Zinc Ion-induced Domain Organization in Metallo-β-lactamases

The reversible unfolding of metallo-β-lactamase from Chryseobacterium meningosepticum (BlaB) by guanidinium hydrochloride is best described by a three-state model including folded, intermediate, and unfolded states. The transformation of the folded apoenzyme into the intermediate state requires only very low denaturant concentrations, in contrast to the Zn2-enzyme. Similarly, circular dichroism spectra of both BlaB and metallo-β-lactamase from Bacillus cereus 569/H/9 (BcII) display distinct differences between metal-free and Zn2-enzymes, indicating that the zinc ions affect the folding of the proteins, giving a larger α-helix content. To identify the regions of the protein involved in this zinc ion-induced change, a hydrogen deuterium exchange study with matrix-assisted laser desorption ionization tandem time of flight mass spectrometry on metal-free and Zn1- and Zn2-BcII was carried out. The region spanning the metal binding metallo-β-lactamases (MBL) superfamily consensus sequence His-X-His-X-Asp motif and the loop connecting the N- and C-terminal domains of the protein undergoes a zinc ion-dependent structural change between intrinsically disordered and ordered states. The inherent flexibility even appears to allow for the formation of metal ion-bridged protein-protein complexes which may account for both electrospray ionization-mass spectroscopy results obtained upon variation of the zinc/protein ratio and stoichiometry-dependent variations of 199mHg-perturbed angular correlation of γ-rays spectroscopic data. We suggest that this flexible “zinc arm” motif, present in all the MBL subclasses, is disordered in metal-free MBLs and may be involved in metal ion acquisition from zinc-carrying molecules different from MBL in an “activation on demand” regulation of enzyme activity.

Variable metal loading states of zinc proteins are attracting increasing interest in the field of cellular regulation processes being a key to the understanding of physiological functions of zinc sensors and metallothioneins as well as regulatory functions of zinc ions. The coexistence of zinc proteins in the metalloaded and the metal-free form, however, requires the regulation of "free" zinc ion concentrations in narrow limits, with nM to pM concentrations in eucaryotes (19) or even much lower concentration in procaryotes (20). The issue, however, that zinc enzymes in their natural environment might be regulated by reversible metal ion binding is infrequently considered.
The impact of metal ion binding on structure and stability of MBL superfamily proteins has been studied in some detail. Zinc was found to be required for the folding of glyoxalase II (21) and arylsulfatases (22) into the native state. For CphA from Aeromonas hydrophila, differential scanning calorimetry and fluorescence spectroscopy demonstrated that zinc binding stabilizes the protein against denaturation with urea. The inactive Zn 2 -CphA proved to be the most stabile species (23). Crystal structures of metal-free and metal-loaded BcII revealed minor structural changes in the active site of the protein (6). 1 H, 15 N heteronuclear single quantum coherence spectra of the backbones amides of BcII resulted in distinct signals for different metal ion/enzyme ratios which allowed discrimination of apoenzyme and metal-loaded states (24). Metal ion binding was considered to be essential for folding of L1 in vivo (25), and variable loading states were described in dependence of the bioavailability of various metal ions (26). We hypothesized earlier that metallo-␤-lactamases are most likely in the metal-free apoenzyme state in the absence of substrates, which is because of the moderate affinity of the enzymes for zinc ions and the very low concentration of free zinc in cellular environments (14). We suggested that substrate availability might induce a spontaneous self-activation by direct transfer of zinc via ligand exchange reactions with delivery systems as substrate presence leads to a drastic increase of zinc affinity. The suggested selfactivation mechanism, however, requires a direct interaction of the apoenzymes with other zinc carriers to allow a ligand exchange reaction to occur. Because such interactions might be considered as unspecific, it seemed reasonable to postulate a high structural flexibility to allow the transient formation of zinc-bridged complexes. To verify this prediction, we initiated an investigation on the role of zinc ions on folding and stabilization of MBLs with BlaB and BcII as test cases. Our results demonstrate that the systems, when unsaturated with metal ions, cannot be correctly described as being composed of variable fractions of the proteins in the loaded and unloaded state alone. We will present indications of the formation of labile, metal-bridged ternary complexes formed under such conditions. The latter may be considered as important intermediate states for metal ion transfer between identical or different zinc binding molecules in general.

EXPERIMENTAL PROCEDURES
Expression, Purification, Characterization, and Metal Depletion of Enzymes-BlaB from C. meningosepticum was produced and purified as described (27). The protein concentration was determined with ⑀ 280 nm (BlaB) ϭ 45670 M Ϫ1 cm Ϫ1 . Metal-free BlaB (apoBlaB) was produced by three dialysis steps of the enzyme (1.3 mg/ml) against a 150-fold volume excess of 30 mM sodium cacodylate, pH 6.5, containing 20 mM EDTA, 0.1 M NaCl (12 h each under stirring at 4°C). EDTA was removed by dialysis against a 150-fold volume excess of 30 mM sodium cacodylate, 1.0 M NaCl, pH 6.5 (three changes), followed by two changes of the same buffer without NaCl. The antibiotic nitrocefin (Unipath Oxoid, Basingstoke, UK) was used as a substrate for BlaB using ⌬⑀ 482 nm ϭ 15,000 M Ϫ1 cm Ϫ1 following hydrolysis. To minimize zinc ion contaminations, buffer solutions were prepared in bi-distilled water and extensively stirred with Chelex 100 (Sigma).
BcII was prepared as describe before (9) with some modifications. BcII was expressed from plasmid pEt9a/BcII in Escherichia coli BL21 (DE3). Cells were grown at 37°C in M9 minimal medium with 10 g/liter of glucose and 1 g/liter of NH 4 Cl. Expression was induced by adding 1 mM isopropyl 1-thio-␤-Dgalactopyranoside at an optical density of 0.6 at 600 nm. After 16 h, cells were harvested by centrifugation, re-suspended in MES buffer (10 mM, 1 mM ZnCl 2 , pH 6), and disrupted using a French press. Cytosolic proteins were separated from cell debris and purified chromatographically (9). Metal-free BcII (apoBcII) was prepared as described (14). Whereas BlaB was only studied as the metal-free or the fully zinc-loaded enzyme (obtained by adding excess of zinc), BcII was additionally studied when loaded with only 1 eq of metal ion. This, however, required a higher precision of the determined enzyme concentration. We used the published absorption coefficient (⑀ 280 nm (BcII) ϭ 30,500 M Ϫ1 cm Ϫ1 ) (9), results from titrations of the apoenzyme with metal ions (see below), and spectrophotometric determination of the thiol group of the single cysteine residue of the protein using 5,5Ј-dithiobis-(2-nitrobenzoic acid) (the assay contained 0.1 M Tris/HCl, 300 M 5,5Ј-dithiobis(nitrobenzoic acid), 1 mM EDTA, 1% SDS). The results from all methods agreed within 10% deviation. We decided to base our calculations on the free thiol concentration determined as the metal binding capacity depends on the presence of an intact active site. On the other hand, the method is sensitive enough to apply it to small protein concentrations and allows for routine controls (e.g. after storage, dialysis, or any other manipulation).
Circular dichroism spectra were recorded with a Jasco J740 at 20°C. To allow direct comparison with binding experiments, the CD spectra were recorded in 5 mM HEPES, pH 7.0, at a protein concentration of 5 M in cuvettes with a 1-mm light path.
Dissociation constants for binding of Zn 2ϩ to BcII were obtained from competition titrations with the chromophoric zinc chelator Mag-fura-2 (MF) (Molecular Probes, Eugene, OR). The experiments were performed in 15 mM HEPES, pH 7.0. Apparent dissociation constants for a first (K mono ) and a second metal ion bound (K bi ) were obtained. A detailed description of the method can be found in de Seny et al. (10).
The activity assay was composed of CB containing 50 M Zn(II), 98 M nitrocefin, and the preincubated protein-GdmCl mixture (final enzyme concentration, 5 nM Zn 2 -BlaB or 10 nM apoBlaB) and was carried out at 25°C. The protein-GdmCl mixtures were diluted into the Zn-containing CB and kept for 30 s to allow refolding and reconstitution of the di-zinc enzyme before the catalytic reaction was started by substrate addition. The data presented in Fig. 1 are shown as the percent of the initial rate value of that in the reference, i.e. the protein preincubated under the same conditions as the samples but in the absence of GdmCl. In the activity assay, the GdmCl concentration was less than 0.03 M because of dilution.
Analysis of the unfolding/folding was performed according to either a two-state (28) or a three-state model (29). A linear dependence of the free energy of unfolding on the GdmHCl concentrations ([D]) was assumed for both transitions (30) according to the empirical relationship (31) ⌬G ϭ ⌬G°Ϫ m [D], in which m is the rate of change of ⌬G°with [D], and ⌬G°is the standard free energy of denaturation in the absence of denaturant. The GdmHCl concentration at the midpoint of each of the unfolding transitions, C m, was determined from ⌬G°/m.
In a two-state model, the signal observed at any denaturant concentration is given by, where R is the gas constant, T is the absolute temperature, and ⌬S N and ⌬S U are the signals for the native and unfolded protein, respectively. In a three-state model the observed signal results from where ⌬S N , ⌬S II , and ⌬S U are the signals for the native, intermediate, and unfolded protein, respectively. ⌬G NI O and ⌬G IU O are the free energies for the N to I and I to U conversions, respectively. ⌬S N and ⌬S U were obtained with the linear extrapolation method of Santoro and Bolen (28) and kept fixed in the three-state model fit. Enzymatic activity (% of the reference) and center of fluorescence (⌺( i F i )/⌺F i calculated from the fluorescence intensities F i at wavelength i with step size 0.1 nm) were fitted to Equations 1 and 2. The intercepts and the slopes of the pre-and post-unfolding regimes were taken from fits of a two-state model to the data (Equation 1) and were kept fixed in the three-state model fit (Equation 2). A significant red shift after incubating the apoenzyme for 24 h at 4°C in the absence of GdmHCl (open circle in Fig. 1B) resulted in a large uncertainty about the pre-unfolding regime ⌬S N . Therefore, we decided to use the value obtained for the fresh, non-incubated apoenzyme to determine ⌬S N in the fit.
ESI-MS Study of Zinc Binding to BcII-Sample preparation, equipment, and methods used have been described before (32). Relative appearances of different species (apo, Zn 1 , Zn 2 species) were calculated from deconvoluted spectra taking the different charge states into account. Percentages given are based on the assumption that all the different species can be detected with the same probability.
Pepsin Digestion and Assignment of BcII Peptides by MALDI-ToF/ToF MS-15 M apoenzyme in 15 mM HEPES, pH 7 (10 l), was diluted 1:11 with 0.1% trifluoroacetic acid solution to decrease the pH to 2.3. Protein digestion was performed by adding 30 l of pepsin bead slurry (Pierce) (washed 4 times with 450 ml of 0.1% trifluoroacetic acid at 4°C before use) and incubated on ice for 10 min with occasional mixing. The resulting peptides were separated from the pepsin beads by centrifugation at 4°C. 10 l of peptide mixture were loaded on a Zip-TipC18 (Millipore Corp., Billerica, MA) for desalting, and the peptides were eluted with 1 l of ␣-cyano-4-hydroxycinnamic acid (CHCA) matrix solution (5 mg/ml CHCA in acetonitrile/ ethanol/ trifluoroacetic acid 20/80/0.1) onto the MALDI plate and dried under a stream of compressed air. We used a 4800 MALDI ToF/TOF TM Mass analyzer (Applied Biosystems, Darmstadt, Germany) equipped with a 200-Hz Nd:YAG-Laser ( ϭ 355 nm, 3-7-ns pulse width). MS data were acquired in the positive ion reflectron mode with 470-ns delayed extraction, accumulating 500 laser shots using the 4000 Series Explorer TM Remote Access Client software (Version 3.5.1). Tandem mass spectrometry (post-source decay with post-acceleration) was performed for the sequencing of all detected peptic peptides; no additional collision gas was used. For MS/MS measurements, the acceleration voltage was 8 kV, and 4000 laser shots were accumulated for each MS/MS spectrum.
Hydrogen Deuterium Exchange (HDX)-For HDX experiments, 1 l of 150 M apoBcII or metal-substituted BcII stock solution was incubated 1:10 with deuterated buffer (D 2 O, 15 mM HEPES, pD 7.4) at 22°C. Deuterium labeling times were between 50 and 5900 s. Each exchange reaction was stopped by the addition of 100 l of 0.1% trifluoroacetic acid on ice, decreasing the pH to 2.3. The digestion with pepsin and the analysis of resulting peptides were performed as described above for the non-deuterated digest. After spotting the deuterated samples on the MALDI target, each exchange experiment was immediately measured by MALDI-MS with less than a 1-min delay. All solutions and ZipTips were kept cold before use. The MALDI plate was kept at room temperature to prevent condensation of water on the plate. All experiments were repeated in triplicate.
The centroid mass of each isotope cluster was calculated using the MagTran software (33) by labeling the left side of the lowest deuterated peak and the right side of the highest deuterated peak. The percentage of deuterium in-exchange of amide groups (%D) at the end of the incubation time in D 2 O was determined using where m t is the observed centroid mass of the deuterated peptide for each in-exchange time, and m 0% corresponds to the non-deuterated mass of the corresponding peptide. Fully deuterated sample m 100% was prepared by incubating 10 l of pepsin-digested apoBcII (15 M) in deuterated buffer (D 2 O, 15 mM HEPES, pD 7.4) for 72 h at 22°C. During sample preparation and transfer to the MALDI target, back exchange of incorporated deuterium atoms to hydrogen cannot be completely suppressed. Therefore, it is essential to correct the experimental data for this back exchange by using the experimentally obtained centroid mass of fully deuterated peptides (m 100% ) after back exchange as the 100% value. The specific back exchange of each peptide was determined separately.
For data representation the experimentally obtained values were further processed. From the experimentally determined masses we calculated hypothetical values for 100% D 2 O in the incubation buffer. For the samples with substoichiometric metal ion, addition corrections for apoenzyme content were introduced.
Evaluation of HDX Kinetics-All the deuteration versus time curves obtained can be described by monoexponential curves. For almost all the peptide fragments investigated, however, a rapid HDX phase, which was completed within the dead time of our experiments, preceded the experimentally resolved kinetic traces. The amplitude of the rapid phase is considered as the dead time in-exchange (%D 0 ) in (Equation 4). Fitting of Equation 4 to the data resulted in %D 0 , the rate constant k, and the amplitude of the time-resolved in-exchange process (%D t ).
Perturbed Angular Correlation of ␥-Ray Spectroscopy-199m Hg-PAC spectroscopy was performed at the ISOLDE facility (CERN, Switzerland). Experimental details for the production of 199m Hg can be found in Iranzo et al. (34), and a detailed description of the perturbed angular correlation of ␥-rays (PAC) method can be found in Hemmingsen et al. (35). The experiments were performed at a temperature of 1 Ϯ 2°C, which was controlled by a Peltier element. The radioactive mercury was produced immediately before starting the experiments and was trapped in 150 l of ice at 77 K. The 199m Hg solution was mixed with nonradioactive mercury acetate, 200 mM BisTris, pH 6.8, and the protein. Finally, sucrose was added to produce a 60% w/w solution.
In the case of identical, static, and randomly oriented molecules, the perturbation function G 2 (t) is G 2 ͑t͒ ϭ a 0 ϩ a 1 cos͑ 1 t͒ ϩ a 2 cos͑ 2 t͒ ϩ a 3 cos͑ 31 t͒ (Eq. 5) with 1 , 2 , and 3 as the three difference frequencies between the three sublevels of the spin 5/2 state of the mercury nucleus. Note that 1 ϩ 2 ϭ 3 . Thus, the Fourier transform of G 2 (t) exhibits three frequencies for each nuclear quadrupole interaction (NQI). Each NQI was modeled by using a separate set of the parameters, Q , , ␦, 1/ c , and A. The parameter Q ( Q ϭ eQV zz /h, in which Q is the nuclear electric quadrupole moment, and V zz is the numerically largest eigenvalue of the electric field gradient tensor) is associated with the strength of the interac-tion between the surrounding electronic environment and the mercury nucleus; is the so-called asymmetry parameter, which is 0 in an axially symmetric complex and has a maximal value of 1; ␦ is the relative frequency spread; c is the rotational correlation time; A is the amplitude of the signal (35). In cases where more than a single NQI is present, the perturbation function is the sum of the different perturbation functions, where each NQI is weighted by its population (36).

RESULTS
Equilibrium Unfolding/Refolding Studies of BlaB-The reversible unfolding of BlaB in GdmHCl was followed by fluorescence spectroscopic measurements. Both apo-and Zn 2 -BlaB show emission maxima at 337 nm (Fig. 1A). The fluorescence intensity, however, is 30% lower for the apoenzyme, indicative of stronger quenching of tryptophan fluorescence. The emission maximum after 24 h of storage at 4°C in the absence of GdmHCl was unchanged for the holoenzyme but red-shifted to 343 nm for the apoenzyme (data not shown), which indicates partial unfolding and, thus, a lower stability of the protein in absence of bound zinc ions. Increasing GdmHCl caused significant changes in both the emission intensity and the red shift of the emission maximum, which occurred in two stages. In the range 1.3-2.0 M for Zn 2 -BlaB and 0.5-1.3 M for apoBlaB, the fluorescence intensity decreased to limiting values of 37 and 58% at 337 nm, respectively, and the emission maximum was red-shifted from 337 to 358 nm for both. A further increase of GdmHCl concentrations resulted in a re-increase of fluorescence intensity for both species, resulting in identical final spectra. Thus, the center of fluorescence may be used as an indicator of the degree of unfolding at various GdmHCl concentrations. Significant spectral changes occur with transitions at GdmHCl concentrations ϳ1.3 and ϳ0.5 M for Zn 2 -and apoBlaB, respectively, indicating a higher susceptibility of the latter to the denaturing agent. In Fig. 1, B and C, the spectral changes with changing GdmHCl concentrations are compared with the enzyme activities obtained under the respective conditions when diluting the preincubated enzyme samples into a reactivation buffer before starting the activity assay. Whereas the fluorescence data represent the unfolding process measured in the presence of GdmHCl, the activity data represent the refolding process measured in buffer containing almost no GdmHCl. The activity data indicate the presence of a non-reactivating intermediate. Whereas the unfolded state U of apoBlaB can undergo 100% refolding to the native state (Fig. 1B), the refolding of the U state of Zn 2 -BlaB only results in an ϳ30% reactivation (Fig. 1C).
Obviously the presence of zinc during the unfolding process influences the refolding step, although the reactivation buffer for both apo-and Zn 2 -BlaB contains 50 M added zinc. To further explore the Zn 2ϩ dependence, we repeated the reactivation experiments of Zn 2 -and apoBlaB preincubated at 6.4 M GdmHCl with zinc concentrations between 50 and 320 M in the reactivation buffer. No influence of increasing zinc concentrations was observed for Zn 2 -BlaB, but the reactivated fraction obtained for apoBcII dropped from initially Ͼ80% to a value below 40%, confirming that the degree of refolding does indeed depend on the Zn 2ϩ concentration. The GdmHCl concentrations yielding the lowest activity (Fig. 1, B and C) and fluores-cence intensity (Fig. 1A) for Zn 2 -and apoBlaB were 1.6 and 1.0 M, respectively.
The lack of reversibility of unfolding observed in the activity measurements indicates aggregation of the protein upon rena-turation. This phenomenon is at least partially zinc-dependent, as could be shown for the U to N transition of the apoenzyme and might appear upon dilution into the zinc-containing activity buffer. If GdmHCl is removed from metal-free BlaB by dialysis, full reversibility of the unfolding process is observed over the whole concentration range of GdmHCl, which could be followed by activity and fluorescence measurements (data not shown). Based on the latter finding, we applied the thermodynamic analysis of the unfolding data, which presupposes reversibility of the unfolding processes. Because similar dialysis experiment with the zinc enzyme did not result in complete reactivation, the respective thermodynamic analysis might be compromised. Interestingly the zinc enzyme required 24 h of incubation with GdmHCl to fully observe the spectroscopic changes, whereas the apoenzyme did show 100% of changes already right after mixing with the denaturing agent. The thermodynamic analysis of the unfolding data required a threestate model of folded, partially unfolded, and unfolded states (Equation 2). The corresponding fits to fluorescence and activity data are shown in Fig. 1, B and C, and the resulting parameters are combined in Table 1. The total stability of the folded state is given by the sum of ⌬G°values for both steps, resulting in 7 Ϯ 1 and 14.6 Ϯ 0.5 kcal/mol for apo-and Zn 2 -BlaB, respectively.

Nano-ESI-MS Investigation of Putative Zinc Loading States of BcII at Variable Zinc/Protein Stoichiometries-ESI-MS spectra
of apoBcII under denaturing conditions (methanol, 0.2% formic acid) showed typical high charge distribution states (ϩ16 and ϩ32) at low m/z values (m/z 800 and 1600) (data not shown), whereas low charge distribution states (ϩ8 and ϩ10) at high m/z values (m/z 2500 and 3100) were measured under "nondenaturing" conditions. The calculated mass from these signals was in agreement with that predicted from the gene sequence of BcII (24,960 Da). A systematic variation of [Zn 2ϩ ]/[BcII] resulted in the data presented in Table 2. A detailed description of the ESI-MS technique used for the investigation of Zn-MBL interactions and the derivation of quantitative results have been published before (32). In the respective publication data for the same BcII batch, as used in the present investigation, are shown which also allow the statement that the enzyme batch did not contain any detectable traces of impurities in terms of modified protein.
Under nondenaturing conditions (NH 4 HCO 3 , pH 7), the assignment of the different metal-loaded species resulted from subtracting the observed mass of apoenzyme from the mass of metal-loaded species (supplemental Fig. 1 ). Surprisingly, all spectra of BcII obtained at Zn 2ϩ Ͻ 2BcII indicated the simultaneous presence of apoBcII, Zn 1 -BcII, and Zn 2 -BcII (supplemental Fig. 1). These results indicate positive cooperativity of zinc ion binding to BcII and are in clear contradiction to our earlier published data (10,14).
This prompted us to repeat the corresponding competition experiments with Mag-Fura-2 (supplemental Fig. 2), resulting in K mono ϭ 0.12 nM and K bi ϭ 0.99 nM. The titrations gave no indication of positive cooperativity for zinc ion binding. Data simulation was extensively used to investigate whether any combination of K mono and K bi with K bi Ͻ K mono might be able to fit the titration data. This turned out to be impossible. Thus, there appears to be a contradiction between the MS and competition experiment data obtained for the same enzyme batch.
It has to be pointed out that the new K bi is significantly lower than earlier published values (10,14). A possible explanation for the deviations in the equilibrium constants results from different preparation protocols. Whereas the enzyme for the present study was obtained from E. coli BL21 (DE3), cells grown in M9 minimal medium, earlier studies were performed with BcII derived from the same E. coli strain but grown in LB medium (37). BcII preparations obtained from LB medium-grown E. coli were shown by ESI-MS to be composed of an ensemble of protein fractions with ragged N termini (38) and additionally contained a pigment of unknown chemical nature that was copurified with the enzyme (37). Neither peculiarities were observed in the minimal medium-derived preparation of BcII used in this work, whereas in some of our earlier studies (9, 10, 14, 39) LB medium-derived BcII was used. SDS-PAGE of the purified enzyme resulted in a single band (supplemental Fig. 3), and neither a previously published ESI-MS study of the BcII batch used here (32) nor the mass spectrometric data shown here for the intact protein (ESI-MS) and the peptide fragments obtained by pepsin digestion (MALDI-ToF/ToF MS) gave any indication of heterogeneity of the protein sample in terms of variable total mass or any kind of chemical modification.
Circular Dichroism Spectroscopy at Variable Zinc/Protein Stoichiometries-To probe structural differences, we performed CD spectroscopic measurements with apoBcII samples with 1 or 2 eq of zinc added in comparison to data obtained for BlaB (Fig. 2). The Zn 2 -enzyme spectra show strongly increased intensities of the negative band at 220 nm as compared with the metal-free enzymes and indicate an increased ␣-helix content of the Zn 2 species. The spectroscopic differences between apoenzyme and fully zinc-loaded species are very similar for BcII and BlaB. Whereas the intensity of the 220 nm band is identical for Zn 1 -and Zn 2 -BcII, an additional shoulder at 210 nm appears in the spectrum of the Zn 1 -enzyme.
Selection of Peptic Peptides for HDX Studies Based on Assignment Using MALDI-ToF-ToF MS-The structural view of hydrogen/deuterium exchange kinetics requires the fragmentation of a protein and the separate investigation of smaller peptide fragments. The experimental protocol was standardized with respect to preparation time and temperature before MALDI measurement. Peptide mass fingerprint analysis of pepsin-digested apoBcII resulted in the identification and assignment of 22 peptides in the mass range of m/z 800 -3500 (supplemental Fig. 4 and supplemental Table  1). The average mass accuracy obtained for these peptides was 3.5 ppm. Examples for experimental data are presented in supplemental Fig. 4. The average back exchange for all  7. B, spectra of apoBlaB (thin line) and Zn 2 -BlaB (thick line). Experimental conditions can be found under "Experimental Procedures."

TABLE 1 Fit results for the three-state equilibrium unfolding of BlaB in GdmHCl
The thermodynamic parameters ⌬G 0 (kcalmol Ϫ1 ), m (kcalM Ϫ1 mol Ϫ1 ) and the characteristics of the intermediate state. ⌬S I , obtained from the fits in Fig. 1, B and C  Some sequence sections are found in several peptides due to alternative cleavage sites of pepsin. We selected 14 peptides with 96% sequence coverage and mapped them on the Zn 2 -BcII structure in Fig. 3. The three metal ion ligands of the C-terminal domain, namely His-149, Cys-168, and His-210, are found on the peptide fragments P-(139 -155), P-(165-188), and P-(205-219), respectively. The typical MBL superfamily sequence motif HXHXD is found on P-(82-110) from the N-terminal domain. In the following analysis, only these 14 peptides are used.  Table 2). The percentage of amide hydrogens not involved in main chain-main chain hydrogen bonds is indicated by full horizontal lines, and the percentage of amide hydrogens involved in neither main chain-main chain nor main chain-side chain hydrogen bonds is represented by broken horizontal lines. The latter data were obtained from an inspection of the crystal structure of Zn 2 -BcII (PDB code 1bvt). Fig. 3. The total time courses consist of a very rapid phase, completed within the dead time of our experiments, and a timeresolved phase which is best described by a monoexponential function (Equation 4).

Quantification and Structural Interpretation of HDX Kinetics-A combined representation of kinetic HDX data is shown in
The inspection of the structure of native Zn 2 -BcII (PDB accession code 1bvt) reveals a correlation of the amplitudes of the rapid HDX phase and the percentage of amide protons not involved in hydrogen bonds. Three types of amide protons are classified as follows: (i) amide protons involved in main chainmain chain (mc-mc) hydrogen bonds. They comprise spatially neighboring peptide bonds found in ␣-helices, ␤-sheets, or turns (126 in total). They are shielded against solvent access and, thus, show generally slow HDX. (ii) The same might hold for amide protons involved in hydrogen bonds to side chains (mc-sc) of spatially neighboring amino acids (15 in total). (iii) Best solvent access and highest rates of HDX are expected for amide protons not involved in hydrogen bonds to protein groups. In supplemental Table 2 we have included the total number of exchangeable amide hydrogens together with the numeric values for mc-mc and mc-sc hydrogen bonds taken from the structure (PDB code 1bvt). In Fig. 3 we used these structural parameters to demonstrate the percent HDX expected if all the amide hydrogens not involved in hydrogen bonds were exchanged (horizontal lines). For fully metal-loaded species, the percentage of non-hydrogen-bonded amide protons correlates well with the first data points obtained (compare Fig. 3). Thus, it might be concluded that at least all the hydrogen-bonded amide protons are largely protected against HDX for incubation times Ͻ50 s. The fitted deuterium in-exchange during the dead time of our experiments, thus, mainly covers non-hydrogen-bonded amide protons. In turn, enzyme species showing a considerably higher percentage of HDX at t ϭ 50 s compared with the fully metal-loaded species might have structures with a decreased number of hydrogen-bonded amide protons within the respective peptide segments.
In general, apoBcII displays a higher HDX within the experimental dead time (%D 0 ) compared with the metal-loaded samples, and this is particularly prominent for three peptide segments involved in the structural organization of the metal binding site, namely P-(82-110), P-(82-110), and P-(115-129). This indicates that the amide protons do not participate to hydrogen bonding to the same extent in the apoprotein as in the fully zinc-loaded species and, thus, that the secondary structure elements are not as well defined in these fragments of the apoenzyme.

DISCUSSION
The present investigation aims to describe MBLs and their metal ion binding capacity in terms of flexibility and dynamics. Our starting hypothesis that MBLs may be highly dynamic proteins resulted from a combination of various observations previously described in the literature. A rapid transfer of zinc ions from strong binding carriers like EDTA to the proteins has been shown to require a direct interaction of MBLs with the zinc carrier (14). The latter, however, presupposes a considerable flexibility of the metal binding site to accommodate a zinc loaded EDTA molecule. In another study, a rapid change of the coordination geometry of MBL-bound cadmium ions was described. Hemmingsen et al. (24) could show that a single bound Cd 2ϩ ion apparently jumped between the two metal sites of BcII in a time regime between 0.1 and 10 s. This finding is especially puzzling when compared with kinetic cadmium binding data. The dissociation constant (K d ) for a first cadmium ion bound to BcII is 8.3 nM (24,39). With the second order rate constant k on ϭ 2.6 ϫ 10 7 M Ϫ1 s Ϫ1 , determined for binding of a first cadmium ion (10), the dissociation rate constant (k off ) of a single cadmium ion k off ϭ K d k on ϭ 0.22 s Ϫ1 results. Assuming that the largest activation energy barrier for metal ion dissociation is the breaking of the three bonds to amino acid side chains, the transfer of the metal ion from a protein-bound state to the bulk water most likely represents the rate-limiting step of the dissociation process. If a single cadmium ion bound to a rigid protein scaffold is supposed to jump between two neighboring binding sites, this process would also require the transient breaking of three bonds to amino acid side chains before translocation can take place. Such a mechanism, however, would require a dissociation rate constant between 7 ϫ 10 4 and 7 ϫ 10 6 s Ϫ1 to explain the fast translocation observed. This means that the experimental k off appears to be 5-7 orders of magnitude too low for such a mechanism. Thus, an alternative explanation including the flexibility of the protein for the rapid transfer between both binding sites is required. For example by moving N-and C-terminal domains relative to each other, a site-to-site metal ion transfer may be possible without breaking all the metal-protein bonds at the same time, i.e. an intramolecular ligand exchange reaction. This hypothesis is supported by the fact that both binding sites are composed of ligands contributed by the N-and C-terminal domains (see Figs. 3 and 5).
If bound metal ions contribute to the stabilization of domain interactions, it may be assumed that the total stability of the enzymes is different for metal-loaded and metal-free states. This was tested for BlaB in an unfolding/folding study with GdmHCl as the chaotrope, as described in the following.
Global Effects of Bound Zinc on Structure and Stability of MBLs-In comparison to Zn 2 -BlaB, the apoenzyme shows less intense tryptophan fluorescence at 337 nm (Fig. 1A) and a decreased intensity of the negative 220-nm band in the CD spectrum (Fig. 2B). Both findings indicate structural differences between the two enzyme states, resulting in stronger fluorescence quenching of one or more tryptophan residues and a potential decrease in e.g. ␣-helix content for the apoenzyme. Circular dichroism spectra ( Fig. 2A) of BcII show very similar results. Thus, zinc binding obviously influences the structure of the proteins. That zinc binding increases the stability against unfolding with chaotropic reagents like GdmHCl is apparent from Fig. 1. Whereas Zn 2 -BlaB resists up to 1 M GdmHCl without significant changes of its fluorescence spectrum and without any change of its activity (Fig. 1C), the apoenzyme shows significantly decreased activity under refolding conditions and a significant shift of the center of fluorescence already with 0.2 M GdmHCl (Fig. 1B) (Fig. 1A). The thermodynamic parameters obtained by fitting Equation 2 to the activity and the fluorescence data are very similar, and average values are used for the following argumentation. Zinc binding stabilizes the enzyme by 8.9 Ϯ 1.2 kcal/mol. The difference in thermodynamic stability exclusively originates from the N to I transition, which is characterized by ⌬G 0 of only 1.7 Ϯ 0.5 kcal/mol for the apoenzyme. This corresponds to an equilibrium constant of K ϭ [I]/[U] ϭ 0.056. At 3 M total protein (compare Fig. 1) and 25°C, this results in ϳ5% of the apoenzyme residing in the I state. This observation might also explain the low stability of the apoenzyme upon storage and the need to use a fresh apoenzyme sample for the GdmHCl ϭ 0 M value in Fig. 1B. Thus, the apoenzyme, which is partly unfolded, as implied by the fluorescence and CD spectra, may bear pronounced characteristics of the intermediate state. To localize the parts of the MBL structure involved in the stabilization by zinc ion binding, we initiated a HDX study on BcII.
Local Effects of Bound Zinc on Structure and Stability of MBLs; Zinc-induced Ordering of a Domain-It has been shown earlier that hydrogen/deuterium exchange studied by mass spectrometry (HDX-MS) may be used to monitor metal ion binding-induced conformational changes of metalloproteins because such changes are coupled to a modified total number of amide protons protected against HDX (42,43). Only few investigations, however, used HDX-MS combined with proteolysis of metalloproteins to identify the individual protein regions affected by metal ion binding (44,45). Here we compare the structural flexibility of metal-free BcII and the enzyme loaded with one or two zinc ions.
In our study the absence of bound metal ions increased the solvent accessibility of a number of regions in the protein; highly buried parts close to the active site as well as regions far from the active site. The kinetic analysis results in the highest in-exchange during the experimental dead time (%D 0 ) and/or time-resolved in-exchange (%D t ) for the apoenzyme. For all the  1bvt). It shows the largest effect of zinc binding on HDX. The side chains forming the hydrophobic core of the motif are depicted in van der Waals representation. B, structural localization of the peptides P-(82-110) (light green) and P-(115-129) (dark green) in the crystal structure of BcII. Both are part of the N-terminal domain (blue) and are connected to the C-terminal domain (orange) via the bound zinc ions (blue spheres) and the connecting loop. The strained ␣-helix II is covered by P-(82-110), whereas ␣-helix III is partly covered by P-(82-110) and P-(115-129). Residues 111-114 (red) are not covered by our peptide mass fingerprint analysis. C, structural alignment of the zinc arm motif composed of the metal binding MBL superfamily consensus sequence HXHXD motif, ␣-helix II, and ␣-helix III. The residues corresponding to amino acid positions 83-118 in BcII are depicted. Zinc ions are shown as spheres, and the metal binding ligands are shown as stick models. As representatives of subclass B1, the structures of BcII from B. cereus strain 569/H/9 (red, PDB accession code 1bvt), CcrA from B. fragilis (green, PDB accession code 1znb), IMP-1 from P. aeruginosa (blue, PDB accession code 1jje), and BlaB from C. meningosepticum (dark blue, PDB accession code 1m2x) are shown. Subclass B2 is represented by CphA from A. hydrophila (yellow, PDB accession code 1x8g), and subclass B3 is represented by L1 from S. maltophilia (magenta, PDB accession code 1slm) and FEZ-1 from Legionella gormanii (dark green, PDB accession code 1k07).
peptides the percentage of un-exchanged amide protons at the end of the time-resolved phase (%H end ϭ ⌺(amide protons) Ϫ %D 0 Ϫ %D t ) is lowest for apoBcII.
Few parts of the structure are almost unaffected by metal ion binding. This is most striking for P-(70 -81), which does not show any statistically relevant differences between the kinetic parameters for the different species. This peptide fragment may, therefore, be considered as an "internal standard" for validating deviations in kinetic parameters between different species observed for other fragments. The most pronounced effects of metal ion binding are observed for three peptides from the N-terminal domain, namely P-(82-110), P-(115-129), and P-(82-110) (see Fig. 3).
In the crystal structure P-(82-110) comprises ␣-helix II spanning residues 88 -101 and the MBL superfamily consensus sequence HXHXD motif, which makes up half of the metal binding ligands for the binuclear zinc site. In Zn 2 -BcII (PDB code 1bvt) 20 of 29 amide protons from P-(82-110) are involved in 16 mc-mc and 4 mc-sc interactions. Consequently the in-exchange during the dead time (%D 0 ) for Zn 2 -BcII is below 30% (compare supplemental Table 2). For apoBcII, however, only 6 of the 29 NH protons in total are un-exchanged after 50 s of incubation in D 2 O, which reflects an almost unstructured state of this peptide segment. This means that the secondary structure elements, including ␣-helix II, are most likely unstructured in apoBcII. The two other peptide fragments, which show similar effects of metal ion loading, namely the peptide connecting the N-and C-terminal domains P-(115-129) and P-(82-110), are hydrogen-bonded to P-(82-110) in the crystal structure of Zn 2 -BcII. The connecting loop P-(115-129) is bound to P-(82-110) via one mc-mc and three mc-sc hydrogen bonds. P-(82-110) is bound to P-(82-110) via three mc-mc and three sc-sc hydrogen bonds including salt bridges of the guanidinium group of Arg-91 to Asp-90 and Asp-56. Apparently the destabilization of the P-(82-110) segment resulting from the absence of bound metal ions induces a coupled destabilization of the sequence segments covered by P-(82-110)and P-(115-129) due to the lack of structure-stabilizing hydrogen bonds between the segments. Thus, the segment spanning residues 82-129 appears to be at least transiently unstructured in apoBcII. The CD results also indicate a reduced ␣-helix content in apoBcII, further supporting the interpretation of the MS data. We conclude that zinc binding is required to stabilize the structure of the motif depicted in Fig.  5C, which harbors peptides P-(82-110) and P-(115-129). We suggest that this motif may undergo zinc ion-induced ordering. A structural alignment of MBLs from all three subclasses shows that the overall structure of this domain is highly conserved (Fig. 5C), indicating that the flexibility of the domain is a common feature of all MBLs.
To further explore the flexibility of the domain spanning residues 82-129 from a theoretical perspective, we applied Dis-Prot VSL2B (46,47), RONN (48), and DisEMBL (49) as predictors for intrinsic protein disorder in BcII and BlaB. All three programs resulted in qualitatively very similar results in predicting at least parts of the zinc switch domain as intrinsically disordered regions. It has to be emphasized that all these predictors do not consider structure-stabilizing effects of bound metal ions and, thus, reflect the properties of the apoproteins. That is, the theoretical predictions agree with the flexibility of the zinc switch domain observed experimentally for the apoenzyme. With few exceptions, the regions predicted by VSL2B to have a high probability of being disordered also show a high HDX within the dead time of our experiments (supplemental Fig. 5). Predictions for the sequences of BcII and BlaB show very similar results. Interestingly, the sequence segment P-(165-188) harboring the single cysteine residue, which is a metal ion ligand in the DCH site, is also predicted to be intrinsically disordered and shows high HDX, which is, however, almost independent of the metal loading state (Fig. 3).
Zn 1 -versus Zn 2 -BcII; Metal Ion-bridged Species May Account for Apparently Contradicting Results-The preceding comparisons intentionally did not discriminate the different loading states of BcII with zinc and mainly compared Zn 2 -enzyme and apoenzyme. In the following we intend to highlight the peculiar characteristics of BcII in case only one zinc ion is available per enzyme molecule. To date all descriptions of the MBL-metal system are based on the assumption that single protein molecules are either existent as metal-free units or units loaded with one or two metal ions. In the following we will demonstrate that this perception is inadequate to explain the apparently contradicting results obtained with different methods and should be reframed by additionally considering metal ion-bridged complexes of two or more MBL molecules.
Surprisingly, all ESI-MS spectra of BcII obtained at Zn 2ϩ Ͻ 2BcII indicated the simultaneous presence of apoBcII, Zn 1 -BcII, and Zn 2 -BcII (Table 2). Even without added metal the residual zinc in the apoBcII preparation was obviously sufficient to allow detection of the Zn 2 species (supplemental Fig. 1). Earlier published dissociation constants for BcII (10,14) resulted in a nanomolar and a micromolar K d for a first and second zinc ion bound, respectively. Thus, one would have expected to observe only apoBcII and Zn 1 -BcII up to Zn 2ϩ / BcII ϭ 1. The most obvious explanation for detection of Zn 2species at low Zn 2ϩ /protein requires that K bi Ͻ K mono , which means positive cooperativity of zinc ion binding. Because of the fact that ESI-MS data neither allow the quantification of free zinc nor give direct access to absolute concentrations of the different enzyme species, it is not possible to use the data for direct determinations of dissociation constants. Whereas the absolute values for both K d values of a sequential binding model cannot be obtained, the apparent ratio of the two constants can be calculated from the relative abundances of metal-free and metal-loaded enzyme species according to Equation 6 (see Table 2).
Although MS is not yet a fully validated method in this context, an average ratio K mono /K bi ϭ 3.7 for zinc ion binding to BcII indicates positive cooperativity. Competition experiments with Mag-Fura-2, however, delivered K mono /K bi ϭ 0.12. Thus, the ESI-MS study and the competition titration gave contradicting results in that they appear to reflect positive cooperativity of zinc ion binding or not, respectively.
An additional apparent discrepancy appears when comparing CD and MS data. The CD spectra with 1 and 2 eq of Zn(II) added to the apoenzyme ( Fig. 2A) are very similar with an additional shoulder at 210 nm in the spectrum of the Zn 1 -enzyme. Under the assumption that the ESI-MS data obtained reflected the species distribution in solution, it is straightforward to derive 37.27, 25.47, and 37.27% for apo, Zn 1 -, and Zn 2 -BcII, respectively, for the experimental conditions used in the CD experiment. This calculation is based on the average K bi /K mono of 3.7 (Table 2). A hypothetical CD spectrum for pure Zn 1 -BcII (see Fig. 2A) can than be calculated according to (Equation 7).
Large deviations from both apoenzyme and Zn 2 -BcII are obvious. Because of the spectral characteristics, only significantly increased ␣-helix content could account for the observed deviations. This, however, would mean that Zn 1 -and Zn 2 -BcII should have significantly different folds which should be reflected in the results obtained from the comparative HDX investigation on the three different enzyme species.
For apoBcII, the three peptides P-(82-110), P-(82-110), and P-(115-129) show increased in-exchange during the dead time (%D 0 ) compared with the Zn 2 species (Fig. 3). The amplitudes of the time-resolved phases for these peptides are quite small, and the number of finally un-exchanged amide protons is 90 for Zn 1 -BcII, which is only slightly less than found for Zn 2 -BcII. Although the three peptides show increased %D 0 values for Zn 1 -BcII, others fall below the values found for Zn 2 -BcII, namely the four peptides from the C-terminal domain spanning residues 156 -219. This is most pronounced for P-(205-219), which comprises the metal ion ligand His-210 and appears to be best protected against HDX in Zn 1 -BcII. Especially the latter results are difficult to understand when assuming only monomeric states of the protein as it appears counterintuitive to assume that the Zn 1 state could experience a stronger stabilization compared with the Zn 2 state. The latter findings, however, might be explained from the formation of BcII-Zn 2 -BcII complexes by stabilization of protein segments upon interaction with the neighboring molecule. Attempts to directly detect dimers of Zn-loaded BcII by ESI-MS upon variation of the experimental conditions (data not shown) failed, however, possibly because of insufficient stabilities to survive the ionization process.
Aiming at the detection of possibly existing dimers of BcII, we studied binding of Hg 2ϩ to BcII with 199m Hg PAC spectroscopy. Considering the high flexibility observed in the HDX study and the apparent lack of defined interactions between Nand C-terminal domains, it appeared possible to create new metal ion binding sites by combining residues from two protein molecules. The idea was that the very strong affinity of Hg 2ϩ for sulfur ligands and its preference for formation of dihedral S-Hg-S complexes might result in the formation of complexes with Hg 2ϩ bridging the single Cys residues of two BcII molecules. It has been shown earlier that such dihedral S-Hg-S coordination can be clearly detected by 199m Hg PAC spectroscopy (34,50). We compared two different stoichiometries, namely 1 and 0.1 eq of Hg(II) added to apoBcII. The resulting spectra (Fig. 4) showed clear differences. Whereas the Hg 1 -BcII spectrum could be fitted with a single nuclear quadripole interaction (NQI 1: Q ϭ 1.09 Ϯ 0.01 GHz and ϭ 0.296 Ϯ 0.016), the Hg 0.1 -BcII spectrum could not. NQI 1 could also be identified in the Hg 0.1 -BcII spectrum, but additionally an NQI 2 with Q ϭ 1.50 Ϯ 0.03 GHz and ϭ 0.215 Ϯ 0.061 was necessary to obtain a satisfactory fit. The contributions to the total amplitude are 68 and 32% for NQI 1 and NQI 2, respectively. At present, further investigations are under way to obtain a structural assignment of NQI 1. A comparison with literature data, however, results in the assignment of NQI 2, which is only seen at low stoichiometry as being due to a dihedral S-Hg-S coordination slightly deviating from a linear arrangement (34,40,41). The formation of such a complex, however, requires the direct interaction of two BcII molecules via a bridging Hg 2ϩ as only one Cys is available per protein molecule. Thus, the 199m Hg PAC data indicate that metal-free BcII is flexible enough to allow the formation of such complexes, and we hypothesize that similar complexes might also be formed at substoichiometric availability of zinc ions.
The fact that we could not find a coherent interpretation of all the experimental data with a binding model only covering three states of BcII, namely apoBcII, Zn 1 -BcII, and Zn 2 -BcII, required extension of the binding model to include additional states of the protein. The HDX study resulted in a proven flexibility of the apoenzyme and the lack of structural organization with respect to the domain-domain interaction and organization of the metal binding site. Additionally HDX demonstrated that some parts of the protein are better stabilized with only 1 eq of zinc bound; e.g. P-(205-219) shows the strongest stabilization against HDX in Zn 1 -BcII, which might be easily explained from protein-protein interactions. This prompted us to hypothesize that metal ionbridged complexes might exist in which e.g. two BcII molecules supply ligands for the same metal ion(s). The 199m Hg PAC studies evidenced that such complexes can be formed at least with mercury as the bridging metal ion.
In the following we will demonstrate how the involvement of such complexes, e.g. BcII-Zn 2 -BcII, might abrogate the apparently contradicting results. The metal ion binding sites of such dimers might be composed such that e.g. the HXHXD motif is contributed by protein molecule A and the C-terminal metal ion ligands by molecule B. This model agrees well with the CD data if the secondary structure elements are already stabilized by the presence of only one zinc ion per enzyme molecule and at the same time accounts for the fact that the spectra of the mono-and di-zinc species display minor but distinct differences.
The competition titrations (supplemental Fig. 2) show that a first zinc equivalent binds stronger compared with a second zinc equivalent bound. The simple binding model (only including apoBcII, Zn 1 -BcII, Zn 2 -BcII) is fully sufficient for the theoretical description as the method is not able to discriminate the formation of Zn 1 -BcII from formation of BcII-Zn 2 -BcII. The only apparently contradicting result remaining at this stage is the observation of Zn 2 -BcII by ESI-MS upon sub-stoichiometric addition of zinc (supple-mental Fig. 1). When assuming that the ESI-MS data exactly reflected the situation in solution before the ionization process took place, the results clearly indicate positive cooperativity of zinc ion binding to BcII. When assuming, however, that complexes like BcII-Zn 2 -BcII exist in solution and that these complexes are to labile to survive the ionization process, they might statistically decompose either into two Zn 1 -BcII complexes or into one molecule of apoBcII and one Zn 2 -BcII complex upon ionization. Thus, the previous existence of monomeric Zn 2 -BcII in solution is not necessarily required to explain its detection in the ESI-MS spectra. If the Zn/BcII system has a strong tendency to form BcII-Zn 2 -BcII dimers at low availability of zinc ions, this could also explain why Zn 2 -BcII can even be detected in apoBcII samples with only residual zinc available for binding (supplemental Fig. 1).
Functional Implications and Conclusions-In an earlier study it was hypothesized that MBLs in the absence of substrates might exist as the metal-free apoenzymes in their native bacterial environment. There the free zinc ion concentration is too low to maintain a metal-loaded state when considering the moderately high affinities of MBLs reported (14). The appearance of substrates in the periplasm of bacteria, which is the native site of action of MBLs, would then require the loading of apoMBLs with zinc to establish activity. Substrates have been shown to increase the affinity of MBLs for a first zinc ion bound to picomolar dissociation constants, whereas a second zinc ion is bound by 3-4 orders of magnitude weaker under such conditions (14). For L1 from S. maltophilia it has been shown that the second metal ion bound might get lost during substrate turnover (Ref. 3 and references therein). Thus, the Zn 1 state of MBLs might be the functional state in vivo.
If free zinc is virtually absent, however, only a direct transfer between zinc-loaded competitors and apoMBLs via intermolecular ligand exchange could allow for a rapid activation in the presence of substrates. For such intermolecular transfer mechanisms, the flexibility of apoBcII observed here may be a prerequisite in that the transient formation of complexes with competing zinc carriers appears more probable than would be the case with a preformed metal binding site in a deep groove of the protein structure. Such a mechanism has been suggested for zinc transfer between MBLs and EDTA (14). It has been shown earlier for zinc finger peptides that substoichiometric availability of metal ions might lead to the formation of metal ion bridged complexes (51). Thus, we propose the ability of MBLs to form such complexes if the metal ion binding site concentration exceeds the concentration of available metal ions. If the transient formation of metal ion-bridged protein-protein complexes is part of the activation mechanism of MBLs at low zinc ion concentrations, the presence of substrate might be required to stabilize the active state of the proteins (14) and trap the metal ion(s) in the active site as activation on demand. If the Zn 1 state is considered the functional state of MBLs under in vivo conditions (3,14), the dynamics observed here might also be important for catalysis. In a present freeze-quench PAC study on Cd(II)-substituted BcII we try to investigate metal binding dynamics during substrate turnover. This method allows following substrate-induced coordination geometry changes of the metal ion on a millisecond time scale. Whether the structural zinc switch domain identified in this work has similar functions in other MBL superfamily proteins remains to be seen.