Mechanism of the Reductive Half-reaction in Cellobiose Dehydrogenase*

The extracellular flavocytochrome cellobiose dehydrogenase (CDH; EC 1.1.99.18) participates in lignocellulose degradation by white-rot fungi with a proposed role in the early events of wood degradation. The complete hemoflavoenzyme consists of a catalytically active dehydrogenase fragment (DHcdh) connected to a b-type cytochrome domain via a linker peptide. In the reductive half-reaction, DHcdh catalyzes the oxidation of cellobiose to yield cellobiono-1,5-lactone. The active site of DHcdh is structurally similar to that of glucose oxidase and cholesterol oxidase, with a conserved histidine residue positioned at the re face of the flavin ring close to the N5 atom. The mechanisms of oxidation in glucose oxidase and cholesterol oxidase are still poorly understood, partly because of lack of experimental structure data or difficulties in interpreting existing data for enzyme-ligand complexes. Here we report the crystal structure of the Phanerochaete chrysosporium DHcdhwith a bound inhibitor, cellobiono-1,5-lactam, at 1.8-Å resolution. The distance between the lactam C1 and the flavin N5 is only 2.9 Å, implying that in an approximately planar transition state, the maximum distance for the axial 1-hydrogen to travel for covalent addition to N5 is 0.8–0.9 Å. The lactam O1 interacts intimately with the side chains of His-689 and Asn-732. Our data lend substantial structural support to a reaction mechanism where His-689 acts as a general base by abstracting the O1 hydroxyl proton in concert with transfer of the C1 hydrogen as hydride to the re face of the flavin N5.

doreductases carrying protoheme and FAD cofactors bound to separate domains. In vitro, CDH from the white-rot Basidiomycete Phanerochaete chrysosporium depolymerizes cellulose, hemicelluloses, and lignin (Refs. 1-3; for review, see Ref. 4) as well as other polymers (5). The exact biological function of CDH has been a subject of lively debate, but recent results suggest that the enzyme is important for invasion and colonization of wood (6).
The catalytic site is located in the flavoprotein domain, where the reductive half-reaction proceeds by oxidation of ␤-cellobiose (apparent k cat 15.7 s Ϫ1 and K m 0.11 mM, see Ref. 7) to yield cellobiono-1,5-lactone ( Fig. 1) and the concomitant twoelectron reduction of FAD. In dilute aqueous solution, cellobionolactone hydrolyzes to cellobionic acid. Results from 1 H NMR spectroscopy show that the product from cellobiose oxidation by CDH is unequivocally cellobionolactone, and thus, cellobionic acid is not formed on the enzyme (8). During the ensuing oxidative half-reaction, the flavin is re-oxidized by an electron acceptor, either directly or via the cytochrome domain (9). At present, the most favored mechanism for in vivo degradation of biopolymers by CDH is the reduction of ferric compounds present in wood in the presence of hydrogen peroxide to form hydroxyl radicals through a Fenton-type reaction (2,10). The cytochrome domain has been implicated in this reaction (10) since the generation of hydroxyl radicals proceeds by oneelectron reduction; however, the flavin domain is also able to generate hydroxyl radicals (1). Besides cellobiose, soluble cellodextrins, mannobiose and lactose, are good or acceptable substrates for CDH, whereas monosaccharides are poor substrates (7). CDH can use a large number of electron acceptors; however, reduction of oxygen is slow (11).
We recently reported the crystal structure of the dehydrogenase fragment of CDH without ligand (DH cdh ) showing that it is closely related to that of members of the glucose-methanolcholine (GMC) family of oxidoreductases (12). The GMC oxidoreductases glucose oxidase (GOx (13,14)) and cholesterol oxidase (ChOx (15,16)) are similar in overall structure and active-site architecture to DH cdh (12). GOx catalyzes the oxidation of ␤-D-glucose to D-glucono-1,5-lactone. Thus, GOx and CDH perform similar chemistry in the reductive half-reaction. The bifunctional enzyme ChOx, however, catalyzes the oxidation and isomerization of cholesterol to 4-cholesten-3-one. Although CDH and GOx catalyze similar chemical reactions, the structural details of the catalytic site in the immediate vicinity of the flavin ring in CDH is more similar to those in ChOx; that is, two conserved residues (His and Asn) at the re face of the flavin ring in a similar conformation and with near identical geometry relative to the flavin N5 (12). In GOx, however, the asparagine is replaced by a histidine residue. Based on modeling of cellobiose in the DH cdh active site, we suggested two glucosyl-binding sites at the re face of the isoalloxazine ring with the reducing end of cellobiose bound to the innermost site C close to the flavin ring and the non-reducing end of cellobiose residing in the distant site B (12).
Over the years, the most widely accepted reaction mechanisms for flavin-assisted dehydrogenation include the carbanion mechanism (17,18), the radical mechanism (19,20), and the hydride-transfer mechanism (21)(22)(23). In general, crystallographic data are not sufficient per se to conclusively discriminate between these mechanisms. However, with the increasing number of available ligand complexes for redox-active enzymes, the hydride-transfer mechanism is gaining in popularity as a general mechanism for dehydrogenation, although the radical mechanism has not been disproved. The only structure available for a GMC enzyme-ligand complex is that of Brevibacterium sterolicum ChOx with bound dehydroisoandrosterone (16). The authors propose a radical mechanism, although the structural details of ligand binding did not exclude any of the possible mechanisms. The situation is, at least partly, complicated by the dual activity of ChOx.
We have used the inhibitor 5-amino-5-deoxy-cellobiono-1,5lactam (Cblm; K i ϳ0.25 mM at 35°C, data not shown), which has a geometry similar to that of the product and of a possible transition state (TST), to study the active-site interactions and delineate a possible reaction mechanism for CDH. This is the first crystal structure of a CDH with a ligand bound in the active site. The reaction mechanism for enzymatic oxidation of cellobiose by CDH is discussed in the light of the present structure of a complex between the dehydrogenase fragment of CDH and Cblm, determined at 1.8-Å resolution.

EXPERIMENTAL PROCEDURES
Crystallization, Data Collection, and Refinement-DH cdh was purified and crystallized as described elsewhere (7,12). Data were collected at 100 K from crystals soaked overnight in reservoir solution containing 1 mM Cblm. Data collection and model refinement statistics are summarized in Table I. Data reduction and scaling were carried out using MOSFLM (24) and SCALA (25), respectively. Our reported structure of DH cdh (Protein Data Bank code 1KDG (12)) was used as starting model for crystallographic refinement against DH cdh ⅐Cblm amplitudes. Initial refinement was done with CNS (26), and manual re-building was done with the program O (27). Starting coordinates for Cblm were generated using CORINA (28) followed by manual fitting of the model to the electron density. Final refinement was done with REFMAC5 (29) at 1.8-Å resolution usg anisotropic scaling, hydrogens in their riding positions, and atomic displacement parameter refinement using the translation, libration, screw-rotation model. The flavin cofactor, inhibitor, glycosylation adducts, and the substrate and flavin binding domains of DH cdh (12) were defined as rigid bodies during translation, libration, screw-rotation refinement. The model contains 2 protein molecules (residues 215-755), two 6-hydroxylated FAD molecules; 5 Nacetylglucosamine residues (3 in molecule A and 2 in B), 2 Cblm molecules, and 1007 water molecules (533 for A, 473 for B, and 1 located on a non-crystallographic symmetry 2-fold axis). The bending angle of the isoalloxazine ring was calculated as described previously (12).
Modeling of Cellobiose-Cellobiose was modeled manually in the active site of DH cdh guided by the observed binding pattern for Cblm. To relieve geometric strain and impose favorable van der Waals contacts, the model was subjected to energy minimization with CNS (26) without the x-ray pseudo-energy term. During the energy minimization, only atoms within a sphere of 5.5 Å from the linking oxygen in the glycosidic bond were allowed to move. Atoms within a cushion of 3.5 Å around the sphere were refined with harmonic restraints.

RESULTS
Overall Structure-The 1.5-Å crystal structure of DH cdh (Protein Data Bank code 1KDG) has been reported elsewhere (12). In brief, DH cdh consists of an FAD binding subdomain and a substrate binding subdomain (Fig. 2a). The entrance to the active site is located at the subdomain interface, where a 12-Å-long tunnel leads down to the flavin ring. The structures of DH cdh and DH cdh ⅐Cblm are nearly identical with root mean square deviation values (all atoms) of 0.35 and 0.34 Å for pair-wise least-squares comparisons of A and B molecules, respectively. The electron density for the protein as well as for the inhibitor is of excellent quality (Fig. 2b). The average residual B factor for the ligand (all atoms) is 8.2 and 8.4 Å 2 for the A and B molecule, respectively.
The Flavin Cofactor-As observed for DH cdh , the non-covalently bound flavin cofactor in DH cdh ⅐Cblm is present as 6-hydroxylated FAD. The butterfly bending angle of the flavin ring is less pronounced in DH cdh ⅐Cblm (molecule A, 7°; molecule B, 11°) compared with DH cdh (22°). The flattening of the flavin ring appears to result from an induced fit where the N5 moves into the isoalloxazine plane to accommodate the incom-  of the DH cdh ⅐Cblm complex Data were collected using synchrotron radiation at station ID14-EH4, European Synchrotron Radiation Facility, Grenoble, France, ϭ 0.977 Å. The outer shell statistics comprising 5% of the reflections are given in parentheses. The percentage of residues that fall outside core regions of the Ramachandran plot are defined according to Kleywegt and Jones (39) where an average model at 2.0-Å resolution or better has 0 -5% outliers. The atomic coordinates and structure factors (code 1NAA) have been deposited with the Protein Data Bank. NCS, non-crystallographic symmetry; r.m.s., root mean square.

Data collection
Resolution ( ing C1 atom of the lactam ring. The angle defined by the flavin N5 and N10 with the backbone nitrogen atom of Gly-310 is 136°a nd the N5-Gly-310 N distance is 3.2 Å. Binding of Cellobionolactam-Inhibitor interactions are outlined in Fig. 3, a and b. The glucosyl-binding sites and the substrate-binding residues are located at the re face of the isoalloxazine ring. The lactam moiety of Cblm, corresponding to the reducing end of cellobiose, is bound in site C with four protein residues and three solvent molecules within hydrogenbonding distance of its exocyclic carbonyl and hydroxyl groups: O1-His-689 N⑀ 2 , O1-Asn-732 N␦ 2 , O2-Ser-687 O, O2-His-689 N⑀ 2 , and O3-Asn-688 N␦ 2 . The C3 and C6 hydroxyl groups can form one and two water-mediated hydrogen (H) bonds, respectively. Thus, a total of eight H-bonds are possible in site C. The endocyclic lactam nitrogen is positioned near the flavin N5 (3.2 Å) and O4 (2.9 Å). The C1 atom of the lactam moiety, which corresponds to the site of oxidative attack in cellobiose, binds in a position 2.9 Å in front of and below the N5-C4a locus of the isoalloxazine ring, defining an angle of 108°(molecule A) and 110°(molecule B) with the N5-N10 atoms of the flavin ring. These values are in agreement with those typically observed in flavoenzymes (30). It should be noted that the lactam C1 and O1 are almost perfectly aligned with the flavin N5 and C4a, respectively (C1-N5, 2.9 Å; O1-C4a, 2.9 Å). The short distance (molecule A, 2.5 Å; molecule B, 2.6 Å) of the H-bond formed between the lactam O1 and His-689 N⑀ 2 suggests that this interaction is strong and that His-689 is suitably positioned to deprotonate the substrate hydroxyl group.
The glucosyl moiety of the ligand resides in site B where it forms a total of five ligand-protein H-bonds: O2-Glu-279 O⑀ 2 , O2-Arg-586 N⑀, O3-Glu-279 O⑀ 1 , O3-Arg-586 N 2 , and O6-Asn-688 N␦ 2 . The aromatic ring of Phe-282 has rotated (molecule A, 7°; molecule B, 11°) compared with the DH cdh structure to stack more planar with the B-site pyranose ring. Five solvent-mediated H-bonds are observed, one each for O2, O3, and O6 and two for O4. A total of 10 H-bonds and one planar hydrophobic stacking interaction contribute to binding in site B. Judged solely from the number of possible interactions, site B is likely to make substantial contribution to the binding of cellobiose, which is in line with reported kinetic constants for di-and monosaccharides (7).
Coupled Tyrosine-Serine Flipping-Only two residues undergo conformational change in response to inhibitor binding (Fig. 3b). In the non-liganded DH cdh structure, Tyr-609 is kept out of the C-site by H-bonds formed by its hydroxyl group with backbone atoms of Asn-732 and the Gln-582 side chain. The active-site tunnel is filled with solvent molecules, of which one water molecule binds in front of the flavin ring within Hbonding distance from His-689 N⑀ 2 .
In the DH cdh ⅐Cblm complex, however, the tyrosine side chain has discarded its H-bonding partners and moved into site C (Fig. 3b). In the new position, Tyr-609 resides below the lactam ring where O coordinates two ordered water molecules not present in the non-liganded structure. One water is located below the lactam O6, where it is involved in a H-bond network with Tyr-609 O, Thr-581 O␥, Asn-732 N␦ 2 , and the C6 hydroxyl group. The second water molecule is below the lactam C3 hydroxyl group and forms H-bonds to Tyr-609 O, the lactam O3, and another water molecule. The tyrosine and its two coordinated water molecules (colored yellow in Fig. 3, b and c) elevates the "floor" of site C.
The flipping of the tyrosine side chain imposes two additional changes in the active site as follows. (i) To accommodate the Tyr-609 side chain in its new position, the side chain of Ser-519 also flips and thereby exchanges its two H-bonding partners (Ser-519 O␥-Ser-687 O␥, and Ser-519-water) for two new ones (Ser-519 O␥-Met-685 O, and Ser-519 O␥-new water). The concomitant flips of Tyr-609 and Ser-519 give rise to a local backbone-C␣ displacement of 0.6 Å and 0.8 Å at residue 609 and 519, respectively; (ii) the re-positioning of the Tyr-609 side chain together with the presence of the ligand effectively traps a water molecule (Wat1366) in a position 3 Å from the C⑀ 1 -N␦ 1 edge of the His-689 imidazole ring (Fig. 3, a-c), i.e. the imidazole side opposite to that interacting with the C1 hydroxyl in cellobiose. This water molecule is not present in the non-liganded structure.

Comparison of Cellobionolactam with Modeled Cellobiose-
Structurally, cellobionolactam resembles the product cellobionolactone, with the only difference that the endocyclic O5 oxygen is replaced by an N-H function (Fig. 1). In both compounds C1 is coplanar with O1, C5, and O5 (or lactam nitrogen). We thus assume that the tetrahedral configuration of the anomeric center of cellobiose is changing toward a planarized FIG. 2. Structure of the DH cdh with bound cellobionolactam. a, overall structure of the DH cdh fragment with bound cellobionolactam in the active site. The polypeptide chain has been color ramped from the N terminus (blue) to the C terminus (red). ␣ helices and ␤ strands are shown as spirals and arrows, respectively. The FAD cofactor and the ligand are shown as ball-and-stick representations. Atom colors are blue for nitrogen, red for oxygen, and for carbon, yellow (FAD) or green (Cblm). The schematic was made with the program PyMOL, www. pymol.org (40). b, A -weighted F o Ϫ F c electron density map calculated to 2.0-Å resolution using the model from the first simulated-annealing refinement with CNS where only protein atoms had been included and refined. The electron density for the ligand is, therefore, free from model bias.
intermediate during its transition to a sp 2 -hybridized carbon in cellobionolactone.
Based on the Cblm binding pattern, we have produced a docking model of cellobiose in DH cdh . Minor steric restraints near the flavin N5 and His-689 N⑀ 2 enable the planar lactam C1AO to penetrate somewhat farther (0.3-0.4 Å) into site C compared with the equatorial O1 group of cellobiose (Fig. 3c). The C1-N5 distance is 2.9 Å in the cellobiose model, and the shortest distance that the C1 hydrogen (1-H) would need to traverse to bind covalently to N5 is, thus, roughly 0.9 -1.0 Å. In the Cblm complex, the corresponding transfer distance (imagining a 1-H as in the TST) would be slightly less, 0.8 -0.9 Å. Although these distances do not differ significantly, the precise geometry of the 1-H relative to N5 is slightly different in cellobiose as compared with a TST in that 1-H in a TST would be closer to being aligned so as to interact with the lowest unoccupied molecular orbital of the flavin (Fig. 3c). The assumption that the observed inhibitor binding is also valid for a TST implies that upon approaching the TST the substrate slides slightly deeper into the active site. This results in a better alignment of the 1-H with the flavin lowest unoccupied molecular orbital. Similar to what has been suggested for several other flavoproteins (30), the resulting negatively charged flavin hydroquinone may be stabilized at the N1-C2AO locus of the isoalloxazine ring by the positive dipole of the C-terminal ␣ helix. These small but distinct differences in binding may be interpreted as the oxidative site favoring a partially planar TST mimicked by the lactam ring in our structure.
Mechanistic Implications-Similar specific relative geometry of the substrate and cofactor-reactive groups has been observed in crystal structures of ligand complexes for nicotinamidedependent (31), flavin-dependent (32), and quinone-dependent (33) oxidoreductases, all of which have been assigned a hydride-transfer mechanism. From a purely structural viewpoint, the close proximity between C1 and N5 together with the relative geometry of the atoms appears to favor a general base-catalyzed hydride-transfer mechanism (Scheme 1, panel a). General base-assisted deprotonation of the C1 hydroxyl group by His-689 in concert with the expulsion of 1-H as hydride via a planar or nearly planar TST would be entirely consistent with the experimentally observed binding of Cblm and modeled cellobiose.
The carbanion mechanism in its classical implementation requires that His-689 abstracts 1-H as a proton, resulting in a substrate carbanion. The carbanion then performs a nucleophilic attack at N5 to form a covalent C1-N5 adduct. The subsequent elimination reaction proceeds by the concomitant formation of a double bound between C1-O1 and uptake of the O1 hydroxyl proton by the flavin O4 and ultimately by N5. To accommodate the carbanion reaction, the C-site glucosyl residue would need to tilt forward toward His-689 by at least 45-90°to position the 1-H for proton abstraction. Such a conformational change in the spatially restricted active site is unlikely, structurally and energetically. The non-reducing end of the substrate is anchored in site B (Fig. 3b), and the conformational change would introduce unreasonable strain in the glycosidic bond between sites B and C. The intimate interaction between His-689 and O1 makes it difficult to find any reasonable incentive for this residue to abstract the more distant 1-H. The carbanion mechanism is also highly unlikely from a purely chemical point of view in that the generated substrate carbanion would be conjugatively destabilized, as has been well established by work of Eliel and co-workers (34). Thus, we suggest that the carbanion mechanism is incompatible with cellobiose oxidation by CDH.
Apart from a hydride-transfer mechanism, a radical mechanism is compatible with the present structure (Scheme 1, panel  b). In this reaction, one electron may be transferred from the substrate O1 to the flavin C4a or N5 concomitantly with the abstraction of the O1-H as a proton by His-689, resulting in a flavin radical and a substrate radical. The subsequent step involves a transfer of the 1-H as a hydrogen radical to N5. Although the structure is compatible with the radical mechanism, the two radical species implied have not been demonstrated. The failure to detect the radical species spectroscopi-cally does not, however, provide conclusive evidence against the radical mechanism. The formation of a cellobiosyl radical may be slow and its subsequent decomposition rapid, making detection of an electron spin resonance signal difficult. We may conclude that a physically meaningful difference between the hydride and the electron-transfer mechanisms hinges upon the temporal sequence, or concertedness, of events.
Studies on Structurally Unrelated Enzymes with Similar Substrate Specificity-The structure of soluble glucose dehydrogenase from Acinetobacter calcoaceticus in complex with its substrate, ␤-D-glucose, has been reported by Oubrie et al. (33). This enzyme catalyzes the oxidation of ␤-D-glucose to gluconolactone but uses pyrroloquinoline quinone as cofactor. Similar to what is discussed here for CDH, a hydride-transfer mechanism was assigned to soluble glucose dehydrogenase (33) based on the specific orientation of 1-H relative to the C5 in pyrroloquinoline quinone (corresponding to the flavin N5 in CDH). In soluble glucose dehydrogenase, the distance between the substrate C1 and pyrroloquinoline quinone C5 is 3.2 Å, implicating a transfer distance of 1.2 Å for 1-H. The overall structure of CDH and soluble glucose dehydrogenase (Protein Data Bank code 1CQ1) and their active sites display no obvious similarity, but nevertheless, interesting details emerge when superimposing the active sites. A superposition with reference to C1 in the two enzyme complexes aligns the C5 of the pyrroloquinoline quinone cofactor in soluble glucose dehydrogenase within 0.7 Å of the flavin N5 in CDH. The N⑀ 2 atoms of the proposed active base in soluble glucose dehydrogenase (His-144) and CDH (His-689) are only 0.4 Å apart, and the distance between Asn-732 N␦ 2 in CDH and Arg-228 N 2 in soluble glucose dehydrogenase is 1.7 Å. Thus, the precise geometry of the tetrad defined by the hydrogen acceptor of the cofactor, the C1, the proton acceptor of the general base catalyst, and the assisting residue coincide remarkably well despite different structure and cofactor dependence.
Studies on Structurally Related GMC Oxidoreductases-Although no results are available from site-directed mutagenesis studies on CDH, the residues proposed to participate in catalysis (12) have been mutated in the related enzymes GOx and ChOx. In B. sterolicum ChOx (35), Streptomyces ChOx (36), and Penicillium amagasakiense GOx (37), replacement of the proposed catalytic base (His-689 in CDH) resulted in enzyme variants with drastically reduced or abolished catalytic performance as measured by k cat , whereas K m values were practically unaffected, thus supporting the assignment of His-689 in CDH as a general base catalyst. On the other hand, mutation of the Asn-732 counterpart in Streptomyces ChOx (N480A, N480Q) and P. amagasakiense GOx (H563A, H563V) resulted in inactive enzymes. In the light of the present structure and the mutant data for GOx and ChOx, we propose a dual role for Asn-732 where it (i) helps to position the substrate with respect to the flavin, and (ii) by offering a H-bond to O1, also facilitates proton abstraction by His-689.
The only crystal structure available for a GMC oxidoreductase-ligand complex is that of B. sterolicum ChOx with bound dehydroisoandrosterone (16). In ChOx, His-447 (His-689 in CDH) has been proposed to activate a water molecule (Wat541) for nucleophilic attack on the substrate. This water molecule occupies the position of the substrate C1-O1 group in CDH. Thus, CDH and ChOx share the same reaction geometry, although the reaction in ChOx is suggested to be relayed through a water molecule. For GOx, no experimentally determined complex with substrate or substrate analogue is available, but ␤-D-glucose has been modeled in the active site (14,37), resulting in a position of the substrate relatively similar to that of the C-site glucosyl moiety of Cblm in CDH; the substrate and the SCHEME 1 catalytic residues are positioned at the re face of the flavin ring, and the C1 hydroxyl group is equidistantly positioned between His-689 and Asn-732 (His-520 and His-563 P. amagasakiense GOx). Molecular dynamics calculations of a glucose-GOx complex with a water-mediated interaction between the substrate O1 and the active histidine similar to that observed for ChOx resulted in expulsion of the water molecule, suggesting that a water-relayed mechanism is unlikely in GOx (14). Thus, direct interaction between the proposed catalytic base and the substrate is in agreement with our observed mode of Cblm binding to CDH.
Induced Fit and Water Trapping-Before another reductive half-reaction can occur, a total of two electrons acquired by the flavin needs to be transferred to an electron acceptor during the ensuing oxidative half-reaction, and the hydrogens transferred to N5 and His-689 N⑀ 2 have to be suitably disposed of. For the N5 hydrogen, the most probable destination of a proton is to bulk water concomitantly with two single-electron or a twoelectron transfer upon flavin re-oxidation (depending on the electron acceptor used). In the case of the proton withdrawn from the substrate 1-hydroxyl group by His-689 N⑀ 2 , the structure provides some hypothetical, but interesting scenarios.
As described above, the conformational change in Tyr-609, assisted by a concomitant flip of the Ser-519 side chain, orchestrates the formation of a highly ordered network of H-bonds below the ligand in site C. This imposes an effective restriction of this site to perfectly accommodate the lactam ring and thereby induce an optimal fit of the TST-like ligand to the protein. The movement of the Tyr-609 side chain into site C appears to occur in response to Cblm binding and results in the entrapment of a water molecule (Wat1366) close to His-689 C⑀ 1 (Fig. 3, a-c). In the non-liganded structure, the position of Wat1366 is occupied by the CZ-OH group of Tyr-609. The water 1366 is particularly interesting in that it is the only water molecule within H-bonding distance to the imidazole group of the proposed catalytic base in the ligand structure, and it may, thus, serve as a secondary proton acceptor. The substrate proton acquired by His-689 can be transferred to Wat1366 by a 180°rotation about 2 of the imidazole. In DH cdh ⅐Cblm, Wat1366 interacts at the center of the aromatic ring of Tyr-609, which is likely to increase its affinity for the extra proton at His-689. A H 3 O ϩ -interaction (38) with the tyrosyl ring may, thus, promote proton transfer to Wat1366 and stabilize the resulting oxonium ion. The next step may be (i) that the oxonium ion triggers the Tyr-609 side chain to swing out from site C and resume its original position and thereby displace the product from the active site, (ii) that product departure itself triggers the tyrosine side chain to leave the active site and release the oxonium ion to exchange a proton with bulk water, or (iii) that protonation of the histidine and the subsequent flip of the imidazole ring forces the product to leave due to an emerging unfavorable contact between the product carbonyl oxygen and the ⑀-1 carbon of His-689 as a consequence of C⑀ 1 assuming the position of N⑀ 2 .
Nevertheless, the tyrosine flip clearly helps to induce an optimal fit of the catalytic site for the inhibitor, and it generates a cavity for a solvent molecule that may accept a proton from the catalytic base. Wat1366 makes no direct contact with the site of oxidative attack, and hence, the C1-N5 pathway of a presumed hydride ion is completely shielded from water. It should be stressed that the position of Wat1366 in DH cdh ⅐Cblm is completely different from that of the water molecule bound in front of His-689 in the non-liganded structure. This stresses the inherent difficulty in assigning catalytic roles to active-site water molecules in the absence of ligand or when non-authentic binding of ligand occurs. Thus, the water molecule implicated in catalysis by ChOx (16) should be carefully evaluated, although the dual function of ChOx may actually justify the presence of a catalytic water positioned between the substrate and the catalytic histidine. CONCLUSIONS The crystal structure of the CDH flavoprotein with bound inhibitor provides support for a hydride-transfer mechanism for dehydrogenation with His-689 acting as a general base catalyst, deprotonating the equatorial 1-hydroxyl group. The axial 1-H is aligned for a concerted hydride transfer from C1 to N5 via a transition state characterized by partial planarization of C1. The binding of the inhibitor is in agreement with hydrogen transfer at the re side of the flavin ring. We also suggest that Asn-732 is bifunctional in that it both participates in the formation of a productive enzyme-substrate complex and that it supports deprotonation by His-689 by serving as a H-bond donor to the 1-hydroxyl group. From a structural and biochemical viewpoint, the reaction mechanism is simple and requires only minor structural changes in the substrate and protein in order for the 1-H to be expelled. The 1-hydrogen would then need to traverse a distance of less than 1 Å for covalent attachment to the flavin N5 atom. Although the perfect setup of the active site for hydride transfer is evident, we cannot rule out the possibility of a radical mechanism using structural data alone. The results provide a structural platform for the use of conventional biochemical and biophysical techniques as well as quantum mechanical and classical molecular mechanics approaches to further investigate the molecular mechanism of cellobiose oxidation by CDH.