Kinetic Analysis and Probing with Substrate Analogues of the Reaction Pathway of the Nitrile Reductase QueF from Escherichia coli*

The enzyme QueF catalyzes a four-electron reduction of a nitrile group into an amine, the only reaction of this kind known in biology. In nature, QueF converts 7-cyano-7-deazaguanine (preQ0) into 7-aminomethyl-7-deazaguanine (preQ1) for the biosynthesis of the tRNA-inserted nucleoside queuosine. The proposed QueF mechanism involves a covalent thioimide adduct between preQ0 and a cysteine nucleophile in the enzyme, and this adduct is subsequently converted into preQ1 in two NADPH-dependent reduction steps. Here, we show that the Escherichia coli QueF binds preQ0 in a strongly exothermic process (ΔH = −80.3 kJ/mol; −TΔS = 37.9 kJ/mol, Kd = 39 nm) whereby the thioimide adduct is formed with half-of-the-sites reactivity in the homodimeric enzyme. Both steps of preQ0 reduction involve transfer of the 4-pro-R-hydrogen from NADPH. They proceed about 4–7-fold more slowly than trapping of the enzyme-bound preQ0 as covalent thioimide (1.63 s−1) and are thus mainly rate-limiting for the enzyme's kcat (=0.12 s−1). Kinetic studies combined with simulation reveal a large primary deuterium kinetic isotope effect of 3.3 on the covalent thioimide reduction and a smaller kinetic isotope effect of 1.8 on the imine reduction to preQ1. 7-Formyl-7-deazaguanine, a carbonyl analogue of the imine intermediate, was synthesized chemically and is shown to be recognized by QueF as weak ligand for binding (ΔH = −2.3 kJ/mol; −TΔS = −19.5 kJ/mol) but not as substrate for reduction or oxidation. A model of QueF substrate recognition and a catalytic pathway for the enzyme are proposed based on these data.

Reduction of a nitrile group to a primary amine is a transformation well known and important to synthetic chemistry (1)(2)(3)(4). Its unique biological equivalent is the reaction catalyzed by QueF enzymes (5). QueF was discovered from the biosynthetic pathway of the tRNA-modified nucleobase queuosine (Q) 2 where it was shown to utilize NADPH for the conversion of 7-cyano-7-deazaguanine (preQ 0 ) into 7-aminomethyl-7deazaguanine (preQ 1 ) (Fig. 1). Until today, QueF remains the only enzyme known to promote the intricate nitrile-to-amine chemistry. The catalytic mechanism of QueF is therefore of significant fundamental interest. In addition, it also has potential application relevance. Q is a nucleoside from the wobble position of the tRNAs for Asn, Asp, His, and Tyr, and it is thus required for translation efficiency and fidelity (6). Despite its ubiquitous occurrence in nature, Q is synthesized de novo only in bacteria (7)(8)(9). The enzymes of the Q pathway, QueF in particular, due to its apparent uniqueness, therefore represent promising drug targets to combat bacterial infections selectively. Moreover, due to the hazardous conditions required in the chemical reaction (1)(2)(3)(4), "greener" routes of nitrile reduction are highly demanded. A biocatalytic route going by the QueF mechanism presents an interesting option.
QueF enzymes have been characterized structurally from unimodular and bimodular classes (10,11), and the herein studied QueF from Escherichia coli is a bimodular protein. The bimodular QueF is a functional homodimer, whereas the unimodular QueF folds into a homodecamer (a dimer of pentamers) (5,10,11). In both types of QueF, the active site is positioned at a structural interface, created from different subunits in the unimodular QueF and from the tandem tunneling-fold domains of the same subunit in the bimodular QueF. The arrangement of catalytic groups is highly conserved in both enzymes (10 -12).
The basic QueF mechanism was delineated in biochemical and structural studies of the unimodular enzyme from Bacillus subtilis ( Fig. 1) (11,13). The binding of preQ 0 involves a large induced fit, resulting in the functional active site to be formed and to become completely secluded from solvent. Cys 55 then attacks the nitrile group to build a covalent thioimide, where it was possible to trap both in the crystal and in solution. QM/MM calculations on the bimodular QueF from Vibrio cholerae suggested a role for Asp 201 (Asp 62 in B. subtilis QueF) * This work was supported by the Federal Ministry of Economy, Family and as a proton shuttle during the thioimide formation (14) (Fig. 1). Hydride reduction of the thioimide by NADPH, also under protonic assistance by Asp, subsequently gives a covalent hemithioaminal. NADP ϩ is exchanged by NADPH and the C-S bond cleaved to yield a non-covalent imine intermediate, which is finally reduced to preQ 1 . The QM/MM calculations suggested that Asp 201 protonates the Cys 194 in V. cholerae QueF on hemithioaminal breakdown and provides electrostatic stabilization during the subsequent hydride transfer (14). Free energy barrier analysis indicated that the imine reduction is slowest among the chemical steps of the computed reaction pathway for the V. cholerae QueF. In B. subtilis QueF, the k cat (0.011 s Ϫ1 ) (5,13) is much smaller than the rate constant of thioimide formation (2.78 s Ϫ1 ) (11). Therefore, this also locates the ratedetermining step in the sequence of steps involved in the NADPH-dependent reductions.
Despite these preceding insights into QueF structure and function, fundamental issues of the enzymatic mechanism remain. Which interactions between QueF and preQ 0 are required in the induced-fit binding of the substrate or in the formation of a competent thioimide intermediate? How does NADPH bind and which of its diastereotopic hydrogens at the nicotinamide C4 is used for hydride transfer in each reduction step? Is the very small k cat of QueF really limited by one of the hydride transfer steps, as QM/MM calculations have suggested (14), or are rather the associated physical steps (which were not analyzed computationally) slow? To address these questions, we prepared tailored analogues of preQ 0 , in which specific recognition sites for QueF binding were removed or altered by targeted organic synthesis (Fig. 2). Together with the native preQ 0 , we used these analogues to characterize the kinetics and thermodynamics of substrate binding by the QueF enzyme from E. coli (ecQueF). We applied transient and steady-state kinetic analysis in combination with kinetic simulations to characterize and to determine rate constants for the individual steps of the enzymatic reaction pathway. We also developed a synthesis of 7-formyl-7-deazaguanine (Fig. 2) as a carbonyl surrogate of the imine intermediate of the QueF reaction. We characterized the binding of this analogue to ecQueF and examined its reactivity as substrate for enzymatic reduction or oxidation. A model of substrate binding recognition by ecQueF is suggested based on the evidence presented. We show that each reduction step by the enzyme involves transfer of the pro-4-R-hydrogen of NADPH. We also show that the hydride transfer steps are rate-determining in the overall reaction.

ITC Study of Substrate and Coenzyme
Binding to ecQueF-ITC experiments were conducted to determine the thermodynamic characteristics of substrate and NADPH binding to  ecQueF (25°C, pH 7.5; 100 mM sodium phosphate buffer). The results are shown in Fig. 3 and the parameters calculated from the data are summarized in Table 1. The binding of preQ 0 was a strongly exergonic process (⌬G Ͻ0) in which a highly favorable enthalpy term (⌬H) overcompensated the non-favorable contribution from the binding entropy (ϪT⌬S). The 39 nM dissociation constant (K d ) thus determined was in good agreement with the K d estimate of 36 nM obtained from a gel filtration analysis of the unimodular QueF from B. subtilis (11). The number of preQ 0 -binding sites occupied in the ecQueF homodimer was determined from the ITC data to be unity (Ϯ5%). Binding models comprising two equivalent or distinct binding sites/enzyme homodimer were also examined but were inconsistent with the experimental results. The half-of-thesites reactivity of ecQueF implied by these findings was in line with the structural suggestion that the minimal catalytic unit of V. cholerae QueF should be the protein homodimer (10).
Besides analyzing preQ 0 binding in the standard phosphate buffer, we also determined the ⌬H of preQ 0 binding in Tris and HEPES buffer (Table 2 and Fig. 3H). Because the three buffers differ strongly in ionization enthalpy (⌬H ion ), the proton release or uptake in conjunction with ligand binding can be determined from the difference in the ⌬H of binding. We showed that in terms of ⌬G, the overall preQ 0 binding to ecQueF was hardly affected by the buffer used (Tables 1 and 2). The binding stoichiometry was also unchanged on variation of the buffer. However, there was a large buffer effect on ⌬H (Table 2), and we determined that 0.78 Ϯ 0.09 protons were taken up by each ecQueF dimer on binding of preQ 0 under the conditions used (pH 7.5, Table 1).
The 2-deamination of preQ 0 resulted in a drastic weakening of the substrate binding, reflected in a 21.5 kJ/mol loss in the ⌬G stabilizing the ecQueF-bound state, comparing 2-deamino-preQ 0 to preQ 0 . This is equivalent to a 6 ϫ 10 3 -fold increase in the K d of the 2-deaminated substrate. The effect in ⌬G was due to a strongly lowered enthalpy contribution to the binding of 2-deamino-preQ 0 as compared with preQ 0 (⌬⌬H ϭ ϩ51.4 kJ/mol). The entropy term was, however, less unfavorable for the binding of 2-deamino-preQ 0 than it was for the binding of preQ 0 . It thus compensated to some degree for the very large binding enthalpy change due to removal of the 2-amino group. We also examined the binding of 6-deoxo-preQ 0 but were unable to obtain a binding isotherm due to the low water solubility of this compound.
Contrary to the binding of preQ 0 and 2-deamino-preQ 0 in which both were enthalpy-driven processes, the binding of 7-formyl-preQ 0 was hardly exothermic. It therefore relied completely on a favorable entropy contribution to become a spontaneous process characterized by a negative ⌬G. Indeed, the ϪT⌬S term exhibited sign change from positive to negative and hence turned from being non-favorable to favorable for binding when the binding of preQ 0 (or 2-deamino-preQ 0 ) was compared with that of 7-formyl-preQ 0 . Despite the completely different thermodynamic signatures of the binding of 2-deamino-preQ 0 and 7-formyl-preQ 0, the dissociation constants of these two ligands were similar ( Table 1).
Removal of the active-site nucleophile in an inactive C190A variant of ecQueF weakened the binding of preQ 0 (K d ϭ 5.5 M) about 141-fold compared with preQ 0 binding to the wild type enzyme. This reflected a drastically lowered enthalpy contribu-tion to the binding (⌬⌬H ϭ ϩ35.6 kJ/mol) in the enzyme variant as compared with wild type ecQueF. The entropy term of preQ 0 binding to the C190A variant was positive (ϪT⌬S ϭ ϩ14.6 kJ/mol). The results are summarized in Table 1. As in the wild type enzyme, the ⌬H of preQ 0 binding to the C190A variant was strongly dependent on the buffer used (phosphate, HEPES, Tris). The ⌬G of binding was however hardly affected by the buffer (Table 2). We determined from the data that 0.46 Ϯ 0.27 protons were taken up by the C190A variant on binding of preQ 0 (pH 7.5; Table 1).
The binding of NADPH by QueF is generally not well understood. We show here ( Fig. 3; Table 1) that NADPH was bound by the free forms of wild type ecQueF and the C190A variant thereof. With both enzymes the binding occurred in an enthalpy-driven fashion. The overall binding was highly exergonic and characterized by 16.0 and 18.1 M dissociation constants for the wild type enzyme and the C190A variant, respectively. These thermodynamic characteristics appear consistent with a binding process involving biological recognition and are unlikely to reflect a merely adventitious binding of NADPH. If so, then the binding of NADPH to ecQueF would not require preQ 0 to have been bound to the enzyme before. Additionally, we determined that NADPH binds to the binary complex of C190A and preQ 0 with about 5-fold enhanced affinity (K d ϭ 3.6 M) compared with its binding to the free enzyme. The binding stoichiometry was 1 NADPH/enzyme dimer. Despite the relatively small difference in ⌬G of binding (⌬⌬G ϭ Ϫ4 kJ/mol), the thermodynamic signature of NADPH binding to the C190A-preQ 0 complex was quite distinct from that of NADPH binding to the free enzyme. The binding of NADPH to the C190A- f Data were obtained with the noncovalent C190A-preQ 0 complex. g The number of protons (n Hϩ ) taken up by the enzyme on preQ 0 binding was obtained from the measured enthalpies in sodium phosphate (shown in this table), HEPES, and Tris buffers at pH 7.5 (for the thermodynamic data of preQ 0 binding in HEPES and Tris buffers, see Table 2). preQ 0 complex lacked a contribution from entropy (ϪT⌬S ϭ ϩ0.07 kJ/mol), and the enthalpy contribution was lowered (⌬⌬H ϭ ϩ16.2 kJ/mol) in comparison with NADPH binding to the free enzyme. Note: NADPH binding to the covalent adduct between wild type enzyme and preQ 0 does not give a stable complex because catalytic reduction takes place. It was therefore not examined with ITC. Substrate Binding and Covalent Thioimide Formation-As in B. subtilis QueF (11), the substrate binding in ecQueF was traceable by both quenching and blue shifting of the protein's Trp fluorescence (Fig. 4, A and B). Trp 57 and Trp 198 in the substrate-binding pocket (Fig. 5) are the likely reporter groups. On addition of preQ 0 , the ecQueF fluorescence (excitation, 280 nm) was quenched, and a slight blue shift in the maximum wavelength of emission ( max , 334 3 330 nm) was observed. The quenching yield increased dependent on the molar ratio of preQ 0 and the enzyme homodimer used. It reached a maximum of about 0.7 when this ratio was approximately unity (Fig. 4A). The titration of ecQueF with 2-deamino-preQ 0 and 6-deoxo-preQ 0 also produced fluorescence quenching, but the corresponding yields were lower, 0.21 and 0.36, respectively (Fig. 4B), than for preQ 0 . The 7-formyl-preQ 0 behaved differently. Its incubation with ecQueF caused a slight enhancement of the protein fluorescence (ϳ10%), and the max was up-shifted (334 3 336 nm), as shown in Fig. 4B. Binding of preQ 0 to the C190A variant of ecQueF caused only a small amount of fluorescence quenching (ϳ15%).
Formation of the covalent thioimide between ecQueF and preQ 0 was detectable by appearance of an absorbance band with maximum absorption at 380 nm (Fig. 4C). A molar extinction coefficient of 10.02 Ϯ 0.14 mM Ϫ1 cm Ϫ1 was determined. With this, we could show that the preQ 0 was linked to each ecQueF dimer in a 1:1 stoichiometry. Results of protein mass analysis ( Fig. 4D) were in good agreement with these findings, showing that the sample from incubation of ecQueF with preQ 0 was composed of about equal amounts of the monomeric enzyme-preQ 0 adduct (43%) and also the monomeric apoenzyme (57%). Reactivity of half-of-the-sites in the ecQueF dimer was thus strongly supported. Time-resolved titration analysis in stopped-flow experiments also revealed that a molar preQ 0 concentration equivalent to that of the ecQueF dimer was sufficient for complete covalent conversion of the available active sites (Fig. 6).
Incubation of ecQueF in the presence of 2-deamino-preQ 0 also gave rise to a new absorbance band, but compared with the preQ 0 thioimide adduct, its maximum absorption was at a lower wavelength of around 340 nm (Fig. 4C). A molar extinction coefficient of 10.50 Ϯ 0.04 mM Ϫ1 cm Ϫ1 was determined. Contrary to preQ 0 , a substrate/enzyme molar ratio of about 13 was required so that covalent conversion of ecQueF with 2-deamino-preQ 0 was complete (Fig. 4E). Protein mass analysis confirmed the presence of the covalent adduct and furthermore revealed that the portion of total ecQueF covalently modified (maximum, ϳ40 -50% of dimer) was strongly dependent on the amount of 2-deamino-preQ 0 offered (Fig. 4D). A K d of 155 M was determined for 2-deamino-preQ 0 binding to ecQueF (Fig.  4F), in useful accordance with the results of the ITC experiments (Table 1). No evidence of covalent linkage formation between ecQueF and 6-deoxo-preQ 0 or ecQueF and 7-formyl-preQ 0 was found in spectrophotometric titration analysis. Likewise, the C190A variant did not show evidence of covalent binding of preQ 0 , as expected.
Addition of preQ 0 to a pre-incubated mixture of ecQueF with any of the three preQ 0 analogues reinstated in each case the effect on fluorescence quenching and "thioimide absorbance band" formation that preQ 0 alone had. Therefore, this showed effective competition of preQ 0 with the other ligands for binding to ecQueF and also demonstrated the reversibility of thioimide covalent linkage formation by 2-deamino-preQ 0 (Fig. 4C).

Stereospecificity of ecQueF for Hydrogen Transfer from
NADPH-This central characteristic of the QueF mechanism has not been elucidated in previous biochemical and structural studies of the enzyme. We used NADPH, (4R)-[ 2 H]NADPH, or (4S)-[ 2 H]NADPH for the reduction of preQ 0 and analyzed with 1 H NMR the products formed, NADP ϩ and preQ 1 . Results are shown in Fig. 7. In all reactions, the conversion of preQ 0 into preQ 1 was traced by monitoring the singlet signal of the deazaguanine C8 hydrogen, which appears at 7.42 ppm in substrate and at 6.76 ppm in product. When NADPH was used, the proton spectra of the product exhibited a singlet peak at 4.08 ppm, which is assigned to the hydrogens of the aminomethyl carbon of preQ 1 (Fig. 7A). This peak was absent from the spectra of the preQ 1 released in the reaction with (4R)-[ 2 H]NADPH (Fig. 7B). Besides, the product of the oxidation of (4R)-[ 2 H]NADPH was NADP ϩ and not 4-[ 2 H]NADP ϩ . The hydrogen at position C4 on the nicotinamide ring was identified unequivocally from the doublet signal at 8.67-8.68 ppm (Fig. 7B). When the alternatively deuterium-labeled (4S)-[ 2 H]NADPH was used, only the hydrogen was transferred in the enzymatic reaction of preQ 0 , and authentic preQ 1 was obtained together with 4-[ 2 H]NADP ϩ (Fig. 7D). These results are clear in showing that ecQueF is specific for transferring the 4-pro-R-hydrogen from NADPH, and it does so in both hydride reduction steps of the catalytic conversion of preQ 0 to preQ 1 .
Steady-state Kinetic Study of ecQueF-Based on the measurement of both NADPH consumption and amine product formation ( Fig. 8), we showed that besides preQ 0 the 2-deamino derivative was also a workable substrate of ecQueF (see below). The 6-deoxo-preQ 0 was inactive within the detection limits of the assay used. The 7-formyl-preQ 0 (100 M) was not reduced by ecQueF in the presence of NADPH (150 or 300 M) nor was it oxidized in the presence of NADP ϩ (300 M). Absence of activity with this substrate was shown at different pH values in the range 5.5-7.5 (100 mM ammonium acetate buffer, pH 5.5 or pH 6.7; 100 mM Tris buffer, pH 7.5). Considering the poor solubility of 7-formyl-preQ 0 in water, we added 7-formyl-preQ 0 (dissolved in DMSO) directly to the enzyme solution contain-  ing NADPH so that enzymatic reduction could precede substrate precipitation. However, no reaction was observed, despite the use of ecQueF concentrations (7.6 -30.0 M) that would have allowed detection of just 0.1 enzyme turnover. Just to note, oxidation by NADP ϩ might have involved the 7-formyl group in native or hydrated form, requiring nucleophilic catalysis from active-site cysteine in the former but not the latter (15)(16)(17)(18). Enzymatic conversion of 7-formyl-preQ 0 was also analyzed in the presence of (NH 4 ) 2 SO 4 or NH 4 Cl (each at 50 mM), considering that addition of the ammonium ions might facilitate the formation of an incipient imine to be reduced by QueF. However, no reaction was observed under these conditions.
We furthermore examined the reverse reaction of ecQueF, measuring reduction of NADP ϩ (100 M), and product formation associated with it, in the presence of preQ 1 (40 M). Using the standard pH of 7.5 but also elevated pH values of up to 9.0 in order to facilitate deprotonation of the 7-amino group, there was no conversion of preQ 1 above the detection limit, which was roughly 0.1 turnover of the molar enzyme concentration used (20 -80 M). To prevent the product attached to the enzyme from escaping detection, we denatured the protein with methanol and repeated the HPLC analysis, thereby confirming the absence of an oxidation of preQ 1 under the conditions used.
Initial rate analysis showed that the K M value for NADPH was not affected by variation of the preQ 0 concentration in the range 5-100 M. The lower end limit of the substrate concentration was defined by the time resolution and sensitivity of the spectrophotometric assay. Considering the 39 nM K d value for preQ 0 , only a negligible portion of free enzyme could have existed at steady state under these conditions. Therefore, this explains a coenzyme K M apparently independent of the substrate concentration. Results are summarized in Table 1. In terms of k cat , 2-deamino-preQ 0 was an equally good nitrile substrate as was preQ 0 . The NADPH K M was, however, elevated about 4-fold in the reaction with 2-deamino-preQ 0 ( Table 1). A huge difference existed in the substrate K M , consistent with results of the ITC binding study of preQ 0 and 2-deamino-preQ 0 .
Transient Kinetic Study and Kinetic Simulation-To characterize individual steps of the ecQueF reaction, a transient kinetic analysis in stopped-flow experiments was performed. Fig. 6 shows representative time courses of thioimide formation upon mixing the enzyme with preQ 0 in the absence of NADPH. The absorbance increased exponentially with an amplitude determined by the limiting concentration of ecQueF dimer or preQ 0 . The rate constants k obs (1.35 Ϯ 0.15 s Ϫ1 ) were independent of the initial concentrations of enzyme and preQ 0 . Using global fitting with COPASI, a two-step binding-reaction model, E ϩ S % E⅐S 3 E-S, was useful to describe the data recorded at varied enzyme and substrate concentrations (Fig. 6, A and B). It involved a relatively fast binding of preQ 0 with a dissociation constant (K d _kinetic) of 2.63 (Ϯ0.01) M that was followed by a slower and, within the limit of accuracy of the method used, effectively irreversible transformation of the non-covalent E⅐S into the covalent complex E-S with a rate constant (k 3 ) of 1.63 (Ϯ 0.01) s Ϫ1 ( Table 3). The obtained k 3 was in a good agreement with the experimentally determined k obs . The relationship between the overall 39 nM dissociation constant determined by ITC and the kinetically determined parameters, i.e. K d _overall ϭ K d _kinetic/(1 ϩ k 3 /k 4 ), can be used to calculate an approximate value of 0.024 s Ϫ1 for the thioimide release rate constant (k 4 ). Fig. 9A shows absorbance time traces at 340 and 380 nm for a reaction under conditions of a single turnover of the enzyme present. Time traces for reactions under multiple turnover conditions in which a limiting enzyme concentration (5 M) was used are shown additionally in Fig. 9, B-D. The single-turnover time traces were characterized by a fast initial increase in absor-   k 3 /k 4 ). The K d _overall is 39 nM. k 4 was not included in the global fitting in COPASI. If it was included, its value was close to zero. Therefore, the k 4 value shown probably gives an upper bound for the dissociation rate constant. b k 10 was obtained from the relationship k 3 k 5 /k 4 k 6 ϭ k 7 k 9 /k 8 k 10 . Note that k 10 was not included in the global fitting in COPASI. If it was included, its value was close to zero. The k 10 value shown probably gives an upper bound for the dissociation rate constant. c KIEs were obtained by comparing rate constants of preQ 0 reduction by NADPH, as shown in the bance, at 380 nm in particular, which reflected formation of the covalent intermediate (Fig. 9A). The subsequent decrease in absorbance was due to NADPH consumption mainly but also included the effect of the concomitant decay of the intermediate, especially when single-turnover conditions were used. Fitting the initial increase in absorbance with an exponential function (Fig. 9D), the corresponding k obs showed a hyperbolic dependence on the NADPH concentration, with a K d of 3.0 (Ϯ 1.6) M and a maximum value of 7.32 (Ϯ 0.46) s Ϫ1 (Fig. 9E). The presence of NADPH therefore resulted in a 4.5-fold speeding up of the formation of the thioimide intermediate. The decrease in absorbance at both 340 and 380 nm was best fit with a single exponential (single-turnover reactions) or a straight line (multiple-turnover reactions). The rate constants calculated from these fits were hyperbolically dependent on the NADPH concentration, with a K d of 10 (Ϯ 2) M and a maximum value of 0.15 (Ϯ 0.01) s Ϫ1 , which is similar to the k cat (Fig. 9F). The kinetic model in Scheme 1 was used to fit the data from all experiments, comprising a large set of averaged stoppedflow time traces from 28 independent reaction conditions, involving variation in NADPH concentration between 10 and 150 M. Fig. 9, A-C, compares the experimental progress curves to the corresponding fitting results, which reveals useful agreement between the two. Table 3 summarizes the rate and binding constants thus determined. The kinetic steps of thio-imide and imine reduction were the slowest in the enzymatic pathway, and their corresponding rate constants were similar. Note that based on a decrease in NADPH absorbance (at 340 nm) alone, it would not have been possible to distinguish the two steps. However, because the thioimide reduction also involves concomitant decrease in absorbance at 380 nm, the first hydride transfer step has a spectral signature different from the second. This could be used to determine the rate constant associated with each step. The rate constants of the microscopic reaction steps (Table 3) were consistent with the k cat measured at steady state (Table 1) and also with the rate constants from stopped-flow experiments. Fig. 9G shows the distribution of different enzyme forms in a single-turnover stopped-flow reaction. The enzyme complexes with preQ 0 prevail during the reaction. Because of the comparably weak binding of NADPH to the covalent enzyme-preQ 0 complex, NADPH that was initially bound to enzyme prior to formation of the covalent adduct was partly released from the enzyme once the thioimide intermediate had been formed. This can be seen in the very early phase of the reaction shown in Fig. 9H.
Three features of the kinetic model in Scheme 1 should be emphasized. First, there are three binding steps for NADPH (K 1 to K 3 ). Simpler models, in which NADPH binding for thioimide reduction occurred only to the covalent enzyme-preQ 0 adduct or NADPH binding occurred to the non-covalent ternary com- Orange solid lines in A-C present fit of the data to the kinetic mechanism described in Scheme 1, and blue solid lines in D indicate single exponential fit of the data. E and F, dependence of stopped-flow rate constants (k obs ) on the NADPH concentration is shown. The circles are the data, and the solid line is a hyperbolic fit. k obs of formation of the thioimide adduct is shown in E. It was obtained from single exponential fits of the data in D. k obs of the NADPH consumption is shown in F. It was obtained from linear fits on the data in B (1-10 s range). G and H, reaction simulation under single-turnover conditions applying the kinetic parameters from Table 3  plex and NADPH dissociation from the covalent adduct was not allowed, failed to describe the experimental time courses properly. Second, the extremely small dissociation constant for NADPH binding to the hemithioaminal/imine intermediate is worth noting. It suggests that NADPH was bound almost irreversibly by ecQueF at this step. Unless having this small value of K 3 , the dependence of the reaction on the NADPH concentration could not be properly explained. Third, the covalent ternary complex showed a much lower absorbance both at 340 nm (ϳ54%) and 380 nm (ϳ60%) than expected from the sum of the corresponding individual absorbances of NADPH and thioimide adduct. Note that the absorbance of NADPH was not affected by binding to ecQueF. The result is interesting as it suggests that the thioimide adduct and NADPH perturb each other electronically in the ternary complex. More interestingly, even the extent of this electronic perturbation was attenuated strongly when the ternary complex was formed from (4R)-[ 2 H]NADPH instead of NADPH. The electronic effect and its dependence on the isotopic substitution in NADPH might be explained by a "near-attack" ground-state conformer that involves close alignment, and perhaps even partial bonding, between the thioimide carbon and the C4 nicotinamide hydrogen of NADPH. Electronic preorganization of the ternary complex for efficient hydride reduction is therefore suggested.
KIE Study-Substitution of the reactive 4-pro-R-hydrogen of NADPH by deuterium caused a slowdown of the enzymatic preQ 0 reduction. The effect was expressed in a substantial primary KIE on both the maximum rate ( D V max ϭ 2.40 Ϯ 0.12) and the catalytic efficiency for coenzyme ( D V max /K M ϭ 2.68 Ϯ 0.41).
Stopped-flow progress curves of preQ 0 reduction by (4R)-[ 2 H]NADPH under multiple-turnover reaction conditions are shown in Fig. 10. The corresponding single-turnover reactions are also presented there. Exponential fit of the initial absorbance increase revealed that the deuteration of NADPH did not cause slowdown of the formation of the thiomide intermediate (k obs ϭ 6.4 Ϯ 0.3 s Ϫ1 ), as expected. The dependence of k obs on the NADPH concentration was also unaffected (K d ϭ 3.0 Ϯ 1.5 M), as shown in Fig. 10, D and E. Rate constants determined from fits of the absorbance decrease in individual stopped-flow traces are shown on Fig. 10F. From the hyperbolic dependence of k obs on the NADPH concentration, a maximum rate constant of 0.055 (Ϯ 0.003) s Ϫ1 was obtained, and the corresponding K d was 14 (Ϯ 3) M. Therefore, the deuteration of NADPH affected strongly the enzymatic rate, although the apparent binding was hardly influenced.
Results of a global fit of Scheme 1 to the data are shown in the figures, and the reaction constants determined are summarized in Table 3. Deuteration of the coenzyme caused slowdown of both reduction steps, but the effect was by far more pronounced on the thioimide conversion. The corresponding KIEs were determined as 3.3 Ϯ 0.1 (thioimide reduction) and 1.8 Ϯ 0.1 (imine reduction). The value of D V max can be explained from these two KIEs (Table 3).

Groups of preQ 0 Required for Binding Recognition and Covalent Adduct Formation by ecQueF-Structural and sequencebased comparisons reveal the bimodular QueFs (from V. chol-
erae and E. coli) to exhibit substrate binding pockets highly similar to that of their unimodular counterpart (from B. subtilis) for which a preQ 0 complex structure has been determined (10 -12). Evidence characterizing the substrate binding in ecQueF is therefore interpreted on the basis of a high degree of functional residue conservation in these enzymes. Comparing the changes in protein fluorescence, the heat released, and the amount of covalent thioimide adduct formed in titrations of ecQueF with preQ 0 (or analogues thereof), the overall substrate binding is found to be composed of three main elements. The initial accommodation of the substrate in the binding pocket lacks a spectroscopic signature in absorbance and fluorescence, but by analyzing the binding of the non-reactive substrate analogue 7-formyl-preQ 0 , it becomes traceable with ITC. Results show that this first step in the binding is driven exclusively by loss in entropy, which probably reflects the desolvation of the substrate, the binding site in ecQueF, or both. The complete lack of an enthalpic contribution to the binding of 7-formyl-preQ 0 ( Table 1) suggests that specific interactions are not developed between this ligand and the enzyme. The second step involves a protein conformational rearrangement, detectable mainly by ITC but also by fluorescence, which, in analogy to the induced-fit binding of preQ 0 in the B. subtilis enzyme (11), is assumed to preorganize and close up the ecQueF active site. Trp 198 of ecQueF (like the corresponding Trp 119 of B. subtilis QueF) would thus be placed in a more hydrophobic microenvironment, with consequent effects on their fluorescence properties. The preQ 0 complex structure of the C55A variant of B. subtilis QueF adopts the "closed" conformation just as the SCHEME 1. Proposed kinetic mechanism of preQ 0 reduction by ecQueF is shown. E is the free ecQueF, and the enzyme-bound preQ 0 and NADPH are indicated by subscript and superscript, respectively. An asterisk on E indicates the covalent thioimide linkage between preQ 0 and enzyme. E imine indicates enzyme with bound imine intermediate. The steps included in kinetic simulations are indicated in black. NADPH binding to E was excluded from the kinetic simulation and is indicated in light gray. The kinetic constants from simulation and fitting of reaction time courses are shown in Table 3. The following rate constants were obtained only as ratios to give the corresponding dissociation constants in rapid equilibrium: K d ϭ k 2 /k 1 (green); K 1 ϭ k 6 /k 5 (light blue); K 2 ϭ k 8 /k 7 (blue); and K 3 ϭ k 13 /k 12 (purple). The chemical and isotope-sensitive steps are thioimide reduction (k 11 ) and imine reduction (k 14 ). These steps were assumed to be effectively irreversible under the reaction conditions used. Likewise, formation of the thioimide intermediate (k 3 and k 9 ) was assumed to be irreversible in the fitting because the rate constants of the reverse reaction (k 4 and k 10 ) are very small in comparison.
wild type enzyme does (11), indicating that thioimide formation is not essential for the induced fit. Evidence from the current study adds to the picture, showing that the cyano group of preQ 0 is crucial in substrate-binding recognition, for its substitution by a formyl group disrupts entirely the ligand-induced fit in the enzyme. The finding appears relevant biologically in that it suggests a mechanism by which QueF discriminates between its substrate and other metabolites of the queuosine biosynthetic pathway that differ from preQ 0 only in the chemical group at position 7 of the deazaguanine core.
The overall ⌬G of preQ 0 binding to the catalytic Cys variant (C190A) of ecQueF includes a substantial contribution from enthalpy (⌬H ϭ Ϫ44.7 kJ/mol; Table 1), which favors binding. Unlike the basal contribution from "desolvation" entropy registered for the binding of 7-formyl-preQ 0 , the contribution from entropy for preQ 0 binding to the C190A variant is nonfavorable (ϪT⌬S Ͼ0; Table 1) in this case. The favorable ⌬H term probably arises from specific non-covalent interactions between the C190A variant and preQ 0 (cf. Fig. 5), formed as a result of the induced-fit conformational change, which, in turn, according to our interpretation also explains the non-favorable ϪT⌬S term in terms of a more compact macromolecular structure. We also note the good agreement between the dissociation constants for non-covalent preQ 0 complexes of the wild type enzyme (K d _kinetic; Table 3) and the C190A variant (Table 1). Most likely in virtue of their effect on substrate positioning, to align precisely the cyano group of the preQ 0 with the thiol group of Cys 190 , these non-covalent interactions are key to promote the covalent thioimide adduct, as shown by the rela-tive inefficacy in that regard of preQ 0 analogues lacking the 2-amino or the 6-oxo group. The overall thermodynamic signature of preQ 0 binding to ecQueF is strongly shaped by the energetics of covalent bond formation, which almost completely masks the effects of the non-covalent interactions.
On binding of preQ 0 at pH 7.5, each ecQueF dimer was shown by ITC to take up 0.78 Ϯ 0.09 protons from the bulk solvent. Preliminary evidence from time-resolved experiments performed in the stopped-flow apparatus and measuring proton uptake with a pH indicator further reveals that the proton uptake occurred concurrently with formation of the thioimide intermediate. A proton is required during conversion of the nitrile to the thioimide group. In the case that the observable proton uptake by ecQueF was an immediate consequence of this reaction, binding of preQ 0 by an enzyme variant that is incapable of thioimide adduct formation would accordingly not involve proton uptake. Clear evidence of proton uptake also by the C190A variant therefore eliminated this mechanistic possibility. An alternative mode of protonation, occurring in consequence of the induced-fit protein conformational change in preQ 0 binding, is therefore suggested. Identification of the groups involved in the protonation requires further study.
Kinetic Pathway of ecQueF Involves Thioimide Reduction as the Rate-limiting Step-Because preQ 0 and NADPH both bind to free ecQueF, the kinetic mechanism appears to be random in principle. However, in the global simulation-fitting analysis with COPASI, binding of NADPH to the free enzyme appeared to be not kinetically significant, at least under the conditions used. In addition, although preQ 0 forms a catalytically compe-  DECEMBER 2, 2016 • VOLUME 291 • NUMBER 49 tent thioimide adduct with ecQueF even in the absence of NADPH, the preferred reaction path when neither preQ 0 nor NADPH is limiting appears to be through the non-covalent ternary complex (Scheme 1). Formation of the thioimide adduct is accelerated about 3.6-fold when NADPH is present, and it is relatively slow compared with the actual substrate binding steps (Table 3). The preQ 0 substrate binds in a stoichiometry of 1 per ecQueF dimer. The possibility of half-of-thesites reactivity was also hinted at by the structure of V. cholerae QueF (10), suggesting that only a single NADPH would be bound in two possible orientations to load one of the two active sites. Based on the structural model of ecQueF, Wilding et al. (12) arrived at the same conclusion, suggesting however that only one of two catalytic centers in the protein dimer was accessible, whereas the other was buried deeply inside the protein.

Reaction Pathway of the Nitrile Reductase QueF from E. coli
Evidence from the study of the inactive C190A variant of ecQueF showed that 1 NADPH was bound to each non-covalent complex between enzyme dimer and preQ 0 .
According to Scheme 1, the observable kinetic parameters from initial rate measurements are related to rate constants as shown in Equations 1 and 2 (19).
Because of the slow breakage of the covalent thioimide linkage (Table 3), the preQ 0 K M value tends to be far too low to be experimentally measurable. The results in Table 3 show that V max was limited, to a similar degree, by the rate constants of thioimide and imine reduction. Rate limitation solely by imine reduction, as suggested by Ribeiro et al. (14) from their computational study of the reaction of V. cholerae QueF, was ruled out for ecQueF based on the kinetic evidence. The stopped-flow progress curves were not consistent with a kinetic mechanism in which the first hydride transfer was faster than the second. The K M (NADPH) mainly reflects the binding of NADPH to the covalent adduct of ecQueF and preQ 0 (K 1 ). A reaction pathway in which the enzyme "loosens its grip" on NADPH while moving from the non-covalent to the covalent complex with the preQ 0 substrate requires comment; however, we believe there might be biological significance to it. Under the premise of ecQueF being present intracellularly in a molar concentration comparable with that of preQ 0 , thioimide adduct formation from an enzyme-preQ 0 -NADPH complex, in which the NADPH is bound tightly (K 2 ), might serve to ensure that the available preQ 0 is trapped fast and completely in a covalently enzyme-bound form. Availability of NADPH in relation to K 1 would determine the portion of total enzyme adduct able to undergo conversion into preQ 1 . The remaining portion, despite being unreactive in the absence of bound NADPH, might still be useful, for it could temporarily "store" part of the preQ 0 until reduction occurs. Because flux through the "normal" reaction pathway (k cat ϭ 0.15 s Ϫ1 , Table 3) exceeds the "off-pathway" thioimide cleavage, loss of adduct to dissociation could thus be prevented effectively. The very small K 3 for the binding of NADPH for imine reduction would help to ensure that any preQ 0 that enters the reduction pathway is taken completely to the preQ 1 product. Considering that the preQ 0 -derived imine might immediately hydrolyze to the corresponding aldehyde on being exposed to water, an important task of the enzyme should be to avoid any partial reduction of the nitrile substrate.
Evidence that in each step of the NADPH-dependent reduction by ecQueF the hydride transfer took place by 4-pro-R stereoselectivity provided the basis for employing primary deuterium KIEs as selective probes of the enzymatic reaction. There are two isotope-sensitive steps in the enzymatic mechanism (Scheme 1). Results of stopped-flow kinetic analysis suggested a much larger KIE on the thioimide reduction than on imine reduction. The directly measured KIE on V max is explicable as the combined result of the two individual KIEs. According to Scheme 1, the V max /K M for NADPH includes in principle all steps from the binding of NADPH up to the product release after the imine reduction. However, because K 3 is much smaller than K 1 , the actual V max /K M value for ecQueF reflects the thioimide reduction, and also the KIE is therefore primarily from this step. Note that the V max /K M value determined at a saturating preQ 0 concentration implies reaction via binding of NADPH to the covalent thioimide adduct. The D V max /K M contained a relatively high standard error, but its value of 2.68 (Ϯ 0.41) appears largely consistent with the calculated KIE of 3.3 on the thioimide reduction step.
An interesting feature of the preQ 0 conversion by ecQueF was that a reverse reaction of preQ 1 and NADP ϩ was not experimentally detectable, despite the use of a broad range of pH conditions to evaluate the reactivity of preQ 1 protonated and unprotonated. Note: the pK a value of the 7-amino group of preQ 1 was estimated to be around 8.6. One possibility is that the oxidation of preQ 1 was disfavored thermodynamically to an extent that prevented products (e.g. NADPH) to accumulate above the detection limit; another possibility is that incorrect protonation of preQ 1 , enzyme, or both caused the reaction rate to decrease below a level appreciable with the assays used.
The complete lack of activity of ecQueF toward reducing 7-formyl-preQ 0 was explicable from the evidence of binding studies that this compound did not elicit a proper substratebinding recognition in the enzyme. Exclusion already at the level of binding therefore superseded the possible use of 7-formyl-preQ 0 as a probe of the imine reduction step in the enzymatic mechanism.
In conclusion, the multistep catalytic reaction pathway of ecQueF was characterized using kinetic analysis and probing with substrate analogues. The key role of the nitrile group for substrate binding recognition was elucidated, and an energetic fingerprint for the different steps of substrate binding was obtained. Hydride reduction of the covalent thioimide intermediate is shown to be the slowest reaction step overall. NADPH binds extremely tightly to the hemithioaminal/imine intermediate compared with its binding to the thioimide intermediate. Therefore, this property of ecQueF may be relevant physiologically, as it ensures that nitrile reduction is always taken completely to the amine when escape of the imine is prevented effectively.
Synthesis of 7-Formyl-7-deazaguanine (7-Formyl-preQ 0 )-A new synthesis of 7-formyl-preQ 0 (systematic name, 2-amino-5formylpyrrolo [2,3-d]pyrimidin-4-one) that, contrary to literature-known procedures (21)(22)(23), does not require protecting group chemistry was developed. PreQ 0 (808 mg, 4.61 mmol) and sodium hypophosphite dihydrate (1.44 g, 13.6 mmol) were dissolved in the solvent mixture (15 ml of pyridine/acetic acid/ deionized water in a ratio of 2:1:1) and stirred under inert atmosphere. Raney nickel catalyst (slurry in water, ϳ500 mg) was added to the stirred reaction mixture. Safety precaution: dry Raney nickel is pyrophoric and should always be handled under inert atmosphere. The resulting reaction mixture was heated to 50°C for 5 h. Subsequently, the reaction mixture was cooled to room temperature and filtered over a pad of celite. The filtrate was reduced in vacuo. The remaining brown residue was dissolved in 6 M aqueous potassium hydroxide solution. The remaining solids were filtered off. The filtrate was cooled to 0°C and neutralized by addition of concentrated aqueous HCl. The product precipitated from the solution and was isolated by filtration and washed with copious amounts of ice water and acetone (brown solid, 770 mg, 94%). 1  Enzyme Preparation-Wild type ecQueF and its C190A variant were obtained as N-terminally His-tagged proteins using expression in E. coli BL21-DE3 (12). Cells induced with isopropyl ␤-D-1-thiogalactopyranoside (1 mM; 25°C, 20 h) were suspended in storage buffer (100 mM Tris, 50 mM KCl, 1 mM tris (2-carboxyethyl)phosphine, and 1% glycerol, pH 7.5) and sonicated. The enzymes were purified from cell extract on a His-Trap TM FF column (GE Healthcare, Buckinghamshire, UK) using a linear gradient of imidazole (10 -500 mM) in 50 mM Tris buffer (pH 7.4, 100 mM NaCl, 1 mM DDT, 3 mM EDTA). Fractions containing the enzyme were gel-filtered on a HiPrep TM 26/10 column or PD-10 desalting columns (GE Healthcare) equilibrated with storage buffer. The purified enzyme was concentrated with Amicon Ultra-15 centrifugal filters (Merck Millipore) to about 75-80 mg/ml. The protein concentration was measured with a Pierce BCA protein assay kit (Thermo Fisher Scientific, Germering, Germany). Protein stock solutions were stored at Ϫ20°C and used up within 3 weeks due to enzyme stability.
ITC-A VP-ITC micro-calorimeter from Microcal (Malvern Instruments Ltd., Malvern, UK) was used at 25°C. The enzyme was gel-filtered twice to sodium phosphate buffer (100 mM NaH 2 PO 4 ⅐Na 2 HPO 4 , pH 7.5, 50 mM KCl) using illustra NAP 5 columns (GE Healthcare). Each experiment consisted of an initial 2 l injection of ligand solution into enzyme solution, followed by 25-29 injections of 6 -10 l of this solution, leaving 300 -330 s between each injection. To avoid heat changes due to the DMSO added from the ligand solution, DMSO was also added to the enzyme solution (Յ2.5%, v/v). Data were normalized and evaluated using ORIGIN with a single binding site model as described previously (24). The enzyme molar concentration was based on the protein concentration assuming a functional ecQueF homodimer with a molecular mass of 71,772 Da (wild type enzyme) or 71,708 Da (C190A variant). Note that 3-deoxo-preQ 0 was not used for ITC measurement because of its prohibitively low water solubility.
ITC measurements were also performed in HEPES and Tris buffer, each 100 mM, at pH 7.5 and 25°C. The buffers additionally contained 50 mM KCl. Data were acquired and processed as described above. It was shown previously that a comparison of the binding enthalpies (⌬H) in HEPES and Tris is applicable to determine the proton uptake or release in conjunction with ligand binding (25,26). Equation 3 was used, where n Hϩ is the number of protons involved in the binding, and ⌬H ion is the ionization enthalpy of the buffer.
The preQ 0 -C190A complex was prepared by mixing enzyme (105 M) with preQ 0 (548 M) and incubating for 40 min at 25°C. The solution of the enzyme complex was then used for titration into the NADPH solution. Using the 5.5 M K d from Table 1, the final concentration of preQ 0 -C190A complex was calculated as 104 M.
Spectrofluorometric and Spectrophotometric Analysis of Substrate Binding in ecQueF-Quenching of the intrinsic Trp fluorescence was shown for B. subtilis QueF to serve as reporter of the formation of the non-covalent complex between enzyme and preQ 0 (11). Fluorescence titrations to study with ecQueF in the binding of preQ 0 and analogues thereof were performed using a fluorescence spectrophotometer F-4500 (Hitachi, Ltd., Tokyo, Japan). Emission spectra were recorded in the range 300 -500 nm at 1200 nm/min with the excitation wavelength at 280 nm. The quenching yield was determined as the ratio (F 0 Ϫ F s )/F 0 , where F 0 and F s are the protein fluorescence intensities in the absence and presence of substrate recorded at the same wavelength of emission.
Formation of the thioimide intermediate is traceable in B. subtilis QueF by the appearance of a new absorbance band at 370 nm (11,13). Using ecQueF, we therefore analyzed substrate binding also by absorbance measurements. A Beckman DU 800 spectrophotometer (Beckman Coulter, Inc.) was used. Before and after each addition, a wavelength scan in the range 300 -800 nm was performed. The total volume change due to multi-ple substrate additions was less than 5%, and the final DMSO concentration in the enzyme solution did not exceed 2%. The substrate concentration range used was dependent on the binding affinity (see under "Results"). In both fluorescent and spectrophotometric measurements, Tris buffer (100 mM, pH 7.5) containing 50 mM KCl and 1.15 mM tris(2-carboxyethyl)phosphine) was used. The fluorescence and spectrophotometric data were averaged in triplicate measurements. It was confirmed that the unbound ligands, in the concentrations used, did not interfere with the fluorescence measurements. Also, all ligands had negligible absorbance in the wavelength range of the experiment.
Protein Mass Analysis-The ecQueF solution (130 M, 0.6 ml) containing preQ 0 (130 M) or 2-deamino-preQ 0 (0.13, 0.52, and 2.6 mM) was incubated at room temperature in 1 h and then it was desalted using Amicon Ultra 0.5 ml Centrifugal filters (Merck Millipore). A final protein concentration of 30 pmol/l was obtained in water containing 5% acetonitrile and 0.1% trifluoroacetic acid. The samples with preQ 0 were separated on a capillary HPLC system (1200 Agilent, Santa Clara, CA) equipped with a PepSwift RP monolithic column (500 m ϫ 50 mm, Thermo Fisher Scientific) at a flow rate of 20 l/min using a gradient prepared from solvent A (0.05% trifluoroacetic acid in water) and solvent B (0.05% trifluoroacetic acid in acetonitrile), 0 -5 min 10% B, 5-55 min 10 -100% B, 55-56 min 100 -10% B, and 56 -71 min 10% B. The injection volume was 5 l. The column temperature was 60°C. Mass analysis was performed in a Thermo LTQ-FT mass spectrometer (Thermo Fisher Scientific) operated with an ESI source in positive mode with mass range of 300 -2000 m/z. The protein mass spectra were deconvoluted (Protein Deconvolution 2.0 software, Thermo Fisher Scientific), using the Xtract algorithm. Enzyme samples incubated with 2-deamino-preQ 0 were not detectable on a Thermo LTQ-FT mass spectrometer. These samples were therefore separated on a capillary HPLC system (Dionex Ultimate 3000, Thermo Fisher Scientific) with the protocol described above. The flow rate was 15 l/min. Solvent A was 0.3% trifluoroacetic acid in water, and solvent B was 0.3% trifluoroacetic acid in acetonitrile. The sample was analyzed in maXis II electron transfer dissociation mass spectrometer (Bruker, Bremen, Germany) operated with the captive spray source in positive mode with a mass range of 250 -3000 m/z. The obtained protein mass spectra were deconvoluted by data analysis software, using the MaxEnt2 algorithm.
1 H NMR Measurement of the Stereochemical Course of Hydrogen Transfer from NADPH-NADPH and (4R)-[ 2 H]-NADPH were prepared by using T. brockii alcohol dehydrogenase to reduce NADP ϩ (2.9 mM) from 2-propanol and 2-propanol-d 8 (each 80 mM), respectively (28). Tris buffer (25 mM, pD 9.0; pD ϭ pH meter reading ϩ 0.4) in D 2 O (99.8% D) was used. The reaction was stopped after 30 min when about 2 mM NADP ϩ had been reduced. Enzyme was filtered off (Amicon Ultra-15 centrifugal filter), and 2-propanol and acetone were evaporated at 40°C and about 20 mbars (Laborota 4000 efficient, Heidolph, Schwabach, Germany). Enzyme and 2-propanol were added fresh to the mixture to reduce all of the remaining NADP ϩ . The NADPH was recovered as described.
Spectra were recorded at 499.98 MHz and 30°C using a Varian INOVA 500 MHz spectrometer (Agilent Technologies). VNMRJ 2.2D software was used to process the spectra. The spectra of coenzymes, preQ 0 and preQ 1 , were compared with the literature-known spectra of these compounds (20,30,31). Considering the possibility of deuterium labeling at the C4 of NADP ϩ , reported in the literature to occur in reactions of the T. brockii alcohol dehydrogenase under the conditions used here (32,33), we checked carefully with LC-MS the (4R)-[ 2 H]NADPH preparation used. Relevant masses for doubly deuterated forms of NADPH (1H ϩ , 746; 2H ϩ , 373; and K ϩ , 784) were not detected.
Kinetic Studies at Steady State-Experiments were performed in 100 mM Tris-HCl buffer, pH 7.5, containing 50 mM KCl and 1.15 mM tris(2-carboxyethyl)phosphine. BSA (0.2 g/liter) was added to stabilize ecQueF (0.5 M). DMSO (1-2%, by volume) was used to increase substrate solubility in all reactions. The assay volume was 0.5 ml, and reactions were started by adding enzyme (10 l) to the substrate (490 l). Initial rates were determined from the NADPH consumed in the enzymatic reaction at 25°C, by absorbance at 340 nm (⑀ NADPH ϭ 6.22 mM Ϫ1 cm Ϫ1 ). It was confirmed by HPLC that the nitrile substrate was converted into the amine product (Fig. 8). Using cuvettes with a 1 cm light path in the Beckman DU 800 spectrophotometer, the lowest substrate or NADPH concentration usable in the assay was about 2 M. The averaged initial rates from triplicate determinations were used. Kinetic parameters were determined from non-linear fits of the Michaelis-Menten equation to the data. When the preQ 0 concentration was constant For the enzymatic conversion of 7-formyl-preQ 0 , Tris buffer (25 mM, pH 7.5) containing 125 mM KCl was used. The 7-formyl-preQ 0 (0.5 mM) was mixed with (NH 4 ) 2 SO 4 or NH 4 Cl (each at 50 mM) prior to attempted reduction (NADPH, 300 M or 1 mM) or oxidation (NADP ϩ , 1 mM) in the presence of ecQueF (30 M) at 30°C for 16 h.
For the oxidation of preQ 1 with ecQueF, preQ 1 was prepared by enzymatic reduction of preQ 0 and then purified from the reaction mixture by HPLC. After removing acetonitrile and ammonium acetate buffer by vacuum evaporation, the preQ 1 was dissolved in DMSO (м98% purity). Reactions were started with NADP ϩ and lasted for 90 min. Tris buffer (pH 7.5, 8.0, 8.5, or 9.0) was used.
KIEs on enzymatic preQ 0 reduction due to deuteration of the coenzyme were obtained by comparing initial rates measured with NADPH and (4R)-[ 2 H]NADPH, both synthesized as described above. The coenzyme concentration was varied (2.0 -36 M) at a constant saturating concentration of preQ 0 (10 M). KIEs were calculated by fitting Equation 4 to the data.