Structural rationalization for the lack of stereospecificity in coenzyme B12-dependent diol dehydratase.

Adenosylcobalamin-dependent diol dehydratase of Klebsiella oxytoca is apparently not stereospecific and catalyzes the conversion of both (R)- and (S)-1,2-propanediol to propionaldehyde. To explain this unusual property of the enzyme, we analyzed the crystal structures of diol dehydratase in complexes with cyanocobalamin and (R)- or (S)-1,2-propanediol. (R)- and (S)-isomers are bound in a symmetrical manner, although the hydrogen-bonding interactions between the substrate and the active-site residues are the same. From the position of the adenosyl radical in the modeled "distal" conformation, it is reasonable for the radical to abstract the pro-R and pro-S hydrogens from (R)- and (S)-isomers, respectively. The hydroxyl groups in the substrate radicals would migrates from C(2) to C(1) by a suprafacial shift, resulting in the stereochemical inversion at C(1). This causes 60 degrees clockwise and 70 degrees counterclockwise rotations of the C(1)-C(2) bond of the (R)- and (S)-isomers, respectively, if viewed from K+. A modeling study of 1,1-gem-diol intermediates indicated that new radical center C(2) becomes close to the methyl group of 5'-deoxyadenosine. Thus, the hydrogen back-abstraction (recombination) from 5'-deoxyadenosine by the product radical is structurally feasible. It was also predictable that the substitution of the migrating hydroxyl group by a hydrogen atom from 5'-deoxyadenosine takes place with the inversion of the configuration at C(2) of the substrate. Stereospecific dehydration of the 1,1-gem-diol intermediates can also be rationalized by assuming that Asp-alpha335 and Glu-alpha170 function as base catalysts in the dehydration of the (R)- and (S)-isomers, respectively. The structure-based mechanism and stereochemical courses of the reaction are proposed.

Kinetic measurements provided some clues to solve the enigma of apparent lack of stereospecificity of diol dehydratase. When each enantiomer of 1,2-propanediol is run independently, the rate with the (R)-isomer is 1.7-1.8 times higher than that with the (S)-isomer (3)(4)(5). However, when racemic 1,2-propanediol is used as substrate, the (S)-isomer reacts at a faster rate than the (R)-isomer (6). This reversal is because of the ratio of K m values (K m (R)/K m (S) ϭ 3.1-3.2). These lines of evidence suggested that the (R)-and (S)-isomers are bound to the enzyme in two different modes with different catalytic efficiency and binding affinity. In other words, the enzyme recognizes the enantiomers as "different" substrates.
The stereochemistry of the diol dehydratase reaction established by the labeling experiments is summarized as follows (Fig. 1A). It was shown by Rétey et al. (7) that [ 18 O]-and unlabeled propionaldehydes are formed from [1][2][3][4][5][6][7][8][9][10][11][12][13][14][15][16][17][18] O]-(S)-and [1-18 O]-(R)-1,2-propanediols, respectively. This indicated that 1,1-gem-diol is formed as an intermediate. The initial migration of an OH group from C(2) to C(1) is stereoselective, and the dehydration of the resulting gem-diol undergoes steric control by the enzyme, with only one of the two OH groups on the prochiral center being eliminated. However, the absolute configurations of the gem-diol intermediates and why and how a specific hydroxyl group undergoes elimination remained to be elucidated. The hydrogen atom moves to the adjacent carbon atom without exchange with solvent protons (8). Abeles and co-workers (3) demonstrated that the pro-S and pro-R hydrogen atoms on C(1) of (S)-and (R)-1,2-propanediols, respectively, migrate to C (2). The migrating OH group is replaced by the hydrogen atom with an accompanying inversion of the configuration at C(2) (3,9). However, when the enantiomeric pair of ethylene glycols labeled stereospecifically with deuterium and tritium is used as substrates, racemic products are obtained (10). This indicated that rapid internal rotation in the intermediate occurs before the hydrogen recombination. Abeles and co-workers (11)(12)(13)(14) demonstrated that the enzyme-bound AdoCbl serves as an intermediate hydrogen carrier, first accepting a hydrogen atom from C(1) of the substrate to C(5Ј) of the coenzyme and then, in a subsequent step, giving a hydro-gen back to C(2) of the product. 5Ј-Deoxyadenosine was postulated to be an intermediate (11,15,16). The formation of cob(II)alamin and an organic radical intermediate during catalysis was demonstrated by optical and electron paramagnetic resonance spectroscopies (17)(18)(19)(20)(21)(22).
From these lines of evidence, the minimal mechanism for diol dehydratase was established (Fig. 1, B and C) (21,(23)(24)(25). In the diol dehydratase reaction, X and H are the OH group on C(2) and a hydrogen atom on C(1), respectively, and a water molecule is subsequently eliminated from a gem-diol intermediate. This has now been accepted as a general mechanism for all the AdoCbl-dependent intramolecular group-transfer reactions (26,27), in which a hydrogen atom migrates from one carbon atom of the substrate to an adjacent carbon atom in exchange for group X that moves in the opposite direction. That is, the enzyme-coenzyme interaction leads to the activation of the Co-C bond of the coenzyme. Substrate binding triggers the homolysis of the Co-C bond, forming the adenosyl radical and cob(II)alamin. The adenosyl radical abstracts a hydrogen atom from the substrate, producing a substrate-derived radical and 5Ј-deoxyadenosine. The substrate radical rearranges to the product radical, which then abstracts a hydrogen atom back from 5Ј-deoxyadenosine. This leads to the formation of the product and regeneration of the coenzyme.
The three-dimensional structures of AdoCbl-dependent enzymes have been solved to understand their fine mechanisms of action (28 -32). We have determined the crystal structures of diol dehydratase in complexes with CN-Cbl (29,33) or adeninylpentylcobalamin (34) with (29,34) and without (33) substrate 1,2-propanediol. The O(1) and O(2) of the substrate are hydrogen-bonded to the respective pair of amino acid residues, namely (Glu-␣170 and Gln-␣296) and (Asp-␣335 and His-␣143), respectively (29). The 1,2-propanediol bound at the active site was assigned to the (S)-isomer, although the crystals were grown in the presence of racemic substrate. The structure-based mechanism of reaction with the (S)-isomer has been proposed (29,33,34).
To confirm this assignment and to solve the mechanism of reaction with the (R)-isomer, we crystallized the diol dehydratase⅐CN-Cbl complexes in the presence of (R)-and (S)enantiomers and analyzed their three-dimensional structures. The conformations of the enzyme-bound 1,1-gem-diol intermediates from enantiomeric substrates were modeled. Based on the crystal structures and the models, we elucidated the stereochemical courses of the diol dehydratase reaction with the (R)and (S)-isomers. We propose here the reason why diol dehydratase is apparently not stereospecific for the enantiomers, although hydrogen abstraction and recombination, as well as OH group migration, proceed in stereospecific manners.

EXPERIMENTAL PROCEDURES
Materials-(R)-1,2-Propanediol was purchased from TCI, Tokyo, Japan. (S)-1,2-Propanediol was synthesized by reduction of L-lactide (TCI, Tokyo, Japan) with LiAlH 4 and purified by distillation under reduced pressure. Crystalline AdoCbl was a gift from Eizai, Tokyo, Japan. All other chemicals were reagent grade commercial products and used without further purification.
Purification of Enzyme and Substitution of Racemic 1,2-Propanediol by Enantiomeric Substrates-Diol dehydratase of Klebsiella oxytoca ATCC 8724 was overexpressed in Escherichia coli and purified by the following modification of the method described previously (35,36). The enzyme extracted from the crude membrane fractions with 1% Brij35 was loaded onto a DEAE-cellulose column, and the column was washed with 0.01 M potassium phosphate buffer (pH 8.0) containing 2% racemic 1,2-propanediol and 1% phenylmethanesulfonyl fluoride, as described previously. The column was washed again with 0.01 M potassium phosphate buffer (pH 8.0) containing 2% (R)-or (S)-1,2-propanediol and then eluted with the same buffer containing 0.15 M KCl, 2% (R)-or (S)-1,2propanediol, and 20 mM sucrose monocaprate.
Data Collection-For both complexes, the diffraction data sets were collected at the BL41XU beamline, SPring-8, Japan. The crystal of (S)-1,2-propanediol-bound form was transferred to a cryoprotectant solution containing 15% PEG20000, 17.5% ethylene glycol, and other components of the protein drop except 1,2-propanediol and PEG6000, which had been used for the crystals grown in the presence of racemic 1,2-propanediol. In the case of (R)-isomer, crystals were dissolved in several cryoprotectant solutions available, but the addition of PEG400 directly to a protein drop to a final concentration of ϳ20% kept crystals intact. For both forms, the crystal was transferred into a cryo-stream with a cryo-loop and was flash-cooled at 100 K. A total of 180°of data were measured. All diffraction data were processed and scaled using DENZO and SCALEPACK (38).
Structure Determination-Two kinds of starting models, substratebound (29) at 100 K and substrate-free (33) forms of the enzyme⅐CN-Cbl complex, for structure determination were used for the first refinement step to cross-check influence of model bias of (S)-1,2-propanediol molecule that had been assigned to the substrate-bound form so far. Coordinates of (S)-1,2-propanediol were deleted from the original substratebound structure. Refinement and model building were carried out with CNS (39) and XFIT (40). After rigid-body refinement, 2mF o Ϫ DF c electron density maps of (R)-1,2-propanediol from both starting models clearly indicated that C(3) atom of (R)-1,2-propanediol shares the similar position to that of (S)-1,2-propanediol. As expected, the (S)-1,2propanediol-bound form clearly displayed the same binding mode as observed in the structures crystallized in the presence of racemic 1,2-propanediol (29,34). Thus we could assign the initial coordinates of substrate molecule for both forms. Each model was built with rounds of manual rebuilding, positional refinement, and individual B-factor refinement, using all data between 30 and 2.3 Å for the (R)-1,2-propanediol-bound form and between 30 and 1.8 Å for the (S)-1,2-propanediol-bound form. Non-crystallographic symmetry restraint was applied for the (R)-1,2-propanediol-bound form. Refinement statistics are shown in Table I.
Geometry Calculation and Modeling Study-The first ␣␤␥ unit of the (␣␤␥) 2 heterohexamer was used for geometry calculation and modeling study of the active-site vicinity of the crystal structures published so far, because the average B-factor of the first unit is much lower than that of the second unit. In this paper, that is the case for the (S)-1,2-propanediolbound form, 23.1 Å 2 for the first unit and 36.0 Å 2 for the second unit. In the case of (R)-isomer-bound form, however, the average B-factor of the (R)-1,2-propanediol molecule of the first unit (25.8 Å 2 ) is slightly higher than that of the second unit (23.5 Å 2 ), and the overall B-factor of the first ␣␤␥ unit (37.6 Å 2 ) is slightly lower than that of the second unit (39.4 Å 2 ), but they are comparable with each other. The part of the final 2mF o Ϫ DF c electron density map covering the second substrate is clearer than that of the first molecule. The second unit, therefore, is used for geometry calculation and modeling study in this paper.
The coordinates of the adenosyl group in the "distal" conformation are derived from the superimposed model of the adenosyl group on the

Modes of Binding of (R)-and (S)-1,2-Propanediols to Diol
Dehydratase-The x-ray structures of diol dehydratase⅐CN-Cbl complexes crystallized in the presence of (R)-and (S)-1,2-propanediols were determined at 2.3-and 1.8-Å resolutions, respectively (Table I). Overall structures of both complexes were essentially identical to each other except for the substrates bound at the active site. As shown in Fig. 2B, the configuration and the conformation of the (S)-1,2-propanediol fitted to the electron density map were the same as those in the enzyme⅐CN-Cbl complex crystallized in the presence of racemic 1,2-propanediol (29). This provided a definite proof for our previous assignment of the (S)-isomer to the electron density map. In contrast, the electron density contoured for the (R)enantiomer was quite different from that for the (S)-isomer ( Fig. 2A). It rather resembles that corresponding to the 1,2propanediol that is bound to the active site of glycerol dehydratase (32), an isofunctional AdoCbl-dependent enzyme.
It should be noted that the hydrogen-bonding interactions between substrate and the active-site amino acid residues are the same in the (R)-and (S)-1,2-propanediol-bound forms of the enzyme. That is, O(1) is hydrogen-bonded to -COO Ϫ of Glu-␣170 and N⑀2 of Gln-␣296 and O(2) to -COO Ϫ of Asp-␣335 and N⑀2 of His-␣143. Previously, we assumed that (R)-and (S)-1,2propanediols are bound to the active site in the opposite orientation as mirror images with their C(1)-C(2) bonds aligned and that Glu-␣170 and Asp-␣335 might at least partially share a common catalytic function (43). Although this pseudo-symmetry was not the case, it became evident that there is another plane of symmetry relating the (R)-and (S)-substrates bound by diol dehydratase.
Stereospecificity of Hydrogen Abstraction from Substrates-Abeles and co-workers (3) demonstrated that the pro-R and pro-S hydrogen atoms on C(1) of (R)-and (S)-1,2-propanediols, respectively, migrate to C(2). The enzyme-bound AdoCbl serves as an intermediate hydrogen carrier, first accepting a hydrogen atom from C(1) of the substrate to C(5Ј) of the coenzyme and then, in a subsequent step, giving a hydrogen back to C(2) of the product (11)(12)(13)(14). Fig. 3, A and B shows the structures of the active sites of the (R)-and (S)-isomer-bound forms of the enzyme⅐cobalamin complex, respectively. The position of the adenosyl radical in the distal conformation was obtained by the modeling study, as described previously (34); that is, the adenine moiety of the coenzyme adenosyl group was superimposed on that of the enzyme-bound adeninylpentylcobalamin, and the ribose moiety was rotated around the glycosidic linkage so that C(5Ј) could come closest to C(1) of the (R)-and (S)-isomers.  (7), stereoselective migration of an OH group from C(2) to C(1) followed by stereospecific dehydration was suggested, although the absolute configurations of 1,2gem-diol intermediates are as yet unknown. Fig. 2, A and B shows the conformations of the enzymebound (R)-and (S)-enantiomers viewed from C(2) along the C(2)-C(1) bond. The torsion angle of O(1)-C(1)-C(2)-O(2) for the (R) and (S)-isomers are Ϫ48.2 and 53.8°, respectively, indicating that both are in the so-called staggered conformations. In Fig. 4, A and B, the conformations of the 1,2-propanediol-1-yl radicals formed by abstraction of specific hydrogen atoms from the substrates are depicted in Newman projections. It is quite reasonable to assume that the OH group in the radical intermediates migrates from C(2) to C(1) by a suprafacial shift, because the energy along this pathway would be minimized by overlapping of the radical p orbital and the O(2)-C(2) orbital. It is thus predictable that the OH group migrates from C(2) to C(1) with the inversion of the configuration at C(1); that is, the 2-OH groups of the (R)-and (S)-isomers become the pro-S and pro-R OH groups on C(1) of the resulting 1,1-gem-diol intermediates, respectively. The stereochemical inversion at C(1) in the substitution of a hydrogen atom by the migrating OH group has been postulated, although not proven, by Diziol et al. (44) and Smith et al. (45).
Modeling Study of 1,1-gem-Diol Intermediates and Hydrogen Recombination-Is it structurally feasible for the 1,1-gem-diol-2-yl radical, the product radical in the diol dehydratase reaction, to back-abstract (recombine) a hydrogen atom from the methyl group of 5Ј-deoxyadenosine? How is it possible to re- place the migrating OH group by a hydrogen atom of the methyl group of 5Ј-deoxyadenosine with the inversion of the configuration at C(2) (3,9)? To solve these problems, the structures of the 1,1-gem-diol intermediates from the (R)-and (S)-1,2-propanediols were modeled.
Because the hydrogen bonds are much stronger than hydrophobic or van der Waals interactions between the C-C-C backbone of the substrates and the enzyme, the OH group on C(2) would migrate to C(1) while maintaining hydrogen bonding to the amino acid residues (34). Thus, if viewed from the position of K ϩ , the C(1)-C(2) bond of the (R)-isomer would rotate clockwise by about 60°around the axis connecting K ϩ and the center of the O(1)-O(2) line in the step of 1,2-OH group migration (Fig. 4A). On the contrary, the C(1)-C(2) bond of the (S)-isomer would rotate counterclockwise by about 70°around the same axis in the 1,2shift of the OH group (Fig. 4B). These considerations enabled us to model the 1,1-gem-diol intermediates that are formed from the (R)-and (S)-1,2-propanediols. If the ribose moiety and the 1,1gem-diol intermediates are rotated around the glycosidic linkage and the O(1)-O(2) line, respectively, so that C(2) could come closest to C(5Ј) of the enzyme-bound 5Ј-deoxyadenosine, the structures shown in Fig. 5, A and B were obtained. Again, the postulated 1,1-gem-diol intermediates from (R)-and (S)-enantiomers are considered to be bound to the active site in a symmetrical mode with respect to the plane including K ϩ , O(1), and O(2). It should be noted that the CH 3 group of 5Ј-deoxyadenosine is on the symmetrical plane including K ϩ , O(1), and O(2) and thus is accessible to C(2), a new radical center, of both 1,1-gem-diol radicals. The distances from C(5Ј) to C(2) of the modeled 1,1-gemdiol radicals from (R)-and (S)-1,2-propanediols are 2.58 and 2.55 Å, respectively. Therefore, it can be concluded that the hydrogen back-abstraction (recombination) from the CH 3 group of 5Ј-deoxyadenosine by the product radical is structurally quite feasible.
Because the CH 3 group of 5Ј-deoxyadenosine is positioned on the opposite side to the OH group that migrated from C(2) to C(1), it is reasonable to predict that the hydrogen atom is abstracted from the CH 3 group of 5Ј-deoxyadenosine by C(2)centered product radicals with the inversion of the configuration at C(2) with both enantiomers of substrate. This prediction is also just as expected from the experimental results (3,9). In contract, ethylene glycols stereospecifically labeled with deuterium and tritium are converted to acetaldehyde with racemization (10), suggesting rapid internal rotation in the product radical with this substrate before hydrogen back-abstraction. In ethanolamine ammonia-lyase, racemization takes place with asymmetrically deuterated ethanolamines (46), which is explained by the torsion symmetry of the trigonal intermediates arising from this substrate. (R)-and (S)-2-aminopropanols are deaminated with the inversion and retention of configuration at C(2), respectively (44), suggesting that the rotameric intermediate formed from the (S)-enantiomer is favored in equilibration. It is likely that, unlike ethanolamine ammonialyase, the active site of diol dehydratase does not permit either symmetrization or equilibration of the 1,1-gem-propanediol-2-yl radical intermediates.
Dehydration of 1,1-gem-Diol Intermediates-It has been reported by Rétey et al. To explain this fact, not only the 1,2-OH group migration is stereoselective, but also the dehydration of the resulting 1,1-gem-diols must be stereospecific. Active-site residues would catalyze the dehydration of the 1,1gem-diol intermediates by serving as a proton acceptor and a proton donor. We have proposed the contributions of the Glu-␣170 -His-␣143 pair to the dehydration of the intermediate from the (S)-1,2-propanediol (34,43). In this case, the oxygen atom originated from O(2) would be lost in water, and the oxygen atom originated from O(1) would be retained in the product as the CϭO oxygen. Radom and co-workers (45) postulated that the Glu-␣170 -His-␣143 pair participates in the dehydration of 1,1-gem-diols from both (R)-and (S)-isomers. If this is true, exchanges of the hydrogen-bonding partners for O(1) and O(2) have to take place to explain the different fates of 18 O on C(1) between the (R)-and (S)-1,2-propanediols. However, this seems less likely, because it must be energetically more favorable for the hydrogen abstraction and recombination as well as the 1,2-OH group migration to proceed while keeping the substrates and intermediates hydrogen bonded to the same amino acid residues. It should be noted that the equilibrium between 1,1-gem-diol and the corresponding aldehyde in aqueous solution is very rapid, and the activation energy for the dehydration of 1,1-gem-diol to aldehyde is quite small, especially in the presence of acid or base catalyst. Therefore, the -COO Ϫ groups of Glu-␣170 and Asp-␣335 might share a common catalytic function as a general base; that is, the Asp-␣335-Gln-␣296 pair might also function effectively for the dehydration of a 1,1-gem-diol intermediate. In this case, the oxygen atoms originated from O(1) and O(2) would be lost in the solvent and retained in the product, respectively.
In the dehydration, deprotonation of an OH group and elimination of the other OH group would proceed in a concerted manner. The energy for the dehydration (formation of CϭO) would thus be lowered by overlapping of the cleaving OH and the C(1)-O orbitals. The extent of overlapping of the orbitals would be dependent on the torsion angle of H-O-C(1)-O eliminating and be minimum and maximum at 90 and 0 or 180°, respectively. Fig. 6, A and B shows the conformation of the modeled 1,1-gemdiol intermediates formed from the (R)-and (S)-enantiomers, respectively, that are viewed from O(1) along the O(1)-C(1) bond. Fig. 6, C and D indicates the same conformations of the intermediates from the (R)-and (S)-isomers, respectively, that are viewed from O(2) along the O(2)-C(1) bond. Conformations of 1,1-gem-diols are schematically illustrated to the right of Fig. 6, A-D in Newman projections. From Fig. 6, A and C, the torsion angles of H-O-C(1)-O eliminating for the 1,1-gem-diol from the (R)-isomer are 63.8 and 148.6°, respectively, when Glu-␣170 and Asp-␣335 serve as bases. Thus, the dehydration by the Asp-␣335-Gln-␣296 pair must be energetically more favorable. It can thus be concluded that, in the dehydration of the 1,1-gem-diol from the (R)-isomer, Asp-␣335 and Gln-␣296 participate as proton acceptor and donor, respectively, and that the oxygen atoms originated from O(1) and O(2) would be lost in the solvent and retained in the product as the CϭO oxygen, respectively. From Fig. 6, B and D, on the other hand, the torsion angles of H-O-C(1)-O eliminating for the 1,1-gem-diol from the (S)-isomer are 127.0 and 87.4°, respectively, when Glu-␣170 and Asp-␣335 serve as bases. Thus, the dehydration by the Glu-␣170 -His-␣143 pair is energetically more favorable. In the dehydration of the 1,1gem-diol from the (S)-isomer, Glu-␣170 and His-␣143 would function as proton acceptor and donor, respectively. The oxygen atoms originated from O(1) and O(2) would be retained in the product as the CϭO oxygen and lost in the solvent, respectively. These predictions are completely consistent with the experimental results of Rétey et al. (7). In conclusion, the pro-R and pro-S OH groups on C(1) of the 1,1-gem-diols from both enantiomers are lost in the solvent and retained in the product aldehyde, respectively.
Overall Mechanism and Stereochemical Course of Diol Dehydratase Reaction-Based on the crystal structures and modeling studies reported in this paper, we propose here a refined overall mechanism for diol dehydratase. The reaction pathways with both (R)-and (S)-1,2-propanediols as substrates are illustrated in Fig. 7. The K ϩ ion in the substrate-free enzyme (1) is hepta-coordinated by five oxygen atoms from the active-site amino acid residues and two oxygen atoms of water molecules (33). As judged from the modeling study on the distortions of the Co-C distance and bond angles using the crystal structure of the substrate-free form of diol dehydratase, it is likely that major conformational changes of AdoCbl takes place upon its binding to apoenzyme even in the absence of substrate, leading to the activation of the coenzyme Co-C bond. At this stage, however, only a small fraction of the enzyme-bound coenzyme is in the dissociated form (5,20,22), because rapid recombination of the adenosyl radical in the "proximal" conformation (34) and cob(II)alamin occurs even if they are formed. The addition of substrate to the holoenzyme brings about a ligand exchange reaction with two water molecules on K ϩ (from 1 to 2). Upon the substrate binding, small but distinct conformational changes take place that trigger the Co-C bond homolysis forming cob(II)alamin and the adenosyl radical in the proximal conformation (33) (2). The adenosyl radical undergoes rotation of the ribose moiety around the glycosidic linkage to the distal conformation (3) (34). As reported in this paper, the (R)-and (S)-isomers of substrate are bound to the active site in a symmetrical mode with respect to the plane including K ϩ and the two oxygen atoms of the substrate. Because C(5Ј), the radical center of the adenosyl radical in the distal conformation, comes on this symmetrical plane, it can abstract the specific hydrogen atoms of the enantiomeric substrates; that is, the pro-R and pro-S hydrogens of the (R)-and (S)-isomers, respectively. The substrate-derived 1,2-diol-1-yl radicals and 5Ј-deoxyadenosine are thus formed (4). The OH group of the substrate radicals then migrates from C(2) to C(1) by a suprafacial shift, resulting in their rearrangement to the product-derived 1,1-diol-2-yl radicals (6) through cyclic transition states (5). The 1,2-OH group shift by a concerted mechanism has been predicted by theoretical calculations (45,(47)(48)(49)(50)(51). The OH groups on C(2) of the (R)and (S)-isomers become the pro-S and pro-R OH groups on C(1) of the product radical, respectively. Because the hydrogen bonds are stronger than hydrophobic or van der Waals interactions, it would be reasonable to assume that the C(1)-C(2) bonds of the (R)-and (S)-isomers rotate clockwise by 60°and counterclockwise by 70°if viewed from K ϩ , respectively, around the axis connecting K ϩ and the center of the O(1)-O(2) line in the step of 1,2-OH group migration. C(2) of the product radicals, the new radical center, then comes close to the CH 3 group of 5Јdeoxyadenosine and back-abstracts a hydrogen atom from it, producing 1,1-gem-diols and the adenosyl radical (7). Because the CH 3 group of 5Ј-deoxyadenosine is positioned on the opposite side to the migrating OH group, this hydrogen recombination proceeds with the inversion of the configuration at C(2). The dehydration of the resulting 1,1-gem-diols would proceed by concerted deprotonation and protonation. By taking the torsion angles of H-O-C(1)-O eliminating in the modeled structures into considerations, it can be concluded that the -COO Ϫ groups of Asp-␣335 and Glu-␣170 play a role as a base catalyst in the dehydration of the 1,1-diols that are derived from the (R)-and (S)-isomers of substrate, respectively. Propionaldehyde formed (8) loses binding affinities for K ϩ and the active-site residues and is displaced from the active site by a water molecule. This causes a conformational change back to the substrate-free form (1), which leads to the recombination of the adenosyl radical in the proximal conformation with cob(II)alamin. As a result, the coenzyme is regenerated, and an energy liberated upon reformation of the Co-C bond would be utilized for ensuring the final steps of the reaction (from 8 to 1). This structure-based mechanism and the stereochemical courses of the diol dehydratase reaction account for all the results of the biochemical and mutational experiments with this enzyme reported so far.