies of ␤-lactamases. The B. fragilis Zn(II)-␤-lactamasecatalyzed nitrocefin hydrolysis is rapid and irreversible. During nitrocefin hydrolysis, the reaction velocity can be determined from either the time-dependent decrease of the absorbance at 390 nm (substrate disappearance) or the timedependent increase of the absorbance at 485 nm (product formation). As both Zn(II) ions in the dinuclear Zn(II) center of the B. fragilis metallo-␤-lactamase are required for the maximum enzymatic activity, a possible mechanism for the breakdown of the ␤-lactam ring is that the carbonyl group activated by one Zn(II) is attacked by a nucleophile followed by breaking of the C-N bond. The fact that k cat and k cat /K m have no pH dependence suggests that the pK a value of the nucleophile is lower than 5.25 or higher than 10.0. A possible candidate for the nucleophile is believed to be the water molecule shared by both Zn(II) ions (19). The pK a value of Zn(II)-bound water is usually between 6 and 10. However, the combination of the hydrogen bond between the bridging water and the carboxylate of Asp-90, the dimetal coordination, and the dipositive effective charge of the Zn(II) site with three histidine ligands may perturb the pK a value of the bridging water to below 5.25 (40,41).
Stopped-flow measurements of single turnover that measured substrate disappearance and product formation were consistent with the accumulation of an enzyme-bound intermediate at pH 5.5, pH 7.0, and pD 7.0. From the simulated kinetic constants, the bindings of substrate and product are under diffusion control. The attack step to form EI designated as k 2 is at least seven-fold greater than its breakdown through step k 3 . The step for formation of the intermediate is pH-and D 2 Oinsensitive, whereas the slight pH dependence of k cat and K m (Table IV) is due primarily to variations in the rate of the breakdown step. In addition, the observed solvent isotope effects (Table II) provide evidence for the involvement of a proton transfer event in the rate-determining step, i.e. the breakdown of the intermediate.
On the basis of the kinetic sequence and chemical precedents, one can envision that, upon binding, the carbonyl group of the ␤-lactam ring is activated for cleavage through polarization by one Zn(II) ion followed by attack of the hydroxide SCHEME 2.
derived from the bridging water to form an acyl-enzyme intermediate. Collapse of this species may proceed through displacement by water of the cleaved lactam and regeneration of the bridged dinuclear active site. This latter step appears to have an associated proton transfer. Alternatively, the enzyme nucleophile may be derived from one of the side chain residues acting as ligand to the Zn(II) ions. Additional experiments, especially spectroscopic characterization of the intermediate are needed to more fully elucidate the catalytic mechanism. 3 The active site of the B. fragilis metallo-␤-lactamase possesses the two conserved functional groups, i.e. a water molecule bridging the two Zn(II) ions and a carboxylate group hydrogen-bonded to this water, found in all the dizinc peptidases (41). This structural feature facilitates the deprotonation of the bridging water molecule, transforming it into a potent nucleophile at neutral pH, and acts to activate the substrate by polarizing its carbonyl group. The Zn(II) ions may also contribute to the electrophilic stabilization of the acyl intermediate and the flanking transition states. A deeper understanding of the catalytic mechanism of the B. fragilis metallo-␤-lactamase will not only provide vital information for the development of clinically useful inhibitors, but will also generate insights into the catalytic mechanism of other bimetallopeptidases.