Mechanism of DOPA radical generation and transfer in metal-free class Ie ribonucleotide reductase based on density functional theory

Graphical abstract


Introduction
DNA is the carrier of genetic information of life, and the replication and repair of DNA is one of the core links in life. In all known organisms, the four 2 0 -deoxyribonucleoside triphosphate (dNTPs) substrates used for DNA synthesis are provided by the catalysis of ribonucleotide reductase (RNR). RNR catalyzes the reduction of C2 0 -OH of ribonucleotides (NDPs/NTPs) to form corresponding deoxyribonucleotides (dNDPs/dNTPs), providing the required precursor for DNA synthesis and repair [1,2]. RNR is a kind of enzyme that catalyzes reactions through biological organic radicals, and can be grouped into three classes, Ⅰ-III [3][4][5][6][7][8][9][10][11][12][13]. Class I RNRs are usually composed of two nonidentical dimeric subunits, a subunit (also known as R1 subunit) and b subunit (also known as R2 subunit), and the two subunits interact to form a 2 b 2 heterotetramer [5]. R1 is the catalytic subunit and R2 is the radical generation subunit. The reaction of ribonucleotide reduction can be started by the cysteinyl radical (CysÅ) located in R1 subunit, and this radical is transferred from the tyrosyl radical (TyrÅ) in R2 subunit over a long distance (more than 35 Å). Depending on the identities of metal cofactors combined by R2 and the mechanism of radical generation, class I RNRs can be divided into four subclasses: Ia (PMID:4337857), Ib (PMID:20698687), Ic (PMID:15247479), and Id (PMID:29609464) [14][15][16][17]. All those subclasses contain dinuclear metal cofactors [18][19][20][21][22][23][24].
Since the discovery of RNR, metal cofactors have been considered to be necessary for the generation and stabilization of the catalytic radical in RNR. However, Högbom et al. [25] and Boal et al. [26] have found a group of RNR (class Ie) with a trace amount of metal cofactors in human pathogens. Compared to the central region of the dimanganese cluster in the R2 subunit of Ib enzymes (Fig. 1a), the active site of class Ie consists of an aspartic acid, a valine, a proline, a lysine, and two histidines (Fig. 1b). For class I RNRs, the six residues being coordinated to the transition metal cofactors in R2 subunits are completely conserved. The six conserved residues and the transition metal cofactors are necessary for enzyme activation. While class Ie is the only subclass that is independent of metal cofactor [25,26]. It is considered to be evolved in response to the extreme metal-limited environments.
These changes imply that the mechanism of radical generation and stabilization in class Ie is completely different from that of other subclasses.
In the R2 protein of class Ie, the tyrosine residue can be oxidized to a stable 3,4-dihydroxyphenylalanine (DOPA) radical during enzyme activation, and this radical can initiate the ribonucleotide reduction in vitro and in vivo. The active site structure of the active R2 is shown in Fig. 1c. Covalent modification of the residue tyrosine to form the DOPA radical is a three-electron oxidation process. It is suggested that the superoxide provided by flavoprotein NrdI takes part in the oxidations [25,26]. How the superoxide participates in the DOPA radical generation is unclear [25,26]. A trace amount of metal ions are detected in R2 protein. It is not clear whether the metal ions are transiently involved in the generation of the DOPA radical [25,26].
In the active state of class Ie-R2 subunit, the tyrosine-derived DOPA radical, which serves as the catalytic initiator, is transferred to a cysteine residue in the R1 subunit through a transfer chain when the ribonucleotide reduction reaction is initiated. After the reaction, the radical is then reversibly transferred back to its initial position at the DOPA residue [25,26]. Stubbe et al. [27] used DOPA as a probe to study the radical transfer chain. They have found that the reduction potential of DOPA was 260 mV lower than that of tyrosine at pH 7.0. It is indicated that DOPA forms a radical trap in the class Ie-R2 subunit. Compared to the tyrosine radical, the DOPA radical is more stable and difficult to be transferred to other residues. Therefore, prior to transferring the DOPA radical, there may be a triggering mechanism that the redox potential difference between DOPA and other residues in the transferring chain is adjusted. This adjustment may be induced by the changes of the protein microenvironment and the conformation of the R2 subunit after binding to R1 subunit. It is no clear that how the redox potential of amino acid residues is adjusted upon the changes of the protein microenvironment and conformation in class Ie. Another concern for Ie RNR is that how to regenerate the DOPA radical directly from the DOPA residue after the radical quenching. Högbom et al. confirmed that once the radical is lost, it can be regenerated by NrdI in an oxygen-dependent process [25]. Hence, the activation step most likely also takes place via superoxide.
As a new discovered RNR enzyme, class Ie is completely different from the previous subclasses in structure and function [25,26]. Understanding the mechanisms of the generation, transfer, and regeneration of the DOPA radical in class Ie is helpful for the combat of pathogens and the development of anti-pathogen drugs. For   (PDB: 6GP3). Green, red, and blue represent carbon, oxygen, and nitrogen, respectively, and water is represented by red sphere. The dotted line in red represents the solvent access channel. (b) Total electrostatic potential (ESP) for the side chains around the solvent access channel. this purpose, in this work, density functional theory (DFT) has been used to study the mechanisms of the generation, transfer, and regeneration of the DOPA radical.

System setup
Two models (models 1 and 2) of the class Ie-R2 subunit were constructed. The initial geometries for building the two models were constructed based on the experimental X-ray crystal structures of chain A of the inactive MfR2 (PDB: 6GP3, model 1) and the active MfR2 (PDB: 6GP2, model 2) [25]. The difference of the crystal structures between the inactive and active states is that, 3,4-dihydroxyphenylalanine (DOPA) at the active site of the active state corresponds to the tyrosine residue Tyr126 at the active site of the inactive state. (Fig. 1b, c). The protonation states of the titratable residues were determined at pH 7.0 based on the protein pKa predictor PROPKA 3.0 [28][29][30] or based on the neighboring hydrogen bond networks. It is reasonable that all arginine (Arg) and lysine (Lys) residues are in their protonated states, whereas all aspartic acid (Asp) and glutamic acid (Glu) residues are in their deprotonated states with the exception of Asp88 (for asp88, we investigated different protonation states). All histidines (His) are in protonated states. His107 and His122 are protonated at their Ne2 positions, and His143 and His216 are protonated at their Nd1 positions. For the amino acid chain, the N-terminal is protonated, whereas the C-terminal is deprotonated.

Molecular dynamics (MD) simulations
All classical molecular dynamics simulations were performed using AMBER99SB-ILDN force field [31] with the GROMACS 2019.5 software package [32,33]. The missing hydrogen atoms were added automatically. The TIP3P model was used for the solvent water molecules [34]. The system was placed in a periodic truncated cubic box, and the closest distance between the surface of the box and the protein atom was set to 8 Å. The protein was hydrated using SPC216 water molecule, and then neutralized by adding ions (Na + and Cl À ) to generate 0.15 mol/L NaCl solution subsequently. The energy minimization was performed to relax the system using steepest descent and conjugate gradient algorithms. Then, the position restriction equilibrium simulations were conducted in following. (i) The minimized system was heated from 0   to 300 K at constant volume using the velocity-rescale thermostat in a NVT canonical ensemble for 100 ps. (ii) The system was equilibrated in a NPT ensemble for 100 ps, employing the Berendsen barostat at the constant temperature of 300 K and 1 bar. Isotropic pressure coupling was applied with a compressibility of 4.5 Â 10 À5bar À1 . All of the solute molecules were restrained to the origin using harmonic potential of 1000 kJ mol À1 nm À2 during the equilibrium simulation process. Finally, a production MD run was performed for continuous 30 ns in a NPT ensemble with a target pressure of 1 bar and a pressure coupling constant of 2 ps. The periodic boundary conditions were employed in all the simulations, with time integration step size of 2.0 fs. LINCS algorithm was used to constrain bonds involving hydrogen [35]. Long-range electrostatic interactions were calculated using the particle-mesh Ewald (PME) summation algorithm [36] with fast Fourier transform (FFT) grid spacing of 1.6 Å and an cubic interpolation order of 4. Lennard-Jones and short-range Coulomb interaction cut off values were both 10 Å. Frames were collected at 10 ps intervals, and the production trajectories were analyzed using Visual Molecular Dynamics (VMD) [37]. Visualization was performed using PyMOL (the PyMOL Molecular Graphics System, Schrödinger, LLC) and CYLview [38]. RMSD changes in skeleton atoms of protein along the MD trajectories for the inactive and active MfR2 are shown in Fig. S1.

QM/MM calculations
The initial structure of the QM/MM calculation was taken from the trajectory of the MD simulations. In order to simplify the research model, the solvent molecules beyond 5 Å of the protein were removed. All QM/MM calculations were performed using ONIOM [38][39][40][41] implemented in the Gaussian 09 program [42]. Model 1 was used to study the mechanism of the generation and regeneration of DOPA radical at the active site of the R2 subunit. The active site residues include Asp88, His122, Tyr126, and Lys213, and the neighboring residues are Leu84, Val119, Leu183, Phe187, and Ile209. O 2 ÅÀ or HO 2 Å are the oxidants. All these consti-tute the QM region of model 1. Model 2 is used to study the transfer mechanism of DOPA radical between DOPA126 and Trp52. Trp52 is closed to Arg211, Asp212, and W558, they have hydrogen-bonding interactions with His122 of the active site. The residues of the QM region of model 1, Trp52, Arg211, Asp212, W558, and W580, constitute the QM region of model 2.
In addition, DOPA126 replaces Tyr126 in model 1. The remaining part of the system was regarded as the MM region. To maintain the contribution of outer residues and water molecules to the system, all amino acid residues and water molecules beyond 6 Å of the QM region were freezed. An electronic embedding scheme [43] was applied to include the polarizing effect of the enzymatic environment on the QM region. Hydrogen link atoms [44] with the charge shift model were employed to treat the QM/MM boundary. The AMBER force field was used for the MM region. The pure QM part was treated by density functional theory (DFT) with the B3LYP exchange-correlation functional method [45][46][47][48]. The B3LYP functional, which is a density-functional theory (DFT) type of calculation based on hybrid functional, has been proved to provide a good balance between speed and accuracy for modeling enzyme reactions [49,50]. Geometry optimizations were performed using a double-f 6-31G (d,p) basis set (labeled B1). All minima were fully optimized without any symmetry restraints. The transition states (TSs) were determined by an initial potential energy surface (PES) scans followed by fully TS optimizations. Frequency calculations were performed using the theory and methodology same to that of geometry optimizations to identify the correct minimum (no imaginary frequency) and transition state (one imaginary frequency) structure, and to obtain zero-point energies (ZPE). To obtain more accurate energies, single-point energies were calculated using a larger all-electron basis set, which is 6-311G (2df,p) for all the atoms (labeled B2). The free energy was obtained based single-point energy, which is corrected by zeropoint energy. Based on the ground-state electronic wave functions, the total electrostatic potential (ESP) analysis, charge analysis, spin density analysis, and spin natural orbitals distribution analysis were performed using Multiwfn 3.6 program [51,52]. Rate constants for the radical transfer between DOPA126 and Trp52 were calculated according to the Marcus-Hush-Levich formula (Eq. (1)) [53][54][55][56][57][58][59].
In this expression, k is the nuclear reorganization energy accompanying electron transfer, H DA is electronic coupling matrix element between the donor and acceptor, DG is the reaction free energy, ⁄ is Planck's constant, k B is Boltzmann's constant, and T is temperature. H DA is estimated by the method in the articles [53,60,61].

Results and discussion
3.1. Generation mechanism of the DOPA radical

) as the oxidant when Asp88 in deprotonated state
Through analysis of the crystal structure of the inactive MfR2 (PDB: 6GP3), it is found that there is a solvent access channel connecting the active site with NrdI (Fig. 2a). This hydrophilic channel exists in both Ib and Ie subclasses and is used to transport superoxide produced by NrdI [26]. At the end point of the channel, there is a water molecule (W530), which has hydrogen bonding interactions with both Tyr126 and Asp88. Superoxide (O 2 ÅÀ ) is generated by the single-electron reduction of O 2 under the catalysis of NrdI [62,63]. The experiments showed that NrdI is involved in the formation of the DOPA radical, therefore O 2 ÅÀ is speculated to be the oxidant for the formation of the DOPA radical [25,26]. Thus, direct oxidation by O 2 ÅÀ was first investigated. Once O 2 ÅÀ is generated, it enters into the active site via the solvent access channel. Based on analysis of the crystal structure, it is suggested that after O 2 ÅÀ enters into the active site, it arrives at the position of W530 and expels the water. In R2 protein, Asp88 was found to be located in the vicinity of Tyr126 and interact with Tyr126 via a hydrogen-bonded water (W530). Asp88 is an acidic amino acid, and its protonation state is depended on the side chain pKa. According to the protein pKa predictor PROPKA 3.0 [28][29][30], the pKa of Asp88 is 5.86 in the inactive MfR2 and 6.20 in the active MfR2. Therefore, we speculate that Asp88 may exhibit different protonation states in the active center, and the protonation state plays an important role in the radical generation and transfer. In order to analyze the protonation state of Asp88, we carried out MD simulations (30 ns) for the crystal structures in different protonation states. Through the overlay of the initial crystal structures and the structures after the MD simulations, it is found that the structure of Asp88 in deprotonated state is more close to the initial structure after MD simulation (Fig. S2). Therefore, we have simulated the direct oxidation by O 2 ÅÀ when Asp88 is in deprotonated state. As illustrated in Fig. 3 Fig. 4. The salt bridge between Asp88 and Lys213 disappears when Asp88 is protonated in IN2. As O 2 ÅÀ enters the active site, it abstracts the proton from Asp88 due to its strong elec-   Fig. 8. Spin natural orbitals distribution of unpaired single-electron for the DOPA radical generation using HO 2 Å as the oxidant in the protonated state of Asp88 (pathway A).
Blue and green represent the negative and positive spin density, respectively.    Fig. 7). The unpaired spin density located HO 2 Å is almost transferred to Tyr126. This step is endother-mic. The calculated free energy barrier is 15.5 kcal/mol and the reaction free energy is 7.1 kcal/mol. In the second step (IM1 ? IM2), the distal oxygen atom Od abstracts the phenol proton of Tyr126, with the formation of a water molecule. In IM1, the hydrogen bond distance between HO 2 Å and the carboxyl group of Asp88 is 1.93 Å. This hydrogen bond is very weak, hence has little effect on the oxidation reaction. From IM1 to IM2, the Od-Op distance is gradually elongated and the Od-H2 distance is gradually shortened. In the intermediate IM2, the  Od-Op and O1-H2 bonds break simultaneously, and the Od-H2 bond is formed with the formation of a water molecule. This step is exothermic. The free energy barrier is 10.5 kcal/mol, and the reaction free energy is À42.9 kcal/mol. In IM2, the unpaired single-electron is essentially located on Op with a spin density of 0.79 (Table 1, Fig. 7), revealing that Op has a strong activity and can receive a hydrogen. Therefore, the third step (IM2 ? PC) involves the abstraction of hydrogen H1 by Op to generate the C3 0 hydroxyl group. This step undergoes the third transition state (TS3). In IM2, the distances of the C3-Op and C3-H1 bonds are 1.35 Å and 1.12 Å, respectively, and the angle \Op-C3-H1 is 106.7°. Due to the abstraction by Op, H2 moves toward Op, and the angle \Op-C3-H1 gradually becomes smaller. In TS3, the angle \Op-C3-H1 becomes 63.1°, and the distances of the C3-Op and C3-H1 bonds are elongated to 1.37 Å and 1.23 Å, respectively. Upon the generation of the hydroxyl group, the DOPA radical is finally formed in the product complex (PC). In PC, the charge and spin density of the oxygen atom O1 are À0.28 and 0.35, respectively ( Table 1). 25% of the unpaired single-electron is located on O1 (Table 1, Fig. 8g), confirming the formation of phenoxylradical. In the final step, the calculated free energy barrier is 18.3 kcal/mol, and the reaction free energy is À43.4 kcal/mol, revealing that this step is highly exothermic. Scheme 4 illustrates the mechanism of the DOPA radical generation. Based on the DFT simulations, it has been found that HO 2 Å , not O 2

ÅÀ
, can directly oxidize Tyr126 to form the DOPA radical. It is suggested that the oxidation does not need the participation of metal ions. The reaction is a three-electron oxidation, which undergoes the tyrosine meta-hydroxylation and the DOPA radical generation. The calculation results are consistent with that proposed by Högbom et al. [25]. Fig. 9 shows the calculated potential energy profile for the pathway A, which undergoes three steps after RC. The third step is the rate-determining step, and it is exothermic. The calculated rate constant in the third step is 1.43 Â 10 2 s À1 , which is close to the experimental reaction rate of HO 2 Å with phenylalanine <180.0 ± 50.0 s À1 [68]. The total reaction energy of the pathway is À79.2 kcal/mol. Therefore, the reaction pathway is thermodynamically favorable. It is noticed that in the pathway, the DOPA radical, not DOPA, is directly formed. When Asp88 is in the deprotonated state, the oxygen atom of the carboxyl group can form a strong hydrogen bond with the surrounding hydrogen atom. The hydrogen bonding interaction is not favorable for the oxidation reaction. Whereas Asp88 in the protonated state, the hydrogen bonding interaction is significantly weakened. Therefore, the prerequisite of the oxidation reaction is that the side chain of Asp88 is protonated. According to the pKa of Asp88 of the inactive MfR2 and the MD simulation of the initial crystal structure (Fig. S2), Asp88 is likely to be deprotonated in the initial state. Therefore, there may be a switch mechanism in class Ie. Once the reaction of the DOPA radical generation is initiated, the switch mechanism can facilitate Asp88 to be protonated. Previous researches have proposed a histidine switch mechanism for viral membrane fusion [69][70][71]. This mechanism shows that, in many viral fusion proteins, histidines are located in the vicinity of positively charged residues in the prefusion conformation, and are singly protonated at neutral pH. However, once stimulated by external conditions such as mildly acidic conditions (pH 4.5-6.5), that are close to the pKa of histidine side-chain protonation, histidines are doubly protonated and positively charged. Many proteins make use of conformational changes to achieve their functional purposes [72]. The conformational change is usually coupled with the protonation states of ionizable residues such as Lys, Glu, Asp, and Arg residues, and the coupling is an effective regulation mechanism of proteins [72]. The internal ionizable residues would adjust their protonation states and pKa values in different microenvironments. For aspartic acids, when they are exposed to a hydrophilic region, they exhibit deprotonation states. However, when they are located in a relatively hydrophobic region of the protein, their pKa values would tend to rise, and they exhibit neutral protonation states. In the active site of the R2 subunit, Asp88 is located in vicinity of the e-amino group of Lys213. As shown in Fig. 2b, the total electrostatic potential (ESP) around the e-amino group of Lys213 is positive. Based on the protein pKa predictor PROPKA 3.0 [28][29][30], the pKa of Asp88 of the inactive MfR2 is 5.86, and that of the active MfR2 is 6.20. Therefore, it is speculated that there is a ''switch phenomenon" in class Ie. When the flavoprotein NrdI binds to the R2 subunit, the microenvironment of the active site is changed from neutral hydrophilic to mildly acidic and relatively hydrophobic, leading to that the pKa of Asp88 is increased from 5.86 to 6.20. In the meantime, Asp88 turns to be in a protonated state. After the generation of the DOPA radical, R2 separates from NrdI. The microenvironment of the active site is changed from relatively hydrophobic to hydrophilic, and the pKa of Asp88 is decreased to 5.86. Thus, Asp88 is deprotonated and returned to its initial state. The possible reason for this mechanism is that when the protein aggregates, the solvent channel between the active site and the external aqueous solution is blocked to some extent, the active site of protein becomes more hydrophobic. O 2 ÅÀ can be protonated to form hydroperoxyl radical HO 2 Å . Therefore, like the protonation of Asp88, it is speculated that when R2 is bonded to NrdI, the microenvironment of the active site turns to be mildly acidic, which benefits the protonation of O 2 ÅÀ . In the active site, the positively charged e-amino group of Lys213 is located in the vicinity of the solvent access channel (Fig. 2a). We speculate that O 2 ÅÀ can be protonated to HO 2 Å at such a mildly acidic condition. HO 2 Å has been demonstrated to be more reactive than O 2 ÅÀ [67]. It is reasonable that HO 2 Å is the oxidant that oxidizes the side chain of Tyr126 to form the DOPA radical.

Transfer mechanism of the DOPA radical
The radical transfer in proteins is accomplished through a proton-coupled electron transfer (PCET) mechanism [73]. Electron is transferred by a long-range electron tunneling. In biological proteins, the amino acids, which can serve as waystations for electron tunneling (hopping) reactions, are almost tyrosine and tryptophan residues [74]. After careful analysis, it is found that Trp52 is the first electron waystation nearby DOPA126. So, when the DOPA radical transfer is initiated, an electron located on Trp52 is transferred to the phenol group of DOPA126 over the radical transfer chain, creating a cationic tryptophan radical Trp52 +Å . The radical transfer from DOPA126 to Trp52 is the first and most important step. The radical transfer in the deprotonated state of Asp88 has been studied. However, we could not obtain Trp52 +Å after optimization. As shown in Fig. 10a, the hydrogen on C3-OH of DOPA126 has been transferred to the carboxyl group of the side-chain of Asp88. This leads to that the radical is still on DOPA126 and cannot be transferred to Trp52, although the C4 0 phenoxyl radical (C-OÅ) has been protonated. Fig. 10b shows that Trp52 has no spin density. This result is consistent with the opinion of Boal et al. [26], who proposed that the retention of this hydrogen is critical for tuning of the redox potential of the DOPA radical. Hence, in order to hold the hydrogen of C3-OH on its position, Asp88 should be protonated before the initiation of the proton/electron transfer.
Further calculations were carried out. The structures are illustrated in Fig. 11, including the reactant which has a neutral DOPA radical in the deprotonated state of Asp88 (DOPA126 Å -Trp52, RC1), the reactant in the protonated state of Asp88 (DOPA126 Å -Trp52, RC2), the transition state (TS), and the product complex (DOPA126-Trp52 +Å , PC). In RC2, there are two intermolecular hydrogen bonds between DOPA126 and Lys213. The optimized distances of O1-H1 and O2-H2 are 1.78 Å and 1.81 Å, respectively. About 98% of the unpaired single-electron is distributed on DOPA126, which has a spin density of 1.0 ( Table 2, Figs. 12, 13). There is almost no spin density resides on the other residues. The charge analysis (Table 2) shows that the total charge of Trp52 is À0.10, indicating that Trp52 is initially electronegative. As the radical transfer is initiated, the atom H1 gradually moves toward O1, and H2 gradually approaches to O2. This step undergoes a transition state (TS). From RC2 to TS, the O1-H1 distance changes from 1.78 Å to 1.11 Å, and the O2-H2 distance changes from 1.81 Å to 1.05 Å. In TS, about 14% of the unpaired single-electron is transferred to Trp52 (Table 2, Figs. 12, 13). As the reaction proceeds,  the O1-H1 and O2-H2 distances are gradually shortened. In PC, the O1-H1 and O2-H2 bonds are formed simultaneously. The lengths of the O1-H1 and O2-H2 bonds are 1.07 Å and 1.03 Å, respectively. As the reaction going from RC2 to PC, the charge of Trp52 changes from À0.01 to 0.06, and the spin density changes from 0.00 to 0.19 (Table 2, Fig. 12). About 19% of the unpaired single-electron is distributed on Trp52 in PC ( Table 2, Fig. 13), revealing that the cationic tryptophan radical Trp52 +Å is formed, although most of  the unpaired single-electron (79%) is still located on DOPA126. The unpaired spin density mainly resides on the aromatic ring of DOPA126 in PC, demonstrating that the electron transfer between these two residues is complicated. The changes of protein environment and conformation may be needed in the transfer process. Our calculations help to understand radical transfer mechanism, and address the concern from the articles that how the DOPA radical is transferred from the radical trap. From RC2 to PC, the reaction is endothermic. The forward and backward energy barriers are 15.2 and 1.0 kcal/mol (Fig. 9), respectively, and the calculated rate constants are 1.17 Â 10 3 and 2.89 Â 10 13 s À1 , respectively. The forward rate constant obtained is in agreement with previous study [74]. The backward energy barrier is slightly endothermic (1.0 kcal/mol), which means that the transfer rate is very fast.
Scheme 5 illustrates the radical transfer between DOPA126 and Trp52. According to the mechanism of the radical transfer between DOPA126 and Trp52 and the hydrogen bonds formed, the residues along the radical transfer chain are identified, (Fig. 14) (Tyr356 was not resolved in the crystal structure).
The e-amino group of Lys213, not only as a bridge for electron transfer but also as a proton donor, provides a proton to the phenoxyl radical (C-OÅ) of DOPA126 to form the phenol group (C-OH). Lys213 is the key residue in the radical transfer chain between DOPA126 and Trp52. Between the side chains of DOPA126 and Lys213, there is a water molecule (W580). The water molecule forms hydrogen bonds with both DOPA126 and Lys213. Thus, an electron and proton channel is formed. It is a water mediated double-proton-coupled electron transfer (dPCET), that occurs     (pathway C in Fig. 9), indicating that the reaction is exothermic. Hence, the regeneration reaction is thermodynamically favorable from the energy point of view. The reaction mechanism is illustrated in Scheme 6.

Hydroperoxyl radical (HO 2 Å ) as the oxidant
DFT calculations also have been carried out using hydroperoxyl radical (HO 2 Å ) as the oxidant, and the optimized structures of the key steps of the process are illustrated in Fig. 18. In RC, HO 2 Å is hydrogen bonded to the C3 0 hydroxyl group of DOPA126. As the reaction going from RC to PC, the phenol proton of DOPA126 is transferred to HO 2 Å , the free radical is transferred concomitantly from HO 2 Å to DOPA126. HO 2 Å is protonated to form a hydrogen peroxide molecule in PC. The reaction also undergoes a transition state (TS). Figs. 19, 20 confirm the transfer of the unpaired singleelectron from HO 2 Å to DOPA126. The calculated energy barrier is 6.7 kcal/mol, and the reaction free energy is À12.7 kcal/mol (pathway D in Fig. 9). The reaction is exothermic and the regeneration reaction is thermodynamically favorable. The reaction mechanism is illustrated in Scheme 7.

Conclusions
For class Ie RNR, the mechanisms of the covalent modification of Tyr126 to the DOPA radical and the radical transfer and regeneration have been studied based on DFT calculations. As a new class of RNR enzyme, class Ie exhibits different mechanisms of radical generation and transfer. The DFT calculations have showed that O 2 ÅÀ cannot directly oxidize Tyr126 to form the DOPA radical in class Ie. It must be protonated to HO 2 Å prior to the oxidation. The DFT calculations suggest that the covalent modification of Tyr126 and the DOPA radical generation can be carried out with no involvement of metal cofactors. The DFT study reveals the steps of bond formation and bond breaking upon the oxidation of Tyr126 by HO 2 Å . This work supports that metal environment and amount may not have an effect on the survival of pathogens containing class Ie [25,26]. The protonation of Asp88 is the prerequisite for the generation and transfer of the DOPA radical in class Ie. The protonation state of Asp88 can be regulated by the change of protein microenvironment, which is induced by the protein aggregation and separation of the R2 subunit with flavoprotein NrdI or the R1 subunit. The e-amino group of Lys213 plays an important role in the DOPA radical generation and transfer. Once the radical is quenched, it can be regenerated via the oxidations by superoxide O 2 ÅÀ and hydroperoxyl radical HO 2 Å .

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.