Nội dung

L-Lactate dehydrogenation in flavocytochrome b2 A first principles molecular dynamics study Gloria Tabacchi1, Daniela Zucchini2, Gianluca Caprini2, Aldo Gamba1, Florence Lederer3, Maria A. Vanoni2 and Ettore Fois1 1 Dipartimento di Scienze Chimiche ed Ambientali and INSTM, Università dell’Insubria, Como, Italy 2 Dipartimento di Scienze Biomolecolari e Biotecnologie, Università degli Studi di Milano, Italy 3 Laboratoire de Chimie Physique, Université Paris XI, Orsay, France Keywords catalysis; dehydrogenation reactions; first principles molecular dynamics; flavocytochrome b2; flavoenzymes Correspondence E. Fois, Dipartimento di Scienze Chimiche ed Ambientali and INSTM, Università dell’Insubria, Via Lucini 3, I-22100 Como, Italy Fax: +39 031 326230 Tel: +39 031 326218 E-mail: fois@fis.unico.it M. A. Vanoni, Dipartimento di Scienze Biomolecolari e Biotecnologie, Università degli Studi di Milano, Via Celoria 26, I-20133 Milan, Italy Fax: +39 0250 314895 Tel: +39 0250 314901 E-mail: maria.vanoni@unimi.it (Received 24 November 2008, revised 27 January 2009, accepted 16 February 2009) doi:10.1111/j.1742-4658.2009.06969.x First principles molecular dynamics studies on active-site models of flavocytochrome b2 (l-lactate : cytochrome c oxidoreductase, Fcb2), in complex with the substrate, were carried out for the first time to contribute towards establishing the mechanism of the enzyme-catalyzed l-lactate oxidation reaction, a still-debated issue. In the calculated enzyme–substrate model complex, the l-lactate a-OH hydrogen is hydrogen bonded to the activesite base H373 Ne, whereas the Ha is directed towards flavin N5, suggesting that the reaction is initiated by a-OH proton abstraction. Starting from this structure, simulation of l-lactate oxidation led to formation of the reduced enzyme–pyruvate complex by transfer of a hydride from lactate to flavin mononucleotide, without intermediates, but with a-OH proton abstraction preceding Ha transfer and a calculated free energy barrier (12.1 kcalÆmol)1) consistent with that determined experimentally (13.5 kcalÆmol)1). Simulation results also revealed features that are of relevance to the understanding of catalysis in Fcb2 homologs and in a number of flavoenzymes. Namely, they highlighted the role of: (a) the flavin mononucleotide–ribityl chain 2¢OH group in maintaining the conserved K349 in a geometry favoring flavin reduction; (b) an active site water molecule belonging to a S371–Wat–D282–H373 hydrogen-bonded chain, conserved in the structures of Fcb2 family members, which modulates the reactivity of the key catalytic histidine; and (c) the flavin C4a–C10a locus in facilitating proton transfer from the substrate to the active-site base, favoring the initial step of the lactate dehydrogenation reaction. Flavoenzymes are a class of oxidoreductases, widespread in nature, which catalyze fundamental oxidoreduction reactions of cell metabolism [1]. The flavin prosthetic group can exist in the oxidized state and in the one- (semiquinone) or two-electron (hydroquinone) reduced forms, with the oxidoreduction potential modulated by the protein environment [2,3]. The versatility of the flavin is responsible for the great number, variety and biological relevance of flavoenzyme-catalyzed reactions and for the mechanistic debates surrounding them. Flavocytochrome b2 (l-lactate cytochrome c oxidoreductase, EC 1.1.2.3, Fcb2) is a homotetrameric yeast enzyme that catalyzes the oxidation of l-lactate to pyruvate with a subsequent reduction of cytochrome c. During the catalytic cycle, lactate binds to the flavodehydrogenase domain and is oxidized to pyruvate with transfer of the reducing equivalents to the bound Abbreviations BP86, Becke-Perdew; DFT, density functional theory; Fcb2, flavocytochrome b2; FMN, flavin mononucleotide; FPMD, first principles molecular dynamics; GO, geometry optimization; HT, hydride transfer; LMO, lactate monoxygenase; PA, proton abstraction; PBE, Perdew–Burke–Ernzerhof; PW, planewaves; QM, quantum mechanical. 2368 FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS Simulations of the flavocytochrome b2 reaction G. Tabacchi et al. flavin mononucleotide (FMN), yielding its hydroquinone form (Eqn 1). Llactate + E-FMNox !Pyruvate + E-FMNred ð1Þ The reduced flavin then transfers electrons, one at a time, to heme b2 in the N-terminal protein domain, with the intermediacy of the flavin semiquinone species. Cytochrome c is the physiological acceptor for heme b2 [4]. Several studies on Fcb2 have focused on the mechanism of lactate dehydrogenation, catalyzed by its flavodehydrogenase domain, which is the structural and mechanistic prototype of a family of l-a-hydroxy acidoxidizing enzymes [4–8]. This class of enzymes includes, as well-characterized members, glycolate oxidase [9], l-lactate monoxygenase (LMO) [10,11], l-lactate oxidase [10], mandelate dehydrogenase [12] and long-chain a-hydroxy acid oxidase [13]. Crystal structures are known for all these enzymes, except LMO. The enzymes show a conserved (a ⁄ b)8-barrel fold with a conserved constellation of residues in the active center. They are believed to share a common mechanism for oxidation of the a-hydroxy acid substrate in the enzyme reductive half reaction, which formally implies the loss of two electrons and two protons during conversion to the corresponding a-keto acid. Two mechanistic hypotheses for the a-hydroxy acid dehydrogenation reaction have been formulated and tested over the past decades, but the mechanism is still debated. The ‘proton abstraction’ (PA) or ‘carbanion’ mechanism predicts that the bound a-hydroxy acid undergoes abstraction of the a-hydrogen (Ha), as a proton, by an active-site base with formation of an a-carbanion intermediate. The latter would evolve, probably via two single-electron transfer steps, to yield the flavin hydroquinone and the a-keto acid product (Scheme 1A). The alternative ‘hydride transfer’ (HT) mechanism predicts that the substrate Ha is directly transferred to the flavin N5 position as a hydride with elimination of the a-hydroxyl proton (Scheme 1B). Opposing views are discussed in Fitzpatrick [5,6] and Lederer et al. [7,8], respectively. We refer to these reviews and others for reference to the original literature whenever possible, although a few key points are summarized here. Among the arguments in support of the PA mechanism is the reported pKa value of 9 in reduced Fcb2 for the group that abstracts the substrate Ha [14]. This value has been considered too low to be that of the reduced FMN N5 atom, which has been estimated to be  20 for free flavin [15]. In support of the hypothesis that this group is H373, a decrease of one pH unit Scheme 1. (A) The carbanion or proton abstraction (PA) mechanism. (B) The hydride transfer (HT) mechanism. The carbanion could evolve via either formation of a covalent intermediate at N5 or two consecutive single electron transfer steps. The active-site base (B) is only shown in the first part of each scheme. of the pKa of this group was induced by the mutation to asparagine of D282, which interacts with H373 [16]. Furthermore, studies of the transhydrogenation reactions between l-[2-2H]-lactate (or l-[2-3H]-lactate) and bromopyruvate led to isotope effects, which could be rationalized only within the frame of a PA mechanism [8,17]. Moreover, in the reaction between LMO and its slow substrate glycolate, two covalent intermediates were observed, one of which had the correct stereochemistry and was catalytically competent [18]. Finally, a spectroscopically detectable intermediate has been observed during the reaction of mandelate dehydrogenase with mandelate [19]. This intermediate has been proposed to be a charge-transfer complex between the oxidized flavin and the mandelate a-carbanion, in view of the sensitivity of its rate of formation on mandelate a-deuteration. Other results suggest that the HT mechanism may be operative. For example, both Fcb2 [4] and LMO [20] can oxidize lactate when FMN is substituted by 5-deaza-5-carba-FMN, which is known to be an obligatory hydride acceptor. Furthermore, a combination of substrate and solvent kinetic deuterium isotope effects from the Fitzpatrick group, using wild-type and mutated enzymes, were interpreted as indicating a HT mechanism with formation of a discrete alkoxide intermediate stabilized by a hydrogen bond from active site Y254 [6]. A number of other results on Fcb2 and related enzymes [8] could be interpreted within the framework FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS 2369 Simulations of the flavocytochrome b2 reaction G. Tabacchi et al. of both the proposed mechanisms. Site-directed mutagenesis of active-site residues in various family members yielded important information about their role in substrate binding and ⁄ or transition state stabilization (Table 1), but these studies did not provide decisive evidence to distinguish between the PA and HT mechanisms. A key piece of missing information is the structure of a complex between the enzyme and a true substrate analog unambiguously defining the structure of the Michaelis complex of Fcb2 and of the enzymes of the family. For these reasons, we wished to address the issue of the mechanism of lactate dehydrogenation catalyzed by Fcb2 by using, for the first time, a computational approach. Thus, we built pure quantum mechanical models of Fcb2 by adopting the ‘active site only’ strategy [21], extended to contain residues up to the third shell, namely residues separated by two other residues or water molecules from the reaction center (i.e. the flavin N5–lactate Ca region). Interaction of the active- site model with l-lactate was investigated by structural optimization (0 K) and finite temperature (300 K) first principles molecular dynamics (FPMD) simulations [22]. Starting from the computed structure of the stable enzyme–lactate complex model, the lactate dehydrogenation reaction was simulated by means of a statistical sampling approach [23] coupled to FPMD. In addition to addressing the mechanistic issue, the calculations presented here provide a set of observations of general interest for understanding how flavoenzyme catalysis takes place. Results and Discussion Choosing the minimal Fcb2 active-site model The first aim of this study was to determine the stable structure of the Fcb2 active-site model in complex with l-lactate. Details of the model-building process and the computational approach are given in Experimental Table 1. Summary of residues included in the flavocytochrome b2 (Fcb2) active-site models and their role as inferred from studies on the wild-type and engineered enzymes. Model Role ⁄ refs Residue FMN FMN Y143 T197-A198 S228 L230 Q252 Y254 T280 D282 L286 R289 D292 K349 H373 R376 Wat609 Wat638 Wat687 Wat763 b (Wat9) (Wat220)b,c (Wat86)b (Wat160)b Side chain OH H-bonded to Pyr a-carboxylate [24,32–34] Ala main chain amide H-bonded to FMN N(5) position [24,32] H-bonded to FMN O4 [24,32] Contributes to substrate specificity [48–50] Side chain amide H bonded to FMN N(3) [24,32] H bond to Pyr O(2) carbonyl; transition state stabilization [8,24,32,48] H bond to FMN O4 [24,32] H bond to H373 in oxidized Fcb2. Ion pair with H373 in reduced Fcb2 [14,16,24,32,37] May contribute to substrate specificity [24,32,48] Two possible orientations a [32,39,40] Ion pair with R289 [24,32] Stabilization of FMNH) [24,32] H bonded to pyruvate carbonyl [24,32]. Proposed active-site base [4,14,51] Ion pair with Pyr carboxylate [4,24,32] Bridge between S371 OH and D282 carboxylate [24,32] Bridge between R289NH2 atom and pyruvate O1 [24] Active site water molecule in subunit A [24] Active site water molecule in subunit A [24] Fragment 0 1 2 3 10-Methyl isoalloxazine 10-Hydroxyethyl isoalloxazine CH2=C(H)OH H-CO-NH-CH2-CH3 CH3-OH CH4 NH2-C(O)-CH3 CH2=C(H)OH X X X X X X X X X X X X X X X X X X X X X X X X X X X X CH3-OH CH3-COO) X X X X X X X X CH4 [CH3-NH-C(NH2)2]+ H-COO) CH3-NH3+ 1-Methyl-imidazole X X X X X X X X X X X X X X X X X X X X [CH3-NH-C(NH2)2]+ H2 O H2 O H2O H2O X X X X X X X X X X X X X a R289 is found in the ‘distal’ (in 1fcb, 1kbi) or ‘proximal’ (1kbj) position, or with partial occupancy of both positions (e.g. in the structure of the recombinant sulfite adduct 1ltd [39] and in the 1kbj subunit A). The R289K substitution led to observing K289 in the ‘proximal’ position in structures of the mutated enzyme–sulfite complex [40]. b The numbering of water molecules in PDB files has been recently changed. The original numbering is maintained here with the new numbering indicated in parentheses. c Replaced by Wat352 from 1kbj, subunit B, in tests with structural models with R289 in the proximal position. 2370 FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS Simulations of the flavocytochrome b2 reaction G. Tabacchi et al. A Fig. 1. The minimum energy structure of Model 2 in complex with L-lactate. The labels indicate the residues of Fcb2 that have been modeled. Hydrogen bonds are shown as dotted lines (assuming 2.2 Å as maximum hydrogen-bond length). Atoms kept fixed in the FPMD simulations are shown as spheres. Atoms color code: carbon, gray; oxygen, red; nitrogen, blue; hydrogen, white. Procedures and Doc. S1. The crystal structure of the Fcb2 flavin domain, where the reaction product pyruvate is observed in front of the flavin semiquinone si side (subunit B from the Protein Data Bank file 1fcb) [24], was used as the starting point for model building. A series of simulations with models of increasing complexity led to the building of a first stable model (Model 0), which was subsequently extended on the basis of the results of geometry optimizations (GO), room temperature molecular dynamics simulations (FPMD) and test runs of the reactivity of the enzyme– substrate complexes obtained. For clarity, only Models 0–3 are described, with Model 2 (Fig. 1) representing the model used to simulate the lactate oxidation reaction. Two sets of GO, differing in the starting orientation of the substrate, were performed on each model. In the first set, lactate was initially positioned in an orientation poised for PA, with Ha oriented towards H373 Ne (Fig. 2A). In the other set, lactate was oriented with Ha pointing towards FMN N5, as required by the HT mechanism (Fig. 2B). These two orientations differ by a rotation of  40 around the C1–C2 bond [4]. Further tests were carried out by starting with the substrate in an intermediate orientation. GO of the initial Model 0 (148 atoms; Table 1), which included the isoalloxazine ring carrying a methyl group in place of the ribityl side chain, did not adequately describe the Fcb2 active site, because the positive K349 side chain reoriented to a position inconsistent with the structural data. This residue, which is believed to favor flavin reduction, is invari- B Fig. 2. (A) PA-prone orientation of L-lactate. The arrow indicates the postulated Ha to Ne transfer. (B) HT-prone orientation of L-lactate. The reaction coordinate Q = r(Ca–Ha) ) r(N5–Ha) chosen for the bluemoon ensemble simulation of the reductive half reaction in the Fcb2 active site Model 2 is also shown, with the Ca, Ha and N5 atoms represented as spheres. ant in the Fcb2 family, or substituted by an Arg or the positive end of an a-helix dipole in many flavoenzymes [2]. However, extension of the FMN model by including the C2¢ hydroxyl group in Model 1 (163 atoms) stabilized K349 in its crystallographic position (Fig. 1), revealing a role for the 2¢-OH position of the flavin ribityl side chain in maintaining the correct geometry of a key active site residue. Interestingly, only a few studies have so far explored the role of the flavin side chain in flavoenzymes [25–29] and our results support its importance also in Fcb2. Regardless of the initial lactate orientation, GO of Model 1 led to structures poised for HT, in which all fragments mimicking the various side chains essentially FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS 2371 Simulations of the flavocytochrome b2 reaction G. Tabacchi et al. maintained the starting (crystallographic) positions. R289 is the only Fcb2 residue that has been found in two alternative orientations in different Fcb2 structures, ‘distal’ and ‘proximal’, as discussed in detail in the section ‘Model building’ and in Doc. S1. The substrate conformation in the Fcb2 active-site model was not affected by changing the R289 orientation. Because the optimized structure with R289 in the proximal position was 11.2 kcalÆmol)1 less stable than that with the residue in the distal position, only models with distal R289 are discussed below. In an FPMD simulation at room temperature starting from the optimized structure of Model 1, the residues and lactate molecule oscillated around the optimized positions, in particular, lactate kept the HT orientation. However, we observed a low energy barrier proton transfer between H373 Nd and D282 O1 [30], which led to the development of a transient imidazolate anion and neutral Asp. Because no evidence of the formation of a histidinate side chain in Fcb2 is available, Model 1 was modified. Inspection of the available Fcb2 crystal structures suggested that no active-site side chain missing from Model 1 is likely to influence the acid–base properties of H373. For this reason, we shifted our attention to crystallographically detected water molecules. In the 1fcb structure, Wat609 (see Table 1 for water molecules numbering) forms a hydrogen-bond bridge between S371 and D282 (Fig. 3). Interestingly, S371 is invariant in Fcb2 family members, being part of the SNHGXRQ signature of this enzyme family. Strikingly, the Ser ⁄ Wat ⁄ Asp ⁄ His geometry is also conserved in the crystal structures of the other Fcb2 family members. The hypothesis that Wat609 may play a role in modulating Fig. 3. The S371–Wat609–D282–H373–pyruvate–FMN arrangement in 1fcb. Carbon atoms are represented by gray spheres, nitrogen by blue spheres and oxygen by red spheres. The dotted lines indicate hydrogen bonds with distances in Å. 2372 the acid–base properties of H373 via D282 was tested by including it in Model 2. Wat687 was also added. This water molecule is detected in 1fcb subunit A, where pyruvate is absent, at a position that could influence lactate orientation. The Michaelis complex GO of Model 2 (169 atoms), with lactate initially oriented for HT, led to a minimum energy structure in which lactate is hydrogen bonded to the Ne atom of the active-site base H373 via its a-hydroxyl proton, whereas its Ha points toward FMN and is 2.73 Å away from N5 (Fig. 1 and Table 2). Analysis of the electronic structure of the optimized system indicated that, as expected, the flavin is in the oxidized state. The lactate carboxylate group forms strong hydrogen bonds with R376 and Y143, and a weaker bond with Table 2. Relevant geometrical parameters for the optimized enzyme–lactate (hydride transfer- and proton abstraction-prone) and enzyme–pyruvate complexes obtained using Model 2. Parameters from crystallographic structures (PDB files: 1fcb and 1kbi) are reported for comparison. In line with the resolution of the X-ray structure, calculated distances are also given to two decimal places. Distances are in Å, angles in degrees. Atoms labeling for L-lactate and active-site residues are as in Table 1 and Fig. 1 (see also PDB files 1fcb and 1kbi). HT, hydride transfer; PA, proton abstraction. O–H Ca–Ha Ca–O O2–C1–Ca–C3 NeH373–HaOH NeH373-OaOH NeH373–Ha NeH373–Ca NdH373–O1D282 H(Nd) H373–O1D282 H(Nd)H373–NdH373 OWat609–O2D282 HY254–OaOH OY254–OaOH Ca–N5 FMN Ha–N5 FMN NR376–O1 NR376–O2 OY143–O2 CR376–CR289 Ca–CR289 C3–CBA198 O2¢OH–N1 FMN NK349–N1 FMN NK349–OC(2)FMN HT PA Pyruvate 1.03 1.11 1.42 )62.1 1.69 2.69 3.87 3.67 2.64 1.56 1.09 2.52 2.28 3.06 3.72 2.73 2.75 2.70 2.74 4.10 6.49 4.74 2.89 2.84 2.68 0.98 1.11 1.44 )3.8 4.87 3.90 2.44 3.51 2.65 1.58 1.08 2.52 3.37 3.94 3.91 3.42 2.65 2.77 2.78 4.13 6.44 3.91 2.88 2.83 2.68 2.95 1.24 )47.6 1.04 2.79 4.41 3.73 2.49 1.18 1.33 2.51 2.29 3.09 3.73 1.02 2.83 2.78 2.79 4.11 6.52 4.64 2.84 2.81 2.66 1fcb 1kbi 1.39 0 1.22 )47.6 2.60 2.90 3.58 2.32 3.72 2.57 2.46 2.65 2.39 3.72 3.21 3.78 3.04 3.68 2.81 4.28 6.90 4.27 2.90 2.80 2.80 3.09 3.02 2.60 4.27 6.47 4.67 2.97 2.77 2.60 FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS Simulations of the flavocytochrome b2 reaction G. Tabacchi et al. Wat638. The a-hydroxyl oxygen is hydrogen bonded to Wat687 and, weakly, to the Y254 hydroxyl group. H373 is hydrogen bonded through its Nd proton to the D282 carboxylate group (which also interacts with Wat609), whereas R289 is hydrogen bonded to D292 and Wat638. The other residues essentially maintain their initial position. The l-lactate HT prone orientation was also kept in a FPMD simulation at room temperature. Remarkably, the inclusion of Wat609 had the hypothesized effect on the protonation state of H373, which remained in the neutral form during the 300 K FPMD simulation. With Model 2, a relative energy minimum was found for the PA-oriented lactate. In this structure, the lactate Ha is 2.44 Å from H373 Ne, and its a-hydroxyl oxygen is hydrogen bonded to Wat687 (Table 2). The lactate carboxylate group is hydrogen bonded to Y143 and R376, as in the HT-prone complex, but no interaction with Y254 is present. However, this structure is 13.4 kcalÆmol)1 higher in energy than the one with lactate in the HT orientation. Moreover, it rapidly converted into the HT-prone complex upon heating to room temperature in an FPMD simulation. Model 2 was further modified (Model 3, 171 atoms) to include Wat763 (Table 1 and Doc. S1 for details). The optimized HT- and PA-prone structures were very similar to the corresponding ones obtained with Model 2: the presence of the Wat638–Wat763–Wat687 chain did not change their stability order. Rather it increased the energy difference in favor of the HT-prone orientation (DE = )16.5 kcalÆmol)1). Also in Model 3, only the complex with lactate oriented as per HT was stable at room temperature. Interestingly, in the structure of the Michaelis complex model, the relative orientation of flavin and protein atoms involved in catalysis resembles that highlighted by Fraaije & Mattevi [2] on the basis of the structural comparison of a series of flavin-dependent enzymes, as well as that found for d-alanine bound in a d-amino acid oxidase active-site model [31]. N5, carries a net negative charge on N1 and is in the two-electron reduced hydroquinone state. Therefore, this structure represents the oxidized substrate–reduced enzyme complex model. The calculated interatomic distances can be compared with those extracted from the available crystal structures (Table 2). The pyruvate keto oxygen is hydrogen bonded to H373 Ne and, more weakly, to Y254, whereas its carboxylate group interacts with R376, Y143 and Wat638. However, the hydrogen-bond distances are longer than in the lactate complex, indicating that the product is less tightly bound than the substrate in the enzyme active-site model. The K349 NH3+ group is near the FMN N1 atom and stabilizes its negative charge. In our model, pyruvate Ca deviates from planarity in excellent agreement with its structure in 1 kbi [32], in which no planarity constraints were introduced during refinement, as opposed to that imposed in 1fcb [24]. An FPMD simulation showed that the structure of the model complex is maintained at 300 K, confirming the stability of pyruvate bound to the two-electron reduced FMN in Model 2. The reaction path The starting point of the simulation of the reaction path [23] was the room-temperature-equilibrated structure of the Michaelis complex model containing the oxidized FMN and lactate oriented as per HT (Fig. 1). The reaction coordinate Q = rCa-Ha ) rN5-Ha (Fig. 2B) was chosen in order to simulate the transfer of the lactate Ha proton to FMN N5 position without making any assumption about the timing of Ha transfer and the nature of the species being transferred (i.e. a proton preceded or followed by the electron pair, or a hydride anion). The reaction products The second goal of this study was the determination of the minimum energy structure of the reaction products, the complex with reduced flavin and pyruvate. To this aim, a guess configuration was built by replacing lactate with pyruvate, and protonating FMN at N5 and H373 at Ne. The resulting optimized structure (in Model 2) is more stable than the reactant state by 12.6 kcalÆmol)1. Analysis of the ground state electronic structure indicates that here the flavin, protonated at Fig. 4. Simulated free energy profile for the lactate dehydrogenation reaction with Model 2. (solid line) Calculated mean force of constraint as a function of the reaction coordinate (dashed line). Black circles indicate the mean force values obtained from the 11 constrained simulations varying Q from Qr ()1.62 Å) to Qp (1.93 Å). FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS 2373 Simulations of the flavocytochrome b2 reaction G. Tabacchi et al. The free energy variation along the path described by Qr fi Qp, where Qr is the Q value corresponding to the room-temperature-equilibrated Michaelis complex and Qp is the Q value relative to the products side, is reported in Fig. 4. The simulated reaction proceeded without intermediates with an activation free energy of 12.1 kcalÆmol)1, a value that compares well with that calculated from the experimentally determined rates of flavin reduction in Fcb2 (13.5–13.6 kcalÆmol)1) [33,34]. The overall reactants-to-products free energy change amounts to )8.1 kcalÆmol)1. As shown in Figs 5 and 6, at the end of the path the unconstrained lactate a-hydroxyl proton is bound to the H373 Ne atom, the ligand Ca–O bond length is that typical of a keto group and Ca is in a sp2 hybridization state. In the flavin ring, the simulated path causes formation of the N5–Ha bond, shortening of the C4a–C10a distance from a single to a double bond, and lengthening of N5–C4a from a double to a single bond. Electronic structure analysis showed that at Qp the flavin is in its two-electron reduced hydroquinone form, protonated at N5, with a negative charge on N1 (stabilized by K349). Therefore, sampling of the Qr fi Qp path has brought about substrate dehydrogenation and flavin reduction. The room temperature structures of the oxidized enzyme–lactate and reduced A C 2374 enzyme–pyruvate model complexes obtained from the Qr and Qp simulations, corresponding to a zero value of the constraint force, compare well with the minimum energy structures of the enzyme–substrate and enzyme–product model complexes described above, respectively. H373 Ne becomes protonated at Q  )0.5 Å (Fig. 6B), indicating that the lactate a-OH proton is abstracted by the active-site base before the Ca–Ha bond is broken. The room temperature transition state, identified by the maximum in the free energy profile, is found at Q = )0.15 Å (Fig. 4), indicating a slightly early transition state. Finally, formation of the pyruvate carbonyl group parallels the transfer of Ha as a hydride to the flavin (see below). Analysis of the electronic structure of the model complex along the calculated free energy profile can be performed by inspecting the maximally localized (Wannier) orbitals [35]. Their centroid positions (W) and spread (d) represent the center of mass of the orbital and the extent of localization of the orbital, respectively. In particular, the distances of WHa, initially localized on the Ca–Ha bond, from Ha, Ca and FMN N5 (Figs 5A and 6A) indicate the transfer of an electron pair from lactate to the isoalloxazine ring, leading to flavin reduction. At Q = )0.15 Å, WHa is shared B D Fig. 5. L-Lactate ⁄ FMN ⁄ H373 arrangement along the simulated lactate dehydrogenation. Centroids of relevant Wannier orbitals localized on the substrate and the flavin ring, representing electron pairs, are shown as green spheres, with the exception of WHa which is represented by a black sphere. Carbon atoms are represented in gray, nitrogen in blue, oxygen in red and hydrogen in white. (A) Electronic structure of the starting point of the reaction corresponding to the HT-prone Michaelis complex, with FMN in the oxidized state. (B) Electronic structure at Q = )0.33 Å (before the transition state) where the lactate a-OH is deprotonated, the Ca–Ha bond is not broken yet and FMN is still in the oxidized state. (C) Electronic structure at the transition state (Q = )0.15 Å), where incipient formation of pyruvate and flavin reduction are observed. The WHa electron pair is localized on the lactate Ha (a white sphere), indicating its transfer to N5 as a hydride moiety. (D) Electronic structure of the final point of the reaction (Q = 1.93 Å) corresponding to pyruvate in complex with the two-electron reduced FMN. FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS Simulations of the flavocytochrome b2 reaction G. Tabacchi et al. A B C -WH N5 WH N5-W Fig. 6. Variation of relevant distances (in Å) and angles in () along the simulated reaction path as a function of the reaction coordinate Q. (A) Distances of the Wannier center WHa from the lactate Ca (Ca–WHa, solid red line) and Ha (Ha–WHa, solid black line) and from the flavin N5 (N5–WHa, dashed red line) nuclei, and Wannier orbital spread dWHa (Å2, dashed black line). (B) Distances of the lactate hydroxyl proton from the middle point of the C4a–C10a bond of the flavin (dashed black line), the lactate a-OH oxygen (solid red line), H373 Ne (dashed red line); distance of lactate a-OH oxygen from Y254 hydroxyl proton (solid black line); distances of Ha from Ca (solid blue line) and FMN N5 (dashed blue line). (C) Ca–O distance in the substrate (dashed black line); C4a–C10a (solid red line) and N5–C4a (solid black line) distances in FMN. (D) C3–Ca–O angle of lactate. H(OH)-FMN FMN C -H N5-H WH O(OH)-HY254 H -WH C O-H C4a-C10a H(OH)-N D C4a-N5 between lactate and flavin, shows its minimum distance from Ha and its maximum spread. Therefore, lactate Ha is transferred to the flavin as an anion. The early lactate a-OH deprotonation by H373 (at Q  )0.5 Å) appears to facilitate hydride transfer to the flavin. Indeed, just before the transition state one of the lone pairs on the deprotonated hydroxyl oxygen is polarized towards the still unbroken Ca–Ha bond (Fig. 7), thus suggesting charge-transfer to the r* (Ca–Ha) antibonding orbital owing to negative hyperconjugation effects, as already pointed out on the basis of higher level of theory calculations on isolated l-lactate [36]. Within the context of a HT mechanism, our calculations predict the formation of a transient alkoxide rather than a discrete intermediate [6]. Lactate alkoxide is formed prior to hydride transfer; however, because of the presence of a single maximum in the free energy profile, such a species cannot be properly defined as a reaction intermediate. During the simulated path, interactions between the lactate a-carboxylate and R376 and Y143 are maintained. The lactate a-OH is engaged in hydrogen bonding with Wat687, Y254 and, initially, H373 (through the H atom). Upon proton transfer to H373-Ne, the hydrogen bond with Y254 is strengthened, and this contact is maintained after conversion to pyruvate (Fig. 6B). The observation of the strengthened interac- C -O tion with Y254 following deprotonation is consistent with a role for Y254 in transition state stabilization, assigned on the basis of the characterization of the Y254F Fcb2 mutant [4,6]. Interestingly, the a-OH proton transfer process from lactate to H373 Ne occurs along a path parallel to the C4a–C10a flavin bond, at 3.1 Å from the isoalloxazine plane (Fig. 6B). This suggests that the isoalloxazine ring may also assist proton transfer to the active-site base H373, which in turn, when protonated, is believed to favor flavin reduction by stabilizing the negatively charged reduced FMN [8]. Recent computational studies on another, structurally unrelated, flavoenzyme (acyl-CoA dehydrogenase) alluded to a similar effect of the flavin in favoring the initial reaction step [38], indicating the path at  3 Å from the flavin ring and over the C4a–C10a positions as the optimal path for substrate–enzyme proton transfer. Thus, it appears that the flavin isoalloxazine ring might contribute to catalysis of flavoenzymes by providing an ‘electronpaved path’ for proton transfer to the active-site base (Fig. 8). Conclusions Using a first principles approach, we built a model of the Fcb2 active site capable of describing bound FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS 2375 Simulations of the flavocytochrome b2 reaction G. Tabacchi et al. A B Fig. 8. Representation of the contour plots of the electronic density of the Wannier orbitals along the C4a–C10a bond of the flavin, forming the ‘electron-paved path’ for a-hydroxyl proton transfer from lactate to H373 shown at Q = )0.7 Å. Color coding: C, gray; O, red; N, blue; H, white. The a-OH proton is represented as a white sphere. Fig. 7. Representation of electronic structure details of the substrate and the flavin ring at Q = )0.33 Å (A) and Q = 0 (B). (A) Contour plots of the three Wannier orbitals localized on the lactate hydroxyl oxygen, on the Ca–Ha bond and on the flavin N5–C4a bond after proton transfer to H373, see also Fig. 5B. (B) Contour plots of the three Wannier orbitals localized along the Ca–O bond, on Ha and on C4a immediately after the transition state. Green and yellow contours represent positive and negative phases of the orbitals, respectively. Carbon atoms are represented in gray, nitrogen in blue, oxygen in red, hydrogen in white. lactate and FMN in the oxidized state and the reaction product pyruvate with reduced flavin. Only complexes between the active-site models and lactate oriented so as to favor the HT mechanism (Scheme 1B) could be obtained at room temperature. 2376 Starting from this Michaelis complex model, the simulated lactate dehydrogenation reaction occurred via an asynchronous HT mechanism, where proton transfer from the lactate a-hydroxyl group to the active-site base (H373) precedes and assists Ha transfer as a hydride to the flavin N5, without formation of a reaction intermediate. The calculated activation free energy value (12.1 kcalÆmol)1) is comparable with those estimated from rapid kinetics experiments (13.5– 13.6 kcalÆmol)1). During this investigation, interesting observations relevant to the general understanding of the catalytic power of flavoenzymes were also made, regarding, in particular, the role of FMN ribityl side chain and the electron rich isoalloxazine ring C4a–C10a region in assisting ⁄ favoring catalysis. Finally, the simulation suggested an unexpected role for the invariant S371 side chain hydroxyl group in positioning a conserved water molecule (Wat609 in 1fcb), which contributes to maintain the D282–H373 pair in the correct protonation state. A study of the effect of S371 substitutions on the Fcb2 properties is currently in progress. In conclusion, besides providing a theoretical contribution to the mechanistic debate on l-lactate dehydrogenation in Fcb2, this study sets the basis for further experimental and theoretical investigations on this and other flavoenzymes. FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS G. Tabacchi et al. Experimental procedures Model building The atomic coordinates of the Fcb2 chain B flavin domain (residues 100–511) were used as the starting point for the building of the active-site model. The detailed description of the rational of the model building process and of the selection of residues is available for inspection in Doc. S1. Briefly, residues to be modeled were selected on the basis of their position in the active center and their proposed role in the reaction. The side chain groups of Y143, S228, L230, Q252, Y254, T280, D282, L286, R289, D292, K349, H373, R376, A198 and the peptide bond connecting it to T197 were included in the model (Table 1 and Figs 1 and S1). Particular care was taken to model R289, which is the only residue close to the active site, the side chain of which has been found in different ‘distal’ (1fcb and 1 kbi for the wild-type Fcb2) [24,32] or ‘proximal’ (i.e. subunit B of the recombinant Fcb2 flavodehydrogenase domain, 1kbj) [32] orientations or discretely disordered between the orientations (subunit A of the Fcb2–sulfite complex structure 1ltd) [39]. From a mechanistic point of view, the distal ⁄ proximal reorientation of R289 was proposed to favor catalysis by providing a net positive charge stabilizing the carbanion intermediate predicted for the PA mechanism (Scheme 1A). Both proximal and distal R289 arrangements have been investigated. Wat638 was included in models with R289 in the distal conformation. With R289 in the proximal position, the central water in the D292–Wat–R289 triad was taken from the 1kbj structure (Wat352). The crystallographic Wat687 and Wat763, detected in 1fcb subunit A, form a hydrogen-bonded chain of water molecules Wat638–Wat763–Wat687, which may interact with the substrate through Wat687. Therefore, Wat687 was included in Model 2 and further calculations were done adding also Wat763 (Model 3). Finally, the crystallographic Wat609 was included in Models 2 and 3. Two molecular FMN models were used. In the first case, the structure was cut at the C1¢ position of the ribityl side chain (Model 0), and the C2¢ hydroxyl group was included in Models 1, 2 and 3. In each model, lactate was initially positioned by superposing its carboxylate group and Ca carbon atom onto the corresponding positions of the bound pyruvate in 1fcb subunit B. Most of the calculations were carried out on Model 2, consisting of 169 atoms (Fig. 1), as detailed in the results section. Computational methods The simulations presented in this study were performed with the planewaves (PW) density functional theory (DFT) Simulations of the flavocytochrome b2 reaction code CPMD (http://www.cpmd.org) (see Doc. S1 for further details). Only valence electrons were treated explicitly and interaction with the ionic cores was modeled by norm conserving pseudopotentials (d-nonlocality for C, O and N, and a local pseudopotential for H) [41,42]. DFT approximation and basis set size were the same adopted in the successful simulation of the reductive half-reaction in d-amino acid oxidase [31], namely the Becke exchange and Perdew correlation (BP86) functionals [43,44] with a 60 Ry cut-off for the PW expansion (BP86 ⁄ PW ⁄ 60). Test calculations were performed with a different gradient corrected DFT approximation and basis set size, namely the Perdew– Burke–Ernzerhof (PBE) functional [45] with a PW cut-off of 70 Ry (PBE ⁄ PW ⁄ 70) (Table S1). Moreover, FPMD simulations of lactate in complex with a Fcb2 model were carried out using the BP86 ⁄ PW ⁄ 60 and PBE ⁄ PW ⁄ 70 schemes [36]. Furthermore, the suitability of such an electronic structure treatment has been validated by comparing the isolated lactate optimized structure with that calculated at higher level of theory [36]. A simulation cell of 24 · 18 · 18 Å3 with periodic boundary conditions was adopted. In order to make the interaction among images negligible, the simulation cell was chosen to be 5 Å longer than the maximum interatomic distance along each axis. Care was taken to maintain electrical neutrality, i.e. the sum of the electronic charges was balanced by the sum of the nuclear charges. GO were carried out by either quasi-Newton or simulated annealing procedures [46], using a convergence criterion of 5 · 10)4 a.u. for the forces on atoms. The thermal stability of the optimized models was checked by FPMD simulations starting from the calculated energy minima and slowly heating the systems to 300 K. The room temperature behavior of the model systems was investigated by analyzing trajectories collected by integrating the Car Parrinello equations of motion with a timestep of 0.121 fs and a fictitious inertia factor of 500 a.u. for the wavefunctions coefficients [22]. Temperature was controlled by means of Nose–Hoover chain thermostats [46]. Local flexibility was taken into account by allowing for reorientation of modeled side chain residues through rotations around several bonds. The atoms to be fixed, shown in Fig. 1, were selected mostly in the peripheral regions. Only one residue fragment (D292) was completely fixed. The CH2=CH nuclear positions of the two (CH2=CH)OH mimicking the functional group of Y143 and Y254 were fixed. The remaining 12 residues were kept in place by fixing 16 nuclear positions. In the FMN group, four carbon atoms of the isoalloxazine ring (C6–C9), two of the methyl groups (C7a, C8a) and three atoms on the ribityl side chain were fixed. Also some of the water molecules were held in place by fixing the oxygen position, with the exception of Wat609 where also the H atoms FEBS Journal 276 (2009) 2368–2380 ª 2009 The Authors Journal compilation ª 2009 FEBS 2377