Reaction Mechanism of Hydroxynitrile Lyases of the α/β-Hydrolase Superfamily

The hydroxynitrile lyases (HNLs) from Hevea brasiliensis (HbHNL) and from Manihot esculenta (MeHNL) are both members of the α/β-hydrolase superfamily. Mechanistic proposals have been put forward in the past for both enzymes; they differed with respect to the role of the active-site lysine residue for which a catalytic function was claimed for the Hevea enzyme but denied for the Manihot enzyme. We applied a freeze-quench method to prepare crystals of the complex of HbHNL with the biological substrate acetone cyanohydrin and determined its three-dimensional structure. Site-directed mutagenesis was used to prepare the mutant K236L, which is inactive although its three-dimensional structure is similar to the wild-type enzyme. However, the structure of the K236L-acetone cyanohydrin complex shows the substrate in a different orientation from the wild-type complex. Finite difference Poisson-Boltzmann calculations show that in the absence of Lys236 the catalytic base His235 would be protonated at neutral pH. All of this suggests that Lys236 is instrumental for catalysis in several ways, i.e. by correctly positioning the substrate, by stabilizing the negatively charged reaction product CN-, and by modulating the basicity of the catalytic base. These data complete the elucidation of the reaction mechanism of α/β-hydrolase HNLs, in which the catalytic triad acts as a general base rather than as a nucleophile; proton abstraction from the substrate is performed by the serine, and reprotonation of the product cyanide is performed by the histidine residues. Together with a threonine side chain, the active-site serine and lysine are also involved in substrate binding.

Hydroxynitrile lyase, HNL 1 (EC 4.1.2.39), catalyzes the cleavage of cyanohydrins to hydrocyanic acid plus the corre-sponding aldehyde or ketone (see Fig. 1A). The release of HCN serves as a defense against herbivores and microbial attack for a variety of plants (1)(2)(3) and may also serve as a nitrogen source for the biosynthesis of asparagine (4,5). In an aqueous solution, cyanohydrin cleavage occurs spontaneously above pH 5, whereas the enzyme reaction also occurs at lower pH values (down to pH ϳ3, Ref. 2). This is technologically exploited for the enantioselective synthesis of chiral cyanohydrins (6 -10), making use of the reverse in vivo reaction. Moreover, this enzyme is an interesting target for the detoxification of cyanogenic food crops forming a potential health risk to its consumers mainly in the third world (11).
Biologically, HNLs constitute an interesting example of convergent evolution (3), as there are at least four completely unrelated classes of proteins with HNL activity belonging to such different fold families as ␣/␤-hydrolases (12,13), FAD-dependent oxidoreductases (14,15), carboxypeptidases (16), and zinc-dependent alcohol dehydrogenases (17). In line with the structural dissimilarity between different HNLs, the activesite architectures are also vastly different, despite similar or identical biological substrates. Moreover, structural similarities with the family of highest sequence homology (e.g. ␣/␤hydrolases, GMC-oxidoreductases, and carboxypeptidases) are not paralleled by similarities in the type and mechanism of the catalyzed reaction.
The HNL from the tropical rubber tree Hevea brasiliensis (HbHNL) belongs to the ␣/␤-hydrolase superfamily (12,13,18). Its three-dimensional structure (Fig. 1B) is known (18), as well as the (very similar, C␣ r.m.s. deviation ϭ 0.42 Å) structure of the highly homologous (77% sequence identity) enzyme from Manihot esculenta (MeHNL, Protein Data Bank code 1dwp). From the three-dimensional structure and from mutagenesis data (20) the active-site residues could be unequivocally identified. They include a catalytic triad consisting of Ser 80 , His 235 , and Asp 207 (18). 2 However, several aspects of the molecular mechanism of the HbHNL-or MeHNL-catalyzed cyanohydrin reaction remained controversial.
Prior to the publication of the three-dimensional structure of HbHNL, sequence and mutational analysis led to the proposition of a catalytic mechanism for the M. esculenta enzyme involving general acid/base catalysis (12). The close structural similarity to the ␣/␤-hydrolases that emerged from the crystal structure of the H. brasiliensis enzyme suggested a mechanism accepted for hydrolases with a covalently bound reaction intermediate (18). This hypothesis was subsequently dismissed on the basis of the structural data on HbHNL-inhibitor complexes (21). In a revised mechanistic proposition (see Fig. 2), residues of the catalytic triad act as a general acid/base, involving (for the cyanohydrin cleavage direction) deprotonation of the OH-Ser 80 by His 235 and concomitant abstraction of a proton from the substrate hydroxyl by Ser 80 . Following cyanohydrin cleavage, His 235 reprotonates the cleaved cyanide. Modeling studies also suggested a role for the activesite residue Lys 236 (22), in which the positive charge was proposed to stabilize the complementary negative charge that is built up on the cyano group.
The crystal structure of the Manihot enzyme was solved by molecular replacement, using the three-dimensional structure of the HbHNL enzyme as a search model. Structural studies were reported for the complex of acetone and trichloroacetone with the native enzyme (23) as well as for the (inactive) S80A mutant in complex with the substrate acetone cyanohydrin (24). They have led to similar but not completely identical mechanistic proposals (see Fig. 3). Although the role of the residues of the catalytic triad is the same in the two mechanisms, there is disagreement with respect to the role of the active-site lysine residue (Lys 236 in the Hevea enzyme or Lys 237 in the Manihot enzyme). A catalytic function for this residue was ruled out for the Manihot enzyme on the basis of the three-dimensional data for the S80A mutant in complex with the substrate acetone cyanohydrin (24).
Both mechanistic proposals suffer from the shortcoming that the three-dimensional structure of the native enzyme with a substrate cyanohydrin has so far been unknown for both HNLs. In view of the rapid turnover, crystallographic observation of such complex structures was deemed a difficult challenge (24). The HNL mechanisms were therefore based on modeling studies (HbHNL) (22) and on the crystallographic observation of complexes between both native enzymes and the biological reaction product acetone (21,23) as well as on the complex between acetone cyanohydrin and the inactive S80A mutant of MeHNL (24). Because the interaction of the (mutated) Ser 80 with the substrate is a central aspect of both mechanisms, conclusions derived from the binding of the substrate to the mutant enzyme have limited significance.
The controversial and open points about the mechanism of HNLs with ␣/␤-hydrolase structure form the subject of the present work. We have developed a method to prepare and freeze-quench complexes between native HbHNL crystals and cyanohydrin substrates and report the crystal structure of the complex with the biological substrate acetone cyanohydrin. Secondly, the mutant K236L was produced by site-directed mutagenesis. It shows no detectable catalytic activity. Its three-dimensional structure shows very little structural difference to the wild-type enzyme. Finally, the structure of the complex between the inactive K236L mutant and acetone cyanohydrin shows that Lys 236 is also responsible for correctly positioning the substrate. Finite difference Poisson-Boltzmann calculations were carried out to rationalize the pivotal role of Lys 236 for the HNL-catalyzed cyanohydrin reaction.
For heterologous expression of HbHNL Pichia pastoris X33 and vector pGAPZB (Invitrogen) were used. P. pastoris strains were cultivated in yeast extract-peptone-dextrose medium, and for selection 100 mg/liter zeocin were added when needed.
DNA Manipulations and Sequence Analysis-DNA manipulations, plasmid transformation into E. coli, and retrieval of DNA from E. coli were performed following standard procedures (25). Transformation of P. pastoris was done by electroporation (26). Sequencing of doublestranded DNA was carried out according to the dideoxy chain termination procedure using an ABI 373 DNA sequencer, the Dye Deoxy Terminator sequencing kit (Applied Biosystems Inc., Faster, CA) and AmpliTaq DNA polymerase (PerkinElmer Life Sciences) (27).
Site-directed Mutagenesis of the K236L Mutant-Specific mutants were prepared using the QuikChange TM site-directed mutagenesis kit (Stratagene, La Jolla, CA) according to the supplied instruction manual. The recombinant plasmid pHNL104 (13), which contains the cDNA of the H. brasiliensis hnl wild-type gene, was used as the template for the mutagenesis reaction. The following oligonucleotides were used to introduce the desired mutations to generate K236L mutant protein: K236L forward, 5Ј-GGTCGAAGGTGGAGATCATCTCCTGCAGCTTA-CAAAGACTAAGG-3Ј and K236L reverse, 5Ј-CCTTAGTCTTTGTAAG-CTGCAGGAGATGATCTCCACCTTCGACC-3Ј. Apart from the amino acid exchange at position 236 of the Hnl protein, a cleavage site for the restriction enzyme PstI was additionally introduced by way of a silent mutation. The temperature profile for plasmid amplification was set as follows: a single denaturation step of 95°C for 1 min followed by 16 cycles of denaturation at 95°C for 30 s, annealing at 55°C for 1 min, and elongation at 68°C for 8 min. After DpnI digestion, a 1-l aliquot of the sample was transformed into E. coli. Plasmid DNA of transformants was checked by restriction analysis, and base replacement was confirmed by sequence analysis. The plasmid containing the mutant gene was named pHNL-K236L.
Construction of P. pastoris HbHNL-K236L Expression Strain-The EcoRI fragment of pHNL-K236L containing the mutated hnl gene was subcloned into the unique EcoRI site of plasmid pGAPZB. Successful insertion was confirmed by restriction analysis and partial sequencing of the resulting expression plasmid pGAPZ-HNL-K236L.
Isolated DNA of pGAPZ-HNL-K236L was then linearized with BglII and electroporated into P. pastoris X33. Transformants were selected for zeocin resistance and tested for their ability to express the Hnl protein, and a suitable expression strain was selected. The integration of the hnl cDNA into the P. pastoris genome and the presence of the specific mutation in the expression strain P. pastoris X33 pGAP-K236L were checked by sequencing a DNA fragment that was amplified by colony PCR using primers specific for the 5Ј-and 3Ј-flanking regions of the hnl gene.
Production and Purification of Recombinant HbHNL-K236L Enzyme-50 ml of YPD in a 300-ml baffled shake flask were inoculated with an overnight culture of the strain P. pastoris X33 pGAP-K236L and incubated overnight at 30°C with shaking. This preculture was then used to inoculate 300 ml of YPD medium in a 2-liter baffled flask to an A 600 of 0.5. This culture was incubated at 30°C with shaking until the A 600 reached 40 (ϳ24 h).
P. pastoris cells were harvested by centrifugation, cell pellets were resuspended in breaking buffer (50 mM sodium phosphate, pH 7.4, 1 mM EDTA, 5% glycerol), and after the addition of an equal amount of acid-washed glass beads cells were disrupted using a Merckenschlager homogenizer. The cytosolic protein fraction was prepared by ultracentrifugation (90,000 g), and HbHNL-K236L protein was purified by anion exchange chromatography on a Resource Q column (Amersham Biosciences). The purity of the isolated HNL protein was determined by SDS-PAGE (28). Protein from the selected fractions was concentrated to 15 mg/ml using Vivaspin filtration units (10-kDa size exclusion, Sartorius, Germany). This enzyme preparation was further purified by size exclusion chromatography using a Superdex 200 column (Amersham Biosciences) and 50 mM phosphate-loading buffer. Protein was concentrated as described above and desalted by dialysis against 10 mM Hepes, pH 7.0. Protein concentrations were determined spectrophotometrically at 595 nm by the Bio-Rad protein assay reagent kit based on the method of Bradford (29) using bovine serum albumin as the standard.
Analysis of HNL Enzyme Activity-Hnl specific activity was determined by following the formation of benzaldehyde from 3.8 mM racemic mandelonitrile in 50 mM Na-citrate buffer, pH 5.0. The increase of absorbance at 280 nm was monitored over 5 min at 25°C. One unit of activity is defined as the amount of enzyme catalyzing the formation of 1 mol/min benzaldehyde from mandelonitrile under test conditions (30). Purified racemic mandelonitrile was a gift from DSM Fine Chemicals and contained less than 0.5% benzaldehyde.

X-ray Crystallography
Crystallization-Native HbHNL was purified and crystallized as described previously (30,31). Large diffraction quality crystals were obtained at room temperature using the vapor diffusion hanging drop method with 2% polyethylene glycol 400, 2.0 M ammonium sulfate in 0.1 M Na-Hepes buffer, pH 7.5, in the reservoir (700 -1000 l) and with 6 -10-l drops consisting of the reservoir solution and a 10 mg/ml solution of the HbHNL (in 50 mM potassium phosphate buffer, pH 7.5) in a 1:1 mixture. Two different crystal forms (orthorhombic C222 1 and tetragonal P4 1 2 1 2) were observed, which occasionally occurred even within the same drop. Tetragonal crystals of HbHNL had a much higher solvent content (70 versus 56%) and produced inferior diffraction patterns. Therefore, all structures communicated in this paper were determined from orthorhombic crystals.
The K236L mutant enzyme (15 mg/ml in 10 mM Na-Hepes buffer, pH 7.0) crystallized under the same conditions and yielded isomorphous orthorhombic crystals. (Tetragonal crystals were not observed for this protein.) However, crystallization events were less frequent, and the crystals took much longer to grow to usable sizes.
Data Collection, Processing, Model Building, and Refinement-Diffraction data sets were all collected at cryogenic temperatures on synchrotron beamlines at the European Molecular Biology Laboratory/ Deutsches Elektronen Synchrotron in Hamburg, Germany and at Elettra in Trieste, Basovizza, Italy (see Table I). The only exception was the data set for the complex of the K236L mutant with acetone cyano-hydrin, which was collected on a Schneider rotating anode generator in our laboratory. Data reduction involved DENZO and SCALEPACK (32) as well as programs from the CCP4 suite (33).
Before flash freezing, crystals were soaked for about 30 s in a cryoprotectant consisting of the reservoir solution with 30% glycerol and the substrate. Acetone and acetone cyanohydrin (2-hydroxy-isobutyronitrile) in particular were found to rapidly damage HbHNL crystals leading to a severe loss in crystallographic resolution. The complexes of native and mutant enzyme with these compounds were thus prepared with a vapor diffusion technique. The crystal was first transferred to a drop of the regular cryoprotectant solution (mother liquor containing 30% glycerol), and this drop was then equilibrated (in a hanging drop fashion) with a reservoir solution containing the volatile substrate in concentrations of 5-10% v/v for about 1 min. This gas diffusion/freeze quenching was a delicate experiment; although a high substrate concentration had to be used to quickly saturate all of the active sites to surpass the ongoing chemical and enzyme-catalyzed cyanohydrin cleavage (the mother liquor of the crystals had pH 7.5), high concentrations of cyanohydrin were found to rapidly damage HNL crystals. Because the gas diffusion was not easily controlled, it necessitated a trial-anderror approach consuming a large number of crystal specimens.
All crystal structures were found to be isomorphous to the published structures of HbHNL (18,21,34), and the coordinates of the rhodanide complex (Protein Data Bank code 2yas) devoid of all ligand and water molecules were used as starting point for the refinement with CNS (35). Topology and parameter files for the substrates were created using the program XPLO2D (36). Model building and fitting steps involved the graphics program O (37), using A -weighted 2F o Ϫ F c and F o Ϫ F c electron density maps (38). R free values (39) were computed from 5-10% randomly chosen reflections not used for the refinement. Water molecules were placed automatically into difference electron density maps and were retained or rejected based on geometric criteria as well as on their refined B-factors (B Ͻ 50 Å 2 ). For all refined structures, typical r.m.s. deviations from ideal values (40) of around 0.005 Å for bond lengths, 1.2°for bond angles, and 23°for dihedrals were calculated. Ramachandran plots showed all residues in all structures in both core and allowed regions with the exception of residues Ser 80 and Lys 129 , which were always observed in disallowed regions of /-space. Ser 80 is known to occur in a somewhat strained main chain conformation in ␣/␤-hydrolases (41), and the density of Lys 129 was found to be well defined throughout.
Details of the data collection, processing, and structure refinement are summarized in Table I. Specific deviations from the general procedure outlined above are described below for each structure. Coordinates have been deposited with the Protein Data Bank; accession numbers are also given in Table I.
Complex of Native HbHNL with Acetone Cyanohydrin-After rigid body refinement, unambiguous difference of electron density for the ligand became visible. The high resolution of 1.8 Å also allowed the unequivocal distinction between two possible species that could have formed during soaking, i.e. acetone cyanohydrin versus acetone plus cyanide or water. The refined B-factors of the substrate molecule were around 10 Å 2 , which is comparable with B-factors of neighboring protein atoms and indicates full occupancy of the binding site. For five residues (Glu 66 , Lys 73 , Gln 206 , Glu 220 , and Glu 251 ) the electron density suggested the presence of discrete alternate side chain conformations. These were included into the model with occupancies of 0.5.
The K236L Mutant-The coordinates used for the initial refinement of this structure still contained Lys 236 . The difference electron density after this step very nicely showed a positive peak in the region that would accommodate one of the terminal methyl groups of Leu 36 as well as pronounced negative features around the ammonium group of Lys 236 thereby yielding independent proof of the nature of the mutation. After introducing the mutation, the structure was refined according to the general procedure. Alternate side chain conformations were refined for Lys 73 , Ser 80 , Gln 206 , Glu 220 , and Glu 251 .
Complex of K236L with Acetone-The acetone molecule was unambiguously placed into difference electron density within the active site. The higher B-factors of the ligand compared with surrounding protein atoms indicated partial occupancy, which was accordingly set to 0.5. Alternate side chain conformations were refined for His 14  Complex of K236L with Acetone Cyanohydrin-Although the data were originally processed to a resolution of 2.7 Å, only reflections to 2.9 Å were used in the refinement. Residual density within the active site indicated the location of the substrate and its orientation, but refined B-factors of the ligand were high (50 Å 2 ). Because of the low resolution of the diffraction data no water molecules were included into the model.
Difference Density in the Active Site of the MeHNL-Acetone Complex-Coordinates and structure factors for the complex of the hydroxynitrile lyase from M. esculenta with the substrate acetone (23) were obtained from the Protein Data Bank (1dwo). A difference electron density map was calculated with the program CNS (35) from the deposited coordinates following a brief positional and B-factor refinement (10 cycles each). A bulk solvent correction was applied, and overall anisotropic B-factors were calculated.
Poisson-Boltzmann Calculations-Finite difference Poisson-Boltzmann calculations were performed to estimate the relative free energies of four protonation (see Fig. 5A) states of the His 235 -Asp 207 dyad in the active site of wt-HbHNL and the K236L mutant. The zwitterionic state was arbitrarily defined as the reference state in all calculations. The computations are based on thermodynamic cycles involving the four ionization states in water and in the protein environment as well as the dyad constituents in water at infinite separation.
The thermodynamic cycles (42) allow the calculation of relative ⌬Gvalues solely from the known pK a values of histidine and aspartic acid in solution (6.5 for His and 4.4 for Asp (43)) plus two parameters that are accessible by finite difference Poisson-Boltzmann methods. They are the electrostatic interaction energy of the two residues in the hydrogen-bonded pair in water and the solvation free energy of the dyad for the transfer from water into the protein environment. This is shown in equations 1-3 below. Electrostatic interaction and solvation-free energies were calculated with the program DelPhi (44) using the following parameters: bulk dielectric constants of 4.0 (protein) and 80.0 (solvent), an ionic strength of 0.145 M, a grid scale of 2 Å Ϫ1 , an ion exclusion radius of 2 Å, and a probe radius (for surface calculations) of 1.4 Å. All calculations were performed with the eight structures of HbHNL currently available in the Protein Data Bank (codes 1qj4, 1yas, 2yas, 3yas, 4yas, 5yas, 6yas, and 7yas), and the resulting energy values were averaged. Aspartate (except Asp 207 ), glutamate, arginine, and lysine side chains were modeled as charged. The Lys 236 to leucine mutation was modeled by neutralizing the charge of Lys 236 . Removal of the terminal atoms of the side chain (to mimic the different volume of Leu compared with Lys) was found to have no effect on the results of the calculations. Tautomers and protonation states of histidine residues (except His 235 ) were assigned by inspection. In contrast with the native enzyme, His 14 was assigned a full positive charge in the K236L mutant, because this histidine forms a salt bridge with Glu 79 in the mutant enzyme structures (see Fig. 4B).

RESULTS AND DISCUSSION
HbHNL and MeHNL Have the Same Molecular Mechanism-For quite some time, the elucidation of structural and molecular mechanisms on ␣/␤-hydrolase-type HNLs has been carried out in parallel by two competing groups for the enzymes from H. brasiliensis (18,20,21) and M. esculenta (12,23,24,45), respectively. Based on three-dimensional structural results, in combination with mutagenesis data and enzyme-kinetic evidence, the two groups arrived at mechanistic proposals shown in Fig. 2 for HbHNL and in Fig. 3 for MeHNL. Although these mechanisms are similar, there are notable differences, which were suggested to be because of the genuine differences in the molecular mechanism between the two enzymes (23,24).
Relevant mutagenesis data are listed in Table II for activesite residues of the two enzymes. In the course of the present investigations, we prepared and characterized a number of HbHNL mutants (C81A, T11A) previously only available for the MeHNL enzyme. We note that there are no contradictory mutagenesis results between the two enzymes. The HbHNL and MeHNL enzymes show 77% sequence identity, and the crystal structures of the acetone complexes superimpose with a r.m.s. deviation of 0.42 for the 253 common Carbon atoms and with an r.m.s. deviation of 0.23 for the atoms of the active-site residues. The striking similarity in the binding of acetone in the active-sites of MeHNL and HbHNL is evident from Fig. 4A. One of the differences in the reaction mechanisms concerns the catalytic role of residue Lys 236 , which will be discussed below for the HbHNL enzyme. Fig. 4A shows that this residue is at the same location in both enzymes, yielding no structural reason supporting a different catalytic role between the two enzymes. The only structural difference between the two enzymes in their complexes with the substrate acetone concerns the observation of a water molecule in the HbHNL-acetone structure. This "central" (21) water molecule, which is present in both enzymes in the substrate-free form, was suggested to occupy the binding site of the cyanide ion in the Hevea enzyme (21). In fact, this water position was observed to have a high affinity for anions such as Cl Ϫ (46). A lack of observation of this water molecule in the MeHNL-acetone structure was suggested to be because of differences in the crystallization conditions for To clarify this only remaining structural difference between the active sites of the two enzymes, we re-calculated the difference electron density for the MeHNL-acetone structure from the deposited coordinates and structure factors (Protein Data Bank code 1dwo and Ref. 23). The density obtained for the active-site region (not shown) exhibits a strong peak (Ͼ3) between the bound substrate and the side chains of His 235 and Lys 236 (Hevea numbering scheme) close to the position occupied by the catalytic water molecule in HbHNL (Fig. 4A). We may thus conclude that there are neither structural nor mutagenesis data supporting differences between the molecular mechanisms of the HbHNL and MeHNL enzymes.  1dwo, chain A). B, superposition of the native HbHNL structure with the HbHNL-K236L mutant. The mutant structure is represented with darker bonds. C, superposition of the complexes of HbHNL with acetone and acetone cyanohydrin. The figure was produced using the programs MolScript (53) and Raster3D (19).
The Three-dimensional Structure of HbHNL in Complex with Acetone Cyanohydrin-Although the native three-dimensional structures (18) and their complexes with the biological reaction product acetone (21,23) have been known for some time for both ␣/␤-hydrolase HNLs, a crystallographic observation of the complex with the biological substrate acetone cyanohydrin was considered a difficult challenge in view of the rapid turnover (24). This was a serious impediment for all mechanistic proposals, which had to be based on modeling studies with HbHNL (22) and on the crystallographic observation of the complex between acetone cyanohydrin and the inactive S80A mutant of MeHNL (24).
Using a gas diffusion technique in combination with rapid freeze quenching, we were able to bind acetone cyanohydrin to crystalline HbHNL and obtain complex crystals suitable for crystallographic structure analysis, as described under "Experimental Procedures." Fig. 6A shows the results of the crystal structure analysis with the biological substrate acetone cyanohydrin, and a superposition of this structure with the HbHNLacetone complex is shown in Fig. 4C. The significance of these results for the mechanism of HNL catalysis is discussed below.
Preparation, Kinetic Characterization, and Structure Analysis of the K236L Mutant and of Its Substrate Complexes-Using site-directed mutagenesis, the gene for the K236L mutant was cloned, and its product was overexpressed in P. pastoris and purified as described under "Experimental Procedures." Hb-HNL-K236L does not show any detectable HNL activity; even 1.5 mg of pure HbHNL-K236L protein in a 1.0-ml reaction assay did not lead to a conversion of mandelonitrile exceeding the chemical background reaction. The crystal structure of the uncomplexed K236L mutant is shown in Fig. 4B, superimposed on the native HbHNL structure. Substantial conformational rearrangement is only observed for the imidazole of His 14 , which is rotated by ϳ90°to compensate for the (now missing) H-bond between Lys 236 and Glu 79 . The structural similarity between the wild-type HbHNL and the K236L mutant suggests that the loss of activity of the latter is not because of gross conformational rearrangements (see below).
In addition to the uncomplexed K236L mutant, we also analyzed crystals soaked in solutions of acetone and acetone cyanohydrin. Observation of the resulting complex structures was facilitated by the catalytic incompetence of the K236L mutant. The three-dimensional structures are shown in Fig. 6, B (acetone) and C (acetone cyanohydrin), respectively.
Poisson-Boltzmann Calculations-Finite difference Poisson-Boltzmann calculations were used to estimate the relative free energies of four protonation states of the His 235 -Asp 207 dyad in the active site of HbHNL. These states (Fig. 5A) include a zwitterionic state (imidazolium/carboxylate, arbitrarily defined as the reference state in all calculations), a neutral state (imidazole/carboxylic acid), a cationic (imidazolium/carboxylic acid), and an anionic state (imidazole/carboxylate). Although the equilibrium between zwitterionic and neutral is inherently pH-independent as it only involves proton transfer among the two residues, the zwitterionic 3 cationic and zwitterionic 3 anionic transitions involve proton exchange with a solvent and therefore depend on the pH. Analogous calculations have recently been performed to explain the appearance of a low barrier hydrogen bond between His 235 and Asp 207 in HbHNL upon binding of a transition-state analog (42). The computations are based on thermodynamic cycles involving the four ionization states in water and in the protein environment as well as the dyad constituents in water at infinite separation. Here, we extend this approach to the analysis of the perturbation caused by the removal of the positive charge of Lys 236 in the (inactive) K236L mutant enzyme.
The pH dependence of the relative free energies of the His 235 -Asp 207 dyad are shown in Fig. 5 for the native HNL enzyme (B) and the K236L mutant (C). The effect of the removal of the positive charge of Lys 236 is clearly detectable from the shifts of the energy curves. Although the anionic state (involving a neutral His 235 -imidazole) is energetically favored over a wide pH range in the native enzyme, its energy curve is shifted to higher pH values in the mutant enzyme leaving the zwitterionic and neutral states as most favored. If His 14 is modeled as neutral, this shift is further intensified resulting in a point of intersection with the zwitterionic (neutral) curve around pH 11 (data not shown). Compared with native HbHNL, the energy difference between the neutral and zwitterionic states is reduced from 3.2 to 0.1 kcal/mol in the mutant. The cationic state does not play a significant role in both enzymes.
The Role of Lys 236 in Catalysis-Even in the absence of mutagenesis data, the catalytic significance of Lys 236 has been predicted from the fact that it is one of the residues surrounding the active-site cavity (21). We are now in a position to substantiate this prediction, because replacement of Lys 236 by Leu completely removes catalytic activity without inducing substantial structural rearrangements. There are at least three ways in which Lys 236 is essential for catalysis. 1) Lys 236 forms a hydrogen bond in the native HNL and in the HbHNL-acetone complex (Fig. 4A) to a water molecule, which has been identified as the cyanide binding site. Evidently, the proximity of the positive charge of Lys 236 will decrease the pK a of HCN via stabilizing the negatively charged CN Ϫ and thus facilitate C-C bond cleavage or, for the synthesis direction, deprotonation of HCN.
2) As shown in Fig. 6A, Lys 236 is involved in binding the cyanohydrin substrate via the formation of a hydrogen bond to the cyanohydrin nitrogen atom. Replacement of Lys 236 by Leu does away with this H-bond resulting in a substantially different orientation of the cyanohydrin within the active-site cavity (Fig. 6C). Interestingly, there is essentially no effect of the K236L replacement on the binding of the acetone substrate to the HNL active site (Fig. 6B).
3) A third effect of Lys 236 on HNL catalysis emerged from the Poisson-Boltzmann calculations for the His 235 -Asp 207 pair. This pair undergoes a protonation/deprotonation during a catalytic cycle. Previously (42) such calculations for native Hb-HNL had yielded a shift of the free energy of the anionic state relative to the zwitterionic reference state by about 10.6 kcal/ mol upon binding of rhodanide, a strong inhibitor mimicking the transition state of the reaction. This destabilization of the anionic state is equivalent to a shift in the apparent pK a of His 235 from 2.5 to 10, which makes sense immensely for the catalytic functioning of the dyad as a base catalyst: a strong base is required for deprotonation of the cyanohydrin hydroxyl (pK a ϳ10.7, Ref. 47), but this base must not be protonated in the substrate-free enzyme in which the active site is exposed to a solvent. The corresponding calculations for the K236L mutant (Fig. 5C) show that such a destabilization of the His-Asp Ϫ pair relative to one of its protonated states already occurs in the absence of inhibitor, i.e. the His-Asp dyad is already protonated under neutral or mildly acidic conditions and can therefore not act as a catalytic base.
The very low (Ͻ 1%) activity of the K236R mutant (Table II), in which the Arg replacing Lys 236 also supplies a positive charge to the active site, is most likely because of steric factors, because the bulky guanidinium group will occupy the binding site of the central water molecule (and hence also of the cyanide ion).
The Molecular Mechanism of the Cyanohydrin Reaction Catalyzed by HNL of the ␣/␤-Hydrolyase Type-The now available structural data allow a reliable description of the main intermediates along the HbHNL catalyzed cyanohydrin reaction. All data are consistent with the four-step mechanism shown for the cyanohydrin cleavage direction in Fig. 2. The data reported in the present work complete the structural and mechanistic characterizations of the four reaction intermediates by supplying structural data for the key intermediate (the complex between HbHNL and acetone cyanohydrin) and by presenting comprehensive evidence for the crucial catalytic role of Lys 236 . The four reaction intermediates are explained below.
(a) The first is the resting state. The catalytic triad is not connected by H-bonding, i.e. the distance between the Ser 80oxygen and the His 235 -N⑀ is 3.30 Å (21). There are several water molecules within the active site, one of them (the central water molecule) is suspended by H-bonds between His 235 and Lys 236 . Upon cyanohydrin binding, this water molecule is replaced by the cyano end of the substrate.
(b) Next is the complex with cyanohydrin substrate (Figs. 4C and 6A). The substrate is fixed by H-bonding to residues Ser 80 , Thr 11 and Lys 236 . There is a hydrogen bond between His 235 and Ser 80 (2.76 Å). The role of Cys 81 has been the subject of concern in the past (13,20) as replacement of this residue by Ala affects the activity only marginally (Table I) in agreement with the fact that the distance to the cyanohydrin oxygen is at the limit for H-bonding (3.67 Å). The low activity of the C81S mutant (Table I) is probably because of the formation of a stronger H-bond to the substrate, which would then be displaced from the active site. Deprotonation of the cyanohydrin is accomplished by Ser 80 , which in turn is deprotonated by His 235 . Cleavage of the C-C bond is facilitated by the stabilization of  (c) Third is the complex with cleavage products (acetone and cyanide). The cyanide occupies the position of the central water molecule. Following protonation by His 235 , it is replaced by a water molecule (48).
(d) Finally, the complex with acetone (Fig. 4A). Binding to the active site is accomplished by H-bonding to residues Ser 80 and Thr 11 possibly with a small contribution of Cys 81 . Lys 236 interacts with the central water molecule occupying the cyanide binding site.
As discussed above, there is no evidence supporting a mechanistic difference between the enzymes from H. brasiliensis and from M. esculenta. Mechanistically, hydroxynitrile lyases of the ␣/␤-hydrolase superfamily (such as the enzymes from H. brasiliensis and M. esculenta) are a curiosity because they belong to a very small group of ␣/␤-hydrolase enzymes (49) in which the catalytic triad does not act as a nucleophile. Among the few other enzymes are a C-C hydrolase (MhpC) from E. coli (50) and possibly a haloperoxidase from Streptomyces aureofaciens (50,51).
Factors facilitating cyanohydrin cleavage via general base catalysis were identified many years ago (52). They include a strong base to deprotonate the cyanohydrin and a positive charge to stabilize the cyanide formed by C-C bond cleavage. From the structural and mechanistic data it is quite clear that the catalytic triad, notably the imidazole of His 235 , serves as strong base, while the positive charge to stabilize the formed cyanide anion is supplied by Lys 236 . Thus, the acid/base-catalyzed cyanohydrin reaction involves a deprotonation-protonation sequence interrupted by the dissociation of the deprotonated cyanohydrin into the corresponding carbonyl compound and cyanide. It is notable that the catalytic cycle of the Hb-HNL-catalyzed cyanohydrin reaction proceeds with very little conformational rearrangement. The only conformational changes involve a small rotation of the imidazole of His 235 as shown in the superposition of the active sites of HbHNL complexed with acetone and acetone cyanohydrin, respectively (Fig. 4C). The figure shows quite convincingly how the proton transfer occurs from the cyanohydrin hydroxyl to the cyanide ion with minimal conformational rearrangement.
The key step of catalysis is obviously the one shown in Fig 2b, i.e. the deprotonation of the substrate followed by C-C-bond cleavage. An obvious question is whether the two events occur concerted or stepwise. In aqueous solution, kinetic and thermodynamic data for the addition of HCN to substituted benzaldehydes (47) are consistent with a stepwise mechanism involving C-C bond cleavage as the rate-limiting step following fast proton transfer from the cyanohydrin. To allow a concerted general acid/base-catalyzed pathway for HNL-catalyzed cyanohydrin cleavage, we have to require that 1) the pK a of the proton-donating catalyst be intermediate between the starting material (cyanohydrin, pK a ϳ11) and the product (carbonyl compound, pK a Ͻ0), and 2) the leaving group properties of cyanide be enhanced. Both requirements appear to be met in the HbHNL catalyzed reaction. The pK a of His 235 was shown to be shifted from Ͻ4 in the free enzyme to higher than 8 in the complex with the transition state analog rhodanide (42), whereas the electrostatic interaction with the positive charge of Lys 236 renders the cyanide ion a better leaving group. Given the data for the reaction in solution (47) and the chemical environment within the active site, it seems likely that Hb-HNL-catalyzed cyanohydrin cleavage follows a concerted reaction pathway, as depicted in Fig. 2b.