Crystallographic Evidence for Substrate-assisted Catalysis in a Bacterial b -Hexosaminidase*

b -Hexosaminidase, a family 20 glycosyl hydrolase, cat-alyzes the removal of b -1,4-linked N -acetylhexosamine residues from oligosaccharides and their conjugates. Heritable deficiency of this enzyme results in various forms of GalNAc- b (1,4)-[ N -acetylneuraminic acid (2,3)]-Gal- b (1,4)-Glc-ceramide gangliosidosis, including Tay-Sachs disease. We have determined the x-ray crystal structure of a b -hexosaminidase from Streptomyces plicatus to 2.2 Å resolution (Protein Data Bank code 1HP4). b -Hexosaminidases are believed to use a substrate-as-sisted catalytic mechanism that generates a cyclic oxazolinium ion intermediate. We have solved and refined a complex between the cyclic intermediate analogue N -acetylglucosamine-thiazoline and b -hexosaminidase from S. plicatus to 2.1 Å resolution (Protein Data Bank code 1HP5). Difference Fourier analysis revealed the

Carbohydrates are involved in many diverse biological functions including cell structural integrity, energy storage, patho-gen defense and invasion mechanisms, viral penetration, and cellular signaling. Therefore, a large number of enzymes dedicated to carbohydrate metabolism have evolved. Enzymes specifically responsible for carbohydrate catabolism are collectively referred to as glycosyl hydrolases and have been classified into 77 families based on amino acid sequence similarity (1)(2)(3). Three-dimensional structures are known for representatives of 30 of the families. Although there are differences in chain length and domain structure between proteins of a single family, all proteins of a family hydrolyze the glycosidic bond with the same stereochemical outcome (4).
Family 20 includes the ␤-N-acetylhexosaminidases (␤-hexosaminidases) 1 (EC 3.2.1.52), enzymes that catalyze the removal of terminal ␤-1,4 linked N-acetylhexosamine residues from the nonreducing ends of oligosaccharides and their conjugates. In humans, there are two major ␤-hexosaminidase isoforms: HexA and HexB. HexA is a heterodimer of subunits ␣ (encoded by HEXA) and ␤ (encoded by HEXB), whereas HexB is a homodimer of ␤ subunits. HexA is essential for degrading GalNAc-␤(1,4)-[N-acetylneuraminic acid (2,3)]-Gal-␤(1,4)-Glcceramide ganglioside; the biological importance of HexA activity is illustrated by the fatal neurodegenerative disorders that result from its heritable deficiency (5). Mutations in HEXA or HEXB cause Tay-Sachs and Sandhoff disease, respectively. These genetic diseases have made the human ␤-hexosaminidase isoforms the subject of much research. A substantial amount of genetic and biochemical information is available for these isozymes (5), but detailed information about their catalytic mechanism is limited. Mechanistic studies have been primarily limited by the difficulties in producing sufficient amounts of recombinant enzyme needed for kinetic analysis (6,7); however, recent improvements in expression and purification procedures have allowed more accurate kinetic measurements to be made (8). Crystals of human HexB have been grown (9); however, attempts at solving its three-dimensional structure have not been successful. Nonetheless, much insight into the mechanism of human HexA and HexB has been provided by structural and functional studies carried out on related family 20 glycosyl hydrolases (10 -12).
Stereochemical outcome studies on the family 20 chitobiase from Serratia marcescens (13) and human ␤-hexosaminidase (14) demonstrated that this family operates via a retaining mechanism. The consensus view for the mechanism of ␤-retaining glycosyl hydrolases involves general acid catalyzed cleavage of the ␤-(1,4)-glycosidic linkage via a transition state with substantial oxacarbenium ion character to form a glycosylenzyme intermediate with an active site carboxylate. General base-catalyzed attack of water at the anomeric center of this intermediate yields a product with the same anomeric stereochemistry as the substrate. This mechanism is commonly referred to as the double displacement mechanism of hydrolysis (1,15,16). Unexpectedly, family 20 ␤-hexosaminidases and chitobiases, as well as the functionally related chitinases of family 18, were found to lack a carboxylate group suitably disposed to stabilize the oxacarbenium ion transition state (10,11,17,18). Instead, x-ray structural analysis of S. marcescens chitobiase (SmCHB) and kinetic studies with inhibitors have provided strong evidence for catalysis involving participation of the neighboring C2-acetamido group on the substrate ( Fig. 1A and Refs. 10, 13, 19, and 20). The 2-acetamido group of the substrate acts in place of the enzyme nucleophile to generate an enzyme-stabilized oxazolinium ion intermediate. The cyclic intermediate is then hydrolyzed by general base catalyzed attack of water at the anomeric center in a manner analogous to the double displacement mechanism described above.
We have determined the three-dimensional crystal structure of a family 20 ␤-hexosaminidase cloned from Streptomyces pli-catus (21). The 55-kDa enzyme, referred to as SpHEX, is a highly active and stable glycosyl hydrolase that functions over a broad pH range. Co-crystallization of SpHEX with the cyclic intermediate analogue N-acetylglucosamine (NAG)-thiazoline ( Fig. 1B and Ref. 19), and subsequent crystallographic analysis has provided decisive structural evidence for a substrate-assisted catalytic mechanism involving 2-acetamido group participation, resulting in the formation of a covalent, cyclic intermediate (Fig. 1).

EXPERIMENTAL PROCEDURES
Protein Expression and Purification-Escherichia coli strain JM109 was used for plasmid amplification, and plasmid purification was carried out using Qiagen purification systems. Restriction enzymes and Vent DNA polymerase were from New England Biolabs. T4 DNA ligase was from Roche Molecular Biochemicals. All cloning procedures are described in Ref. 22. SpHEX is a 506-amino acid protein having a predicted molecular mass of 55010 Da (GenBank TM accession number AF063001). It was expressed as a recombinant, N-terminal His 7 -tagged fusion protein. Briefly, the plasmid psHEX-1.8 (11) contained the SpHEX open reading frame. The first 100 base pairs of the 5Ј-end of the SpHEX open reading frame was amplified by the polymerase chain reaction using the sense primer (5Ј-GGAATTCCATATGCATCATCAT-CATCATCATCACACCGGCGCCGCCCCGGACCGGAAG-3Ј) and the antisense primer (5Ј-TGGCGCGCCGCCGGGGTCGACCGAGGCGGG-3Ј). This polymerase chain reaction product was restriction digested with AscI and NdeI for ligation into the final expression plasmid. To obtain the remaining 1.7-kilobase pair fragment of the SpHEX open reading frame, a further aliquot of psHEX-1.8 was restriction digested with AscI and BamHI. The 100-base pair (NdeI/AscI) and 1.7-kilobase pair (AscI/BamHI) fragments were then ligated into the T7 expression plasmid pET-3a (Novagen) that had been linearized by digestion with NdeI and BamHI. The ligation product resulted in the expression plasmid p3AHEX-1.8 whose sequence was verified prior to use in fusion protein expression.
The His 7 -SpHEX fusion protein was expressed in E. coli strain BL21 (DE3). Transformed cells were grown at 37°C to an A 600 ϭ ϳ0.5 and then induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 3 h at 25°C. Cells were pelleted by centrifugation, resuspended in a lysis buffer (20 mM Tris-Cl, pH 8.0, 300 mM NaCl, 10 mM imidazole, 10 mM ␤-mercaptoethanol) and lysed by French press. After centrifugation at 20,000 ϫ g for 1 h, the supernatant was loaded onto a nickelnitrilotriacetic acid superflow (Qiagen) column pre-equilibrated with lysis buffer. Once loaded, the column was washed with the lysis buffer supplemented with 80 mM imidazole (pH 8.0). The fusion protein was eluted from the column using lysis buffer supplemented with 250 mM imidazole, pH 8.0, and precipitated with 55% ammonium sulfate for storage at 4°C. Aliquots of the precipitated protein were routinely resuspended and dialyzed twice against 50 mM trisodium citrate, pH 6.0, 300 mM NaCl, and 0.5 mM dithiothreitol and then concentrated to ϳ10 mg/ml with a Millipore concentrator. Approximately 40 -60 mg of pure fusion protein was routinely obtained per liter of culture. Electrospray ionization mass spectrometric analysis using a VG Quattro triple quadrupole mass spectrometer (VG Biotech, Altringham, UK) determined the mass of the purified fusion protein to be 56,054 Da, in good agreement with the theoretical mass of 56,049 Da.
Seleno-Met-substituted His 7 -SpHEX was expressed in E. coli strain BL21 (DE3) pLys S using the method described in Ref. 23. Transformed cells were grown at 37°C in M9 minimal medium until mid-log phase growth was reached. The culture was then supplemented with 0.5 mM Lys, 0.8 mM Thr, 0.6 mM Phe, 0.8 mM Leu, 0.8 mM Ile, and 0.8 mM Val to inhibit endogenous Met biosynthesis. After a 30-min incubation, the culture was further supplemented with 0.25 mM seleno-Met and induced with 0.5 mM isopropyl-1-thio-␤-D-galactopyranoside for 10 h. Seleno-Met substituted His 7 -SpHEX was purified in the same manner as native His 7 -SpHEX except that the protein was dialyzed against 3 mM dithiothreitol before concentrating to avoid selenium oxidation. Electrospray mass spectrometric analysis verified that all 6 Met residues in the 512-amino acid SpHEX protein had been substituted with seleno-Met. All purified fusion protein was visualized for purity by SDS-polyacrylamide gel electrophoresis.
Crystallization and Data Collection-Both native and seleno-Met substituted His 7 -SpHEX crystallized in the hexagonal space group P6 1 22 within 2 weeks by vapor diffusion at room temperature. The mother liquor consisted of 2.2 M ammonium sulfate, 100 mM trisodium citrate, pH 6.0, and 20 -25% glycerol. Hanging drops were set up by mixing an aliquot of SpHEX (concentrated to 10 mg/ml) with an equal amount of the mother liquor. Crystals of His 7 -SpHEX in complex with NAG-thiazoline were obtained by co-crystallization of the native fusion protein (from which dithiothreitol had been removed by dialysis) with 2-5 mM NAG-thiazoline. Diffraction data for a MAD phasing experiment were collected at the Advanced Photon Source, BioCARS sector beamline BM-14-C and BM-14-D on native and seleno-Met substituted His 7 -SpHEX crystals flash cooled to 100 K, respectively (see Table I). Diffraction data from crystals of the complex between His 7 -SpHEX and NAG-thiazoline were collected at Stanford Synchrotron Radiation Laboratory, beamline 9-2 (see Table I). All diffraction data were processed using DENZO and SCALEPACK (24).
Structure Determination and Refinement-A solution to the crystal structure of the protein was obtained by a MAD phasing experiment performed on seleno-Met-substituted protein crystals (25). A combination of data derived from the MAD phasing experiment at beamline BM-14-D with data collected from native SpHEX crystals at beamline BM-14-C allowed for the determination of the three-dimensional structure of S. plicatus ␤-hexosaminidase to 2.2 Å resolution. Although the SpHEX crystals diffracted to slightly higher resolution than 2.2 Å, data collection was restricted to this resolution to avoid excessive data rejection caused by spot overlap. The program SOLVE (26) was used for local scaling of the data and to calculate the anomalous and dispersive differences needed to find selenium sites and to determine phase probability distributions. Patterson maps, calculated from the anomalous and dispersive differences, allowed us to find clearly five of the six selenium atoms present in the SpHEX structure. The missing selenium atom was part of the initiation Met whose position could not be determined because of disorder of the first 14 residues of the His 7 -tagged N terminus.
Electron density maps, generated using structure factor phases obtained from the MAD phasing experiment (initial figure of merit 0.8), were improved only slightly by solvent flattening using Density Modification ( Fig. 2 and Ref. 27). Map boundaries were extended beyond the CCP4 asymmetric unit using EXTEND (28) and skeletonized using MAPMAN (29). A molecular model of the enzyme was built from the skeletonized map using O (30). Residues 8 -512 were readily fit into the density as one continuous chain. Coordinates for the small molecules glycerol and SO 4 Ϫ were obtained through the HIC-UP world wide web site, and their geometries were optimized by X-PLOR (31) prior to use in model building.
The molecular model of native SpHEX was refined using a maximum likelihood target function during both simulated annealing and conjugate gradient minimization as implemented in Crystallography and Nuclear Magnetic Resonance System (32). Prior to refinement, 10% of the diffraction data was randomly flagged for cross-validation using the free R factor. After each round of refinement, the model was manually inspected with O using 2F o Ϫ F c and F o Ϫ F c maps. The final refinement statistics for the model reflect the high quality data (see Table I).
NAG-thiazoline Complex-An F o Ϫ F c map, used to visualize NAGthiazoline in the active site, was obtained using structure factor phases calculated from the native SpHEX model that had been positioned into the unit cell of the NAG-thiazoline complex using rigid body refinement followed by conjugate gradient minimization. Solvent molecules were removed from the model before placing it into the new cell and were relocated during later rounds of refinement. Any waters found in the active site were deleted until the NAG-thiazoline had been modeled into the electron density ascribed to it. The initial NAG-thiazoline model and its geometrical parameters were based on the x-ray crystal structure of N-acetylgalactosamine-thiazoline (GalNAc-thiazoline). 2 Refinement of the NAG-thiazoline complex was carried out using Crystallography and Nuclear Magnetic Resonance System as described for the native SpHEX model above. The final refinement statistics are presented in Table I.
Coordinates-The coordinates and structure factors have been deposited into the Protein Data Bank (native SpHEX Protein Data Bank code 1HP4; SpHEX/NAG-thiazoline complex Protein Data Bank code 1HP5).

RESULTS AND DISCUSSION
Structure of ␤-Hexosaminidase-Excellent crystallographic data (Table I) produced easily interpretable electron density maps into which a model of SpHEX was built (Fig. 2). The enzyme is a kidney shaped, two-domain protein having overall dimensions of ϳ68 ϫ 58 ϫ 56 Å (Fig. 3). The two domains of SpHEX have a similar fold to domains II (residues 214 -335) and III (residues 336 -818) of SmCHB (Fig. 4); however, significant deviations between the two structures exist. The most striking structural difference between SpHEX and SmCHB is the absence in SpHEX of two of the four domains that compose SmCHB (Fig. 4). This results in a solvent-exposed active site at the C-terminal end of the (␤/␣) 8 barrel forming domain II. Such a solvent-exposed active site appears to explain why ␤-hexosaminidases, such as human ␤-hexosaminidase A, can accommodate large glycoconjugates like GalNAc-␤(1,4)-[N-acetylneuraminic acid (2,3)]-Gal-␤(1,4)-Glc-ceramide ganglioside.
Domain I of SpHEX is composed of residues 1-151. As in SmCHB, this domain has an ␣/␤ topology consisting of a solvent exposed, seven-stranded anti-parallel ␤-sheet that buries two, roughly parallel, ␣-helices (Fig. 3). Similar topologies have been found in matrix metalloproteinases (10) and collagenases.  The amino acid sequence identity between SpHEX and SmCHB is lowest throughout this domain. A structure-based alignment using SwissPDBviewer (33) indicated only a 16.1% amino acid identity. Nonetheless, the fold is well conserved, with 87 C ␣ atoms of the two homologous domains having a rms difference of only 1.34 Å. A multiple sequence alignment of all family 20 glycosyl hydrolases indicates that domain I is conserved throughout the entire family. Such conservation suggests a functional requirement for this domain by family 20 glycosyl hydrolases; ironically, however, its function remains unknown.
Domain II of SpHEX is composed of residues 151-512 and is folded into a (␤/␣) 8 barrel with the active site of the enzyme residing at the C termini of the 8 ␤-strands of the barrel. This domain is homologous to domain III in SmCHB, and a structure-based sequence alignment demonstrated there to be a 29.5% sequence identity between the two domains, where 236 of the C ␣ atoms had a rms difference of 1.30 Å. What may indeed be a common feature of this (␤/␣) 8 barrel domain in family 20 glycosyl hydrolases is the conspicuous absence of regular helices at positions ␣5 and ␣7 in the (␤/␣) 8 barrel. In both SpHEX and SmCHB helix ␣5 consists of only a single turn of a 3/10 helix, whereas helix ␣7 is completely absent and is instead replaced by an extended loop. Overall, this domain in SpHEX contains shorter surface loops and is much more compact than its homologous counterpart, domain III, in SmCHB. Multiple sequence alignments of family 20 glycosyl hydrolases suggest that such a compact (␤/␣) 8 barrel domain may be a common feature among many family 20 ␤-hexosaminidases, including the human isoforms (10,11).
Unlike the basic (␤/␣) 8 barrel motif, domain II of SpHEX contains three major loop structures that extend from the C termini of three of the 8 ␤-strands of the barrel. First, loop Lp7 replaces helix ␣7 as described above (Fig. 3). Second, a 36amino acid loop, Lp2, extends from the C terminus of strand ␤2 and contains a short helical segment that packs against and stabilizes the third major loop Lp3. Lp3 is a 41-amino acid loop that extends from the C terminus of strand ␤3 and contains a helical segment that is complimentary to and packs against the helical segment found in Lp2 (Fig. 3). There is only one disulfide bond in SpHEX (Cys 263 -Cys 282 ), and its presence close to the base of Lp3 may help to stabilize the conformation of this loop. Lp3 and Lp7 act in concert to form the hydrophobic faces of sugar binding site ϩ1 described below (Fig. 5). There are two homologous loops in SmCHB; however, they are longer and perform an additional function by interacting with a domain not present in SpHEX (SmCHB domain I) (Fig. 4). Finally, an extra helix continues on from helix ␣8 of the (␤/␣) 8 barrel to complete the C terminus of SpHEX. This extra helix stabilizes domains I and II with respect to each other (Fig. 3). It is interesting to observe that the relative orientation of domains I and II of SpHEX is the same as the homologous domains II and III in SmCHB.
The Complex with NAG-Thiazoline: Mechanistic Implications-According to our x-ray structure of SpHEX and that of the SmCHB-chitobiose complex (10), family 20 glycosyl hydrolases do not appear to contain a side chain in a position suitable to act as a catalytic nucleophile that would stabilize developing oxacarbenium ion character. Instead, it has been observed that, in the conformation bound by the enzyme, the C2 acetamido oxygen of the nonreducing sugar in subsite Ϫ1 is held within 3 Å of its C-1 anomeric carbon. When in this position, it is  believed that the acetamido oxygen can act as a nucleophile and attack the anomeric center to form a cyclic NAG-oxazolinium ion intermediate (10). We have determined the three-dimensional structure of an analogue of the proposed NAG-oxazolinium ion intermediate bound to SpHEX. Because NAG-oxazoline itself is too hydrolytically unstable for use in structural studies, a relatively stable analogue, NAG-thiazoline, has been synthesized and shown to be a potent competitive inhibitor of jack bean ␤-hexosaminidase (K i ϭ 280 nM) (19). NAG-thiazoline also acts as an excellent competitive inhibitor of both SpHEX and human ␤-hexosaminidase B. Fig. 6 shows NAG-thiazoline bound in the SpHEX active site and the quality of the electron density into which it was modeled. Excluding O-4 and O-6 because of differences in C-4 chirality and enzyme packing effects, respectively, the remaining atoms in NAG-thiazoline had an rms difference of only 0.071 Å compared with the equivalent atoms in the small molecule structure of GalNAc-thiazoline. NAG-thiazoline was bound in the Ϫ1 subsite of SpHEX and adopts a conformation that is close to a 4 C 1 chair, although the current data do not exclude small distortions toward a sofa or skew boat conformation. There are no significant changes in the SpHEX structure upon binding NAG-thiazoline except for a slight opening of the active site pocket. Fig. 7 clearly shows Trp residues 344, 361, and 442 of the Ϫ1 subsite of SpHEX and the homologous residues in SmCHB (Trp 616, Trp 639, and Trp 737) forming a tight hydrophobic pocket into which the nonreducing GlcNAc residue binds. This pocket appears to help force the C2 acetamido oxygen into close proximity with the anomeric carbon, and the tight packing between the acetamido group and the enzyme helps ensure a precise alignment of the acetamido oxygen with the anomeric carbon. The hydrophobic pocket will also protect the otherwise reactive oxazolinium ion intermediate from hydrolysis by attack at the oxazolinium carbon atom originally derived from the amide. Indeed, such a protected hydrophobic pocket is highly reminiscent of that found around the catalytic nucleophile in the structures of glycosyl-enzyme intermediates in "normal" retaining glycosidases. In both cases such an environment would protect the intermediate from solvolysis via unwanted pathways. Importantly, the conformation of the sugar in this intermediate is a 4 C 1 chair in both covalent glycosyl-enzyme and cyclic oxazoline.

is the i th intensity measurement and ͗I(h)͘ is the weighted mean of all measurements of
Numerous hydrogen-bonding interactions lock NAG-thiazoline into the active site of SpHEX and disperse the positive charge distributed into the thiazoline ring upon cyclization (Fig. 6B). These include at least one hydrogen bond to every hydroxyl group on the pyranose ring. However, no hydrogen bonds to the ring oxygen O-5 are evident; indeed, a hydrogen bond to O-5 would be counter-catalytic because it would decrease the extent of lone pair donation by O-5 to the antibonding orbital of the scissile bond (34,35).
NAG-thiazoline is held in place particularly strongly by Arg 162 , which forms hydrogen bonds to both O-3 and O-4 of the inhibitor. The mutation R162H results in a 40-fold increase in K m relative to wild type SpHEX and a 5-fold decrease in V max when assayed using 4-methylumbelliferyl-␤-N-acetylglucosaminide (11). The resultant 200-fold decrease in V max /K m confirms that this residue is involved in stabilization of the transition states occurring along the reaction coordinate. The analogous mutation in the ␣-subunit of human HexA (R178H) is associated with the B1 variant form of Tay-Sachs disease in which the enzyme appears to be normally folded and processed but lacks sufficient enzymatic activity and thus results in disease (36,37). Recently, the mutation R211K (homologous to Arg 178 of the ␣-subunit of human HexA) was created in human HexB (8). The mutation resulted in a 10-fold increase in K m , paralleling the findings with SpHEX (Arg 162 ). Furthermore, the k cat value for the R211K mutation was 500-fold less than that of the wild type enzyme, suggesting that it may serve a more important role in transition state stabilization than its counterpart in SpHEX (8). Such bidentate hydrogen bonding from an Arg side chain to two vicinal hydroxyl groups on the substrate has been seen previously in other glycosyl hydrolases. For example, functionally equivalent hydrogen bonding is seen in SmCHB (Arg 349 ) (10) and in Bacillus circulans xylanase, where Arg 112 hydrogen bonds to both O-2 and O-3 (34), and mutation of this residue to Asn results in a 35-fold decrease in k cat /K m . 3 Two particularly important hydrogen-bonding interactions are formed with the thiazoline ring of NAG-thiazoline when it binds to SpHEX. First, the OH of Tyr 393 donates a hydrogen bond to the sulfur atom of the thiazoline ring. In the substrate complex such a hydrogen bond would orient the carbonyl oxygen into position for nucleophilic attack on the anomeric carbon C-1. A similar role is envisioned for Tyr 669 of SmCHB (10). Second, upon formation of the cyclic intermediate, the nitrogen atom N-2 develops a positive charge and SpHEX appears to stabilize this positive charge by delocalizing it through a hydrogen-bonding network between Asp 313 , Asp 246 , and the main chain NH group of Met 247 . This is seen in the two short hydrogen bonds of 2.5 and 2.4 Å from the nitrogen N-2 of the thiazoline ring and the carboxylate oxygens of Asp 313 and Asp 246 , respectively (Fig. 6). These short hydrogen bond distances indicate that the carboxylate of Asp 313 is likely deprotonated and possesses a delocalized negative charge during catalysis.
The other key residue in the active site of retaining glycosidases is the acid/base catalyst, which adopts a dual role, functioning to protonate the departing aglycone in the first step and then to deprotonate the incoming water in the second step. In the structure of the complex of SmCHB with chitobiose, a 2.9 Å hydrogen bond was seen between the glycosidic oxygen of chitobiose and Glu 540 , leading to the assignment of Glu 540 as the acid catalyst (10). Comparative molecular modeling combined with site-directed mutagenesis and kinetic studies of SpHEX and human ␤-hexosaminidase subunits ␣ and ␤ have shown Glu 314 , Glu 323 , and Glu 355 to be homologous to SmCHB Glu 540 , respectively (10,11,38,39). The mutation E314Q in SpHex decreases both V max and K m for 4-methylumbelliferyl ␤-Nacetylglucosaminide by 296-and 7-fold, respectively, confirming an important role for this residue in catalysis (11). Superposition of the crystal structures of SpHEX and SmCHB confirms that Glu 314 of SpHEX is indeed positioned within the active site such that it too would make a hydrogen bond to the glycosidic oxygen of the superimposed chitobiose model (Fig. 7).
The second and final step in the double displacement mechanism is the hydrolysis of the intermediate by general basecatalyzed attack of water at the anomeric center C-1, resulting in overall retention of the anomeric configuration. Figs. 5 and 8 show the position of a glycerol molecule bound in the ϩ1 subsite. This glycerol superimposes onto half of the pyranose ring of chitobiose and suggests that subsite ϩ1 in SpHEX causes the sugar in this subsite to be twisted ϳ90°relative to the sugar bound in subsite Ϫ1 (Fig. 5). Furthermore, one of the hydroxyl groups of this glycerol is within 3.4 Å of the anomeric C-1 of NAG-thiazoline and forms a hydrogen-bonding interaction with the carboxylate of the general acid/base Glu 314 . We postulate that this hydroxyl group occupies the position that an incoming water molecule would take to nucleophilically attack C-1, thereby hydrolyzing the oxazolinium ion intermediate, with release of ␤-N-acetylglucosamine. Abstraction of the proton from water by Glu 314 is assisted by a hydrogen-bonding network formed between its carboxylate group, the imidazole nitrogens of His 250 , the carboxylate of Asp 191 and the main chain NH group of Asp 192 (Fig. 8). The active site water molecule seen in the SmCHB structure, and proposed to be the reactant species (10), is indeed conserved in the SpHEX structure and is indicated in Figs. 6 and 8 as WAT. However, this water molecule is buried within the active site of both structures, and it seems more plausible that the incoming water enters directly from the bulk solvent after departure of the aglycone rather than occupying this site first. The role of buried 3 M. Joshi, personal communication. water is unclear, but structured waters that mediate the binding of sugars with proteins are quite common and may provide some of the flexibility required to accommodate substrates of both gluco and galacto configuration.
A ␤-retaining mechanism utilizing acetamido group participation in family 20 ␤-hexosaminidases and chitobiases is consistent with observations from glycosyl hydrolases from the functionally related family 18 (17). In this family, there is also no apparent enzyme nucleophile, and crystallographic analysis of the family 18 plant chitinase hevamine in complex with the chitinase inhibitor allosamidin suggests that a similar cyclic reaction intermediate is formed in chitinases by C2-acetamido group participation (18,40). Further examples of enzymes possibly utilizing substrate assisted catalysis include soluble lytic transglycosylase (41), and goose lysozyme (42). Hence, it appears that substrate-assisted catalysis is a common feature between glycosyl hydrolase families 18 and 20 and potentially other families.
Mechanistic Conclusion-A combination of the results from this study, in which the structure of a complex with an intermediate analogue is presented, with those from a previous study of the structure of the substrate (chitobiose) complex with SmCHB allows interesting insights into the reaction mechanism and particularly into the substrate conformational changes that occur along the reaction coordinate.
The substrate binds to the enzyme with the sugar in the Ϫ1 subsite in a distorted sofa/boat conformation, as seen in the bound chitobiose structure (Fig. 7). This places the scissile bond in a pseudo-axial orientation similar to that seem for the complex of lysozyme with NAM-NAG-NAM bound as a product (43). Such a conformation allows atoms C-5, O-5, C-1, and C-2 of the sugar in the Ϫ1 site to adopt the coplanar configuration and glycerol (Gol) bound to sugar binding subsites ؊1 and ؉1 of SpHEX, respectively. Semitransparent surfaces have been drawn around hydrophobic residues using GRASP (50). The catalytic triad (Glu 314 , His 250 , and Asp 191 ) has been drawn along with its hydrogen-bonding network. The glycerol hydroxyl group hydrogen bonding to the carboxylate of Glu 314 is believed to occupy the position that an incoming water molecule would take to nucleophilically attack C-1. WAT indicates the conserved incoming water molecule proposed by Ref. 10. required for effective overlap of the nonbonding lone pair of electrons on O-5 with the antibonding orbital at the electrondeficient anomeric center of the oxacarbenium ion. This conformation not only satisfies the requirements of stereoelectronic theory, it also obeys the principle of least nuclear motion and the need to minimize 1,3-diaxial repulsive interactions between the approaching nucleophile and H3 and H5 of the substrate (15,35,44). A similar conformational distortion of the analogous sugar has been observed in a nonhydrolyzable thiooligosaccharide mimic of cellulose bound to endoglucanase I from family 7 (45). Upon cleavage of the glycosidic bond, with concerted proton donation from Glu 314 , the sugar ring relaxes to the 4 C 1 chair conformation, as evidenced by the structure of the spHEX-thiazoline complex. Hydrolysis of this intermediate then follows a similar conformational itinerary, with formation of a product complex in a skew boat conformation, and finally product release. A very similar conformational itinerary has been shown for a normal retaining ␤-glycosidase in which a covalent glycosyl-enzyme intermediate is formed.
Interestingly, these crystal structures reveal that as the bound substrate proceeds along the reaction coordinate to yield the enzyme-bound product, the greatest nuclear motion of heavy atoms occurs at C-1, as shown in Fig. 9. As the reaction proceeds, the C-1 atom scribes an arc from its initial position (position 1) as it breaks a covalent bond to the glycosidic oxygen to form a new bond with the acetamido oxygen (position 3). Approximately halfway along this arc is the transition state where C-1, C-2, C-5, and O-5 are coplanar (position 2). During hydrolysis of the intermediate, C-1 traces the reverse path as it breaks the bond with the oxazolinium ion ring oxygen, proceeds through the transition state, and forms a covalent bond with a suitably positioned water molecule. Thus the motion of C-1 through the catalytic cycle can be described as a "wagging" back and forth from positions below and above the plane of the sugar ring. Very little motion of protein atoms is required; all changes occur within a site that has been optimized for this minimized motion. This behavior is entirely consistent with both the antiperiplanar lone pair hypothesis and the principle of least nuclear motion and appears to be general for retaining glycosyl hydrolases (44,46).