An Active Site Tyr Residue Guides the Regioselectivity of Lysine Hydroxylation by Nonheme Iron Lysine-4-hydroxylase Enzymes through Proton-Coupled Electron Transfer

Lysine dioxygenase (KDO) is an important enzyme in human physiology involved in bioprocesses that trigger collagen cross-linking and blood pressure control. There are several KDOs in nature; however, little is known about the factors that govern the regio- and stereoselectivity of these enzymes. To understand how KDOs can selectively hydroxylate their substrate, we did a comprehensive computational study into the mechanisms and features of 4-lysine dioxygenase. In particular, we selected a snapshot from the MD simulation on KDO5 and created large QM cluster models (A, B, and C) containing 297, 312, and 407 atoms, respectively. The largest model predicts regioselectivity that matches experimental observation with rate-determining hydrogen atom abstraction from the C4–H position, followed by fast OH rebound to form 4-hydroxylysine products. The calculations show that in model C, the dipole moment is positioned along the C4–H bond of the substrate and, therefore, the electrostatic and electric field perturbations of the protein assist the enzyme in creating C4–H hydroxylation selectivity. Furthermore, an active site Tyr233 residue is identified that reacts through proton-coupled electron transfer akin to the axial Trp residue in cytochrome c peroxidase. Thus, upon formation of the iron(IV)-oxo species in the catalytic cycle, the Tyr233 phenol loses a proton to the nearby Asp179 residue, while at the same time, an electron is transferred to the iron to create an iron(III)-oxo active species. This charged tyrosyl residue directs the dipole moment along the C4–H bond of the substrate and guides the selectivity to the C4-hydroxylation of the substrate.


■ INTRODUCTION
−3 However, high expression of lysine hydroxylases may lead to cancer development. 4,5−9 Thus, some lysine hydroxylases bind a collagen or protein chain on the interface of an enzymatic dimer and hydroxylate individual lysine residues of the chain, while others operate on isolated lysine amino acids. 10As a consequence, these lysine hydroxylases show major differences in substrate activation and product distributions.
One class of lysine hydroxylases is the nonheme iron/αketoglutarate (αKG, also called 2-oxoglutarate)-dependent lysine dioxygenases (KDOs).−21 They all have a central iron(II) atom that is bound to the protein through interactions with the side chains of two histidine and one carboxylate-based, e.g., Asp or Glu, residue.Indeed, the structure of KDO5 displayed in Figure 1 shows the facial 2-His/1-Glu binding orientation of iron(II), where the metal is covalently bound to His 176 , Glu 178 , and His 312 . 22,23The cosubstrate αKG binds as a bidentate ligand trans to the His 176 and Glu 178 groups.Substrate Lys is missing from the Xray crystal structure, but the active site has a number of polar residues that are likely involved in substrate binding and positioning, namely, Asp 179 , Asp 260 , and Arg 338 .
There are various KDO isozymes labeled KDO1−KDO6 that all react with a free lysine substrate at a nonheme iron center.However, KDO1 gives S-3-hydroxylysine as a product, whereas the other isozymes react with a free lysine amino acid on the iron(II) center with αKG and dioxygen to form R-4-hydroxylysine regio-and stereoselectively.To understand this high regio-and stereoselectivity, we performed a detailed computational study of KDO5 enzymes.−21,24−26 For various nonheme iron dioxygenases, the iron(IV)-oxo species has been experimentally trapped and characterized with spectroscopic methods.−30 Despite the fact that the iron(IV)-oxo species of KDO5 has never been trapped and characterized, biochemical studies have been reported on its activity, substrate scope, and biochemistry. 31,32Moreover, engineered KDOs gave insight into substrate binding and orientation. 33o gain insight into the high selectivity of KDO5 enzymes, we decided to perform a computational study into lysine activation by KDO5 enzymes using the PDB 6EUR X-ray crystal structure as a starting point. 22,23The work shows that key hydrogen bonding and polar residues in the substrate binding pocket direct the local electric field and dipole moment and trigger a proton-coupled electron transfer from an active site Tyr residue to a neighboring Asp residue that guides the reaction to C 4 -hydroxylation even though it is not a thermodynamically favored pathway.
■ METHODS Model Setup.The model setup follows previous quantum cluster model calculations from our group and will be briefly summarized here. 34,35We started from the 6EUR PDB file, 22,23 which is a substrate-devoid but iron(II)-and α-ketoglutarate (αKG)-bound enzymatic tetramer of KDO.We selected chain A of the protein for our work and removed glycerol molecules from the structure.The missing loop in the protein from residues 230−240 was inserted using a homology match found with the UniProt software package for patching missing loops from the AlphaFold model AF-J3BZS6-F1. 36ext, substrate lysine was docked into the substrate binding pocket using the Autodock Vina software package, 37 and the most stable conformation was selected.Thereafter, hydrogen atoms were added in Chimera under pH 7 conditions, 38 whereby all Lys and Arg residues of the protein chain were in their protonated states, while all Asp and Glu residues were in their deprotonated state.Histidine residues were manually inspected, and all internal histidine residues were taken in their singly protonated states.The structure was solvated with TIP3P-  Molecular Dynamics Simulations.Parameters of the metal with its first-coordination sphere were obtained from the MCPB program, while the Amber force field was used on protein atoms. 39,40The enzyme structure was equilibrated for 1 ns and heated to 298 K over a period of 1 ns using the ff14SB force field. 41Thereafter, molecular dynamics simulation was performed for 50 ns in Amber without constraints under NVT conditions; see the Supporting Information for details. 40The MD simulation converged the protein to a stable structure, and an analysis of geometries during the MD simulation shows that the system is highly rigid and, particularly, the substrate is bound tightly within the substrate-binding pocket.Geometrically, the last point of the MD simulation shows only minor differences with respect to the original crystal structure coordinates as shown from the overlay of the two structures in Figure S4, Supporting Information.
Cluster Model Setup.After the MD simulation, the last step of the equilibrated structure was used to create several cluster models.−47 The cluster models were created based on the first-and second-coordination spheres of the iron and substrate, and they are shown in Figure 2. We initially did exploratory calculations using cluster model A of 297 atoms and an overall charge of −1.Model A included the iron(IV)-oxo(succinate) group, where we truncated succinate to acetate.The axial histidine residue (His 312 ) was shortened to methylimidazole, while the other protein ligands of the metal were part of the peptide chain His 176 -Thr 177 -Glu 178 -Asp 179 -Ala 180 -Phe 181 -Leu 182 with Thr 177 truncated to Gly.In addition, part of the substrate binding pocket was included in the model, namely, the chains Asp 230 -Ala 231 -Asn 232 -Tyr 233 and Arg 338 -Met 339 -Met 340 with the residues Ala 231 and Met 339 truncated to a glycine residue.The Gln 144 , His 152 , and Asp 260 side chains were also included in the model, as well as four water molecules.We further expanded the model to 312 atoms with the amino acid residue Arg 145 as it has been implicated in substrate recognition and may be involved in the assistance of pulling the substrate into the protein structure. 22Model B therefore has an overall neutral charge.Models A and B were calculated in the lowest energy singlet, triplet, quintet, and septet spin states.None of the models use constraints during the geometry optimizations.A comparison of the optimized geometries with crystal structure coordinates showed minor deviations with respect to the starting structures.
Finally, a larger model was created, namely, model C with 407 atoms, which included the iron(IV)-oxo(succinate) active site with succinate truncated to acetate.The environment was incorporated through the peptide chain from Leu 173 to Leu 182 and the protein chains Gln 144 -Arg 145 , His 152 -Ala 153 -Val 154 , Thr 165 -Gly 166 -Ser 167 , Asp 230 -Ala 231 -Asn 232 -Tyr 233 , and His 312 -Gly 313 -Arg 314 .The residues Phe 174 , Val 175 , Thr 177 , and Ala 231 were truncated to Gly residues.The model also included the side chain of Arg 338 and seven water molecules.Model C had an overall charge of +2 and was calculated in the quintet spin state only (see below).A final model C2 was explored, which includes model C expanded with the full succinate group and the guanidinium group of Arg 334 .
Computational Methods.The Gaussian-09 software package was used for all quantum chemical calculations discussed here. 48ollowing previous experience with cluster models of nonheme iron dioxygenases, 34,49 we utilized the unrestricted B3LYP density functional method in combination with LANL2DZ (with electron core potential) on iron and 6-31G on the rest of the atoms, denoted basis set BS1. 50−54 Dispersion corrections were included with the GD3 approach and Becke−Johnson damping factor. 55To correct the energetics, single point calculations with LACV3P+ (with electron core potential) on iron and 6-311+G* on the rest of the atoms (BS2) were performed.Frequency calculations were performed on all local minima and transition states, and it was confirmed that local minima had real frequencies only, while the transition states had a single imaginary mode for the expected vibration along the reaction coordinate.To test the effect of the basis set, we also did geometry optimizations with def2-SVP on all atoms followed by single point calculations with def2-TZVP.These calculations obtained results similar to those with basis sets BS1 and BS2.Test calculations using an implicit solvent model with a dielectric constant mimicking chlorobenzene were performed on key transition state structures for model C, which did not change the ordering of the transition states and predicted the same selectivity; see the Supporting Information.

■ RESULTS
System Setup and MD Simulations.We started the work with a set of molecular mechanics and molecular dynamics (MD) simulations on the protein structure based on chain A of the X-ray crystal structure of KDO5 solved with bound αKG (PDB: 6EUR).Initially, we docked substrate lysine into the active site; see the Supporting Information for details.The docking pose with the highest score was then selected for further setup and MD simulations.The root-meansquare deviation (RMSD) of the structure converges within a few nanoseconds of MD simulation as shown in Figure 3 and reaches a stable plateau, which applies to the RMSD of the protein, substrate, and ligand components.In general, the structure is highly rigid and little deviation between the crystal structure and the final structure of the MD simulation is seen.The only motion seen during the MD simulation relates to the position of the termini of the protein chain; however, these ends are far from the active site.We then analyzed the substrate positions from each individual snapshot during the MD simulations and measured distances between the substrate and active site amino acids, as defined in Tables S2 and S3 (Supporting Information).Thus, during the full MD simulation, the carboxylic acid group of Asp 179 is within 2 Å of the substrate ammonium terminus that holds the substrate in a tight orientation and position (Figure 3b).This is common in amino acid hydroxylases and seen before in, e.g., viomycin biosynthesis enzyme VioC. 22,23,56Another close amino acid residue from the substrate is the side chain of Arg 338 that is held within 2−3 Å of the substrate and forms a strong salt bridge interaction with its carboxylate group.Although the Asn 232 side chain points into the active site, its distance to the substrate is further than 3 Å, and therefore, it may not be an important residue for substrate positioning and catalysis.
As MD trajectories sometimes see major changes in the position of the substrate, 47 DFT Cluster Calculations on the Reactants.Based on the last snapshot of the MD simulation, we created three QM cluster models that describe the metal and substrate and their first-and second-coordination spheres, namely, model A (297 atoms, total charge of −1), model B (312 atoms, total charge of 0), and model C (407 atoms, total charge of +2).DFToptimized geometries of the iron(IV)-oxo species of KDO for models A, B, and C are shown in Figure 4.The quintet spin state is the ground state for all models, whereby for models A and B, the system has four unpaired electrons in metal-based molecular orbitals and configuration π* xy These orbitals represent interactions of the atomic 3d orbitals on the iron with first-coordination sphere atoms.In particular, the antibonding interactions of the atomic 3d xy , 3d xz , and 3d yz orbitals with 2p orbitals on the oxo group yield the molecular π* xy , π* xz , and π* yz orbitals.In addition, there is a singly occupied orbital (σ* x2−y2 ) for the σ-type antibonding interaction in the equatorial plane with the side chains of His 176 , Glu 178 , and succinate.The σ* z2 orbital is virtual and located along the N ax −Fe−O axis with N ax being the axial nitrogen atom from His 312 .−83 The alternative singlet and triplet spin states were also calculated but found to be higher in energy for model C by 22.3 and 5.3 kcal mol −1 , respectively.These large spin state energy gaps imply that KDO will react on a dominant quintet spin state surface through single-state reactivity.
Interestingly, in model C, electron transfer has taken place from an active site Tyr residue (Tyr 233 ) to the iron to create an iron(III)-oxo species and tyrosyl radical.Thus, Tyr 233 has donated a proton to the Asp 179 carboxylic acid group, and at the same time, an electron is transferred from the Tyr side chain to iron.As a consequence, the metal has five unpaired electrons, and there is considerable spin density found on Tyr 233 , where a down-spin electron is located with ρ = −0.97,while the Tyr residue shows no unpaired spin density in the iron(IV)-oxo complexes in models A and B. As such, in model C, the metal is in the oxidation state iron(III) with orbital occupation π* xy The reactant complex 5 Re C therefore can be described as [Fe III (O)---Tyr • ].We tried to swap molecular orbitals for 5 Re C and create an iron(IV)-oxo with a neutral Tyr residue; however, during geometry optimization, the proton moved back to the carboxylate of Asp 179 and the electron transferred to iron.The electronic configuration was also obtained with alternative computational approaches and protocols.Furthermore, expanding model C with the complete succinate and the Arg 334 side chain (model C2) did not change the structure dramatically and resulted in a similar electronic configuration with a radical on Tyr 233 , a proton on Asp 179 , and an iron(III)oxo group; see Figure 4. Therefore, in our large KDO5 model, the iron(IV)-oxo species is a fleeting intermediate that rapidly converged to an iron(III)-oxo species due to deprotonation of Tyr 233 by the Asp 179 residue and electron transfer.Interestingly, the Tyr-Asp couple does not seem conserved within the class of lysine dioxygenases; 84 hence, these enzymes may operate through a different mechanism.Unfortunately, there are no reports of the iron-oxo species in KDO5 being trapped and characterized by electron paramagnetic resonance (EPR) and/or Mossbauer spectroscopy.However, a ground state for [Fe III (O)---Tyr • ] should give it a distinct EPR and Mossbauer spectrum that is different from an [Fe IV (O)---Tyr] configuration.
In the heme enzyme cytochrome c peroxidase (CcP), similar valence tautomerism occurs and the active species is an iron(IV)-oxo(heme) with a nearby Trp radical, i.e., [Fe IV (O)-(heme)---Trp •+ ]. 85 DFT calculations showed this state to be close in energy to an electronic state with an iron(IV)-oxo heme cation radical and closed-shell Trp group, i.e., [Fe IV (O)-(heme •+ )---Trp], whereby external perturbations, e.g., a cation binding site, near the heme determined its ground state. 86,87−93 As such, valence tautomerism should also be possible in enzymatic systems despite the fact that there is no experimental evidence in the literature of iron(III)-oxo species.−98 The four optimized iron(IV)-oxo species have a very similar first-coordination sphere of the metal with its direct ligands particularly in the equatorial plane.The axial nitrogen atom of His 312 (N ax ) is at a distance of 2.279 Å from the iron atom in model C, while it is at 2.090 and 2.123 Å for models B and A, respectively.This is not surprising as 5 Re C has the σ* z2 orbital occupied with one electron that causes antibonding interactions along the O−Fe−N ax axis.As a result, also the Fe−O interaction is elongated to 1.784 Å for model C, while it is 1.631 Å for model B and 1.617 Å for model A. Therefore, the different models give a first-coordination sphere geometry of the active site with bond lengths that match the orbital occupation and configuration.
The lysine substrate is bound in the substrate binding pocket and held by a salt bridge of its carboxylate group with the side chain of Arg 145 , and it forms hydrogen bonding interactions with the hydroxyl groups of Thr 165 and Ser 167 .The terminal ammonium group of the substrate forms hydrogen bonding interactions with the carboxylic acid group of Asp 230 and the phenol group of Tyr 233 and two water molecules.A further hydrogen bonding network in the substrate and oxidant bonding pocket includes the side chains of Asp 179 , Asn 232 , and Arg 338 .As such, the substrate is tightly bound in the substrate binding pocket, which creates a highly rigid structure.These DFT calculations therefore support the results from the MD simulations that the substrate is tightly bound and highly rigid within the substrate binding pocket.
An overlay of the 5 Re C -optimized geometry with the X-ray crystal structure coordinates of the 6EUR PDB file is shown on the right-hand side of Figure 4, while the overlay with the last snapshot from the MD simulation is shown in Figure S4 (Supporting Information).The match between the two structures is very good, and most protein backbone groups are in the same position in both systems.Several residues on the edge of the cluster model are highlighted in Figure 4 and are close to their starting position obtained from the crystal structure coordinates.The overlay of the two structures therefore confirms that the protein structure is tight and rigid and shows little flexibility as a function of time.Moreover, our models reflect the enzyme structure well and are expected to give an accurate reflection of the real system.
Hydrogen Atom Abstraction from Lys. Next, we calculated the hydrogen atom abstraction pathways of substrate Lys on the C 3 −H, C 4 −H, and C 5 −H positions as Journal of the American Chemical Society described in Scheme S1 (Supporting Information).In previous work, we explored pro-R and pro-S hydrogen atom abstraction of taurine in taurine/αKG dioxygenase. 45As those calculations gave close energy barriers for pro-R and pro-S pathways, we considered one stereoisomer only here.In particular, we investigated hydrogen atom abstraction via transition state TS1 to form radical intermediate IM1.Thereafter, an OH rebound barrier via the transition state TS2 leads to alcohol products (Pr).The labeling system gives the spin multiplicity in superscript before the label and the position of the hydrogen on the substrate and the cluster model in subscript after the label.In all cases, hydrogen atom abstraction is ratedetermining and the OH rebound has a much lower barrier.Table 1 summarizes the calculated barrier heights for hydrogen atom abstraction from the C 3 −H, C 4 −H, and C 5 −H positions of the Lys substrate by the iron-oxo species of KDO for models A, B, and C.
In models A and B, the lowest free energy of activation for the hydrogen atom abstraction barrier is via 5 TS1 HA,C5 with magnitudes of ΔG ‡ = 13.1 kcal mol −1 (model A) and ΔG ‡ = 18.6 kcal mol −1 (model B).In model B, the other hydrogen atom abstraction barriers have a free energy that is >9 kcal mol −1 higher.Consequently, model B predicts selective hydrogen atom abstraction from the C 5 −H position, and little or no alternative products are predicted for this model.This is in contrast to experimental data on this KDO structure that measured dominant C 4 −H hydroxylation products. 6,32n model A, the margins between the hydrogen atom abstraction barriers are much smaller and the C 3 −H hydrogen atom abstraction is only ΔG = 1.5 kcal mol −1 above the one for C 5 −H, and therefore, some C 3 −H may be expected, although the C 4 −H barrier is still ΔG = 5.4 kcal mol −1 higher in free energy than the C 5 −H hydrogen atom abstraction barrier.−9 These models give the dominant C 5 -hydroxylation of Lys rather than the expected C 4hydroxylation.Models A and B differ from each other through the additional Arg residue in model B. The large rise in barrier heights in model B therefore is the result of the inclusion of this residue, which fixes the substrate in a tighter orientation that makes hydrogen atom abstraction from some positions more difficult.In particular, Arg 145 strongly destabilizes the pathway for C 3 −H hydrogen atom abstraction.As both models A and B predict the wrong products, these models do not accurately mimic KDO activity.Therefore, we explored further expanded model C, which is an even larger cluster model with an additional Arg residue in the second coordination sphere and several chains surrounding the substrate binding pocket.
In model C, we find the lowest energy transition state via 5 TS1 HA,C4 with a free energy of activation of ΔG ‡ = 24.8kcal mol −1 , while the barriers for C 3 −H and C 5 −H activation are 3.8 and 2.2 kcal mol −1 higher lying, respectively.Consequently, model C predicts dominant C 4 −H hydrogen atom abstraction, in agreement with experimental product distributions.Therefore, we will focus the rest of the paper on the model C results.Model C includes one additional Arg residue, namely, Arg 314 , and its electrostatic perturbations appear to destabilize the C 5 − H hydrogen atom abstraction pathway.As such, the two Arg residues (Arg 145 and Arg 314 ) are positioned so that they disfavor the hydrogen atom abstraction pathways from the C 3 −H and C 5 −H positions and enable efficient hydrogen atom abstraction from the C 4 −H position.These perturbations, however, do raise the barrier heights for all processes significantly but will succeed in a regioselective reaction pathway.
The full energy landscape for Lys hydroxylation at the C 4 −H and C 5 −H positions as calculated for model C is given in Figure 5.After the hydrogen atom abstraction, the structures relax to a radical intermediate; however, there is a small barrier for OH rebound, leading to alcohol products with large exothermicity.Consequently, we expect the radical intermedi- During the hydrogen atom abstraction step of the Lys hydroxylation mechanism, the Tyr group is reprotonated again and accepts an electron from substrate Lys so that we form an iron(III)-hydroxo group with a Lys radical.In all radical intermediates ( 5 IM1 C3 , 5 IM1 C4 , and 5 IM1 C5 ) for model C, there is a down-spin electron on the substrate and five up-spin unpaired electrons in metal-type orbitals.After the radical intermediates, facile OH rebound (with a negligible barrier) leads to alcohol product complexes with large exergonicity.Therefore, the radical intermediates will be short-lived and quickly convert into alcohol product complexes in an irreversible step.
Optimized geometries of 5 TS1 HA,C4 and 5 TS1 HA,C5 for model C are shown in Figure 5.In both structures, the Fe−O bond is long, probably due to occupation of the σ* z2 orbital with one electron.The 5 TS1 HA,C4 structure has the transferring hydrogen atom almost midway between the donor and the acceptor atom, i.e., the O−H distance is 1.277 Å and the C 4 −H distance is 1.261 Å.This is similar in 5 TS1 HA,C5 , where the transferring hydrogen atom is at an O−H distance of 1.320 Å, while the C 5 −H distance is at 1.259 Å.Therefore, the C 5 −H hydrogen atom abstraction structure is reactant-like, whereas the C 4 −H structure is neither reactant-nor product-like but intermediate.Overall, the structural differences between the two transition states are small and similar angles and first-coordination sphere metal−ligand interactions are found.−83,99−110 In both transition state structures, the angle between the substrate and the oxidant is similar.Thus, the Fe−O−C angle of the carbon atom that loses a hydrogen atom is 137°for 5 TS1 HA,C4 , while it is 130°f or 5 TS1 HA,C5 .Also, the O−H−C angles are similar at 166°a nd 169°, respectively.Therefore, the mechanistic and energetic differences observed here are not the result of structural differences in the transition states.

■ DISCUSSION
In this work, the regio-and chemoselective hydroxylation of a free Lys residue by KDO enzymes is explored.Three models were investigated, but only the largest model (model C) shows a preference for C 4 -hydroxylation products, while models A and B predict C 5 -hydroxylation.To understand the differences between the cluster models and, particularly, the hydrogen atom abstraction energetics, we analyzed the structure and electronic configuration of all species in detail.
Figure 6 shows the electron transfer pathways calculated for the hydrogen atom abstraction mechanisms for model C of KDO.As reported above, instead of a reactant that is iron(IV)oxo with a quintet spin ground state as is commonly observed f o r α K G -d e p e n d e n t n o n h e m e i r o n d i o x y g enases, 27,45,[61][62][63][64][65][66][67][68][69][70][71][72][73][74][75][76][77]111,112 we find an iron(III)-oxo reactant species.This is the result of proton-coupled electron transfer from Tyr 233 that has released a proton to the carboxylate group of Asp 179 and an electron donated into the iron set of orbitals. Thus, the arboxylate group of Asp 179 is within the hydrogen bonding distance (1.596 Å in 5 Re C ) from the phenol group of Tyr 233 , while on the other side of the Tyr residue, the terminal ammonia group of Lys is at a short distance in the other direction (1.95 Å).The terminal ammonia group of substrate Lys via a bridging water molecule connects to the oxo group of the iron(III)-oxo species through a hydrogen bonding network.We therefore anticipate that the mutation of Asp 179 by, e.g., Asn 179 will prevent the proton-coupled electron transfer from happening and presumably affects product distributions dramatically.
The reaction mechanism then continues with another proton-coupled electron transfer, whereby the iron(III)-oxo species abstracts the proton from the C 4 -position.At the same time, an electron is transferred from substrate Lys to the tyrosyl radical of Tyr 233 , which results in proton transfer from Asp 179 to Tyr 233 .In the radical intermediate, therefore, the metal stays in the iron(III) oxidation state, but the radical disappears from the Tyr 233 residue and accumulates on the substrate.Indeed, the group spin densities on iron are 4.1 in 5 Re C and 4.2 in 5 IM1 C .On the other hand, a spin of −0.97 is found on the protein (mostly on Tyr 233 ) in 5 Re C , while no spin on the substrate is observed.In the radical intermediate, the spin on the substrate is −0.95, while the spin on Tyr has disappeared.Recent combined experimental and computational studies on a biomimetic iron(III)-oxo species, as compared to its iron(IV)-oxo analog, showed that both oxidation states react with toluene with similar hydrogen atom abstraction barriers. 75However, iron(III)-oxo by itself can only react as a one-electron oxidant and is not able to perform OH rebound reactions.In the KDO system discussed here, iron(III)-oxo has a nearby protein radical (on Tyr 233 ) that can be used as the second oxidation equivalent during the reaction mechanism.Indeed, the OH rebound step transfers the final electron to the metal to form the iron(II)-alcohol complex to complete the catalytic cycle.It would be interesting to investigate what effect site-specific mutations of the Asp 179 and Tyr 233 groups will have on the general catalysis by KDO5 enzymes.We predict that mutation of the residues Asp 179 and Phe 233 will significantly affect the product distributions and regioselectivity of lysine dioxygenase as it will influence the dipole moment and local electric field effect inside the active site.
The active site structure and overall protein fold of KDO5 are not dramatically different from other nonheme iron dioxygenases and hence do not explain the PCET observation.In particular, a search for analogous enzyme structures shows that the closest enzyme structures of KDO5 are the structures of asparagine hydroxylase (AH, 25.9% homology) and taurine/ α-ketoglutarate dioxygenase (TauD, 87.1% homology).Both AH and TauD are α-ketoglutarate-dependent nonheme iron dioxygenases that bind an iron(II) to a 2-His-1-carboxylate orientation of the protein.As such, all three enzymes utilize molecular oxygen and convert α-ketoglutarate to succinate concomitant to the C−H hydroxylation of a substrate.Figure 7 shows the comparison of the structures of KDO5 and asparagine dioxygenase as taken from the 6EUR and 2OG7 PDB files. 22,23,113The overlay of the two PDB files is shown in the top right corner of Figure 7 and shows the same structural features and protein fold for the two enzymes with loops, helices, and chains in similar position and orientation.The only difference appears to be an additional loop on the edge of the protein.When we zoom in into the active site region, the strong similarities between the two proteins become even more apparent (left-hand side of Figure 7).Two active site chains are highlighted in Figure 7, namely, part of the chain that includes the equatorial ligands and the part surrounding the axial ligand of the iron(II) ion.Thus, both proteins have the same 2-His/1-Glu coordination environment that holds the metal in position.They also both bind αKG in roughly the same position with the α-keto and carboxylate groups in the same plane as the equatorial ligands.The two protein structures have a protein loop surrounding the αKG  cosubstrate that position it in the active site through interactions with the side chains of Leu and Arg residues: Leu 152 and Arg 289 in asparagine hydroxylase and Leu 173 and Arg 314 in KDO5.Therefore, the αKG binding loop is conserved among αKG-dependent nonheme iron dioxygenases as reported before. 114The only major difference between the two protein structures is the position of a carboxylatecontaining residue (Asp and Glu) in the substrate binding pocket.Thus, in the position of Glu 125 in the substrate binding pocket of asparagine dioxygenase, there is a Gln 144 residue in KDO5, while in KDO5, there is a carboxylate in position 158 (Asp 158 ), whereas asparagine dioxygenase has an Asn amino acid in that position.The difference in the position of an anionic amino acid in the substrate binding pocket will lead to differences in the charge and local electric field orientation in the protein.Previously, a similar difference between the VioC and NapI enzymes was identified as a cause to the dipole moment change and to explain the differences in reactivity of the two enzymes, whereby VioC acts as a C 3 -hydroxylase of free Arg substrate, while NapI reacts with L-Arg through desaturation of the C 4 −C 5 bond. 115ext, we explored the electrostatics of the second coordination sphere of the model and the influence of the electric field effects.The overall dipole moment of structure 5 Re C is shown in Figure 8, while those in models A and B are shown in Figure S8 (Supporting Information).The dipole moment of model C in the 5 Re C state points perpendicular to the chain of atoms of the substrate along the C 4 −H bond.In recent work on the viomycin biosynthesis enzyme VioC, we showed that substrate arginine has its C 3 −H bond aligned with the dipole moment in the active site. 116This results in the weakening of the C 3 −H bond and strengthening of the C 4 −H bond of the substrate.In VioC, the dipole moment therefore guides the reaction selectivity and triggers a lesser thermodynamic favorable pathway.By contrast, in the naphthyridinomycin biosynthesis enzyme NapI, the dipole is along the substrate backbone and appear not to influence the C−H bond strengths in the substrate. 117As a consequence, in NapI, the weakest C−H bond in the gas phase is activated by the protein, whereas in VioC, one of the stronger bonds is activated instead.The dipole moment in model C is also perpendicular to the aromatic ring of Tyr 233 and hence is likely to stabilize a radical center on this group and trigger an electron transfer from Tyr 233 to iron(IV)-oxo and a simultaneous proton transfer from Tyr 233 to Asp 179 .In model A, on the other hand, the dipole vector is perpendicular to the one in 5 Re C and therefore keeps the nonheme center in an iron(IV)-oxo state that reacts differently with the substrate.In particular, in 5 Re A , the C 5 −H bond is weakened, and abstraction of this hydrogen atom is preferential.
To understand how the substrate C−H bond strengths are influenced by the environment and external perturbations, we calculated the C−H bond strengths of an isolated lysine molecule versus those of a hydrogen atom and lysine with one hydrogen atom removed.In the gas phase, the C 3 −H, C 4 −H, and C 5 −H bond dissociation free energies (BDEs) are ΔG = 91.8,93.3, and 92.7 kcal mol −1 , respectively.As such, in the gas-phase for an isolated Lys amino acid, the weakest C−H bond is the C 3 −H bond, and hence, hydrogen atom abstraction from the C 3 −H position should give the most exergonic reaction step.Based on the Bell−Evans−Polanyi principle, the most abundant product should therefore originate from C 3 −H hydrogen atom abstraction.Often, enzymes overcome the Bell−Evans−Polanyi rules by substrate and oxidant positioning and external perturbations in the substrate binding pocket and lead to products that are not expected from the thermochemistry of the reactions.This is sometimes termed negative catalysis. 118To overcome the large exergonicity of hydrogen atom abstraction from the C 3 −H position, external perturbations and substrate binding orientations may block this reaction channel.
To investigate the effect of external perturbations on lysine C−H bond strengths, we recalculated the C 3 −H, C 4 −H, and C 5 −H BDEs in the presence of an applied electric field along either the molecular x-, y-, or z-axis of the substrate.−127 The definitions of the x-, y-, and z-axes in the substrate are shown in Figure 8 with the y-axis along the carbon chain and the x-and z-axes along the C−H bonds.Figure 8b shows the BDE patterns as a function of the applied electric field along the x-and z-axes, which are both perpendicular to the carbon chain of the substrate.In the gas phase, i.e., with zero electric field, the C 5 −H position has the weakest bond followed by the C 3 −H and C 4 −H positions.A field along the positive x-or zdirection keeps these orderings, although the energy separations change.With a negative direction of the electric field (based on the definition in Gaussian) along the x-axis, however, C 4 −H becomes the weakest bond and C 5 −H the strongest bond.Under a local electric field effect in the protein, therefore, C 4 −H hydrogen atom abstraction can be favored over the C 3 −H and C 5 −H channels.On the other hand, a field along the z-axis in the negative electric field direction makes the C 3 −H bond of the substrate the weakest and the C 5 −H bond the strongest.These electric field calculations show that the charge distributions in the substrate binding pocket can be crucial in determining the weakest C−H bond of the substrate within the substrate binding pocket.The charge distributions ultimately determine the regio-and stereoselectivity of the reaction process.
To find out whether the BDEs of the substrate C−H bonds are indeed affected by the cluster models, we took the 5  Therefore, the dipole moment in model C and the local electric field strengths polarize the substrate and lead to preferential C 4 −H hydroxylation of substrate lysine, while that is not the case with the other cluster models.
We further applied electric field calculations on a minimal cluster model of iron(IV)-oxo with its first coordination sphere only.These calculations give a strong field effect on the reduction potential of the iron(IV)-oxo species with a field located along the molecular z-axis.This is not surprising as iron(IV)-oxo reduction leads to filling of the σ* z2 orbital with one electron, which is an orbital along the molecular z-axis.In addition, electric field effect calculations were done on a Journal of the American Chemical Society minimal model containing p-cresol and acetate.These calculations show that proton transfer can be triggered by an electric field effect parallel to the hydrogen bond axis.Our calculations therefore indicate that subtle perturbations from protein residues can trigger a proton transfer, electron transfer, or a combination of the two to obtain valence tautomerism in the active oxidant.

■ CONCLUSIONS
In this work, a computational study is presented on a 4-lysine dioxygenase.We set up three cluster models based on an MD simulation of the substrate-bound iron(IV)-oxo species.The MD shows tight substrate binding with strong interactions between the substrate and polar active site residues.For all three clusters, a full reaction mechanism for the hydroxylation of L-Lys on the C 3 -, C 4 -, and C 5 -positions was calculated.In general, the hydrogen atom abstraction transition state is ratedetermining and the OH rebound step is much lower in energy.Only the largest cluster model predicts C 4 -hydroxylation, while the other two models predict C 5 -hydroxylation.As such, the polarity of the models and the secondcoordination sphere of the substrate and oxidant influences reactivities and kinetics of the substrate activation processes.We then analyzed the structures and found that the dipole moment is different in the three cluster models and particularly points along the C 4 −H bond in model C, while it does not point along the C 4 −H bond in the other models.Further analysis of the structures shows that KDO reacts through negative catalysis where the electric dipole moment and charge distributions in the substrate binding pocket affect the various substrate C−H bond strengths and trigger the C 4 -hydroxylation mechanism.Our work shows that KDOs can be selectively engineered to give the preferred reaction pathways with substrate lysine or alternative substrates.Finally, in the largest cluster model, we observe unusual proton-coupled electron transfer where an active site Tyr residue is involved in the catalysis and reduces the iron from iron(IV) to iron(III) while donating its proton to a nearby Asp residue.This implies that in KDO5, substrate activation is performed by an iron(III)-oxo species rather than an iron(IV)-oxo species.Despite the lower valency of the metal center, the reaction still proceeds through substrate hydroxylation, which further shows that iron(III)-oxo species are capable of substrate oxidation events.This unusual proton-coupled electron transfer with assistance of an active site Tyr residue may be common in highly selective and energy-demanding (negative catalysis) reaction processes in nature, and we are searching for further applications of this interesting phenomenon.

■ ASSOCIATED CONTENT
* sı Supporting Information

Figure 1 .
Figure 1.Active site of KDO5 (PDB ID: 6EUR) and the general reaction catalyzed by the enzyme.

Figure 2 .
Figure 2. Cluster models A, B, and C of KDO5 as studied in this work.Model C is based on model B with the second-coordination sphere expanded, as highlighted in purple.Wiggly lines identify where bonds were cut and replaced by C−H.

Figure 3 .
Figure 3. (a) RMSD plot of the MD simulation on a substrate-bound KDO5 structure.(b) Distance plots of key active site amino acid residues with respect to the substrate.(c) Substrate−oxidant C 3 −O, C 4 −O, and C 5 −O distances during the MD simulation.
,57−60 we analyzed the substrate position with respect to the iron(IV)-oxo species during the MD simulation, and we display the C 3 −O, C 4 −O, and C 5 −O interactions in Figure 3c.After about 40 ns, the substrate− oxidant interaction stabilizes with the C 4 atom closest to the iron(IV)-oxo species.In the MD run between 100 and 130 ns, however, the C 4 and C 5 atoms of the substrate are at similar distances.In the final stages of the MD simulation, i.e., after 130 ns, the substrate falls into a position that is closest to the iron(IV)-oxo species with the C 4 atom of the substrate nearest to the oxidant.In those structures, the system is clearly set up for C 4 −H activation as the C 3 and C 5 atoms are much further away.

Figure 4 .
Figure 4. (a) UB3LYP-D3/BS1-optimized geometries of the iron(IV)-oxo species of KDO5 for models A, B, and C with bond lengths in Å.(b) Overlay of the 5 Re C structure (blue) with the crystal structure coordinates of the 6EUR PDB (gold).

Figure 6 .
Figure 6.Electronic configuration of the intermediates along the hydroxylation mechanism for model C.

Figure 8 .
Figure 8.(a) Electric dipole moment in 5 Re C .(b) Electric field effects along the x-or z-direction on the C−H BDEs of an isolated lysine substrate molecule with ΔG values (in kcal mol −1 ) as a function of the applied electric field.
Re A (model A),5 Re B (model B), and5 Re C (model C) structures and ran a single point energy calculation in the sextet spin state with one hydrogen atom removed from the substrate from position C 3 −H, C 4 −H, or C 5 −H.Using 5 Re A as a starting point, all substrate BDE values are calculated to be within 0.3 kcal mol −1 of each other.By contrast, for5 Re C , the weakest substrate bond is the C 4 −H interaction with a BDE of ΔE = 109.3kcal mol −1 .On the other hand, the C 3 −H and C 5 −H BDEs are ΔE = 112.1 and 112.9 kcal mol −1 .Finally, for model B, the weakest C−H bond is the C 3 −H bond at 109.2 kcal mol −1 , while the C 4 −H bond has a BDE of 110.1 kcal mol −1 .

Table 1 .
Hydrogen Atom Abstraction Transition States Calculated for C 3 −H, C 4 −H, and C 5 −H Abstraction from Lysine by KDO5 Iron(IV)-oxo Species aEnergies are calculated at the BS2 level of theory, while ZPE, thermal, and entropic corrections are obtained at the BS1 level of theory. a