EXAFS Simulation Refinement Based on Broken-Symmetry DFT Geometries for the Mn(IV)-Fe(III) Center of Class I RNR from Chlamydia trachomatis

Abstract

Ribonucleotide Reductases (RNRs) catalyze the reduction of ribonucleotides into deoxyribonucleotides necessary for DNA biosynthesis. Unlike the conventional class Ia RNRs which use a diiron cofactor in their subunit R2, the active site of the RNR-R2 from Chlamydia trachomatis (Ct) contains a Mn/Fe cofactor. The detailed structure of the Mn/Fe core has yet to be established. In this paper we evaluate six different structural models of the Ct RNR active site in the Mn(IV)/Fe(III) state by using Mössbauer parameter calculations and simulations of Mn/Fe extended X-ray absorption fine structure (EXAFS) spectroscopy, and we identify a structure similar to a previously proposed DFT-optimized model that shows quantitative agreement with both EXAFS and Mössbauer spectroscopic data.

Introduction

Ribonucleotide reductases (RNRs) are crucial enzymes. All organisms rely on these enzymes to produce deoxyribonucleoside triphosphates (dNTPs) – the building blocks of deoxyribonucleic acid (DNA)¹. The structure and assembly of RNRs have been investigated by a variety of experimental^2–11 and theoretical^12–18 methods. Three main classes of RNRs have been discovered that display a common reaction mechanism using metals and free radical chemistry.¹⁹ Although these classes differ in composition and cofactor requirements, they all possess a conserved cysteine residue at the active site that is converted (during the catalytic cycle) into a thiyl radical (Cys-S^•), which initiates substrate turnover by abstracting a hydrogen atom from the ribose ring of the substrate.^20–22 There is a dinuclear metal center in class I RNRs, a cobalt containing cobalamin cofactor (adenosylcobalamin) in class II RNRs, and a 4Fe-4S cluster in class III RNRs.²² Each of these cofactors generates a radical that transfers to produce Cys-S^•. Class I RNRs are found in all eukaryotes as well as in some microorganisms like Escherichia (E.) coli and have two dissimilar protein subunits R1 (α₂-homodimer) and R2 (β₂-homodimer). R1 contains the substrate binding site and the conserved cysteine residue, and acts as a catalyst for the dehydroxylation of the 2’-hydroxyl group of the ribose ring. R2 contains the dinuclear metal cluster that generates a stable radical (except for class Ic RNRs where the cluster itself is the oxidant). This radical then transfers (through a long-range proton-coupled-electron-transfer propagation mechanism) to create Cys-S^• which initiates the ribonucleotide-to-deoxyribonucleotide reaction in R1. In E. coli class Ia RNR, a tyrosine residue (Tyr122 in R2) is the radical bearer closest to the diiron center in R2.^{21, 23, 24} The fairly stable tyrosyl radical is generated by an Fe(III)Fe(IV) intermediate state X^{2, 5, 6, 8, 10, 11, 25–32} following the reaction of the reduced Fe(II)Fe(II) center with molecular O₂.²⁵ The active form of class Ia R2 is described as an Fe(III)Fe(III)-Tyr^• state.^{1–10, 19, 20, 33–38} The radical bearing tyrosine is conserved among more than 200 sequenced R2s. Mutants with a phenylalanine in this position are enzymatically inactive^{39, 40} with the exception of native RNR-R2 from the pathogenic bacteria Chlamydia trachomatis (Ct). There, a phenylalanine (Phe127) resides in place of the tyrosine residue which carries the radical in conventional RNRs.^l Later Bollinger’s group discovered that the Ct-RNR-R2 contains a Mn-Fe center rather than a diiron center in its functional form, and it uses the Mn(IV)-Fe(III) cofactor (intermediate X) directly for radical initiation instead of the tyrosine radical.^{41, 42}

Currently, a variety of Mössbauer, hyperfine, and EXAFS data are available for the Ct-R2 center in different oxidation states.^{11, 41, 43–46} However, explicit structures of the Mn/Fe active site in each oxidation state are incompletely known. The first X-ray structure (pdb entry 1SYY) obtained for Ct-R2 was in the Fe(III)-Fe(III) diiron inactive form.⁴⁷ This diiron center is very similar to the diferric center of the hydroxylase component of soluble methane monooxygenase (MMOH).⁴⁸

While we were preparing the current manuscript, two relevant experimental papers on Ct-R2 were published.^{49, 50} Andersson et al.⁴⁹ performed X-ray crystallographic anomalous diffraction experiments and concluded that Mn occupies site 1 (the metal site which is closer to Phe127, see Fig. 1) in Ct-R2. Dassama et al.⁵⁰ also studied the location of Mn(IV) vs. Fe(III) by correlating X-ray crystallographic anomalous scattering intensities with catalytic activity for Ct-R2 protein samples reconstituted in vitro by two different procedures. They found that Mn occupies site 1 preferentially to site 2 with Fe in the alternate site, but with some heterogeneity in Mn and Fe sites. With diminished Mn(II) loading in the reconstruction, no evidence for Mn occupancy of site 2 was found.⁵⁰

Figure 1.

Ct RNR active site model 3. In 3, Mn(IV) occupies site 1, the metal position closer to Phe127, and Fe(III) occupies site 2, which is the metal position further from Phe127. The protein environment surrounding the Fe(III)/Mn(IV) cofactor is the same in each model. The differences between the active site cores are shown in Fig. 2.

Multiple theoretical studies have been performed in order to characterize the Ct-R2 active site structures and the positions of the Mn and Fe sites.^{11, 16, 18} For the Mn(IV)-Fe(III) state, it is now commonly agreed that the active site likely contains a bridging hydroxo- and a bridging oxo-ligand.^{11, 16, 18}. In Ref. 16, two µ-oxo, µ-hydroxo structures (see Fig. 1 for a full active site structure, and models 3 and 4 in Fig. 2 for first coordination shell) for the Mn(IV)-Fe(III) state of the Ct-R2 active site have been examined in addition to two di-µ-oxo structures (see structures 1 and 2 in Fig. 2). In these µ-oxo-µ-hydroxo models, oxygen O2 (compare Figure 1) on the side of the negatively charged carboxylates of Glu89 and Glu227 is protonated because protonation of the oxygen atom O1 nearer to the neutral His123 and His230 is energetically less favourable.¹⁸ Structures 1 and 2 differ principally in the positions (site 1 or 2) of the Mn/Fe metals in the cofactor relative to the position of Phe127 (see Fig. 1). This difference in the positions of the metal atoms also differentiates structures 3 and 4.

Figure 2.

Ct RNR active site model cores of structures 1–4 as presented in Ref. 16. The subscripts 1 and 2 on the metal atoms indicate which site these atoms occupy. Site 1 is closer to Phe127 (not shown, see Fig. 1) than site 2. The subscripts 1 and 2 on the bridging oxygen atoms simply differentiate the two positions (i.e., cis and trans to His123, respectively).

This paper evaluates the quantitative agreement with EXAFS of the four Ct RNR active site model structures (1–4) examined by Han et al.¹⁶ as well as of models 5 and 6 (see Fig. 3) which have been derived from structures 3 and 4, respectively, through additional fitting of interatomic distances comparing simulated and experimental¹¹ Mn and Fe extended X-ray absorption fine structure (EXAFS) spectra. Calculated Mössbauer parameters for the Mn(IV)-Fe(III) states are also compared with experimental parameters for Fe isomer shift (δ) and quadrupole splittings (ΔE_Q). There are several possibilities to change the atomic positions for a good fit to the experimental EXAFS data. We have found optimally fit structures with reasonably close atomic distances starting from structures 3 and 4 whose computed Mössbauer parameters are in good agreement with experimental data.

Figure 3.

Ct RNR active site model core of structures 5 and 6. The subscripts 1 and 2 on the metal atoms indicate which site these atoms occupy. Site 1 is closer to Phe127 (not shown, see Fig. 1) than site 2. The subscripts 1 and 2 on the bridging oxygen atoms simply differentiate the two positions (i.e., cis and trans to His123, respectively).

Experimental EXAFS and Mössbauer spectroscopy data

Experimental EXAFS and Mössbauer spectroscopy data are from Younker et al.¹¹

EXAFS calculations

All EXAFS spectra were computed using the program FEFF version 8.30, and all fits were performed using the program IFEFFIT version 1.2.11, which optimizes the fitting parameter with a least-square fit to the experimental data.^{51, 52} For each structure, unfitted FEFF EXAFS spectra were generated. Fitting of these EXAFS spectra to experimental data was performed in reduced distance (r)-space after Fourier transformation over a range of k=3.05 – 11.6 Å⁻¹ (k is the photoelectron wave vector) using the program IFEFFIT. For the Fourier transformation, a Hanning window with Δk=1 was applied to the k³-weighted EXAFS data. The grids of k points, which are equally spaced at 0.05 Å⁻¹, were then used for the Fourier transformation to reduced distance (r)-space.

As in Ref. 11, three shells of Debye-Waller factors σ², which account for thermal disorder in the structure,⁵³ were considered, whereby one shell is used for the Mn-Fe distance and the other shells account for the six nearest atomic neighbours around the scattering center factors (coordination numbers of 2 and 4 for the first and second coordination shell, respectively, were employed). The negligibly small contribution of hydrogen atoms was not considered. In each fit, the passive electron reduction factor S₀ for each shell was held constant while the K-edge energy E₀ and each shell’s Debye-Waller factor σ² were varied. As in Ref.11, fits were performed using values of 0.8, 0.9, and 1.0 for S₀ in case of Mn and Fe K-edge EXAFS, respectively. The initial guess values for E₀ have been chosen as 6551.6 eV for the Mn EXAFS and 7122.1 eV for the Fe EXAFS calculations, also based on Ct RNR EXAFS fitting data from Younker et al.¹¹.

The fitted EXAFS spectra presented in the following have been chosen since they show positive and reasonable low Debye-Waller factors and give rise to the smallest R-factors. The R-factor which measures the absolute misfit between the computed and experimental spectrum, is in general given as follows⁵³ (for details, we refer to Ref. ⁵⁴,⁵⁵:

$$R=\frac{{\sum }_{i=1}^{N_{\mathit{\text{fit}}}}{[x_i^{\mathit{\text{exp}}}-x_i^{\mathit{\text{fit}}}(p)]}^2}{{\sum }_{i=1}^{N_{\mathit{\text{fit}}}}{[x_i^{\mathit{\text{exp}}}]}^2}$$

N_fit is the number of data points, x^exp is the experimental, and x^fit the calculated EXAFS function which depends on the set of fitting variables p. The fit of the calculated data to the experimental one is the closer, the smaller the R-factor value is.

For the derivation of structures 5 and 6, an iterative procedure was used: two or three interatomic distances were fitted in addition to the Debye-Waller factors and E₀ for one metal scattering center. The structures were chosen such that the interatomic distances are close to the DFT optimized ones of structures 3 and 4, respectively, and chemically reasonable. As additional criteria, a good agreement between the experimental and calculated Mössbauer parameters and a good match between the experimental and computed EXAFS spectrum with the other metal as scattering center were considered.

DFT calculations

The calculations were performed using the Amsterdam Density Functional (ADF) package^{55, 56} with the Vosko, Wilk, and Nusair (VWN) parametrization⁵⁷ for the local density approximation and the OPBE density functional^58–60 for the nonlocal exchange and correlation. In analogy to Ref. 16, the triple-zeta TZP basis set with frozen core as implemented in ADF has been employed in all calculations except of the ones for the Mössbauer properties where no frozen core has been applied. The COSMO conductor-like screening model^{16, 61–64} has been used with the dielectric constant set to 4 and van der Waals radii for the atoms Fe, Mn, C, N, O, and H have been chosen as 1.5, 1.5, 1.7, 1.55, 1.4, and 1.2 Å, respectively¹⁶. A value of 2.0 Å has been applied for the probe radius of the contact surface between the molecule under study and the solvent. Single-point energy calculations have been performed for models 1–4, whose optimized geometries as presented in Ref. 16 have been used. The structures obtained using the EXAFS optimizer starting from the DFT-optimized models 3 and 4 by best fit to the experimental EXAFS spectra produce models 5 and 6. At this stage, the six nearest atomic neighbours around the Mn and Fe centers and the Mn and Fe positions are fixed from the EXAFS optimization calculations. The remaining atoms of the residues Glu89, Glu227, Glu193, His230, Glu120, and His123 as well as the hydrogen atoms of the water molecule bound to the Mn/Fe atom and of the µ-hydroxo bridge are optimized by DFT for the purpose of comparative energy calculations. All other residues, which are further away from the metal centers, have also been frozen. The energies given in this paper are spin-projected energies for the whole model shown in Fig. 1 and calculated as described in Ref. 16. The Mössbauer quadrupole splitting, isomer shift, and asymmetry parameter η were calculated for each structure with the same computational settings as those described in Ref. 16.

Results

The active site cores of the structures studied in this paper are shown in Figs. 2 and 3, while an example of a core and the local protein environment is shown in Fig. 1. A terminal water ligand at metal site 1 has been chosen since previous results based on EXAFS fitting¹¹ and quantum chemical calculations^{11, 18} showed that a water ligand is more reasonable than a hydroxo ligand. The xyz coordinates of model structures 1–4 were taken from Ref. 16. Structure 5 has been derived from structure 3 by fitting of interatomic distances to obtain a better consistency of calculated and experimental EXAFS data (for details, see above). In an analogous way, structure 6 has been obtained from structure 4. The computed Mn K-edge EXAFS spectra in k-space and r-space for structures 1–4 are shown in Fig. 4 and the analogous spectra for structures 5 and 6 in Fig. 6.

Figure 4.

Mn K-edge EXAFS spectra in k-space (left-hand side) and r-space (right-hand side) for model structures 1–4. A passive electron reduction factor S₀ of 0.8 was used as well as the K-edge energy shifts and Debye-Waller factors given in Table 1.

Figure 6.

Mn (upper part) and Fe (lower part) K-edge k-space (left-hand side) and r-space (right-hand side) EXAFS spectra for model structures 5 and 6. A passive electron reduction factor S₀ of 0.8 and 1.0 for the Mn and Fe EXAFS calculations, respectively, was used. The K-edge energy shifts and Debye-Waller factors employed are given in Tables 1 and 2.

Values of 0.8, 0.9, and 1.0 for S₀ were employed where S₀ = 0.8 gave the smallest deviation of the calculated Mn EXAFS spectrum for each model structure compared to the experimental data. This is consistent with fitting results from Younker et al.¹¹ The Fe K-edge EXAFS spectra for structures 1–4 are shown in Fig. 5, and the analogous ones for structures 5 and 6 in Fig. 6. In accordance to Ref. 11, an S₀ value of 1.0 gave the closest agreement of the calculated Fe EXAFS spectra to the corresponding experimental ones for each model structure, compared to other S₀ values of 0.8 and 0.9. As expected, the simulated spectra of structures 5 and 6 match the experimental Ct RNR EXAFS spectra the best. The Mn and Fe EXAFS fitting data confirms this conclusion (compare Tables 1 and 2) since structures 5 and 6 have the lowest R-factor. The data also show that the next best models are 3 and 4.

Figure 5.

Fe K-edge k-space (left-hand side) and r-space (right-hand side) EXAFS spectra for model structures 1–4. The passive electron reduction factor S₀ was set to 1.0 and the K-edge energy shifts and Debye-Waller factors employed are presented in Table 2.

Table 1.

Mn EXAFS data for Ct RNR model structures 1–6. In each fit the passive electron reduction factor S₀=0.8 was held constant while the K-edge energy E₀ shift (in eV) and each coordination shell’s Debye-Waller factor σ² (in 10⁻³ Ų) were varied. Shell 3 corresponds to the Mn-Fe distance, shells 1 and 2 to Mn-O/N distances. The dimensionless R-factor (R) is the absolute misfit between the experimental and the computed data.

Structure	E0 shift	Shell 1 σ2	Shell 2 σ2	Shell 3 σ2	R
1	−27.2	8.3	5.0	1.4	0.88
2	−38.0	8.9	9.8	6.8	1.03
3	10.4	0.4	2.8	1.7	0.20
4	13.2	8.5	3.1	0.0	0.31
5	5.1	0.8	3.7	0.7	0.08
6	3.2	6.2	3.4	1.1	0.07

Table 2.

Fe EXAFS data for Ct RNR model structures 1–6. In each fit the passive electron reduction factor S₀=1.0 was held constant while the K-edge energy E₀ shift (in eV) and each coordination shell’s Debye-Waller factor σ² (in 10⁻³Å²) were varied. Shell 3 corresponds to the Mn-Fe distance, shells 1 and 2 to Fe-O/N distances. The dimensionless R-factor (R) is the absolute misfit between the experimental and the computed data.

Structure	E0 shift	Shell 1 σ2	Shell 2 σ2	Shell 3 σ2	R
1	−2.8	4.2	7.3	8.5	0.53
2	7.3	4.3	6.8	8.9	0.73
3	6.4	4.1	8.6	4.9	0.40
4	11.5	4.6	6.7	6.1	0.27
5	3.6	1.3	3.7	4.4	0.02
6	4.6	4.3	4.8	4.5	0.03

Table 3 shows computed Mössbauer properties for each model (the results for structures 1–4 are taken from Ref. 16) as well as experimental values from Refs. 41, 42. The model structures showing best agreement between the calculated and experimental isomer shifts (δ) are 2, 3, and 5. The smallest deviation between calculated and experimental Mössbauer quadrupole splitting (ΔE_Q) values has been obtained for structure 5, followed by models 4 and 3. The uncertainty in the calculated Mossbauer parameters can be estimated by considering the training sets used in Ref.⁶⁵ which contain isomer shift fits for 30 Fe sites and quadrupole-splitting calculations for a corresponding but larger set of 59 Fe sites, all using the OPBE exchange-correlation potential.

Table 3.

Computed Mössbauer properties [isomer shift δ (in mm s⁻¹), quadrupole splitting ΔE_Q (in mm s⁻¹), and asymmetry parameter η] of Ct-R2 model structures 1–6 with oxidation state Mn(IV)-Fe(III)^a. The experimental data is taken from Refs. 41, 42.

	1	2	3	4	5	6	Exp.
	16	16	16	16	Calc.	Calc.	41, 42
δ	0.58	0.53	0.54	0.49	0.54	0.47	0.52
ΔEQ	−0.82	−0.78	−1.23	−1.38	−1.29	−1.63	−1.32
η	0.93	0.74	0.24	0.83	0.45	0.91	0.11
Г b							0.3

^aComputed results for models 1–4 are taken from Ref. 16.

^bLine width Г (in mm s⁻¹).

The calculated standard deviation (SD) for the isomer shift test set was 0.072 mm s⁻¹ and for the quadrupole splitting 0.25 mm s⁻¹. Using these SD's in Table 3 of the manuscript, we find that the quadrupole splittings are more diagnostic in separating the best models from the rest, since model 6 differs from model 5 by more than 1 SD (also the sign of the quadrupole splitting of model 6 is wrong), while both models 3 and 4 are acceptable. Models 1 and 2 are poor because their quadrupole-splitting values deviate more than 2 SD compared to experiment.

The experimental line width Г = 0.3 mm s⁻¹ (Table 3)^{41, 42} should approximately correspond to the experimental SD of the quadrupole splitting. For Table 3, the experimental uncertainty (SD) is thus similar to the calculated SD. In Table 4, the experimental SD is larger, since the line with Г is approximately 0.5 mm s^{−1 41}, while the SD obtained with the OPBE density functional remains the same (SD=0.25 mm s⁻¹).

Table 4.

Computed Mössbauer Properties [isomer shift δ (in mm s⁻¹), quadrupole splitting ΔE_Q (in mm s⁻¹), and asymmetry parameter η] of Ct-R2 model structures containing Mn(III)-Fe(III) and Mn(IV)-Fe(IV) oxidation states,^a “1” and “2” refer to the site the metal occupies (see Fig. 1 for details).

	Mn1(III)-Fe2(III)	Fe1(III)-Mn2(III)	Exp. b
δ	0.56	0.53	0.53
ΔEQ	0.37	0.60	0.73
η	0.66	0.74	0.40
	Mn1(IV)-Fe2(IV)	Fe1(IV)-Mn2(IV)	Exp. c
δ	0.24	0.23	0.17
ΔEQ	0.47	0.76	0.75
η	0.43	0.79	0.64

^aCalculated Mössbauer properties are taken from Ref. 16.

^bTaken from Ref. 41; the experimental quadrupole splitting ΔE_Q was determined at T = 190 ˚K, large line width Г ~ 0.5 mm s⁻¹

^cTaken from Ref. 43, experimental isomer shift δ and quadrupole splitting ΔE_Q were not precise because the magnetic spectrum persists up to T = 120 ˚K.

Discussion

Our objective of this paper is to determine which structure among models 1–6 agrees best with experiment based on evaluation of their simulated EXAFS and Mössbauer spectroscopy data. Structures 1 and 2, which both have di-µ-oxo bridging between Mn(IV) and Fe(III), can be eliminated immediately because their Mn and Fe EXAFS spectra give rise to the highest R-factors (see Tables 1 and 2 and Figs. 4 and 5). These EXAFS fits were obtained by constraining the Debye-Waller factors to be positive and less than 0.010 Ų because otherwise negative or unreasonably large Debye-Waller factors were computed. The high values of the K-edge energy E₀ shifts also indicate that structures 1 and 2 are not good models for fitting the experimental EXAFS data. In addition, the calculated quadrupole splitting values of structures 1 and 2 deviate significantly from the experimental one leading to an overall better agreement of the µ-oxo-µ-hydroxo models (3–6) with experiment. This is consistent with results from Han et al.¹⁶ Therefore, the remainder of this discussion will evaluate the relative accuracy of the µ-oxo-µ-hydroxo models 3, 4, 5, and 6.

Comparison of the R-factors from the Mn EXAFS fits for each structure shows that structure 6 has the lowest value, closely followed by structure 5. In case of the Fe EXAFS fits, structure 5 leads to a slightly lower R-factor compared to model 6. The calculated Fe-Mn distances of 2.92 and 2.91 Å in models 5 and 6, respectively, are in agreement with the findings in Ref. 11. One short Mn-O distance (1.68 Å for structure 6 and 1.71 Å for structure 5) is obtained, similar to a single oxygen scatterer at 1.74 Å found in Ref. 11. In addition, structures 6 and 5 contain four scatterers around the Mn atom with distances between 1.85 Å and 1.96 Å, close to the result in Ref. 11 where four scatterers at 1.95 Å have been determined. Since models 3 and 4 are not EXAFS-refined structures, their R-factors are significantly larger compared to the ones of structures 5 and 6: the R-factors of models 6, 5, 4, and 3 are 0.07, 0.08, 0.31, and 0.20 in the Mn EXAFS fits and 0.03, 0.02, 0.27, and 0.40 in the Fe EXAFS fits. The Mn and Fe EXAFS data for structure 3 agrees better with experiment than the one corresponding to structure 4 in case of the Mn EXAFS fit but the situation is reversed in case of the Fe EXAFS fit. Thus, based on analysis of the Fe and Mn computed EXAFS data alone, structure 3 cannot be preferred to structure 4.

Considering the Mössbauer calculations, for the isomer shifts, the errors for structures 3 and 5 relative to experiment are 3.8%; for structures 4 and 6, 5.9% and 9.6%, respectively. For the calculated quadrupole splittings, we mainly focus on the absolute values when comparing with experiment, since our previous studies show that if η is close to 1, the sign of the calculated quadrupole splitting can change with the chosen atomic basis sets, computational methods, and the size of the quantum cluster.⁶⁶ The computed quadrupole-splitting values of 1.29, 1.38, 1.23, and 1.63 mm s⁻¹ for structures 5, 4, 3, and 6 give rise to an error of 1.5%, 4.5%, 6.8%, and 23.5% compared to experiment, respectively.

Comparing structures with the same metal site specificity, the most significant change in structure 5 compared to model 3 occurs for the distance of the Fe center to the oxygen atom of the µ-hydroxo group, which is shortened by 0.13 Å, as well as the bond length of Fe to an oxygen atom of the Glu120 carboxylate group, which is elongated by 0.20 Å. Structure 6 contains a bond between the Fe center and the oxygen of the µ-hydroxo bridge which is elongated by 0.13 Å compared to structure 4. Another remarkable modification is the distance between the Mn center and the nitrogen atom of His230, which is 0.20 Å longer in structure 6 than in structure 4.

The major structural difference between structures 3 and 5 vs. 4 and 6 is the position of the metal atoms in the cofactor. In models 3 and 5, Fe(III) occupies the metal site 2 that is further away from Phe127, while in model structures 4 and 6 Fe(III) occupies the metal site 1 that is closer to Phe127 (see Fig. 1). Considering the EXAFS data, model 3 cannot be favored over model 4. Based on the simulated Mössbauer data alone, Han et al. concluded that model 4 is a better structure than model 3 because its Fe1(III)-Mn2(III) and higher oxidation state Fe1(IV)-Mn2(IV) models’ calculated ⁵⁷Fe quadrupole splitting values (Table 4) agreed better with experiment.¹⁶ The DFT calculations performed by Younker et al. showed that placing Fe(III) in site 1 gave better isomer shifts for the mixed valence Fe(III)-Mn(IV) state.¹¹ Nevertheless, the computed isomer shift of structure 3 is somewhat closer to experiment than that of structure 4. Thus, considering the EXAFS, Mössbauer isomer shift, and quadrupole splitting results, it cannot be decided clearly if structure 3 or 4 is in better agreement with experiment, and no statement about the exact position of the metal atoms can be made based on these results. This issue is also not resolved if spin-projected energies are taken into account since they solely differ by about 0.9 kcal/mol for structures 3 and 4 with structure 3 lower in energy.¹⁶ The situation is different for models 5 and 6. Although the EXAFS fitting results of course favor structures 5 and 6 compared to models 3 and 4, they do not clearly favor either model 5 or 6 based on the R-factor alone. We note, however that, the Debye-Waller factors are mainly smaller for model 5 than for model 6 or any of the other models 1–4. Since Debye-Waller factors are treated as free fitting parameters, and the experimental sample temperature is low (near that of liquid N₂), the fit with the smaller Debye-Waller factors, model 5 is preferred. Also the computed Mössbauer properties (δ, ΔE_Q, η) of model 5 deviate less from the experiment values than those of structure 6. This leads to the conclusion that model 5 is the one in best agreement with the experimental EXAFS and Mössbauer data used in this study. The spin-projected energy of model 5 is found to be 1.7 kcal/mol higher than the corresponding energy of model 3 but it is 2.7 kcal/mol lower than that of model 6.

In earlier work,¹⁶ we proposed that a model where Fe is nearer Phe127, and Mn is farther from Phe127 was likely based on calculated mono-Mn(II) site binding energies, and on comparisons of calculated versus experimental Mössbauer parameters for two oxidation states Mn(III)-Fe(III) and Mn(IV)-Fe(IV) (see Table 4). In particular, the predicted quadrupole splittings ΔE_Q are in better agreement with experiment for the Fe1-Mn2 site arrangement than for the opposite Mn1-Fe2 ordering. However, for experimental reasons, the accuracy of Mössbauer parameter determination has significant limitations for these two oxidation states (see Table 4, footnotes b and c for more details) in comparison with the mixed valence Mn(IV)-Fe(III) state. The latter has a much sharper line width Г = 0.30 mm s⁻¹, and is well resolved at low temperature T = 4.2 ˚K.

Turning to the question of metal ion binding to form an Fe-Mn heterodinuclear center, a close reading of Younker et al.¹¹ and the very recent paper by Dassama et al.⁵⁰ shows that our earlier calculations and analysis used an overly simple model. One standard protocol (used to prevent formation of Fe(II)-Fe(II) and Fe(III)-Fe(III) after reaction with O₂) is to start with an excess of Mn(II) in the solution in air and then to add Fe(II) slowly. Like Fe(II)-Fe(II), Mn(II)-Fe(II) reacts with O₂, while Mn(II)-Mn(II) is labile and does not react with O₂. It is clear that the Mn-Fe complex formation is complicated and involves kinetic as well as thermodynamic issues. The rationale for the site specificity of the Mn(II)-Fe(II) complex and for the higher oxidation states leading to Mn(IV)-Fe(III) cannot be solved by considering the mono-Mn(II) site binding affinity in isolation.

Conclusions

Based on comparison of experimental data with simulated Mn and Fe EXAFS spectra as well as calculated Mössbauer properties, we find that structure 5 is the model out of the six models considered in this paper that agrees best with experimental data. The EXAFS simulations of models 5 and 6 show that these models are, in contrast to the ones of structures 3 and 4, consistent with the experimental EXAFS data. A significant difference is observed in case of the computed Mössbauer properties where model 5 agrees better with experiment than model 6. In addition, structure 5 is found to be lower in energy than model 6.

Structure 5, like structure 6, has been derived by fitting of interatomic distances to experimental EXAFS data. The calculated Mössbauer data further support this structure as a good model for the Ct-R2 Mn(IV)-Fe(III) active site.

The metal site binding affinity calculations on Ct-R2 for Mn(II)-binding¹⁶ and on E. coli-R2 for Fe(II)-binding⁶⁷ show that site 2 in both proteins has higher metal-binding affinity. A single-Fe containing mouse R2 structure was crystallized at pH 4.7 (PDB code: 1XSM), and Fe was also found bound at site 2.⁶⁸ Current experimental X-ray^{49, 50} and EXAFS data¹¹ and our DFT calculated energies⁶⁷ versus structure leads to the proposal that, for RNR-R2 structures the metal site 2 has higher metal-binding affinity. In Ct-R2, site 2 has higher Fe(II)-binding affinity than Mn(II)-binding based on X-ray anomalous diffraction data.^{49, 50} Even if site 2 is occupied by Mn(II) when Mn(II) is added to apo-R2 prior to the addition of Fe(II),⁵⁰ the Mn(II) on site 2 will be replaced by Fe(II) when a smaller amount of Fe(II) is subsequently added. If more Fe(II) is added, the Mn(II) on site 1 will also be replaced by Fe(II) — forming the inactive diiron form of Ct-R2. The relative Mn(II)- and Fe(II)-binding affinities for sites 1 and 2 in Ct-R2 and O₂ reaction merit further work. Our current EXAFS calculations support the recent experimental observations that Fe mainly occupies site 2 and Mn site 1 in the Ct-R2 Mn/Fe active site. Further, the combined DFT-EXAFS simulation refinement methodology we have developed should prove valuable for analyzing structures and metal site specificity in metalloenzymes.