A new tyrosyl radical on Phe208 as ligand to the diiron center in Escherichia coli ribonucleotide reductase, mutant R2-Y122H. Combined x-ray diffraction and EPR/ENDOR studies.

The R2 protein subunit of class I ribonucleotide reductase (RNR) belongs to a structurally related family of oxygen bridged diiron proteins. In wild-type R2 of Escherichia coli, reductive cleavage of molecular oxygen by the diferrous iron center generates a radical on a nearby tyrosine residue (Tyr122), which is essential for the enzymatic activity of RNR, converting ribonucleotides into deoxyribonucleotides. In this work, we characterize the mutant E. coli protein R2-Y122H, where the radical site is substituted with a histidine residue. The x-ray structure verifies the mutation. R2-Y122H contains a novel stable paramagnetic center which we name H, and which we have previously proposed to be a diferric iron center with a strongly coupled radical, Fe(III)Fe(III)R.. Here we report a detailed characterization of center H, using 1H/2H -14N/15N- and 57Fe-ENDOR in comparison with the Fe(III)Fe(IV) intermediate X observed in the iron reconstitution reaction of R2. Specific deuterium labeling of phenylalanine residues reveals that the radical results from a phenylalanine. As Phe208 is the only phenylalanine in the ligand sphere of the iron site, and generation of a phenyl radical requires a very high oxidation potential, we propose that in Y122H residue Phe208 is hydroxylated, as observed earlier in another mutant (R2-Y122F/E238A), and further oxidized to a phenoxyl radical, which is coordinated to Fe1. This work demonstrates that small structural changes can redirect the reactivity of the diiron site, leading to oxygenation of a hydrocarbon, as observed in the structurally similar methane monoxygenase, and beyond, to formation of a stable iron-coordinated radical.

N-and 57 Fe-ENDOR in comparison with the Fe III Fe IV intermediate X observed in the iron reconstitution reaction of R2. Specific deuterium labeling of phenylalanine residues reveals that the radical results from a phenylalanine. As Phe 208 is the only phenylalanine in the ligand sphere of the iron site, and generation of a phenyl radical requires a very high oxidation potential, we propose that in Y122H residue Phe 208 is hydroxylated, as observed earlier in another mutant (R2-Y122F/E238A), and further oxidized to a phenoxyl radical, which is coordinated to Fe1. This work demonstrates that small structural changes can redirect the reactivity of the diiron site, leading to oxygenation of a hydrocarbon, as observed in the structurally similar methane monoxygenase, and beyond, to formation of a stable iron-coordinated radical.
Over the past decades a number of structurally related proteins that contain oxygen-bridged dinuclear iron centers have been discovered and characterized (1,2). Among these are the hydroxylase protein component of methane monooxygenase (MMO), 1 called MMOH (3)(4)(5), and the R2 protein component of class I ribonucleotide reductase (RNR) (6 -8). These proteins reveal a strikingly similar coordination arrangement of their diiron centers, but fulfill entirely different biological functions; in MMOH, the diiron center is located directly at the substrate binding catalytic site and is directly responsible for the hydroxylation of methane or other substrates, whereas in R2, the diiron center is more than 35 Å away from the substrate binding active site located in the second protein subunit R1 (9 -11). The catalytic role of the diiron center in R2 is to generate and stabilize a tyrosyl radical (at position Tyr 122 in Escherichia coli R2), and this tyrosyl radical is connected to the active site in R1 via a chain of H-bonded amino acid residues. It has been proposed that upon substrate binding in R1 Tyr 122 transfers its radical character via this chain to a cysteine (Cys 439 in E. coli R1), which in turn starts substrate turnover from ribonucleotide to deoxyribonucleotide (9 -12).
The R2 protein is a homodimer which in its active form contains two -oxo bridged diferric iron centers and a substoichiometric amount of a tyrosyl radical. The active form can be generated in vitro by the so-called reconstitution reaction by adding a 6-fold molar excess of Fe 2ϩ and molecular oxygen to the iron-free apoR2 protein (13). The tyrosyl radical is stabilized by a surrounding cluster of hydrophobic side chains as well as by the diiron center, and survives for a couple of days at room temperature. During the reconstitution reaction, several intermediate states of the diiron center have been proposed (7, 14 -17), some of which have been observed by spectroscopy, such as a high valence Fe III Fe IV state, known as intermediate X. Intermediate X has a net electron spin of S ϭ 1 ⁄2, and exhibits a 1.8 mT broad singlet EPR spectrum at X-band frequency and cryogenic temperatures (14 -17).
To investigate the redox chemistry of R2 and the functional abilities of diiron centers in general, it is of great interest to modify the proteins by site-directed mutagenesis and to trap and characterize paramagnetic intermediates during the reconstitution reaction of the apoprotein with ferrous iron and molecular oxygen. The substitution of the radical site at Tyr 122 in E. coli R2 with other amino acids has been performed for several reasons: (i) A protein-linked radical generated on another aromatic amino acid, e.g. tryptophan, is expected to be similarly well shielded and stabilized by the hydrophobic environment as is the tyrosyl radical. Such radicals and their interactions with the protein may be studied as models for functional amino acid radicals in other enzymes, where radicals might be more difficult to detect. (ii) A radical from another amino acid residue at this site may or may not assume the function of Tyr 122 as a temporary electron and proton acceptor for the active site in the R1 protein in the catalytic cycle. (iii) By altering the normal target for oxidation by the diiron center, the latter may perform a different chemistry, which could help in understanding the different reactions of other structurally related diiron proteins, such as MMO.
Results for several Tyr 122 mutations have been reported previously. In R2-Y122F the precursor intermediate X was observed to have an increased lifetime compared with the wild type, and it was succeeded by transient tryptophan radicals at Trp 111 and Trp 107 near the iron site, which have been analyzed by EPR and ENDOR spectroscopy (18 -20). The center X found in R2-Y122F has identical spectroscopic properties to the X found in wild-type R2 (17). In protein R2 of mouse RNR the mutation Y177W (Tyr 177 is complementary to Tyr 122 in E. coli R2) leads to a tryptophan radical, which has been studied by EPR and ENDOR and assigned to residue Tyr 177 (20,21). The catalytic activity of mouse R2-Y177W, measured by the 5Ј-[ 3 H]CDP assay (22), was found to be quenched, with a residual activity less than 1% under multiturnover conditions. Whether a small residual catalytic activity is present during the short lifetime of Tyr 177 of about 1 min (21) remains an open question.
The mutant R2-Y122H of E. coli RNR described in this work was made in an attempt to generate a histidine radical at the shielded position 122. Histidine residues in proteins play important roles in coordination of metal ions and may also mediate electron and proton transfer. So far there is no evidence for the involvement of histidine radicals in the catalytic reaction of a protein. However, transient histidine radicals have been generated in Cu/Zn superoxide dismutase during a reaction with hydroperoxide and were indirectly detected by EPR using the spin trap technique (23).
The mutant protein E. coli R2-Y122H contains a small amount (ϳ5%) of a novel and rather stable paramagnetic species, center H (24). It exhibits an isotropic EPR signal at Xband, which is similar to the Fe III Fe IV intermediate X. However, based on ENDOR studies of 57 Fe-enriched R2-Y122H, we proposed in an earlier report that center H is a diferric iron species with a strongly coupled, probably coordinated radical, Fe III Fe III R ⅐ (24). Here, we present the crystal structure of R2-Y122H and a more detailed spectroscopic characterization of the paramagnetic center H in comparison with intermediate X from R2-Y122F. The crystal structure shows the majority (ϳ95%) of the protein, and information on the structure of the paramagnetic minority species H can be obtained only from spectroscopy. Detailed EPR and ENDOR studies of R2-Y122H with specific isotope labels on the ligands of the diferric cluster indicate that the stable paramagnetic center H in R2-Y122H is a diferric iron center strongly coupled to a phenoxyl radical derived from the phenylalanine Phe 208 . This finding is supported by comparison of MALDI-TOF mass spectra of trypsindigested R2-Y122H and wild-type R2. Trypsin Gold (Mass Spectrometry Grade) was purchased from Promega (Madison, WI), dithiothreitol from Biomol, Hamburg, Germany, and iodoacetamide from Bio-Rad. The oligonucleotide used for generating the R2 mutant Y122H d(5Ј-TTCATTCTAGATCCCATACTCATATC-3Ј) was synthesized and purified by Scandinavian Gene Synthesis AB, Köping, Sweden. A restriction cleavage site for XbaI (TCTAGA), which does not alter the resulting amino acid sequence was also included in the primer. The histidine codon is indicated by bold face letters and the underlining denotes the changed nucleotides.
Oligonucleotide-directed Mutagenesis-Site-directed mutagenesis of pTB2 was performed with the uracil-DNA method described by Kunkel (25) using the Muta-Gene in vitro mutagenesis kit purchased from Bio-Rad. The new construction, plasmid pTB2-Y122H, was verified with dideoxyribonucleotide sequencing.
Expression and Purification-The cells carrying the R2-Y122H mutation were grown either in Luria-Bertani (LB) medium, or in minimal medium (27). Both media contained carbenicillin 50 mg/ml and kanamycin 50 mg/ml. R2-Y122H was purified using the standard methods as described previously (28). Protein purified by these methods was normally 90% pure. For crystallization the mutant was further purified on FPLC Mono Q 10/10 and Superdex 200 columns. The molar absorption index (⑀ 280 -310 ) for R2, 120,000 M Ϫ1 cm Ϫ1 , was used to determine the protein concentration. The iron content was determined colorimetrically with bathophenanthroline as an iron-chelating agent (29). The cells expressing R2-Y122F were only grown in the minimal medium to produce the iron-free apo form.
Isotope Labeling of R2-Y122H-An acidic 57 Fe stock solution was prepared as described by Sturgeon et al. (16). The iron content in the 57 Fe solution was determined using the bathophenanthroline assay (29) after removing undissolved fragments. Cells containing the mutation R2-Y122H were grown in minimal medium as described by Åberg et al. (27) except that the trace metal solution was omitted, and substituted with either 1 ml of 0.5% 56 Fe III Cl 3 ϫ 6H 2 O for the unlabeled control, or a corresponding amount of the 57 Fe stock solution for the labeled sample. This modified minimal medium was also used for global 15 N labeling and for 13 C, and 2 H isotope labeling of amino acid residues, where the medium was supplemented with sterile filtered solution of the isotope labeled L-amino acid to a final concentration of 160 M. The cells were harvested, and the protein was purified as described above.
Reconstitution of Apoprotein R2 with Ferrous Iron and Oxygen-The reconstitution of the iron site in R2 carried out at room temperature as follows: a solution of 2 mM apoR2 (Y122H or Y122F) in air-saturated 50 mM Tris-HCl, pH 7.6 was mixed in an EPR tube with an anaerobic solution of Mohr's salt (NH 4 ) 2 Fe(SO 4 ) 2 (6ϫ molar excess) dissolved in 50 mM Tris-HCl, pH 7.6. The reaction was stopped by immersing the EPR tube into cold isopentane (Ϫ120°C). To trap the Fe III Fe IV intermediate X in R2-Y122F, a rapid freeze quench apparatus (System 1000 from Update Instruments) was used to mix equal volumes of aerobic apoR2 and anaerobic Fe 2ϩ solutions. The mixture was sprayed into a cold isopentane bath (Ϫ120°C) to quench the reaction at time points from 8 ms to 2 s after mixing. The frozen spray was then tightly packed into an EPR tube.
Hydroxyurea Treatment-In order to suppress the superimposing EPR doublet signal of Y122 ⅐ from chromosomally encoded wild-type protein, the R2-Y122H protein was incubated with 10 mM hydroxyurea at 4°C. After 30 min of incubation time the hydroxyurea was removed by running R2 through a Sephadex G-25 column. Hydroxyurea did not react with center H.
Chemical Reduction-Reduction of R2-Y122H was performed in septum-sealed vials with alternating evacuation and argon flushing to minimize the presence of oxygen. To a solution of 0.2-1.0 mM protein in 50 mM Tris-HCl, pH 7.6, an equimolar concentration of phenazine methosulfate was added as redox mediator as well as a four times excess of either dithionite or ascorbate as reducing agent (30). Samples were extracted by an airtight Hamilton syringe and transferred to a 4-mm outer diameter EPR tube before freezing in liquid nitrogen after incubation at room temperature from 1 to 10 min.
Trypsin Digestion-The trypsin digestion was performed on wildtype R2 and R2-Y122H using the following protocol (31): 10 l of the 0.35 mM protein R2 was added to 100 l of 6 M urea in 0.1 mM Tris-HCl, pH 7.8, to unfold the protein. Then 5 l of 200 mM dithiothreitol was added to reduce the cysteine residues, and the mixture was incubated for 60 min at room temperature. Then 20 l of 200 mM iodoacetamide was added to block all cysteine thiol groups and thereby prevent formation of unwanted disulfide bridges. After 60 min of incubation at room temperature, 20 l of 200 mM dithiothreitol was added, which was allowed to react with any excess of iodoacetamide for another 60 min at room temperature. The solution was then diluted with 775 l of 0.1 mM Tris-HCl, pH 7.8, to reduce the urea concentration to below 0.6 M where trypsin is active, and then 100 l of 0.2 ng/l trypsin was added. The reaction mixture was incubated at 37°C overnight. The reaction was terminated by addition of 5 l of 100% acetic acid, and the samples frozen for later analysis by mass spectrometry.
Continuous Wave (CW) EPR and ENDOR Instrumentation-X-band (9.5 GHz) EPR spectra were recorded on a Bruker ESP 300E spectrometer using a standard rectangular (TE 102 ) EPR cavity (Bruker ER4102ST) equipped with an helium flow cryostat (Oxford, ESR9). CW-ENDOR spectra were measured on a Bruker ESP 300E spectrometer using a self-built ENDOR accessory, which consists of a Rhode & Schwarz RF synthesizer (SMT02), an ENI A200L solid state RF amplifier, and a self-built high-Q TM 110 ENDOR cavity (32). The cavity was adapted to an Oxford helium flow cryostat (ESR 910). The measured g-values were calibrated using the known g-value standard Li:LiF, with g ϭ 2.002293 Ϯ 0.000002 (33). Spin concentrations were determined by comparison of the double integrals of the EPR spectra with that of a 0.200 mM Cu(II)(ClO 4 ) 2 standard.
Pulse ENDOR-The pulse ENDOR spectra were recorded at X-band using a Bruker ESP 380E spectrometer equipped with an Oxford Instruments helium flow cryostat (CF935) and a Bruker ER 4118X-MD5-EN resonator. For all measurements the Davies pulse sequence (34) was used. The lengths of the microwave (MW) pulses ( and /2), the radiofrequency pulse (RF), , and the shot repetition rate are given in the figure captions for each experiment.
Simulation of EPR Powder Spectra-The EPR powder spectra have been analyzed using a program for simulation and fitting of EPR spectra with anisotropic g and hyperfine tensors based on the work of Rieger (35), which is described in Ref. 19 and references therein. Thereby, resonant field positions were calculated to second order for arbitrary orientations of the g and hyperfine tensor principal axes.
Trypsin-digested R2 proteins were analyzed using a MALDI-TOF mass spectrometer (Voyager-RP DE) in the linear positive ion mode. The total acceleration voltage was 25 kV. The voltage on the first grid and the delay time between ion production and extraction were adapted to the mass of the different samples. For each spectrum 100 single scans were accumulated. The matrix, ␣-cyano-4-hydroxycinnamic acid was dissolved at a concentration of 15 mg/ml in a 1:1 (v/v) mixture of acetonitrile and 0.1% (v/v) aqueous trifluoroacetic acid. An aliquot of 1 l of sample (3.5 pmol) and 1 l of matrix solution were mixed on the sample plate and air-dried. All data were calibrated by using an external calibration standard mixture (Applied Biosystems, Foster City, CA).
Crystallization and Structure Determination of R2-Y122H-The mutant R2-Y122H was crystallized as described previously for metR2 (36). Crystals belonged to space group P2 1 2 1 2 1 with unit cell dimensions a ϭ 74.1 Å, b ϭ 85.0 Å, c ϭ 114.8 Å. A single crystal was soaked in a cryoprotectant solution consisting of crystal mother liquor containing an additional 20% glycerol for 5 min then flash-cooled directly in liquid N 2 . Data to 1.9-Å resolution were collected at the Swiss-Norwegian beamline of the ESRF synchrotron, Grenoble, France. The data were integrated and scaled using XDS (37). Further data reduction and processing utilized programs from the CCP4 package (38). Details of the data quality are presented in Table I. The starting model for refinement was the 1.7-Å structure of diferrous R2 (39) with PDB accession number 1XIK. The programs TNT (40) followed by Refmac5 (41) were used for refinement. 5 cycles of rigid body refinement were carried out at 4-Å resolution followed by full refinement of atomic positions and B-factors, first at 2.5 Å then at 1.9 Å. Occupancies for Tyr 122 side chain atoms were set to zero. The density for a histidine side chain was clearly visible in the difference electron density maps. Refinement and model quality statistics are presented in Table I.

RESULTS
Expression of R2-Y122H-The mutant protein R2-Y122H of E. coli RNR behaves like the wild-type protein during the cell growth and purification procedures; we obtained ϳ6-7 mg of pure protein R2 per liter of bacterial culture grown in minimal medium and about 6 times more pure protein when cells were grown in LB medium. The mutation Tyr 3 His at position 122 was verified by dideoxy DNA sequencing (not shown), mass spectrometric analysis (see above) as well as x-ray crystallography (see below). The specific activity of R2-Y122H during substrate turnover in R1 was determined using the 5Ј-[ 3 H]CDP assay at room temperature in the presence of excess protein R1, as described by Thelander et al. (22); wild-type R2 was used as control. A residual activity of less than 10 units/mg (nmol of product ([ 3 H]dCDP) min Ϫ1 mg Ϫ1 ) was detected in our preparations of R2-Y122H. The specific activity of the pure wild-type R2 is 2800 -2900 units/mg. The specific activity in the R2-Y122H preparations is hence less than 0.5%, and probably arises from contamination of chromosomally encoded wild-type R2.
X-ray Structure-The x-ray structure of R2-Y122H at 1.9-Å resolution clearly confirms the mutation Tyr 3 His at position 122 of the R2 structure (see Fig. 1). All ligands of the iron center are well defined by the electron density, and their conformations are comparable to the structure of the wild-type oxidized diferric R2 (metR2) (42,43). There are, however, some noticeable differences. In the R2-Y122H dimer, the Fe-Fe distance is 3.5 Å in monomer A and 3.4 Å in monomer B, whereas in the wild-type metR2 dimer at 1.4-Å resolution, the distances are smaller and more equal, (3.39 and 3.35 Å, PDB ID 1MXR) where I j,k are the k individual observations of each reflection j and ͗I͘ j is the weighted mean.  (43). In wild-type metR2 the carboxyl group of Asp 84 is a monodentate ligand to Fe1 with the free carboxylate oxygen making a H-bond to the water coordinated to Fe2, the Fe1-bound oxygen making a second weak H-bond to Y122-OH (43). In R2-Y122H, D84 is rotated further away from the iron site, leaving one oxygen atom ligated to Fe1 and the free oxygen forming a new short hydrogen bond to N⑀ of His 122 (2.7-Å O-N distance). The position of Asp 84 in diferric R2-Y122H resembles the situation in the diferrous reduced form of wild-type R2 (39) and mutant R2-Y122F (44) (not shown). Another commonly flexible side chain in R2, Glu 238 , is found in the same conformation in both monomers of R2-Y122H as in wild-type metR2, i.e. with one carboxyl oxygen coordinated to Fe2 and the second oxygen hydrogen-bonding to a water molecule that is coordinated to Fe1 (Wat1 in Fig. 1). There is also an additional water molecule located in the enlarged pocket around His 122 in R2-Y122H (Wat4 in Fig. 1) in both monomers. The iron centers in the two crystallographically independent protein chains A and B of the R2-Y122H dimer are almost identical. Differences between chains A and B have been observed in other R2 mutants (39,44) and were explained by a different accessibility for small molecules in the two halves due to crystal packing effects. However the only differences near the diiron center of R2-Y122H are in the conformation of Phe 208 (see Fig. 1), and in the coordination of Fe1. The phenyl ring of Phe 208 is shifted between the two monomers because of a 16°r otation around the C ␣ -C ␤ bond ( 1 torsion angle). The 1 values are Ϫ115°in chain A and Ϫ131°in chain B, compared with Ϫ105°and Ϫ103°in chains A and B of wild-type metR2. Rotation about 1 keeps the plane of Phe 208 approximately at the same distance from Fe1. For instance, in wild-type metR2 the distance between the Phe 208 C 4 (C ) carbon and Fe1 is 4.6 Å in both monomers and it is 4.4 -4.5 Å in R2-Y122H chain A and B, despite the different 1 torsions. Crystallographic B-factors and electron density maps indicate somewhat higher flexibility of Phe 208 in chain A than in chain B. In subunit B of R2-Y122H, elongated electron density could allow the placement of two water molecules, one coordinating Fe2 (Wat2) and the other coordinating Fe1 (Wat3), in addition to Wat1 in wild-type metR2 (see Figs. 1 and 2). However the B-factor for the Fe1 ligand is higher (27 versus 21 Å 2 ), the coordinating distance longer (2.5 versus 2.4 Å), and the distance between the two water molecules is only 2.0 Å. This suggests that the water molecule Wat3 coordinating Fe1 is only partially occupied and probably not present at the same time as the water molecule Wat2 coordinating Fe2. The differences between these two water ligands are even more pronounced in subunit A (B-factor 33 versus 14 Å 2 , coordinating distance 2.7 versus 2.2 Å) indicating lower occupancy for Wat3 in subunit A. The high resolution structure of wild-type metR2 indicates 5-coordinated Fe1 (43). The presence of the extra water ligand to Fe1 may be made possible partly by the shift of Asp 84 toward His 122 described above.
Optical Spectra-The optical absorption spectrum of R2-Y122H obtained from cells grown in LB medium shows two broad bands at 325 and 370 nm, which are characteristic for a -oxo-bridged diferric iron center, and is identical to the spectrum of wild-type R2 after hydroxyurea treatment (i.e. metR2) (45) (compare middle spectra in Fig. 3). The intensity of these bands suggests that R2-Y122H has fully occupied iron sites, i.e. 4 irons per dimer. This was confirmed by quantification of iron from acid-denatured protein using the bathophenanthroline assay (29), which yielded 3.6 -4.2 irons per dimer in different preparations. The characteristic narrow peak at 410 nm resulting from the tyrosyl radical in the active wild-type R2 (upper trace in Fig. 3) is missing in the optical absorption spectra of R2-Y122H. Upon treatment with the strong iron chelator 8-hydroxyquinoline in the presence of imidazole 2 as a mild denaturing agent, and subsequent gel filtration to isolate the protein, the optical bands at 325 and 370 nm disappear (Fig. 3, lower trace), as also observed for wild-type R2 (13). However, traces of up to 0.2 irons/dimer could still be detected in R2-Y122H using the more sensitive bathophenanthroline assay. When R2-Y122H was isolated from cells grown in iron-free minimal medium, we obtained a similar optical spectrum without the bands at 325 and 370 nm (not shown). In consistency with previous studies of wild-type R2 from iron-free growth medium (27); however, a residual amount of up to 0.2 irons/ dimer could also be detected here using the bathophenanthroline assay.
EPR Spectra of the Paramagnetic Center in R2-Y122H-Purified R2-Y122H contains a small amount of a novel and stable paramagnetic species, called center H (24). Quantification of center H showed that it appears in Ϸ3-5% of the R2-Y122H dimers at pH 7.6. In contrast to the tyrosyl radical Y122 ⅐ of wild-type R2, which gives rise to a doublet EPR spectrum (g iso ϭ 2.0047 Ϯ 0.0002) (46) (not shown), center H exhibits a Gaussian-shaped isotropic singlet EPR signal at Xband (g iso ϭ 2.0029 Ϯ 0.0002, linewidth, 2.2 Ϯ 0.1 mT) (Ref. 24, see also Fig. 4C). Center H exhibits a microwave power saturation behavior and temperature dependence that is unusual for a free radical in a protein. The EPR spectra of R2-Y122H recorded at increasing temperatures exhibit increasing EPR linewidths, and above 70 K the spectrum is no longer observable (24).
The overall lineshape of the EPR spectrum of center H at X-band (9.5 GHz) is similar to the EPR spectrum of intermediate X (g iso ϭ 2.000 and linewidth 1.8 mT) (see Fig. 4A), which is a short-lived Fe III Fe IV intermediate that appears in wild-

FIG. 2. Electron density at the two iron centers of the crystallographically independent monomers of the R2-Y122H dimer.
The electron density (yellow) is a SIGMAA-weighted 2Fo-Fc map from REFMAC, contoured at 1.0 . The electron density for Phe 208 is contoured in light blue. The coordinating water molecules, the -oxo bridge and the extra water molecule in the enlarged pocket near His 122 are drawn as red spheres, except for the extra water molecule coordinating Fe1, which is drawn as a blue sphere in monomer A to highlight its presumed lower occupancy and longer coordination distance (see text). The figure was made with Bobscript (82,83) and Raster3D (84). type R2 and mutant R2-Y122F during the oxygen activation reaction (14 -17). However, the spectrum of center H is somewhat broader (2.2 mT) and rather stable. EPR signals were observed from Y122H protein that had been kept for 2 weeks at room temperature (not shown).
The g tensor principal components were obtained from 94 GHz (W-Band) high-field EPR spectra of frozen solutions of R2-Y122H, published in a previous report and are given in Table II (24). The g 1 -and g 2 -values are similar to the respective values reported for the wild-type tyrosyl radical in E. coli (20,47); however, the smallest g-value, g 3 , which is below 2.0, is untypical for an isolated organic radical and resembles the g 3 -value of the Fe III Fe IV intermediate X of 1.994 (from Ref. 16) (see Table II), providing evidence for a metal nature of center H. As observed at X-band EPR, no hyperfine structure was resolved either in the W-band EPR spectra. Therefore ENDOR experiments were required, see below.
Intriguingly, center H was also visible in samples obtained from cells grown in minimal medium, when no iron was added. However, as mentioned above and also reported for the wildtype, these preparations still contain traces of up to 0.2 irons/ dimer as detected in R2-Y122H using the colorimetric bathophenanthroline assay (13,27). Attempts to generate center H in vitro by reconstitution of apoR2-Y122H with iron(II) and oxygen did not lead to a significant increase of the yield of H.
Chemical Stability of Center H-The paramagnetic center H in R2-Y122H exhibits unusual stability. We have treated R2-Y122H with different chemical agents and compared the results with wild-type R2. The X-band EPR spectra recorded at low microwave power (not shown) are partly superimposed by a signal resulting from the tyrosyl radical Tyr 122 , confirming a small contamination of chromosomally encoded wild-type R2, as detected in the activity assay (see above). Therefore we used hydroxyurea, a well-known scavenger of the tyrosyl radical in wild-type R2 (45), to quench selectively the superimposed doublet signal from wild-type Tyr 122 . Hydroxyurea had no effect on the intensity or lineshape of the EPR spectrum of center H. Treatment with hydroxyquinoline/imidazole and subsequent gel filtration removes 95% of the iron coordinated to the protein (see Fig. 3). Interestingly however, the amount of center H, as measured by EPR, was merely reduced to 40% of its original level after this treatment. Addition of 6ϫ molar excess of Fe 2ϩ and molecular oxygen to the iron-depleted R2-Y122H restored the optical absorption bands at 325 and 370 nm. However, the magnitude of the EPR signal of center H was unaffected by this reconstitution.
Chemical reduction of R2-Y122H using dithionite (EЈ 0 ϭ Ϫ460 mV) and phenazine methosulfate (PMS) as mediator (EЈ 0 ϭ Ϫ270 mV) under anaerobic conditions led to a decrease of the EPR signal of center H. After 5 min reaction time 80% of the original EPR spectrum of center H vanished, and a new EPR spectrum exhibiting large g-anisotropy, already resolved at X-Band, appeared. (Fig. 5). The g tensor values obtained from a simulation (Fig. 5, dotted trace) are listed in Table II. Based on these g tensor values we assign the new signal to a mixed valence Fe II Fe III center as observed in the structurally similar enzyme MMOH and in mammalian R2 after chemical reduction (30,8). After removing the PMS and dithionite by gel filtration we were able to regain the signal of center H through treatment with hydrogen peroxide, although with slightly lower spin concentration than before reduction. Treatment of R2-Y122H with sodium ascorbate/PMS, or dithionite alone, also led to the reduction of center H and simultaneous appearance of the mixed valence Fe II Fe III center (not shown). As can be seen in Fig. 5, the integrated intensity of the broad mixed valence Fe II Fe III EPR signal exceeds by far that assigned to center H. This clearly shows that the mixed valence signal is generated from the Fe III Fe III state in the majority of the protein (total spin S ϭ 0) by one electron reduction. In center H, dithionite reduction leads to an EPR silent species, e.g. a Fe III Fe III state with a phenoxylate ligand, or a Fe II Fe III state with a coordinated radical, coupled to an S ϭ 0 ground state. Whatever the nature of this state is, it is at least in part reoxidized to the radical ligated Fe III Fe III by hydrogen peroxide.
X-band EPR of Center H and X Labeled with 57 Fe-In a previous study (24), we observed a significant broadening of the EPR spectra of center H labeled with 57 Fe. Here, we compare  this result with the mutant R2-Y122F, which is known to exhibit a paramagnetic diiron species formally described as an Fe III Fe IV center, called intermediate X (14 -16) and which is in the wild type the precursor of the active Fe III Fe III -radical state. Because the paramagnetic center H in R2-Y122H could not be generated by reconstitution, this protein had to be expressed in E. coli cells grown in minimal media enriched with 57 Fe, whereas for R2-Y122F an 57 Fe II solution was mixed with the purified iron-depleted apoprotein sample, which was then frozen in liquid nitrogen.
The EPR spectra of both 57 Fe-labeled intermediate X (Fig.  4B) and center H (Fig. 4D) exhibit a clear isotope effect of 57 Fe compared with the non-labeled spectra (Fig. 3, A and C, respectively). In R2-Y122F, we observed a doublet splitting of the EPR signal of about 2.6 mT due to the nuclear spin of I ϭ 1 ⁄2 of 57 Fe (Fig. 4B), consistent with earlier studies (16). In contrast, for the EPR spectrum of center H in R2-Y122H, only an increase of the linewidth was observed after incorporation of 57 Fe (Fig. 4D) from 2.2 mT ( 56 Fe) to 4.3 mT ( 57 Fe). A significant broadening of all three g-components of center H was also observed in the 94 GHz EPR spectra (24). 57 Fe-ENDOR of Center H and X-In order to determine the 57 Fe hyperfine tensor components of center H we performed 57 Fe-ENDOR experiments on R2-Y122H (24), and here we compare these results with intermediate X from R2-Y122F. The CW ENDOR spectra at 10 K of the non-labeled samples of R2-Y122F, containing intermediate X (Fig. 6A), and R2-Y122H containing center H (Fig. 6C), are both dominated by strong signals from weakly coupled protons placed symmetrically around the free proton frequency ( 1H ϭ 14.1 MHz at 330 mT) and signals in the low frequency range (1-10 MHz) resulting from nitrogen nuclei ( 14N ϭ 1.02 MHz at 330 mT) (see below). ENDOR lines from hydrogens occur as pairs in the spectra according to the first-order resonance condition in Equation 1, where n represents the Larmor frequency of a nucleus with spin I Ͼ 0 nucleus, and A j the electron-nuclear hyperfine coupling for this nucleus. In frozen solution the hyperfine coupling is anisotropic, defined by three tensor components, A 1 , A 2 , and A 3 (j ϭ 1, 2, 3), which all give rise to spectral features in the ENDOR spectra (49). X in Y122F-The spectra of the samples labeled with 57 Fe exhibit large couplings from 57 Fe ( 57 Fe ϭ 0.45 MHz at 330 mT). The ENDOR spectrum of the labeled intermediate X ( 57 Fe III 57 Fe IV ) in R2-Y122F shows two groups of 57 Fe-ENDOR lines, one group at Ϸ35 MHz, and a second group at frequencies below 20 MHz partly overlapping with the proton ENDOR lines (Fig. 6B). The larger 57 Fe tensor obtained from the group of lines at 35 MHz exhibits only small anisotropy, indicative of high spin Fe III , whereas the group of lines below 20 MHz shows a larger 57 Fe hyperfine anisotropy, indicative of Fe IV . In high spin Fe III , one unpaired electron is found in each of the five d-orbitals, leading to an overall spherical spin density distribution. Therefore only a small 57 Fe hyperfine anisotropy is expected for high spin Fe III compared with Fe IV (16,24). This assignment is consistent with the earlier reported Q-band EN-DOR data (16). The 57 Fe hf tensor components obtained from simulations (Fig. 6B, dotted trace) are given in Table III and are consistent with the 57 Fe-splitting seen in the EPR spectra, Fig. 4B.
H in Y122H-The appearance of two new groups of ENDOR lines around 25 and 35 MHz in the ENDOR spectra of R2-Y122H labeled with 57 Fe (Fig. 6D) compared with the nonlabeled protein (Fig. 6C) clearly shows that center H is also a diiron center. The 57 Fe hf tensor values obtained from ENDOR simulations (dotted line in Fig. 6D) are presented in Table III. As can be seen from comparing Figs. 6B and 5D, the large 57 Fe tensor of center H is comparable to the large 57 Fe tensor of Fe III in intermediate X in R2-Y122F. However the small 57 Fe tensor of center H is far more isotropic, is significantly larger than that of Fe IV in intermediate X and is also different from 57 Fe tensors found for Fe II in Fe II Fe III centers (48), indicating that the small 57 Fe tensor of center H is neither Fe II nor Fe IV (16,24). The similarly small anisotropy of the two 57 Fe tensors of center H suggests that both tensors derive from Fe III , and we therefore describe center H as an antiferromagnetically coupled Fe III Fe III center. However, this diferric center needs to be strongly coupled to a radical, R ⅐ , in order to obtain a total spin of S ϭ 1 ⁄2. Interestingly, such a model was originally suggested for intermediate X (14,50), but later revised (16). 14 N and 15 N ENDOR-For nuclei with a nuclear spin I Ͼ 1 ⁄2, such as 14 N with I ϭ 1, the observed ENDOR resonance frequencies ( ENDOR , see Equation 1) are also influenced by the anisotropic quadrupolar splitting, which may result in an additional splitting, according to the first-order expression (51) in Equation 2, ENDOR ϭ ͉ n Ϯ A j /2 Ϯ 3P j /2͉ (Eq. 2) P j (j ϭ x,y,z) is an element of the traceless quadrupole coupling tensor P, with P z ϭ e 2 qQ/2h, where Q is the nuclear quadrupole moment, q the electric field gradient tensor at the nucleus, e the elementary charge, and h is the Planck constant. The traceless tensor is usually described by the value e 2 qQ/h, and the asymmetry parameter ϭ (P x Ϫ P y )/P z . In the CW ENDOR spectra of R2-Y122F containing interme- diate X (Fig. 6A), ENDOR lines from only one 14 N nucleus are observed. Previous X-band (52) and Q-band (17) ENDOR studies of intermediate X have assigned these lines to the N ␦ of the histidine ligand of the Fe III . From the crystal structure of R2, each of the two irons are known to be coordinated by a separate histidine residue (6) (see Fig. 1), however, the hyperfine interactions from the histidine ligated to Fe IV were believed to be too small to be observed by ENDOR spectroscopy (16,17,52).
In the ENDOR spectrum of R2-Y122H (Fig. 6C), the 14 N ENDOR line pattern is clearly different from those in intermediate X (Fig. 6A). In fact, four resonance positions can easily be observed, which were unambiguously identified as 14 N lines by recording spectra at different microwave frequencies (9.55 and 9.15 GHz) and correspondingly different magnetic fields (data not shown). Due to the larger nuclear g-value of protons (g 1H ϭ 5.5857), and hence larger nuclear Larmor frequency ( n ), the spectral positions from proton ENDOR lines are more susceptible to changes in the magnetic field than those from nitrogens, which have a smaller nuclear g-value (g 14N ϭ 0.4038).
To unravel the hyperfine and quadrupolar interactions of center H, a sample of R2-Y122H was prepared where all nitrogens were uniformly labeled with 15 N, a stable isotope with nuclear spin I ϭ 1 ⁄2, and therefore without quadrupole moment. Thus, the 15 N hyperfine tensor values could be obtained from the CW ENDOR spectra shown in Fig. 7B (dotted traces, simulations). Despite the low signal/noise ratio of the nitrogen ENDOR signals in the 15 N sample, it is evident that the 14 N lines observed in the unlabeled samples are missing in these spectra, and ENDOR lines from two 15 N nuclei (termed N 1 and N 2 ) emerge at different spectral positions. 15 N has a different g n than 14 N (g 15N ϭ Ϫ0.5664) which influences the nuclear Larmor frequency ( 15N ϭ 1.42 MHz at 330 mT) and the hyperfine interaction with the electron, and therefore, the 15  is observed at 9 -10 MHz (Fig. 7B). Both ENDOR lines, ENDOR ϩ and ENDOR Ϫ are observed for the smaller 15 N 2 hyperfine tensor. The best simulation of the 14 N ENDOR spectrum (Fig. 6B) is obtained using the 15 N hyperfine tensor values of two nitrogens, N 1 and N 2 , scaled by the smaller nuclear g-value of 14 N and introducing for both 14 N nuclei weak quadrupolar couplings. The deduced 14 N hf tensor components and quadrupole parameters are presented in Table III.
Because of the weak intensities of the lines at frequencies below 7 MHz in CW ENDOR spectra, we also recorded pulsed Davies 14 N-ENDOR spectra (34,53), see Fig. 7A. This spectrum confirms all four 14 N-ENDOR lines. The dashed line represents a simulation of the nitrogen ENDOR lines in the pulsed ENDOR spectrum of center H using the parameters in Table  III. It is important to note that the small anisotropy of the two 57 Fe hf tensors and the detection of a second 14 N hf tensor, which is missing in the Fe III Fe IV center X, both indicate that center H is made up of two Fe III ions.

Experiments to Identify the Ligand Radical Site of Center H: Selective Isotope Labeling of Iron Site Ligands of R2-Y122H-
Candidates for the location of a ligand radical in center H are the amino acid residues or small molecule ligands in close vicinity to the iron site, including histidine, aspartate, and glutamate residues. In a first attempt, we prepared R2-Y122H proteins labeled with [␥-13 C]aspartate and [␦-13 C]glutamate by adding labeled amino acids to the minimal medium used for overexpressing the E. coli cells. Specific 15 N-labeling of histidine residues is more difficult in a minimal medium, since histidine will also be used for purine biosynthesis (54). The incorporation of labeled amino acids was checked by MALDI mass spectrometry determining the difference between theoretical and measured mass of the whole molecule (for data, see "Experimental Procedures"). None of these samples displayed any change in the EPR or ENDOR spectra of center H, expected from electron spin density at the magnetic isotope 13 C (data not shown). This excludes aspartates and glutamates as the origin of the radical site.
We then replaced the phenylalanine residues in R2-Y122H We introduced a small rhombicity in the A 2 and A 3 values from Fe(IV) for better agreement between simulated and experimental Fe(IV) ENDOR spectra (Fig. 6). Assignment of A 1,2,3 to g-values, g 1,2,3 (Table II), and to Fe(III) and Fe(IV) in analogy to Ref. 16. c Isotropic value (A x ϩ A y ϩ A z )/3. d Assignment to Fe1 and Fe2 could be exchanged. However, the larger 14 N hf-tensor is assigned to the histidine coordinated to the iron with the larger hf-tensor. with fully deuterated phenylalanine D 8 -Phe, since hydroxylation of Phe 208 by high valence diiron-oxygen intermediates has been observed in other mutants (44,55). The incorporation of D 8 -Phe was verified by MALDI-MS. In the pulse ENDOR spectrum of R2-Y122H labeled with D 8 -Phe (Fig. 8B) two new features appeared compared with the non-labeled sample (Fig.  8A, see also expanded inset). These new features are centered around the nuclear frequency of deuterium, D ϭ 2.16 MHz at 330 mT according to the resonance condition (Equation 1). The ENDOR difference spectrum (Fig. 7C, Phe H 8 Ϫ Phe D 8 ) reveals corresponding proton couplings that are missing in the spectrum from samples with deuterated Phe (Fig. 8B). The deuterium lines are split by 0.54 MHz, which corresponds to a splitting of 3.5 MHz for the respective proton ENDOR lines (Fig. 8C). The ratio of 6.5 between these values represents the difference in the magnetic moments of protons and deuterons. The deuterium quadrupole coupling is usually very small (Յ 200kHz) and is here not resolved. These data clearly show that the deuterons of a labeled phenylalanine interact with the paramagnetic center H. Since Phe 208 is the only phenylalanine located within 5 Å of the diiron site (6, 43), we conclude that residue Phe 208 carries significant spin density giving rise to the observed deuteron ENDOR lines.

DISCUSSION
The Nature of Center H EPR Spectra of Center H and Intermediate X-To characterize the novel center H, observed in R2-Y122H, we compared its spectroscopic parameters with those of the short-lived intermediate X, that has been observed both in wild-type R2 and in the mutant R2-Y122F (16,56). Q-band ENDOR experiments of 17 O-and 57 Fe-enriched R2 protein (16,56) have shown that intermediate X is a spin-coupled Fe III -O-Fe IV center with significant spin density on both iron ions and on oxygen ligands. The unusually fast spin relaxation for center H, deduced from the microwave saturation behavior and temperature dependence of the X-band EPR spectra (see "EPR Spectra of the Paramagnetic Center in R2-Y122H" in the "Results" section), as well as the small g 3 component obtained from high-field EPR at W-band (see Table II), provided strong evidence for center H being a metal-centered spin system, similar but not identical to intermediate X. 57 57 Fe hf tensors in center H exhibit relatively small anisotropy, typical for Fe III , indicating an Fe III Fe III center. In order to obtain an S ϭ 1 ⁄2 ground state for center H, a third spin needs to be coupled to the diiron center, resulting in a three spin system of S 1 (Fe III ) ϭ 5/2, S 2 (Fe III ) ϭ 5/2, and a coupled radical S 3 (R ⅐ ) ϭ 1 ⁄2, which may be located on a ligand. The radical spin can couple to one Fe III , either ferromagnetically, leading to an intermediate spin  Table IV, which are derived from a tyrosyl radical using the proper spin projection factor (2/9, see text). The ENDOR lines flanking the proton Larmor frequency of the protons (14.8 MHz), which correspond to the small hf couplings, appear with larger intensity in the simulation than in the experiment because of the fact that Davies ENDOR intensities are inherently weak when the hf couplings are very small (53). Spectra were recorded at 10 K: radiofrequency pulse, 8 s; microwave frequency, 9.7 GHz; magnetic field, 345.8 mT; microwave -pulse length, 296 ns. leads to different "spin projection factors" c for the different spins within the three spin system. These spin projection factors are c 1 ϭ Ϫ5/3, c 2 ϭ 20/9, c 3 ϭ 4/9 for ferromagnetic coupling, when S 23 ϭ 3, and c 1 ϭ 7/3, c 2 ϭ Ϫ14/9, c 3 ϭ 2/9 for antiferromagnetic coupling, when S 23 ϭ 2 (24,50). These factors scale all nuclear hyperfine couplings from the respective iron ions ( 57 Fe III -and ligand 14/15 N-and proton-hyperfine couplings) and those from the radical, compared with the respective uncoupled species. In particular, for the antiferromagnetic case, a rather small value, c 3 ϭ 2/9, and hence small hyperfine couplings, are expected for the radical. When assuming the same intrinsic hyperfine couplings for both 57 Fe III nuclei Fe1 and Fe2, the ratio of the isotropic parts of their 57 Fe hf tensors should reflect their different spin projection factors. This hyperfine ratio (69.6 MHz/48.3 MHz, see Table III) is 1.44, which is closer to the expected ratio of the spin projection factors of 1.50 for the case of antiferromagnetic coupling compared with 1.33 for the case of ferromagnetic coupling.
14/15 N-ENDOR of Center H-In view of the structure of the diiron site in R2-Y122H (Fig. 1), 14/15 N-ENDOR lines may be expected from N ␦ and N ⑀ of the histidines ligating the two irons ions, and perhaps also from nitrogens of the protein backbone (58 -63). For the case of a backbone N-H ligating a quinone, quadrupole couplings (e 2 qQ/h) of 3.2 MHz, ϭ 0.52 (Q A in Rhodopseudomonas (R.) viridis) and 3.05 MHz, ϭ 0.54 (Q A in R. sphaeroides) have been reported (58). These e 2 qQ/h values are much larger than the quadrupole couplings for the ligating nitrogens in center H (see Table III). A similarly high value of e 2 qQ/h ϭ 3.36 MHz for histidine N ⑀ is reported in reference (59) and similar values were reported for heme Fe III ligating imidazole e 2 qQ/h ϭ 2.3 Ϫ 3.2, ϭ 0.1 Ϫ 0.3 (60,61). The magnitudes of the observed nitrogen couplings in center H are, however, in better agreement with those observed for histidine coordinated in N ␦ position to heme Fe III , and also histidine N ␦ ligating a quinone (60,62) which lie in a range of e 2 qQ/h ϭ 1.44 Ϫ 1.65 and ϭ 0.69 Ϫ 0.91. Values of e 2 qQ/h Ϸ 1.4 MHz and Ϸ 0.9 have been reported for the remote N of histidine coordinated to copper in azurin (63). However, the e 2 qQ/h values for 14 N of center H in R2-Y122H (1.6 and 1.3 MHz) are in remarkably good agreement with those reported for the histidine ligands in the structurally very similar diiron complexes of MMOH and semimethemerythrin sulfide (1.4 MHz), for which also comparable 14 N hf tensors were observed (64). This indicates that the resonances in center H indeed derive from two N ␦ of two histidines. It can also be excluded that the couplings arise from the non-coordinating remote nitrogens of the histi-dines, since these are known to exhibit only small hyperfine coupling of about 2 MHz (63, 65). The 14 N hf tensors observed for center H have isotropic parts of 16.9 and 9.9 MHz, and a low hyperfine anisotropy, which is typical for the 14 N nucleus of histidines directly coordinated to a metal. We therefore assign the nitrogen resonances to two coordinating N ␦ of histidines His 118 and His 241 . The position of the mutation, His 122 , is too remote, for the large and rather isotropic nitrogen hyperfine couplings observed in ENDOR.
The ratio of the isotropic part of the 14 N hyperfine tensors (16.9 MHz/9.9 MHz) is 1.7, which is even larger than expected for the case of antiferromagnetic radical-Fe1 coupling (1.5) and is not in agreement with the ratio of 1.33 expected for ferromagnetic coupling (see above). For intermediate X (Fe III Fe IV ), where spin projection factors of 7/3 and Ϫ4/3 with a ratio of 1.75 are expected, only the histidine N ␦ ligated to Fe III , and not the one ligated to Fe IV , could be observed in the ENDOR spectra (see Fig. 6, A and B, and Refs. 16, 17, and 50).
MALDI-TOF Experiments, Hydroxylation of Phe 208 -Evidence for the nature of the ligand radical, R ⅐ , came from pulsed ENDOR data of R2-Y122H labeled with D 8 -Phe, which clearly demonstrates significant spin density on a phenylalanine residue, as expected for a radical on this residue. From the x-ray structure of R2-Y122H, it is clear that Phe 208 is the only phenylalanine residue that could become a direct ligand of the diiron center. In view of the remarkable stability of center H, a phenyl radical at Phe 208 seems however highly unrealistic, since this type of radical is known to be highly reactive and unstable. Interestingly, for another mutant of R2 (Y122F/ E238A), hydroxylation of Phe 208 has been observed (44). Therefore, we assume a similar hydroxylation reaction for R2-Y122H, which leads in ϳ5% of the protein to a further oxidation to a phenoxyl radical at residue 208. In order to prove this assumption, we have trypsin-digested the R2-Y122H protein and investigated the obtained fragments by MALDI-TOF mass spectrometry. The trypsin fragment with Phe 208 ( 208 FYVSFACSFAFAER 221 ) has a theoretical mass of 1645 daltons. This value will be increased by 57 due to the acetamide group at the cysteine (R-S-CH2-CONH2, see "Experimental Procedures") to 1702 daltons. Hydroxylation of Phe 208 in a fraction of the protein will increase the mass of this fragment by 16 to a value of 1718 daltons. In the recorded MALDI-TOF spectra of R2-Y122H fragments we found indeed a peak at 1702 daltons, and a weaker peak at 1718, which had only about 15% of the intensity. The corresponding spectra from R2 wild-type protein showed only the peak at 1702 daltons (Fig. 9). This strongly supports our assumption of hydroxylation of Phe 208 in a fraction of the R2-Y122H protein. The different yield of hydroxylation (ϳ15%) found by MALDI-TOF compared with the yield for center H (ϳ5%) from EPR might result from different sensitivity in the MALDI-TOF spectra of the corresponding two protein fragments, or it might indicate that not all of the hydroxylated Phe 208 becomes a radical.
H/D-ENDOR of Center H-Individual proton hf tensor values could not be directly obtained from the proton ENDOR difference spectrum (Fig. 8C), due to overlapping of signals from several protons, and simulations of the spectra were required, based on hyperfine interactions found in similar systems. On the above assumption of a phenoxyl radical F208-O ⅐ , we used the proton hyperfine couplings from Y122 ⅐ (66) and scaled these down by the spin projection factor, c 3 ϭ 2/9, which we found for the ligand radical from the 57 Fe ENDOR data for an antiferromagnetic Fe III -radical coupling. The fact that none of the observed proton couplings in Fig. 8 are larger than ϳ6 MHz excludes the alternative value, c 3 ϭ 4/9, for the ferromagnetic coupling. This is in line with the observed ratios of 57 Fe hyperfine couplings (Fe1/ a From Hoganson et al. (66). b This work, from simulations of the proton ENDOR spectrum of the radical in the difference spectrum Fig. 8, assumes a dihedral angle of 60°for both ␤ -methylene protons. Fe2) and 14 N hyperfine values (N1/N2), which both favor the antiferromagnetic Fe III -radical coupling.
However, the large ␤-methylene 1 H-coupling observed in Y122 ⅐ remains too large even for the spin projection factor c 3 ϭ 2/9, which indicates that the ␤ protons have a different orientation in the F208-O ⅐ radical of center H. The magnitude of the hf coupling of ␤ protons as a function of the spatial orientation of the side chain is given by the McConnell relation (19,59,66) in Equation 3, where B 2 represents the maximum isotropic hyperfine coupling of the ␤ proton, and is the dihedral angle between the H ␤ -C ␤ bond and the -orbital axis of C 1 , which is perpendicular to the plane of the phenol ring (Fig. 10B). For the phenoxyl (or tyrosyl) radical a value for B 2 of 162 MHz has been reported (66,67). In Fig. 10, the magnitude of the expected isotropic hf couplings for the two ␤ protons, using B 2 ϭ 162 MHz, ϭ 0.38 (66), and taking the spin projection factor 2/9 into account, is plotted for all possible orientations of the tyrosine side chain. The difference of the dihedral angles 1 and 2 for the two ␤ protons is ϳ120°because of the tetrahedral sp 3 orbital of C ␤ . Fig. 10 shows that for 1 ϭ 50 -70°, and 2 ϭ Ϫ70°to Ϫ50°, isotropic hf couplings (A iso ) of the two ␤ protons are predicted between 1.5 and 6 MHz, which is within the range of values deduced from the pulsed ENDOR spectrum of center H (Fig. 8).
The dihedral angles of the side chain ␤ protons of Phe 208 in the crystal structure of R2-Y122H are 85°and Ϫ35°for subunit A and 80°and Ϫ40°for subunit B (Fig. 10B), somewhat outside the range estimated from the experiments, and the distance between the Phe 208 C4 carbon and Fe1 is larger than 4 Å in both subunits A and B. However, the x-ray structure shows the majority oxidized species, which is EPR silent. For those protein molecules carrying center H, which are not seen in the x-ray structure because of the low yield, Phe 208 is expected to move closer to Fe1 when it becomes oxidized and coordinated to Fe1. The dihedral angles of the side chain ␤ protons of residue Tyr 208 in the structure of R2-F208Y, where residue Tyr 208 is further oxygenated to DOPA and coordinated to Fe1 (48°and Ϫ72°for subunit A, and 42°and Ϫ78°for subunit B), agree better with the range of values estimated from the experiments. We thus propose a very similar orientation for the oxidized coordinated phenoxyl radical of center H in R2-Y122H, except for the second oxygen at the meta position of residue 208 in R2-F208Y, which is absent in center H. Furthermore, the oxygen at C4 (para position) of Tyr 208 in R2-F208Y is very close to the extra water Wat3 in R2-Y122H (Fig. 1). This suggests that Phe 208 is hydroxylated at the para position in H, as also suggested by the spectroscopic results (see below). Two other mutants showing hydroxylation of Phe 208 at the meta position and coordination to Fe2 rather than Fe1 have dihedral angles of the side chain ␤ protons of residue 208 outside the range estimated from the ENDOR experiments (Fig. 8A), excluding these structures as models for center H: in subunit A of R2-Y122F/E238A (44) the angles are 105°and Ϫ15°and in both subunits of R2-D84E/W48F the angles are 132°and 12°.
The simulation of the ENDOR spectrum in Fig. 8C based on a para-phenoxyl radical fits reasonably well with the experiment. For a meta-phenoxyl radical large spin densities are expected for C 2 , C 4 , and C 6 . In this case, there would be only a very small hf coupling from the ␤-methylene protons of the side chain due to the low spin density on C 1 . However, there would be an additional large anisotropic hf tensor expected from the ␣ proton on C 6 with estimated values of Ϫ9 MHz, Ϫ7 MHz, and Ϫ2.6 MHz. (Fig. 10C, caption). In the ENDOR difference spectrum ( Fig. 8) we found no evidence for such a large hyperfine splitting of Ϫ9 MHz. However the largest hf tensor value of such an ␣ proton would appear in the ENDOR spectrum as a broad and weak wing of an anisotropic pattern, and we cannot rigorously role out that such a large anisotropic component might be broadened beyond detection. Nevertheless, both the ENDOR spectra and the overlaid structures of R2-Y122H and R2-F208Y (Fig. 1) suggest oxidation of Phe 208 at the para position. A model of this center is presented in Fig. 11.
Remarkably, it is almost impossible to remove the iron ions of center H with chelating agents, in contrast to the majority of the R2-Y122H protein, which does not contain center H and for which the diiron center can be reversibly removed (Fig. 3). The x-ray structure (Fig. 1) of the diferric form, which represents the majority species, shows two terminal water molecules for Fe1 and one for Fe2. We propose that in those protein molecules that form center H there is no second terminal ligand at Fe1. Instead, Phe 208 is oxygenated and oxidized to a phenoxyl radical, which then occupies the place of the second water ligand to Fe1. The additional coordination to Fe208-O ⅐ may explain the unusually strong iron binding of center H. Relatively stable phenoxyl ligand radicals have been reported for Fe III model complexes (68). The terminal water ligands of center H are expected to give rise to large anisotropic hyperfine couplings from their protons, which should be exchangeable against deuterium. In an extensive study by Willems et al. (69) on intermediate X in R2 of E. coli, it was shown that from a terminal water ligand on Fe III a dipolar proton hyperfine tensor with approximate components of Ϫ10 MHz, Ϫ10 MHz, and ϩ20 MHz is expected. Thus, similar large dipolar hf tensors are expected from the protons of the two terminal water molecules in center H. However, such a large anisotropic hf tensor leads to a very broad, weak pattern in the ENDOR spectra, extending over a range of 15 MHz, with signal amplitudes lowered by at least a factor of 10 compared with the other proton ENDOR lines in Fig. 8. This, together with the low yield of center H in R2-Y122H (typically 3-5%) explains why we were not able to detect these proton hyperfine couplings in our ENDOR spectra. However, simulations of the EPR spectrum of H (Fig. 4C)  spectrum of center H is consistent with the presence of large anisotropic proton hyperfine couplings from terminal water ligands, not detected in the ENDOR spectra.
Hydroxylation of Phe 208 in Other R2 Mutants-Hydroxylation of Phe 208 has also been observed in several other R2 mutants. In the double mutant R2-Y122F/E238A residue Phe 208 is hydroxylated at the meta position and becomes a phenolate ligand to Fe2, replacing the normal ligand E238 in wild-type R2 (44). A recent density functional study comparing the centers in R2 and MMOH with focus on the shifting carboxylate E238 (E234 in MMOH) concluded that the carboxylate shifts have very low energy barriers in both cases and may be essential for the oxygen activation (70). In mutant R2-F208Y, residue Tyr 208 is converted to a DOPA (27), where, the para-OH group of DOPA is ligated to both iron Fe1 and Fe2, and the meta-OH group is only ligated to Fe1 (44). Interestingly, in R2-F208Y a stable paramagnetic species called center Z, with an EPR spectrum similar to that of center H, has been generated in an alternative route to the formation of DOPA when high concentrations of ascorbate were present during the reconstitution reaction (71). A ␤-d 2 -Tyr labeled preparation of this protein showed a narrowing of the X-band EPR signal of center Z (71,72). This indicates that centers Z in R2-F208Y and H in R2-Y122H may be very similar, both most probably involving a tyrosyl radical on residue 208, which, however, has to be formed by hydroxylation of Phe 208 in the case of R2-Y122H.
Hydroxylation of Phe 208 at the meta position has also been observed in R2 mutant D84E/W48F (55), where the D84E mutation was introduced to correct the only difference of the iron ligands between R2 and the structurally related protein MMOH, and the W48F mutation served to block the radical transfer chain, which is absent in MMOH. In these mutants hydroxylation was observed at the meta position of Phe 208 . Our data indicate hydroxylation of the para position of Phe 208 for mutant R2-Y122H.
Mechanistic Implications-While in mutant R2-Y177W of mouse RNR, a tryptophan radical was observed at Trp 177 , replacing the normal tyrosyl radical at Tyr 177 (21), we found no evidence for generation of a histidine radical in E. coli R2- FIG. 10. A, angular dependence (Equation 3) of the isotropic part of the hyperfine coupling of two ␤ protons of a para-phenoxyl (i.e. tyrosyl) radical when the phenol ring is rotated around the C ␤ -C 1 axis (B). The hf coupling is calculated from Equation 3 for a radical spin projection factor c 3 ϭ 2/9 (S 2,3 ϭ 2), and a spin density ϭ 0.38 on C 1 (66), see text.
i is the dihedral angle between the C ␤ -H i bond and an axis perpendicular to the phenol ring (B, orientation for F208 in R2-Y122H, see text). Angles limited to the marked range are consistent with the observed ENDOR spectrum in Y122H. This is most probably due to the significantly higher redox potential of histidine (1.4 V) in comparison with tryptophan (0.9 V) (73). However, using strongly oxidizing OH ⅐ radicals, generated in a Fenton reaction, transient histidine OHadduct cation (74) and neutral radicals (75) have been generated and investigated in a liquid aqueous solution of histidine. Recently, histidine radicals were also observed in a reaction of superoxide dismutase with hydroperoxide (23). In R2 mutant Y122H, either the oxidation power of the intermediate X (Fe III Fe IV ) is not sufficient for generating a histidine radical, or the hydroxylation of Phe 208 , and subsequent formation of a coordinated phenoxyl radical as a rather stable product, is energetically more favorable.
In mutant R2-Y122F of E. coli, where the active site residue Phe 122 is also very difficult to oxidize, transient tryptophan radicals were observed at residues Trp 107 and Trp 111 (19 -21), and not hydroxylation of Phe 208 as in case of R2-Y122H. This difference in the reaction pathways must rely on structural differences in the iron coordination. Indeed, the histidine His 122 in R2-Y122H forms a hydrogen bond to aspartate Asp 84 making Asp 84 a monodentate iron ligand, whereas in R2-Y122F, Asp 84 is bidentately ligated to the iron (44). Interestingly, the corresponding aspartate in the met-R2 form of mouse RNR is monodentate (76), similar to the reduced diferrous form of E. coli RNR, indicating that this residue is rather flexible. In MMOH from Methylosinus trichsporium OB3b the complementary Glu 114 forms a monodentate bond to Fe1 (3). This residue is less flexible due to the extended length of the side chain, which probably ensures that the substrate oxidation reaction in MMOH always happens directly at the iron site, and not at a more distantly located amino acid residue, as in R2. Furthermore, in R2-Y122H, as well as in MMOH, there is an extra water ligand to Fe1 at the place that is normally occupied by one of the Asp 84 carboxyl oxygens. This structural difference, may also explain why the rhombic g tensor of the mixed valence species, Fe II Fe III , of the R2-Y122H protein shows greater similarities with the g tensor of the Fe II Fe III center in MMOH than with the g tensor of Fe II Fe III in wild-type R2 (Table II). It could be further speculated that in the diferrous state R2-Y122H has also one terminal water ligand at Fe1 as observed for MMOH (77).
In MMOH, the hydroxylation of the substrate is performed by the Fe IV Fe IV intermediate Q (5,78), which has so far not been observed in R2. In the wild-type R2, intermediate X performs the one-electron oxidation of Y122 (15). This intermediate was originally described as an Fe III Fe III species with a strongly coupled hydroxyl radical ligand bound to Fe1 (50). Later, this model was changed to a Fe III Fe IV species, based on ENDOR and Mössbauer data with a substantial portion of the spin delocalized to the oxygen ligands (16). However, the actual electronic structure of X could be a Fe III Fe IV state with a hydroxyl ligand with some small admixture of a Fe III Fe III -hydroxyl radical character. In the case of R2-Y122H this equilibrium might be shifted to the hydroxyl radical form, which could react with residue Phe 208 forming a stable phenoxyl radical.
Investigation of the mechanism of hydroxylation of Phe 208 and generation of the paramagnetic center H in R2-Y122H using the iron reconstitution reaction was not possible, since center H was already present in all preparations of R2-Y122H, and its yield could not be further increased by this reaction. However, it seems reasonable that it is generated in an alternative route of the oxygen activation reaction leading to oxidation of a hydrophobic amino acid residue in the close vicinity of the iron. Thereby different possible reaction pathways may be considered. Center H could be generated in a reaction with one molecule of oxygen. In order to explain the experimental data, a branching has to be assumed. The main branch leads for the majority to the diamagnetic -oxo-bridged Fe III Fe III center and to formation of an additional terminal water ligand at Fe1, which is seen in the x-ray structure. This reaction requires two extra-electrons (in addition to those from the two irons) and two protons, to form the water molecule. A second, much less frequent branch leads also to formation of a -oxo-bridged Fe III Fe III center, but to oxygenation and oxidation of Phe 208 to a phenoxyl radical, which is coordinated to Fe1. The formation of the diferric -oxo bridged iron center requires two electrons from the irons, whereas for the oxygenation and generation of the phenoxyl radical one electron and one proton has to be removed from Phe 208 and replaced by the second oxygen from the split O 2 molecule.
Alternatively, two rounds of reactions with molecular oxygen could be involved in generation of center H. The first round leads for the main branch as above to the -oxo-bridged Fe III Fe III center and an additional water ligand to Fe1. A second less frequent branch leads to formation of the -oxobridged Fe III Fe III center and to hydroxylation of Phe 208 , as observed in other self-hydroxylating mutants. When in this subensemble of R2-Y122H protein the diiron center is thereafter reduced to the diferrous form, a second round of reaction with molecular oxygen could generate the -oxo-bridged diferric state and the radical on the previously hydroxylated residue Phe 208 , which then coordinates to Fe1. This latter scheme is attractive. It could explain, why center H is not generated in the normal iron reconstitution reaction, since after hydroxylation of Phe 208 a reduction to the diferrous form is required before the second round can start. Such a mechanism would represent a very interesting combination of the function of MMOH, hydroxylation, in the first round with the function of RNR, radical generation, in the second round. However, in contrast to wildtype R2, in R2-Y122H the phenoxyl (or tyrosyl) radical is generated, not at residue 122, but at the hydroxylated residue Phe 208 , where it is much more stable, and less reactive compared with the radical at Tyr 122 , because of its coordination to Fe1.
From an enzymology point of view it is interesting to note that the mutant R2-Y122H does not exhibit any measurable enzymatic activity except that of remaining chromosomally encoded wild-type R2 in the sample. A complete loss of enzymatic activity was also found in mouse R2-Y177W even when a tryptophan free radical was formed (W177 ⅐ ) (21). These observations support the unique role of the tyrosyl radical Y122 ⅐ in protein R2 of E. coli RNR, and Y177 ⅐ in mouse RNR, respectively, for the catalytic activity. Interestingly, in the intracellular bacterial parasite Chlamydia trachomatis, the RNR has been classified as class Ic since there is no tyrosyl residue at the expected position for the tyrosyl radical (79,80); however, the RNR reaction is sensitive to hydroxyurea and an alternative radical mechanism involving the intermediate X has been suggested (80). The present combined EPR/ENDOR and x-ray diffraction study underlines the diversity of redox reactions possible in mutants of the diiron carboxylate protein R2 of E. coli, which is important for the understanding of the functional intermediate states in other native diiron carboxylate enzymes.