Mono- and Binuclear Zn-b-Lactamase ROLE OF THE CONSERVED CYSTEINE IN THE CATALYTIC MECHANISM*

From ‡Fachrichtung 12.4 Biochemie, Universitaet des Saarlandes, D-66041 Saarbruecken, Germany, §Centre d’Ingéniérie des Protéines, Institut de Chimie B6, Université de Liège, Sart-Tilman, B-4000 Liège, Belgium, iEMBL-Outstation Hamburg at DESY, Notkestrasse 85, D-22603 Hamburg, Germany, ¶Department of Physics, The Royal Veterinary and Agricultural University, Dk-1871 Frederiksberg C, Denmark, and the **Dipartimento di Biologia Moleculare, Sezione di Microbiologia, Universita di Siena, Siena 53100, Italy

Zn 2ϩ -␤-lactamases catalyze the hydrolysis of ␤-lactam antibiotics by cleaving their ␤-lactam rings. The production of Zn 2ϩ -␤-lactamases most often renders bacteria resistant to almost all ␤-lactam drugs so far designed, including carbapenems. Some of these organisms like Bacteroides fragilis, Serratia marcescens, Stenotrophomonas maltophilia, Pseudomonas aeruginosa and Aeromonas hydrophilia are human pathogens (1), and the search for useful inhibitors for clinical purposes has become of major importance.
The structures of Zn 2ϩ -␤-lactamases from Bacillus cereus strain 569/H/9 and B. fragilis have been solved by x-ray crys-tallography (2)(3)(4). Both enzymes contain two metal-binding sites. The zinc ligands are His-86, His-88, and His-149 at the first site (the "three His" site) and those of His-210, Asp-90, and Cys-168 at the second, Cys, site. These residues are highly conserved in almost all the enzymes of the family for which sequence data are available. The first crystal structure of the B. cereus enzyme, solved at pH 5.6 and 293 K, showed one zinc ion in the first site (2) but that of the B. fragilis enzyme highlighted an oxygen-bridged two-zinc center (3), a result in agreement with the observation that the latter enzyme binds two zinc ions with dissociation constants below 10 M and reaches its maximum activity when two zinc ions are bound (5). Earlier studies of the B. cereus enzyme suggested a much weaker binding of a second equivalent of zinc with marginal effects on the activity (6 -7), but further crystallographic studies, performed at 100 K revealed a fully occupied second site (4). The crystallographic data which indicate that Cys-168 is not involved in Zn 2ϩ coordination at the high affinity site are apparently in contradiction with spectroscopic studies on the B. cereus Co 2ϩ and Cd 2ϩ derivatives that suggest sulfur ligation at the first site (8).
Despite the different pH conditions used in the crystallographic and biochemical studies, the B. cereus and B. fragilis enzymes have been hypothesized to be mono-and binuclear Zn 2ϩ enzymes, respectively.
The present report investigates this problem for the B. cereus Zn 2ϩ -␤-lactamase and analyzes the catalytic mechanisms of the mono-and binuclear Zn 2ϩ enzymes. The results indicate that the conserved Cys-168 is essential for the activity of the mono-Zn 2ϩ species but not for the binuclear enzyme. We further present EXAFS 1 data that reconcile the crystallographic and spectroscopic results concerning Zn 2ϩ ligation.

EXPERIMENTAL PROCEDURES
Site-directed Mutagenesis-The C168A mutant of the B. cereus 569/ H/9 Zn 2ϩ -␤-lactamase was constructed by PCR. Two partially overlapping fragments were amplified using the following primers: 5ЈGCGT-CCTCGAGAAAGGGTTGATGACATGAA3Ј(␣) plus 5ЈGATTTCACTAA-AGAGCCTCCAACTAA3Ј; and 5ЈAGTTGGAGGCTCTTTAGTGAAAT-C3Ј plus 5ЈGCGGCTCTAGACGTAATCAACAGATTCAGCAT3Ј (␤). The PCR fragments were gel-purified and combined by overlap PCR in a total volume of 100 l using 10 ng of each fragment, 100 ng of each oligonucleotide ␣ and ␤, 2 units of Goldstar polymerase, 1.5 mM MgCl 2 , 200 M dNTPs, 50 pmol of primer, and 1 ng of pRTWH012. The corresponding amplimer was digested with PstI and ClaI restriction enzymes. The 0.24-kilobase pair fragment was introduced in pET-BcII plasmid 2 to yield pET-BcIICA. Finally, the gene coding for the mature form of the Zn 2ϩ -␤-lactamase wild-type and C168A mutant were introduced by PCR into the pTrxFus plasmid after the gene coding for thioredoxin. Two unique restriction sites (KpnI and BamHI) were introduced before the gene segment coding for the mature form of the ␤-lactamase and after the STOP codon, respectively. The primers were 5ЈCACAATTTCTTCTGTACAGGTACCACAAAAGGTAGAGAAAAC3Ј and 5ЈCCCGGGATCCTTAAATATAGTTAGAAGAAAGAGAGGAGAA-3Ј. 25 ng of pRTWHO12 (for the wild-type) and pETBcIICA (for the C168A mutant) were used as templates. Reaction conditions were 4 min at 95°C, 30 times (30 s at 95°C, 1 min at 55°C, and 1 min at 72°C). The KpnI-BamHI PCR fragment was cloned into pTrxfus in order to create pCIP32 (wild-type) and pCIP33 (C168A mutant). The gene was then completely sequenced with the help of an automated laser fluorescent DNA sequencer (Amersham Pharmacia Biotech) to verify that no unwanted mutation had been introduced during the mutagenesis process.
Purification of the Enzymes-The wild-type and C168A mutant from B. cereus, strain 569/H/9, was produced by introduction of pCIP32 and pCIP33, respectively, in Escherichia coli GI724. The bacteria were grown at 30°C in 1 liter of induction medium (Invitrogen, San Diego) containing 100 g/ml ampicillin as selection agent. At an A 550 of 0.5, tryptophan (100 g/ml final concentration) was added, and the culture was further grown for 120 min. The bacteria were harvested after centrifugation of the culture at 5,000 ϫ g during 15 min. The pellet was resuspended in 100 ml of 10 mM sodium cacodylate buffer, pH 6.5. The cells were broken with the help of a cell disintegrator (Series Z, Constant System, Warwick, UK). After centrifugation at 20,000 ϫ g during 30 min, the supernatant was collected and loaded on a SP Sepharose column (2.5 ϫ 30 cm, Amersham Pharmacia Biotech, Uppsala, Sweden) pre-equilibrated in 10 mM sodium cacodylate, pH 6.5. The hybrid protein was eluted at a rate of 5 ml/min by a linear salt gradient (0 -0.6 M NaCl) in 10 mM sodium cacodylate, pH 6.5. The active fractions were concentrated to 5 ml by ultrafiltration and were dialyzed overnight against 50 mM Tris-HCl, pH 8.0, 1 mM CaCl 2 , 0.1% Tween 20. Enterokinase (0.1 unit/20 g of hybrid protein) was added, and the reaction mixture was incubated at 37°C for 16 h. The solution was loaded on a MonoS column pre-equilibrated in 10 mM sodium cacodylate, pH 6.5. The ␤-lactamase was eluted by a linear salt gradient (0 -0.5 M) in 10 mM sodium cacodylate, pH 6.5. The active fractions were concentrated to 1 mg/ml as determined by the absorption at 280 nm. The mutant protein was characterized as follows. 1) The mass spectrum, obtained by electrospray mass spectrometry, gave a molar mass of 25,040 Ϯ 10 g/mol (the theoretical value is 25,037 g/mol). 2) The N-terminal sequence (6 residues) was identical to that of the WT enzyme. 3) The CD spectra, both in the far and near UV, were superimposable on those of the WT enzyme. 4) The melting temperature, determined according to the modification of the fluorescence spectrum, was identical to that of the WT enzyme (67 Ϯ 0.1°C) as was the guanidinium chloride concentration resulting in 50% denaturation.
Metal Content Analysis and Preparation of Apoenzymes-To determine the Zn 2ϩ content under various conditions, 0.35-0.5-ml samples of the different enzymes at a concentration of either 30 or 50 M were dialyzed against 100 ml of the specified buffers containing different concentrations of Zn 2ϩ at 4°C. The protein concentrations were determined after dialysis by measuring the absorbance at 281 nm for the various B. cereus enzymes using the following extinction coefficients determined by five different methods including the determination of total amino acid content: 32,700 M Ϫ1 cm Ϫ1 for B. cereus 5/B/6; 30,500 M Ϫ1 cm Ϫ1 for the WT B. cereus 569/H/9; and 31,000 M Ϫ1 cm Ϫ1 for the B. cereus 569/H/9 C168A mutant. The values are accurate to 5% and were used in all protein concentration determinations. Zinc concentrations in samples and in the final dialysis buffers were measured with a Perkin-Elmer 2100 AAS spectrometer in the flame mode or by inductively coupled plasma mass spectroscopy as described by Hernandez Valladares et al. (10).
To produce "metal-free" buffers, buffer solutions were purified by extensive stirring with 0.2-0.5% (v/v) of iminodiacetic acid-agarose (Affiland, Liège, Belgium). The residual Zn 2ϩ content of the buffers after treatment was approximately 20 nM (10). Standard precautions were taken when the experiments required metal-free conditions (11).
Apoenzymes from strain 5/B/6, 569/H/9 WT, and from the C168A mutant were prepared by dialysis of the corresponding enzymes against 2 changes of 20 mM sodium cacodylate buffer, pH 6.5, containing 1 M NaCl and 20 mM EDTA over 24 h under stirring. EDTA was removed from the resulting apoenzyme solution by 5-7 dialysis steps against the same buffer without metals. In all preparations the remaining zinc content did not exceed 5% as judged by AAS.
Equilibrium Dialysis with 65 Zn-65 Zn in 0.1 M HCl (15 mCi/mol) was from Amersham Pharmacia Biotech. Radioactive 65 Zn was measured with a Canberra detector model GC1018 connected to a PC via an amplifier and a Canberra ACCUSPEC MCA board.
The dissociation constant relevant for binding of the first zinc ion to the 5/B/6 B. cereus enzyme in substoichiometric amounts was determined by dialyzing a 20 -400-fold excess of the apoenzyme against radioactive 65 Zn. Under these conditions only the mononuclear species of the enzyme can be formed. The samples containing the apoenzyme were placed in a 200-l dialysis button (Hampton Research), sealed with a membrane tubing for dialysis and placed in a volume of 2.5 ml of 50 mM HEPES buffer, pH 7.5, in Linbro plate reservoirs for 1-3 days at room temperature under orbital shaking. The final concentration of the enzyme was between 0.15 and 2.75 M (with respect to the total volume of dialysis). The samples were dialyzed against 0.36 Ci of 65 Zn, mixed either with the apoenzyme in the button or in the 2.5-ml reservoir solution.
For studying the binding of zinc in stoichiometric amounts, the 5/B/6 apoenzyme was dialyzed at two different concentrations (6.75 and 14.1 M) against various concentrations of isotopically diluted 65 Zn.
Equilibrium Model for Zinc Binding-When two sites can bind metals, the following four microscopic equilibria describe metal binding as shown in Equations 1 and 2.
where M denotes zinc ions, E the enzyme, EM the enzyme with zinc bound in the three His site, ME the enzyme with zinc bound in the Cys site, and MEM the enzyme with zinc ions bound to both sites. In equilibrium dialysis one cannot differentiate between binding to the two sites. Instead macroscopic equilibrium constants are derived. Under substoichiometric (no extra zinc besides 65 Zn) and stoichiometric conditions K mono ϭ 1/(1/K EM, E ϩ 1/K ME, E ) and K bi ϭ K MEM, EM ϩ K MEM, ME can be determined, respectively. Note that K EM, E K MEM, EM ϭ K ME, E K MEM, ME . For a given set of equilibrium constants, the five different equilibrium

concentrations [E], [M], [ME], [EM], and [MEM]
can be derived by solving the above equations numerically. From these concentrations one can form the ratio of protein-bound Zn 2ϩ to total Zn 2ϩ (substoichiometric conditions) or protein-bound Zn 2ϩ to total protein concentration (stoichiometric conditions). Such calculated ratios were compared with the experimentally determined ratios, and the dissociation constants K mono and K bi were derived by standard nonlinear least squares fitting. Kinetic Measurements and a Model for the Mechanism-Nitrocefin and benzylpenicillin were from Unipath (Oxford, UK) and Rhone Poulenc (Paris, France), respectively. The hydrolysis of substrates was followed by monitoring the change in absorbance with a Perkin-Elmer Lambda 2 UV/VIS spectrometer at 482 nm for nitrocefin and 235 nm for benzylpenicillin. K m(app) and k cat(app) values were obtained by the use of initial rates (the complete time courses of hydrolysis were used when the values within the uncertainties were identical to the values obtained with initial rates (12)). The reported k cat(app) and K m(app) values are the means of at least three single experiments in which the different enzymes were added to the substrate solutions prepared in buffers containing the stated Zn 2ϩ concentrations. All experiments were performed at 25°C in 25 mM HEPES, pH 7.5.
A steady state model in which k cat(app) and K m(app) contains contributions from both the mononuclear and binuclear Zn 2ϩ enzyme via k cat,1 , k cat,2 , K m,1 and K m,2 where the subscripts 1 and 2 refer to the mono-and binuclear Zn 2ϩ enzymes, respectively, is presented in Equations 3 and 4.
Steady State Model for the Mono and Binuclear Zn 2ϩ -␤-Lactamase-2 M. Galleni, unpublished observations. The kinetic data were analyzed according to the following steady state model involving catalysis by both the mono-and the binuclear Zn 2ϩ enzymes as follows.
where EZn 2ϩ , EZn 2ϩ 2 , EZn 2ϩ S, and EZn 2ϩ 2 S are the mononuclear and binuclear enzyme without and with bound substrate, respectively. In addition the binding of the second Zn 2ϩ ion to both EZn 2ϩ and EZn 2ϩ S is assumed to be in rapid equilibrium as shown in Equations 5 and 6. ϭ k cat,1 ͓EZnS͔ ϩ k cat,2 ͓EZn 2 S͔ (Eq. 7) where is the steady state velocity. The k cat and K m values for the mono-and binuclear zinc enzyme are k cat,1 and k cat,2 and K m,1 and K m,2 , respectively. Here K m,1 ϭ (k Ϫs,1 ϩ k cat,1 )/k s,1 and K m,2 ϭ (k Ϫs,2 ϩ k cat,2 )/k s,2 . Solving Equations 7-11 yields /E 0, k cat(app) , and K m (app) .
If k s,2 /k s,1 is equal to 1 the equation for K m(app) simplifies, and furthermore, a simple linear relation between K m(app) and k cat(app) can be derived by elimination of [Zn 2ϩ ] in Equations 13 and 14 for K m(app) and k cat(app) . We thus get brings the steady state model closer to the equilibrium situation, whereas a decrease of the same constant increases the differences between the two models and can result in significant deviations from linearity in the K m(app) versus k cat(app) plot. Standard nonlinear least squares fittings were applied in fitting the kinetic data to the model. EXAFS Spectroscopy-The EXAFS studies were performed with the enzyme produced by strain 5/B/6. The sample was prepared by dialysis of the native Zn 2ϩ enzyme against two changes of 25 mM Bis-Tris buffer, pH 6.5, containing 1 M ammonium acetate and 10 M Zn 2ϩ followed by an additional dialysis against the same buffer without Zn 2ϩ . The presence of a high ionic strength was necessary to avoid precipitation of the highly concentrated enzyme. After centrifugation the enzyme concentration was 710 Ϯ 35 M. The [Zn 2ϩ ]/[E] ratio was 1.2 Ϯ 0.1 as determined by AAS. The EXAFS data were collected at beamline D2 at the European Molecular Biology Laboratory Outstation Hamburg, and samples were measured as frozen solutions at 18 K in fluorescence mode (13). The energy resolution was better than 2.5 eV. The data were analyzed using the computer program packages EXPROG (14)   Analysis of data (Fig. 1) from equilibrium dialysis of the 5/B/6 enzyme against 65 Zn (no extra Zn 2ϩ added) gave the values for the equilibrium constants K mono shown in Table I. Stoichiometric binding of Zn 2ϩ ( 65 Zn) to the metal-free 5/B/6 B. cereus ␤-lactamase was also studied in equilibrium dialysis experiments with 14.1 M apoenzyme against 10, 20, 40, and 80 M Zn 2ϩ (no NaCl added) and with 6.75 M apoenzyme (1 M NaCl) against 20, 40, and 80 M Zn 2ϩ . For the fitting of these data, K mono was constrained to the value obtained under substoichiometric conditions. The results are shown in Table I. The dissociation constants K mono and K bi do not depend on the presence of NaCl within the experimental error and K bi is about 10 times larger than K mono .
For comparison with the crystallization conditions used by Carfi et al. (2), the Zn 2ϩ content of the enzyme from strain 5/B/6 was determined by AAS after dialyzing 0.35 ml of 50 M apoenzyme against 100 ml of 25 mM citrate buffer, pH 5.6, containing 1 M NaCl. The [Zn 2ϩ ]/[E] ratio was Ͻ0.1 without added Zn 2ϩ ; 0.6 and 0.8 with 13.3 and 62.5 M external Zn 2ϩ , respectively. As the [Zn 2ϩ ]/[E] ratio was less than 1 even at 62.5 M Zn 2ϩ K bi was ignored in analyzing the data. If a second zinc ion binds it does so very weakly. The fitted value of K mono is given in Table  I. Table II shows the kinetic parameters obtained for the 5/B/6 and 569/H/9 enzymes at different Zn 2ϩ concentrations. For the two WT enzymes the K m values do not significantly vary with the Zn 2ϩ concentration. Note that measurements in metal-depleted buffer (20 nM [Zn 2ϩ ] or less) with final enzyme concentrations far below the dissociation constant K mono results in k cat(app) values not significantly different from zero with benzylpenicillin as substrate in contrast to the results obtained with nitrocefin where the 5/B/6 enzyme exhibits full activity in metal-depleted buffer, even in the presence of 10 M EDTA in the assay buffer. For benzylpenicillin this could simply be explained by the release of Zn 2ϩ (fast on the kinetic time scale) and for nitrocefin by a strong increase in affinity for Zn 2ϩ upon binding of nitrocefin most likely by a decrease of the rate constant for the release of Zn 2ϩ . Additional evidence comes from the observation that the apoenzyme at a concentration of 36 nM is partly reactivated in the presence of nitrocefin prepared in metal-depleted buffer (20 nM [Zn 2ϩ ] or less). For the 5/B/6 enzyme with nitrocefin as substrate, k cat also appears to be independent of the Zn 2ϩ concentration (above 20 nM, Table  II). Thus, binding of a second Zn 2ϩ ion, if it occurs, has, in this case, no influence on the kinetic parameters. That only one Zn 2ϩ ion is necessary for full activity was confirmed by titrating the apoenzyme with increasing zinc concentration. The result demonstrates a virtual linear dependence upon zinc concentration up to the apoenzyme concentration of 10 M followed by a plateau (Fig. 2 (‚)). Fig. 2 also demonstrates that the specific activity versus Zn 2ϩ concentration for benzylpenicillin changes essentially in the submicromolar range of Zn 2ϩ , a result reflecting the formation of the mononuclear enzyme. Thus, as with nitrocefin, the mononuclear enzyme is likely to be fully active. No significant difference was observed if the reaction was started by adding the apoenzyme to the reaction mixture. The specific activity for the hydrolysis of benzylpenicillin for the 569/H/9 WT enzyme is shown in Fig. 3. Again as with the 5/B/6 enzyme, the major changes in specific activity occur at submicromolar values of Zn 2ϩ concentrations. However, both the specific activity and the k cat(app) values further increase above 1 M Zn 2ϩ . Kinetic parameters fitted using the model developed under "Experimental Procedures" also highlight a k cat,2 value about twice as high as k cat,1 (Table III). For the 569/H/9 WT enzyme with 150 M nitrocefin as substrate, similar rates with residual and 1 M Zn 2ϩ were observed but increased 2-fold upon further addition of Zn 2ϩ (Fig. 3). As the values of K m(app) are about 20 times lower than 150 M, the specific activity supplies a very good approximation of k cat(app) . Therefore the specific activity shown in Fig. 3 was fitted to Equation 13 for k cat(app) given under "Experimental Procedures." The results are given in Table III. The specific activity of the 569/H/9 enzyme versus 150 M nitrocefin in citrate buffer, pH 5.6, containing 100 M Zn 2ϩ is 46% that in HEPES buffer, pH 7.5, also containing 100 M Zn 2ϩ demonstrating that at this pH only the mononuclear enzyme is active.   Table I. The kinetic properties of the C168A mutant are completely different from that of the WT enzyme. The activity in the presence of residual Zn 2ϩ is negligible for both substrates and both the k cat(app) and K m(app) values for benzylpenicillin increase with increasing Zn 2ϩ concentration (Fig. 4). The kinetic parameters for the mononuclear species for the C168A mutant were derived by mixing 2 M apoenzyme with solutions con-taining 1.9 M Zn 2ϩ and different concentrations of benzylpenicillin and nitrocefin. The results are given in Table III. When fitting the values of k cat(app) for benzylpenicillin (Fig. 4) to the equation for k cat(app) , k cat,1 was fixed to 1.8 s Ϫ1 (Table III). At Zn 2ϩ concentrations above 1 M, the specific activity versus nitrocefin also increased (Fig. 5), and the shape of the curve was close to that of the k cat(app) for benzylpenicillin. In Fig. 6 the values of K m(app) are plotted versus k cat(app) (from Table II for the 569/H/9 enzyme and from Fig. 4 for the mutant). Fig. 6 demonstrates a linear dependence of K m(app) versus k cat(app) , and corresponding least squares fitting to straight lines gave the values for K m,1 and K m,2 presented in Table II.

Correlation between Zinc Concentration and Hydrolysis for
EXAFS Spectroscopy-The rigid structure of systems like imidazole is well known. Therefore, in EXAFS restrained re-  finement is applied to such problems (15). We first modeled the coordination sphere of zinc with three histidine residues and a water molecule as ligands as would be expected from the crystal structure data if zinc is not coordinated to the site with cysteine as a ligand. The interpretation of the extracted k 2 weighted fine structure by these assumptions result in fit 1 shown in Fig. 7, where the corresponding Fourier transform clearly indicates a missing contribution at about 2.25 Å. This can be accounted for by the contribution of a cysteine ligand (16). A fractional contribution of sulfur was also observed in EXAFS spectroscopy on the B. cereus, 569/H/9, enzyme at pH 6.0 also containing about 1 eq of Zn 2ϩ (17). However, the authors did not give any interpretation of the presence of this sulfur. From the amplitude it was obvious that, on average, less than one sulfur atom was present. To obtain an upper limit for the number of sulfur ligands, the corresponding Debye-Waller parameters were fixed, because of their strong correlation with coordination numbers in EXAFS spectra. The Debye-Waller parameter of the sulfur atom accounts only for dynamic disorder and the static disorder between all the enzyme units, whereas the Deby-Waller parameter for the nitrogen also bound to the central Zn 2ϩ additionally accounts for the static disorder within this unit (between the three imidazole ligands). Thus the Debye-Waller parameter for sulfur should be much smaller than for nitrogen. To estimate the upper limit for the presence of sulfur atoms, it was fixed to an even slightly lower value (0.003 Å 2 ). Analyzing the data with this model resulted in a significant improvement of the fit and a maximum coordina-  Table II  tion number for sulfur of 0.5 (Fig. 7). The corresponding parameters given in Table IV show that the improvement of the fit is only due to this sulfur contribution, because all other parameters were identical within their errors. The difference between the Fourier transforms of experiment and theory clearly indicated the absence of any further contribution above the noise level. The structures derived from x-ray diffraction data show that the Cys-168 residue is not close enough to coordinate the zinc at the 3-histidine site. As the EXAFS data show a fractional zinc coordination by sulfur, the only solution is a partial occupancy of the second site with cysteine, aspartate, and histidine as zinc ligands in the mononuclear species. However, because the Zn 2ϩ /enzyme ratio was 1.2 Ϯ 0.1 in the present case, part of the sulfur signal could also arise from a weakly occupied binuclear zinc enzyme.

Existence of a Mono-and a Binuclear Zinc Enzyme with
Different Kinetic Properties-Equilibrium dialysis in citrate buffer, pH 5.6, provided an estimation of K mono of 10 M for the 5/B/6 enzyme and no evidence for binding of a second Zn 2ϩ ion. The fact that the activity of the 569/H/9 enzyme at pH 5.6 versus nitrocefin is only about 50% that observed in HEPES buffer, pH 7.5 (when both contain 100 M Zn 2ϩ ), correlates well with the absence of a second enzyme-bound Zn 2ϩ ion. The preferential occupancy of the three-His site in the mononuclear species revealed in the crystal structure (Fig. 8) (2) is then satisfactorily explained by the much weaker binding of Zn 2ϩ to the Cys site at pH 5.6. However, already at pH 6.5 the EXAFS data indicate a significant occupancy of the Cys site in the so-called mononuclear species. This together with the observation of the binding of a second zinc at pH 7.5 with a weaker binding (Table I) is consistent with a dominant population of the three-His site together with a relatively lower population of the Cys site at pH values higher than or equal to 6.5 and at stoichiometries close to or below 1. The mononuclear zinc enzyme thus corresponds to a protein with only 1 zinc ion per molecule which could be either in the three-His site or the Cys site. Dialyzing the enzymes against a large concentration of zinc always results in the formation of a binuclear enzyme. In agreement with this, recent crystallographic studies of the 569/H/9 enzyme at pH 7.5 show a fully occupied second site. 3 The k cat value of the 569/H/9 enzyme increases 2-fold upon binding of the second Zn 2ϩ ion for both substrates (Table II, see also Fig. 3). From this alone, it is not possible to assign different mechanisms for the mononuclear and the binuclear zinc enzymes because half activity for an average of one zinc ion bound per protein molecule could equally well be explained by the coexistence of enzyme molecules with no zinc ions or enzyme molecules with two zinc ions. However, the data for the 5/B/6 enzyme changes this for two reasons. First, there is no increase in the k cat value with increasing zinc concentrations for nitrocefin (Table II). Second, the activity recovery curve with nitrocefin starting with the 5/B/6 apoenzyme shows unambiguously that maximum activity is obtained with only one zinc ion bound, i.e. no further increase in activity occurs upon formation of the binuclear enzyme (Fig. 2, see also Table II). This is further confirmed by the full activity of the 5/B/6 enzyme with the same substrate when no extra zinc is added (Table II). It is obvious that upon binding to the enzyme, nitrocefin (but not benzylpenicillin) increases the affinity for the first zinc ion as demonstrated by the activity with no extra zinc added (Table II)-With benzylpenicillin as substrate and the 5/B/6 enzyme, the data are also consistent with a 2-fold rise in k cat from the mononuclear zinc enzyme to the binuclear enzyme (Table II, see also Fig. 2) as for the 569/H/9 enzyme with both substrates. The conclusion is then that the kinetic properties of the mononuclear and the binuclear enzymes can differ according to the substrate and the enzyme. In a recent work (18) a similar conclusion was drawn for the Zn 2ϩ -␤lactamase from B. fragilis. Despite the differences observed in the kinetic parameters between the mono-and binuclear spe-  cies, a large proportion of activity persists for the mononuclear enzyme. Thus the formation of a binuclear enzyme is not necessary for efficient catalysis.
Note that the 5/B/6 and 569/H/9 enzymes differ only by 17 substitutions. Although the Thr-173 3 Ala and Ala-175 3 Ser substitutions in the 5/B/6 enzyme relative to the 569/H/9 enzyme are not far from the active site, they fail to explain, at the present time, the different values of the kinetic parameters as well as the different dependences on zinc concentrations of the two WT enzymes.
The Role of the Conserved Cysteine Residue-The kinetic analysis of the 569/H/9 WT enzyme and its C168A mutant shows that the affinity for benzylpenicillin is identical for the WT and the mutant for the mononuclear but different for the binuclear Zn 2ϩ enzyme, whereas hydrolysis of benzylpenicillin is strongly reduced for the mono-Zn 2ϩ mutant relative to its wild-type counterpart but not for the binuclear Zn 2ϩ species. With nitrocefin, it is also clear that the mutant activity is higher than that of the WT at high Zn 2ϩ concentrations (compare Figs. 3 and 5). This suggests that the mononuclear and binuclear Zn 2ϩ enzyme function via different mechanisms. Indeed, Cys-168 is essential for efficient hydrolysis by the mononuclear enzyme but not by the binuclear species.
As suggested by Concha et al. (3) the two zinc ions could be bridged by a shared hydroxyl which would attack the carbonyl carbon of the ␤-lactam ring, but the exact role of the Cys residue in the mononuclear enzyme remains to be elucidated.
Nevertheless, the crucial importance of Cys-168 in catalysis by the mononuclear Zn 2ϩ enzyme and its suggested irrelevance for hydrolysis by the binuclear enzyme is supported by the fact that the C168A mutant is able to bind two Zn 2ϩ ions at pH 7.5 with dissociation constants below 10 M. Cys-168 is thus not essential for binding of the second zinc ion. Interestingly the Pseudomonas maltophilia enzyme where the otherwise conserved cysteine residue is a serine residue also does bind two Zn 2ϩ ions but the third ligand of the second zinc ion is now a His side chain situated in a completely different part of the polypeptide chain (His-89) (19). The possible formation of the binuclear Zn 2ϩ enzyme may represent a kind of sophistication in an alternative mechanism that does not require Cys-168.
The present work shows that the catalytic mechanism of Zn 2ϩ enzymes requiring one metal ion for activity may become somewhat more efficient by acquisition of co-catalytic sites with two zinc ions in close proximity acting as a unit center (20). However, as shown by studies performed with the B. fragilis enzyme (18,21), the catalytic efficiency of the binuclear enzyme is only marginally superior to that of its mononuclear counterpart with some substrates and is even lower with other ones (18).