Function of the diiron cluster of Escherichia coli class Ia ribonucleotide reductase in proton-coupled electron transfer

Abstract

The class Ia ribonucleotide reductase (RNR) from Escherichia coli (Ec) employs a free-radical mechanism, which involves bidirectional translocation of a radical equivalent or “hole” over a distance of ∼35 Å from the stable diferric/tyrosyl-radical (Y₁₂₂•) cofactor in the β subunit to cysteine 439 (C₄₃₉) in the active site of the α subunit. This long-range, inter-subunit electron transfer occurs by a multi-step “hopping” mechanism via formation of transient amino acid radicals along a specific pathway and is thought to be conformationally gated and coupled to local proton transfers. Whereas constituent amino acids of the hopping pathway have been identified, details of the proton-transfer steps and conformational gating within the β sununit have remained obscure; specific proton couples have been proposed, but no direct evidence has been provided. In the key first step, the reduction of Y₁₂₂• by the first residue in the hopping pathway, a water ligand to Fe₁ of the diferric cluster was suggested to donate a proton to yield the neutral Y₁₂₂. Here we show that forward radical translocation is associated with perturbation of the Mössbauer spectrum of the diferric cluster, especially the quadrupole doublet associated with Fe₁. Density functional theory (DFT) calculations verify the consistency of the experimentally observed perturbation with that expected for deprotonation of the Fe₁-coordinated water ligand. The results thus provide the first evidence that the diiron cluster of this prototypical class Ia RNR functions not only in its well-known role as generator of the enzyme's essential Y₁₂₂•, but also directly in catalysis.

Introduction

Ribonucleotide reductases (RNRs) catalyze the reduction of ribonucleotides to deoxyribonucleotides in all organisms, thereby providing and controlling the only de novo pathway to the four precursors required for DNA replication and repair.^1,2 RNRs use a free-radical mechanism, in which a transient cysteine thiyl radical (C•) in the active site of the enzyme initiates substrate reduction by abstraction of a hydrogen atom (H•) from C3′.^1,3-9 In class Ia RNRs, such as the RNR from aerobically growing Escherichia coil (Ec), a stable tyrosyl radical (Y₁₂₂• in the Ec ortholog) in close proximity to a μ-oxo-(Fe^III)₂ cluster in the β subunit^10-14 reversibly generates the transient C• in the active site of the enzyme's α subunit in the functional α₂β₂ complex.^5-7,15,16 A model of the complex, constructed by computer docking of the structures of the individual subunits⁷ and subsequently validated by electron-electron double resonance spectroscopic experiments,^17-19 suggests a distance of ∼35 Å between Y₁₂₂• in β and the H•-abstracting C₄₃₉ in α. Electron transfer (ET) between C₄₃₉ and Y₁₂₂ by a single tunneling step over that distance would be far too slow to account for the enzyme's turnover rate (2-10 s^-1).^15,20 Instead, this long-range inter-subunit ET is mediated by a chain of strictly conserved aromatic amino acids, which form transient radicals in a “hopping” mechanism (Scheme 1A).^7,15,21-28 Direct detection of these pathway radicals in the wild-type enzyme has been hampered by a preceding rate-limiting conformational change within the α₂β₂ complex.²⁰ This slow conformational change, which occurs upon binding of substrate and allosteric effector to α and allows for translocation of the radical from its resting location on β-Y₁₂₂ to where it functions in catalysis on α-C₄₃₉, masks the subsequent, fast chemical events.²⁰ Substitution of pathway tyrosines by unnatural amino acids with altered redox properties led to the first detection of pathway radicals and provided the most direct evidence that these residues are redox-active “stepping stones” in the long-range ET (Scheme 1A).^21-26 The individual ET hopping steps in the overall 35-Å hole-translocation process are thought to be coupled to multiple short-range proton transfer steps (i.e., proton-coupled ET or PCET), which effectively tune the thermodynamics of the component steps for efficiency and reversibility of the overall process.^15,29 The coupling of ET to proton transfer (PT) steps could therefore permit radical translocation to be controlled by the PT steps, which could, due to their more stringent distance and orientation requirements, be controlled by the conformation of the protein: engagement of a proton-transfer pathway upon substrate binding could be the basis for the conformational gating in RNR.

Scheme 1.

Proposed radical translocation pathway (A) and schematic representation of the two approaches used to trap the forward radical translocation product in Ec RNR (B) and (C). In (A) W₄₈ and D₂₃₇ are shown in gray because there is currently no direct evidence for involvement of W₄₈ in radical translocation. Trapping of the forward radical translocation product induced by N₃UDP in the wild-type α₂β₂ complex is shown in (B), and by CDP in a complex with an α variant containing the radical-stabilizing unnatural amino acid 3-aminotyrosine (NH₂-Y) at residue α-Y₇₃₁ (C). EPR- and Mössbauer-observable species are illustrated in (B) and (C) by red and blue outlines, respectively.

Whereas the radical-hopping nature of the process and the identities of the mediating residues have (with the exception of tryptophan 48 in β, which may or may not be a mediator) been established, much less is known about the details of the proton transfers. It has been proposed that PT proceeds orthogonally to ET in the β subunit and collinearly in α.^15,29 A recent electron-nuclear double resonance (ENDOR) spectroscopic study provided evidence for hydrogen bonds among the three ET pathway residues of α in the active α₂β₂ complex, consistent with collinear PCET.³⁰ Within the β subunit, specific proton coupling partners have been proposed (Scheme 1A), but little experimental evidence has been provided.^27,31 Specifically, in the first step of forward (β-Y₁₂₂• → α-C₄₃₉) radical translocation, the neutral Y₁₂₂• is reduced to the neutral Y₁₂₂ by β-Y₃₅₆ (or perhaps β-W₄₈) in the pathway, and it has been suggested that the proton required to maintain neutrality of the Y₁₂₂ side chain is delivered orthogonally to ET by the water ligand on the iron ion in site 1 (Fe₁).^15,32 In this study, we have trapped the cofactor in β in its product state of the forward radical translocation process by using either the radical-trapping substrate analog 2′-azido-2′-deoxyuridine 5′-diphosphate (N₃UDP) with the wild-type enzyme or a natural substrate (CDP) with a variant of the α-subunit containing the radical-stabilizing unnatural amino acid 3-aminotyrosine (NH₂-Y) at the subunit-interfacial pathway residue α-Y₇₃₁ (Scheme 1B and C). We show that the Mössbauer spectrum of the (Fe^III)₂ cluster, especially the quadrupole splitting parameter (ΔE_Q) associated with Fe₁, changes while the oxidation state of the cluster remains unchanged and that this spectral perturbation is specific to the form of the enzyme that has engaged in forward radical translocation. The nature of the observed perturbation – a ∼ 0.5 mm/s diminution in |ΔE_Q| of Fe₁ with much smaller changes to |ΔE_Q| of Fe₂ and the isomer shifts (δ) of both sites – agrees remarkably well with the effect predicted by simple DFT calculations for removal of a proton from the Fe1-OH₂ ligand. The results provide the first direct evidence that the diiron cluster of the prototypical class Ia RNR from Ec not only serves its well-known role as generator of the Y₁₂₂•^11,15,33 but also actively participates in the enzyme's catalytic cycle.

Results and Discussion

To initiate forward radical translocation and trap the enzyme in the product state of this step, the α and β subunits were incubated in the presence of the positive allosteric effector, thymidine triphosphate (TTP), with the substrate analog N₃UDP, which brings about the irreversible reduction of Y₁₂₂• in β along with the formation of a meta-stable, nucleotide-based, nitrogen-centered radical (N^•) in the active site of α (Scheme 1B).^34-36 Conversion of β-Y₁₂₂• to the N• was confirmed by comparison of the X-band EPR spectrum at 14 K of the reaction sample (Fig. 1 A, red) to that of a control sample from which N₃UDP was omitted (Fig. 1 A, green). Subtraction of the features of the unreacted β-Y₁₂₂• from the spectrum of the N₃UDP-treated sample yields the spectrum of the N• (Fig. 1A).^36,37 Its intensity accounts for ∼36% of the Y₁₂₂• originally present in the control (− N₃UDP) sample. Conversions of ≤ 50% have generally been observed in such experiments with Ec RNR and have been attributed to the facile reaction of only one αβ pair in the α₂β₂ hetero-tetramer (“half-of-sites reactivity”), a property thought to be intrinsic to the enzyme. ^17,23-26

Figure 1.

EPR and Mössbauer spectra of Ec RNR before and after N₃UDP-induced trapping of the product of forward radical translocation. (A) EPR spectra of uniformly ⁵⁷Fe labeled α₂β₂ with TTP and in the absence (green) or presence (red) of N₃UDP. Subtraction of the features of the unreacted Y₁₂₂^• (green) from the spectrum of the N₃UDP-treated sample (red) yields the spectrum of the N• (blue). 4.2-K/53-mT Mössbauer spectra of uniformly ⁵⁷Fe labeled (B), site 1 ⁵⁷Fe enriched (C) and site 2 ⁵⁷Fe enriched (D) α₂β₂ with TTP and in the absence (I) or presence (II) of N₃UDP. The solid lines in I are simulations of the two quadrupole doublets (red and blue) of the resting (Fe^III)₂ cluster with parameters quoted in the text and the sum of the two doublets (black). The solid line in II is the spectrum from I scaled appropriately to remove the contribution of the unreacted cofactor, and subtraction of this contribution from II reveals the spectrum of the product of the forward radical translocation (III), which can be simulated with two quadrupole doublets (red and blue; parameters quoted in the text). The sum of the two quadrupole doublet simulations is shown as a black line. The spectrum in IV is the total difference of I-II and the solid line is a simulation of the difference spectrum as a sum of the four quadrupole-doublet components.

The changes to the (Fe^III)₂ cluster in β upon this N₃UDP-induced forward radical translocation were monitored by Mössbauer spectroscopy on two samples identical to the aforementioned EPR samples. The 4.2-K/53-mT Mössbauer spectrum of the control sample in the absence of N₃UDP [i.e., of the resting (Fe^III)₂ cluster]^38-40 consists of two resolved quadrupole doublets with parameters [isomer shift, δ₁, of 0.46 mm/s and quadrupole splitting parameter, ΔE_Q1, of 2.43 mm/s (Fig. 1B,I, red), and δ₂ = 0.54 mm/s, ΔE_Q2 = 1.63 mm/s (Fig. 1B, I, blue)] nearly identical to those reported for active β alone and previously assigned to Fe₁ and Fe₂, respectively.^11,39,41-42 Incubation with N₃UDP changes the Mössbauer spectrum significantly, and the spectrum exhibits new features (Fig. 1B, II, and clearly apparent in the difference spectrum, IV). Removal of the contribution of the unreacted (Fe^III)₂ cluster by subtraction of the appropriately scaled spectrum of the control (− N₃UDP) sample (solid line) resolves the new spectrum of the radical translocation product (Fig. 1B, III). The new features account for ∼41% conversion of the active cofactor, which agrees well with the conversion of Y₁₂₂• to N• determined by EPR (see also Table S2). The derived spectrum of the radical translocation product can be analyzed as two symmetrical quadrupole doublets with equal intensity, equal line width, and parameters δ₁ = 0.43 mm/s, ΔE_Q1 = 1.89 mm/s, and δ₂ = 0.57 mm/s, ΔE_Q2 = 1.76 mm/s (Fig. 1B, III, red and blue, respectively; Table S1).⁴³ The modest changes to the isomer shifts of both sites and the fact that the spectrum still comprises quadrupole doublet features indicative of an integer-electron-spin ground state imply that no change in oxidation state of the (Fe^III)₂ cluster is effected by the N₃UDP treatment,⁴⁴ as expected from previous studies employing this compound.^34-36 In addition, the large magnitudes of the ΔE_Q-values further suggest that the oxo bridge remains intact.⁴⁵

The spectra of control samples containing β, effector, either CDP or N₃UDP, and an α variant having the hopping pathway disabled by substitution of the subunit-interfacial Y₇₃₁ in α with F lack these new features and are essentially identical to the spectrum of the sample with wild-type α and β before reaction with N₃UDP (Fig. S2). The spectrum of an additional control sample, in which the Y₁₂₂• in β was reduced in the absence of α with hydroxyurea (HU) to yield the inactive (Fe^III)₂/Y₁₂₂ met form (Fig. S2),⁴⁶ is also very similar to that of the active form. These results imply that the observed perturbation to the spectrum of the (Fe^III)₂ cluster in the wild-type enzyme caused by N₃UDP is related to radical translocation and not to either the absence of Y₁₂₂• per se or nucleotide binding events.

The two quadrupole doublets that make up the spectrum of the radical translocation product were unambiguously assigned to Fe₁ and Fe₂ by site-selective labeling of β with ⁵⁷Fe and ⁵⁶Fe. The two sites in Ec β have different affinities for Fe^II, and this property can be exploited to obtain β with the Mössbauer active ⁵⁷Fe enriched in one or the other site.⁴¹ The spectrum of a sample of the complex prepared with β subunit having site 1 enriched with ⁵⁷Fe can be simulated with the same parameters used for spectra of the uniformly ⁵⁷Fe labeled samples above, and the relative intensities of the two quadrupole doublets indicate that ∼73% of the ⁵⁷Fe resides in site 1 and ∼27% in site 2 (Fig. 1C, I, red and blue). Treatment of this sample with N₃UDP yielded a β-Y₁₂₂• to N• conversion of ∼38%, as determined by EPR spectroscopy (Fig. S3, Table S2), similar to that achieved with uniformly ⁵⁷Fe-labeled complex. The Mössbauer spectrum of the radical translocation product, obtained after subtraction of the unreacted component (Fig. 1C, II and III), constitutes 38% of the active cofactor, in agreement with the EPR quantification. It can be simulated with the same parameters used for the spectrum of the product in the uniformly labeled complex (Fig. 1C, III, red and blue). Owing to the site-selective labeling, the two quadrupole doublets have different intensities and can therefore be unambiguously assigned to Fe₁ (red) and Fe₂ (blue). Samples with ⁵⁷Fe enriched in site 2 confirm this assignment (Fig. 1D). These samples contain ∼24% of the ⁵⁷Fe in site 1 and ∼76% in site 2 (Fig. 1D, I, red and blue). Reaction with N₃UDP resulted in conversion of ∼47% of initial β-Y₁₂₂• to N•, as quantified by EPR (Fig. S3, Table S2), and produced the Mössbauer spectrum shown in Fig. 1D, II. The Mössbauer spectrum of the radical translocation product, obtained after removal of the unreacted component (Fig. 1D, III) and accounting for 52% conversion of the active cofactor, can again be simulated with the same parameters used for the spectrum of the product with the uniformly and site-1-enriched complexes (Fig. 1D, III, red and blue). Assignment of the two quadrupole doublets to Fe₁ (red) and Fe₂ (blue) shows that the spectrum of Fe₁, the site with the water proposed to be the proton donor, changes much more (|ΔE_Q| decreases by 0.5 mm/s) than that of Fe₂ upon N₃UDP-induced trapping of the radical translocation product.

In addition to the use of the radical-trapping substrate analog, N₃UDP, the product of forward radical translocation can also be trapped by using an α variant that contains an unnatural amino acid incorporated into the hopping pathway (Scheme IC). The radical produced from 3-aminotyrosine (NH₂Y•) has a reduction potential estimated to be ∼190 mV less than that of Y•, causing the radical to reside on this unnatural residue during catalysis by Ec RNR variants.^23,47 The variant α protein having NH₂Y in place of Y₇₃₁ is capable of nucleotide reduction, and, upon its incubation with the β subunit, the effector ATP and the CDP substrate, a NH₂Y• accumulates to ∼50% of the total spin in the sample.²³ The dependence on the presence of nucleotides and β, and the ability of this enzyme to catalyze deoxynucleotide production, show that the NH₂Y• forms by gated radical translocation in the functional holoenzyme complex. Translocation of the radical from β-Y₁₂₂• to α-NH₂Y₇₃₁ is expected to be accompanied by the same change to the (Fe^III)₂ cluster observed above in the wild-type complex upon reaction with N₃UDP. Indeed, the experimental Mössbauer spectrum after reaction of β with α-Y₇₃₁NH₂Y, ATP, and CDP and the product spectrum (45% conversion) obtained after removal of the unreacted component (Fig. 2) are almost identical to the spectra of the radical translocation product trapped by N₃UDP (Fig. 1B). The observation of the same perturbed quadrupole-doublet spectrum in samples in which the Y₁₂₂•-reduced β was trapped by either N₃UDP or the α-Y₇₃₁NH₂Y variant, but not in samples prepared either with a pathway-disabled α or by reduction of Y₁₂₂• by HU in the absence of α, strongly suggests that the change to the (Fe^III)₂ cluster is associated specifically with functional translocation of the radical from Y₁₂₂• into the hopping pathway.

Figure 2.

4.2-K/53-mT Mössbauer spectra of the product of forward radical translocation trapped in the (Y₇₃₁NH₂Y-α)₂β₂ complex. (I) Sample prepared with β, Y₇₃₁NH₂Y-α, ATP and CDP. The green line is the scaled control spectrum of β that had been incubated with pathway-blocked Y₇₃₁F-α, ATP and CDP, and subtraction of this spectral contribution reveals the spectrum of the product of forward radical translocation (II). The green line overlaid in II is the spectrum of the radical translocation product induced with N₃UDP from Fig.1B, III, and is shown to illustrate the near identity of the perturbed spectra resolved by the two different trapping approaches. The spectrum in III is the total difference of the spectrum of (Y₇₃₁F-α)₂β₂ in the presence of ATP and CDP subtracted from I, and the green line is the total difference spectrum from the α₂β₂ ±N₃UDP experiment in Fig.1B, IV.

To evaluate whether the change to the Mössbauer spectrum of the diferric cluster observed upon radical translocation is consistent with the proposed deprotonation of the Fe₁-OH₂ (Scheme 1B and C), we performed a series of DFT calculations. Remarkably, the calculations, performed by the broken-symmetry DFT methodology,⁴⁸ predict the same qualitative change to the Mössbauer parameters upon removal of a proton from the Fe₁-OH₂ [0.47-0.74 mm/s decrease in |ΔE_Q| for Fe₁, a much smaller (≤ 0.25 mm/s) increase or decrease in |ΔE_Q| of Fe₂, and almost no change (0.04 mm/s) to the isomer shift of either site; Fig. 3 and Table S3] as is observed experimentally upon radical translocation. The calculations were performed by starting from the published high-resolution structure of the Ec β protein,⁴⁹ which is purportedly of the met form. All first-sphere (ligand) residues and non-protein oxygen ligands [an HO(H) ligand to each iron and the oxo bridge] were included in the models (Fig. S4). Two sets of models were considered. One set includes both the diferric cluster and the radical tyrosine (Y₁₂₂) in either the resting state of the cofactor (Fe₁-OH₂/Y•) or the postulated radical-translocation-product state (Fe₁-OH/Y). A second set includes the diferric cluster in either its resting (Fe₁-OH₂) or radical translocation (Fe₁-OH) state but omits the tyrosine (Fig. S4). To limit the number of atoms while still preventing ligand motions that would be precluded by their attachment to the protein backbone from occurring during geometry optimization, the approach used by Roos and Siegbahn was adopted.⁵⁰ With these geometric constraints, none of the calculated models diverged markedly from the experimental structure during optimization (Fig. S4). In addition to calculations in the gas phase, the effect of the protein environment on the calculated Mössbauer parameters was evaluated using the COSMO solvation model^51,52 with various dielectric constants (ε) of 4, 10, and 40. Calculations using ε = 4 reproduce the experimental parameters remarkably well, to within ± 0.18 mm/s for the models that include Y₁₂₂ (Fig. 3A). Moreover, all the calculations, whether in the gas phase or with the COSMO solvent model and with Y₁₂₂ included or omitted, reproduce the essential features of the experimental spectral perturbation, giving a relatively large decrease in |ΔE_Q| and much smaller change in δ for site 1and small changes in |ΔE_Q| and δ for site 2 (Fig. 3B and Table S3). The results imply that the observed effect arises directly from the change in the charge of the Fe₁ HO(H) ligand rather than some interaction of the cluster with the Y₁₂₂/Y₁₂₂•. Consistent with this conclusion and the experimental Mössbauer spectra, parameters calculated for a model having the reduced, neutral Y₁₂₂ and the Fe₁-OH₂ ligand (corresponding to the met form of the protein) do not deviate significantly from those calculated for the active state (with Y₁₂₂• and Fe₁-OH₂). These DFT calculations thus establish that the change to the Mössbauer spectrum associated with radical translocation is consistent with the proposed donation of a proton to Y₁₂₂ by the Fe₁–OH₂ in the first PCET step.

Figure 3.

Comparison of the experimental quadrupole splitting (ΔE_Q) parameters for the resting and radical-transolcation-product states of the cofactor to the values calculated by DFT. (A) Absolute values of ΔE_Q from experiment (gray bars) and the DFT calculations using a dielectric constant of ε = 4 for models of either the full cofactor including Y₁₂₂ (solid colored bars) or just the diferric cluster without Y₁₂₂ (striped colored bars). The blue bars correspond to Fe₁ and the red bars to Fe₂. (B) The changes in the absolute values of ΔE_Q associated with the radical-translocation/deprotonation event either from the experiments (dashed gray lines) or from the calculations on the models including (solid bars) or omitting (striped bars) Y₁₂₂. The color coding is the same as in panel A.

We anticipated that the Fe₁–OH radical translocation product would be meta-stable and eventually undergo protonation (with the ultimate source being bulk solvent) to generate the Fe₁–OH₂ species of the stable met form (Fe₁-OH₂/Y). To test this notion, the radical translocation product trapped with N₃UDP was thawed and incubated on ice to permit decay of the N• and enable subsequent adjustments of the protein complex and the (Fe^III)₂ cluster. Periodically, the sample was re-frozen for acquisition of its Mössbauer spectrum. Spectra acquired after total incubation times of 10 to 260 minutes demonstrate the return of the spectrum of the resting (presumably Fe₁–OH₂) form of the cluster (Fig. 4). This result further confirms that the observed perturbation to the (Fe^III)₂ cluster upon reaction with N₃UDP or Y₇₃₁NH₂Y-α and CDP is specific to the complex actively engaged in radical translocation and catalysis. The regeneration is relatively slow (t_1/2 of ∼60 min at ∼ 0 °C), consistent with the hypothesis that it reflects the slow diffusion of an extra proton from bulk solvent to the Fe₁–OH in the protein interior to convert it to Fe₁–OH₂ of the resting met form (perhaps subsequent to decay of the N• and disengagement of β from α with a t_1/2 of 23 min at 25 °C ⁵³). That this proton transfer from bulk solvent would be slow is supported by studies in which the kinetics of electrochemical reduction of the β subunit in the presence of a facile ET mediator (methyl viologen) were monitored. These experiments revealed that the (Fe^III)₂ center in the met form can be reduced relatively rapidly (within seconds), whereas reduction of the active (Fe^III)₂/Y₁₂₂• form of the protein proceeds with the initial fast reduction of Y₁₂₂• followed by much slower reduction of the (Fe^III)₂ cluster.⁴⁰ The observations suggest that the cluster in the met form and the one generated from the active (Fe^III)₂/Y₁₂₂• cofactor upon rapid Y₁₂₂• reduction are somehow different. A reasonable explanation is that the cluster after fast Y₁₂₂• reduction has the same number of protons as the radical translocation product, possessing the Fe₁–OH generated by transfer of a proton from the Fe–OH₂ to Y₁₂₂ (Scheme 1B and C), whereas the stable met form produced by HU reduction of Y₁₂₂• in the absence of α has the cluster in the Fe₁–OH₂ form. The absence of this proton in the initial Y₁₂₂•-reduced β and presence in the stable met form would make the charge of the buried cluster different by one unit, altering the electrostatics of the site and potentially causing the difference in reduction kinetics of the two forms.

Figure 4.

Disappearance of the spectrum of the N₃UDP-trapped radical translocation product and return of the spectrum of the resting (Fe^III)₂ cluster upon prolonged incubation on ice. 4.2-K/53-mT Mössbauer spectra of α₂β₂ with TTP and in the absence or presence of N₃UDP. The +N₃UDP sample was thawed, incubated on ice, and periodically re-frozen after total incubation times of 10 to 260 minutes for re-acquisition of its Mössbauer spectrum. The spectrum at t = 260 minutes is almost identical to the spectrum of α₂β₂ with TTP in the absence N₃UDP (overlaid as a solid line).

Conclusion

The diiron cluster in class Ia Ec RNR has long been known to generate the Y₁₂₂• in the initial activation of the β subunit by reaction of its (Fe^II)₂ form with O₂.^11,15,33 Our results now strongly suggest that the (Fe^III)₂ cluster also actively functions in the catalytic cycle, specifically during translocation of the oxidizing equivalent or hole from its resting position on Y₁₂₂• in β to the nucleotide reduction site in α. Reduction of Y₁₂₂• upon forward radical translocation requires transfer of a proton to yield a neutral Y₁₂₂, and, upon reverse radical translocation and reoxidation to Y₁₂₂•, the proton should be returned. The Mössbauer-detected change to the cluster seen upon use of either the substrate analog or the α variant is almost certainly associated specifically with propagation of the radical into the hopping pathway, because the perturbation (1) is observed when the radical on Y₁₂₂ translocates in a functionally relevant reaction into α, but not when it is reduced in a nonfunctional context by HU, (2) relaxes upon decay of the N• in α, and (3) is not observed in a complex having the hopping pathway blocked by the α-Y₇₃₁F substitution. The nature of the spectral perturbation (significant decrease in |ΔE_Q| of site 1 and much smaller changes to the other three parameters) implies that the oxidation state of the cluster does not change and matches that predicted by DFT calculations for removal of a proton from the Fe₁-OH₂ ligand. Our data are thus consistent with the previous suggestion that this water ligand serves as the proton-coupling partner to Y₁₂₂ for radical translocation (Scheme 1).^15,32 The reduction of Y₁₂₂• constitutes the first step of forward radical translocation and might, therefore, be a key step in the gating of the process by the protein. If one or more protein side chain is required to mediate this proton transfer (e.g., the nearby D₈₄ carboxylate ligand), a conformational change in the α₂β₂ complex occurring upon substrate binding could engage this proton-transfer pathway and thereby open the gate to reduction of Y₁₂₂• by either W₄₈ or Y₃₅₆ in the initial step of the long-distance radical translocation.

Methods

Materials, protein production and purification, reconstitution of RNR-β with ⁵⁷Fe and/or ⁵⁶Fe, and activity assays are described in the Supporting Information.

Reactions with N₃UDP

The reaction was carried out in a final volume of 0.6 mL and contained 0.3 mM α₂ (or 0.29 mM Y₇₃₁F-α₂), 0.3 mM β₂ (0.27 mM β₂ in the case of uniformly ⁵⁷Fe labeled β and 0.29 mM β₂ in the case of Y₇₃₁F-α₂), 0.8 mM TTP, 1 mM N₃UDP, 15 mM MgSO₄, 1 mM EDTA, and 1 mM DTT in HEPES buffer. The reaction was initiated by the addition of N₃UDP and β₂ and allowed to proceed at RT (21 ± 2 °C) for a total reaction time of 2.5 min. Aliquots (0.3 mL) were transferred to Mössbauer and EPR sample cells and frozen in liquid N₂.

Reaction of Y₇₃₁NH₂Y-α₂ with substrate

Pre-reduced Y₇₃₁NH₂Y-α₂ and ⁵⁷Fe reconstituted β in assay buffer were concentrated to 0.3 mL, and mixed with an aliquot (0.05 mL) of ATP and CDP in assay buffer. The final reaction solution (0.35 mL) contained 0.25 mM Y₇₃₁NH₂Y-α₂, 0.25 mM β₂, 3 mM ATP and 1 mM CDP and was incubated at RT (22°C) for a total reaction time of 20 s and frozen in liquid N₂.

EPR spectroscopy and analysis

EPR spectra were recorded on a Bruker ESP300 spectrometer equipped with a ER 041 MR Microwave Bridge and a Bruker 4102ST TE₁₀₂ X-band resonator. Spectrometer configuration and data acquisition was performed by an external PC via GPIB interface using EWWIN 6.1 software from Scientific Software Services. Spectra were acquired at a temperature of 14.0 ± 1 K, a microwave power of 8 μW a microwave frequency of 9.45 GHz, a modulation amplitude of 3 G, a modulation frequency of 100 kHz, a receiver gain of 5 × 10⁴, and a conversion time of 0.029 s. Four scans were averaged for the spectra acquired from 2,000 to 4,000 G. For spectra collected over the narrower field range from 3,180 to 3,580 G, a modulation amplitude of 1.0 G was used, and two scans were averaged.

The total electron spin concentration in each sample was determined by integrating its EPR absorption spectrum and comparing the integrated area to that of the spectrum of a standard containing 1.025 mM CuSO₄, 2 M NaClO₄, and 0.01 M HCl.⁵⁴ The first derivative spectra recorded from 2,000 to 4,000 gauss G were integrated and the baselines were corrected by using a linear function. The second integral was formed and integrated areas were corrected for differences in g-values, as previously described.⁵⁵

Samples treated with N₃UDP contain a mixture of Y₁₂₂• and the nitrogen-centered radical (N^•). Removal of the features of the Y₁₂₂• by subtraction of the appropriately scaled spectrum of the control (– N₃UDP) sample yields the spectrum of the N•. Quantification of the subspectra of Y₁₂₂• (scaled control spectrum) and N• in the samples was accomplished by integrating the two subspectra, applying a linear baseline correction, and forming the second integral to determine the area. These areas were used to calculate the percent area of the two subspectra, which were then multiplied by the measured total spin concentration of the samples to determine the concentration of each radical. The values are reported in Table S2. The concentrations of the Y₁₂₂• and N• in the N₃UDP treated samples can also be calculated using the scaling factor of the Y₁₂₂• control spectrum that was subtracted. This scaling factor was multiplied by the spin concentration of the control sample to determine the concentration of Y₁₂₂• in the sample. This was then subtracted from the concentration of the reacted sample to determine the remaining spin concentration, which was attributed to N•. The values obtained by this procedure agree well with the values using the double integral of the two subspectra reported in Table S2.

Mössbauer spectroscopy and analysis

Mössbauer spectra were recorded at a temperature of 4.2 K and an externally applied magnetic field of 53 mT oriented parallel to the γ-beam on a SVT-400 spectrometer from WEB Research (Edina, MN). Data analysis was performed using WMOSS (WEB Research, Edina, MN). Isomer shifts are quoted relative to the centroid of a spectrum of a metallic foil of α-Fe at room temperature.

Parameters of the spectrum of the active α₂β₂ complex before reaction

The Mössbauer spectrum of the active α₂β₂ complex consists of two symmetrical quadrupole doublets of equal intensity. The spectrum was fitted by two quadrupole doublets with the linewidths (Γ) of the doublets constrained to be the same. The resulting parameters are cited in the main text and in Table S1.

Analysis of the spectrum of the α₂β₂ complex after reaction with N₃UDP

Removal of the contribution of the unreacted (Fe^III)₂ cluster (71% of total Fe of the uniformly ⁵⁷Fe labeled sample) from the spectrum of the sample reacted with N₃UDP resolves the spectrum of the radical translocation product. The subtraction was guided by visual inspection of the resolved left line at -0.76 mm/s and by the objective of making the resulting spectrum symmetric. The spectrum of the radical translocation product was then analyzed as two symmetrical quadrupole doublets of equal intensity and linewidth. Two sets of physically meaningful parameters (left-right solution and inner-outer solution, Table S1) fit the data equally well and cannot be distinguished owing to the 1:1 ratio of the two quadrupole doublets.

The dependence of the Mössbauer parameters obtained for the radical translocation product on the fraction of the spectrum of unreacted complex subtracted from the experimental spectrum was evaluated. We found that δ and ΔE_Q vary by only 0.001-0.02 mm/s, which is within the intrinsic uncertainty limit for the parameters (0.02 mm/s), when reference spectra for the radical translocation product corresponding to 29 ± 3% of total Fe (71 ± 3% of the experimental spectrum attributed to the unreacted cluster and removed in the subtraction) were analyzed.

Analysis of the spectra with site selective ⁵⁷Fe labeled β

To determine the occupancies of ⁵⁷Fe in sites 1 and 2 in the site-selectively labeled βs, Mössbauer spectra of the ⁵⁷Fe₁ and ⁵⁷Fe₂ enriched β (1.3 mM β₂) used in all the experiments were recorded. The spectra were simulated by fixing the parameters to those obtained from analysis of the spectrum of the uniformly ⁵⁷Fe labeled β in the sample of the α₂β₂ complex and allowing only the areas of the two doublets to vary. The relative areas for the two doublets correspond to the relative amounts of ⁵⁷Fe in sites 1 and 2, respectively. These occupancies were kept constant for all subsequent analyses of the spectra of reaction samples.

Subtractions of the unreacted component (scaled spectra of α₂β₂ before reaction) from the spectra after reaction with N₃UDP to obtain the spectrum of the radical translocation products were carried out as described above for the uniformly ⁵⁷Fe labeled samples. The product spectrum was simulated with the same parameters as used for the spectrum of the product in the uniformly labeled complex but using the areas for the two quadrupole doublets determined above. To evaluate the robustness of the percentage of intensity assigned to the unreacted complex, theoretical spectra for the two quadrupole doublets of the radical translocation product (using the quoted parameters) were subtracted in appropriate fractions from product spectra generated by subtraction of different amounts of the unreacted component.

The spectra of the site-selectively labeled samples allow unambiguous assignment of the two quadrupole doublets to Fe₁ (red) and Fe₂ (blue). In addition, they should, in principle, allow differentiation between the two possible sets of parameters (left-right solution and inner-outer solution, Table S1). However, the parameters of the two solutions are very similar, and we cannot make this assignment with confidence. While the two low-energy lines are at least partially resolved, both solutions (inner-outer and left-right) assign the -0.52 mm/s and -0.31 mm/s lines to the quadrupole doublets associated with Fe₁ and Fe₂, respectively. For the high-energy lines, the two lines are nearly at the same position (1.37 mm/s and 1.45 mm/s). The small difference in line position of only of 0.08 mm/s precludes a distinction between the two solutions, because variations in the linewidth influence the simulations to a greater extent than the choice of inner-outer versus left-right parameters.

DFT calculations

All geometry optimizations were performed using the GAUSSIAN 09⁵⁶ revision C.01 package. The initial geometry guess was based on the crystal structure of Ec RNR β, which is purportedly the diferric met form (PDB 1MXR⁴⁹). First shell residues around the irons and Y₁₂₂ were cut at the α carbon. To mimic the strain imposed by the protein environment, the α carbon and two out of the three adjacent hydrogen atoms were frozen in all calculations, similarly to the procedures described by Roos and Siegbahn.^32,50 Four different models were explored: the active resting state (Fe₁-OH₂/Y•) and the predicted radical translocation product (Fe₁-OH/Y) of the cofactor, both with and without Y₁₂₂ to account for the tyrosine's influence on the Mössbauer parameters. In addition, the met form (Fe₁-OH₂/Y) of the enzyme was modeled with Y₁₂₂ and calculated for reference. Calculations were performed in the gas phase as well as in three distinct dielectric environments implemented via a conductor-like screening model (COSMO)⁵¹ in the polarizable continuum model (PCM)⁵² framework, termed as C-PCM, with ε = 4, 10 and 40. Optimizations were performed within the unrestricted DFT formalism with the three-parameter Becke–Lee–Yang–Parr (B3LYP) hybrid functional.^57,58 The antiferromagnetic coupling between the irons was achieved using broken-symmetry methods based on the ones developed by Noodleman et al.⁴⁸ Pople's 6-31g basis set⁵⁹ was used on all atoms except the irons, which were represented with the 6-311+g* basis set. Calculations of spectroscopic parameters were performed using the optimized geometries with the unrestricted DFT formalism using the ORCA⁶⁰ 2.9.1 package with the B3LYP functional and the 6-311g* basis set on all atoms. In the case of the iron atoms, a diffuse function was added to the basis set (6-311+g*). The broken symmetry state was realized via the “flipspin” feature implemented in ORCA. All ORCA calculations utilized the resolution of the identity Coulomb density fitting approximation⁶¹ with the chain of spheres exchange (RIJCOSX).⁶² The ⁵⁷Fe isomer shifts (δ) were calculated from the electron density at the iron nucleus (ϱ(0)) using linear correlation response theory.^63,64 Calibration of ϱ(0) values was performed by correlating DFT calculated ϱ(0) values of iron complexes using the geometries presented by Römelt et al. with the experimental isomer shifts of these complexes.⁶⁵ The linear fit yielded the equation δ = 0.0963 – 0.3857 ϱ(0) – 11616.5) with δ in mm/s and ϱ(0) in au^-3, and RMS = 0.9899 (Fig. S5).