On how the binding cavity of AsqJ dioxygenase controls the desaturation reaction regioselectivity: a QM/MM study

On how the binding cavity of AsqJ dioxygenase controls the desaturation reaction... The Fe(II)/2-oxoglutarate-dependent dioxygenase AsqJ from Aspergillus nidulans catalyses two pivotal steps in the synthesis of quinolone antibiotic 4′-methoxyviridicatin, i.e., desaturation and epoxidation of a benzodiazepinedione. The previous experimental results signal that, during the desaturation reaction, hydrogen atom transfer (HAT) from the benzylic carbon atom (C10) is a more likely step to initiate the reaction than the alternative HAT from the ring moiety (C3 atom). To unravel the origins of this regioselectivity and to explain why the observed reaction is desaturation and not the “default” hydroxyla- tion, we performed a QM/MM study on the reaction catalysed by AsqJ. Herein, we report results that complement the experi- mental findings and suggest that HAT at the C10 position is the preferred reaction due to favourable interactions between the substrate and the binding cavity that compensate for the relatively high intrinsic barrier associated with the process. For the resultant radical intermediate, the desaturation/hydroxylation selectivity is governed by electronic properties of the reactants, i.e., the energy gap between the orbital that hosts the unpaired electron and the sigma orbital for the C–H bond as well as the gap between the orbitals mixing in transition state structures for each elementary step. Graphical abstract Regiospecificity of the AsqJ dehydrogenation reaction is dictated by substrate–protein interactions. 82 × 44 mm (300 × 300 dpi) Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s0077 5-018-1575-3) contains supplementary material, which is available to authorized users. Extended author information available on the last page of the article Vol.:(0123456789) 1 3 JBIC Journal of Biological Inorganic Chemistry Keywords ODD · AsqJ dioxygenase · 4′-methoxyviridicatin · Hydrogen atom abstraction · C–H bond activation · Desaturation/hydroxylation selectivity · QM/MM · DFT Abbreviations 2OG 2-oxoglutarate HAT Hydrogen atom transfer MD Molecular dynamics ODD 2-oxoglutarate-dependent dioxygenases QM/MM Quantum mechanics/molecular mechanics TauD Taurine/α-ketoglutarate dioxygenase Introduction The Fe(II)/2-oxoglutarate-dependent dioxygenases (ODD) are mononuclear enzymes that catalyse a wide variety of reactions, i.e., hydroxylations, desaturations, halogenations, or cyclisations [1, 2]. These oxidative transformations are initiated by reactive Fe(IV)-oxo (oxoferryl) species formed Fig. 1 Desaturation and epoxidation reactions catalysed by AsqJ and when 2-oxoglutarate (2OG) is oxidatively cleaved into suc- the final nonenzymatic rearrangement. The oxidised fragment is indi- cated in grey. Marvin was used for drawing chemical structures [18] cinate and CO with use of molecular oxygen [3]. A typical iron-binding site of the ODD is composed of three amino acids from the HXDX H triad motif characteristic for the the metal ion in the oxoferryl complex is hexacoordinated family [4], 2OG bound to Fe(II) in the bidentate mode, and with one water molecule occupying the equatorial position a water molecule [5]. and another water molecule in a second coordination shell Dioxygenase AsqJ from Aspergillus nidulans catalyses [10]. In contrast, in the QM/MM study on the desaturation two key steps in biosynthesis of a quinolone antibiotic. First, mechanism, a pentacoordinated complex in trigonal bipy- the substrate, 4′-methoxycyclopeptin, is desaturated to form ramidal geometry was used. In such a model, the oxo ligand 4′-methoxydehydrocyclopeptin (see Fig. 1), which is, later, was located far from the substrate (trans to His-134) and, in a second reaction catalysed by AsqJ, epoxidized at the consequently, the computed reaction energy profiles might same position [6]. The resultant epoxide departs from the not reflect the actual interactions within the AsqJ active site. binding cavity to the solution where it undergoes an equally Therefore, it seems pertinent to revisit the mechanism of the fascinating acid/base catalysed rearrangement leading to the desaturation reaction with QM/MM methods for an experi- tricyclic quinolone antibiotic molecule, 4′-methoxyviridi- mentally consistent geometry of the activated metal cofactor catin [7]. The recent experimental results suggest that desat- and to address the still open questions concerning product uration is initiated by hydrogen atom transfer (HAT) at the specificity and regioselectivity of the reaction. benzylic (C10) position by the high-spin oxoferryl species The experimentally observed reaction outcome is the (consult Fig. 1) and, once the radical/Fe(III)-OH intermedi- dehydrogenated product [6], which raises a question about ate is formed, it might proceed through the hydroxylated the roots of selectivity in a system that can facilitate both or carbocation intermediate [8] or via a second HAT from desaturation and hydroxylation. The selectivity of the ODD the C3 atom from the ring moiety. Results of a previous catalysed reactions is a widely investigated yet still puzzling QM/MM study on the reaction mechanism of AsqJ sug- issue. Results of a hybrid DFT study on the cluster model of gested that the energetically favourable desaturation path- TauD and a synthetic shape-selective catalyst (TpOBzFe) way involves HAT at the C3 position (with energy barrier of indicated that the switch between the desaturation and −1 19.3 kcal mol ), whereas the alternative HAT from C10 is hydroxylation channels depends on intrinsic factors such as −1 hindered by a high barrier of 30 kcal mol [9]. This result C–H bond strength and delocalisation of the unpaired elec- is at odds with the experimental findings, which indicate tron on the radical intermediate; moreover, a gap between the opposite regioselectivity of the initial HAT process [8]. the electron donating orbitals, i.e., σ (operative in desatu- C–H The differences could stem from the choice of Fe(IV) coor - ration) and φ (operative in hydroxylation) alongside ener- dination geometry that was assumed in the computational getic proximity of these orbitals to the acceptor one ( ), xz∕yz study. More specifically, the previous joint experimental and plays a major role in reaction selectivity [11]. Apart from computational study on the activation of AsqJ revealed that 1 3 JBIC Journal of Biological Inorganic Chemistry 2+ this energetic factor, the extent of overlap between the donor was replaced with Fe . Hydrogen atoms were added with and acceptor orbitals can promote either radical (OH) the LEaP program from the AmberTools 14 package [19] rebound or HAT. It was shown for synthetic manganese cata- and the system was solvated by placing it in a box of TIP3P lysts that relative orientation of the radical and the water [20] with the minimal distance between the atoms of Mn(IV)–OH moiety governs the desaturation/hydroxylation the protein and the wall of the box equal to 10 Å. Water mol- partitioning [12]. Flexibility of the substrate radical and ecules present in the crystal structure were retained. Charge steric accessibility of the active site was also shown to be a of the protein was neutralised with 4 Na ions. The mini- factor that favours the desaturation pathway in non-heme misation and MD simulation were performed with use of diiron hydroxylase AlkB [13]. Similarly, the 2-oxoglutarate- Amber ff03 force field [21]. Missing parameters for residues 2+ dependent halogenase, SyrB2, controls the hydroxylation/ligating Fe were calculated based on the model of the first 2+ halogenation selectivity by positioning the fragment of the coordination shell of F e , i.e., His-134, Asp-136, and His- molecule that host the unpaired electron closer to the halo- 211, 2OG, and water molecule. The backbone of the ligating gen ligand that to the hydroxyl moiety [14, 15]. In the later amino acids was truncated at the Cα of the neighbouring res- study, it was shown, using frontier molecular orbital argu- idues. The model was optimised with the spin-unrestricted ments, that even though hydroxylation is thermodynamically B3LYP-D3 [22, 23] method and LanL2DZ basis set [24]; the favoured, chlorination is more accessible due to lower electrostatic potential (ESP) was calculated with Gaussian energy of the orbital that accepts the electron in the halo- 09 [25] at the B3LYP/cc-pVTZ/IEFPCM (ε = 4, r = 1.4 Å) genation reaction, i.e., d . [16]. level [26]. Atomic charges of the residues were calculated Fe−Cl In addition, desaturation versus hydroxylation bifurcation with use of restrained electrostatic potential (RESP) formal- was investigated for heme-containing cytochrome P450 with ism as implemented in antechamber [27]. the use of valence bond theory. This study revealed that the Prior to molecular dynamics (MD) simulation, the system preference for a given reaction stems from intrinsic proper- was minimised in three steps. First, the solvent molecules ties of the reactants, i.e., energies of the bonds formed in were optimised, whereas positions of all atoms of the solute −1 the competing processes (O–H and O–C) energy of the C–H were restrained with a 500 kcal mol harmonic potential. bond and the π-conjugation energy. Interestingly, compari- In the next step, the restraining potential was reduced to −1 son of QM and QM/MM results indicated that the electroni- 10 kcal mol . The final minimisation was performed with- cally driven preference can be altered by steric constraints out restrains for the whole solute except the first coordina- 2+ −1 imposed by the enzyme-binding pocket [17]. tion shell of F e , which was fixed with a 500 kcal mol The aim of this study was to elucidate the origins of HAT potential to retain the geometry consistent with the one from regioselectivity as well as unravel the factors that determine the crystal structure. These restraints were present through- the preference for desaturation over hydroxylation displayed out all minimisation and MD steps. Such an approach is by AsqJ. The beneath reported computational results show commonly used to treat non-heme iron containing enzymes that the reaction is initiated by HAT from the C10 atom [28–30]. MD simulation started with heating of the system and the preference for this pathway stems from favourable from 0 to 300 K in 100 ps followed by 1 ns-long density interactions between the substrate molecule and the bind- equilibration (NTP). The MD production run was per- ing cavity of AsqJ in the relevant transition structure. In the formed for 10 ns in constant temperature equal to 300 K energetically favoured path, the selectivity for desaturation and a constant pressure equal to 1 atm under periodic bound- over hydroxylation is rooted in a smaller energy gap between ary conditions. Time step was equal to 2 fs and all bonds electron donating and electron accepting orbitals for the for- involving hydrogen atoms were constrained with use of the mer process. This inherent property is not compensated by SHAKE algorithm [31]. The electrostatic energy of a cell the protein environment, but the analysis of protein–ligand was calculated by the particle-mesh Ewald method [32]. The interactions indicates that the AsqJ reaction selectivity could system reached equilibrium after 2.5 ns, as the root-mean- be changed by appropriate amino acid substitutions. square displacement (RMSD) of the backbone of the protein revealed and the stable part of the trajectory (2–10 ns, con- sult Fig. S1) was further clustered with use of agglomerative Computational methods method as implemented in CPPTRAJ [33]. The representa- tive of the dominant cluster was optimised with the ONIOM Model preparation and MD simulation method with Gaussian 09 (vide infra). The active site of the optimised structure was modi- The model of the dimeric protein was constructed based on fied in the following way: 2OG was replaced with succi- 2+ crystal structure for Ni -substituted AsqJ with 2OG and nate positioned in the same plane, oxo ligand was located 4′-methoxycyclopeptin bound in the active site (PDB ID: in axial position trans to His-211 and water molecule was 2+ 5DAQ) [7]. To obtain the model of the active protein, Ni added as an equatorial ligand trans to His-134, thus forming 1 3 JBIC Journal of Biological Inorganic Chemistry octahedral geometry as reported previously for AsqJ [10]. Single-point MM energy calculations for the structures mut The geometry of the system was optimised again with ( E ) were performed with Amber force field in Gaussian MM Gaussian 09 and prepared for MD simulation in the same 16. As the change is restricted to the MM part of the system, M M way as the AsqJ:2OG complex. The bonded parameters values of MM ( E ) and QM energy ( E ) of the model MM QM involving Fe were derived based on calculations for a bigger system were the same as in the respective optimised wild- model that incorporated three additional residues (Gln-131, type structures. The ONIOM energies of the ES and TS Thr-172, and Arg-223) necessary to retain the orientation of structures for the mutant were calculated following the sub- succinate. The production run of the MD was elongated to tracting coupling scheme: 21 ns and the representative snapshot for the QM/MM cal- culations was chosen based on clustering of the stable part of mut mut M M E = E − E + E , QM∕MM MM MM QM the trajectory (7–21 ns, the RMSD plot is shown in Fig. S2). mut where E stands for single-point MM energy of a mutant MM ONIOM calculations structure. E is a value calculated at the B3LYP-D3/def2- QM TZVP level with mechanical embedding for the QM region The QM/MM calculations were performed with the ONIOM of wild-type structure. In the further analysis, the barrier method in the Gaussian 09 program [25] on a model that heights, computed as E(TS)-E(ES), were compared with consisted of the protein and all water molecules within 20 those obtained for the wild-type form of the enzyme to Å of the Fe ion. Positions of residues with atoms located assess the stabilisation/destabilisation effect of the chosen further than 15 Å from the ion were fixed. The MM region residue. was described with the Amber force field as implemented in Gaussian [34]. The QM part consisted of Fe(IV)-oxo Cluster calculations species, side chains of residues ligating Fe(IV) (His-134, Asp-136, and His-211), equatorial water molecule, and suc- The cluster calculations were performed on a model involv- cinate modelled by acetate. Geometries were first optimised ing the iron ion and its first coordination shell (His-134, with the spin-unrestricted B3LYP method [22] and the def2- Asp-136, His-211, water molecule, acetate, and the oxo SVP basis set [35–37] employing the mechanical embed- ligand) and the substrate molecule. The positions of the Cβ ding scheme. To take into account the impact of the partial and the methyl C of acetate were fixed during optimisations. charges of the MM region on the QM part, the re-optimisa- Optimisations and frequency calculations were performed tion procedure was employed. First, the ESP of the QM part with the spin-unrestricted B3LYP and def2-SVP basis set was calculated at the same level of theory as the optimisation [22, 36]. Solvent corrections for the optimised geometries was performed; MM part was modelled as point charges; were calculated at the same level of theory and the IEFPCM however, the link atoms and their nearest (bonded) neigh- model [39] with ε = 4.0 to model the protein environment. bours were omitted. The new atomic charges of the QM part Energy was also calculated in a single-point manner with were obtained with the RESP procedure. The structures were def2-TZVP basis set. The reported energy is B3LYP-D3/ re-optimised with the new atomic charges at the B3LYP- def2-TZVP energy with ZPE and solvent corrections. D3/def2-SVP level [23] with Gaussian 16 [38], which was followed by frequency calculation. Finally, single-point energy calculations for stationary points were performed QM calculations for the nonenzymatic reaction with the same functional and the def2-TZVP basis set [36] with mechanical and electronic embedding. The reported The X-ray structure of 3′-hydroxycyclopenin was obtained energy values, unless stated otherwise, are ONIOM(B3LYP- from the CSD [40] (CSD reference code POHBEV [41]), D3/def2-TZVP, Amber) energies computed with electronic the C3′-bound hydroxyl group was replaced by hydrogen embedding plus Gibbs free energy corrections obtained at atom and methoxy group was introduced at the C4′ position the ONIOM(B3LYP-D3/def2-SVP, Amber) and mechanical to construct 4′-methoxycyclopenin. The Brønsted base and embedding level. The reported orbital gap energies as well acid compounds that take part in the reaction were mod- as distortion energies were calculated using the B3LYP-D3/ elled as acetate and acetic acid or ascorbate and ascorbic def2-TZVP method for the QM part of the system (or its acid. The geometry optimisations and analytical frequency fragments). calculations were performed at the B3LYP-D3/6-31G level [42] with solvent (water) modelled using the IEFPCM Alanine screening model with dielectric constant of the homogeneous dielec- tric medium equal to 78. For stationary points, single-point In the structures of the mutant proteins, side chain of a cho- energy calculations were performed with the 6-311G(d,p) sen amino acid was replaced with the methyl group. basis set [43] and PCM model. The reported energy values 1 3 JBIC Journal of Biological Inorganic Chemistry are B3LYP-D3/6-311G(d,p) energies with Gibbs free energy structure, the distance between the C3-bound or the C10- corrections. bound H and the O atom totals to 2.43 and 3.38 Å, respec- tively. In the optimised structure of the activated form, i.e., with 2OG substituted by succinate and Fe(II) oxidised to Results and discussion Fe(IV)=O, the distance between the reactants shortened by ca. 0.2 Å (for the C3-bound H) and by 0.6 Å (for the C10- Enzymatic reaction bound H) and the respective Fe–O–H angles increased to 148° and 92° (Fig. S3C, D). During the subsequent MD Enzyme–substrate complex: MD approach simulation, all of the monitored values increased slightly. Snapshots of the final 11 ns-long part of the trajectory were In the first step of the desaturation reaction, the hydrogen clustered. In this procedure, the distance between frames atom could be abstracted from either the benzylic position was calculated as RMSD for the fragment of the substrate (C10) or from the C3 atom of the bicyclic ring (see Fig. 2). molecule that takes part in the reaction. The frames were In the crystal structure [7] (with hydrogens added with divided into four clusters (the Davies–Bouldin Index: 0.69, LEaP; see Computational Methods), the distance between the pseudo-F statistic: 60.82; clustering metrics obtained for the C3- and C10-bound hydrogen atoms and the oxygen three and five clusters were of lower quality). The dominant atom of the carboxylate group of 2OG is 2.24 and 3.14 Å, cluster covered 95.3% of the analysed trajectory; the average respectively. The relevant Fe–O–H angles total to 113° distance between points in the cluster was 0.068 (the average and 77°. During the MD simulation, the substrate moves distance to centroid was 0.048) with standard deviation of slightly away from the metal cofactor (consult Table S1 and 0.3, whereas the distance between this cluster and the other Fig. S3A, B), and in the subsequently optimised ONIOM three was 0.125. Fig. 2 Possible pathways for AsqJ-catalysed oxidation reactions of 4′-methoxycyclopeptin. C3-bound hydrogen is shown in green; C10-bound hydrogen in blue. Marvin was used for drawing chemical structures [18] 1 3 JBIC Journal of Biological Inorganic Chemistry Hydrogen atom transfer and the spin density is located mostly on C3 (− 0.70) and high-spin Fe(III) (4.27). The Fe-bound hydroxide forms a The optimised enzyme–substrate complex (S) can be hydrogen bond with the carboxyl group of succinate. described as a high-spin Fe(IV) (spin population of 3.18) The Gibbs free energy barrier associated with the alterna- −1 coordinated by an oxo ligand (spin population of 0.59) and tive route, passing through TS-1b, is by 0.4 kcal mol lower side chains of His-134, Asp-136, His-211, succinate, and a than the barrier connected with TS-1a (see Fig. 4). This is in water molecule in octahedral geometry. Such a complex is line with the results of a recent stopped-flow/UV–Vis study the most likely species to initiate the oxidation of the pri- of AsqJ-catalysed desaturation, which suggest that the reac- mary substrate, as the previous Mössbauer and QM/MM tion is most likely initiated by abstraction of the C10-bound study on AsqJ has revealed [10].⁠ hydrogen by the Fe=O core [8]. Our computational results In the optimised structure, the C3- and C10-bound hydro- suggest that the first HAT is the rate-limiting step with a free −1 gens are positioned, respectively, 2.13 and 2.73 Å away from energy barrier of 13.4 kcal mol (TS-1b). In this transi- the oxo ligand of the oxoferryl species (Fig. S3F). Such close tion structure, the distance between the hydrogen atom and contacts are expected to facilitate hydrogen atom transfer the oxygen of the oxoferryl species totals to 1.36 Å and the (HAT) from both C10 (TS-1b) and C3 (TS-1a); however, Fe–O–H angle is equal to 114°. Typically, for such a sharp values of the Fe–O–H angle (146° for C3-bound H, 90° for angle, the electron is expected to be transferred from the C10-bound H) suggest that HAT from C3 may be preferred, σ(C–H) to π*(Fe=O) orbital (π-pathway), as their overlap at least for the sigma channel, which usually requires angles is significantly better than for the σ(C–H) and σ*(Fe=O) larger than 120° [44–46]. orbitals (σ-pathway) [48]. The spin population on C10 is The transition state structure associated with HAT at the − 0.25 and the Fe(III) ion is in the high-spin state (spin den- C3 position (TS-1a) starts the σ-pathway as manifested by sity totals to 4.16) (Fig. 3). Examination of natural orbitals spin populations: 4.03 on Fe and − 0.27 on C3. For TS- for spin density (see Fig. S4C, D) indicates a transfer of 1a, the Fe–O distance is slightly elongated as compared to the α electron between the C–H bond and the π*(Fe=O) S, 1.75 vs 1.65 Å (see Fig. 3a). The distance between the orbital. It is consistent with the π-pathway involving high- C3-bound hydrogen and the oxygen atom of oxoferryl spe- spin Fe(III) (S = 5/2) and a β radical intermediate, which cies (1.29 Å), as well as the Fe–O–H angle (140°) are within occurs alternatively to the π channel leading to intermediate- a typical range for TS of HAT proceeding in the sigma chan- spin Fe(III) and is possible due to mixing of the electronic nel (consult natural orbitals for spin density presented in Fig. states enabled by breaking the symmetry of the system [49]. S4A, B). The computed Gibbs free energy barrier for this The TS associated with the π-pathway with Fe(III) (S = 3/2) −1 −1 process equals to 13.9 kcal mol (Fig. 4). Formation of the lies ca. 5 kcal mol higher in energy. The resultant inter- −1 radical intermediate (RI-1a) is exoergic by 11.9 kcal mol mediate (RI-1b) is a radical with the unpaired electron (β Fig. 3 Optimised structures for TS-1a (a) and TS-1b (b). Distances are given in Å and spin populations larger than 0.1 are given in italics. Fig- ure rendered with PyMOL [47] 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 4 Reaction energy profile for AsqJ-catalysed desatura- tion and plausible hydroxyla- tion. The energy values, i.e., Gibbs free energy (ΔG), relative potential energy calculated at the ONIOM(B3LYP-D3/ def2-TZVP, Amber) level with electronic embedding and the relative potential energy of the QM part obtained with the B3LYP-D3/def2-TZVP method −1 are given in kcal mol −1 spin) delocalised over the C10 atom (− 0.68) and the anisyl to 22.4  kcal  mol and its height is most likely caused ring (total spin population on the ring is − 0.27). The spin by the excitation required to enter the π channel with delocalisation most likely contributes to the stability of the S(Fe(III)) = 5/2. −1 intermediate, which lies 16.9 kcal mol lower in energy To investigate the effect of protein environment for the than the initial reactant complex S. HAT barriers, the amino acid residues with atoms within 5 Å of the substrate and 2OG molecule were consecutively Regioselectivity of HAT replaced by alanine (alanine screening; for details of the pro- cedure, see Computational methods). For such mutants, the The regioselectivity of HAT can be analysed with the use ONIOM energy of S, TS-1a, and TS-1b was recalculated of Marcus theory [50] to obtain intrinsic barriers, i.e., in a single-point manner. As shown in Fig. 5b, the position- ones unaffected by thermodynamic contributions. Such ing of the substrate in optimised structure of TS-1b allows an approach was successfully employed by Srnec et al. to for favourable (mostly) interactions with hydrophobic resi- analyse the reaction selectivity of SyrB2 [16]. The relation dues lining the cavity. Comparison of optimised structures between the observed ΔG barrier and the intrinsic ΔG is of TS-1a and TS-1b reveals that, in TS-1a, the substrate int given by the equation: moves slightly away from the roof (upper part; as shown in Fig. 5a) of the binding pocket, which weakens interactions ΔG ΔG 0 0 with hydrophobic residues (for details, see Figs. S5, S6, and ΔG =ΔG + + , int 2 Table S2), but this movement is necessary to facilitate HAT 16ΔG int at the C3 position. Interestingly, TS-1a, contrary to TS-1b, is stabilised by hydrogen bonds between the ligands of the where ΔG is the reaction energy, ΔG and ΔG are known 0 0 metal cofactor (Asp-136, Fe-bound carboxyl group of succi- from the reaction-free energy profile. Thus, calculated nate) or substrate with nearby polar residues (Asn-157, Gln- intrinsic barrier associated with HAT from C3 (TS-1a; −1 −1 131, and Asn-70). This effect, however, does not compensate 19.4 kcal mol ) is by 1.6 kcal mol lower than the one −1 for weakened van der Waals interactions. for HAT from C10 (TS-1b; 21.0 kcal mol ). These values Additional computations employed a cluster model con- indicate that the observed reaction preference is driven by sisting of the first coordination shell of Fe and the substrate. higher thermodynamic driving force for the formation of The results obtained with this minimal QM model indi- RI-1b as compared to RI-1a. cate that the position of the substrate as observed in TS-1b Moreover, the preference for TS-1b over TS-1a can be results solely from its interactions with the binding pocket observed only when interactions with the binding pocket of the protein. During optimisation of the cluster variant of are included in the computational model. The picture is TS-1b, the substrate migrated from its initial position and different when the QM energies of these transition struc- formed a nearly linear Fe–O–H angle. The activation energy tures are analysed; in this case, the preferred process is −1 −1 of such a TS is 12.7 kcal mol and it is by 2.7 kcal mol HAT at the C3 position with electronic energy barrier −1 higher than the barrier associated with a QM-cluster variant equal to 11.2 kcal mol (see Fig. 4). The electronic energy of TS-1a (see Fig. S7). Therefore, it is the protein environ- barrier associated with TS-1b is relatively high, it totals ment, not electronic properties of the reactants, that governs 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 5 Binding cavity of AsqJ for TS-1a (a) and TS-1b (b). The amino acids lining the cavity are coloured from blue (stabilising the transition state) to red (destabilising the transi- tion state) the regioselectivity of HAT, which results in preferential to be operational when HAT is unfeasible for geometrical formation of RI-1b. reasons (as in the case of the C3-epimer). The TS-2b and TS-2b are both the early transitions H OH Desaturation vs. hydroxylation structures. The Fe(III)–OH distance is elongated from 1.86 Å (in RI-1b) to 1.96 Å in TS-2b , whereas it remains almost After the radical intermediate RI-1b is formed, the reac- unchanged (1.87 Å) in TS-2b . Spin density population OH tion can follow three scenarios. The C3-bound hydrogen can analysis indicates that five unpaired α electrons are located be abstracted from the intermediate to form a desaturated on Fe(III) (spin population of 4.17 and 4.13 for TS-2b and product P , which is the experimentally observed product TS-2b , respectively) and the β electron is delocalised over H OH [6]. The radical can also donate an unpaired electron to the C10 and the anisyl ring (see Fig. 6). Fe(III) site and form a carbocation, which is later deproto- The preference for desaturation over hydroxylation is nated yielding the same P . Finally, Fe-bound hydroxide most likely rooted in electronic properties of the substrate/ can recombine with the radical resulting in a hydroxylated metal cofactor pair. The electronic energy barriers calculated product P-b . for the QM part of the system (ΔE ) show that desaturation OH QM In the optimised structure of RI-1b, the distance is a more feasible process, which is consistent with ONIOM between the C3-bound hydrogen and the oxygen atom of ΔG results. The difference between the two barriers is even −1 −1 the Fe(III)–OH moiety is 1.97 Å and the Fe–O–H angle is larger, it totals to 10.5 kcal mol (cf. 1.4 kcal mol for 130°, whereas C10 atom, that hosts the unpaired electron, is ONIOM ΔG). This intrinsic preference for desaturation can located 3.48 Å away from the oxygen atom. Such a geometry be attributed to difference in energy gaps between orbitals is predisposed to desaturation, yet the hydroxylation cannot that mix in the transition states, i.e., β-dπ*(Fe–OH) (accep- be ruled out. tor for an electron in both reactions) and the β-p(C10 ) for The computed Gibbs free energy barriers indicate that hydroxylation or β-π*(C10–C3) for desaturation, which are HAT is the most favourable process (see Fig. 4). The transi- the electron source in hydroxylation and desaturation reac- tion structure associated with the second HAT (TS-2b ) lies tions, respectively. To estimate the orbital energy, we per- −1 1.4 kcal mol lower than TS for OH rebound (TS-2b ). formed single-point calculations separately for the Fe(III) OH Despite numerous attempts (QM/MM calculations, com- site and 4′-methoxycyclopeptin radical, their geometries bined also with TD-DFT and calculations done for cluster being the same as in transition structures (TS-2b and TS- models with fragmental guess), so far we have not managed 2b ) and identified orbitals most similar to ones mixing OH to optimise a species with substrate derived carbocation and in the respective transition state structures. As shown in Fe(II)–OH cofactor. Fig. 7, the energy gap between fragment orbitals that mix −1 The recent experimental study of AsqJ desaturation activ- in TS-2b totals to 60.9 kcal mol , whereas the gap for −1 ity showed that the C3-epimer of cyclopeptin is converted to TS-2b is by 6.3 kcal mol higher. The total fragments’ OH the desaturated product, and thus, the hydroxylated species distortion energies computed for these two TS are 12.0 and −1 P-b or a carbocation was suggested as intermediates [8]. 1.5 kcal mol for TS-2b and TS-2b , respectively. The OH H OH Our results show that, for the native substrate, the formation preference for desaturation might also be supported by delo- of P-b is less favourable than HAT; however, the barrier calisation of the radical over the anisyl ring, which usually OH for hydroxylation is low enough for hydroxylation pathway hinders radical rebound and hence supports entering the alternative reaction channel [11]. 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 6 Optimised structures for TS-2b (a) and TS-2b (b). Distances are given in Å and spin populations larger than 0.1 are given in italics H OH Fig. 7 Orbitals mixing in TS-2b (a) and TS-2b (b). Figure rendered using VMD [51] H OH Notably, the energy difference (ΔG and ΔE) between radical intermediate accompanied by stabilising interactions TS-2b and TS-2b decreases when interactions with the between the bicyclic ring and the cavity (Fig. 8b). Neverthe- H OH protein are taken into account; comparison of ΔG or ΔE less, the stabilisation effect of TS-2b does not compensate OH to ΔE (see Fig. 4) reveals that TS-2b is significantly for the higher electronic energy barrier for hydroxylation. QM OH stabilised by the cavity, whereas TS-2b is slightly destabi- That is why, desaturation is the experimentally observed lised. To analyse it further, we performed alanine screening process. It is possible that replacing some small residues for these two elementary steps. The results (presented in lining the cavity (e.g., Val-72 or Thr-229, see Table S3 and Fig. 8a, b) show that in, TS-2b , the substrate takes a posi- Fig. S8A, B) with more bulky ones might provide additional tion that facilitates HAT from the C3 atom at the expense stabilisation for TS-2b and, consequently, change the OH of weakening its interactions with the cavity (Fig. 8a). On reaction selectivity of AsqJ. the other hand, formation of the bond between C10 and The formation of P , as well as P-b , is an exoergic H OH OH in TS-2b requires only very modest shift of the process. The free energy of desaturation is comparable to the OH 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 8 Binding cavity of AsqJ for TS-2b (a), TS-2b (b), H OH TS-2a (c), and TS-2a (d). H OH The amino acids lining the cavity are coloured from blue (stabilising the transition state) to red (destabilising the transi- tion state) free energy of hydroxylation at the C10 position, whereas the latter process. The preference for hydroxylation is even energies of the QM system indicate that desaturation is more more pronounced in the results obtained with the cluster exothermic (see Fig. 4). The thermodynamic cycles for the model, where OH rebound occurs without a barrier (Fig. hydroxylation and desaturation show that the OH rebound S7). Analysis of the energy gap between orbitals that mix −1 · is by 4.7 kcal mol more favourable than hydrogen atom in the transition structure [β-dπ*(Fe–OH) and the β-p(C3 ) abstraction at the C3 position by the OH radical. However, for hydroxylation or β-π*(C3–C10) for desaturation] shows this is compensated by the subsequent binding of the water that the gap between orbitals mixing in TS-2a totals −1 molecule (formed in the desaturation process) to the active to 70.2  kcal  mol , and for TS-2a the gap is only by OH, −1 site, which results in the stabilisation of the system by 3 kcal mol higher (see Fig. S12), which indicates a smaller −1 11.4 kcal mol (see Figs. S9, S10). preference towards desaturation than in the TS-2b /TS- −1 The hydroxylated product P-b adapts such a position 2b pair (6 kcal mol ). The total distortion energy for OH OH −1 within the binding cavity that it is additionally stabilised by TS-2a is 10.4 kcal mol , which is slightly lower than the −1 the protein environment, which results in comparable free one calculated for TS-2b (12.0 kcal mol ), whereas the energy of desaturation and hydroxylation. total distortion energy for TS-2a is larger than for TS- OH −1 −1 2b , and it totals to 4.8 kcal mol (cf. 1.5 kcal mol for OH Back to path A: final view on the role of the protein TS-2b , shown in Table 1). Moreover, in the electronic OH in reaction selectivity structure of RI-1a (calculated for radical together with the active site), the energy gap between the β-p(C3 ) and the −1 The less favoured radical intermediate, RI-1a, when formed, lower in energy β-σ(C10–H) totals to 135.2  kcal  mol , can also undergo a subsequent hydrogen atom abstraction which is a larger value than the one calculated for β-p(C10 ) −1 (proceeding via TS-2a , shown in Fig. S11A) or hydroxyla- and β-σ(C3–H) of RI-1b (131.0 kcal mol , as shown in tion (via TS-2a , Fig. S11B). The electronic energy bar- Fig. 9). Therefore, the inherent preference for OH rebound OH −1 rier ΔE for the former is by 3 kcal mol higher than for QM 1 3 JBIC Journal of Biological Inorganic Chemistry Table 1 Values of distortion energy obtained for Fe(III) site and involves the formation of a bond between C10 and the oxo 4′-methoxycyclopeptin radical in the same geometries as in transition ligand of the oxoferryl species that yields a C3-centered rad- structures associated with desaturation/hydroxylation ical [10]. Comparison of these results with the ones reported −1 Path A (kcal mol ) Path B here for desaturation shows that both AsqJ-catalysed reac- −1 (kcal mol ) tions are initiated at the C10 position. In case of epoxidation, the reported QM/MM barrier for attack at C10 is slightly Desaturation (TS-2a/b ) 10.4 12.0 lower that the one calculated at the QM level, which indi- Hydroxylation (TS-2a/b ) 4.8 kcal 1.5 OH cates an additional stabilisation of the transition structure by the binding cavity, similar to TS-1b and TS-2b reported OH after formation of RI-1a can also stem from the increased here. stability of the σ orbital with respect to the β-p(C3 ). Interactions with the binding cavity of the protein change Nonenzymatic rearrangement the selectivity of the reaction via destabilisation of TS-2a . OH The results of alanine screening (TS-2a vs RI-1a) show The final stage of the reaction is a nonenzymatic rearrange- OH again that shortening the distance between the C3 atom and ment that is supposed to take place in the solvent outside of the iron site results in such a re-positioning of the interme- the enzyme active site. It is a two-step process, which, via diate that its favourable van der Waals interactions with the elimination of methyl isocyanate, results in formation of the binding cavity are weakened (see Fig. 8c, d). keto form of 4′-methoxyviridicatin (Fig. 10). This observation together with analysis presented above The reaction starts with formation of a bridged tricy- for the pathway B leads to a general conclusion that due to a clic intermediate (BI). In water modelled with polarizable slight change in the position of a substrate/radical interme- continuum model, this step requires crossing a high barrier −1 −1 diate transition structures for breaking the C3-H bond (TS- of 30.5 kcal mol . The barrier is lowered by 5 kcal mol 1a and TS-2b ) or formation of the C3-O bond (TS-2a ) in the presence of two water molecules forming hydrogen H OH feature weaker contacts with the amino acid residues lining bonds with oxirane fragment and the NH group of the ben- the cavity than the preceding intermediates, which results in zodiazepinedione moiety (see Fig. S13). increase of the free energy barriers. The effect of acid/base catalysis for the rearrangement The desaturated intermediate P undergoes a subsequent was investigated with the use of models consisting of the (AsqJ-catalysed) epoxidation. The process was investigated epoxide and acetic acid/acetate or ascorbic acid/ascorbate by QM/MM calculations and the proposed mechanism serving as acid H–B and base B . The general observation Fig. 9 Orbitals in RI-1a (a) and RI-1b (b) that donate and accept the electron in desaturation and hydroxylation steps 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 10 Reaction scheme for the nonenzymatic rearrangement favourable C10 keto group. In systems involving ascorbic −1 acid, the elimination goes over a barrier of ca. 6 kcal mol and imaginary frequency associated with this transition state reveals that the proton is transferred back to the ascorbate anion. Conclusions The proposed mechanism for AsqJ-catalysed desatura- tion involves HAT at the C10 position followed by second hydrogen abstraction from the neighbouring C3 atom. The regioselectivity of the first HAT stems from favourable inter - actions with the hydrophobic residues lining the binding Fig. 11 Reaction profiles for rearrangement in presence of implicit cavity, whereas product selectivity is dictated by electronic water (black line), ascorbate as B (blue), ascorbic acid as H–B properties of the reactants. However, the electronic prefer- (green), and both ascorbate and ascorbic acid (magenta) ence for desaturation is partially reduced by stabilising effect of the binding pocket that lowers the barrier for the hydroxy- is that, to a very good extent, the barrier lowering effects lation reaction. This observation opens up new possibilities are additive, and the acid molecule, which donates a pro- for switching reaction selectivity of AsqJ by introducing ton to the epoxide oxygen, is responsible for most of the mutations within the binding pocket. catalytic effect (see Figs.  11, S14). The ascorbic acid lowers −1 the barrier by 17.7 kcal mol , whereas the acetic acid by Acknowledgements This research project was supported by Grant −1 8.8 kcal mol . This difference can be partly attributed to no. UMO-2014/14/E/NZ1/00053 from the National Science Centre, Poland, and by PL-Grid Infrastructure. Calculations were performed larger pK of acetic acid (4.7 as compared to 4.2 for ascorbic at the Academic Computer Centre Cyfronet AGH. acid). Visual inspection of the imaginary frequency normal mode shows that, along with the C–C bond formation, pro- Open Access This article is distributed under the terms of the Crea- ton migrates from the donor to the epoxide oxygen, thus tive Commons Attribution 4.0 International License (http://creat iveco forming a hydroxyl group.mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- tion, and reproduction in any medium, provided you give appropriate Regardless of the presence of base molecule, that can credit to the original author(s) and the source, provide a link to the accept proton from the NH group, the N–H bond remains Creative Commons license, and indicate if changes were made. intact, and thus, the resulting bridged bicyclic intermediate has a cationic character. However, the presence of a nega- tively charged entity interacting with the cationic moiety References lowers the Gibbs free energy of the BI intermediate by ca. 2 −1 (ascorbate) − 5 (acetate) kcal mol . 1. Martinez S, Hausinger RP (2015) Catalytic mechanisms of The final step of the reaction, elimination of methyl iso- Fe(II)- and 2-oxoglutarate-dependent oxygenases. J Biol Chem cyanate, in the absence of acid molecule occurs without any 290:20702–20711 energy barrier as it leads to the formation of energetically 1 3 JBIC Journal of Biological Inorganic Chemistry 2. Abu-Omar MM, Loaiza A, Hontzeas N (2005) Reaction mecha- 20. Jorgensen WL, Chandrasekhar J, Madura JD et al (1983) Com- nisms of mononuclear non-heme iron oxygenases. Chem Rev parison of simple potential functions for simulating liquid water. 105:2227–2252 J Chem Phys 79:926–935 3. Price JC, Barr EW, Glass TE et al (2003) Evidence for hydrogen 21. Case DA, Cheatham TE, Darden T et al (2005) The Amber bio- abstraction from C1 of taurine by the high-spin Fe(IV) intermedi- molecular simulation programs. J Comput Chem 26:1668–1688 ate detected during oxygen activation by taurine: α-ketoglutarate 22. Becke AD (1993) Density-functional thermochemistry. III. The dioxygenase (TauD). J Am Chem Soc 125:13008–13009 role of exact exchange. J Chem Phys 98:5648–5652 4. Hegg EL, Que L (1997) The 2-His-1-carboxylate facial triad– 23. Grimme S, Ehrlich S, Goerigk L (2011) Effect of the damping an emerging structural motif in mononuclear non-heme iron(II) function in dispersion corrected density functional theory. J Com- enzymes. Eur J Biochem 250:625–629 put Chem 32:1456–1465 5. Hausinger RP (2004) Fe(II)/α-ketoglutarate-dependent hydroxy- 24. Hay PJ, Wadt WR (1985) Ab initio effective core potentials for lases and related enzymes. Crit Rev Biochem Mol Biol 39:21–68 molecular calculations. Potentials for K to Au including the out- 6. Ishikawa N, Tanaka H, Koyama F et al (2014) Non-heme dioxy- ermost core orbitals. J Chem Phys 82:299–310 genase catalyzes atypical oxidations of 6,7-bicyclic systems to 25. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, form the 6,6-quinolone core of viridicatin-type fungal alkaloids. Cheeseman JR, Scalmani G, Barone V, Petersson GA, Nakat- Angew Chemie Int Ed 53:12880–12884 suji H, Li X, Caricato M, Marenich A, Bloino J, Janesko BG, 7. Bräuer A, Beck P, Hintermann L, Groll M (2016) Structure of Gomperts R, Mennucci B, Hratchian HP, Ortiz JV, Izmaylov AF, the dioxygenase AsqJ: mechanistic Insights into a one-pot mul- Sonnenberg JL, Williams-Young D, Ding F, Lipparini F, Egidi tistep quinolone antibiotic biosynthesis. Angew Chemie Int Ed F, Goings J, Peng B, Petrone A, Henderson T, Ranasinghe D, 55:422–426 Zakrzewski VG, Gao J, Rega N, Zheng G, Liang W, Hada M, 8. Liao H-J, Li J, Huang J-L et al (2018) Insights into the desatura- Ehara M, Toyota K, Fukuda R, Hasegawa J, Ishida M, Nakajima tion of cyclopeptin and its C3 epimer catalyzed by a non-heme T, Honda Y, Kitao O, Nakai H, Vreven T, Throssell K, Montgom- iron enzyme: structural characterization and mechanism elucida- ery JA, Peralta JE Jr, Ogliaro F, Bearpark M, Heyd JJ, Brothers tion. Angew Chemie Int Ed 57:1831–1835 E, Kudin KN, Staroverov VN, Keith T, Kobayashi R, Normand 9. Su H, Sheng X, Zhu W et al (2017) Mechanistic Insights into the J, Raghavachari K, Rendell A, Burant JC, Iyengar SS, Tomasi J, decoupled desaturation and epoxidation catalyzed by dioxygenase Cossi M, Millam JM, Klene M, Adamo C, Cammi R, Ochterski AsqJ involved in the biosynthesis of quinolone alkaloids. ACS JW, Martin RL, Morokuma K, Farkas O, Foresman JB, Fox DJ Catal 7:5534–5543 (2013) Gaussian 09, revision D.01. Gaussian Inc., Wallingford 10. Song X, Lu J, Lai W (2017) Mechanistic insights into dioxygen 26. Duan Y, Wu C, Chowdhury S et al (2003) A point-charge force activation, oxygen atom exchange and substrate epoxidation by field for molecular mechanics simulations of proteins based on AsqJ dioxygenase from quantum mechanical/molecular mechani- condensed-phase quantum mechanical calculations. J Comput cal calculations. Phys Chem Chem Phys 19:20188–20197 Chem 24:1999–2012 11. Usharani D, Janardanan D, Shaik S (2011) Does the TauD 27. Wang J, Cieplak P, Kollman PA, Kollman PA (2000) How well enzyme always hydroxylate alkanes, while an analogous syn- does a restrained electrostatic potential (RESP) model perform thetic non-heme reagent always desaturates them? J Am Chem in calculating conformational energies of organic and biological Soc 133:176–179 molecules? J Comput Chem J Comput Chem 21:1049–1074 12. Hull JF, Balcells D, Sauer ELO et al (2010) Manganese catalysts 28. Kumar D, Thiel W, de Visser SP (2011) Theoretical study on the for C–H activation: an experimental/theoretical study identi- mechanism of the oxygen activation process in cysteine dioxyge- fies the stereoelectronic factor that controls the switch between nase enzymes. J Am Chem Soc 133:3869–3882 hydroxylation and desaturation pathways. J Am Chem Soc 29. Wang B, Cao Z, Sharon DA, Shaik S (2015) Computations reveal 132:7605–7616 a rich mechanistic variation of demethylation of N-methylated 13. Cooper HLR, Mishra G, Huang X et al (2012) Parallel and com- DNA/RNA nucleotides by FTO. ACS Catal 5:7077–7090 petitive pathways for substrate desaturation, hydroxylation, and 30. Wang X, Su H, Liu Y (2017) Insights into the unprecedented radical rearrangement by the non-heme diiron hydroxylase AlkB. epoxidation mechanism of fumitremorgin B endoperoxidase J Am Chem Soc 134:20365–20375 (FtmOx1) from Aspergillus fumigatus by QM/MM calculations. 14. Matthews ML, Neumann CS, Miles LA et al (2009) Substrate Phys Chem Chem Phys 19:7668–7677 positioning controls the partition between halogenation and 31. Miyamoto S, Kollman PA (1992) Settle: an analytical version of hydroxylation in the aliphatic halogenase, SyrB2. Proc Natl Acad the SHAKE and RATTLE algorithm for rigid water models. J Sci USA 106:17723–17728 Comput Chem 13:952–962 15. Huang J, Li C, Wang B et al (2016) Selective chlorination of 32. Essmann U, Perera L, Berkowitz ML et al (1995) A smooth par- substrates by the halogenase SyrB2 is controlled by the protein ticle mesh Ewald method. J Chem Phys 103:8577–8593 according to a combined quantum mechanics/molecular mechan- 33. Roe DR, Cheatham TE (2013) PTRAJ and CPPTRAJ: software ics and molecular dynamics study. ACS Catal 6:2694–2704 for processing and analysis of molecular dynamics trajectory data. 16. Srnec M, Solomon EI (2017) Frontier molecular orbital contribu- J Chem Theory Comput 9:3084–3095 tions to chlorination versus hydroxylation selectivity in the non- 34. Cornell WD, Cieplak P, Bayly CI et al (1995) A second genera- heme iron halogenase SyrB2. J Am Chem Soc 139:2396–2407 tion force field for the simulation of proteins, nucleic acids, and 17. Ji L, Faponle AS, Quesne MG et al (2015) Drug metabolism by organic molecules. J Am Chem Soc 117:5179–5197 cytochrome P450 enzymes: what distinguishes the pathways lead- 35. Feller D (1996) The role of databases in support of computational ing to substrate hydroxylation over desaturation? Chem A Eur J chemistry calculations. J Comput Chem 17:1571–1586 21:9083–9092 36. Weigend F, Ahlrichs R (2005) Balanced basis sets of split valence, 18. Marvin was used for drawing, displaying and characterizing chem- triple zeta valence and quadruple zeta valence quality for H to ical structures, substructures and reactions. Marvin 17.28.0, 2017, Rn: design and assessment of accuracy. Phys Chem Chem Phys ChemAxon. http://www.chema xon.com 7:3297–3305 19. Wang J, Wang W, Kollman PA, Case DA (2006) Automatic atom 37. Schuchardt KL, Didier BT, Elsethagen T et al (2007) Basis set exchange: a community database for computational sciences. J type and bond type perception in molecular mechanical calcula- Chem Inf Model 47:1045–1052 tions. J Mol Graph Model 25:247–260 1 3 JBIC Journal of Biological Inorganic Chemistry 38. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, molecular-orbital studies of organic molecules. J Chem Phys Cheeseman JR, Scalmani G, Barone V, Petersson GA, Nakat- 54:724–728 suji H, Li X, Caricato M, Marenich AV, Bloino J, Janesko BG, 43. Krishnan R, Binkley JS, Seeger R, Pople JA (1980) Self-consistent Gomperts R, Mennucci B, Hratchian HP, Ortiz JV, Izmaylov AF, molecular orbital methods. XX. A basis set for correlated wave Sonnenberg JL, Williams-Young D, Ding F, Lipparini F, Egidi functions. J Chem Phys 72:650–654 F, Goings J, Peng B, Petrone A, Henderson T, Ranasinghe D, 44. Bernasconi L, Baerends EJ (2008) The EDTA complex of Zakrzewski VG, Gao J, Rega N, Zheng G, Liang W, Hada M, oxidoiron(IV) as realisation of an optimal ligand environment for Ehara M, Toyota K, Fukuda R, Hasegawa J, Ishida M, Nakajima high activity of FeO +. Eur J Inorg Chem 2008:1672–1681 T, Honda Y, Kitao O, Nakai H, Vreven T, Throssell K, Montgom- 45. Geng C, Ye S, Neese F (2010) Analysis of reaction channels ery JA, Peralta JE Jr, Ogliaro F, Bearpark MJ, Heyd JJ, Brothers for alkane hydroxylation by nonheme iron(IV)-oxo complexes. EN, Kudin KN, Staroverov VN, Keith TA, Kobayashi R, Normand Angew Chemie Int Ed 49:5717–5720 J, Raghavachari K, Rendell AP, Burant JC, Iyengar SS, Tomasi J, 46. Janardanan D, Usharani D, Chen H, Shaik S (2011) Modeling Cossi M, Millam JM, Klene M, Adamo C, Cammi R, Ochterski C–H abstraction reactivity of nonheme Fe(IV)O oxidants with JW, Martin RL, Morokuma K, Farkas O, Foresman JB, Fox DJ alkanes: what role do counter ions play? J Phys Chem Lett (2016) Gaussian 16, Revision A.03. Gaussian Inc., Wallingford 2:2610–2617 39. Miertus S, Scrocco E, Tomasi J (1982) Approximate evaluations 47. Schrödinger L (2015) The PyMOL molecular graphics system. of the electrostatic free energy and internal energy changes in Version 1:8 solution processes. Chem Phys 65:239–245 48. Ye S, Geng C-Y, Shaik S, Neese F (2013) Electronic structure 40. Groom CR, Bruno IJ, Lightfoot MP, Ward SC (2016) The Cam- analysis of multistate reactivity in transition metal catalyzed reac- bridge structural database. Acta Crystallogr Sect B Struct Sci tions: the case of C–H bond activation by non-heme iron(iv)–oxo Cryst Eng Mater 72:171–179 cores. Phys Chem Chem Phys 15:8017–8030 41. Li J, Wang J, Jiang C-S et al (2014) (+)-Cyclopenol, a new natu- 49. Srnec M, Wong SD, England J et al (2012) Frontier molecular rally occurring 7-membered 2,5-dioxopiperazine alkaloid from orbitals in S = 2 ferryl species and elucidation of their contribu- the fungus Penicillium sclerotiorum endogenous with the Chi- tions to reactivity. Proc Natl Acad Sci 109:14326–14331 nese mangrove Bruguiera gymnorrhiza. J Asian Nat Prod Res 50. Marcus RA, Sutin N (1985) Electron transfers in chemistry and 16:542–548 biology. Biochim Biophys Acta Rev Bioenerg 811:265–322 42. Ditchfield R, Hehre WJ, Pople JA (1971) Self-consistent molec- 51. Humphrey W, Dalke A, Schulten K (1996) VMD: visual molecu- ular-orbital methods. IX. An extended Gaussian-type basis for lar dynamics. J Mol Graph 14:33–38 Affiliations 1 1 Zuzanna Wojdyla  · Tomasz Borowski * Zuzanna Wojdyla Jerzy Haber Institute of Catalysis and Surface Chemistry, ncwojdyl@cyf-kr.edu.pl Polish Academy of Sciences, Niezapominajek 8, 30239 Kraków, Poland * Tomasz Borowski ncborows@cyf-kr.edu.pl 1 3 http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png JBIC Journal of Biological Inorganic Chemistry Springer Journals

On how the binding cavity of AsqJ dioxygenase controls the desaturation reaction regioselectivity: a QM/MM study

Free
14 pages
Loading next page...
 
/lp/springer_journal/on-how-the-binding-cavity-of-asqj-dioxygenase-controls-the-m004VZHbCP
Publisher
Springer Berlin Heidelberg
Copyright
Copyright © 2018 by The Author(s)
Subject
Life Sciences; Biochemistry, general; Microbiology
ISSN
0949-8257
eISSN
1432-1327
D.O.I.
10.1007/s00775-018-1575-3
Publisher site
See Article on Publisher Site

Abstract

The Fe(II)/2-oxoglutarate-dependent dioxygenase AsqJ from Aspergillus nidulans catalyses two pivotal steps in the synthesis of quinolone antibiotic 4′-methoxyviridicatin, i.e., desaturation and epoxidation of a benzodiazepinedione. The previous experimental results signal that, during the desaturation reaction, hydrogen atom transfer (HAT) from the benzylic carbon atom (C10) is a more likely step to initiate the reaction than the alternative HAT from the ring moiety (C3 atom). To unravel the origins of this regioselectivity and to explain why the observed reaction is desaturation and not the “default” hydroxyla- tion, we performed a QM/MM study on the reaction catalysed by AsqJ. Herein, we report results that complement the experi- mental findings and suggest that HAT at the C10 position is the preferred reaction due to favourable interactions between the substrate and the binding cavity that compensate for the relatively high intrinsic barrier associated with the process. For the resultant radical intermediate, the desaturation/hydroxylation selectivity is governed by electronic properties of the reactants, i.e., the energy gap between the orbital that hosts the unpaired electron and the sigma orbital for the C–H bond as well as the gap between the orbitals mixing in transition state structures for each elementary step. Graphical abstract Regiospecificity of the AsqJ dehydrogenation reaction is dictated by substrate–protein interactions. 82 × 44 mm (300 × 300 dpi) Electronic supplementary material The online version of this article (https ://doi.org/10.1007/s0077 5-018-1575-3) contains supplementary material, which is available to authorized users. Extended author information available on the last page of the article Vol.:(0123456789) 1 3 JBIC Journal of Biological Inorganic Chemistry Keywords ODD · AsqJ dioxygenase · 4′-methoxyviridicatin · Hydrogen atom abstraction · C–H bond activation · Desaturation/hydroxylation selectivity · QM/MM · DFT Abbreviations 2OG 2-oxoglutarate HAT Hydrogen atom transfer MD Molecular dynamics ODD 2-oxoglutarate-dependent dioxygenases QM/MM Quantum mechanics/molecular mechanics TauD Taurine/α-ketoglutarate dioxygenase Introduction The Fe(II)/2-oxoglutarate-dependent dioxygenases (ODD) are mononuclear enzymes that catalyse a wide variety of reactions, i.e., hydroxylations, desaturations, halogenations, or cyclisations [1, 2]. These oxidative transformations are initiated by reactive Fe(IV)-oxo (oxoferryl) species formed Fig. 1 Desaturation and epoxidation reactions catalysed by AsqJ and when 2-oxoglutarate (2OG) is oxidatively cleaved into suc- the final nonenzymatic rearrangement. The oxidised fragment is indi- cated in grey. Marvin was used for drawing chemical structures [18] cinate and CO with use of molecular oxygen [3]. A typical iron-binding site of the ODD is composed of three amino acids from the HXDX H triad motif characteristic for the the metal ion in the oxoferryl complex is hexacoordinated family [4], 2OG bound to Fe(II) in the bidentate mode, and with one water molecule occupying the equatorial position a water molecule [5]. and another water molecule in a second coordination shell Dioxygenase AsqJ from Aspergillus nidulans catalyses [10]. In contrast, in the QM/MM study on the desaturation two key steps in biosynthesis of a quinolone antibiotic. First, mechanism, a pentacoordinated complex in trigonal bipy- the substrate, 4′-methoxycyclopeptin, is desaturated to form ramidal geometry was used. In such a model, the oxo ligand 4′-methoxydehydrocyclopeptin (see Fig. 1), which is, later, was located far from the substrate (trans to His-134) and, in a second reaction catalysed by AsqJ, epoxidized at the consequently, the computed reaction energy profiles might same position [6]. The resultant epoxide departs from the not reflect the actual interactions within the AsqJ active site. binding cavity to the solution where it undergoes an equally Therefore, it seems pertinent to revisit the mechanism of the fascinating acid/base catalysed rearrangement leading to the desaturation reaction with QM/MM methods for an experi- tricyclic quinolone antibiotic molecule, 4′-methoxyviridi- mentally consistent geometry of the activated metal cofactor catin [7]. The recent experimental results suggest that desat- and to address the still open questions concerning product uration is initiated by hydrogen atom transfer (HAT) at the specificity and regioselectivity of the reaction. benzylic (C10) position by the high-spin oxoferryl species The experimentally observed reaction outcome is the (consult Fig. 1) and, once the radical/Fe(III)-OH intermedi- dehydrogenated product [6], which raises a question about ate is formed, it might proceed through the hydroxylated the roots of selectivity in a system that can facilitate both or carbocation intermediate [8] or via a second HAT from desaturation and hydroxylation. The selectivity of the ODD the C3 atom from the ring moiety. Results of a previous catalysed reactions is a widely investigated yet still puzzling QM/MM study on the reaction mechanism of AsqJ sug- issue. Results of a hybrid DFT study on the cluster model of gested that the energetically favourable desaturation path- TauD and a synthetic shape-selective catalyst (TpOBzFe) way involves HAT at the C3 position (with energy barrier of indicated that the switch between the desaturation and −1 19.3 kcal mol ), whereas the alternative HAT from C10 is hydroxylation channels depends on intrinsic factors such as −1 hindered by a high barrier of 30 kcal mol [9]. This result C–H bond strength and delocalisation of the unpaired elec- is at odds with the experimental findings, which indicate tron on the radical intermediate; moreover, a gap between the opposite regioselectivity of the initial HAT process [8]. the electron donating orbitals, i.e., σ (operative in desatu- C–H The differences could stem from the choice of Fe(IV) coor - ration) and φ (operative in hydroxylation) alongside ener- dination geometry that was assumed in the computational getic proximity of these orbitals to the acceptor one ( ), xz∕yz study. More specifically, the previous joint experimental and plays a major role in reaction selectivity [11]. Apart from computational study on the activation of AsqJ revealed that 1 3 JBIC Journal of Biological Inorganic Chemistry 2+ this energetic factor, the extent of overlap between the donor was replaced with Fe . Hydrogen atoms were added with and acceptor orbitals can promote either radical (OH) the LEaP program from the AmberTools 14 package [19] rebound or HAT. It was shown for synthetic manganese cata- and the system was solvated by placing it in a box of TIP3P lysts that relative orientation of the radical and the water [20] with the minimal distance between the atoms of Mn(IV)–OH moiety governs the desaturation/hydroxylation the protein and the wall of the box equal to 10 Å. Water mol- partitioning [12]. Flexibility of the substrate radical and ecules present in the crystal structure were retained. Charge steric accessibility of the active site was also shown to be a of the protein was neutralised with 4 Na ions. The mini- factor that favours the desaturation pathway in non-heme misation and MD simulation were performed with use of diiron hydroxylase AlkB [13]. Similarly, the 2-oxoglutarate- Amber ff03 force field [21]. Missing parameters for residues 2+ dependent halogenase, SyrB2, controls the hydroxylation/ligating Fe were calculated based on the model of the first 2+ halogenation selectivity by positioning the fragment of the coordination shell of F e , i.e., His-134, Asp-136, and His- molecule that host the unpaired electron closer to the halo- 211, 2OG, and water molecule. The backbone of the ligating gen ligand that to the hydroxyl moiety [14, 15]. In the later amino acids was truncated at the Cα of the neighbouring res- study, it was shown, using frontier molecular orbital argu- idues. The model was optimised with the spin-unrestricted ments, that even though hydroxylation is thermodynamically B3LYP-D3 [22, 23] method and LanL2DZ basis set [24]; the favoured, chlorination is more accessible due to lower electrostatic potential (ESP) was calculated with Gaussian energy of the orbital that accepts the electron in the halo- 09 [25] at the B3LYP/cc-pVTZ/IEFPCM (ε = 4, r = 1.4 Å) genation reaction, i.e., d . [16]. level [26]. Atomic charges of the residues were calculated Fe−Cl In addition, desaturation versus hydroxylation bifurcation with use of restrained electrostatic potential (RESP) formal- was investigated for heme-containing cytochrome P450 with ism as implemented in antechamber [27]. the use of valence bond theory. This study revealed that the Prior to molecular dynamics (MD) simulation, the system preference for a given reaction stems from intrinsic proper- was minimised in three steps. First, the solvent molecules ties of the reactants, i.e., energies of the bonds formed in were optimised, whereas positions of all atoms of the solute −1 the competing processes (O–H and O–C) energy of the C–H were restrained with a 500 kcal mol harmonic potential. bond and the π-conjugation energy. Interestingly, compari- In the next step, the restraining potential was reduced to −1 son of QM and QM/MM results indicated that the electroni- 10 kcal mol . The final minimisation was performed with- cally driven preference can be altered by steric constraints out restrains for the whole solute except the first coordina- 2+ −1 imposed by the enzyme-binding pocket [17]. tion shell of F e , which was fixed with a 500 kcal mol The aim of this study was to elucidate the origins of HAT potential to retain the geometry consistent with the one from regioselectivity as well as unravel the factors that determine the crystal structure. These restraints were present through- the preference for desaturation over hydroxylation displayed out all minimisation and MD steps. Such an approach is by AsqJ. The beneath reported computational results show commonly used to treat non-heme iron containing enzymes that the reaction is initiated by HAT from the C10 atom [28–30]. MD simulation started with heating of the system and the preference for this pathway stems from favourable from 0 to 300 K in 100 ps followed by 1 ns-long density interactions between the substrate molecule and the bind- equilibration (NTP). The MD production run was per- ing cavity of AsqJ in the relevant transition structure. In the formed for 10 ns in constant temperature equal to 300 K energetically favoured path, the selectivity for desaturation and a constant pressure equal to 1 atm under periodic bound- over hydroxylation is rooted in a smaller energy gap between ary conditions. Time step was equal to 2 fs and all bonds electron donating and electron accepting orbitals for the for- involving hydrogen atoms were constrained with use of the mer process. This inherent property is not compensated by SHAKE algorithm [31]. The electrostatic energy of a cell the protein environment, but the analysis of protein–ligand was calculated by the particle-mesh Ewald method [32]. The interactions indicates that the AsqJ reaction selectivity could system reached equilibrium after 2.5 ns, as the root-mean- be changed by appropriate amino acid substitutions. square displacement (RMSD) of the backbone of the protein revealed and the stable part of the trajectory (2–10 ns, con- sult Fig. S1) was further clustered with use of agglomerative Computational methods method as implemented in CPPTRAJ [33]. The representa- tive of the dominant cluster was optimised with the ONIOM Model preparation and MD simulation method with Gaussian 09 (vide infra). The active site of the optimised structure was modi- The model of the dimeric protein was constructed based on fied in the following way: 2OG was replaced with succi- 2+ crystal structure for Ni -substituted AsqJ with 2OG and nate positioned in the same plane, oxo ligand was located 4′-methoxycyclopeptin bound in the active site (PDB ID: in axial position trans to His-211 and water molecule was 2+ 5DAQ) [7]. To obtain the model of the active protein, Ni added as an equatorial ligand trans to His-134, thus forming 1 3 JBIC Journal of Biological Inorganic Chemistry octahedral geometry as reported previously for AsqJ [10]. Single-point MM energy calculations for the structures mut The geometry of the system was optimised again with ( E ) were performed with Amber force field in Gaussian MM Gaussian 09 and prepared for MD simulation in the same 16. As the change is restricted to the MM part of the system, M M way as the AsqJ:2OG complex. The bonded parameters values of MM ( E ) and QM energy ( E ) of the model MM QM involving Fe were derived based on calculations for a bigger system were the same as in the respective optimised wild- model that incorporated three additional residues (Gln-131, type structures. The ONIOM energies of the ES and TS Thr-172, and Arg-223) necessary to retain the orientation of structures for the mutant were calculated following the sub- succinate. The production run of the MD was elongated to tracting coupling scheme: 21 ns and the representative snapshot for the QM/MM cal- culations was chosen based on clustering of the stable part of mut mut M M E = E − E + E , QM∕MM MM MM QM the trajectory (7–21 ns, the RMSD plot is shown in Fig. S2). mut where E stands for single-point MM energy of a mutant MM ONIOM calculations structure. E is a value calculated at the B3LYP-D3/def2- QM TZVP level with mechanical embedding for the QM region The QM/MM calculations were performed with the ONIOM of wild-type structure. In the further analysis, the barrier method in the Gaussian 09 program [25] on a model that heights, computed as E(TS)-E(ES), were compared with consisted of the protein and all water molecules within 20 those obtained for the wild-type form of the enzyme to Å of the Fe ion. Positions of residues with atoms located assess the stabilisation/destabilisation effect of the chosen further than 15 Å from the ion were fixed. The MM region residue. was described with the Amber force field as implemented in Gaussian [34]. The QM part consisted of Fe(IV)-oxo Cluster calculations species, side chains of residues ligating Fe(IV) (His-134, Asp-136, and His-211), equatorial water molecule, and suc- The cluster calculations were performed on a model involv- cinate modelled by acetate. Geometries were first optimised ing the iron ion and its first coordination shell (His-134, with the spin-unrestricted B3LYP method [22] and the def2- Asp-136, His-211, water molecule, acetate, and the oxo SVP basis set [35–37] employing the mechanical embed- ligand) and the substrate molecule. The positions of the Cβ ding scheme. To take into account the impact of the partial and the methyl C of acetate were fixed during optimisations. charges of the MM region on the QM part, the re-optimisa- Optimisations and frequency calculations were performed tion procedure was employed. First, the ESP of the QM part with the spin-unrestricted B3LYP and def2-SVP basis set was calculated at the same level of theory as the optimisation [22, 36]. Solvent corrections for the optimised geometries was performed; MM part was modelled as point charges; were calculated at the same level of theory and the IEFPCM however, the link atoms and their nearest (bonded) neigh- model [39] with ε = 4.0 to model the protein environment. bours were omitted. The new atomic charges of the QM part Energy was also calculated in a single-point manner with were obtained with the RESP procedure. The structures were def2-TZVP basis set. The reported energy is B3LYP-D3/ re-optimised with the new atomic charges at the B3LYP- def2-TZVP energy with ZPE and solvent corrections. D3/def2-SVP level [23] with Gaussian 16 [38], which was followed by frequency calculation. Finally, single-point energy calculations for stationary points were performed QM calculations for the nonenzymatic reaction with the same functional and the def2-TZVP basis set [36] with mechanical and electronic embedding. The reported The X-ray structure of 3′-hydroxycyclopenin was obtained energy values, unless stated otherwise, are ONIOM(B3LYP- from the CSD [40] (CSD reference code POHBEV [41]), D3/def2-TZVP, Amber) energies computed with electronic the C3′-bound hydroxyl group was replaced by hydrogen embedding plus Gibbs free energy corrections obtained at atom and methoxy group was introduced at the C4′ position the ONIOM(B3LYP-D3/def2-SVP, Amber) and mechanical to construct 4′-methoxycyclopenin. The Brønsted base and embedding level. The reported orbital gap energies as well acid compounds that take part in the reaction were mod- as distortion energies were calculated using the B3LYP-D3/ elled as acetate and acetic acid or ascorbate and ascorbic def2-TZVP method for the QM part of the system (or its acid. The geometry optimisations and analytical frequency fragments). calculations were performed at the B3LYP-D3/6-31G level [42] with solvent (water) modelled using the IEFPCM Alanine screening model with dielectric constant of the homogeneous dielec- tric medium equal to 78. For stationary points, single-point In the structures of the mutant proteins, side chain of a cho- energy calculations were performed with the 6-311G(d,p) sen amino acid was replaced with the methyl group. basis set [43] and PCM model. The reported energy values 1 3 JBIC Journal of Biological Inorganic Chemistry are B3LYP-D3/6-311G(d,p) energies with Gibbs free energy structure, the distance between the C3-bound or the C10- corrections. bound H and the O atom totals to 2.43 and 3.38 Å, respec- tively. In the optimised structure of the activated form, i.e., with 2OG substituted by succinate and Fe(II) oxidised to Results and discussion Fe(IV)=O, the distance between the reactants shortened by ca. 0.2 Å (for the C3-bound H) and by 0.6 Å (for the C10- Enzymatic reaction bound H) and the respective Fe–O–H angles increased to 148° and 92° (Fig. S3C, D). During the subsequent MD Enzyme–substrate complex: MD approach simulation, all of the monitored values increased slightly. Snapshots of the final 11 ns-long part of the trajectory were In the first step of the desaturation reaction, the hydrogen clustered. In this procedure, the distance between frames atom could be abstracted from either the benzylic position was calculated as RMSD for the fragment of the substrate (C10) or from the C3 atom of the bicyclic ring (see Fig. 2). molecule that takes part in the reaction. The frames were In the crystal structure [7] (with hydrogens added with divided into four clusters (the Davies–Bouldin Index: 0.69, LEaP; see Computational Methods), the distance between the pseudo-F statistic: 60.82; clustering metrics obtained for the C3- and C10-bound hydrogen atoms and the oxygen three and five clusters were of lower quality). The dominant atom of the carboxylate group of 2OG is 2.24 and 3.14 Å, cluster covered 95.3% of the analysed trajectory; the average respectively. The relevant Fe–O–H angles total to 113° distance between points in the cluster was 0.068 (the average and 77°. During the MD simulation, the substrate moves distance to centroid was 0.048) with standard deviation of slightly away from the metal cofactor (consult Table S1 and 0.3, whereas the distance between this cluster and the other Fig. S3A, B), and in the subsequently optimised ONIOM three was 0.125. Fig. 2 Possible pathways for AsqJ-catalysed oxidation reactions of 4′-methoxycyclopeptin. C3-bound hydrogen is shown in green; C10-bound hydrogen in blue. Marvin was used for drawing chemical structures [18] 1 3 JBIC Journal of Biological Inorganic Chemistry Hydrogen atom transfer and the spin density is located mostly on C3 (− 0.70) and high-spin Fe(III) (4.27). The Fe-bound hydroxide forms a The optimised enzyme–substrate complex (S) can be hydrogen bond with the carboxyl group of succinate. described as a high-spin Fe(IV) (spin population of 3.18) The Gibbs free energy barrier associated with the alterna- −1 coordinated by an oxo ligand (spin population of 0.59) and tive route, passing through TS-1b, is by 0.4 kcal mol lower side chains of His-134, Asp-136, His-211, succinate, and a than the barrier connected with TS-1a (see Fig. 4). This is in water molecule in octahedral geometry. Such a complex is line with the results of a recent stopped-flow/UV–Vis study the most likely species to initiate the oxidation of the pri- of AsqJ-catalysed desaturation, which suggest that the reac- mary substrate, as the previous Mössbauer and QM/MM tion is most likely initiated by abstraction of the C10-bound study on AsqJ has revealed [10].⁠ hydrogen by the Fe=O core [8]. Our computational results In the optimised structure, the C3- and C10-bound hydro- suggest that the first HAT is the rate-limiting step with a free −1 gens are positioned, respectively, 2.13 and 2.73 Å away from energy barrier of 13.4 kcal mol (TS-1b). In this transi- the oxo ligand of the oxoferryl species (Fig. S3F). Such close tion structure, the distance between the hydrogen atom and contacts are expected to facilitate hydrogen atom transfer the oxygen of the oxoferryl species totals to 1.36 Å and the (HAT) from both C10 (TS-1b) and C3 (TS-1a); however, Fe–O–H angle is equal to 114°. Typically, for such a sharp values of the Fe–O–H angle (146° for C3-bound H, 90° for angle, the electron is expected to be transferred from the C10-bound H) suggest that HAT from C3 may be preferred, σ(C–H) to π*(Fe=O) orbital (π-pathway), as their overlap at least for the sigma channel, which usually requires angles is significantly better than for the σ(C–H) and σ*(Fe=O) larger than 120° [44–46]. orbitals (σ-pathway) [48]. The spin population on C10 is The transition state structure associated with HAT at the − 0.25 and the Fe(III) ion is in the high-spin state (spin den- C3 position (TS-1a) starts the σ-pathway as manifested by sity totals to 4.16) (Fig. 3). Examination of natural orbitals spin populations: 4.03 on Fe and − 0.27 on C3. For TS- for spin density (see Fig. S4C, D) indicates a transfer of 1a, the Fe–O distance is slightly elongated as compared to the α electron between the C–H bond and the π*(Fe=O) S, 1.75 vs 1.65 Å (see Fig. 3a). The distance between the orbital. It is consistent with the π-pathway involving high- C3-bound hydrogen and the oxygen atom of oxoferryl spe- spin Fe(III) (S = 5/2) and a β radical intermediate, which cies (1.29 Å), as well as the Fe–O–H angle (140°) are within occurs alternatively to the π channel leading to intermediate- a typical range for TS of HAT proceeding in the sigma chan- spin Fe(III) and is possible due to mixing of the electronic nel (consult natural orbitals for spin density presented in Fig. states enabled by breaking the symmetry of the system [49]. S4A, B). The computed Gibbs free energy barrier for this The TS associated with the π-pathway with Fe(III) (S = 3/2) −1 −1 process equals to 13.9 kcal mol (Fig. 4). Formation of the lies ca. 5 kcal mol higher in energy. The resultant inter- −1 radical intermediate (RI-1a) is exoergic by 11.9 kcal mol mediate (RI-1b) is a radical with the unpaired electron (β Fig. 3 Optimised structures for TS-1a (a) and TS-1b (b). Distances are given in Å and spin populations larger than 0.1 are given in italics. Fig- ure rendered with PyMOL [47] 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 4 Reaction energy profile for AsqJ-catalysed desatura- tion and plausible hydroxyla- tion. The energy values, i.e., Gibbs free energy (ΔG), relative potential energy calculated at the ONIOM(B3LYP-D3/ def2-TZVP, Amber) level with electronic embedding and the relative potential energy of the QM part obtained with the B3LYP-D3/def2-TZVP method −1 are given in kcal mol −1 spin) delocalised over the C10 atom (− 0.68) and the anisyl to 22.4  kcal  mol and its height is most likely caused ring (total spin population on the ring is − 0.27). The spin by the excitation required to enter the π channel with delocalisation most likely contributes to the stability of the S(Fe(III)) = 5/2. −1 intermediate, which lies 16.9 kcal mol lower in energy To investigate the effect of protein environment for the than the initial reactant complex S. HAT barriers, the amino acid residues with atoms within 5 Å of the substrate and 2OG molecule were consecutively Regioselectivity of HAT replaced by alanine (alanine screening; for details of the pro- cedure, see Computational methods). For such mutants, the The regioselectivity of HAT can be analysed with the use ONIOM energy of S, TS-1a, and TS-1b was recalculated of Marcus theory [50] to obtain intrinsic barriers, i.e., in a single-point manner. As shown in Fig. 5b, the position- ones unaffected by thermodynamic contributions. Such ing of the substrate in optimised structure of TS-1b allows an approach was successfully employed by Srnec et al. to for favourable (mostly) interactions with hydrophobic resi- analyse the reaction selectivity of SyrB2 [16]. The relation dues lining the cavity. Comparison of optimised structures between the observed ΔG barrier and the intrinsic ΔG is of TS-1a and TS-1b reveals that, in TS-1a, the substrate int given by the equation: moves slightly away from the roof (upper part; as shown in Fig. 5a) of the binding pocket, which weakens interactions ΔG ΔG 0 0 with hydrophobic residues (for details, see Figs. S5, S6, and ΔG =ΔG + + , int 2 Table S2), but this movement is necessary to facilitate HAT 16ΔG int at the C3 position. Interestingly, TS-1a, contrary to TS-1b, is stabilised by hydrogen bonds between the ligands of the where ΔG is the reaction energy, ΔG and ΔG are known 0 0 metal cofactor (Asp-136, Fe-bound carboxyl group of succi- from the reaction-free energy profile. Thus, calculated nate) or substrate with nearby polar residues (Asn-157, Gln- intrinsic barrier associated with HAT from C3 (TS-1a; −1 −1 131, and Asn-70). This effect, however, does not compensate 19.4 kcal mol ) is by 1.6 kcal mol lower than the one −1 for weakened van der Waals interactions. for HAT from C10 (TS-1b; 21.0 kcal mol ). These values Additional computations employed a cluster model con- indicate that the observed reaction preference is driven by sisting of the first coordination shell of Fe and the substrate. higher thermodynamic driving force for the formation of The results obtained with this minimal QM model indi- RI-1b as compared to RI-1a. cate that the position of the substrate as observed in TS-1b Moreover, the preference for TS-1b over TS-1a can be results solely from its interactions with the binding pocket observed only when interactions with the binding pocket of the protein. During optimisation of the cluster variant of are included in the computational model. The picture is TS-1b, the substrate migrated from its initial position and different when the QM energies of these transition struc- formed a nearly linear Fe–O–H angle. The activation energy tures are analysed; in this case, the preferred process is −1 −1 of such a TS is 12.7 kcal mol and it is by 2.7 kcal mol HAT at the C3 position with electronic energy barrier −1 higher than the barrier associated with a QM-cluster variant equal to 11.2 kcal mol (see Fig. 4). The electronic energy of TS-1a (see Fig. S7). Therefore, it is the protein environ- barrier associated with TS-1b is relatively high, it totals ment, not electronic properties of the reactants, that governs 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 5 Binding cavity of AsqJ for TS-1a (a) and TS-1b (b). The amino acids lining the cavity are coloured from blue (stabilising the transition state) to red (destabilising the transi- tion state) the regioselectivity of HAT, which results in preferential to be operational when HAT is unfeasible for geometrical formation of RI-1b. reasons (as in the case of the C3-epimer). The TS-2b and TS-2b are both the early transitions H OH Desaturation vs. hydroxylation structures. The Fe(III)–OH distance is elongated from 1.86 Å (in RI-1b) to 1.96 Å in TS-2b , whereas it remains almost After the radical intermediate RI-1b is formed, the reac- unchanged (1.87 Å) in TS-2b . Spin density population OH tion can follow three scenarios. The C3-bound hydrogen can analysis indicates that five unpaired α electrons are located be abstracted from the intermediate to form a desaturated on Fe(III) (spin population of 4.17 and 4.13 for TS-2b and product P , which is the experimentally observed product TS-2b , respectively) and the β electron is delocalised over H OH [6]. The radical can also donate an unpaired electron to the C10 and the anisyl ring (see Fig. 6). Fe(III) site and form a carbocation, which is later deproto- The preference for desaturation over hydroxylation is nated yielding the same P . Finally, Fe-bound hydroxide most likely rooted in electronic properties of the substrate/ can recombine with the radical resulting in a hydroxylated metal cofactor pair. The electronic energy barriers calculated product P-b . for the QM part of the system (ΔE ) show that desaturation OH QM In the optimised structure of RI-1b, the distance is a more feasible process, which is consistent with ONIOM between the C3-bound hydrogen and the oxygen atom of ΔG results. The difference between the two barriers is even −1 −1 the Fe(III)–OH moiety is 1.97 Å and the Fe–O–H angle is larger, it totals to 10.5 kcal mol (cf. 1.4 kcal mol for 130°, whereas C10 atom, that hosts the unpaired electron, is ONIOM ΔG). This intrinsic preference for desaturation can located 3.48 Å away from the oxygen atom. Such a geometry be attributed to difference in energy gaps between orbitals is predisposed to desaturation, yet the hydroxylation cannot that mix in the transition states, i.e., β-dπ*(Fe–OH) (accep- be ruled out. tor for an electron in both reactions) and the β-p(C10 ) for The computed Gibbs free energy barriers indicate that hydroxylation or β-π*(C10–C3) for desaturation, which are HAT is the most favourable process (see Fig. 4). The transi- the electron source in hydroxylation and desaturation reac- tion structure associated with the second HAT (TS-2b ) lies tions, respectively. To estimate the orbital energy, we per- −1 1.4 kcal mol lower than TS for OH rebound (TS-2b ). formed single-point calculations separately for the Fe(III) OH Despite numerous attempts (QM/MM calculations, com- site and 4′-methoxycyclopeptin radical, their geometries bined also with TD-DFT and calculations done for cluster being the same as in transition structures (TS-2b and TS- models with fragmental guess), so far we have not managed 2b ) and identified orbitals most similar to ones mixing OH to optimise a species with substrate derived carbocation and in the respective transition state structures. As shown in Fe(II)–OH cofactor. Fig. 7, the energy gap between fragment orbitals that mix −1 The recent experimental study of AsqJ desaturation activ- in TS-2b totals to 60.9 kcal mol , whereas the gap for −1 ity showed that the C3-epimer of cyclopeptin is converted to TS-2b is by 6.3 kcal mol higher. The total fragments’ OH the desaturated product, and thus, the hydroxylated species distortion energies computed for these two TS are 12.0 and −1 P-b or a carbocation was suggested as intermediates [8]. 1.5 kcal mol for TS-2b and TS-2b , respectively. The OH H OH Our results show that, for the native substrate, the formation preference for desaturation might also be supported by delo- of P-b is less favourable than HAT; however, the barrier calisation of the radical over the anisyl ring, which usually OH for hydroxylation is low enough for hydroxylation pathway hinders radical rebound and hence supports entering the alternative reaction channel [11]. 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 6 Optimised structures for TS-2b (a) and TS-2b (b). Distances are given in Å and spin populations larger than 0.1 are given in italics H OH Fig. 7 Orbitals mixing in TS-2b (a) and TS-2b (b). Figure rendered using VMD [51] H OH Notably, the energy difference (ΔG and ΔE) between radical intermediate accompanied by stabilising interactions TS-2b and TS-2b decreases when interactions with the between the bicyclic ring and the cavity (Fig. 8b). Neverthe- H OH protein are taken into account; comparison of ΔG or ΔE less, the stabilisation effect of TS-2b does not compensate OH to ΔE (see Fig. 4) reveals that TS-2b is significantly for the higher electronic energy barrier for hydroxylation. QM OH stabilised by the cavity, whereas TS-2b is slightly destabi- That is why, desaturation is the experimentally observed lised. To analyse it further, we performed alanine screening process. It is possible that replacing some small residues for these two elementary steps. The results (presented in lining the cavity (e.g., Val-72 or Thr-229, see Table S3 and Fig. 8a, b) show that in, TS-2b , the substrate takes a posi- Fig. S8A, B) with more bulky ones might provide additional tion that facilitates HAT from the C3 atom at the expense stabilisation for TS-2b and, consequently, change the OH of weakening its interactions with the cavity (Fig. 8a). On reaction selectivity of AsqJ. the other hand, formation of the bond between C10 and The formation of P , as well as P-b , is an exoergic H OH OH in TS-2b requires only very modest shift of the process. The free energy of desaturation is comparable to the OH 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 8 Binding cavity of AsqJ for TS-2b (a), TS-2b (b), H OH TS-2a (c), and TS-2a (d). H OH The amino acids lining the cavity are coloured from blue (stabilising the transition state) to red (destabilising the transi- tion state) free energy of hydroxylation at the C10 position, whereas the latter process. The preference for hydroxylation is even energies of the QM system indicate that desaturation is more more pronounced in the results obtained with the cluster exothermic (see Fig. 4). The thermodynamic cycles for the model, where OH rebound occurs without a barrier (Fig. hydroxylation and desaturation show that the OH rebound S7). Analysis of the energy gap between orbitals that mix −1 · is by 4.7 kcal mol more favourable than hydrogen atom in the transition structure [β-dπ*(Fe–OH) and the β-p(C3 ) abstraction at the C3 position by the OH radical. However, for hydroxylation or β-π*(C3–C10) for desaturation] shows this is compensated by the subsequent binding of the water that the gap between orbitals mixing in TS-2a totals −1 molecule (formed in the desaturation process) to the active to 70.2  kcal  mol , and for TS-2a the gap is only by OH, −1 site, which results in the stabilisation of the system by 3 kcal mol higher (see Fig. S12), which indicates a smaller −1 11.4 kcal mol (see Figs. S9, S10). preference towards desaturation than in the TS-2b /TS- −1 The hydroxylated product P-b adapts such a position 2b pair (6 kcal mol ). The total distortion energy for OH OH −1 within the binding cavity that it is additionally stabilised by TS-2a is 10.4 kcal mol , which is slightly lower than the −1 the protein environment, which results in comparable free one calculated for TS-2b (12.0 kcal mol ), whereas the energy of desaturation and hydroxylation. total distortion energy for TS-2a is larger than for TS- OH −1 −1 2b , and it totals to 4.8 kcal mol (cf. 1.5 kcal mol for OH Back to path A: final view on the role of the protein TS-2b , shown in Table 1). Moreover, in the electronic OH in reaction selectivity structure of RI-1a (calculated for radical together with the active site), the energy gap between the β-p(C3 ) and the −1 The less favoured radical intermediate, RI-1a, when formed, lower in energy β-σ(C10–H) totals to 135.2  kcal  mol , can also undergo a subsequent hydrogen atom abstraction which is a larger value than the one calculated for β-p(C10 ) −1 (proceeding via TS-2a , shown in Fig. S11A) or hydroxyla- and β-σ(C3–H) of RI-1b (131.0 kcal mol , as shown in tion (via TS-2a , Fig. S11B). The electronic energy bar- Fig. 9). Therefore, the inherent preference for OH rebound OH −1 rier ΔE for the former is by 3 kcal mol higher than for QM 1 3 JBIC Journal of Biological Inorganic Chemistry Table 1 Values of distortion energy obtained for Fe(III) site and involves the formation of a bond between C10 and the oxo 4′-methoxycyclopeptin radical in the same geometries as in transition ligand of the oxoferryl species that yields a C3-centered rad- structures associated with desaturation/hydroxylation ical [10]. Comparison of these results with the ones reported −1 Path A (kcal mol ) Path B here for desaturation shows that both AsqJ-catalysed reac- −1 (kcal mol ) tions are initiated at the C10 position. In case of epoxidation, the reported QM/MM barrier for attack at C10 is slightly Desaturation (TS-2a/b ) 10.4 12.0 lower that the one calculated at the QM level, which indi- Hydroxylation (TS-2a/b ) 4.8 kcal 1.5 OH cates an additional stabilisation of the transition structure by the binding cavity, similar to TS-1b and TS-2b reported OH after formation of RI-1a can also stem from the increased here. stability of the σ orbital with respect to the β-p(C3 ). Interactions with the binding cavity of the protein change Nonenzymatic rearrangement the selectivity of the reaction via destabilisation of TS-2a . OH The results of alanine screening (TS-2a vs RI-1a) show The final stage of the reaction is a nonenzymatic rearrange- OH again that shortening the distance between the C3 atom and ment that is supposed to take place in the solvent outside of the iron site results in such a re-positioning of the interme- the enzyme active site. It is a two-step process, which, via diate that its favourable van der Waals interactions with the elimination of methyl isocyanate, results in formation of the binding cavity are weakened (see Fig. 8c, d). keto form of 4′-methoxyviridicatin (Fig. 10). This observation together with analysis presented above The reaction starts with formation of a bridged tricy- for the pathway B leads to a general conclusion that due to a clic intermediate (BI). In water modelled with polarizable slight change in the position of a substrate/radical interme- continuum model, this step requires crossing a high barrier −1 −1 diate transition structures for breaking the C3-H bond (TS- of 30.5 kcal mol . The barrier is lowered by 5 kcal mol 1a and TS-2b ) or formation of the C3-O bond (TS-2a ) in the presence of two water molecules forming hydrogen H OH feature weaker contacts with the amino acid residues lining bonds with oxirane fragment and the NH group of the ben- the cavity than the preceding intermediates, which results in zodiazepinedione moiety (see Fig. S13). increase of the free energy barriers. The effect of acid/base catalysis for the rearrangement The desaturated intermediate P undergoes a subsequent was investigated with the use of models consisting of the (AsqJ-catalysed) epoxidation. The process was investigated epoxide and acetic acid/acetate or ascorbic acid/ascorbate by QM/MM calculations and the proposed mechanism serving as acid H–B and base B . The general observation Fig. 9 Orbitals in RI-1a (a) and RI-1b (b) that donate and accept the electron in desaturation and hydroxylation steps 1 3 JBIC Journal of Biological Inorganic Chemistry Fig. 10 Reaction scheme for the nonenzymatic rearrangement favourable C10 keto group. In systems involving ascorbic −1 acid, the elimination goes over a barrier of ca. 6 kcal mol and imaginary frequency associated with this transition state reveals that the proton is transferred back to the ascorbate anion. Conclusions The proposed mechanism for AsqJ-catalysed desatura- tion involves HAT at the C10 position followed by second hydrogen abstraction from the neighbouring C3 atom. The regioselectivity of the first HAT stems from favourable inter - actions with the hydrophobic residues lining the binding Fig. 11 Reaction profiles for rearrangement in presence of implicit cavity, whereas product selectivity is dictated by electronic water (black line), ascorbate as B (blue), ascorbic acid as H–B properties of the reactants. However, the electronic prefer- (green), and both ascorbate and ascorbic acid (magenta) ence for desaturation is partially reduced by stabilising effect of the binding pocket that lowers the barrier for the hydroxy- is that, to a very good extent, the barrier lowering effects lation reaction. This observation opens up new possibilities are additive, and the acid molecule, which donates a pro- for switching reaction selectivity of AsqJ by introducing ton to the epoxide oxygen, is responsible for most of the mutations within the binding pocket. catalytic effect (see Figs.  11, S14). The ascorbic acid lowers −1 the barrier by 17.7 kcal mol , whereas the acetic acid by Acknowledgements This research project was supported by Grant −1 8.8 kcal mol . This difference can be partly attributed to no. UMO-2014/14/E/NZ1/00053 from the National Science Centre, Poland, and by PL-Grid Infrastructure. Calculations were performed larger pK of acetic acid (4.7 as compared to 4.2 for ascorbic at the Academic Computer Centre Cyfronet AGH. acid). Visual inspection of the imaginary frequency normal mode shows that, along with the C–C bond formation, pro- Open Access This article is distributed under the terms of the Crea- ton migrates from the donor to the epoxide oxygen, thus tive Commons Attribution 4.0 International License (http://creat iveco forming a hydroxyl group.mmons.or g/licenses/b y/4.0/), which permits unrestricted use, distribu- tion, and reproduction in any medium, provided you give appropriate Regardless of the presence of base molecule, that can credit to the original author(s) and the source, provide a link to the accept proton from the NH group, the N–H bond remains Creative Commons license, and indicate if changes were made. intact, and thus, the resulting bridged bicyclic intermediate has a cationic character. However, the presence of a nega- tively charged entity interacting with the cationic moiety References lowers the Gibbs free energy of the BI intermediate by ca. 2 −1 (ascorbate) − 5 (acetate) kcal mol . 1. Martinez S, Hausinger RP (2015) Catalytic mechanisms of The final step of the reaction, elimination of methyl iso- Fe(II)- and 2-oxoglutarate-dependent oxygenases. J Biol Chem cyanate, in the absence of acid molecule occurs without any 290:20702–20711 energy barrier as it leads to the formation of energetically 1 3 JBIC Journal of Biological Inorganic Chemistry 2. Abu-Omar MM, Loaiza A, Hontzeas N (2005) Reaction mecha- 20. Jorgensen WL, Chandrasekhar J, Madura JD et al (1983) Com- nisms of mononuclear non-heme iron oxygenases. Chem Rev parison of simple potential functions for simulating liquid water. 105:2227–2252 J Chem Phys 79:926–935 3. Price JC, Barr EW, Glass TE et al (2003) Evidence for hydrogen 21. Case DA, Cheatham TE, Darden T et al (2005) The Amber bio- abstraction from C1 of taurine by the high-spin Fe(IV) intermedi- molecular simulation programs. J Comput Chem 26:1668–1688 ate detected during oxygen activation by taurine: α-ketoglutarate 22. Becke AD (1993) Density-functional thermochemistry. III. The dioxygenase (TauD). J Am Chem Soc 125:13008–13009 role of exact exchange. J Chem Phys 98:5648–5652 4. Hegg EL, Que L (1997) The 2-His-1-carboxylate facial triad– 23. Grimme S, Ehrlich S, Goerigk L (2011) Effect of the damping an emerging structural motif in mononuclear non-heme iron(II) function in dispersion corrected density functional theory. J Com- enzymes. Eur J Biochem 250:625–629 put Chem 32:1456–1465 5. Hausinger RP (2004) Fe(II)/α-ketoglutarate-dependent hydroxy- 24. Hay PJ, Wadt WR (1985) Ab initio effective core potentials for lases and related enzymes. Crit Rev Biochem Mol Biol 39:21–68 molecular calculations. Potentials for K to Au including the out- 6. Ishikawa N, Tanaka H, Koyama F et al (2014) Non-heme dioxy- ermost core orbitals. J Chem Phys 82:299–310 genase catalyzes atypical oxidations of 6,7-bicyclic systems to 25. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, form the 6,6-quinolone core of viridicatin-type fungal alkaloids. Cheeseman JR, Scalmani G, Barone V, Petersson GA, Nakat- Angew Chemie Int Ed 53:12880–12884 suji H, Li X, Caricato M, Marenich A, Bloino J, Janesko BG, 7. Bräuer A, Beck P, Hintermann L, Groll M (2016) Structure of Gomperts R, Mennucci B, Hratchian HP, Ortiz JV, Izmaylov AF, the dioxygenase AsqJ: mechanistic Insights into a one-pot mul- Sonnenberg JL, Williams-Young D, Ding F, Lipparini F, Egidi tistep quinolone antibiotic biosynthesis. Angew Chemie Int Ed F, Goings J, Peng B, Petrone A, Henderson T, Ranasinghe D, 55:422–426 Zakrzewski VG, Gao J, Rega N, Zheng G, Liang W, Hada M, 8. Liao H-J, Li J, Huang J-L et al (2018) Insights into the desatura- Ehara M, Toyota K, Fukuda R, Hasegawa J, Ishida M, Nakajima tion of cyclopeptin and its C3 epimer catalyzed by a non-heme T, Honda Y, Kitao O, Nakai H, Vreven T, Throssell K, Montgom- iron enzyme: structural characterization and mechanism elucida- ery JA, Peralta JE Jr, Ogliaro F, Bearpark M, Heyd JJ, Brothers tion. Angew Chemie Int Ed 57:1831–1835 E, Kudin KN, Staroverov VN, Keith T, Kobayashi R, Normand 9. Su H, Sheng X, Zhu W et al (2017) Mechanistic Insights into the J, Raghavachari K, Rendell A, Burant JC, Iyengar SS, Tomasi J, decoupled desaturation and epoxidation catalyzed by dioxygenase Cossi M, Millam JM, Klene M, Adamo C, Cammi R, Ochterski AsqJ involved in the biosynthesis of quinolone alkaloids. ACS JW, Martin RL, Morokuma K, Farkas O, Foresman JB, Fox DJ Catal 7:5534–5543 (2013) Gaussian 09, revision D.01. Gaussian Inc., Wallingford 10. Song X, Lu J, Lai W (2017) Mechanistic insights into dioxygen 26. Duan Y, Wu C, Chowdhury S et al (2003) A point-charge force activation, oxygen atom exchange and substrate epoxidation by field for molecular mechanics simulations of proteins based on AsqJ dioxygenase from quantum mechanical/molecular mechani- condensed-phase quantum mechanical calculations. J Comput cal calculations. Phys Chem Chem Phys 19:20188–20197 Chem 24:1999–2012 11. Usharani D, Janardanan D, Shaik S (2011) Does the TauD 27. Wang J, Cieplak P, Kollman PA, Kollman PA (2000) How well enzyme always hydroxylate alkanes, while an analogous syn- does a restrained electrostatic potential (RESP) model perform thetic non-heme reagent always desaturates them? J Am Chem in calculating conformational energies of organic and biological Soc 133:176–179 molecules? J Comput Chem J Comput Chem 21:1049–1074 12. Hull JF, Balcells D, Sauer ELO et al (2010) Manganese catalysts 28. Kumar D, Thiel W, de Visser SP (2011) Theoretical study on the for C–H activation: an experimental/theoretical study identi- mechanism of the oxygen activation process in cysteine dioxyge- fies the stereoelectronic factor that controls the switch between nase enzymes. J Am Chem Soc 133:3869–3882 hydroxylation and desaturation pathways. J Am Chem Soc 29. Wang B, Cao Z, Sharon DA, Shaik S (2015) Computations reveal 132:7605–7616 a rich mechanistic variation of demethylation of N-methylated 13. Cooper HLR, Mishra G, Huang X et al (2012) Parallel and com- DNA/RNA nucleotides by FTO. ACS Catal 5:7077–7090 petitive pathways for substrate desaturation, hydroxylation, and 30. Wang X, Su H, Liu Y (2017) Insights into the unprecedented radical rearrangement by the non-heme diiron hydroxylase AlkB. epoxidation mechanism of fumitremorgin B endoperoxidase J Am Chem Soc 134:20365–20375 (FtmOx1) from Aspergillus fumigatus by QM/MM calculations. 14. Matthews ML, Neumann CS, Miles LA et al (2009) Substrate Phys Chem Chem Phys 19:7668–7677 positioning controls the partition between halogenation and 31. Miyamoto S, Kollman PA (1992) Settle: an analytical version of hydroxylation in the aliphatic halogenase, SyrB2. Proc Natl Acad the SHAKE and RATTLE algorithm for rigid water models. J Sci USA 106:17723–17728 Comput Chem 13:952–962 15. Huang J, Li C, Wang B et al (2016) Selective chlorination of 32. Essmann U, Perera L, Berkowitz ML et al (1995) A smooth par- substrates by the halogenase SyrB2 is controlled by the protein ticle mesh Ewald method. J Chem Phys 103:8577–8593 according to a combined quantum mechanics/molecular mechan- 33. Roe DR, Cheatham TE (2013) PTRAJ and CPPTRAJ: software ics and molecular dynamics study. ACS Catal 6:2694–2704 for processing and analysis of molecular dynamics trajectory data. 16. Srnec M, Solomon EI (2017) Frontier molecular orbital contribu- J Chem Theory Comput 9:3084–3095 tions to chlorination versus hydroxylation selectivity in the non- 34. Cornell WD, Cieplak P, Bayly CI et al (1995) A second genera- heme iron halogenase SyrB2. J Am Chem Soc 139:2396–2407 tion force field for the simulation of proteins, nucleic acids, and 17. Ji L, Faponle AS, Quesne MG et al (2015) Drug metabolism by organic molecules. J Am Chem Soc 117:5179–5197 cytochrome P450 enzymes: what distinguishes the pathways lead- 35. Feller D (1996) The role of databases in support of computational ing to substrate hydroxylation over desaturation? Chem A Eur J chemistry calculations. J Comput Chem 17:1571–1586 21:9083–9092 36. Weigend F, Ahlrichs R (2005) Balanced basis sets of split valence, 18. Marvin was used for drawing, displaying and characterizing chem- triple zeta valence and quadruple zeta valence quality for H to ical structures, substructures and reactions. Marvin 17.28.0, 2017, Rn: design and assessment of accuracy. Phys Chem Chem Phys ChemAxon. http://www.chema xon.com 7:3297–3305 19. Wang J, Wang W, Kollman PA, Case DA (2006) Automatic atom 37. Schuchardt KL, Didier BT, Elsethagen T et al (2007) Basis set exchange: a community database for computational sciences. J type and bond type perception in molecular mechanical calcula- Chem Inf Model 47:1045–1052 tions. J Mol Graph Model 25:247–260 1 3 JBIC Journal of Biological Inorganic Chemistry 38. Frisch MJ, Trucks GW, Schlegel HB, Scuseria GE, Robb MA, molecular-orbital studies of organic molecules. J Chem Phys Cheeseman JR, Scalmani G, Barone V, Petersson GA, Nakat- 54:724–728 suji H, Li X, Caricato M, Marenich AV, Bloino J, Janesko BG, 43. Krishnan R, Binkley JS, Seeger R, Pople JA (1980) Self-consistent Gomperts R, Mennucci B, Hratchian HP, Ortiz JV, Izmaylov AF, molecular orbital methods. XX. A basis set for correlated wave Sonnenberg JL, Williams-Young D, Ding F, Lipparini F, Egidi functions. J Chem Phys 72:650–654 F, Goings J, Peng B, Petrone A, Henderson T, Ranasinghe D, 44. Bernasconi L, Baerends EJ (2008) The EDTA complex of Zakrzewski VG, Gao J, Rega N, Zheng G, Liang W, Hada M, oxidoiron(IV) as realisation of an optimal ligand environment for Ehara M, Toyota K, Fukuda R, Hasegawa J, Ishida M, Nakajima high activity of FeO +. Eur J Inorg Chem 2008:1672–1681 T, Honda Y, Kitao O, Nakai H, Vreven T, Throssell K, Montgom- 45. Geng C, Ye S, Neese F (2010) Analysis of reaction channels ery JA, Peralta JE Jr, Ogliaro F, Bearpark MJ, Heyd JJ, Brothers for alkane hydroxylation by nonheme iron(IV)-oxo complexes. EN, Kudin KN, Staroverov VN, Keith TA, Kobayashi R, Normand Angew Chemie Int Ed 49:5717–5720 J, Raghavachari K, Rendell AP, Burant JC, Iyengar SS, Tomasi J, 46. Janardanan D, Usharani D, Chen H, Shaik S (2011) Modeling Cossi M, Millam JM, Klene M, Adamo C, Cammi R, Ochterski C–H abstraction reactivity of nonheme Fe(IV)O oxidants with JW, Martin RL, Morokuma K, Farkas O, Foresman JB, Fox DJ alkanes: what role do counter ions play? J Phys Chem Lett (2016) Gaussian 16, Revision A.03. Gaussian Inc., Wallingford 2:2610–2617 39. Miertus S, Scrocco E, Tomasi J (1982) Approximate evaluations 47. Schrödinger L (2015) The PyMOL molecular graphics system. of the electrostatic free energy and internal energy changes in Version 1:8 solution processes. Chem Phys 65:239–245 48. Ye S, Geng C-Y, Shaik S, Neese F (2013) Electronic structure 40. Groom CR, Bruno IJ, Lightfoot MP, Ward SC (2016) The Cam- analysis of multistate reactivity in transition metal catalyzed reac- bridge structural database. Acta Crystallogr Sect B Struct Sci tions: the case of C–H bond activation by non-heme iron(iv)–oxo Cryst Eng Mater 72:171–179 cores. Phys Chem Chem Phys 15:8017–8030 41. Li J, Wang J, Jiang C-S et al (2014) (+)-Cyclopenol, a new natu- 49. Srnec M, Wong SD, England J et al (2012) Frontier molecular rally occurring 7-membered 2,5-dioxopiperazine alkaloid from orbitals in S = 2 ferryl species and elucidation of their contribu- the fungus Penicillium sclerotiorum endogenous with the Chi- tions to reactivity. Proc Natl Acad Sci 109:14326–14331 nese mangrove Bruguiera gymnorrhiza. J Asian Nat Prod Res 50. Marcus RA, Sutin N (1985) Electron transfers in chemistry and 16:542–548 biology. Biochim Biophys Acta Rev Bioenerg 811:265–322 42. Ditchfield R, Hehre WJ, Pople JA (1971) Self-consistent molec- 51. Humphrey W, Dalke A, Schulten K (1996) VMD: visual molecu- ular-orbital methods. IX. An extended Gaussian-type basis for lar dynamics. J Mol Graph 14:33–38 Affiliations 1 1 Zuzanna Wojdyla  · Tomasz Borowski * Zuzanna Wojdyla Jerzy Haber Institute of Catalysis and Surface Chemistry, ncwojdyl@cyf-kr.edu.pl Polish Academy of Sciences, Niezapominajek 8, 30239 Kraków, Poland * Tomasz Borowski ncborows@cyf-kr.edu.pl 1 3

Journal

JBIC Journal of Biological Inorganic ChemistrySpringer Journals

Published: Jun 6, 2018

References

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off