Observation of a Short, Strong Hydrogen Bond in the Active Site of Hydroxynitrile Lyase from Hevea brasiliensis Explains a Large p K a Shift of the Catalytic Base Induced by the Reaction Intermediate*

The hydroxynitrile lyase from Hevea brasiliensis ( Hb-HNL) uses a catalytic triad consisting of Ser 80 -His 235 Asp 207 to enhance the basicity of Ser 80 -O (cid:1) for abstracting a proton from the OH group of the substrate cyanohydrin. Following the observation of a relatively short distance between a carboxyl oxygen of Asp 207 and the N (cid:2) 1 (His 235 ) in a 1.1 Å crystal structure of Hb HNL, we here show by 1 H and 15 N-NMR spectroscopy that a short, strong hydrogen bond (SSHB) is formed between the two residues upon binding of the competitive inhibitor thiocyanate to Hb-HNL: the proton resonance of H-N (cid:2) 1(His 235 ) moves from 15.41 ppm in the free enzyme to 19.35 ppm in the complex, the largest downfield shift observed so far upon inhibitor binding. Simultaneously, the D/H fractionation factor decreases from

Short, strong hydrogen bonds (SSHBs) 1 occur when the pK a of a hydrogen-bonded donor matches the one of the acceptor. If the hydrogen bond is sufficiently short, the central maximum of the (more or less symmetrical) hydrogen atom potential along the line connecting donor and acceptor may fall below the vibrational groundstate, resulting in a delocalized hydrogen atom. This situation is then referred to as a "low-barrier hydrogen bond" (LBHB). 2 Low-barrier hydrogen bonds may occur in solutions and crystals of organic and inorganic compounds (1). Their observation in the active sites of enzymes (2)(3)(4) has led to controversy concerning their significance for the mechanism of enzyme catalysis (2)(3)(4)(5)(6)(7)(8)(9)(10)(11). It has been suggested that the energy gained by the transient conversion of a "normal" hydrogen bond into a short, strong hydrogen bond can be used to stabilize transition states of enzymatic reactions (2)(3)(4)(5)(6)(7). However, this energy gain (12) and its relevance for enzyme catalysis did not remain undisputed (8 -11). There are indeed enzymes where a LBHB was observed that was shown not to be an inherent requirement for substantial rate enhancement (13). Irrespective of their energetic relevance, low-barrier hydrogen bonds are useful diagnostic tools to indicate the matching pK a s of a hydrogen bonded donor-acceptor pair, which was used to rationalize enzymic mechanisms (14).
The occurrence of SSHBs has so far been reported for a considerable number of enzymes, including ketosteroid isomerase (15)(16)(17), triosephosphate isomerase (18), serine proteases (3,6,19), tryptophan synthase (20), 2-amino-3-ketobutyrate-CoA ligase (21), and cholinesterases (22,23). Their occurrence has typically been suggested by NMR spectroscopy (24,25). The exposure of the delocalized proton decreases the electron density around the proton nucleus, which shifts the NMR signal to very low field (higher frequency; typically 18 -22 ppm (26)). Another NMR-observable quantity used to identify a short, strong hydrogen bond is the fractionation factor for the associated proton, defined as the equilibrium constant for the exchange of hydrogen by deuterium with the solvent (27). depends on the relative strength of a hydrogen bond compared with the solvent, with values lower than 1 indicating stronger hydrogen bonds (24).
Hydroxynitrile lyase (HNL) (EC 4.1.2.39) catalyzes the cleavage of cyanohydrins to hydrocyanic acid plus the corresponding aldehyde or ketone (Fig. 1A). The release of HCN serves as a defense against herbivores and microbial attack for a variety of plants (28 -30). In aqueous solution, cyanohydrin cleavage occurs spontaneously above pH 5, whereas the enzyme reaction also occurs at lower pH values (down to ϳpH 3; Ref. 29). This is technologically exploited for the enantioselective synthesis of chiral cyanohydrins (31)(32)(33)(34)(35), making use of the reverse in vivo reaction. Structurally, the HNL from the tropical rubber tree Hevea brasiliensis (HbHNL) belongs to the ␣/␤-hydrolase superfamily (36) and has a prototypical catalytic triad consisting of residues Asp 207 , His 235 , and Ser 80 . However, the environment of this triad differs from other hydrolases. Based on a series of crystal structures of HbHNL with various substrates and inhibitors (37) in combination with modeling studies (38) and kinetic data (39) a general acid/base reaction mechanism ( Fig. 1B) was proposed. Key elements of this mechanism are (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 . The subsequent cleavage of the cyanohydrin is facilitated by the stabilization of the charge of the nascent cyanide through interaction with the positive charge of Lys 236 . A very similar mechanism devoid of a catalytic role for the active site lysine has been proposed independently for the similar and highly homologous HNL from Manihot esculenta (40,41). Irrespective of these differences, the question about the pK a of the catalytic base in the protein interior (42) and its modulation during the enzyme-catalyzed reaction (17) remained open. Thus, despite its similar threedimensional fold, the reaction mechanism of this type of HNLs differs fundamentally from that of other ␣/␤ hydrolases, as it constitutes one of the rare examples where the triad does not act as a nucleophile.
To obtain evidence about the protonation states in the active site, the HbHNL structure was refined against diffraction data collected to very high (1.1 Å) resolution (43). The refinement yielded a rather short distance between the two residues of the catalytic triad (2.67 Å between N␦(His 235 ) and O⑀(Asp 207 )). Difference density indicated a proton attached to N␦(His 235 ), with no density near N⑀(His 235 ) or O⑀(Asp 207 ). Thus, the proton between His 235 and Asp 207 did not appear to be shared among the two heteroatoms, as it was observed in some other hydrolases (44).
Here we report NMR experiments that show that this "catalytic" hydrogen bond converts into a short, strong hydrogen bond upon addition of a strong competitive inhibitor mimicking the rate-limiting transition state of the enzyme reaction. To rationalize this NMR result, we performed Poisson-Boltzmann computations, which yield a consistent explanation for the occurrence of the SSHB. In fact, the results of the computations show how the occurrence of the SSHB is paralleled by a catalytically relevant increase in basicity of the general base upon approach of the transition state.

EXPERIMENTAL PROCEDURES
Materials-Crude natural abundance HbHNL was obtained by Roche Applied Science in a 20 mM citrate phosphate buffer pH 6.5 (5300 units/ml and 59 mg/ml) and purified as reported (45). For NMR measurements the protein was dialyzed against 50 mM potassium phosphate buffer, pH 7.4. For the isotopically labeled protein, 15 N-ammonium sulfate, 13 C-glycerol (Martek, Columbia, MD), and 13 C-methanol (Cambridge Isotope Laboratories, Andover, MA) were used for the minimal medium. Anhydrous potassium thiocyanate (KSCN), acetone, benzaldehyde, mandelonitrile, HCl, sodium hydroxide, and deuterium oxide (D 2 O, 99.8% at % D) were purchased from Merck. All solvents and reagents were analytical or reagent grade and were used without further purification.
Enzyme Preparation-Wild-type HbHNL was labeled with 15 N or 15 N and 13 C by expression from the Pichia pastoris strain 2386 GS115 H ϩ M Ϫ pHILD1.17-HNL (Technical University Graz) in a minimal medium that has been slightly modified from the published fermentation procedure (45). Instead of a bioreactor, shake flasks containing 0.4 M KH 2 PO 4 /K 2 HPO 4 , pH 5, 0.1 g ( 15 NH 4 ) 2 SO 4 per 200 ml minimal medium were used. In addition, the concentrations of trace elements were lower by a factor of four. The protein was purified by anion exchange (HiTrap Q HP, Amersham Biosciences) and gel filtration chromatography (Su-perdex200, Amersham Biosciences) as reported previously (45). All samples were Ͼ95% pure based on SDS-PAGE analysis. Protein concentration was determined by the BCA assay (Bio-Rad). The 15 N Hb-HNL sample was concentrated to 10 mg/ml (activity ϭ 200 units/ml) and the 15 N/ 13 C double labeled HbHNL to 19 mg/ml with an activity of 321 units/ml.
Hydroxynitrile Lyase Activity Test-Enzyme activity (39) was measured following the cleavage of racemic mandelonitrile into benzaldehyde and HCN, at a substrate concentration of 12 mM and at an enzyme concentration of about 0.74 g/ml. The assay was performed in 20 mM glutamate buffer at pH 5 and 25°C, and the formation of benzaldehyde was monitored spectrophotometrically at 280 nm. In addition to the enzymatic reaction the base-catalyzed chemical reaction has to be measured separately and subtracted. The reaction was monitored for 2 min, and the enzyme activity was calculated from the slope in a plot of ⌬E versus time.
NMR Sample Preparation-For all measurements 600 l of between 0.33 and 0.63 mM HNL in 50 mM potassium phosphate buffer, pH 7.4, 90:10% H 2 O/D 2 O were used. In the titrations with acetone, benzaldehyde, and thiocyanate, concentrations of 1-50 mM, 1-6 mM, and 0.2-1 mM were used, respectively. For the determination of the fractionation factors we used H 2 O/D 2 O buffer mixtures of 550:50 l, 420:180 l, 300:300 l and 180:420 l. The pH value was changed within the range pH 4 -10 for the acid/base titrations. Starting from a sample with pH 7.01, titration involved steps of 3 to 5 l of 0.1 M HCl or NaOH. The pH in the 5 mm NMR sample tube was monitored using an Orion Model 9803 NMR pH electrode.
NMR Experiments-All one-dimensional 1 H-NMR spectra were collected on a Bruker Avance DRX 500 MHz NMR spectrometer, where the water signal was suppressed using the WATERGATE sequence (46). Due to the fast relaxation of the protein signals, at least 1024 scans had to be acquired for each one-dimensional spectrum consisting of 16,000 data points. A 60°phase-shifted squared sine-bell window function was applied prior to Fourier transformation. The 1 H-15 N HSQC was acquired on a Varian Unity INOVA 600 MHz spectrometer with a 5 mm HCN triple resonance probe at 27°C. For each of the 128 increments, 64 transients of 1000 complex points were accumulated and the data set zero filled to a final size of 2048 ϫ 512 complex data points. In both dimensions 60°phase-shifted squared sine-bell window functions were used. To achieve maximum intensity for the signal belonging to the proton in the SSHB the carrier frequencies were shifted to 15.41 and 190 ppm for the 1 H and 15 N dimension, respectively.
For the measurement of the fractionation factors of the deshielded proton resonances in free HbHNL and in the HbHNL⅐SCN Ϫ complex, the enzyme was dissolved in mixtures of protonated and deuterated water with the above H 2 O/D 2 O ratios. We started with the mixture containing the lowest D 2 O concentration, lyophilized it and resuspended it with the next mixture of H 2 O/D 2 O. The sample was equilibrated for 2 h at room temperature to allow complete exchange of the deshielded proton before the next spectrum was recorded. Each sample was measured twice and the signal intensities were averaged. A high field-shifted methyl signal was used as an internal standard to calibrate the signal intensities to compensate for the loss of enzyme during lyophilization and re-dissolution.
The fractionation factor is defined as the equilibrium constant of the reaction (27) as seen in the Equation 1 as follows.
Here, I max is the maximal peak intensity at X ϭ 1.00 (24). The fractionation factor was calculated from the slope of a straight line obtained by least squares fitting the reciprocal measured signal intensities (1/I) as a function (1 Ϫ X)/X in the linearized form of Equation 2 as follows.
Calculation of Electrostatic Interactions-Electrostatic interaction and solvation-free energies were calculated using the finite-difference-Poisson-Boltzmann (FDPB) method as implemented in the program DelPhi (47). The two values for the bulk dielectric constant were set to 4.0 (protein) and 80.0 (solvent), and an ionic strength of 0.145 M was assumed throughout. The scale of the grid was 2 Å Ϫ1 . The ion exclusion radius was set to 2 Å and the probe radius (for surface calculations) to 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) (36,37,43), and the resulting energy values were averaged. Water molecules and ligand atoms present in the structures were removed. In the case of alternate conforma-tions, the higher occupied conformer was kept. Aspartate (except Asp 207 ), glutamate, arginine, and lysine side-chains were modeled as charged. Tautomers and protonation states of histidine residues (except His 235 ) were assigned by inspection.
A special situation was encountered for the cationic pair, in which the aspartate is neutral but still acts as hydrogen bond acceptor. For the protein-bound pair, on either of the two carboxyl oxygen a proton would not find a suitable hydrogen bond acceptor, but possibly lead to unacceptable interactions with the protein (main-chain amide nitrogen atoms of Ile 209 and Phe 210 ; side-chain of Asn 104 ). Neutralization of the Asp 207 side-chain was therefore accomplished by modifying the charges only, without the addition of a hydrogen atom.
For the FDPB-calculations in water, the entire amino acids for His 235 and Asp 207 were used in a conformation as observed in the respective HbHNL crystal structure (36,37,43) and the obtained energy values for the various structures were averaged as described for the calculations in the protein environment.

RESULTS
Observation of a Downfield-shifted 1 H-NMR Signal-The observation of a short distance (2.67 Å) between O⑀(Asp 207 ) and N␦ 1 (His 235 ) in the high-resolution crystal structure of HbHNL (43) indicated the possibility for a short, strong hydrogen bond in the active site of this enzyme. A 1 H-NMR spectrum was acquired between Ϫ3 to 22 ppm, and a signal was indeed observed at 15.41 ppm (Fig. 2). Because the large molecular size (60 kDa) of this homodimeric (48) protein prohibited a complete sequential NMR assignment, we had to resort to indirect ways to corroborate the assignment of the observed signal at 15.41 ppm to the proton attached to N␦ 1 (His 235 ).
Signal Assignment-The following empirical correlation between proton chemical shifts and hydrogen bond lengths in imidazolium⅐carboxylate complexes (23) is as follows D ϭ 1.99 ϩ 0.198 ln(␦) ϩ (10.14/␦) 5 (Eq. 4) was used to deduce a heteroatom separation from the observed chemical shift and compare it with the crystallographically observed value for the N␦ 1 (His 235 )-O⑀(Asp 207 ) distance. The distance D obtained from the proton chemical shift ␦ ϭ 15.41 ppm is 2.65 Å, in good agreement with the crystallographically observed (43) distance of 2.67 Å.
To ensure that the downfield signal belongs to a proton attached to a nitrogen nucleus, a 1 H-15 N-HSQC was acquired on a sample of 0.33 mM 15 N-isotopically enriched HbHNL. The proton and nitrogen carriers were placed at 15.41 and 190 ppm, respectively, to optimize the signal intensities in the region of the 15.41 ppm signal. The resulting two-dimensional spectrum is shown in Fig. 3, where the signal at 15.41 and 180.08 ppm is clearly visible. Attempts to correlate this signal to a nearby 13 C nucleus in 15 N, 13 C doubly labeled HbHNL in a two-dimensional H-(N)-C correlation experiment (49) were unsuccessful, presumably due to the fast relaxation within this large protein and to the relatively small 1 J NC coupling constants within histidine side-chains (50).
The assignment of the 15.41 ppm signal was further confirmed by determining the shift induced by the presence of benzaldehyde in the active side. The crystal structure of Hb-HNL complexed with benzaldehyde has not yet been reported, but modeling studies (38) yield a stable solution which shows the aromatic ring 7.1 Å away from the H-N␦(His 235 ), as shown in Fig. 4A. This orientation was used to calculate (program SHIFTS (51)) the change in chemical shift of the proton attached to N␦(His 235 ) induced by the ring current of benzaldehyde. The computed shift of Ϫ0.13 ppm compares well with the experimentally observed shift of the proton signal from 15.41 ppm to 15.22 ppm upon titrating 6 mM benzaldehyde into a solution of 0.6 mM HbHNL (Fig. 2C). In a control experiment, acetone was titrated into a solution of 0.6 mM HbHNL up to a concentration of 50 mM, which did not cause any shift of the downfield signal (data not shown).
The Effect of the inhibitor Thiocyanate and of the Substrate Mandelonitrile-Although a chemical shift of 15.41 ppm for a H-N␦(His) is clearly outside the "regular" range (10.1 Ϯ 3.5 ppm, BioMagResBank, www.bmrb.wisc.edu (52)), it is not yet a low barrier hydrogen bond, whose 1 H-NMR signal is expected between 18 -22 ppm (26). Because short, strong hydrogen bonds are often observed in protein conformations mimicking a transition state, we also titrated the strong inhibitor thiocyanate (K i ϭ 5.5 M (37, 53)) into a solution of 0.5 mM HbHNL. Upon addition of thiocyanate, the signal at 15.41 ppm disappears and a new signal appears at 19.31 ppm (Fig. 5), i.e. at a value characteristic for low barrier hydrogen bonds.
A different behavior is observed when adding mandelonitrile, one of the substrates of HbHNL (K m ϭ 3 mM (39)). Upon titration of 0.1-3 mM mandelonitrile into a solution of 0.5 mM HbHNL, the signal at 15.41 ppm disappears and no signal beyond 13 ppm is observable. We assume that the dynamics of the HNL reaction prohibit the observation of H-N␦ 1 (His 235 ), whose resonance shifts to several different positions during one reaction cycle.
Hydrogen-Deuterium Fractionation Factors-We have determined the fractionation factors of unliganded HbHNL and of HbHNL complexed with SCN Ϫ by monitoring the intensities of the downfield-shifted signal in varying ratios of D 2 O/H 2 O. The reciprocal normalized intensities 1/I plotted versus (1 Ϫ X)/X are shown in Fig. 6 for HbHNL and HbHNL-SCN. From the slope of this normalized plot, a fractionation factor of 0.98 is obtained for HbHNL in the absence of thiocyanate and of 0.35 in the presence of 1 mM thiocyanate. The latter value is typical for a short, strong hydrogen bond.
pH Titration-The protonation properties of His 235 were investigated by pH titrations in the range between pH 4 and 10 for HbHNL in the presence and in the absence of SCN Ϫ . These studies were complicated by the pH instability of the protein.
In the uncomplexed form, HbHNL starts to precipitate below ϳpH 5, and it denatures above pH 10 as indicated by the complete collapse of the one-dimensional 1 H spectrum to random-coil values (54).
Between ϳpH 4 and 9, titration of free HbHNL did not lead  to detectable spectral changes. In the presence of thiocyanate, HbHNL appears to be slightly more stable at lower pH values, as judged from the later onset of precipitation, which now occurs below pH Ͻ4. Although at the low pH side, no pH-dependent spectral changes are observed, the signal at 19.35 ppm starts to diminish above pH 7 with concomitant reappearance of the 15.41 ppm signal in the presence of thiocyanate. This process is complete at pH 9.
Poisson-Boltzmann Calculations-FDPB calculations were used to estimate the relative free energies of the four protonation states of the His 235 -Asp 207 diad shown in Fig. 7A. These states included a zwitterionic state (arbitrarily defined as the reference state in all calculations), a neutral state, a cationic, and an anionic state. Although the transition 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 solvent and therefore depend on the pH of the solution.
The estimation of relative free energy values of the different protonation states of the diad within the protein was accomplished through three sets of thermodynamic cycles. One of them is shown in Fig. 7B. They enabled the calculation of ⌬G-values solely from the pK a values of histidine and aspartic acid in solution plus two parameters that are accessible by FDPB methods. They are the electrostatic interaction energy of the two residues in the hydrogen-bonded pair and the solvation free energy of the diad for the transfer from water into the protein environment.
Each thermodynamic cycle starts from the two residues in water at infinite separation, where the relative free energies ⌬G 0 depend only on the pK a values of the two residues (6.5 for His and 4.4 for Asp (55)) and, for the transitions from zwitterionic to cationic or anionic state, on the solution pH as shown in the following equation.
The approach of the two solvated groups results in the formation of different types of hydrogen bonds (neutral H-bond or salt bridge). Thereby the electrostatic interaction of His 235 with the negative charge of Asp 207 in the anionic state raises the pK a of the former. Likewise, the pK a of Asp 207 is lowered by the same amount due to the interaction with the positive charge of His 235 in the cationic state. These shifts correspond to differences in electrostatic interaction energy ⌬⌬G elec , with the inherent symmetry of these mutual interactions (causing equal pK a shifts in opposite directions) causing a factor of two in the equation for ⌬G 1 neutral .
The pH dependence of the relative free energies of the diad are shown in Fig. 8 for the protein environment. For the calculation of ⌬G 2 the pH dependence of ⌬⌬G solv was ignored, i.e. ⌬⌬G solv was calculated neglecting the pH dependence of the charges of surrounding amino acid residues. This approximation was considered appropriate for the pH range of about 4 -9, which was used in the NMR titration experiments. DISCUSSION The above NMR evidence leaves no doubt that the proton between the two active-site residues His 235 and Asp 207 gives rise to a downfield-shifted NMR signal, which appears at 15.41 ppm at all accessible pH values in the absence of inhibitor. In the presence of the strong inhibitor thiocyanate, the signal shifts to 19.35 ppm, and the D/H fractionation factor of 0.35 classifies it as originating from a proton involved in a short, strong hydrogen bond. This short, strong hydrogen bond persists at low pH, but reverts back to a normal H-bond (with the signal at 15.41 ppm) under basic conditions (pH Ͼ8). Similar results were reported for acetylcholinesterase (22), ␣-lytic protease (10), and chymotrypsin (3), but with considerably smaller inhibitor-induced shifts (about 3 ppm versus 3.8 ppm) and with a different pH dependence.
These observations can be rationalized on the basis of free energy calculations for a small number of protonation states. Such estimates of free energies (56) and similar continuum solvent approaches have successfully been used to compute pK a shifts of residues in proteins as well as solvation free energies (57). The calculations yield ramifications for the mechanism of the HbHNL-catalyzed cyanohydrin reaction, which are discussed below.
Free Energy Calculations-The catalytic diad (consisting of residues His 235 and Asp 207 ) comprises at least 4 distinct but coupled protonation sites. A discussion of its pH-and inhibitordependent protonation behavior in terms of individual pK a values for each site is complicated because the pK a of each site depends on the state of protonation of the other sites. In other words, there are 16 formally different ways how a system of 4 different sites can be protonated. Not all of these possibilities are equally probable.
To simplify the problem, we have selected four protonation states of the His 235 -Asp 207 pair on the basis of chemical plausibility, as defined in Fig. 7A. Using an appropriately chosen system of thermodynamic cycles in combination with finite difference Poisson-Boltzmann calculations, the relative free energies of these four protonation states of the His 235 -Asp 207 diad were estimated as a function of pH for HbHNL as observed crystallographically in the absence (Fig. 8A) and in the pres- ence (Fig. 8B) of thiocyanate. To our knowledge, this constitutes a novel approach for the quantitative analysis of short, strong hydrogen bonds in proteins, although similar cycles have been used qualitatively to analyze low-barrier hydrogen bonds in proteins (8,9) and in the gas phase (58). Most importantly, the use of such cycles and the treatment of the His-Asp diad as an entity does away with the difficulties in discussing distinct pK a values of the components of the diad.
The titration curves for the His-Asp diad in HbHNL (Figs. 8,   A and B) show that for the free enzyme the anionic state has the lowest free energy except at very low pH. This is in full agreement both with NMR (one signal at 15.41 ppm between pH 4 and 9) and high-resolution crystallographic evidence (density originating from one proton located near N␦ of His 235 ). Among the two states whose relative free energy is pH-independent, the neutral state is lower than the zwitterionic state as a result of the low dielectric constant of the medium and the interaction of the zwitterionic form with the positive charge of Lys 236 . The pK a of His 235 amounts to about 2.5 as judged from the point of intersection of anionic and neutral states (Fig. 8A), which agrees well with a pK a of less than 4 obtained from the NMR experiments. At very low pH the neutral state would be favored, but its observation is prevented by enzyme instability. The cationic state is energetically unfavorable at all pH values.
The (delocalized) negative charge of the thiocyanate inhibitor has two distinct effects on the relative free energies of the four protonation states (Fig. 8B); whereas the free energy of the anionic state is increased (relative to the zwitterionic state) by about 10.6 kcal/mol (equivalent to a shift of the apparent pK a of His 235 from 2.5 to about 10), the better solvation of the neutral compared with the zwitterionic state is compensated as a result of the favorable electrostatic interaction between zwitterion and thiocyanate. The very similar free energy of neutral and zwitterionic states is tantamount to an equality of the pK a s of the two residues involved in the diad, which is a prerequisite for the formation of a short, strong hydrogen bond (3). The latter is indeed observed at all but very high pH, where the SSHB signal at 19.35 ppm reverts to the 15.41 ppm signal of the anionic state. Thus, within the accuracy of the FDPB calculations, the theory provides a perfect rationalization of the NMR results. The quality of the calculations can be estimated from the apparent pK a of His 235 , which is predicted around 10 compared with the value of 8 estimated from NMR.
Reaction Mechanism of HbHNL-Our current view on the molecular mechanism of the HbHNL-catalyzed cyanohydrin cleavage, as deduced from crystallographic (36,37,43), enzyme-kinetic (39), and molecular modeling (38) data, involves the following key aspects (Fig. 1B). 1) The substrate cyanohydrin is attached to the active site by hydrophobic interactions as well as by hydrogen bonding between its hydroxy group and the OH groups of Thr 11 and Ser 80 . 2) Following substrate binding, the OH-Ser 80 is deprotonated by His 235 . This induces the concomitant deprotonation of the substrate hydroxyl by Ser 80 . 3) The subsequent cleavage of the cyanohydrin is facilitated by the stabilization of the charge of the nascent cyanide through interaction with the positive charge of Lys 236 . 4) The formed cyanide ion is eventually protonated by His 235 .
The Complex with Thiocyanate Mimics the Transition State of HbHNL-X-ray crystallographic data are available for the native enzyme (36) as well as for the complex with the reaction product acetone and the inhibitor thiocyanate (37). The threedimensional structure of the complex between HbHNL and a cyanohydrin is only known from modeling studies (38). It is important to note that very little rearrangement occurs in the protein conformation as a result of substrate/inhibitor binding; instead, solvent molecules, some of which are only partially ordered in the native enzyme, become replaced. Of special relevance is a water molecule located near the Lys 236 and His 235 side-chains, which becomes sequestered from solvent upon binding of acetone to the active site. In fact, this water (termed Wat in Fig. 4B) would be ideally positioned for nucleophilic attack on the carbonyl system of the bound acetone. The conjecture that this water molecule in fact occupies the position of the cyanide ion following C-C bond cleavage is also in accord with the stereochemical course of the HbHNL-catalyzed reaction (37). Not surprisingly, this water molecule is replaced by the nitrogen atom of the bound thiocyanate in the corresponding complex structure. A superposition of the active sites of the crystal structures of the complexes of HbHNL with acetone and thiocyanate is shown in Fig. 4B.
Mechanistic Implications-His 235 acts as the general base in HbHNL, because it effects (via Ser 80 ) the crucial step of cyanohydrin deprotonation. This step has to involve a substrateinduced pK a -shift of its side-chain, because in the substratefree form of the enzyme, the His 235 side-chain is in partial contact with solvent and, if its pK a were already in the substrate-free form high enough to deprotonate the cyanohydrin OH (with a pK a around 10.7 (59)), it would become protonated by solvent. In fact, a substrate-induced pK a -shift is exactly what is implied by the above free energy computations and the NMR results. Concommitant with the inhibitor-induced equalization of the energy levels of zwitterionic and neutral protonation states (leading to the observation of a SSHB), the free energy of the anionic state is increased by about 10.6 kcal/mol (i.e. the apparent pK a of His 235 shifts from 2.5 to 10). Insofar as thiocyanate mimics the transition state (see above), this amounts to an increase in basicity of the catalytic base upon approach of the transition state, brought about by the nascent negative charge on the substrate upon deprotonation and subsequent cleavage of the cyanohydrin with formation of a (also negatively charged) cyanide ion. These observations parallel the ones made for ketosteroid isomerase (17).
We showed that a SSHB forms at or near the transition state of the rate-determining step of the enzyme-catalyzed cyanohydrin cleavage, which we consider relevant for understanding the factors enabling proton abstraction from the substrate by the catalytic base. Similar effects also have to be relevant for other enzymes with general acid/base catalysis. We wish to emphasize, however, that our results neither imply nor exclude the possibility for extra stabilization of the transition state due to the formation of a SSHB.