Rapid X-ray Photoreduction of Dimetal-Oxygen Cofactors in Ribonucleotide Reductase

Background: Typical FeFe and MnFe cofactors bind to numerous enzymes such as ribonucleotide reductases. Crystallographic data suggest x-ray photoreduction (XPR) effects. Results: Rapid XPR-induced cofactor changes were monitored using time-resolved x-ray absorption spectroscopy. Conclusion: The XPR-induced cofactor states differ significantly from the native configurations, but comply with crystallographic structures. Significance: Structure determination for high-valent dimetal-oxygen cofactors requires free electron-laser protein crystallography combined with x-ray spectroscopy. Prototypic dinuclear metal cofactors with varying metallation constitute a class of O2-activating catalysts in numerous enzymes such as ribonucleotide reductase. Reliable structures are required to unravel the reaction mechanisms. However, protein crystallography data may be compromised by x-ray photoreduction (XRP). We studied XPR of Fe(III)Fe(III) and Mn(III)Fe(III) sites in the R2 subunit of Chlamydia trachomatis ribonucleotide reductase using x-ray absorption spectroscopy. Rapid and biphasic x-ray photoreduction kinetics at 20 and 80 K for both cofactor types suggested sequential formation of (III,II) and (II,II) species and similar redox potentials of iron and manganese sites. Comparing with typical x-ray doses in crystallography implies that (II,II) states are reached in <1 s in such studies. First-sphere metal coordination and metal-metal distances differed after chemical reduction at room temperature and after XPR at cryogenic temperatures, as corroborated by model structures from density functional theory calculations. The inter-metal distances in the XPR-induced (II,II) states, however, are similar to R2 crystal structures. Therefore, crystal data of initially oxidized R2-type proteins mostly contain photoreduced (II,II) cofactors, which deviate from the native structures functional in O2 activation, explaining observed variable metal ligation motifs. This situation may be remedied by novel femtosecond free electron-laser protein crystallography techniques.

Prototypic dinuclear metal cofactors with varying metallation constitute a class of O 2 -activating catalysts in numerous enzymes such as ribonucleotide reductase. Reliable structures are required to unravel the reaction mechanisms. However, protein crystallography data may be compromised by x-ray photoreduction (XRP). We studied XPR of Fe(III)Fe(III) and Mn(III)-Fe(III) sites in the R2 subunit of Chlamydia trachomatis ribonucleotide reductase using x-ray absorption spectroscopy. Rapid and biphasic x-ray photoreduction kinetics at 20 and 80 K for both cofactor types suggested sequential formation of (III,II) and (II,II) species and similar redox potentials of iron and manganese sites. Comparing with typical x-ray doses in crystallography implies that (II,II) states are reached in <1 s in such studies. First-sphere metal coordination and metal-metal distances differed after chemical reduction at room temperature and after XPR at cryogenic temperatures, as corroborated by model structures from density functional theory calculations. The intermetal distances in the XPR-induced (II,II) states, however, are similar to R2 crystal structures. Therefore, crystal data of initially oxidized R2-type proteins mostly contain photoreduced (II,II) cofactors, which deviate from the native structures functional in O 2 activation, explaining observed variable metal ligation motifs. This situation may be remedied by novel femtosecond free electron-laser protein crystallography techniques.
Dinuclear transition-metal cofactors play important roles in small molecule conversions by a broad range of biological enzymes. A prominent example are the prototypic so-called dimetal-oxygen cofactors, which are found, for example, in ribonucleotide reductases (RNRs) 5 (1)(2)(3), methane monooxygenases (4,5), and in various other oxidases (6 -8). RNRs are essential for DNA synthesis in all organisms (9). In these enzymes, the metal cofactor is located in the R2 subunit, whereas the R1 subunit houses the active site of ribonucleotide reduction. In the classical R2 of Escherichia coli class-Ia RNR, two iron atoms form the cofactor, which are bound to the protein by the amino acid side chains of two histidine residues, three glutamates, and one aspartate, in a highly conserved binding motif (10,11).
The metal cofactor in R2 proteins functions in dioxygen (O 2 ) reduction (12,13). Activation of the enzymes is achieved by O 2 reduction at the initially divalent metal ions. Thereby the Oatoms become incorporated for example, as metal-bridging O or OH species (12)(13)(14) so that a high-valent Fe(IV)Fe(III) site (intermediate X) is formed (15)(16)(17). Subsequent oxidation of a nearby tyrosine residue to a radical (Y ⅐ ) leaves the cofactor in an Fe(III)Fe(III) state. The catalytic cycle starts with the generation of a cysteine radical at the ribonucleotide binding site in R1 via long-range proton-coupled electron transfer to Y ⅐ , initiating ribonucleotide reduction, in the course of which Y ⅐ is restored by reverse ET for further turnover (18 -23).
Recently, a new class-Ic RNR has been described in the human pathogenic bacterium Chlamydia trachomatis (Ct), in which a hetero-dinuclear MnFe cofactor is assembled instead of an FeFe site and furthermore, the radical-forming tyrosine is replaced by a redox-inert phenylalanine (Phe-127) (24 -27). Extensive investigations on the Ct enzyme have revealed that a Mn(IV)Fe(III) site, as formed after O 2 reduction at divalent manganese and iron ions, is the target of the proton-coupled electron transfer from R1 (25)(26)(27)(28)(29)(30). Also in CtR2 under certain conditions an FeFe cofactor can be incorporated, but shows significantly lower activity (31,32). Similar MnFe sites meanwhile have been found in a number of related enzymes (7,8,(33)(34)(35)(36) and in addition, MnMn cofactors were described in the tyrosine-radical forming class-Ib RNRs (37)(38)(39). An important common feature of all three types of cofactors is the participation of high-valent intermediates, for example, M(III)M(III) and M(IV)M(III), in the catalytic reactions (25,28,35,36).
The Protein Database was used for structures. Most R2 structures are for FeFe-or MnMn-containing proteins. Only very recently, three structures have been reported for CtR2 including a MnFe cofactor (41,42). In previous studies (see Refs. 29 and 30), we have surveyed the R2 structures. In none of the structures did the apparent mean metal oxidation state exceed ϳ2.5 and the average value was close to 2, suggesting the predominance of M(III)M(II) and Mn(II)M(II) sites. The latter holds true also for structures that have been obtained for initially oxidized R2 proteins. In addition, considerable observed variability in metal coordination by oxygen species and amino acid groups is unexplained on a functional basis. Important structural parameters, such as the metal-metal distance, which is determined, e.g. by the type of metal-bridging oxide species (4,30,43,44), are not correlated to the ligation environment or to the increasing crystallographic resolution over the last 20 years. Typical metal(II)-ligand bond lengths are often observed (29,30).
These observations suggest that the metal sites in the crystal structures in fact do not correspond to high-valent states (29,30). The likely reason for this is x-ray photoreduction (XPR) of initially high-valent cofactors during diffraction data collection (29,30,45). However, for well founded attributions, for example, on the O 2 -activation process, reliable structural information is also required for the high-valent cofactors.
X-ray irradiation induced modifications are a long-known problem in protein crystallography (46 -49). XPR of metal cofactors often occurs at x-ray doses that are orders of magnitudes lower than used in x-ray crystallography causing radiation damage of the protein matrix (50 -62). XPR therefore has even been used to study reduced states of metal centers (51)(52)(53)56). However, because of a limited, but significant mobility of atoms in proteins at the cryogenic temperatures (around 100 K) usually employed in crystallography, XPR-induced and native structures of cofactors in reduced states may be expected to be different (58,60). By x-ray absorption spectroscopy (XAS), structural (metalligand bond lengths, metal-metal distances) and electronic (oxidation state) parameters also for high-valent metal sites can be determined (63)(64)(65) and furthermore, this technique allows studies of the XPR kinetics (58 -60, 66). Using XAS, for CtR2 we have shown previously that under high-intensity x-ray irradiation the Mn(IV)Fe(III) cofactor becomes reduced by XPR within seconds to minutes even at 20 K (30). However, quantitative relationships between XPR-induced structural changes and metal oxidation states so far have not been obtained for the dimetal-oxygen cofactors. This is necessary to reconcile the site configurations in the crystal data with deviating structural fea-tures, for example, usually much shorter metal-metal distances, as determined by XAS (29, 30, 66 -68). In principle, this would allow for the reversion of modifications due to XPR in the crystal structures in silico, using quantum chemical calculations (64,69,70), to obtain improved models of the high-valent sites.
In the present investigation, we used XAS to monitor XPR in CtR2 initially containing Fe(III)Fe(III) or Mn(III)Fe(III) cofactors at cryogenic temperatures. The XPR kinetics were determined by time-resolved XANES measurements and the structural changes, as derived from EXAFS analysis, were put in relationship to the XPR-induced metal oxidation states. Rapid XPR was observed for both cofactor types. This gave rise to site configurations, which differ from the ones in the native oxidized or reduced CtR2 proteins, but are in agreement with crystallographic data.

MATERIALS AND METHODS
Protein Sample Preparation-Recombinant Ct R2 protein was overexpressed in E. coli and R2 protein was purified and concentrated as described previously (see Refs. 29 and 30 and references therein). R2 protein containing an FeFe cofactor was obtained without addition of metal ions to the TB growth medium. R2 protein containing predominantly a MnFe cofactor was obtained by the addition of 30 M MnCl 2 to the LB medium (iron concentration 8 M) after induction with isopropyl isopropyl-␤-D-thiogalactopyranoside. Reduction of R2 was achieved by the addition of 100 M TCEP (Tris(2-carboxyethyl)phosphine hydrochloride) to the cell lysis buffer. R2 protein (monomer) concentrations were derived photometrically using an extinction coefficient at 280 nm of 57,750 M Ϫ1 cm Ϫ1 for Ct R2 (27 . Metal Content Quantification-Metal contents of R2 samples were quantified by total-reflection x-ray fluorescence analysis (TXRF) (29, 30, 72) on a PicoFox spectrometer (Bruker) using a gallium metal standard (Sigma) and the respective R2 polypeptide concentrations.
X-ray Absorption Spectroscopy-XAS was performed at the SuperXAS beamline of the Swiss Light Source (SLS) at Paul Scherrer Institute (Villigen, Switzerland) using a double-crystal Si [111] monochromator. Higher harmonics were suppressed by a silicon mirror in grazing incidence mode. The synchrotron was operated in top-up mode at 400 mA ring current, providing constant x-ray flux at the sample position, which was attenuated by carbon foil absorbers in the beam when necessary. A rapid shutter in front of the I 0 ion chamber prevented x-ray irradiation of samples prior to the XAS scans. The spot size on the samples was ϳ0.2 mm (vertical) ϫ ϳ1.0 mm (horizontal) as set by slits. Samples were positioned with micrometer precision in the beam by moving the cryostat on a computer-controlled stage. K␣ fluorescence-detected XAS spectra at the iron and manganese K-edges were collected using an energy-resolving 13-element germanium detector (Canberra), which was shielded by 3-m chromium foil (Mn) or 6-m manganese foil (iron) against scattered x-rays. Samples were held in a liquidhelium cryostat (Oxford). Detector dead time-corrected XAS spectra were averaged after energy calibration using an iron metal foil (iron) or a KMnO 4 powder sample (manganese) as energy standards (73,74). XANES spectra were normalized and EXAFS oscillations were extracted as described previously (30). The duration of EXAFS scans up to a k value of 12.5 Å Ϫ1 was ϳ22 min; the scan duration up to the K-edge was ϳ1-2 min. EXAFS spectra were derived using E 0 values of 7112 eV (iron) and 6540 eV (manganese); E 0 was refined to 7120 Ϯ 2 eV (iron) and 6547 Ϯ 2 eV (manganese) in the least-squares simulations of unfiltered k 3 -weighted spectra (in-house program SimX (75), phase functions were calculated by FEFF8 (76,77), amplitude reduction factors, S 0 2 , of 0.9 for iron and 0.85 for manganese). The error sum (R F ) was defined as described elsewhere (75). Fourier transforms (FTs) of EXAFS spectra were calculated using k values of 2-12 Å Ϫ1 and cos 2 windows extending over 10% at both k range ends. K-edge energies were determined at the 50% level of the normalized XANES. Time scans of the x-ray fluorescence were recorded at given monochromator energies at the K-edges to monitor XPR (60), using an acquisition period of 2 s/data point. Kinetic XPR data were simulated by sums of monoexponential functions.
Bond Valence Sum Calculations-BVS calculations (78) were done as previously described (29 -30), using in-house software for the analysis of crystal structure files from the Protein Database, and included atoms within a radius of 2.7 Å around the metal ions. Given BVS values represent the average of calculations using R 0 values for metal(II) and metal(III) species (30).
Density Functional Theory Calculations (DFT)-Spin-unrestricted geometry optimizations and calculations of electronic parameters of structural models of the metal sites were performed using the ORCA program package (79) as described previously (30,80). Geometry optimizations involved the BP86 exchange-correlation functional (81) with a triplevalence (TZVP) basis set (82). One set of polarization functions was used for all atoms. The resolution of identity approximation was used with the auxiliary TZV/J Coulomb fitting basis set (83) and a dielectric constant of ⑀ ϭ 4 in a COSMO solvation model (84). To derive the correct spin coupling of the two metal atoms, broken symmetry formalism using the flip-spin technique as implemented in ORCA (85, 86) was applied as described before (30). The initial geometry optimizations of (III,III) cofactors were based on crystal structure PDB code 1SYY of wild type Ct FeFe R2. For further restrictions in the DFT calculations see the "Results."

Properties of Cofactors in CtR2
Crystals-Two crystal structures of the wild type CtR2 MnFe protein have been published (24,32,41,42). Earlier structures of CtR2 proteins produced in E. coli without addition of or reconstitution with manganese should relate to CtR2 FeFe proteins (wt and the F127Y mutant) (24,32,41,42). The structures were initially obtained using oxidized R2, and were expected to contain mainly Fe(III)Fe(III) or Mn(III)/(IV)Fe(III) cofactors. All structures show different coordination motifs of the metal ions, in particular with respect to metal-bridging oxides and orientation of the carboxylate group of Glu-227 ( Fig. 1). BVS calculations reveal apparent oxidation states, which are close to the divalent level even for the 6-coordinated metal ions ( Fig. 1, see legend); the average metalligand bond lengths thus are in the range for Mn(II) and Fe(II) species.
The metal-metal distances span about 3.07-3.47 Å (Fig. 1) and are seemingly unrelated to the metal coordination. The coordinate error may be expected to be on the order of at least 10% of the crystallographic resolution (87). This means that the metal-metal distances in the CtR2 structures at a resolution of ϳ2 Å may be considered as equal within error limits. These distances are within the range of metal-metal distances observed in other R2 structures, which show an even broader distance distribution centered around ϳ3.6 Å, and similar site heterogeneity has also been observed for related cofactors in other enzymes (see Refs. 29 and 30) and references therein). We consider these results as evidence that during diffraction data collection XPR altered the cofactor structures.
Metal and MnFe/FeFe Cofactor Contents-FeFe-and MnFecontaining CtR2 samples were prepared, which contained airoxidized (ox) or chemically reduced (red) proteins and are denoted CtR2 FeFe ox , CtR2 FeFe red , CtR2 MnFe ox , and CtR2 MnFe red . The sample conditions (29,30) were chosen to allow for comparison of the cofactor properties in the native oxidized and reduced states with those of XPR-induced species.
Metal contents were in the range of about 0.5-0.7 mM manganese and 0.9 -3.2 mM iron in the XAS samples, which contained about 1-2 mM R2 polypeptide ( Table 1). The binding sites in CtR2 FeFe were occupied near quantitatively with close to two iron atoms per R2 polypeptide; manganese was at the detection limit. A slightly lower apparent occupancy was observed for CtR2 MnFe , in which the manganese content was ϳ55% of the iron content. The metal contents of oxidized and reduced samples were rather similar. These observations are in agreement with previous results (25,29,30,71). An elevated Zn content was found only in the CtR2 red samples. The maximal relative amounts of MnFe and FeFe sites were calculated from the metal and R2 concentrations (Table 1), yielding close to 100% FeFe sites in CtR2 FeFe . In CtR2 MnFe , sufficiently high amounts (70%) of MnFe sites were obtained (Table 1), which allowed for determination of the properties of the hetero-metal sites by XAS.
Structure and XPR of the FeFe Cofactor-XANES spectra at the Fe K-edge were recorded for the CtR2 FeFe samples ( Fig. 2A). They revealed K-edge energies, which indicated close to 100% Fe(III) in CtR2 FeFe ox , but about 70% Fe(III) and 30% Fe(II) in CtR2 FeFe red . These values were obtained by comparison to iron reference compounds, revealing a K-edge energy difference of ϳ2.8 eV between the Fe(II) and Fe(III) levels ( Table 2). Combining the XANES and TXRF data, ϳ100% Fe(III)Fe(III) sites for CtR2 FeFe ox , but ϳ57% Fe(III)Fe(III) and ϳ43% Fe(III)Fe(II) for CtR2 FeFe red were calculated. For monitoring of XPR of the initial Fe(III)Fe(III) cofactor, a series of consecutive. XAS scans were performed on single spots of CtR2 FeFe ox samples. Respective XANES spectra for about 2, 90, and 265 min of x-ray exposure at 20 K ( Fig. 2A) revealed K-edge energies, which decreased from an initial mean oxidation state of 3 to a level of about 2.2, suggesting ϳ80% of Fe(II) after prolonged XPR (Table 2). A plot of the K-edge energies of all obtained XANES spectra versus the x-ray exposure periods revealed biphasic XPR kinetics at 20 K (Fig. 2B). A similar behavior was observed when the normalized x-ray fluorescence level at a constant excitation energy of 7123 eV in the iron K-edge was used as an alternative measure of XPR (Fig. 2B). Both data sets were well simulated using sums of two exponential functions with similar time constants () of about 60 and 260 min and amplitudes close to 50% each (Fig. 2B, Table 3). We attribute these kinetic phases (i) to the more rapid formation of Fe(III)Fe(II) states, terminated after about 1.5 h, and (ii) to the slower reduction to the Fe(II)Fe(II) level, finished after about 7 h at 20 K. Kinetic XPR data for CtR2 FeFe ox at an increased temperature of 80 K still revealed biphasic behavior, but acceleration of both reduction phases compared with the 20 K data (Fig. 2B, inset); so that 1 and 2 were decreased to ϳ1 and ϳ18 min, respectively ( Table 3).
Changes of the coordination geometries of the iron sites were assayed by analysis of the pre-edge features in the XANES (88) (Fig. 2A, inset). For CtR2 FeFe ox , the small pre-edge amplitude  Table 3). Upper inset, K-edge spectra around the 50% level of CtR2 FeFe ox and CtR2 FeFe x260 , the arrows mark the edge downshift (⌬E) and corresponding fluorescence intensity increase (⌬F) for an excitation energy of 7123 eV. Lower inset, expanded view of isolated pre-edge peaks in the XANES (asterisk). B, K-edge energies (triangles, left y axis) and x-ray fluorescence intensities at 7123 eV (circles, right y axis) at 20 K, together with the fit curves calculated using parameters in Table 3. Inset, time scan trace of the x-ray fluorescence intensity at 7123 eV at 80 K, K-edge data at 20 K, and fit curves (Table 3) on a logarithmic axis. Traces were normalized to unity amplitude after offset subtraction on the basis of the fit results. The given dose values (in Gy, gray ϭ J kg Ϫ1 ) are approximate and correspond to the incident dose, the absorbed dose in 1 mm of water for an energy of 7125 eV is ϳ77% of the incident dose. Dose values were calculated for an energy of 7125 eV ϭ 1.14 ϫ 10 Ϫ14 joule, a flux (upper limit) of 10 11 photons s Ϫ1 , and an irradiated volume of 0.2 mm 3 corresponding to 2.15 ϫ 10 Ϫ7 kg (aqueous buffer plus R2 protein contributing about 0.15 ϫ 10 Ϫ7 kg at 1.8 mM and a molecular mass of ϳ43 kDa).

TABLE 1 Metal and protein concentrations and relative amounts of cofactor species
Metal concentrations were determined by TXRF. Relative amounts of cofactor species represent maximal figures, which were calculated on the basis of the metal and R2 polypeptide concentrations and assuming that all Mn ions were incorporated in MnFe sites. Values in parentheses were calculated under the assumption that all zinc ions were included in FeZn sites (114). Other transition metal species were negligible in the R2 samples. suggested predominately 6-coordinated Fe(III) ions, whereas the increased pre-edge peak at ϳ0.4 eV lower energy for CtR2 FeFe red suggested 5-coordinated iron at least in the 30% Fe(II) fraction. A pronounced increase and shift to lower energy of the pre-edge was observed for increasing x-ray exposure periods. Accordingly, this was explained by a change from 6-coordinated Fe(III) to 5-coordinated Fe(II) in the major sample fraction after prolonged XPR.

Sample
EXAFS on the native oxidized and reduced states, and on photoreduced CtR2 FeFe samples was performed to study changes of interatomic distances at the metal cofactor (Fig. 3). The FTs of the EXAFS spectra (Fig. 3A, inset) of CtR2 FeFe ox and CtR2 FeFe red revealed relatively small differences, e.g. changes of the FT maxima reflecting Fe-Fe distances. Visual inspection of the FTs for increasing x-ray exposure periods revealed systematic changes, for example, an initial shift to shorter distances and later diversification of the main FT maximum due to Fe-Fe distances and shifts of the main FT maximum due to Fe-ligand interactions (Fig. 3A).
By simulation of the EXAFS spectra (curve fitting) the structural changes were quantified (Fig. 3A). The main difference in CtR2 FeFe red in the first iron coordination sphere was an increase of the Fe-O,N bonds lengths and a slightly increased coordina-

TABLE 3 XPR kinetics for manganese and iron
Relative amplitudes (A) and time constants () resulted from double-exponential simulations of K-edge energy (E) and X-ray fluorescence intensity (F) data for the two metals (Figs. 3B and 5), using appropriate offset values.  Table  4 (Fe-O,N distances); and B (Fe-Fe distances) for the indicated CtR2 FeFe states and mean x-ray exposure periods. Spectra were vertically displaced for comparison. B, coordination numbers (N Fe-Fe ) for EXAFS data in A of 3 Fe-Fe distances for oxidized (ox) and reduced (red) CtR2 FeFe samples (depicted at "0" levels, meaning for negligible XPR) and for CtR2 FeFe ox after the indicated mean XPR periods. Fe-Fe distance margins (in parentheses) and vertical bars denote the full range of R and N variations for different fit approaches (simulations of individual spectra or joint simulations of all spectra including variations in the fit restraints for Fe-O,N shells (40)). Inset, mean Fe-O,N distances calculated from the data in Table 4. Smooth curves show fits of the N values by sums of two exponential functions using time constants in Table 3 and appropriate offsets and amplitude signs. Given dose values are approximate and were calculated as detailed in the legend of Fig. 2. tion number (N) but largely increased Debye-Waller parameter (2 2 ) for the shorter bonds, compared with CtR2 FeFe ox ( Table  4). In particular the 2 2 increase suggested significant amounts of 5-coordinated iron ions in CtR2 FeFe red . Fe-Fe distances (R) of about 2.95, 3.05, and 3.40 Å readily accounted for the Fe-Fe interactions. The 3.05-Å distance represented the main species in CtR2 FeFe ox (ϳ70%), whereas for CtR2 FeFe red it was the ϳ3.40 Å distance (ϳ55%) (Fig. 3B). The ϳ2.95-Å distance (ϳ10%) was at the detection limit. The Fe-Fe distance changes and coordination numbers in oxidized and reduced samples suggested that in the Fe(III)Fe(III) site the iron atoms are separated by ϳ3.05 Å and ϳ3.40 Å distance accounts for an Fe(III)Fe(II) site (29,30).
Simulations of a series of EXAFS spectra for the x-ray-exposed CtR2 FeFe samples revealed the following (Fig. 3B, Table  4). The structural parameters for ϳ15 min XPR were similar to those for CtR2 FeFe ox , revealing a predominant contribution of the ϳ3.05 Å Fe-Fe distance. For increasing x-ray exposure periods up to ϳ240 min, for the main Fe-O shell a simultaneous increase of the coordination number (N), the Debye-Waller parameter (2 2 ), and the bond-length (R) were determined, whereas for the initially longer second Fe-O,N interaction decreased N and R values (for constant 2 2 ) were obtained ( Table 4). The mean Fe-ligand bond lengths, however, revealed an initial decrease up to ϳ100 min XPR and an increase for longer periods (Fig. 3B, inset). Also for CtR2 FeFe red , increased N and 2 2 values for the Fe-O shell compared with CtR2 FeFe ox were observed. These effects were explained by accumulation of the 5-coordinated Fe(II) during XPR.
For the Fe-Fe distances (Fig. 3B), an initial increase of the relative coordination number of the ϳ2.95 Å distance to ϳ50% after ϳ70 min of x-ray exposure occurred at the expense of the ϳ3.05 Å distance. It was followed by a decrease of the ϳ2.95-Å contribution and a concomitant increase of a distance close to ϳ3.45 Å to ϳ55% after 240 min. The kinetic behavior of the three Fe-Fe distances was reasonably well described using the same time constants as determined for XPR from the XANES analysis (Fig. 3B). This suggested an Fe-Fe distance shortening from ϳ3.05 Å to ϳ2.95 Å for the initial Fe(III)Fe(III) 3 Fe(III)-Fe(II) reduction step, which was followed by an elongation from ϳ2.95 Å to ϳ3.45 Å upon the Fe(III)Fe(II) 3 Fe(II)Fe(II) transition.
Structure and XPR of the MnFe Cofactor-The XPR kinetics of CtR2 MnFe ox samples were monitored using similar approaches as described above (Fig. 4). Biphasic reduction behavior was observed both at the manganese and iron K-edges. Kinetic simulations of the 20 K data showed that the first phase was by a factor of about 5-10 faster than the second phase for manganese and iron ( Table 3). The time constants of the two phases for manganese reduction were by factors of about 1.5-2.5 smaller than the ones for iron reduction. At 80 K, pronounced acceleration of reduction resulted in quite similar kinetics for both metal species (Table 3). Overall, the time constants for reduction of manganese and iron for CtR2 MnFe and CtR2 FeFe samples, initially containing mostly Mn(III) and Fe(III), were similar within a factor of ϳ2 at both 20 and 80 K.
We note that a minor Mn(IV) contribution in CtR2 MnFe ox was suggested by the XANES spectrum (see below). However, the initial reduction phase of Mn(IV)Fe(III) sites has been found to be ϳ150 faster than the next step (30). This suggested values of about 0.3 and 0.01 min for Mn(IV)Fe(III) reduction at 20 and 80 K under the present conditions, which was at the kinetic resolution limit. Small amplitudes due to Mn(IV)Fe(III) reduction accordingly were not discernable for CtR2 MnFe .

TABLE 4 EXAFS simulation parameters for the first coordination shells of manganese and iron
N, coordination number, R, interatomic distance; 2 2 , Debye-Waller factor, R F , fir error sum (75) Table 3. Inset, data at 80 K on a linear time axis. The incident dose values are approximate (the absorbed dose in 1 mm of water for 6550 eV is ϳ85% of the incident dose); dose values were calculated using an energy of 6550 eV ϭ 1.05 ϫ 10 Ϫ14 joule, an irradiated volume of 2.11 ϫ 10 Ϫ7 kg (buffer plus about 0.11 ϫ 10 Ϫ7 kg of R2 protein at 1.3 mM, and further parameters given in the legend of Fig. 2. XANES spectra at the manganese and iron K-edges for CtR2 MnFe ox , CtR2 MnFe red , and for a state obtained after prolonged XPR denoted CtR2 MnFe x-ray are shown in Fig. 5. The spectrum of CtR2 MnFe x-ray resulted from summation of spectra for about 30 -210 min x-ray exposure (mean of ϳ130 min) and thus represents the average over several manganese and iron oxidation states. Compared with manganese reference compounds, the manganese and iron edge energies of CtR2 MnFe ox were close to the trivalent level; a minor Mn(IV) contribution (ϳ20%) was implied by the somewhat higher manganese edge energy ( Table 2) Metal site geometry changes again were deduced from the pre-edge features (Fig. 5, insets). Small pre-edge amplitudes for CtR2 MnFe ox suggested predominantly 6-coordinated (i.e. nearoctahedral) Mn(III) and Fe(III) ions. The diminished manganese and iron pre-edge peaks at Ͼ1 eV lower energies for CtR2 MnFe red implied a further symmetrization of the coordination of the divalent metal ions. At variance with the latter behavior, CtR2 MnFe x-ray showed increased manganese and iron pre-edge features at lower energies compared with CtR2 MnFe ox . This suggested formation of 5-coordinated manganese and iron ions after prolonged XPR.
Visual inspection of the EXAFS spectra of the three CtR2 MnFe states (Fig. 6) revealed that the main FT maximum of the manganese EXAFS due to first-sphere Mn-O,N distances was at higher distances in CtR2 MnFe red and CtR2 MnFe x-ray compared with CtR2 MnFe ox , suggesting an overall increase of the bond lengths for Mn(II). Less pronounced changes of the main FT maximum due to Fe-O,N bonds were observed. In the manganese EXAFS, two FT maxima at reduced distances of 2-3 Å were attributed to Mn-Fe interactions. For CtR2 MnFe ox the first maximum was larger than the second one, whereas for CtR2 MnFe red this was reversed. For CtR2 MnFe x-ray the main Mn-Fe FT peak was even at larger distances than for the other two samples. The iron EXAFS of CtR2 MnFe ox showed one main Fe-Mn,Fe distance, whereas for CtR2 MnFe red and CtR2 MnFe x-ray , several Fe-Mn,Fe distances were resolved (Fig. 6).
The simulation results for the manganese and iron EXAFS spectra are summarized in Table 4 (first coordination sphere) and Fig. 7 (metal-metal distances) (Table 2). Insets, corresponding isolated pre-edge features in magnification. At least three metal-metal distances were discernable both in the manganese and iron EXAFS spectra (Fig. 7). The main Mn-Fe distance in CtR2 MnFe ox was ϳ2.90 Å (ϳ55%) and minor contributions of a ϳ3.30-Å distance (ϳ35%) were observed. CtR2 MnFe red showed opposite magnitudes (ϳ25%, ϳ50%) of contributions from these two distances. Similar metal-metal distances in CtR2 MnFe have previously been attributed to Mn(IV)/(III)Fe(III) (ϳ2.90 Å) and Mn(II)Fe(III) (ϳ3.3 Å) sites (29,30,66). Here, the ϳ2.90 Å distance mostly reflected the Mn(III)Fe(III) cofactor, because Mn(IV) contributions were small. We note that in the iron EXAFS the ϳ3.3 Å distance was less well defined than in the manganese EXAFS so that several equivalent fits with distances around 3.1-3.4 Å were possible. This reflects contributions from the minor amounts of FeFe species (compare Table 1 and Fig. 3B).
In CtR2 MnFe x-ray , the main metal-metal distance was ϳ3.65 Å (ϳ55%) and contributions from the other two distances apparently were small (Fig. 7). A ϳ3.65 Å distance was close to the detection limit in CtR2 MnFe ox and CtR2 MnFe red . Accordingly, we assign the ϳ3.65 Å distance to the main XPR-induced state of the MnFe site in CtR2 MnFe x-ray . Comparison of the TXRF, XANES, and EXAFS results revealed that the ϳ3.65 Å distance most likely reflects a Mn(II)Fe(II) state of the cofactor, as preferably produced by XPR at cryogenic temperatures. That the ϳ2.9 Å distance in CtR2 MnFe x-ray remained at a level, which was comparable with CtR2 MnFe red , may be explained by a similar initial metal-metal distance shortening as observed for CtR2 FeFe .
Model Structures for XPR from DFT-Density functional theory calculations were employed to generate model structures for the native and XPR-modified cofactors. The restricted mobility of amino acids during XPR at cryogenic temperatures was simulated by fixation of carbon atoms during geometry optimization after addition of 1 electron to the initial Fe(III)-Fe(III) and Mn(III)Fe(III) sites or 2 electrons to the singly reduced structures. Structures were calculated for additions of electrons only or for balancing of respective surplus negative charges by protons.
Comparison of more than 20 DFT-optimized structures revealed the following (supplemental Table S1, Fig. 8). When only one C-atom of the side chains (C1 in Fig. 8) was fixed during geometry optimizations of single and double reduced states and no protons were added, the resulting structures showed excessive changes, for example, breaking the Glu-120 carboxylate and O(H) bridges, detachment of histidine ligands, and pronounced Glu ligand reorientations, leading, e.g. to 4-coordinated manganese and iron ions. We consider such large scale geometry changes as unlikely at cryogenic temperatures. Fixation of two C-atoms resulted in less severe structural changes, but the tendency for carboxylate bridge breaking and histidine detachment remained, still causing mostly low coor-   Table S1. dination numbers as well as metal-metal distance elongation already for single-electron reduction in some cases. The structural changes after additions of electrons and protons depended on the actual protonation site. As a trend, less pronounced changes were observed so that higher metal coordination numbers prevailed (supplemental Table S1).
Best agreement with the structural parameters from XAS was achieved by proton addition first to terminal OH groups at Fe2 or manganese in the single reduced sites and second to O(H) bridges in the double reduced sites, both for the FeFe and MnFe structures (Fig. 8). In this case, the first reduction for both structures resulted in a shortening of the metal-metal distance by ϳ0.1 Å in the mixed-valent states, similar to the experiment. This was followed by elongations to Ͼ3 Å of Fe-O(H) and Fe/Mn-OH 2 bonds, resulting in 5-coordinated Fe(II) and Mn(II) ions after the second reduction step and in a pronounced elongation of the metal-metal distance to about 3.4 -3.5 Å, as also inferred from the experimental data (Fig. 8). For a brief account on the electronic configurations of the DFT structures and their possible relationship to the redox potentials see supplemental Table S1.  (Fig. 9). This is in agreement with previous assignments (29,30,44). Chemical reduction of the (III,III) states by one electron at room temperature likely induces breaking of a O(H) bridge, presumably accompanied by a protonation, which explains the ϳ0.4 Å metal-metal distance increase (4,43,63,75,89) and metal-ligand bond elongations in the (III,II) cofactors. Fe(III) reduction leads to 5-coordinated Fe(II), at least in the FeFe site, whereas in particular for the MnFe site, Mn/Fe(II) formation may lead to the (partial) loss of one ion and to a remaining 6-coordinated site. For the (II,II) sites, the absence of O(H) metal bridges is likely, resulting in metal-metal distances exceeding 3.7 Å (Fig. 9), similar to the distances observed in crystal structures of reduced R2 proteins (29,30).

Structures of FeFe and MnFe Cofactors-Our
The structural differences between the FeFe and MnFe sites in CtR2 presumably are relatively subtle. On the structural basis, the reasons for the preferential use of differently metallated sites in the class-Ia RNRs (FeFe), in the class-Ib RNRs (MnMn), and in the tyrosine-radical lacking R2 protein of class-Ic Ct RNR (MnFe) therefore remain elusive. Tuning of the redox potential, for example, by different protonation states of bridging oxides (29,30,44,90,91), thus may play an important role in the functional diversification of the dimetal-oxygen cofactors.
Structural Changes during XPR-The main structural changes at the FeFe and MnFe cofactors of CtR2 that are induced by XPR at cryogenic temperatures apparently differ from the changes occurring during chemical reduction at room temperature ( Fig. 9). At low temperatures, single electron reduction of the (III,III) cofactors causes even a ϳ0.1 Å short-ening of the metal-metal distance and only relatively minor bond length changes in the first-sphere metal coordination, but leaves the O(H) bridging intact. A possible explanation for this behavior is the associated protonation, for example, of a terminal OH group. Reduction of the (III,II) states induces more severe changes, such as formation of 5-coordinated Fe/Mn(II) ions, which likely is explained by the breaking of at least one O(H) bridge and/or of a terminal metal-OH 2 bond. This leads to a Ն0.5 Å elongation of the metal-metal distances in the (II,II) states (Fig. 9). Apart from minor differences in the overall site geometries, this sequence of events seems to hold for FeFe and MnFe cofactors.
The structural changes during XPR at 20 K have to involve movements of metal ions and (first-sphere) ligand atoms on the order of 1 Å or less, to explain the observed bond length changes. Changes of similar magnitude have been inferred from XPR studies of other metalloproteins and synthetic complexes (53,54,58,60,61,(92)(93)(94). XPR even has been suggested as a tool for studying reduced states of metal cofactors (52,53,56,57). In the case of CtR2, our data suggest that both the mixed valence and fully reduced states of the FeFe and MnFe cofactors differ from their counterparts formed at ambient conditions. Presumably, this reflects the limited mobility of amino acid groups and water species in the frozen state. The XPR-induced states of dimetal-oxygen cofactors thus may have limited physiological relevance.
Our DFT results suggest that protonation events may play a role in the compensation of XPR-induced metal-centered charges even at 20 K. Indeed, proton movements in proteins and model systems can occur at low temperatures at least within a single hydrogen bond, if not over longer distances within extended hydrogen-bonded networks (95)(96)(97). This may be taken as evidence that terminal and/or metal-bridging O(H) ligands in CtR2 are integrated in hydrogen-bonding interactions, for example, with water molecules. Water molecules in hydrogen-bonding distance, for instance, to metal-binding carboxylates, indeed were found in high-resolution R2 structures (39,99,100). The proton distribution in XPR-induced structures thus may significantly differ from the respective native oxidation states.
XPR Kinetics and Redox Potentials-Metal site reduction during XPR mostly is anticipated to result from electron transfer from radical sites, which have captured photoelectrons produced after core level excitations mainly of C-, O-, and N-atoms of the protein matrix (46). In addition, the XPR rates for the R2 cofactors may be expected to be proportional to their redox midpoint potentials (E m ) (30,60,101).
The XPR rates for the Fe(III)Fe(III) and Mn(III)Fe(III) cofactors differ by less than a factor of 2 (30), pointing to E m values for the FeFe and MnFe cofactors, which are similar within about 150 mV (74,(102)(103). The absolute E m values for the Mn(IV)Fe(III) states presumably are on the order of 1 V (30) and the slower XPR of the (III,III) states thus suggests a significant E m drop (104). A similar E m implies relatively minor structural differences between the FeFe and MnFe sites at least in the (III,III) states, as observed.
Biphasic XPR kinetics were found for both the FeFe and MnFe sites. For the Fe(III)Fe(III) cofactor, this is straightforwardly explained by sequential Fe(III)Fe(II) and Fe(II)Fe(II) formation. For the MnFe cofactor, similar XPR rates for Mn(III) and Fe(III), in particular at 80 K, imply reduction of either ion at about equal probability. At first glance, the Mn(III) and Fe(III) ions thus seem to possess a similar E m . However, our DFT results suggested reduction first of the iron ion (supplemental Table S1). The even faster manganese reduction at 20 K therefore may be explained by the ϳ1.5-times larger absorption of the protein at the lower manganese K-edge energies, i.e. creation of more reducing sites within a given XPR period. This could lead to the apparent inversion of manganese and iron reduction rates.
Implications for X-ray Crystallography of Dimetal-Oxygen Cofactors-Our present and previous data (29,30,66) indicate that XPR of high-valent FeFe and MnFe sites in CtR2 and also in the standard type R2 protein from mouse can occur within sec-onds to minutes already under XAS conditions. The respective doses (at 80 K) are at least 100 times smaller than those resulting in radiation damage of amino acid groups (46,105).
Using an x-ray flux of ϳ10 11 photons s Ϫ1 at the iron and manganese K-edges as the upper limit and a spot size on the sample of ϳ0.2 mm 2 , time constants of about 1-1.5 min (mean of ϳ70 s) for the (III,III) states and 20 -40 min (mean of ϳ2000 s) for the (III,II) states were observed at 80 K, which is close to the 100 K often used in crystallography. A linear dose-rate relationship for the R2 cofactor reduction is suggested by the ϳ10fold decrease of the time constant from ϳ3500 to ϳ350 s (at 20 K) of (III,III) reduction for a flux increase from ϳ10 11 photons s Ϫ1 used here to ϳ10 12 photons s Ϫ1 used previously (30). A typically more focused beam of 50 m diameter for crystallography, i.e. a spot size of 0.002 mm 2 , and a similar flux of ϳ10 11 photons s Ϫ1 therefore are expected to cause about 100 times faster XPR, i.e. time constants of 0.7 s for the (III,III) states and 20 s for the (III,II) states. A flux of ϳ10 13 photons s Ϫ1 (or more) of undulator crystallography beamlines thus decreases both time constants to far below 1 s. That these estimates are realistic is exemplified by the observed reduction of the Mn(IV)Fe(III) site in CtR2 at 20 K within ϳ2 s at ϳ10 12 photons s Ϫ1 (30).
These results imply that for typical x-ray crystallography conditions, XPR of dimetal-oxygen cofactors is unavoidable. Moreover, this will not improve much for the use of liquidhelium temperatures, radical quenching agents (106), acquisition of only a few frames per crystal spot, moderately diminished doses, and faster detector readout. Accordingly, the initial high-valent sites in the crystals were mostly reduced to the (II,II) level. The respective metal-metal distance spread thus may in part reflect the remnants of (III,II) states. Further evidence for this view comes from our result that the Fe/Mn-Fe distances of 3.4 -3.7 Å of the XPR-induced (II,II) states match the distances in most R2 structures, irrespective of the initial valence state (29,30). Notably, the XAS-derived metal-metal distances for the (III,III) states of ϳ3 Å are considerably shorter than most of the crystallographic distances even for an upperlimit coordinate error of ϳ0.4 Å. The broad distribution of metal coordination motifs in the crystals may thus be due to the superimposition of several XPR-induced conformations.
Crystallographic information for initial (II,II) sites, which will not become further reduced, should be more reliable. However, in these states the metal complex is labilized, leading to single-metal sites and/or to heterogeneity in amino acid and oxygen species ligation geometries. The question arises why the significant XPR-induced structural changes at the metal site still allow for high-resolution crystallographic coordinates. The likely reason for this is that XPR is terminated already during the first few seconds of diffraction data collection and thereafter, the low-valence configuration of metal ions and ligands is stable.
In conclusion, crystal structures of higher valence states exceeding (II,II) or perhaps (III,II) levels of highly oxidizing dimetal-oxygen cofactors cannot be reliably obtained by conventional crystallography. A similar situation has been encountered for other types of high-potential metal cofactors (70,107). The recent inauguration of fourth generation synchrotron radiation sources based on free electron-laser techniques seems to offer a way to overcome the XPR problem (108 -110). Using femtosecond x-ray pulses, XPR may be outrun by termination of the scattering process prior to initiation of the photoelectron cascade. The first structures for the free electron-laser approach using protein nanocrystals at room temperature have been reported even for radiation-sensitive systems, albeit at moderate resolution (111,112). However, there is no principal limitation toward high resolution structures (113). Femtosecond crystallography on the superfamily of enzymes containing dimetal-oxygen cofactors appears to be an excellent scientific case for this emerging technique to obtain structures of highvalent intermediates.