Conformationally Gated Multisite Proton-Coupled Electron Transfer in the Ribonucleotide Reductase $\beta$ Subunit

Abstract

Ribonucleotide reductase (RNR) is an essential enzyme that converts ribonucleotides into deoxyribonucleotides, enabling DNA synthesis and repair in all living organisms. Central to class Ia RNR activity is a long-range radical transport pathway spanning ~32Å across the $\alpha$ and $\beta$ subunits by a series of proton-coupled electron transfer (PCET) reactions. Although the collinear PCET reactions in the $\alpha$ subunit have been extensively studied, the multisite, orthogonal PCET reactions in the $\beta$ subunit are less well understood. This work focuses on orthogonal PCET between the redox-active tryptophan, W48, and interfacial tyrosine, Y356, in the $\beta$ subunit. Multiscale modeling strategies are employed to explore this PCET reaction. The simulations show that radical transfer from W48 to Y356 is thermodynamically favorable and is likely to occur by electron transfer from Y356 to the W48 cationic radical in conjunction with proton transfer from Y356 to a glutamate, E52, which forms a hydrogen-bonding interaction with Y356 following oxidation of W48. The conformational gating motion of Y356 is shown to be critical for allowing this residue to participate in PCET with W48 in the $\beta$ subunit and with a tyrosine in the $\alpha$ subunit. Application of vibronically nonadiabatic PCET theory highlights the significance of hydrogen tunneling and conformational motions that shorten the distance between Y356 and E52. This work demonstrates how conformational gating, hydrogen-bonding networks, and hydration at the $\alpha /\beta$ interface modulate PCET in RNR. These fundamental insights are also applicable to other biomolecular systems and may guide therapeutic and protein engineering applications.

Ribonucleotide reductase (RNR) catalyzes the conversion of ribonucleotides to deoxyribonucleotides and is an essential enzyme for DNA synthesis and repair (1–4). Due to its vital role in the maintenance of genome and DNA replication fidelity, RNR has become a key therapeutic target for both antibacterial and anticancer drug design (5, 6). Among the different classes of RNR, E. coli class Ia RNR (7) is widely recognized as the canonical model system for studying class Ia-type RNR. It is chemically homologous to human RNR (8, 9) in key structural and mechanistic features while offering unparalleled experimental accessibility for obtaining reliable structural, kinetic, and mechanistic data. The recently solved high-resolution cryo-electron microscopy (cryo-EM) structures (Fig. 1A) (10, 11) of the active ${\alpha }_2{\beta }_2$ complex make it possible to further understand the underlying molecular mechanism of this critical enzyme. These structures are composed of two pairs of $\alpha$ and $\beta$ subunits that form an asymmetric functional unit.

Fig. 1.

RNR structure and PCET pathway. (A) The 3.6-Å resolution cryo-EM structure of the E. coli class Ia RNR ${\alpha }_2{\beta }_2$ complex (PDB ID 6W4X) (10), where the $\alpha /\beta$ pair of subunits is in the pre-turnover state, prior to radical transport. (B) The PCET pathway spans ~32 Å The radical transport process starts from Y122 in the $\beta$ subunit, and the pathway consists of Y122 ↔ [W48] ↔ Y356 ↔ Y731 ↔ Y730 ↔ C439. The blue arrows show proton transfer (PT), the red arrows show electron transfer (ET), and PCET steps are highlighted with purple arrows. W48 has been proposed to participate in the PCET pathway, but its role has not been demonstrated conclusively. In this work, we focus on radical transfer between W48 and Y356 and the associated PT processes. The radical is on Y122 in the resting state, but here we also show the radical on Y356 as it undergoes PCET with Y731 and on Y731 as it undergoes PCET with Y730 for visualization purposes. This figure was modeled after a figure in Ref. (40). Copyright 2020 American Chemical Society.

RNR catalysis requires a long-range radical transport process to enable a cysteine radical to react with the ribonucleotide substrate (Fig. 1B) (12, 13). The directionality and reversibility of this long-range radical transport, as well as the substrate specificity, are tightly regulated at multiple levels (8, 14–16). This process is initiated in class Ia RNR by the activation and reduction of molecular oxygen, which generates a diferric-tyrosyl cofactor (Y122^•) (17–19) in the $\beta$ subunit (7, 20). To initiate the catalytic step, the radical on Y122 in the $\beta$ subunit must travel over ~32 Å across the $\alpha /\beta$ interface to reach C439 in the active site of the $\alpha$ subunit (21). This long-range radical transport is achieved through a series of proton-coupled electron transfer (PCET) reactions (22–26) along a pathway of conserved redox-active residues (Fig. 1B) (27–30). This pathway has been structurally resolved in the interacting $\alpha /\beta$ subunits of the cryo-EM structures (10, 11).

The overall pathway can be divided into three interconnected parts: 1) PCET within the $\beta$ subunit, 2) PCET across the $\alpha /\beta$ interface, and 3) PCET within the $\alpha$ subunit. Based on the properties of the redox-active amino acids along the pathway, each of these PCET reactions is presumed to be concerted, in that the electron and proton transfer simultaneously without forming a stable intermediate (24, 31–33). In the $\alpha$ subunit, the geometric alignment of the redox residues facilitates collinear, concerted PCET, where the electron and proton move in the same direction. Our group has previously characterized the reaction mechanisms (34–37), nonadiabaticity (37), and kinetics (38, 39) of these collinear, concerted PCET steps.

The $\beta$ subunit features an orthogonal PCET pathway, where the proton and electron move in different directions. Due to the distinct proton and electron transfer paths, weak vibronic coupling, and complex underlying multidimensional free energy landscapes, the PCET process in the $\beta$ subunit remains poorly understood from a theoretical perspective. The tyrosine residues shown in Fig. 1B have been identified experimentally as participating in the radical transport process through mutation studies using unnatural tyrosine analogs (4, 12, 30, 41). The tryptophan residue W48, which resides between Y122 and Y356, has been proposed to participate in the PCET pathway (4, 5, 25). W48 is surrounded by several $\alpha$ helices of the $\beta$ subunit and flexible loops at the $\alpha /\beta$ interface, and it is well positioned to hydrogen bond with a nearby D237. However, W48 has not yet been demonstrated definitively to participate in the PCET pathway, either experimentally or computationally.

The $\alpha /\beta$ interface consists of flexible loops and a water channel (10, 11). Y356 is located at the highly flexible $\beta$-tail loop and interacts with the $\alpha$ subunit. A comparison of the two cryo-EM structures, which correspond to the pre-turnover and mid-turnover states (10, 11), shows that Y356 exhibits significant side-chain rearrangement between these two states. Due to its critical location, Y356 in the $\beta$ subunit has been extensively characterized and shown to form a transient Y356-O^• radical based on experiments using unnatural tyrosine analogs (29, 30, 42). A central mechanistic question concerns the identification of the proton donor and acceptor enabling the interconversion between Y356-OH and Y356-O^• during the radical transport process. As discussed further below, experimental and theoretical studies have suggested a direct interaction between Y356 and interfacial water molecules (29, 40, 43, 44). Moreover, experimental studies have shown that mutation of a nearby glutamate, E52, to glutamine (E52Q) leads to loss of function (45, 46). The structural and mutagenesis experimental results have been interpreted to suggest that E52 regulates proton transfer from Y356 indirectly by participating in a water and charged residue network connected to bulk solvent (45). However, direct experimental evidence verifying the role of E52 is still lacking, leaving its precise mechanistic function unresolved.

Another important open question involves the role of the interfacial water channel connecting the $\alpha$ and $\beta$ subunits, namely how this hydration network contributes to long-range radical transport. Electron nuclear double resonance (ENDOR) and electron paramagnetic resonance (EPR) spectroscopies, together with density functional theory (DFT) studies and classical molecular dynamics (MD) simulations, have suggested a direct interaction between Y356 and interfacial water molecules (29, 40, 43, 44). With different orientations of the Y356 side chain, the residue appears to engage with different water channels (11). However, the driving forces and molecular mechanism underlying this conformational change, as well as the collective impact on radical transfer, remain elusive. A recent cryo-EM study (47) mapped the conformational landscapes of class I RNRs and demonstrated that radical transport is tightly gated by asymmetric $\alpha /\beta$ dynamics. What remains unclear is how local interactions and the interfacial water network enable this conformational gating.

These intriguing questions motivate our computational investigation aiming to elucidate multisite, orthogonal PCET in the $\beta$ subunit and to understand how local hydration and conformational motions at the $\alpha /\beta$ interface regulate long-range radical transport in RNR. Specifically, we provide computational evidence that W48 participates directly in the radical transport pathway and clarify the mechanism of orthogonal PCET between W48 and Y356. Our investigation of the proton transfer reaction from W48 to the nearby D237 shows that proton transfer (PT) from W48 to D237 is thermodynamically and kinetically possible but is unlikely to play a key role in the radical transport process. Furthermore, our calculations provide evidence that the direct proton acceptor for Y356 may be E52, which becomes accessible after conformational rearrangements upon oxidation of W48. Finally, we explore how conformational rearrangements and water networks at the $\alpha /\beta$ interface can modulate the radical transport process.

Results and Discussion

To tackle such a complex system and these challenging questions, we employed multiscale computational and theoretical strategies. This work focuses on radical transfer between W48 and Y356 in the $\beta$ subunit and the associated PT reactions and conformational changes. Starting with unbiased atomistic MD simulations of different radical states of W48 and Y356, we can mimic different stages of the radical transfer process and generate conformational ensembles for systematic analysis. These simulations allowed us to identify key conformational changes and to select appropriate reaction coordinates for subsequent enhanced sampling simulations. Considering the conformational flexibility of the system and the potential involvement of water molecules, we performed metadynamics simulations to explore conformational rearrangements and grand canonical nonequilibrium candidate Monte Carlo (GCNCMC) simulations to sample water configurations near the reactive residues. Next, representative conformations from classical MD trajectories were used to initiate hybrid quantum mechanical/molecular mechanical (QM/MM) finite-temperature string simulations to elucidate the thermodynamics and minimum free energy paths (MFEPs) of the relevant PCET and PT reactions. Finally, we applied PCET nonadiabatic rate theory to further clarify the mechanism and kinetics of this multisite, orthogonal PCET reaction.

Conformational and Hydration Changes Occurring upon W48 Oxidation.

To understand the conformational changes that occur upon oxidation of W48, we compared the conformational ensembles when W48 is in its standard form (Fig. 2A) and cationic radical form (Fig. 2B). Before oxidation, the cavity is relatively dry and tightly packed, as indicated by the small “empty volume” (blue region in Fig. 2C) and the relatively low number of cavity water molecules sampled from the GCNCMC simulations (top blue histogram and curve in Fig. 2E). Oxidation of W48 introduces a positive charge inside the cavity, while at the same time the Y122 radical becomes a standard Y122 and the H₂O ligand becomes an OH⁻ ligand bound to the di–iron cofactor (see Fig. 1B). The oxidized W48 forms a stronger hydrogen-bonding interaction with D237, as indicated by the slightly shorter distance between W48 and D237 (Fig. S1A). The electrostatic environment also allows water molecules to penetrate into the region, resulting in more water near the oxidized W48 (top orange histogram and curve in Fig. 2E). Concurrently, the size of the cavity expands, thereby accommodating the enhanced hydration level (blue region in Fig. 2D). The distance between D237 and D736 is another semi-quantitative descriptor of the cavity compactness and therefore can characterize the corresponding conformational changes (Fig. S1B). Upon oxidation of W48, conformations with a longer distance between D237 and D736 are observed with relatively high population, suggesting expansion of the cavity.

Fig. 2.

Changes in conformation and hydration upon oxidation of W48. Representative conformations from classical MD when (A) W48 is in the standard form and (B) W48 is oxidized in the cationic radical form. Residues other than D237, W48, E52, and Y356 are colored according to the residue types: positively charged (blue), negatively charged (red), polar uncharged (green), and hydrophobic (white). Water molecules are represented with oxygen atoms shown as orange spheres. Cavity ”empty volume” shown by blue region of overlapping spheres for representative conformations from classical MD when (C) W48 is in the standard form and (D) W48 is in the cationic radical form. (E) Number of water molecules within 6Å of W48 or Y356 from GCNCMC simulations when W48 is in the standard form (blue) and the cationic radical form (orange). Histograms represent the occurrence of a particular number of water molecules averaged over three replicas, with error bars indicating the standard deviation. The curves are Gaussian distributions fitted to the histograms. (F) Schematic depiction of the changes in conformation and hydration upon oxidation of W48. The hydration change and the flipping motion of the Y356 side chain are depicted.

These changes in hydration and cavity volume are accompanied by significant changes in the local structure. The conformational changes around Y356 are of particular significance. Y356 is located on the flexible $\beta$-tail region at the $\alpha /\beta$ interface and is directly involved in the PCET reaction across the interface. Before W48 oxidation, when the cavity remains hydrophobic, Y356 is mainly stabilized by R236 (Fig. 2A and F). The hydroxyl oxygen in Y356 forms a stable hydrogen-bonding interaction with the guanidinium group in R236 (Fig. S1C). Before oxidation, R236 is stabilized by the nearby D237, D736, and F46 backbone (Fig. S1D–F). When the cavity becomes more hydrated and expanded after W48 oxidation, R236, D237, D736, and F46 are stabilized by water to compensate for the loss of some of these other electrostatic interactions (Fig. 2B and Fig. S2). Moreover, the hydrogen-bonding interaction between Y356 and R236 is lost, and Y356 moves further away (Fig. S1C). Specifically, the Y356 side chain flips toward bulk solvent upon oxidation of W48, contributing to the cavity expansion (Fig. 2B and F). Note that these simulations do not provide information about the order in which these conformational changes occur.

The dominant rotamer of Y356 and the distance between W48 and Y356 change significantly upon W48 oxidation, demonstrating the conformational flexibility of Y356. The potentials of mean force (PMFs) obtained from metadynamics simulations (Fig. S3) show that the dominant conformation of Y356 before W48 oxidation (Fig. S3A) is at ${\chi }_1\approx 60°$, which corresponds to Y356 pointing away from the $\alpha$ subunit, located closer to W48. The second state, which has a lower population, is at ${\chi }_1\approx 310°$, which corresponds to Y356 pointing toward the $\alpha$ subunit, located closer to Y731 in the $\alpha$ subunit. Although interconversion between the two conformations is possible, the much larger population of the dominant conformation and the relatively high free energy barrier connecting them suggest that Y356 is likely to remain in the thermodynamically more favorable conformation with ${\chi }_1\approx 60°$ prior to W48 oxidation. This conformation is conducive to radical transfer from W48 to Y356. After W48 oxidation (Fig. S3B), although the dominant conformation of Y356 is still close to W48 (i.e, ${\chi }_1\approx 40°$), the population of the other conformation with ${\chi }_1\approx 310°$ increases and the barrier for the dihedral transition decreases, suggesting that the Y356 flipping motion may become more thermodynamically and kinetically favorable at this stage. Nevertheless, the thermodynamically favorable conformation of Y356 is still conducive to radical transfer from the oxidized W48. Notably, another metastable state at ${\chi }_1\approx 190°$, which corresponds to Y356 transitioning from pointing away from to pointing toward the $\alpha$ subunit, was also observed in the oxidized W48 state (Fig. S3B).

Proton Acceptor of Y356 in the Orthogonal PCET between W48 and Y356.

Analysis of the conformational changes occurring upon W48 oxidation also allowed us to identify the most likely proton transfer acceptor of Y356. The classical MD simulations revealed that, after its flipping motion, Y356 can form a stable hydrogen-bonding interaction with E52 in the $\beta$ subunit (Fig. 3A). To confirm that a short Y356–E52 hydrogen-bonding distance is the most stable conformation, we carried out metadynamics simulations using the distance between the phenolic hydroxyl oxygen of Y356 and the closer carboxylate oxygen of E52 as the collective variable. The PMFs obtained from these metadynamcis simulations clearly show that the distance corresponding to the most stable conformation shifts from 6.6 Å to 2.6 Å upon oxidation of W48 (Fig 3B). Therefore, E52 is a plausible proton acceptor of Y356 associated with radical transfer from W48 to Y356. A previous hypothesis based on mutagenesis studies and structural information from cryo-EM is that E52 is part of a water channel and may indirectly regulate the proton release from Y356 to the bulk solvent (45). However, our simulations support an alternative mechanism in which E52 acts as the proton acceptor from Y356 and thus directly participates in the PT reaction. The higher $\text{p}{K}_a$ of glutamic acid compared to H₃O⁺ would lead to a larger driving force and shorter equilibrium proton donor–acceptor distance for E52 compared to a nearby water, making the PT process more thermodynamically and kinetically favorable (48). Another well-established example of proton exchange between a glutamate and tyrosine residue, most likely involving a water network, occurs in cytochrome c oxidase, although the specific mechanism does not appear to be the same (49, 50).

Fig. 3.

Analysis of the orthogonal PCET mechanism for radical transfer from W48 to Y356. (A) A representative conformation sampled from classical MD when the radical is on W48, which is in its cationic radical form. (B) The PMF along the distance between the hydroxyl oxygen atom of Y356 and the closest carboxylate oxygen atom of E52, denoted as the Y356–E52 distance in the figure. The solid line shows the block-averaged PMF, and the shaded areas show the statistical error estimated from block analysis. (C) A schematic depiction of the orthogonal PCET mechanism involving W48, Y356, and E52. The red arrow indicates ET, and the blue arrow indicates PT. (D) The projected two-dimensional free energy surface obtained from QM/MM free energy simulations for the PCET reaction shown in panel C with the QM region containing these side chains. The MFEP for the converged string is shown as a black line. The PT coordinate is the difference between the distance between the Y356 hydroxyl oxygen and the transferring hydrogen and the distance between the E52 carboxylate oxygen and the transferring hydrogen. The Y356–E52 distance is the distance between the Y356 hydroxyl oxygen and the E52 carboxylate oxygen. (E) The free energy profile along the MFEP. The solid line represents the average value, and the shaded area shows the 95% confidence interval obtained from bootstrapped samples. Reaction progress represents points along the MFEP, where 0 and 1 denote the reactant and product states, respectively. (F) Spin densities calculated from representative conformations corresponding to the reactant and product states along the MFEP shown in E. Spin densities are depicted with an isovalue of 0.003 e/Bohr³.

Radical Transfer from W48 to Y356 through Orthogonal PCET.

After investigating the changes in conformations and hydration near the $\alpha /\beta$ interface and identifying the likely proton acceptor of Y356, we performed QM/MM finite temperature string simulations to study the orthogonal PCET reaction involving W48, Y356, and E52 (Fig. 3C). We used three reaction coordinates for the string simulation: the distance between the oxygen of the Y356 hydroxyl group and the transferring hydrogen, the distance between the oxygen of the E52 carboxylate group and the transferring hydrogen, and the distance between these two oxygen atoms. Upon convergence, we constructed the two-dimentional and three-dimensional (Fig. 3D and S8A) free energy surfaces and estimated the MFEP from the three-dimensional free energy surface. The one-dimensional free energy profile along the MFEP was also calculated (Fig. 3E). The free energy difference between the last and first points of this free energy profile represents the reaction free energy, $\Delta G°$, for radical transfer from W48 to Y356, and the free energy difference between the top of the barrier and the first point of the free energy profile represents the free energy barrier, $\Delta G^{‡}$. See SI for additional data and convergence analysis.

The free energy profile indicates that $\Delta G°=-1.4\pm 0.2$ kcal/mol and $\Delta G^{‡}=5.0\pm 0.1$ kcal/mol for radical transfer from W48 to Y356 (Fig. 3E). Note that these uncertainties represent only statistical errors for the computed data and do not account for systematic errors related to the level of theory and limitations of conformational sampling. The exoergic nature of the reaction suggests that the forward radical transfer from W48 to Y356 is thermodynamically favorable. Representative conformations corresponding to the reactant and product along the MFEP and the associated spin densities are shown in Fig. 3F. The spin density is found only on W48 in the reactant state and only on Y356 in the product state, consistent with an orthogonal concerted PCET mechanism in which electron transfer from Y356 to W48 occurs simultaneously with proton transfer from Y356 to E52 (Fig. 3C). As discussed below, this PCET reaction is vibronically nonadiabatic, and therefore the rate constant cannot be determined from the adiabatic free energy barrier.

The reversibility of the radical transport process is critical to the function of RNR. The cryo-EM structures of RNR (10, 11), as well as our previous simulations (34, 40) and the results herein, suggest that protein conformational changes occur between the forward and reverse radical transport processes. These findings indicate that differences in protein conformation and hydration may regulate the directionality of radical transfer. Given the short timescale of our QM/MM simulations, however, the slow conformational changes occurring between the forward and backward radical transport processes cannot be sampled. As the current simulations were initiated from the cryo-EM structure corresponding to the pre-turnover state, these results should be considered to be relevant only to forward radical transfer.

Conformational Motions of Y356 Before and After Radical Transfer from W48 to Y356.

Radical transfer from W48 to Y356 within the $\beta$ subunit is followed by interfacial radical transfer from Y356 in the $\beta$ subunit to Y731 in the $\alpha$ subunit. Our analysis above has revealed significant conformational flexibility across the interface. Therefore, it is important to examine the changes in the relative orientations of Y356 and Y731 during these processes. Our simulations indicate that a Y356–Y731 orientation favoring radical transfer between them becomes more populated as the radical approaches the interface. Before oxidation of W48, the hydroxyl group oxygen atoms of the two tyrosines are typically around 18 Å apart (Fig. S4A). Upon oxidation of W48, the two tyrosines become closer. Although the distance associated with the most populated conformation is still around 8.2 Å, a second conformation with a much shorter distance of 3.4 Å also has a notable population (Fig. S4B). After radical transfer from W48 to Y356, at which point Y356 is a neutral radical, Y356 can flip toward Y731 for subsequent radical transfer to the $\alpha$ subunit (Fig. S4D and Fig. S12).

A previous ENDOR spectroscopic study (44) revealed a distance between the hydroxyl oxygen atoms of radical Y356 and Y731 of around 8 Å, which is consistent with the corresponding distance in the cryo-EM structure (10, 11). A conformation with a much shorter distance of around 3.0 Å was also revealed in this ENDOR study, suggesting the possibility of direct PCET between Y356 and Y731. However, the experimental ENDOR data did not provide information regarding the relative populations of the different Y356 conformations. In our results, the dominant conformation sampled from metadynamics simulations with the radical on Y356 features a distance of around 6.4 Å between the Y356 and Y731 hydroxyl oxygen atoms (Fig. S4D and S12E). Although this distance is slightly shorter than that observed in the ENDOR study, one caveat is that both Y122 and Y731 were fluorine-substituted in the ENDOR study, whereas our simulations used the standard form for both tyrosines. The most productive conformation for direct PCET between radical Y356 and Y731, where the distance between the two hydroxyl oxygen atoms is around 2.9 Å, only has a minor population (Fig. S4D and Fig. S12E). It is likely that the productive conformation with the shorter oxygen–oxygen distance between radical Y356 and Y731 is relatively high in free energy, but the PCET reaction between Y356 and Y731 will occur quickly when the shorter distances are sampled.

Proton Transfer from W48 to D237.

Up to this point, our calculations and analysis of the orthogonal PCET reaction have been based on the assumption that a W48 cationic radical participates in the PCET reaction. However, the distance between the nitrogen on the indole side chain of W48 and the oxygen of the carboxylate group of D237 decreases upon W48 oxidation (Fig. S1A). The well-positioned geometries of W48 and D237 enable direct PT from the oxidized W48 to D237 (Fig. 4A). Given the feasibility of such a PT, it is important to determine whether W48 is in its cationic radical or neutral radical form when it undergoes PCET to Y356.

Fig. 4.

Analysis of the possible PT between W48^•+ and D237 with a water molecule in the QM region. (A) A schematic depiction of PT between W48^•+ and D237, where the question mark indicates that the feasibility of this reaction had not been established prior to this work. (B) The location and occupancy of water identified from clustering analyses based on GCMC simulations. Water molecules are represented with oxygen atoms shown as spheres. Occupancy is defined as the fraction of simulation frames in which the water site is occupied. Color changes from purple to blue to red with increasing occupancy. The water molecule included in the QM region of the QM/MM free energy simulations is circled and labeled. (C) The two-dimensional free energy surface obtained from QM/MM free energy simulations of PT between W48^•+ and D237 with the QM region containing these two side chains and the water molecule circled in panel B. The MFEP for the converged string is shown as a black line. The PT coordinate is the difference between the distance between the W48 indole nitrogen and the transferring hydrogen and the distance between the D237 carboxylate oxygen and the transferring hydrogen. The W48–D237 distance is the distance between the W48 indole nitrogen and the D237 carboxylate oxygen. (D) The free energy profile along the MFEP. The solid line represents the average value, and the shaded area shows the 95% confidence interval obtained from bootstrapped samples. Reaction progress represents points along the MFEP, where 0 and 1 denote the reactant and product states, respectively.

To directly answer this question, we performed QM/MM finite temperature string simulations of this PT step. Based on our GCNCMC simulations, a water molecule was identified to be in close proximity of both W48 and D237 with a high occupancy (Fig. 4B), and thus we included this water molecule in the QM region in our QM/MM simulations. We then constructed the multidimensional free energy surfaces and identified the MFEP (Fig. 4C and S9). The free energy profile along the MFEP (Fig. 4D) indicates that $\Delta G°=-0.7\pm 0.1$ kcal/mol and $\Delta G^{‡}=2.2\pm 0.1$ kcal/mol. As a comparison, if the GCNCMC-identified water molecule is not included in the QM region, $\Delta G°$ and $\Delta G^{‡}$ obtained from the free energy profile along the MFEP become 0.6 ± 0.1 kcal/mol and 2.9 ± 0.1 kcal/mol, respectively (Fig. S9 and S10). Again, the reported uncertainties reflect only statistical error from the computed data and do not reflect systematic errors. The different signs of $\Delta G°$ with and without the water in the QM region suggest that the thermodynamics of the reaction can be perturbed by the local environment, such as a nearby water molecule. However, the difference between their values, as well as the difference between the free energy barriers, is within the expected uncertainty for two independent finite-temperature string simulations with different starting conformations.

The magnitude of $\Delta G°$ for this PT reaction is sufficiently small to imply that the reaction is nearly isoergic. Because $\Delta G°$ is similar to $k_{B}T$, both the cationic W48 radical and the neutral W48 radical are expected to have non-negligible populations at equilibrium. In addition, both the forward and backward reactions are predicted to be kinetically fast given the low free energy barrier. Thus, the cationic and neutral radical forms of W48 are expected to interconvert rapidly in this system. Another perspective is that the interaction between the W48 radical cation and D237 could be a low-barrier hydrogen bond (51, 52).

Both the tryptophan cationic radical and the tryptophan neutral radical have higher standard reduction potentials $(E°\prime )$ than does tyrosine and hence can oxidize tyrosine under physiological conditions (32, 53). However, $E°\prime$ of the tryptophan cationic radical is higher than that of the tryptophan neutral radical, indicating that the cationic radical is a stronger oxidizing agent, as expected for a positively charged species compared to a neutral species. Therefore, the driving force for the oxidation of Y356 should be greater when W48 is a cationic radical than when it is a neutral radical.

In addition, less favorable conformations for radical transfer from W48 to Y356 were frequently observed when W48 is a neutral radical, in conjunction with a protonated D237. For example, the distance between Y356 and E52, which are hypothesized to undergo PT during radical transfer from W48 to Y356, is longer in the simulations with the neutral W48 radical (Fig. S6C versus Fig. S6B and Fig. S11B). Specifically, the distance between the proton donor and acceptor oxygen atoms of Y356 and E52 for the conformation with the highest population for the W48 neutral radical is 4.0 Å, compared to 2.6 Å for the W48 cationic radical. The Y356 ${\chi }_1$ angle of this dominant conformation for the W48 neutral radical is around 312° (Fig. S3C), indicating that the hydroxyl group points toward the $\alpha$ subunit and away from E52 (Fig. S11B). The other highly-populated conformation has a distance of 5.6 Å, corresponding to the rotation of the carboxylate group of E52. This longer distance prevents direct PT from Y356 to E52, and thus the PT between Y356 and E52 would most likely need to be mediated by a water molecule (Fig. S11A). Although there is a third conformation for the W48 neutral radical with a shorter distance of 2.7 Å (Fig. S6C and S11B), its population is significantly smaller compared to the two other conformations, suggesting that direct PT from Y356 to E52 is much less likely when W48 is in its neutral form.

Next, we consider the subsequent radical transfer from Y356 to Y731. The conformation with the shortest distance of 3.4 Å between the hydroxyl oxygen atoms of Y356 and Y731 has a much lower population when W48 is in its neutral radical form compared to when W48 is in its cationic radical form (Fig. S4C and Fig. S11C). This finding suggests that Y356 and Y731 can rarely form a reactive geometry when W48 is a neutral radical. Combining all these parts of the analysis, although direct PT from W48 to D237 is feasible, the PCET reaction between W48 and Y356, and possibly the subsequent PCET reaction between Y356 and Y731, is more likely to occur when W48 is in its cationic radical form. The radical form of W48 can be further probed by incorporating unnatural tryptophan analogs to trap the radical on W48, together with UV–Vis spectroscopy and redox potential measurements.(32, 54–56).

Kinetics of Orthogonal PCET between W48 and Y356.

We used our theoretical framework for PCET (23, 33, 57–60) to analyze the PCET reaction involving W48, D237, and E52 in the $\beta$ subunit of RNR. Analytical expressions (23, 57, 58) for PCET rate constants have been derived for various well-defined regimes. The first step is to use our systematic classification scheme (33) to determine which form of the PCET rate constant is appropriate for this system. The calculated parameters, such as the vibronic couplings and adiabaticity parameter (61, 62), are given in Table S4 of the SI. Our analysis suggests that this PCET reaction is vibronically and electronically nonadiabatic. The vibronic coupling is much smaller than the thermal energy, $k_{\text{B}}T$, indicating vibronic nonadiabaticity. Moreover, the electron-proton adiabaticity parameter is much smaller than unity, indicating electronic nonadiabaticity.

In this regime, the rate constant can be calculated using the golden-rule expression given by

$$k(R)=\sum _{\mu }{P}_{\mu }\sum _{\nu }\frac{{\left|V_{\mu \nu }\right|}^2}{\hslash }\sqrt{\frac{\pi }{\lambda k_{\text{B}}T}}\text{exp}\left[-\frac{{\left(\Delta G_{\mu \nu }^{\text{o}}+\lambda \right)}^2}{4\lambda k_{\text{B}}T}\right]$$ [1a]

$$V_{\mu \nu }=V^{\text{el}}{S}_{\mu \nu }$$ [1b]

$$\Delta G_{\mu \nu }^{°}=\Delta G°+{\epsilon }_{\nu }-{\epsilon }_{\mu },$$ [1c]

where $\mu$ and $\nu$ represent the vibronic states of the reactant and product, respectively, $P_{\mu }$ is the Boltzmann population of the reactant vibronic state $\mu ,\lambda$ is the total reorganization energy, $\Delta G°$ is the reaction free energy for the ground vibronic states, ${\epsilon }_{\mu }$ and ${\epsilon }_{\nu }$ are the energies of the reactant and product vibronic states relative to the energies of their respective ground vibronic states, $\hslash$ is the reduced Planck constant, $k_B$ is the Boltzmann constant, and $T$ is the temperature. In the electronically nonadiabatic limit, the vibronic coupling is the product of the electronic coupling, $V^{\text{el}}$, and the overlap integral, $S_{\mu \nu }$, of the reactant and product proton vibrational wave functions. The overall PCET rate constant is obtained by thermal averaging over the proton donor-acceptor distance $R$,

$$k^{\text{PCET}}={\int }_0^{\infty }k\left(R\right)P\left(R\right)\text{d}R,$$ [2]

where $P\left(R\right)$ is the probability distribution function for $R$. Note that quantitative calculations of the absolute energetics for reactions in RNR are not possible with current computational methods, and therefore these results should be viewed in terms of providing qualitative insights. Details about how to compute the input quantities needed for the calculation of the PCET rate constant are described in Methods and in the SI.

The mechanism of this orthogonal, concerted PCET reaction is shown in Fig. 3C. The electron transfers from Y356 to W48^•+, and the proton transfers from Y356 to E52. The QM/MM free energy simulations show that the reaction free energy $\Delta G°$ for this reaction is −1.4 kcal/mol (Fig. 3E). We then computed the diabatic proton potential energy curves and associated proton vibrational wavefunctions for a model system composed of the side chain atoms of Y356 and E52. In this model system, W48 is not included explicitly but rather is treated as an external oxidant for the purposes of generating the proton potential energy curves. In this case, the reactant corresponds to the negatively charged Y356-E52 model system due to the carboxylate anion on E52, and the product corresponds to the neutral Y356-E52 system, where electron transfer (ET) from Y356 to W48^•+ has occurred.

We calculated the proton potential energy curves for the Y356-E52 model system for a series of proton donor–acceptor distances $R$. The initial conformation was obtained from an image at the top of the free energy barrier in the QM/MM string simulations. This geometry was further optimized for the calculations of the proton potential energy curves following the protocol given in the SI. Fig. 5A shows the average reactant/product structure obtained at $R=2.53$ Å and the proton potential energy curves for the reactant and product diabatic states at different $R$ values. The minimum of each proton potential energy curve is on the proton donor or acceptor side for the reactant and product, respectively. The proton potential energy curves depend strongly on $R$, ranging from a single-well potential at small $R$ to an asymmetric double-well potential at larger $R$. The barrier height and separation between the two minima increase as $R$ increases. These proton potential energy curves were used to compute the proton vibrational wave functions and energy levels, providing the Boltzmann populations $P_{\mu }$, the energy levels ${\epsilon }_{\mu }$ and ${\epsilon }_{\nu }$, and the overlap integrals $S_{\mu \nu }$ required to compute the vibronic coupling (Fig. S13 and S14, Table S5). The proton vibrational wavefunctions and associated overlap for the $(\mu ,\nu )=\left(\mathrm{0,0}\right)$ pair of vibronic states at $R=2.53$ Å are shown in Fig. 5B.

Fig. 5.

Kinetic analysis of the nonadiabatic PCET reaction between W48^•+ and Y356, as shown in Fig. 3C. (A) Upper: The gas-phase model used to compute the proton potential energy curves. Only Y356 and E52 are included, and the reactant and product correspond to the negatively charged and neutral systems, respectively, representing ET from Y356 to W48^•+. The average reactant/product structure is shown for $R=2.53$ Å. The dashed black line illustrates the proton transfer coordinate axis, which is defined as the line connecting the optimized positions of the proton on the donor and acceptor oxygen atoms. Lower: Proton potential energy curves for the reactant (left) and product (right) diabatic states computed with DFT at different proton donor–acceptor distances $R$. The negative and positive values of the proton transfer axis $r_{\text{p}}$ correspond to the proton closer to the Y356 hydroxyl oxygen or the E52 carboxylate oxygen, respectively. (B) Diabatic proton potential energy curves for the reactant (blue) and product (red) with their corresponding ground-state proton vibrational wave functions at the proton donor–acceptor distance $R=2.53$ Å. The curves are shifted to align their zero-point energy levels. (C) $k\left(R\right)$, $P\left(R\right)$, and their product $k\left(R\right)P\left(R\right)$, which is the integrand to calculate the thermally averaged rate constant, $k^{\text{PCET}}$. The y-axis has no units, as all quantities have different units and are scaled to be displayed on the same plot.

We used a different model system to compute the reorganization energy and electronic coupling, assuming that these quantities are dominated by ET. This model system represents the ET reaction from Y356 to W48^•+ in the absence of PT from Y356 to E52. Note that Y356^•+ is not a stable species during PCET and was used only in this model system to compute the reorganization energy and electronic coupling. The PCET reorganization energy is dominated by ET because the ET distance is significantly longer than the PT distance. Moreover, we do not expect the reorganization energy or the electronic coupling to depend significantly on the position of the transferring proton.

To compute the reorganization energy, distributions of energy gaps between the relevant diabatic states and the corresponding free energy profiles (Fig. S15A and Table S3) were obtained by performing constrained density functional theory with configuration interaction (CDFT-CI) single point calculations on conformations sampled with classical MD. The CDFT-CI calculations were performed using QM/MM electrostatic embedding to include the effects of the protein and solvent. The QM region consisted of the side chain atoms of W48 and Y356, and the MM region included the remainder of the protein and solvent atoms, which were represented as MM point charges. The free energy profiles were constructed as functions of the energy gap using the average value of the variances obtained from Gaussian fits to the energy gap distributions (upper panel in Fig. S15A). Using methods discussed in the SI, the reorganization energy, $\lambda$, was determined from the two free energy profiles to be around 25.5 kcal/mol. A reorganization energy of 25 kcal/mol for this multisite, orthogonal PCET reaction seems reasonable given that the electron is transferring around 10 Å, whereas the proton is transferring only a few tenths of an Angstrom. As mentioned above, the reorganization energy is dictated primarily by the ET reaction over this relatively long distance. The physical basis for this reorganization energy is mainly the changes in the electrostatic environment due to conformational motions of the protein and solvent.

The magnitude of the electronic coupling between the relevant diabatic states, $\left|V^{\text{el}}\right|$, was plotted against the energy gap, $\Delta E$ (Fig. S15B). Because the electronic coupling at the crossing point (i.e., where $\Delta E=0$) is the most relevant coupling for ET within Marcus theory and the golden-rule formalism, we considered the electronic coupling distribution for the energy gap between −10 and 10 kcal/mol. The range of this distribution is 0.06 −1.21 kcal/mol, with the most probable value of 0.36 kcal/mol and a median value of 0.63 kcal/mol. Note that this estimate of the electronic coupling is associated with considerable uncertainty.

All of these input quantities were used to compute the vibronically nonadiabatic PCET rate constant at each $R$ value using Eq. 1. The $k\left(R\right)$ curve in Fig. 5C shows that the rate constant increases dramatically as $R$ decreases, mainly because the overlap integral increases as the proton donor and acceptor become closer to each other (Fig. S14 and Table S5). The probability distribution $P\left(R\right)$ and the product $P\left(R\right)k\left(R\right)$ are also shown in Fig. 5C. The equilibrium proton donor–acceptor distance, corresponding to the maximum of $P\left(R\right)$, is 2.61 Å (gray dashed line in Fig. 5C), whereas the dominant proton donor–acceptor distance for the PCET reaction, corresponding to the maximum of the product $P\left(R\right)k\left(R\right)$, is 2.52 Å (black dashed line in Fig. 5C). Conformational fluctuations of $R$ are essential to sample the shorter distances that enable effective hydrogen tunneling via a sufficiently large proton vibrational wavefunction overlap integral (Fig. 5B). The overall PCET rate constant, $k^{\text{PCET}}$, is estimated to be 1.6 × 10⁷ s⁻¹, although the various approximations lead to considerable uncertainty in this quantitative value.

We then analyzed the contributions from reactant/product pairs of vibronic states to the PCET rate constant at different $R$ values (Table S5). We found that for the $R$ values near the dominant proton donor–acceptor distance of 2.52 Å, the $(\mu ,\nu )=\left(\mathrm{0,0}\right)$ pair is the dominant pair of vibronic states. The contribution from this pair decreases from 93% to 70% as $R$ increases from 2.43 to 2.53 Å. Moreover, for the even longer equilibrium $R$ value of 2.63 Å, the contribution of the $(\mu ,\nu )=\left(\mathrm{0,0}\right)$ pair further decreases to 25%, and the dominant pair of vibronic states becomes the $(\mu ,\nu )=\left(\mathrm{1,0}\right)$ pair with a contribution of 65%. Thus, excited vibronic states play a small but non-negligible role in this PCET reaction.

Conclusion

Our multiscale simulations reveal several key mechanistic features of multisite, orthogonal PCET in the $\beta$ subunit of E. coli class Ia RNR. Although W48 consistently lies between Y122 and Y356 across crystallographic and cryo-EM structures, direct evidence for its participation in radical transport has remained elusive. Our simulations support its involvement in the radical transport pathway. We find that forward radical transfer from W48 to Y356 is thermodynamically favorable and that this process preferentially occurs when W48 adopts the cationic radical form. Due to the conformational flexibility of interfacial residues and the nearby water channel, the proton acceptor of Y356 has yet to be clearly identified, with only indirect experimental hints. The extensive conformational sampling from our metadynamics simulations and our QM/MM free energy simulations identify E52 as the likely proton acceptor for Y356 in the PCET reaction associated with forward radical transfer from W48 to Y356.

More generally, we find that Y356 undergoes distinct conformational rearrangements during different stages of the radical transport process in the $\beta$ subunit. Its conformational motions are strongly influenced by nearby charged residues, particularly R236, as well as the local hydration environment. Recent cryo-EM studies have emphasized the importance of conformational dynamics in regulating radical transport in RNR (47). In conjunction with experimental data, our work provides insights into these processes by elucidating how conformational motions and hydration regulate multisite, orthogonal PCET in the $\beta$ subunit.

Additionally, we used vibronically nonadiabatic PCET theory to calculate the rate constant for the orthogonal PCET reaction involving W48, Y356, and E52. Our kinetic analysis shows that this reaction is electronically and vibronically nonadiabatic. Our analysis also shows that relatively short proton donor–acceptor distances accessed through conformational fluctuations are essential for effective hydrogen tunneling. Overall, the computational strategies developed here are also broadly applicable to other enzymatic systems that exhibit multisite PCET mechanisms.

Although this study provides useful insights, a complete understanding of how conformational fluctuations, hydration changes, and proton transport enable the long-range radical transport process in RNR will require further computational and experimental studies. In particular, future work could be aimed at elucidating the driving forces for the conformational changes that regulate the PCET reactions, specifically the conformational changes of Y356, and understanding how the Y356 radical is stabilized in its different conformational states and during the transitions between them.

Understanding how RNR controls radical transfer provides fundamental insights into enzyme design principles governing long-range radical transport and radical chemistry in biology (13, 32, 63–66). Beyond the specific case of RNR, these results offer broader insights into how enzymes orchestrate proton and electron transfer across large distances, balancing reactivity and regulation through a combination of local hydrogen-bonding rearrangements, solvation dynamics, and conformational gating. These insights are generally applicable to a wide range of enzymes and redox-active systems that rely on PCET and radical transport for function. This enhanced understanding provides the foundation for designing biological systems that harness PCET for catalysis and regulation.

Materials and Methods

Unbiased MD Simulations of Different Radical States.

Unbiased MD simulations were conducted with the Amber24 (67) package to compare key structural properties in multiple radical and/or protonation states. These states include W48 in its standard form, neutral radical form, and cationic radical form, as well as Y356 in its standard form, tyrosyl radical form, and cationic radical form. In these simulations, E52 is deprotonated, and D237 is deprotonated except when W48 is a neutral radical, at which point D237 is protonated because it is presumed to have accepted the proton from W48. The MD setup was the same as that used in our previous work on RNR (36). The MD trajectories were analyzed with the MDAnalysis package (68, 69). The trajectories were visualized with VMD (70) and PyMOL (71). Other simulation details and additional data analysis for the unbiased trajectories are provided in the SI.

GCNCMC Simulations.

GCNCMC simulations (72, 73) were performed with the grand (73) package to sample the water configurations and number of water molecules near selected residues. The simulation details and additional data analysis are provided in the SI.

Metadynamics Simulations.

Multiple-walker well-tempered metadynamics simulations (74–76) were carried out using the PLUMED (77, 78)-OpenMM (79) interface. To probe the conformational changes and flexibility of Y356, the ${\chi }_1$ angle of Y356 and the distance between the hydroxyl oxygen atoms of Y356 and Y731 were chosen as the two collective variables. Additionally, to confirm that E52 is a plausible proton acceptor of Y356 associated with radical transfer from W48 to Y356, the distance between the Y356 hydroxyl oxygen atom and the E52 carboxylate oxygen atom was chosen as the collective variable. Simulation details and additional data analysis are provided in the SI.

QM/MM Finite Temperature String Simulations with Umbrella Sampling.

We performed QM/MM free energy simulations (80, 81) of the orthogonal PCET reaction involving W48, Y356, and E52 and the PT reaction between oxidized W48 and D237. These QM/MM finite temperature string simulations (82) with umbrella sampling (83) were performed using the Amber/Q-Chem interface (67, 84, 85). The spin densities of the reactant and product states for the PCET reaction were calculated using Q-Chem 6.2.2 (84). Simulation details and additional data analysis are provided in the SI.

PCET Kinetics.

Calculations of the nonadiabaticity parameters and vibronic couplings (33, 37, 61) were performed to probe the electronic and vibronic nondiabaticity of the orthogonal PCET reaction. The CDFT-CI method (86–89) with electrostatic embedding of the MM charges was used to construct the diabatic states and to compute the reorganization energy and electronic coupling for ET between W48^•+ and Y356 using Q-Chem 6.2.2 (84). Calculations of the proton potential energy curves for obtaining the proton vibrational wave functions and energy levels associated with PT from Y356 to E52 upon ET from Y356 to W48, as well as the corresponding nonadiabatic PCET rate constant, followed the protocol given in Ref. (90). Computational details and additional data analysis are provided in the SI.

Supplementary Material

Supporting Information Appendix (SI). Additional simulation and calculation details, as well as supplementary results are included in the SI.

Significance Statement.

Efficient and regulated proton and electron movement underlies many life-sustaining enzymatic processes. The enzyme ribonucleotide reductase, which is critical for DNA synthesis and repair, relies on a precisely tuned network of redox-active residues that transfer a radical to the catalytic site through a series of proton-coupled electron transfer steps. Using multiscale simulations, we reveal the molecular mechanism of a key step in which the electron and proton move concertedly between different sites in different directions. Our simulations show that conformational changes within the enzyme and hydration by interfacial water molecules modulate this process. These atomistic-level mechanistic insights advance the understanding of how enzymes control the movement of electrons and protons and inform strategies for rational therapeutic design and protein engineering.