A 2.8 Å Fe-Fe Separation in the Fe₂^III/IV Intermediate (X) from Escherichia coli Ribonucleotide Reductase

Abstract

A class Ia ribonucleotide reductase (RNR) employs a µ-oxo-Fe₂^III/III/tyrosyl radical cofactor in its β subunit to oxidize a cysteine residue ~ 35 Å away in its α subunit; the resultant cysteine radical initiates substrate reduction. During self-assembly of the Escherichia coli RNR-β cofactor, reaction of the protein’s Fe₂^II/II complex with O₂ results in accumulation of an Fe₂^III/IV cluster, termed X, which oxidizes the adjacent tyrosine (Y) 122 to the radical (Y₁₂₂•) as the cluster is converted to the µ-oxo-Fe₂^III/III product. As the first high-valent non-heme-iron enzyme complex to be identified and the key activating intermediate of class Ia RNRs, X has been the focus of intensive efforts to determine its structure. Initial characterization by extended X-ray absorption fine structure (EXAFS) spectroscopy yielded a 2.5 Å Fe-Fe separation (d_Fe-Fe), which was interpreted to imply the presence of three single-atom bridges (O²⁻, HO⁻, and/or µ-1,1-carboxylates). This short d_Fe-Fe has been irreconcilable with computational and synthetic models, which all have d_Fe-Fe ≥ 2.7 Å. To resolve this conundrum, we revisited the EXAFS characterization of X. Assuming that samples containing increased concentrations of the intermediate would yield EXAFS data of improved quality, we applied our recently developed method of generating O₂ in situ from chlorite using the enzyme chlorite dismutase to prepare X at ~ 2.0 mM, > 2.5 times the concentration realized in the previous EXAFS study. The measured d_Fe-Fe of 2.78 Å is fully consistent with computational models containing a (µ-oxo)₂-Fe₂^III/IV core. The correction of d_Fe–Fe brings the experimental data and computational models into full conformity and thus informs analysis of the mechanism by which X generates Y₁₂₂•.

Ribonucleotide reductases (RNRs) catalyze the conversion of ribonucleotides to deoxyribonucleotides, thus providing all organisms with precursors for the de novo synthesis and repair of DNA.^1,2 All RNRs identified to date utilize a free-radical mechanism. A transient cysteine thiyl radical (C•),³ generated in situ in the first step of the reaction, abstracts a hydrogen atom from the 3'-position of the bound nucleotide. The mechanism by which the C• is generated in each turnover is the basis for the division of RNRs into classes I-III.^1,2

A class Ia RNR, such as the prototypical orthologue from aerobically-growing Escherichia coli (Ec), functions as a 1:1 complex of homodimeric subunits, α₂ and β₂. The α subunit binds substrates and allosteric effectors and contains the C residue (C₄₃₉ in Ec RNR) that is oxidized to the C•, whereas the β subunit self-assembles a µ-oxo-Fe₂^III/III/tyrosyl radical cofactor that functions to generate the C• reversibly in each catalytic cycle.^4,5 The functional cofactor is produced by reaction of the Fe₂^II/II complex of β with O₂.⁶ Addition of O₂ yields a µ-peroxo-Fe₂^III/III (P) complex^7–9 that is reduced upon cleavage of the O−O bond of the peroxo moiety. In the Ec β reaction, the O–O-cleavage step results in the one-electron oxidation of the solvent-accessible W₄₈ to a cation radical (W₄₈⁺•)¹⁰ with concomitant formation of an Fe₂^III/IV form of the diiron cluster termed cluster X.¹¹ The Ec W₄₈⁺• can be reduced in vitro by small-molecule reductants including ascorbate and thiols,¹⁰ but it is possible that an accessory protein serves as the reductant in vivo.¹² The decay of the W₄₈⁺• leaves X to oxidize the nearby Y₁₂₂ residue to the stable Y₁₂₂•. In the process, X is reduced to the µ-oxo-Fe₂^III/III cluster of the active β subunit.^6,11,13 The Y• is strictly conserved among all class Ia and Ib RNRs and is absolutely required for their activity.^1,4,14

The importance of X to the function of class Ia RNRs (which include the Homo sapiens orthologue) has made it a prime target for structural characterization. For Ec RNR, the rapid rate at which X decays (~ 1 s⁻¹ in the wild type β; 0.2 s⁻¹ in the Y₁₂₂F variant at 5 °C^11,13) has thus far prevented characterization by X-ray crystallography. Instead, the freeze-quench technique has been used to trap the intermediate, and it has then been characterized by a variety of spectroscopic methods.^13,15–21 Density functional theory (DFT) calculations have afforded models for its diiron core, and these models have been evaluated for consistency with the spectroscopic data.^22–26 This approach, now commonplace in investigations of reactive metalloenzyme intermediates,²⁷ has thus far failed to forge a consensus regarding the structure of X. The primary reason is that the short Fe-Fe separation (d_Fe-Fe ~ 2.5 Å)¹⁹ of the intermediate determined by extended X-ray absorption fine structure (EXAFS) spectroscopy seemingly requires a structure with three single-atom bridges provided by some combination of the protein carboxylate ligands and O₂/solvent-derived hydr(oxo) ligands. Such a structure has been disfavored in computational studies on energetic grounds. Indeed, structures favored in these studies have values of d_Fe-Fe ≥ 2.7 Å and no more than two single-atom oxygen bridges.^16,23–25 Furthermore, none of the available synthetic models for X has had such a short d_Fe-Fe.^28–30 Additionally, structural metrics determined by EXAFS for a Mn^IV/Fe^III homolog of X in the RNR from Chlamydia trachomatis have agreed with those derived by DFT, ³¹ indicating that current DFT methods are capable of accurately predicting the structures of such enzyme-bound dinuclear complexes.

We sought to resolve the conundrum concerning the structure of X by revisiting the irreconcilably short d_Fe-Fe determined in the initial EXAFS study. The kinetics of the activation reaction preclude trapping of X in pure form, with a maximum fraction of ~ 0.7 having been achieved in published studies. The challenging kinetics had conspired with the poor solubility of O₂ in aqueous solutions (< 2 mM at 1 atm) to limit the concentration of X that could be trapped to < 0.8 mM. Recent technological advancements now permit accumulation of O₂-derived intermediates at concentrations exceeding 2 mM,³² and we reasoned that the ability to trap X at elevated concentrations might yield samples of higher quality to permit the re-characterization of the intermediate by EXAFS.

Samples for this study were prepared by the method of generating O₂ in situ from chlorite (ClO₂⁻) with the enzyme chlorite dismutase (Cld). A reactant solution containing a high concentration of the pre-formed Fe₂^II/II complex of Ec RNR-β-Y₁₂₂F and a catalytic concentration (12.5 µM) of Cld was mixed with 0.25 equivalent volumes of a second reactant solution containing ClO₂^−, and the reaction was freeze-quenched after 0.3 s (at 5 °C). The 4.2-K/53-mT Mössbauer spectra of the freeze-quenched samples reveal the presence of ~ 65% X, comparable to the maximum fraction obtained in the previous EXAFS study, along with ~ 18 % unreacted Fe^II species and ~ 18 % of µ-oxo-Fe₂^III/III product cluster (Figures S1-S3 of the Supporting Information). This fraction of X corresponds to 2.0 mM, more than 2.5 times the maximum concentration attained in the previous EXAFS study.

The X-ray absorption near-edge structure (XANES) spectra (Figure S4) show a higher K-edge absorption energy (the energy at which the 1s-core electron is ejected) for samples containing X than for the samples of the unreacted Fe₂^II/II-β starting material. However, the edge of the X samples lies at a lower energy than for samples of the µ-oxo-Fe₂^III/III product. This phenomenon, also observed by Riggs-Gelasco et al.,¹⁹ may result from the contribution of the unreacted Fe₂^II/II component in the freeze-quenched samples of X. Alternatively, the skewing of the edge energy to a lower value may be a feature inherent to X that remains to be explained theoretically.

Fe K-edge EXAFS data over k = 0.3 – 14 Å⁻¹ (for samples containing the Fe₂^II/II reactant complex and the µ-oxo-Fe₂^III/III product state) and k = 0.3 – 16 Å⁻¹ (for samples containing X) are shown in Figure 1, along with fits to the raw data based on the parameters given in Tables 1 and S1-S3. The EXAFS data of the Fe₂^II/II reactant complex (Figure 1A, left panel) are best fit with a model that contains a total coordination number of four oxygen/nitrogen (O/N) ligands per Fe. This value is consistent with the crystal structure of that enzyme form.³³ There is no evidence for an Fe-Fe scatterer in the Fourier transform (FT) of the EXAFS (Figure 1A, right panel), presumably because d_Fe-Fe is too large (~ 3.8 ų³) in this form of the cluster. The EXAFS data of the µ-oxo-Fe₂^III/III product cluster (Figure 1B, left panel) can be fit with a model that contains six O/N ligands per Fe.³⁴ Furthermore, the FT of the EXAFS data exhibits a prominent Fe scattering interaction at R = 3.2 Å (Figure 1B, right panel), a value that is also consistent with the reported d_Fe-Fe of this form.^35,36

Figure 1.

Fe K-edge EXAFS data (left panel) and their Fourier transforms (right panel) for samples containing the Fe₂^II/II reactant complex (A), the µ-oxo-Fe₂^III/III product (B) and X (C, D). Fit parameters are provided in Tables 1, S1, S2, and S3.

Table 1.

Fe K-edge EXAFS (k = 0.3 − 16 Å⁻¹) fitting results of samples containing X for which the fit model includes two discernable Fe−Fe interactions. Occupancies were fixed during the fit, but distances, Debye-Waller factors, and the threshold energy shift were allowed to vary

Scatterer Type	N	R	σ2
Fe−O/N	5	2.02	0.0097
Fe−O	0.65	1.75	0.0020
Fe−C	3	2.97	0.0046
Fe−C	1	3.24	0.0060
Fe−Fe	0.65	2.79	0.0033
Fe−Fe*	0.18	3.22	0.0043
F		0.366
E0		−11.067
Resolution		0.099 Å

N: occupancy; R: distance (Å); σ²: Debye-Waller factor (Ų); E₀: threshold energy shift (eV); F: fit error.

^*Parameters for this scattering interaction were constrained to the values obtained from fits of the diferric EXAFS

The EXAFS data for the samples containing X (Figure 1C, left panel) can be fit by a model with 3 O/N at 2.01 Å, 2 O/N at 2.11 Å, and 0.65 O at 1.75 Å per Fe. This fit also includes two Fe-Fe scattering interactions: 0.65 Fe at 2.78 Å, and 0.18 Fe at 3.22 Å (Table S3). The occupancies of the two Fe-Fe scattering interactions account for the heterogeneity of the sample, specifically the fractions of X and µ-oxo-Fe₂^III/III cluster determined from the Mössbauer data. The contribution to the Fe-Fe scattering interactions in the EXAFS data from the ~ 18% Fe₂^II/II component in the samples is not obvious and is not accounted for in the analysis. The Fe-O/N interaction at 1.75 Å is likely to arise from an oxo bridge, and the occupancy of 0.65 is consistent with the presence of two such µ-oxo interactions in an asymmetric di-µ-oxo-Fe₂ core (see Table S3 for additional fitting results).

The agreement between the fit and the data at R ≥ ~3.0 Å can be improved through the inclusion of additional non-nearest-neighbor scattering interactions. Examination of the crystal structures of the reactant and product complexes reveals the presence of carbon atoms from the bridging and terminal carboxylates and histidines that could contribute to scattering interactions at ~ 3.0 and ~ 3.2 Å. In the crystal structure of the Fe₂^III/III product state (1RIB), there are a total of 6 carbons (three per Fe) at ~ 3.0 Å and 2 carbons (one per Fe) at 3.2 Å away from the Fe ions.³⁴ Assuming that these atoms might also be present at similar distances in the samples containing X, we included three Fe-C scattering interactions at 3.0 Å and one at 3.2 Å into the fit model. Their inclusion significantly improves the agreement between the fit and the data (Figure 1D; parameters provided in Table 1). It is unclear why these interactions would be required for fits of X but not the diferric EXAFS. It is possible that the high-valent X contains a tighter core, making these scattering interactions pronounced. Irrespective of the origin of the additional interactions, the dominant scattering interaction at ~2.8 Å can be assigned to an Fe scatterer, and there is no evidence for an interaction at the previously reported d_Fe-Fe of ~ 2.5 Å.

To determine the structure of the diiron core of X and rationalize the 2.8 Å d_Fe-Fe, we generated a series of structural models by broken-symmetry DFT methods, following previous work by Noodleman and coworkers^22–26 (see Supporting Information for a more detailed description). The models were derived from the X-ray crystallographic data (1RIB) of Ec β by modifying the ligation. Two main candidates were examined in detail, a di-(µ-oxo)-(µ-1,3-carboxylato) core structure and a (µ-oxo)(µ-hydroxo)(µ-1,3-carboxylato) structure (Figure 2). The di-(µ-oxo) model has distance parameters that closely match the experimentally determined values, including, most notably, the d_Fe-Fe of 2.8 Å (Figure 2).

Figure 2.

Structural models for the Fe₂^III/IV core of X derived from broken-symmetry DFT calculations. Left: (µ-oxo)₂ core; Right: (µ-oxo)(µ-hydroxo) core.

It is noteworthy that the DFT calculations imply that protonation of one of the µ-oxo bridges should result in an elongation of the Fe-Fe separation to ~ 3.0 Å, the distance at which a minor scatterer is detectable in the data. The results of magnetic circular dichroism studies on X suggested a model in which one of the bridging oxo groups is protonated.¹⁶ The inclusion of an Fe scatterer at ~3.0 Å (in lieu of additional C scatters at 3.0 and 3.2 Å) also improves the fit in this region. However, such a structure has remained inconsistent with data from ²H-electron-nuclear double resonance experiments, which do not detect a bridging hydron.^18,20,21

The effect of including the essential Y₁₂₂ that is oxidized by X in the DFT calculations was also evaluated (see Supporting Information for a more detailed description of the computational methodology). Y₁₂₂ forms a hydrogen bond to aspartate 84, which ligates Fe1 of the Fe−Fe cluster. Thus, its presence provides a proton transfer pathway by which one of the µ-oxo bridges in X might be protonated, thereby altering the core structure of the intermediate. The results of our DFT calculations, similar to those of Noodleman and co-workers, show that the presence of Y₁₂₂ has a minor effect on the optimized geometries, resulting in only a slight increase (~ 0.02 Å) in d_Fe−Fe (Table S9). Thus, it seems unlikely that the structure of X in the β-Y₁₂₂F variant could be significantly different from that formed in the wild-type protein.

In an effort to understand the basis for the discrepancy between the d_Fe-Fe of 2.8 Å determined here and the previously reported distance of 2.5 Å, we considered that third-generation synchrotron technology and the increased (2.5 ×) concentration of X (obtained through the use of Cld and ClO₂⁻) could result in a critical increase in signal-to-noise ratio. Interestingly, this was not the case. The data from both studies have effectively the same signal-to-noise ratio. We also considered that the increased resolution provided by the extended k-range of our measurements might be critical to the observation of the 2.8 Å Fe-Fe distance. Whereas the data analyzed in the previous study were limited to k = 2 – 12.6 Å⁻¹, the data reported here were fit from k = 0.3 – 16 Å⁻¹. To evaluate whether this difference might be a plausible explanation for the discrepant results, we examined FTs of unfiltered EXAFS data with cutoffs at k = 11, 12, 13, 14, 15, and 16 Å⁻¹. Figure 3 shows that the intensity of the 2.8 Å peak does decrease with k, becoming a shoulder when k_max = 11.0 Å⁻¹. However, fits over five of the six k ranges listed in Figure 3 yield an Fe-Fe distance of 2.8 Å. (Fits of the shortest range, k=0–11 Å⁻¹, yield an Fe-Fe distance of 2.37 Å, with a large Debye-Waller factor of 0.01.) In no case does truncation of our data lead to the assignment of a 2.5 Å Fe-Fe scattering interaction. It appears then that the data reported herein and the previous study are inherently different, suggesting that they were obtained from inherently different samples (i.e., not from the same species). To illustrate this point, we overlay the EXAFS obtained from both studies in Figure S8.

Figure 3.

FT of EXAFS data of samples containing X plotted with different cutoffs of the k-range. The dashed line is drawn at the middle of the ~ 2.8 Å peak. Overlaid in red is the FT resulting from the fit reported in Table 1.

This re-examination of the structure of X and subsequent upward adjustment of its d_Fe−Fe calls into question the short d_Fe−Fes reported for other O₂-derived diiron intermediate complexes. For example, the high-valent Fe₂^IV/IV complex, Q, that accumulates during the conversion of methane to methanol by the soluble methane monooxygenase from Methylosinus trichosporium OB3b was characterized by EXAFS, and the measured d_Fe−Fe of 2.46 Å led to the proposal of a [(µ-oxo)₂Fe₂] "diamond core" structure.³⁷ Subsequently, EXAFS characterization of the µ-peroxo-Fe₂^III/III complexes that accumulate in the reactions of M ferritin from frog³⁸ and the D₈₄E/W₄₈A variant of Ec RNR-β³⁹ led to reports of similar values of d_Fe−Fe (~ 2.5 Å) even in these mid-valent complexes. In general, the structures dictated by these surprisingly short Fe-Fe separations have been irreconcilable with synthetic and computational models, which predict d_Fe−Fes of ~ 2.7 Å for Q and > 3.0 Å for the µ-peroxo-Fe₂^III/III complexes.^40–43 Re-examination of these other complexes and re-determination of their Fe-Fe separations would seem to be warranted.

Scheme 1.

Schematic description of the activation Ec class Ia RNR. Intermediate X is the precursor to the active Fe₂^III/III/Y• cofactor.

ABBREVIATIONS

RNR

ribonucleotide reductase

Ec

Escherichia coli

DFT

Density Functional Theory

EXAFS

extended X-ray absorption fine structure

k

photoelectron wave vector

XANES

X-ray absorption near structure

FT

Fourier transform

BS

broken symmetry

COSMO

conductor-like screening model

ε

dielectric constant