Redox Intermediates of the Mn-Fe Site in Subunit R2 of Chlamydia trachomatis Ribonucleotide Reductase

The R2 protein of class I ribonucleotide reductase (RNR) from Chlamydia trachomatis (Ct) can contain a Mn-Fe instead of the standard Fe-Fe cofactor. Ct R2 has a redox-inert phenylalanine replacing the radical-forming tyrosine of classic RNRs, which implies a different mechanism of O2 activation. We studied the Mn-Fe site by x-ray absorption spectroscopy (XAS) and EPR. Reduced R2 in the R1R2 complex (R2red) showed an isotropic six-line EPR signal at g ∼ 2 of the Mn(II)Fe(II) state. In oxidized R2 (R2ox), the Mn(III)Fe(III) state exhibited EPR g values of 2.013, 2.009, and 2.015. By XAS, Mn-Fe distances and oxidation states of intermediates were determined and assigned as follows: ∼4.15 Å, Mn(II)Fe(II); ∼3.25 Å, Mn(III)Fe(II); ∼2.90 Å, Mn(III)Fe(III); and ∼2.75 Å, Mn(IV)Fe(III). Shortening of the Mn/Fe-ligand bond lengths indicated formation of additional metal bridges, i.e. μO(H) and/or peroxidic species, upon O2 activation at the site. The structural parameters suggest overall configurations of the Mn-Fe site similar to those of homo-metallic sites in other R2 proteins. However, the ∼2.90 Å and ∼2.75 Å Mn-Fe distances, typical for di-μO(H) metal bridging, are shorter than inter-metal distances in any R2 crystal structure. In diffraction data collection, such bridges may be lost due to rapid x-ray photoreduction of high-valent metal ions, as demonstrated here for Fe(III) by XAS.

Ribonucleotide reductases (RNRs) 3 are the only enzymes that, in all organisms, catalyze the reduction of ribonucleotides to their deoxy forms essential for DNA synthesis (1)(2)(3). RNRs also are important targets in cancer and antiviral therapy (4,5).
Class I RNRs found in eukaryotes and microorganisms (6) are heterotetrameric enzymes of R1 2 R2 2 organization. The R1 protein contains the nucleotide binding site and R2 houses a dinuclear metal center, which is the site of dioxygen (O 2 ) activation and, in conventional RNRs, is of the Fe-Fe type (7).
Extensive investigations on Fe-Fe RNRs from, e.g. Escherichia coli, Saccharomyces cerevisiae, Mus musculus, and Homo sapiens have established that the catalytic reactions involve activation of an O 2 molecule at the di-metal cluster to generate a high potential site, which oxidizes a nearby tyrosine residue to a tyrosyl radical, Y ⅐ (8 -10). In E. coli R2 this Tyr-122 is at ϳ6 Å distance to the nearest iron (11,12). Subsequent proton-coupled electron transfer (13) leads to the re-reduction of Y ⅐ and to the oxidation of a cysteine at the substrate binding site in R1 to a radical (C ⅐ ) (14,15). C ⅐ initiates ribonucleotide reduction involving disulfide formation by two additional cysteines (16). Regeneration of reduced cysteines requires electron input from external thio-or glutaredoxins and ultimately from NADPH (17).
At least the Fe(II) 2 , Fe(III) 2 , Fe(IV)Fe(III), and Fe(IV) 2 oxidation states of the metal center seem to be involved in the electron transfer reactions (18,19) of classic RNRs. The Fe(III)-Fe(IV) state, termed "intermediate X" (20,21), is crucial because it oxidizes the tyrosine to Y ⅐ , leaving the di-iron site in the Fe(III) 2 state. Y ⅐ usually survives a large number of catalytic cycles, but when it is lost, the inactive Fe(III) 2 -Met form remains (22). Reactivation of the enzyme first requires reduction of the metal site to Fe(II) 2 , which then must react with O 2 , leading to the cleavage of the O-O bond and again to the formation of Fe(III) 2 and Y ⅐ (22,23).
According to the above reaction sequences, a tyrosine radical and the Fe(IV)Fe(III) state (X) have been anticipated to be decisive for RNR function. This view has been challenged recently because R2 proteins of RNRs in several species have been discovered (24,25), containing a redox-inert phenylalanine instead of the tyrosine. One enzyme is found in the important human pathogenic bacterium Chlamydia trachomatis (Ct) (25,26). It is a fully functional RNR and the only RNR encoded in the genome of this organism (27,28). In E. coli, however, the Tyr 3 Phe exchange abolishes RNR activity (20,30). A tyrosyl radical still is not observed in the Phe 3 Tyr mutant of Ct RNR (31).
Ct R2 can be reconstituted with Fe(II) ions so that a typical Fe-Fe cofactor is formed, but in this case, RNR activity is low (28). More recent studies have revealed that a much higher enzyme activity (at least ten times higher) is obtained in the presence of stoichiometric amounts of manganese and iron (27,32). Thus, the Ct enzyme now is believed to represent the first RNR that contains a hetero-bimetallic Mn-Fe cluster in its native state. Presumably closely related Mn-Fe sites have been found in purple acid phosphatase from sweet potato (33) and in an N-oxygenase of Streptomyces thioluteus (34). These proteins seem to belong to a growing family of O 2 -activating Mn-Fe enzymes.
A large number of crystal structures of Fe-Fe-containing R2 proteins is available in the PDB data base. Consistently, the metal ions are coordinated by conserved amino acids, i.e. four glutamates and two histidines (9,35). In addition, a variable number of metal-bound oxygen species (H 2 O or OH) was detected. In all structures at least one carboxylate group bridges between the metal ions and furthermore, bridging oxo (O) or hydroxo (OH) species were observed, their number likely depending on the iron oxidation state. Similar structural features were observed in R2 proteins reconstituted with a Mn-Mn site (36,37).
Crystal structures of Ct R2 containing an Fe-Fe site recently have been reported (26,31). Overall, the metal coordination seems to be similar to, e.g. Escherichia coli R2; the significance of certain structural differences is unclear. In previous studies on Ct Mn-Fe RNR, high valent states, i.e. Mn(IV)Fe(III) and Mn(IV)Fe(IV), have been proposed to be involved in catalysis (27,32,38). Direct structural information on the Mn-Fe site of Ct R2, in particular in its high valent states, is indispensable to unravel the oxygen activation and ribonucleotide-reduction mechanisms of the tyrosine-less Mn-Fe RNRs. It also may serve as a benchmark for quantum-chemical calculations aiming at optimized structures of the Mn-Fe site.
In the present investigation, we use x-ray absorption spectroscopy (39,40) at the manganese and iron K-edges and EPR to characterize the Mn-Fe site of Ct RNR in the oxidized R2 protein (R2 ox ) and in reduced R2 in the R1R2 complex (R2 red ). The first information on the atomic structure (metal-ligand and metal-metal distances) and electronic configuration of several oxidation states is reported, individually for the manganese and iron ions. Possible structures of the Mn-Fe site are discussed.

MATERIALS AND METHODS
Protein Sample Preparation-Ct RNR expression vectors for truncated wild-type R1⌬1-248 and wild type R2 proteins were a kind gift of G. McClarty (University of Manitoba, Canada). Recombinant proteins were expressed and purified as described previously (28,31,32). To obtain Mn-Fe R2 protein, Mn(II) was added as 40 M MnCl 2 to the LB growth medium at the moment of R2 overexpression in the Ct cells, and the Mn(II) level in the medium (30 -50 M) subsequently was controlled by monitoring its six-line EPR signal in the supernatant of the bacterial suspension and by the addition of MnCl 2 . Protein R2 and R1⌬1-248 concentrations were determined photometrically using extinction coefficients of 116 mM Ϫ1 cm Ϫ1 and 274 mM Ϫ1 cm Ϫ1 at 280 nm. The total amount of Mn-Fe sites in R2 was quantified by recording EPR spectra (20 K, 0.8 milliwatt) of the Mn(III)Fe(III) state in a catalytic mixture (50 M R2, 100 M R1, 0.3 mM ATP, 1.8 mM MgCl 2 , 2 mM CDP, 10 mM DTT; incubated for 20 min at room temperature with 0.1 mM hydroxyurea) and signal comparison to a standard solution of 1 mM CuSO 4 in 10 mM EDTA (32). Reduction of as-isolated (oxidized) Mn-Fe R2 was achieved by incubation of R2 (ϳ50 M and ϳ16 ml) with ϳ100 M of R1, 0.3 mM ATP, 1.8 mM MgCl 2 , and 10 mM DTT at 0°C overnight under argon. Protein samples were concentrated (Amicon), filled into EPR and XAS sample holders, and frozen in liquid nitrogen.
Metal Content Quantification-Metal contents of protein samples were assayed by total-reflection x-ray fluorescence detection (TXRF) (41) on a PicoFox spectrometer (Bruker). Protein samples were mixed with a galium standard (Sigma, 5 mg/liter), 5 l of the samples were dried on quartz disks for ϳ1 h at ϳ50°C, and disks were mounted on the sample changer of the spectrometer. The excited x-ray fluorescence was recorded for 10 min for five samples each of R2 red and R2 ox . Iron and manganese concentrations were determined using the routines of the spectrometer and molecular weights of manganese of 54.93 g/mol and iron of 55.84 g/mol.
EPR Spectroscopy-9.5-GHz X-band EPR spectroscopy was carried out on a Bruker ESP300E spectrometer equipped with a rectangular microwave cavity. Samples were kept in quartz tubes in an Oxford ESR900 helium cryostat. For further spectrometer settings see the figure legends. For determination of g values the magnetic field was calibrated with a LiF standard (42). Spin quantification was performed by comparison of the double-integrated signal of RNR samples with that of a CuSO 4 standard.
X-ray Absorption Spectroscopy-K ␣ fluorescence-detected XAS spectra at the iron and manganese K edges were collected at 20 K using energy-resolving germanium detectors and helium cryostats as previously described (43,44) at beamline D2 of the EMBL outstation (at HASYLAB, DESY, Hamburg, Germany) and at beamline KMC-1 of BESSY (Berlin, Germany). Harmonic rejection was achieved by detuning of the Si(111) double-crystal monochromators to 50% (Fe edge) and 70% (Mn edge) of their peak intensities. Deadtime-corrected XAS spectra were averaged (11-28 scans) after energy calibration of each scan (for iron spectra using the peak at 7112 eV in the first derivative of the absorption spectrum of an Fe-foil and for manganese spectra using the pre-edge peak at 6543.3 eV in the absorption spectrum of KMnO 4 as energy standards (45); estimated accuracy Ϯ 0.1 eV). The respective scan ranges and durations were 6900 -8200 eV (up to k ϭ 16 Å Ϫ1 ) and ϳ1 h for iron spectra and 6400 -7100 eV (up to k ϭ 12 Å Ϫ1 ) and ϳ30 min for manganese spectra. One to two scans were performed per sample spot (x-ray spot size of 5 ϫ 1 mm 2 set by slits). XAS spectra were normalized, and EXAFS oscillations were extracted as described before (46). The energy scale of EXAFS spectra was converted to the wavevector scale (k scale) using E 0 values of 7112 eV (iron) and 6540 eV (manganese). Unfiltered k 3 -weighted spectra were used for least-squares curve-fitting with the in-house software SimX (46) using phase functions calculated by using FEFF-7 (47) and for calculation of Fourier transforms (FTs). An amplitude reduction factor, S 0 2 , of 0.9 was used in the EXAFS fits. From experimental XANES spectra the pre-edge peak region was extracted by subtraction of a polyno-mial spline through the main edge rise using the program Xanda. 4 The line shape of the pre-edge features was reproduced by fitting the intensities, I, by a least-squares algorithm to sums of Gaussian functions (Equation 1, E C , center energy).

RESULTS
Metal Content and Site Occupancy-Two types of Ct RNR protein samples were used in this spectroscopic study: (a) asisolated, oxidized R2 protein, denoted R2 ox , and (b) the same R2 protein extensively reduced with DTT in an R1R2 complex, R2 red . The final R2 ox protein concentration was ϳ0.85 mM R2-dimer. The R2 red sample was prepared using the same R2 preparation as for R2 ox . After mixing with protein R1 and reduction with DTT, the final R2 red protein concentration was ϳ0.75 mM R2-dimer.
The metal contents of samples were quantified by TXRF and also by previously described spectroscopy methods (32); briefly, the manganese content was determined by EPR in an acid-denatured protein sample, and the iron content was spectrophotometrically determined using an iron complex assay). Table 1 summarizes the results. The TXRF and spectroscopy methods to determine the total manganese and iron concentrations were in good agreement, as seen for the data on R2 ox . In the estimations shown below we have used the data from the spectroscopy methods. The as-isolated R2 protein (R2 ox ) used in the XAS experiments was produced under conditions of constant and moderate manganese concentration (40 M) in the overexpression medium; the iron concentration in the medium was ϳ10 M (supplemental Table S1). This resulted in R2 protein with ϳ0.86 manganese per R2-dimer. The amount of Mn-Fe sites in the protein was estimated from the specific EPR signal of the Fe(III)-Mn(III) state (not shown) in the enzymatic mixture after incubation with hydroxyurea (32). The results showed that the R2 ox protein contained ϳ0.75 Mn-Fe sites per R2-dimer. A slightly larger value can be calculated from the TXRF data (Table 1). We therefore concluded that almost all (close to 90%) of manganese was incorporated in the Mn-Fe sites. The excess iron in the samples (total ϳ1.6 iron per R2-dimer) was estimated to ϳ0.75 per R2-dimer. At this stage we cannot judge whether the iron not present in the Mn-Fe sites is bound to specific (i.e. dimetal) or unspecific sites.
We found that a lower manganese concentration in the overexpression medium resulted in lower overall manganese concentrations in the protein, whereas higher manganese concentrations led to Mn-Mn clusters with a broad EPR signal at g ϳ 2.5 (data not shown, see Ref. 50). Supplemental Table S1 shows the characteristic properties of varying preparation procedures. The dominance of the Mn-Fe site and sufficient metal concentrations in the R2 ox and R2 red samples allowed for x-ray absorption experiments both at the iron and manganese K-edges to assess the individual metal site structures.
EPR Characterization of Samples- Fig. 1A shows EPR spectra at X-bands of R2 ox and R2 red , recorded under non-saturating conditions. The signal of R2 ox (at high microwave power) is similar, but not identical to the one previously assigned to the Mn(III)Fe(III) state in the R1R2 samples (32). It reflects antiferromagnetic coupling of the high spin Mn(III) (S ϭ 2) and Fe(III) (S ϭ 5/2) ions to produce a ground-state S ϭ 1/2 system (32). Line splittings reflect hyperfine interactions of the 55 Mn nucleus (nuclear spin of 5/2). Simulation of the spectrum (Fig. 1B) yielded the g values and hyperfine (hf) tensors given in Table 2.
In a previous study (32), the Mn(III)Fe(III) state has been obtained in a mixture containing, besides of R2 ox protein, also R1 protein, reductant, and nucleotides, so that enzymatic turnover occurred until hydroxyurea was added as an inhibitor to stop the reaction and accumulate the inactive Mn(III)Fe(III) state. Thus, this Mn(III)Fe(III) state was generated after several catalytic cycles. Here, the Mn(III)Fe(III) state in the as-isolated oxidized R2 protein was studied, which was never in a complex with R1 or reacted with hydroxyurea. Interestingly, there are subtle differences in the EPR g-and hf-parameters of the respective Mn(III)Fe(III) states (Table 2). In addition, weak features on the high field side of the EPR spectrum ( Fig. 1B) seem to indicate contributions from a second Mn(III)Fe(III) species. We did not attempt to simulate these spectral contributions because of their comparably small magnitude. Evidence for a second Mn(III)Fe(III) species, however, also was obtained from XAS (see below).
R2 red samples showed an EPR spectrum of manganese (at low microwave power) with a different shape of its six main lines (Fig. 1, A and C). Such a spectrum was expected for a Mn(II) site with significant zero-field splitting D (Table 2, legend), indicat-  FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7

Mn-Fe Site Structure of C. trachomatis RNR
ing binding of the Mn(II) to the protein.
Remarkably, there is no indication in the spectrum for a coupling between Mn(II) (S ϭ 5/2) and a high spin Fe(II) (S ϭ 2). However, the clearly visible "forbidden" transitions in the EPR spectrum of Mn(II) (Fig. 1C) show that the Mn(II) is high spin. In line with this observation, the spectrum readily was simulated using an isotropic g-value, an isotropic 55 Mn hf-value, and most importantly, a D-value (including D-strain) typical for high spin (hs) S ϭ 5/2 Mn(II) (  Table 3). The fraction of EPR-silent R2 protein was obtained as the difference between total Mn-Fe clusters (Table 1) and those observed by EPR. For R2 ox samples, the fraction of EPRvisible species is 0.69, comprising ϳ75% Mn(III)Fe(III) and ϳ25% Mn(II)Fe(II). For R2 red , the fraction of EPR-visible species is smaller, 0.48, and comprised ϳ85% Mn(II)Fe(II) and ϳ15% Mn(III)Fe(III). The EPR-invisible states are suggested to be the high potential Mn(IV)Fe(III) in R2 ox and the low potential Mn(III)Fe(II) in R2 red .
Electronic Features and Nuclear Geometry of the Manganese and Iron Sites from XANES-The XANES region of XAS spectra is sensitive to the number of metal ligands, the site geometry, and to metal oxidation and spin state. Fig. 2 shows XANES spectra of R2 red and R2 ox collected at the iron (top) and manganese (bottom) K-edges.
Iron K-edge-The iron edge energy in R2 ox (7123.6 eV) and its downshift by ϳ1.5 eV in R2 red (to 7122.1 eV), compared with iron compounds with known oxidation state (51-53), indicate Fe(III) in 80 -100% of R2 ox and Fe(II) in 70 -90% of R2 red , in line with the EPR data. Thus, the reductive treatment in the presence of R1 induced a near-quantitative single-electron iron reduction in R2.
The pre-edge feature of the XANES (Fig. 2, insets) reflects formally dipole-forbidden 1s33d electronic transitions, which may gain intensity, e.g. by 3d4p orbital mixing. The small iron pre-edge amplitude in R2 red and R2 ox suggested a relatively centro-symmetric coordination of Fe(II) and Fe(III) (54). Fig. 3 shows the isolated iron pre-edges. In R2 red , curve-fitting FIGURE 1. EPR spectra at X-band. A, spectra mainly of Mn(III)Fe(III) in R2 ox (upper trace) and Mn(II)Fe(II) in R2 red (lower trace). The temperature was 20 K, and the microwave power was 12.5 milliwatts for R2 ox and 50 microwatts for R2 red . B, spectrum due to the Mn(III)Fe(III) state in R2 ox (after subtraction of a small Mn(II) contribution) and its simulation. C, spectrum of Mn(II)Fe(II) in R2 red (after subtraction of a small Mn(III)Fe(III) contribution) and its simulation. Simulation parameters were derived using the program Easy-Spin (82) and are given in Table 2. Fig. 1 For Mn(II)Fe(II) simulations, additionally a D-value (79), ͉D͉ ϭ 11.8 milliteslas (mT) with 10.7 mT strain, was used. g-tensor; g x , g y , g z (؎0.002) 55 (Fig. 2, bottom) is well compatible with Mn(III), compared with model compounds (Fig. 2, bottom, upper inset) (39,55). However, the dip in the edge rise (arrow) points to a significant Mn(IV) fraction in R2 ox (Ref. 56 and see below). The edge energy of 6548.0 eV in R2 red suggests a mean oxidation state of ϩ2.5 (39), due to the presence of ϳ50% of Mn(II) and Mn(III). Thus, R2 red mainly contained Mn(II)Fe(II) and Mn(III)Fe(II) sites, in agreement with the EPR data. The small pre-edge magnitudes (Fig. 2) are explained by Mn(II/III) ions in near-octahedral geometries (39,57,58), but the flat edge maxima suggest marked Mn-ligand distance heterogeneity.

TABLE 2 The g-and 55 Mn hf-tensor parameters for Mn(III)Fe(III) in R2 ox and Mn(II)Fe(II) in R2 red obtained from simulations of the EPR spectra in
The Mn(IV)Fe(III) state is EPR-silent due to its total integer spin of 1 (interaction of high spin Fe(III) with S ϭ 5/2 and Mn(IV) with S ϭ 3/2) (27,32). We searched for Mn(IV) by comparison of the pre-edge features (Fig. 4A) and XANES (Fig.  4B) of R2 ox and comparable Mn(II, III, IV) compounds (58,59). Indeed, the pre-edge spectrum of R2 ox was well reproduced by summation of the Mn(II, III, IV) spectra with relative weight-   ings of 0.14, 0.62, and 0.24 (Fig. 4A). A similar Mn(IV) contribution also clearly was discernable in the first derivative of the R2 ox manganese edge spectrum (Fig. 4B). Thus, the XANES suggests the presence of ϳ25% Mn(IV) in R2 ox . In R2 red , Mn(IV) was negligible, and this spectrum was reproduced using 50% each of Mn(II) and Mn(III) (Fig. 4A).
The EXAFS analysis (below) revealed short Mn-O distances in R2 ox due to Mn(O)Fe or even MnϭO motifs, which may cause an increased pre-edge amplitude (60,61). Indeed, it was increased in R2 ox (Fig. 2). Ab initio XANES calculations using a full multiple scattering, self-consistent field approach (47) confirmed that this increase was compatible with both the exchange of a terminal Mn-O against an Mn-O or MnϭO motif (not shown). Thus, the pre-edge data suggest surplus Mn(III, IV)-oxide interactions in R2 ox .
X-ray Photoreduction-Modification of high valent metal sites due to photoreduction is a concern when using intense X-rays (62,63). The rate of Fe(III) reduction was estimated by recording series of XAS scans on the same R2 ox sample spot (not shown) and determination of the edge energies of the resulting spectra. Surprisingly rapid Fe(III) reduction was observed (Fig. 5), suggesting that the Fe(II) level was reached after ϳ10 h of x-ray exposure. On this basis and using an x-ray flux of ϳ10 9 photons s Ϫ1 mm Ϫ2 , a dose for the Fe(III)3 Fe(II) transition of ϳ10 13 photons mm Ϫ2 (at 20 K) is calculated. The x-ray dose applied in crystallography (46,62) typically is at least an order of magnitude higher and thus may result in significant iron reduction. Reduction of high valent manganese may be similarly rapid and will be assessed in a forthcoming study (see Ref. 48). At least during the first two XAS scans, there was no evidence for a change in the manganese oxidation state due to the x-ray exposure in R2 ox and R2 red samples (supplemental Fig. S1). The shown XAS spectra of Fe(Mn) were obtained after Յ2 h (Յ1 h) of x-ray irradiation where photoreduction was negligible.
Interatomic Distances from EXAFS-Simulation (curve fitting) of EXAFS spectra was employed to investigate the atomic structure of the manganese and iron sites (numbers of metal ligands, metal-metal/ligand distances). Fig. 6 shows iron and manganese EXAFS spectra of R2 red and R2 ox (black lines).
Iron EXAFS-Two resolved FT peaks at reduced distances of ϳ1.5 Å and ϳ1.8 Å in R2 red (the reduced distance is the true absorber-backscatterer distance minus ϳ0.4 Å due to a phase shift) indicate at least two significantly different Fe-ligand distances, likely due to longer terminal and shorter metal-bridging Fe-ligand interactions. In R2 ox , both FT peaks were at smaller distances, due to bond-shortening at Fe(III). Additional clear FT peaks were observed in the range of ϳ2.3-3.0 Å (Fig. 6, asterisks), corresponding to Fe-Mn interactions as revealed by the EXAFS simulations (below). These vectors also were  (45,59,74). B, first derivatives of XANES spectra (in the inset) of the manganese references compared with R2 red and R2 ox spectra. The arrows mark the Mn(IV) contribution in the R2 ox spectra. shorter in R2 ox , suggesting changes in the Fe-Mn bridging motif.
Manganese EXAFS-Several FT peaks due to first-sphere ligands at manganese were observed. Their comparably small magnitudes indicate interference of EXAFS oscillations from Mn-ligand vectors with significantly different lengths. In R2 ox , the enhancement of the peaks at Ͻ1.8 Å suggests formation of additional short Mn-ligand bonds in the presence of Mn(III) and Mn(IV). Further FT peaks in the range of 2.3-4 Å exist (Fig.  6, asterisks), reflecting Mn-Fe distances as revealed by the EXAFS simulations (below). The FT peak proportions indicate that shorter Mn-Fe distances prevail in R2 ox , again suggesting changes in the metal coordination and bridging motif.
EXAFS Simulations-The spectra in Fig. 6 were analyzed using (in part) a knowledge-based simulation approach. Best fits in Fig. 6 (colored lines) represent parameters in Table 4. The crystal structure of Ct R2 (1SYY (26, 31)), containing an Fe-Fe site, reveals the coordination of one nitrogen from histidine and of three oxygen atoms from carboxylate groups of glutamates or water molecules to each iron, two oxygens in metal-bridging positions, and thus two six-coordinated iron ions. Near-octahedral iron and manganese ions in R2 red and R2 ox are in agree-ment with the XANES analysis. Accordingly, the simulations were based on a similar coordination as in the crystal, including Fe/Mn binding to two O-shells (the shorter one accounting for O-species) and one N-shell (Table 4), to yield a total coordination number, N O,N , of six. Notably, in all cases N O,N was in the range 5-6 if it was allowed to vary in the curve fitting.
The XANES and EPR data suggested ϳ50% Mn(II)Fe(II) and Mn(III)Fe(II) in R2 red and Mn(III)Fe(III) (dominant) and Mn(IV)Fe(III) (ϳ25%) in R2 ox , plus minor amounts of other species in both cases. Each valence couple could exhibit a different Mn-Fe distance, contributing to the EXAFS. In the EXAFS fits, the Mn-Fe distances of the dominant species (from spectral features above noise level, Fig. 6) were assessed (Table 4).   (Table 4, fits b for manganese). FTs were calculated for k-ranges of 2.5-16 Å Ϫ1 (iron) and 1.6 -12 Å Ϫ1 (manganese). Insets: back transforms of FTs over reduced distances of 0.5-4.5 Å (iron) and 1-3.5 Å (manganese) into the k space. *, FT peaks of metal-metal interactions.

TABLE 4 Simulation parameters of EXAFS spectra
Each simulation component (absorber-backscatterer interaction) is represented by three numbers: N i , coordination number (per metal atom); R i , absorber-backscatterer distance (in Å); and 2 2 i Debye-Waller factor (ϫ10 3 , in Å 2 ). The quality of each simulation is characterized by the error sum R F (80). R F was calculated over reduced distances of 0.5-4.5 Å (manganese spectra) and 1-3.5 Å (iron spectra). For the simulations of the manganese spectra: (a) represents simulations with three metal-metal distances, and (b) represents simulations with four metal-metal distances. Restraints used in the simulations: The sum of the N i -values of the metal-O,N shells was 6; the sum of the N i -values of the Fe-Mn(Fe)/Mn-Fe interactions was 1; 2 2 of the first O-shell, the N-shell, and of the metal-metal interactions were fixed to the given values.  FEBRUARY 13, 2009 • VOLUME 284 • NUMBER 7 4) may be due to Mn-O bonds, but also is compatible with a double bond, i.e. a Mn(IV)ϭO motif (61,64).

Mn-Fe Site Structure of C. trachomatis RNR
Metal-Metal Distances-The EXAFS simulations revealed Mn-Fe distances in a wide range of ϳ2.75-4.15 Å (Table 4). Their relative contributions to the EXAFS were in qualitative agreement with the redox state populations from the EPR and XANES analyses. The two most prominent Mn-Fe distances (ϳ2.9 Å and ϳ3.25 Å) were observed consistently both in the manganese and iron EXAFS data. Particularly long and short Mn-Fe distances were only discernable in the manganese data. This result and the slightly longer metal distances in the iron data may be caused by contributions of Fe-Fe sites; in addition, single occupancy iron sites may diminish the relative contributions of metal-metal distances to the iron EXAFS, rendering minor contributions invisible.
Below we summarize the immediate conclusions on the EXAFS-derived metal-metal distances: (i) The longest Mn-Fe distance of 4.15 Å observed only in R2 red is beyond the upper limit of inter-metal distances of divalent ions in R2 crystal structures (Fig. 7). For this long distance, metal-bridging O-species are not expected. Thus, it presumably reflects the Mn(II)Fe(II) state giving rise to magnetically uncoupled metal ions. (ii) The ϳ3.7 Å Mn-Fe distance in R2 ox is close to those observed in crystal structures containing a bridging-chelating carboxylate or two carboxylate bridges (Fig. 7). (iii) The ϳ3.25 Å Mn-Fe distance is similar to that of Fe-Fe in the structure of wild-type Ct R2 (1SYY). It also is at the lowest limit of distances in any R2 structure (Fig. 7) and compatible with a mono-O(H) bridge between the manganese and iron ions (65,66). (iv) The ϳ2.9 Å and ϳ2.75 Å Mn-Fe distances are shorter than intermetal distances in any R2 crystal structure (Fig. 7). The ϳ2.9 Å was prominent both in the manganese and iron EXAFS of R2 ox . The 2.75-Å distance was only derived from the manganese EXAFS of R2 ox when four Mn-Fe distances were employed in the simulations. However, the fit quality (as judged by the almost two times lower error sum, R F , of this R2 ox simulation) was significantly better than for the fit with three distances. A similar improvement was not observed for R2 red (Table 4, manganese fits a and b; the Mn-Fe distance for R2 red of fit b of 2.64 Å refers to a very minor component, which is not reliably determined and hence not further discussed). Mn-Fe distances Ͻ3 Å are found for di-O(H) metal bridging in other proteins (Fig. 7) and models (46, 66 -69). The ϳ25% contribution of the ϳ2.75-Å distance in R2 ox relates it to Mn(IV).
On the basis of their relative contributions to the EXAFS and taking the above considerations into account, the observed Mn-Fe distances (Table 4) are assignable to oxidation states; in R2 red : ϳ4.15 Å, Mn(II)Fe(II) and ϳ3.25 Å, Mn(III)Fe(II), and in R2 ox : ϳ2.9 Å, Mn(III)Fe(III) and ϳ2.75 Å, Mn(IV)Fe(III). The ϳ3.7-Å distance may belong to Mn(III)Fe(III) with a bridgingchelating carboxylate, i.e. to a second Mn(III)Fe(III) species (evidence for such a species comes from EPR) or to a second Mn(II)Fe(II) state found only in R2 ox .
BVS Calculations-The consistency of the EXAFS-determined metal-ligand bond lengths with the metal oxidation states was verified by BVS calculations (49) using Equation 2. For the iron data the parameters in Table 4 yielded mean iron oxidation states (BVS values) of 2.18 for R2 red (assuming Fe(II) in the calculation) and 3.16 for R2 ox (assuming Fe(III)), in agreement with near-quantitative amounts of Fe(II) in R2 red and Fe(III) in R2 ox . (A match of BVS and experimental oxidation state within Ϯ15% is considered as a reasonable result (70).) From the manganese parameters (fits a), BVS values of 2.62 for R2 red (assuming Mn(II) and neglecting the shortest Mn-O distance due to Mn(III)) and of 3.06 for R2 ox (neglecting the shortest Mn-O distance due to Mn(IV) and assuming Mn(III)) were calculated, also in agreement with the manganese oxidation levels from the XANES analysis. Interestingly, for the metal-ligand distances in the Ct R2 crystal structures (26,31) we calculate mean iron oxidation states of ϳ2.5 (1SSY) and ϳ1.9 (2ANI). The structures were determined from protein initially containing Fe(III). Thus, x-ray photoreduction of Fe(III) may have occurred during diffraction data collection, comparable to the rapid Fe(III) reduction observed in the present XAS study.

DISCUSSION
Mn-Fe Electronic Features-EPR signals were detected of Mn(II)Fe(II) and Mn(III)Fe(III), as previously observed (27,32,71). The Mn(IV)Fe(IV) state attributed to a precursor of Mn(III)Fe(III) in Refs. 71 and 72 was not detected in the present study. This state is not accumulated under our aerobic purification conditions of R2 protein. The g and hyperfine tensors of the Mn(III)Fe(III) state in as-isolated oxidized R2 differ to some extent from that formed in the R1R2 complex during catalytic turnover. Formation of the holoenzyme complex and substrate binding to R1 during catalytic turnover seems to influence the electronic properties of the metal site, as previously proposed (27). The two Mn(III)Fe(III) states may differ in their metalbridging motifs.
XAS suggested high spin (hs) Fe(II) and Fe(III) ions. This result is in agreement with the previously observed antiferromagnetic coupling of Fe(III) hs and Mn(III) hs , yielding a groundstate S ϭ 1/2 system (27,32,48). Magnetic uncoupling of Mn(II) hs from Fe(II) hs and the large metal distance (ϳ4.15 Å) imply the absence of metal-bridging O-species in the Mn(II)Fe(II) state.
Apparently, the Fe(III) is more rapidly reduced than the Mn(III). Thus, the midpoint potential of the Fe(III) seems to be more positive. The velocity of x-ray photoreduction suggests a value close to 1 V (62). Interestingly, Mn(IV) is more rapidly reduced than Fe(III) (27,48). These results point to a Mn(IV) Ͼ Fe(III) Ͼ Mn(III) order of the oxidizing potentials. The Mn(IV) ion hence is the primary target of electron transfer from R1.
Structure of the Mn-Fe Site-Overall, the XAS-derived metal-metal distances and metal-ligand bond lengths and the metal ligation by histidines and glutamates observed in the available crystal structures of Ct R2 (26,31) are compatible with structural motifs of the hetero-metallic site similar to those of homometallic sites in other R2 proteins and related enzymes. Four basic motifs are shown in Fig. 7B. Deviations from these motifs with respect to the presence of certain non-amino acid ligands, overall site geometry, and bonding mode of carboxylic groups were observed in crystals and, e.g. may be due to site heterogeneity resulting from x-ray photoreduction of metal ions.
In the reduced R2 sample studied here, two main oxidation states were present, Mn(II)Fe(II) and Mn(III)Fe(II). In the Mn(II)Fe(II) state, O(H) metal bridges seemed to be absent, in line with Fe(II)Fe(II) crystal structures in other R2 proteins showing only carboxylate bridges (Fig. 7). However, some structures reveal metal ions coordinated only by four ligands (tetra-coordinated ions). The XAS data argues against tetracoordinated Mn(II) and Fe(II) in Ct R2, but suggests penta-to hexa-coordinated ions. Thus, there may be metal-bound water species, which are replaced by O 2 and its cleavage products upon the metal site oxidation.
The largely reduced Mn-Fe distance (ϳ3.25 Å) in the Mn(III)Fe(II) state is compatible with a mono-carboxylato, O(H)-bridging motif (66,67,74) (Fig. 7). For a similar Fe-Fe distance (ϳ3.3 Å) in the structure of wild-type Ct R2 (31), two O atoms in bridging positions were found, but two Fe(III) ions were anticipated. Due to limited resolution of crystal structures the assignment of such O-species to true O(H) bonds often is ambiguous (31,78). The BVS calculations suggest Fe(III)Fe(II) in the structure, i.e. one iron may have become singly reduced during data collection. In future studies x-ray photoreduction should be quantified under crystallography conditions.
In R2 oxidized at ambient O 2 , two main oxidation states were observed, Mn(III)Fe(III) and Mn(IV)Fe(III); only a small fraction was in the Mn(II)Fe(II) state. The metal distance of ϳ3.7 Å is similar to that of ϳ3.5 Å in the structure of the Phe-127 3 Tyr mutant of Ct R2 (31) where a bridging-chelating carboxylate was observed, but also similar to metal(II) 2 sites (Fig. 7). It thus may represent a Mn(III)Fe(III) site similar to the Ct mutant or a bis-carboxylato bridged Mn(II)Fe(II). Further studies are required to clarify this issue.
In the main Mn(III)Fe(III) and Mn(IV)Fe(III) species, metalmetal distances (ϳ2.9 and ϳ2.75 Å) are shorter than those in any R2 crystal structure (Fig. 7). An even shorter Fe-Fe distance of ϳ2.5 Å has been reported for E. coli R2 (65). We did not observe such a short metal distance for Ct R2. According to our results it could rather belong to a metal-N(His) bond. The ϳ2.9-Å distance is similar, e.g. to that of the Fe(III) 2 state in methane monooxygenase (73). Metal distances of ϳ2.65-2.80 Å are typical for mono-carboxylato, di-O(H)-bridged metal pairs (46,66,67,74). Density functional theory studies on the Fe(IV)Fe(III) state yielded metal distances of ϳ2.7 Å (76, 77). A distance of 2.63 Å for a mono-carboxylato, di-O-bridged site, but of 2.82 Å for a mono-carboxylato, (O)(OH)-bridged site was found in (21), similar to the ϳ2.75-Å distance here assigned to Mn(IV)Fe(III). In summary, these results likely imply di-O(H) metal bridging at least in the Mn(IV)Fe(III) state of Ct R2 ("diamond core"). Mn(IV)Fe(III) formation may lead to deprotonation of OH bridges (74,75) or reduction of a peroxide (OOH) in the Mn(III)Fe(III) state.
Recently, an XAS study on the Mn(IV)Fe(III) state reported a longer Mn-Fe distance of ϳ2.9 Å (48), similar to the one here observed for Mn(III)Fe(III). In Ref. 48, the Mn(IV)Fe(III) state was populated after reconstitution of apo-R2 with Mn(II) and Fe(II) ions, whereas in the present study R2 was purified, which already contained the Mn(IV) and Fe(III) ions. A possible explanation for the two Mn-Fe distances is the formation of several different Mn(IV)Fe(III) intermediates in the O 2 activation path (Fig. 8), which differ, e.g. with respect to the protonation of O species. One additional protonated bridge (OH) in Mn(IV)Fe(III) can explain the elongation of the Mn-Fe distance from ϳ2.75 Å (this study) to ϳ2.9 Å (48). The formation of several Mn(IV)Fe(III) intermediates also has been suggested in a recent Mössbauer investigation on Ct R2 (29) and related to redox transitions of amino acids (Trp-51 andTyr-222) and protolytic reactions. These results point to the presence of several configurations of the active Mn(IV)Fe(III) state, which all could be of functional relevance.
Particularly short Mn-O bonds observed in R2 ox are expected for metal-O(H) bonds, but also are compatible with a Mn(IV)ϭO motif, i.e. due to a terminal oxo/peroxo group or a (semi)bridging peroxide. A metal-bridging peroxide has been proposed for the Fe(III) 2  Construction of models of the Mn-Fe site requires knowledge on the manganese and iron binding positions. The metalto-protein stoichiometries suggest that manganese preferentially is bound to a site already containing one iron. The ion ligated by, e.g. His-230, may be bound more strongly, because it is coordinated by four amino acids and hence may be iron. Accordingly, at present we favor manganese binding to, e.g.
Reaction Sequence of O 2 Activation-In this investigation, starting from aerobically prepared Ct R2 protein grown in the presence of manganese and iron (31) and found to adopt two main states (Mn(III)Fe(III) and Mn(IV)Fe(III)), two reduced states of R2 (Mn(III)Fe(II) and Mn(II)(Fe(II)) were produced by the addition of R1, nucleotides, and reductant (DTT). The reverse process, formation of states after the addition of O 2 to R2 apo-protein reconstituted with Mn(II) and Fe(II) ions, also has been studied (27,38,72). Consistently, it was found that the active state, Mn(IV)Fe(III), is formed in R2 in the presence of O 2 only and does not require the input of electrons, e.g. from the R1 subunit. Mn(III)Fe(III) formation was observed after single electron donation to Mn(IV)Fe(III) from R1, exogenous peroxide, or dithionite (27,38,72). Here, the Mn(III)Fe(III) state was dominant in as-isolated R2, in the absence of R1 and reductants. Thus, this Mn(III)Fe(III) state may be a precursor of the Mn(IV)Fe(III) state and the primary product of O 2 binding to Mn(II)Fe(II) (Fig. 8). It thus may carry a bridging peroxide. Whether the second minor Mn(III)Fe(III) state also is an intermediate in O 2 activation is unclear. It may also result from the formation of an inactive species (22,81) or from unspecific reduction of Mn(IV)Fe(III). In any event, the differences of the Mn(III)Fe(III) states produced by various treatments may reflect subtle, but functionally important changes at the metal site.
Apparently, the Mn(II)Fe(II) state of Ct R2 is hard to reach even by prolonged R2 ox reduction. Mn(III)Fe(II) is formed more rapidly. This state represents a novel intermediate in the reduction pathway. Effective Mn/Fe reduction presumably involves electron transfer from R1. It is an option that the Mn(II)Fe(II) state is not involved in catalysis in vivo and instead, Mn(III)Fe(II) is the starting state of O 2 activation, preformed to adopt a bridging peroxide by its shorter Mn-Fe distance.
In Fig. 8, we tentatively summarize the information from our XAS and EPR data, also taking into account previous crystallographic (26,31) and spectroscopic results (27,29,38,48,72). A potential role of the Mn(III)Fe(III) state in oxidation as well as reduction of the metal cluster is suggested. Future investigations will aim at determination of the structural basis for preferential iron or manganese binding, of the resistance of Ct R2 toward reactive oxygen species (26,72), and on the complete sequence of redox intermediates in O 2 activation and reduction of the unusual Mn-Fe site in the tyrosine-less Ct RNR.