Structural and Kinetic Analysis of Bacillus subtilis N-Acetylglucosaminidase Reveals a Unique Asp-His Dyad Mechanism*

Three-dimensional structures of NagZ of Bacillus subtilis, the first structures of a two-domain β-N-acetylglucosaminidase of family 3 of glycosidases, were determined with and without the transition state mimicking inhibitor PUGNAc bound to the active site, at 1.84- and 1.40-Å resolution, respectively. The structures together with kinetic analyses of mutants revealed an Asp-His dyad involved in catalysis: His234 of BsNagZ acts as general acid/base catalyst and is hydrogen bonded by Asp232 for proper function. Replacement of both His234 and Asp232 with glycine reduced the rate of hydrolysis of the fluorogenic substrate 4′-methylumbelliferyl N-acetyl-β-d-glucosaminide 1900- and 4500-fold, respectively, and rendered activity pH-independent in the alkaline range consistent with a role of these residues in acid/base catalysis. N-Acetylglucosaminyl enzyme intermediate accumulated in the H234G mutant and β-azide product was formed in the presence of sodium azide in both mutants. The Asp-His dyad is conserved within β-N-acetylglucosaminidases but otherwise absent in β-glycosidases of family 3, which instead carry a “classical” glutamate acid/base catalyst. The acid/base glutamate of Hordeum vulgare exoglucanase (Exo1) superimposes with His234 of the dyad of BsNagZ and, in contrast to the latter, protrudes from a second domain of the enzyme into the active site. This is the first report of an Asp-His catalytic dyad involved in hydrolysis of glycosides resembling in function the Asp-His-Ser triad of serine proteases. Our findings will facilitate the development of mechanism-based inhibitors that selectively target family 3 β-N-acetylglucosaminidases, which are involved in bacterial cell wall turnover, spore germination, and induction of β-lactamase.

Glycosidases that catalyze the hydrolysis of glycosidic linkages under retention of anomeric configuration of substrate and product (retaining glycosidases) proceed via a two-step double displacement mechanism (1,2). In most cases, the catalytic machinery of these enzymes involves two carboxylic acids, which are located ϳ5.5 Å apart in the active site (Fig. 1A) and function as catalytic nucleophile and catalytic acid/base (also general acid/base catalyst). In the first step (glycosylation), a carboxylic acid that hydrogen bonds to the glycosidic oxygen acts as a general acid facilitating the leaving group departure simultaneously with a nucleophilic attack by the second carboxylate that forms a covalent glycosyl enzyme intermediate. In the second step (deglycosylation), the first residue then functions as a general base to activate an incoming water molecule for nucleophilic attack that hydrolyzes the glycosyl enzyme to form a sugar hemiacetal product with overall retention of stereochemistry (1,2). This mechanism, for instance, is performed by the well studied lysozyme that hydrolyzes the ␤-glycosidic linkage between N-acetylmuramic acid (MurNAc) 4 and N-acetylglucosamine (GlcNAc) of the backbone polysaccharide of the bacterial cell wall compound peptidoglycan (murein). It was shown only recently that lysozyme proceeds though a covalent ␣-glycosyl enzyme (3) and not a long-lived oxocarbenium-ion intermediate as was proposed earlier (4). We are studying a group of bacterial ␤-N-acetylglucosaminidases, which hydrolyze the other glycosidic linkage in peptidoglycan, between GlcNAc and MurNAc, and are involved in turnover and recycling of the bacterial cell wall (5)(6)(7)(8)(9)(10). These enzymes are classified on the basis of amino acid sequence and secondary structure to family 3 of glycosidases (according to the carbohydrate active enzymes (CAZY) data base), which comprises a heterogeneous group of exo-acting, retaining ␤-glycosidases that besides ␤-N-acetylglucosaminidases (EC 3. Family 3 ␤-N-acetylglucosaminidase from Vibrio furnisii (VfExoII) as well as related ␤-glucosidases were shown to proceed through a glycosyl enzyme intermediate. A conserved aspartate residue was identified in all cases as the catalytic nucleophile by trapping the glycosyl enzyme intermediate using slow substrates, proteolytic digestion, and subsequent mass spectrometry of the labeled peptide (11)(12)(13). Moreover, for some ␤-glucosidases of family 3, good evidence is provided by structural or kinetic analyses for a glutamate acting as the acid/base catalyst, e.g. exo-␤-glucanase of Hordeum vulgare (14 -16), ␤-glucosidase of Flavobacterium meningosepticum (17,18), and the ␤-glucosylceramidase of Paenibacillus sp. TS12 (13). Intriguingly, the structure of the exo-␤-glucanase of H. vulgare (HvExoI) reveals that the glutamate acid/base catalyst resides on a short helix on the less conserved C-terminal domain of the protein that comes into close contact to the active site region of the N-terminal domain (Fig. 1, B and C). A conserved glutamate, which may act as the acid/base catalyst, however, was never identified in the subset of ␤-N-acetylglucosaminidases contained in family 3 (Fig. 2). Moreover, some ␤-N-acetylglucosaminidases of this family even completely lack a C-terminal domain, and therefore need to provide a different residue for acid/base catalysis. It was shown earlier that family 3 ␤-N-acetylglucosaminidases are characterized by the highly conserved sequence pattern KH(F/I)PG(H/L)GX(4)D(S/ T)H, which lays on the N-terminal domain (Fig. 2) and an involvement in acetamido group binding of the substrate was proposed (11).
Here we show that the Asp and His within this pattern (bold letters in the sequence shown above) that reside on the N-terminal domain of BsNagZ are directly involved in the mechanism of the ␤-N-acetylglucosaminidases subfamily of family 3 glycosidases. We present the structure of NagZ of Bacillus subtilis (BsNagZ), the first structure of a two-domain ␤-N-acetylglucosaminidase, along with kinetic analyses, which provide evidence for participation of the side chains of the conserved Asp and His residues during catalysis. Our results indicate that the histidine, instead of a glutamate, acts as acid/base catalyst, which undergoes hydrogen bonding with the aspartate residue, thereby forming a catalytic dyad that protonates the glycosidic oxygen in the first (glycosylation) step and deprotonates and activates water for nucleophilic attack of the glycosyl enzyme in the second (deglycosylation) step of the reaction (Fig. 1D). The function of this unique Asp-His dyad in glycoside hydrolysis resembles that of the Asp-His-Ser triad of serine proteases (19) as well as the Asp-His dyad of ribonucleases (20). A, schematic of the general two-step double displacement mechanism of retaining ␤-glycosidases as proposed also for HvExoI. Two catalytic carboxyl groups, which are located ϳ5.5 Å apart in the active site, act as the catalytic nucleophile and general acid/base catalyst, respectively. B, ribbon model of HvExoI (PDB identifier 1X38). View of the top of the catalytic N-terminal (␤/␣) 8 (TIM-) barrel domain (magenta) to which the inhibitor glucophenylimidazole is bound in the active site (gray sticks; for chemical structures see supplemental Fig. S1), which carries the catalytic nucleophile (orange stick, left). The C-terminal domain, shown in green, is in close contact with the TIM-barrel domain and also contributes to the active site (15). C, a short helix of the C-terminal domain approaches the bound inhibitor and carries a glutamate residue (Glu 491 ) that acts as the catalytic acid/base residue (orange stick, right). The amino acid sequence of ExoI shows 22% identity with BsNagZ. D, schematic of the modified double displacement mechanism as proposed for BsNagZ. Similar to HvExoI, an asparate residue acts as the catalytic nucleophile (Asp 318 ), however, instead of an acid/base glutamate residue an Asp-His catalytic dyad was now identified as the general acid/base catalyst of ␤-N-acetylglucosaminidases of family 3. The catalytic groups are located ϳ6.3 Å apart in the active site. E, ribbon model of BsNagZ (PDB identifier 3NVD). View on the top of the catalytic N-terminal (␤/␣) 8 (TIM-barrel domain (magenta)) to which the inhibitor PUGNAc is bound (gray sticks; for chemical structures see supplemental Fig. S1) in the active site that carries the catalytic nucleophile (orange stick, left). In contrast to HvExoI, the C-terminal domain of BsNagZ, shown in green, is further apart from the TIM-barrel and does not contribute to the active site. F, in BsNagZ an Asp-His dyad mediates the acid/base function that superimposes with the acid/base glutamate of HvExoI (cf. C). Residues Asp 232 and His 234 (shown in orange) are in H-bond distance to each other and His 234 , acid/base catalyst, as well as Asp 318 , the catalytic nucleophile (shown in orange), are H-bonding the PUGNAc inhibitor.
Cloning and Site-directed Mutagenesis-BsnagZ was cloned as an cytoplasmic construct in which the N-terminal signal sequence was removed and exchanged by a His tag. DNA preparation, restriction enzyme digest, ligation, and transformation were performed according to standard techniques using 5 units of PWO DNA polymerase (Genaxxon Biosciences, Biberach, Germany) and 30 ng of chromosomal DNA. BsnagZ from B. subtilis 168 was cloned into pET16b without signal sequence (pET16b-BsNagZ) using the following primers: BsNagZ/FP, 5Ј-AAA ACC ATG GGC CAT ATG TTT TTC GGG GCC AGA CAG AC-3Ј; and BsNagZ/RP, 5Ј-T TTT CTC GAG TTA AAG CGG TCT TCC CGT TTT G-3Ј (underlined are the recognition sites for the endonucleases NdeI and XhoI, respectively). Thirty cycles (15 s at 94°C, 30 s at 55°C, and 120 s at 72°C) were performed in a thermal cycler and revealed a single 1.9-kb fragment (BsnagZ). BsnagZЈ missing the N-terminal signal sequence and the C-terminal domain was amplified using primers with recognition sites for restriction endonucleases NdeI and XhoI, respectively (underlined): BsNagZЈ/FP, 5Ј-AAA ACC ATG GGC CAT ATG TTT TTC GGG GCC AGA CAG-3Ј and BsNagZЈ/RP 5Ј-GGG CTC GAG TGC TTT TTC AGC TAA TTT TTT CTC TGC-3Ј (Thermo Fisher). Thirty cycles (30 s at 94°C, 30 s at 50°C, and 60 s at 72°C) were performed and the 1.3-kb fragment was cloned into vector pET29b (pET29b-BsNagZЈ). The vector pET16b-BsnagZ was used as template for site-directed mutagenesis of H234G. The following degenerated primers (Thermo Fisher) were used: H234G/FP as well as D232G/FP, 5Ј-GG AGA TAT ACC ATG GGC CAT C-3Ј; H234G/RP, 5Ј-C TTG GCC ATG GGA AAC GAG CGG CAG TCC ATA AYM GCT GTC AAC GTC CGT GTC TCC-3Ј; and D232G/RP, 5Ј-C TTG GCC ATG GGA AAC GAG CGG CAG TCC ATA ATG GCT CBC AAC GTC CGT GTC TCC ATG TC-3Ј (the mutated codon is shown in bold with Y ϭ C or T, M ϭ A or C, and B ϭ C, G, or T; the underlined residue is the restriction site for NcoI). Thirty cycles (30 s at 94°C, 30 s at 50°C, and 60 s at 72°C) were performed. The amplified 758-bp fragments carrying the mutation were cleaved with NcoI and exchanged with the wild type fragment of pET16b-BsNagZ. Two mutants were further processed that give rise to H234G (AYM ϭ ACC) and D232G (CBC ϭ CCC) protein variants.
Overproduction and Purification of Proteins-E. coli strain ) gal dcm) harboring pET16b-BsNagZ, pET16b-H234G, pET16b-D232G, or pET29b-BsNagZЈ were grown at 37°C in Luria-Bertani (LB) medium supplemented with ampicillin (100 g/ml) or kanamycin (50 g/ml), respectively, under vigorous shaking to logphase (A 600 0.5-0.6). Production of the enzyme was induced by the addition of isopropyl ␤-D-thiogalactopyranoside to a final concentration of 1 mM. Incubation was continued for a further 3 h and the cells were harvested by centrifugation (4000 ϫ g for 30 min at 4°C), resuspended in ice-cold phosphate buffer (20 mM Na 2 HPO 4 ϫ 2H 2 O, 500 mM NaCl, pH 7.5), and lysed by passing the cells three times through a French pressure cell. The lysates were clarified by centrifugation at 100,000 ϫ g for 1 h at 4°C. The His-tagged proteins in the supernatant were purified by Ni 2ϩ -affinity chromatography on a 5-ml HisTrap column (Amersham Biosciences) pre-equilibrated with phosphate buffer and eluted from the column with imidazole (phosphate buffer supplemented with 500 mM imidazole, pH 7.5). To avoid cross-contamination of the proteins, new columns were used for each protein. The purity of the enzymes was assessed by SDS-PAGE. Fractions containing homogeneous protein were pooled, concentrated, and desalted by dialysis against 10 mM sodium acetate, pH 4.5, for crystallization or against 20 mM phosphate buffer, pH 7.5, for enzymatic assay studies at 4°C. The protein concentrations were measured according to the method of Bradford with bovine serum albumin as a standard. 3 to 5 mg of pure protein was obtained from a 1-liter culture.
Crystallization and Data Collection-Purified BsNagZ was concentrated to 14 mg/ml using an Amicon Ultracentricon (Millipore) with a 10-kDa cut off. Complexes of BsNagZ with the inhibitor for co-crystallization were prepared by mixing PUGNAc with BsNagZ solution at a 10:1 molar ratio and incubating for 30 min at room temperature. Crystals of unliganded BsNagZ and its co-crystals with inhibitor PUGNAc were grown in 0.1 M sodium acetate, pH 4.9, using 24-well hanging drop plates. Presaturated protein drops were prepared by mixing protein and reservoir solution at ratios 1:1, 1:2, and 2:1, yielding final drop volumes of 2-3 l. The reservoir contained 0.1 M sodium acetate, pH 4.9, and varying concentrations of polyethylene glycol 1000. Crystals were shock-frozen in liquid nitrogen. Data sets were collected at the Swiss Light Source beamline X06SA of the Paul Scherrer Institut, Villigen, Switzerland. Data processing was done with XDS (21) (see Table 1).
Structure Determination-The structure of BsNagZ without ligand was solved first. An ensemble of PDB files (1EX1, 1IEX, 1J8V, 1LQ2, and 1X38 (14 -16)) was defined in PHASER (22) as the starting point for molecular replacement. Overall amino acid sequence identities between the target enzyme and the related search models was low (23% and less) (23). Buccaneer (24) was used to build parts of two molecules of BsNagZ (996 out of 1284 residues as poly-Ala model). ARP/wARP (25) and Resolve (26) were used for completing the model. By using Resolve, 2-fold non-crystallographic symmetry (NCS) could be determined and used for phase improvement, leading finally to an almost complete model with 1139 amino acids of the BsNagZ sequence. The resulting electron density map was used for further manual model building in COOT (27) and refinement with Refmac5 and Phenix (28). After modeling water molecules and adding hydrogen atoms, the last anisotropic refinement resulted in a final R-factor of 12.7% (R free , 16.6%). The structure of BsNagZ in complex with its inhibitor PUGNAc was determined by using the first structure of BsNagZ as the search model in Molrep (29). The pseudo-merohedral twinning law was identified with the help of the CCP4 (30) program SFCHECK (31). Model refinement was done in Refmac5 and in Phenix, which allows refinement of the twinned structures. Manual modifications were done in COOT. After adding waters and modeling the inhibitor PUGNAc, the last isotropic refinement resulted in a final R-factor of 18.7% (R free , 24.6%). Refinement statistics are presented in Table 1. Atomic coordinates and structure factors for unliganded and liganded N-acetylglucosaminidases of B. subtilis have been deposited in the Protein Data Bank under accession code 3BMX and 3NVD, respectively.
Kinetic Studies-All Michaelis-Menten kinetics were carried out at least in triplicates using a discontinuous assay measuring the release of 4-methylumbelliferone from 4-Mu ␤-GlcNAc. were initiated by addition of 1.5 mM 4-Mu ␤-GlcNAc and incubated over a period of 5-30 min at 37°C. Protein amounts used were 4.5 ϫ 10 Ϫ5 , 0.0114, and 0.01465 mg for BsNagZ, H234G, and D232G, respectively. All reactions were stopped by addition of 270 l of 0.2 M Na 2 CO 3 , pH 10.0, to 30 l of the reaction mixture. Kinetic parameters were obtained by direct fit of the rate versus substrate concentration data to the Michaelis-Menten equation using Prism 4 software (Graph Pad Software, Inc., La Jolla, CA). The extinction coefficient was determined by calibration using 4Ј-methylumbelliferone. One unit is defined as the amount of enzyme that hydrolyzes 1 mol of substrate per min at pH 5.8 at 37°C.
Mass Spectrometry-H234G was incubated with pNP ␤-GlcNAc over a period of 10 min until the enzyme/substrate solution exhibited a light yellow color indicating the release of pNP. The mass spectrometry analysis of the trapped intermediate (glycosyl-enzyme) was performed by electrospray ionization time of flight mass spectrometry (ESI-TOF-MS) (at the ZMBH, University of Heidelberg).
HPLC Analysis-High performance liquid chromatography (HPLC) was used for analysis of the ␤-GlcNAc azide generated by reaction of BsNagZ-H234G and BsNagZ-D232G with pNP ␤-GlcNAc (6 mM) in 100 mM sodium phosphate, pH 7.5, and 500 mM sodium azide, pH 7.5. In brief a reversed-phase column (Gemini RP-C18, 150 ϫ 4.6 mm, 5 M particle size, Phenomenex) was used at a flow rate of 0.5 ml/min, isocratic elution with 0.05% trifluoroacetic acid for 5 min, followed by a gradient from 0.05% trifluoroacetic acid to 10% acetonitrile containing 0.035% trifluoroacetic acid over a period of 50 min according to Ref. 32.

RESULTS
Purification, Crystallization, and Structure Determination-BsNagZ was overproduced in E. coli cytoplasm and purified by Ni 2ϩ -affinity chromatography to apparent homogeneity (23). The protein retained activity for several months at 4°C and in the pH range between pH 4.0 and 8.0. BsNagZ tends to precipitate at a pH above 6 and low ionic strength but can be kept soluble at 12-13 mg/ml and high ionic strength (e.g. 20 mM sodium phosphate, pH 7.5, 500 mM sodium chloride). At acidic pH solubility behavior is reversed but the protein shows only low enzymatic activity; BsNagZ is soluble at low ionic strength (e.g. 20 mM sodium acetate, pH 4.5) but precipitates upon addition of 100 mM sodium chloride. Crystals of unliganded BsNagZ and its complex with the transition state-like inhibitor PUGNAc, O-(2-acetamido-2-deoxy-D-glucopyranosylidene)amino-Nphenylcarbamate (chemical structures of inhibitors and substrates of BsNagZ are shown in supplemental Fig. S1) were grown at pH 4.9 and 18°C by vapor-phase equilibration as described under "Experimental Procedures." The best results were obtained at polyethylene glycol 1000 reservoir concentrations between 27 and 33% (w/v). Within 2-3 weeks crystals reached a size of approximately 50 m 3 and exhibited triangular, rhombic, or cuboid habits. Data were collected at the Swiss Light Source synchrotron to resolutions of 1.4-Å and 1.84-Å for the native (unliganded) and inhibitor complex (liganded) protein crystals, respectively. All crystals belonged to space group P1 and possessed very similar unit cells with two BsNagZ molecules per cell (cf. Table 1). The inhibitor complex crystals displayed pseudo-merohedral twinning with a twinning fraction of 13.7%. The structures of native BsNagZ and its inhibitor complex were solved by molecular replacement as described under "Experimental Procedures." BsNagZ Has a Unique Two-domain Structure-The final atomic model comprises residues 26 -642 (native enzyme numbering, corresponding to residues 31-647 of the engineered BsNagZ, lacking the N-terminal His tag) from both monomers of the asymmetric unit (structural parameters and refinement statistics, see Table 1). BsNagZ structure reveals two separate domains (Fig. 1E). The N-terminal domain (residues 26 -420) adopts a (␤/␣) 8 -barrel fold (TIM-barrel), which is typical for catalytic domains of glycosidases and contain the conserved aspartate, the catalytic nucleophile of family 3 glycosidases. The C-terminal domain (residues 421-642) displays an ␣␤␣-sandwich fold. The surface buried between the domains (1407.5 Å 2 ) indicates that both domains are intimately associated (33), however, no part of the C-terminal domain comes into close contact with the substrate/inhibitor binding site on top of the (␤/␣) 8 -barrel domain (Fig. 1, E and F). This distinguishes BsNagZ from HvExoI although both display weak sequence similarity (22% overall amino acid sequence identity) (15). As mentioned above, in HvExoI a short helix of the C-terminal domain, which carries the general acid/base catalyst (Glu 491 ), protrudes into the active site (Fig. 1, B and C). This helix is missing in BsNagZ (Fig. 1, E and F). NOVEMBER 12, 2010 • VOLUME 285 • NUMBER 46

JOURNAL OF BIOLOGICAL CHEMISTRY 35679
An Asp-His Dyad in the Inhibitor Binding Site-The overall structures of BsNagZ with and without the inhibitor (liganded and unliganded structures) are basically identical, including the inhibitor binding site arrangement. An acetate molecule in the unliganded BsNagZ structure (Fig. 3A) superimposes with the N-acetamido group of PUGNAc of the liganded structure (Fig.  3B). In the unliganded structure, a sodium ion lies within H-bond distance (2.6 Å) to the catalytic nucleophile Asp 318 and an active site water localizes at 2.8 Å distance to His 234 (Fig. 3A). In the liganded structure, the inhibitor sits in a cavity formed by the N-terminal (␤/␣) 8 -barrel and the GlcNAc moiety of PUGNAc makes multiple H-bond contacts with amino acid site chains constituting the active site of BsNagZ. The average B-factor of 12.9 for the pyranose ring of PUGNAc including the imine nitrogen is consistent with the observed set of H-bonding contacts to amino acids within the binding pocket. The high average B-factor of 56.4 and lack of continuous electron density of the aglycon of PUGNAc is likely caused by high flexibility due to the lack of H-bonding contacts. Intriguingly, N⑀2 of His 234 comes within hydrogen bonding distance (2.9 Å) to the imine nitrogen of PUGNAc (Fig. 3B). The histidine superimposes with the glutamate residue of HvExoI (cf. Fig. 1) and is a likely candidate for the general acid/base catalyst. N␦1 of His 234 , furthermore, forms an H-bond to Asp 232 (distance of 2.9 Å) (Fig. 3B).
Kinetic Properties-Besides natural peptidoglycan substrates (cf. Ref. 23, depicted in supplemental Fig. S1), BsNagZ readily hydrolyzes the chromogenic and fluorogenic substrates pNP ␤-GlcNAc and 4-Mu ␤-GlcNAc), respectively, which are convenient substrates for continuous assays and kinetic studies. Kinetics were determined with 4-Mu ␤-GlcNAc, because this substrate generates a fluorogenic product that can be measured with much higher sensitivity than chromogenic products. Hydrolysis of 4-Mu ␤-GlcNAc by wild type BsNagZ and mutants obey Michaelis-Menten kinetics and the kinetic constants are given in Table 2. To investigate the role of the Asp-His dyad in family 3 ␤-N-acetylglucosaminidases, His 234 and Asp 232 of BsNagZ were exchanged by a glycine. Both mutants showed severely reduced activity. As shown in previous studies with ␤-glycosidases, glycosylation (the first irreversible step that is reflected through k cat /K m ) requires major assistance in protonation of the glycosidic oxygen by the general acid/base catalyst for cleavage of the substrates that have poor leaving groups. By contrast, the second step (deglycosylation, reflected through k cat ) depends on the general acid/base catalyst functioning as base at this stage independent of the leaving group of the substrate (34). Exchanging His 234 with glycine resulted in a 1900-fold reduction in k cat but only a 80-fold reduction in k cat /K m compared with wild type BsNagZ and with 4-Mu ␤-GlcNAc as substrate (a fairly good substrate; pK a of methylumbelliferone ϭ 7.79 compared with the natural muropeptide substrates with an estimated pK a of about 14). The apparent second-order rate constant (k cat /K m ) for substrate hydrolysis of 4-Mu ␤-GlcNAc by H234G thus was much less affected than the apparent first-order rate constant (k cat ), indicating less impairment of the glycosylation step (reflected in k cat /K m ) than the deglycosylation step (reflected in k cat ), indicating a particularly important role of His 234 in base catalysis. This leads to a significant accumulation of the covalent glycosyl-enzyme intermediate (cf. Fig. 4), which is reflected in a 24-fold reduced Michaelis-Menten constant (K m ) for H234G compared with the wild type enzyme.
The rate of 4-Mu ␤-GlcNAc hydrolysis by the D232G mutant was 4500-fold reduced compared with wild type, which is an even larger reduction compared with the effect of the H234G mutation (Table 2). However, the K m was only a little affected by the D232G mutation (Table 2). This indicates a shift in the rate-determining step in the D232G mutant compared with H234G from deglycosylation in the latter to glycosylation in the former. This might be explained by a larger impairment of the protonation of the leaving group of the substrate (glycosylation step) in D232G compared with H234G, presumably because Asp 232 mostly is required for His 234 to function as a proton donor. Protonation of the glycosidic oxygen in catalysis by the H234G mutant might be substituted by small organic acids of the buffer (phosphate, acetate), which, however, cannot substitute general base catalysis (deglycosylation step). Congruently, the activity of the H234G mutant of BsNagZ in the Tris buffer was lower than in phosphate or acetate buffer but to low to determine kinetic parameters.
pH Dependence of Catalysis-The pH activity profile of wild type BsNagZ was compared with H234G and D232G mutants. The plots of log k cat as a function of pH are shown in Fig. 5 and supplemental Fig. S3. The pH activity profile for the BsNagZcatalyzed hydrolysis of 4-Mu ␤-GlcNAc resembles a bellshaped curve as expected due to involvement of two ionizable  This was also the case in the H234G mutant as reflected by a small K m value. Furthermore, direct evidence that His 234 functions as the acid/ base catalyst was performed by ESI-MS measurement. After reaction of the H234G mutant with the substrate pNP ␤-GlcNAc two protein species were observed (Fig. 4). The mass of 71.133 kDa corresponds to the unmodified protein, whereas the mass of 71.336 kDa corresponds to the glycosyl-enzyme intermediate. The mass difference of 203 Da is in accordance with the mass of bound N-acetylglucosaminyl.
Chemical Rescue-As shown above, deglycosylation is the ratelimiting step in hydrolysis of 4-Mu ␤-GlcNAc by BsNagZ H234G, which leads to accumulation of the glycosyl-enzyme intermediate. It has been shown for many glycosidases that the addition of external small anionic nucleophiles like azide can compensate for the loss of the base residue, resulting in rescue of activity (13,18). However, activity of the H234G mutant of BsNagZ could not be restored by the addition of azide, although replacement of His 234 by the small glycine should allow accommodation of small compounds in the active site (see supplemental Table S1). Possibly, binding of the small anionic azide to the active site is disfavored by its negative charge. However, ␤-azide product formation was identified upon pNP ␤-GlcNAc cleavage by D232G (and to some extent also by H234G) in the presence of azide (supplemental Fig. S4). With D232G a 2-fold rate enhancement was observed in the presence of azide (supplemental Table S1), possibly due to accelerating the rate-determining glycosylation step in these mutant.

DISCUSSION
N-Acetylglucosaminidases of family 3 of glycosidases are involved in peptidoglycan turnover and cell wall recycling of bacteria (23,32). In this process they liberate, for instance, inducers of chromosomal ␤-lactamases in Gram-negative bacteria (35,36) as well as spore germinants in the Gram-positive bacterium B. subtilis (37). A detailed understanding of the mechanism of these enzymes is a prerequisite for the design of potent selective inhibitors that may serve as novel therapeutic agents. The identification of evolutionary and structural rela-  3BMX and 3NVD, respectively). The F o Ϫ F c omit maps of the ligands within the active sites are shown, as calculated from the models and contoured at the 2.5level. The orientation of catalytic residues Asp 318 and Asp 232 /His 234 as well as further residues in the binding pocket (Asp 123 , Glu 125 , Arg 131 , Arg 191 , Lys 221 , and His 222 ) that are highly conserved among ␤-N-acetyl-glucosaminidases of family 3 (54), are almost identical to the native, unliganded (A) and liganded structure (B). Putative hydrogen bonds between protein and ligands, identified by using the criteria of proper geometry and a distance cut off of 3.2 Å are indicated. The blue mesh shows the maximum likelihood electron density map for acetate, sodium, and active water (A) and PUGNAc (B) contoured at 0.2 e Ϫ /Å 3 . N⑀2 of His 234 forms a hydrogen bond (2.8 Å) to a water molecule (blue sphere) that is situated in proximity of a bound acetate molecule in A or to the imine nitrogen (2.9 Å) of the transition state inhibitor PUGNAc (gray sticks) in B. The acetate molecule in the binding site in the native structure (A) is located at a position that superimposes with the acetamido group of the PUGNAc in the liganded structure (B). Arg 57 , which is highly conserved among members of the glucosidase, might be involved in binding of the natural substrate (muropeptides). The catalytic nucleophile (O␦1 of Asp 318 ) is in 2.6 or 3.2 Å distance to an enzyme bound sodium ion (red sphere) or C1 of the GlcNAc part of the inhibitor, respectively. His 234 is in hydrogen bond distance to the nitrogen atom between GlcNAc and the aglycon ring of the inhibitor and N␦1 of His 234 is H-bonded to O␦1 of Asp 232 . tionships within glycosidases, which led to the classification into families (CAZY), has greatly facilitated the efforts toward the rational design of such inhibitors. Given that structure is more conserved than sequence, it is assumed that the mechanism of glycoside hydrolysis is identical for all members of a family, which makes this classification particularly valuable. Catalytic residues identified in one member of a family, in general, allowed the prediction of catalytic residues in others. In almost all retaining glycosidases, in which the general acid/base catalyst has been reliably identified to date, it was an enzymic glutamic or aspartic acid residue (cf. CAZY). This, however, does not hold for all members of family 1 glycosidases. Myrosinases of this family cleave the highly reactive S-glycoside sinigrin without requiring protonic assistance for the intermediate glycosylation step in catalysis, hence lacking a general acid/ base residue. The subsequent deglycosylation step, however, depends on the unusual coenzyme ascorbic acid that acts as a base catalyst in the hydrolysis of the glycosyl-enzyme intermediate (38). We recognized that a glutamate residue that acts as acid/base residue is also missing in the ␤-N-acetylglucosaminidase subfamily of family 3 glycosidases (cf. Fig. 2). It would have been possible that substrate participation in the hydrolysis of natural substrates (see supplemental Fig. S1) by these enzymes may not require assistance by a general acid/ base catalyst.
This study now showed that family 3 ␤-N-acetylglucosaminidases, instead of a glutamate general acid/base, involve an Asp-His catalytic dyad, which is unique for glycosidases. The crystal structure of BsNagZ, the first structure of a two-domain ␤-Nacetylglucosaminidase of family 3 glycosidases, revealed that the Asp-His dyad superimposes with a glutamate residue that had been identified as the general acid/base catalyst in other members of family 3 of glycosidases. Most intriguing evidence for the Asp 232 -His 234 dyad functioning as acid/base catalyst in family 3 ␤-N-acetylglucosaminidases was obtained from kinetic studies with protein variants. Removal of His 234 or Asp 232 of the dyad by site-directed mutagenesis resulted in disappearance of the characteristic bell-shaped pH profile of glycosidases and rendered activity pH-independent at higher pH ranges. It therefore suggests that a group was deleted that is responsible for pH dependence in the basic range. Moreover, the first-order rate constants were decreased compared with wild type enzyme by 1900-and 4500-fold for 4-Mu ␤-GlcNAc on exchange of His 234 or Asp 232 with glycine, respectively, which are reductions in k cat values that are typically determined for general acid/base mutant glycosidases (13,18). The second-order rate constants k cat /K m , which reflects the first irreversible step (34), is relatively little affected in the H234G mutant but severely affected in the D232G mutant. This indicates a shift in the ratelimiting step in the His mutant toward deglycosylation, which is also indicated by accumulation of the N-acetylglucosaminylenzyme intermediate by the H234G mutant. The glycosylation step, which involves acid catalysis, however, is the rate-limiting step of wild type BsNagZ as well as BsNagZ-D232G. Together, all these results, structural as well as kinetic data, provide evi-   dence that His 234 of the Asp-His dyad of family 3 ␤-N-acetylglucosaminidases acts as the general acid/base that is assisted by Asp 232 . This contrasts with a study on Clostridium paraputrificum M-21 ␤-N-acetylglucosaminidase (Nag3A). A conserved aspartate residue (Asp 175 ) on the N-terminal domain was proposed as the acid/base catalyst (39) and replacement of Asp 175 with Ala abolished the activity of Nag3A. However, clear kinetic evidence for a role as acid/base catalyst of the above mentioned residue is lacking. The amino acid histidine is perfectly qualified as a general acid/base because it has a pK a near neutrality. Nevertheless, involvement of a histidine residue, respectively, in an Asp-His dyad, is unique in glycosidases. It is, however, commonly found in enzymes cleaving phosphodiester bonds, e.g. ribonucleases (20,40). The major role of the aspartate of the dyad in ribonucleases may be to orient the proper tautomer of the histidine for catalysis. Notably, in bovine ribonucleases N␦1 rather that N⑀2 faces the phosphodiester (20). The His-Asp-Ser catalytic triad is renowned for serine proteases (19) and lipases (41) and the Asp-His hydrogen bond in the catalytic triad is known to contribute greatly to catalysis, potentially via forming of a short, strong (low-barrier) hydrogen bond (42). There is no evidence for the formation of a low-barrier hydrogen bond between His 234 and Asp 232 for the family 3 ␤-N-acetylglucosaminidases. Possibly the major role of Asp 232 in BsNagZ is its influence on proton dissociation of N⑀2 of His 234 for catalysis. Frank and Wen (43) suggested a cooperative behavior in chains of hydrogen-bonded molecules that may sharpen the acid/base behavior of the His for catalysis (reviewed in Ref. 44). The catalytic triad His 57 -Asp 102 -Ser 195 of chymotrypsin (bovine chymotrypsin numbering) operates in this way: Ser 195 functions as a nucleophilic catalyst, assisted by N⑀2 of His 57 that serves as acid/base catalyst, and residue Asp 102 assists in acid/base catalysis by hydrogen bonding to N␦1 of His 57 increasing the pK a of His 57 . In the native BsNagZ structure (Fig. 3A), Asp 232 is in short H-bond distance (2.6 Å) to His 234 and electron density that can be attributed to a water molecule in the H-bond distance to N⑀2 of His 234 . Although, at 1.4-Å resolution this has to be taken with caution, the situation clearly resembles that of chymotrypsin, in which the serine of the triad forms a short H-bond with the His 57 (19). In the liganded structure His 234 H-bonds to the glycosidic oxygen-mimicking nitrogen of PUGNAc and the Asp 232 -His 234 distance is significantly larger (2.9 Å). According to the suggested mechanism (Fig. 1D), in BsNagZ the catalytic His 234 forms an ion pair with Asp 232 thereby allowing protonation of the extracyclic oxygen of the glycosidic bond and facilitating removal of the leaving group upon nucleophilic attack of Asp 318 . In a second step, the His removes a proton of an incoming water molecule thereby hydrolyzing the glycosyl-enzyme intermediate.
The catalytic nucleophile and the Asp-His dyad are located ϳ6.3 Å apart, which is in the range of catalytic residues involved in bond cleavage via a two-step double displacement mechanism in glycosidases (1). They reside on the N-terminal domain of BsNagZ and are positioned at the carboxyl-terminal ends of ␤-strands 5 and 7, respectively. The Asp-His dyad lays on an extended loop that occupies a position after strand 4 in the BsNagZ molecule due to the shortened strand 4 loop, thereby resembling the situation of the 4/7-superfamily of glycoside hydrolases (clan GH-A glycoside hydrolases) in which the nucleophile is positioned at the C terminus of ␤-strand 7 as in family 3 glycosidases but the acid/base catalyst lays in a loop extending ␤-strand 4 (45,46).
Recently, the crystal structure of a single domain ␤-N-acetylglucosaminidases of family 3 of glycosidases from Vibrio cholerae, VcNagZ, in the presence of the competitive transition state-like inhibitor PUGNAc has been reported (36). This structure, however, provided no information regarding the identity of a putative acid/base catalyst. Apparently in this structure, a flexible loop carrying the Asp-His dyad is flipped outward. Moreover, the aspartate nucleophile within this structure is distorted, indicating a non-physiological conformation of the active site in the crystal or an unproductive binding of the inhibitor (cf. Fig. 1E). We recently solved a further BsNagZ structure, in which a short loop of the protein that carries the Asp-His dyad is moved outwards away from the active site. 5 It can be speculated that the proper orientation of the Asp-His dyad in this enzymes is induced upon substrate binding, providing a high degree of substrate specificity.
The obvious question is why the sub-family 3 ␤-N-acetylglucosaminidases apparently are the only glycosidases that act by a catalytic mechanism involving an Asp-His dyad. One rational for the replacement of an acid/base glutamate for a dyad might be the negative charge of the natural substrates of these enzymes, which are MurNAc or 1,6-anhydroMurNAc containing cell wall fragments as mentioned above. The negative charge of the carboxylic acid of these molecules might interfere with the use of a negative charged acid/base catalyst in the active site. A similar situation holds for sialidases, which have been shown to utilize tyrosine as a catalytic nucleophile rather than a carboxylate nucleophile (47,48). It was argued that the anomeric center of the sialic acid sugars bears an anionic carboxylate residue and the nucleophile attack by an anionic nucleophile is therefore disfavored. Sialytransferases of family 42 glycosyltransferases have been reported to utilize a histidine as the base catalyst that abstracts the proton from the nucleophilic hydroxyl group of the sugar acceptor, thereby facilitating attack on the CMP-Neu5Ac donor nucleotide (49,50). Further precedence for the role of histidine residues as base catalysts are reported (51,52).
The residues His 234 and Asp 232 are completely conserved in the subfamily of ␤-N-acetylglucosaminidases of family 3 glycosidases located in the conserved sequence pattern KH(F/ I)PG(H/L)GX(4)D(S/T)H, which is used as an identifier for members of the subfamily. The Asp-His dyad is suitably positioned to act as the acid/base catalyst (cf. supplemental Fig. S2) and may substitute the "normal" carboxylic acid acid/base catalyst to act on substrates bearing a negative charged residue. Our findings will facilitate the development of mechanismbased inhibitors that selectively target family 3 ␤-N-acetylglucosaminidases, which are involved in cell wall turnover, release of spore germinants, and induction of ␤-lactamase in bacteria.