Kinetic Evidence That a Radical Transfer Pathway in Protein R2 of Mouse Ribonucleotide Reductase Is Involved in Generation of the Tyrosyl Free Radical*

Class I ribonucleotide reductases consist of two subunits, R1 and R2. The active site is located in R1; active R2 contains a diferric center and a tyrosyl free radical (Tyr⋅), both essential for enzymatic activity. The proposed mechanism for the enzymatic reaction includes the transport of a reducing equivalent, i.e. electron or hydrogen radical, across a 35-Å distance between Tyr⋅ in R2 and the active site in R1, which are connected by a hydrogen-bonded chain of conserved, catalytically essential amino acid residues. Asp266 and Trp103 in mouse R2 are part of this radical transfer pathway. The diferric/Tyr⋅ site in R2 is reconstituted spontaneously by mixing iron-free apoR2 with Fe(II) and O2. The reconstitution reaction requires the delivery of an external reducing equivalent to form the diferric/Tyr⋅ site. Reconstitution kinetics were investigated in mouse apo-wild type R2 and the three mutants D266A, W103Y, and W103F by rapid freeze-quench electron paramagnetic resonance with ≥4 Fe(II)/R2 at various reaction temperatures. The kinetics of Tyr⋅ formation in D266A and W103Y is on average 20 times slower than in wild type R2. More strikingly, Tyr⋅ formation is completely suppressed in W103F. No change in the reconstitution kinetics was found starting from Fe(II)-preloaded proteins, which shows that the mutations do not affect the rate of iron binding. Our results are consistent with a reaction mechanism using Asp266 and Trp103 for delivery of the external reducing equivalent. Further, the results with W103F suggest that an intact hydrogen-bonded chain is crucial for the reaction, indicating that the external reducing equivalent is a H⋅. Finally, the formation of Tyr⋅ is not the slowest step of the reaction as it is in Escherichia coli R2, consistent with a stronger interaction between Tyr⋅ and the iron center in mouse R2. A new electron paramagnetic resonance visible intermediate named mouseX, strikingly similar to species X found inE. coli R2, was detected only in small amounts under certain conditions. We propose that it may be an intermediate in a side reaction leading to a diferric center without forming the neighboring Tyr⋅.

Class I ribonucleotide reductases consist of two subunits, R1 and R2. The active site is located in R1; active R2 contains a diferric center and a tyrosyl free radical (Tyr ⅐ ), both essential for enzymatic activity. The proposed mechanism for the enzymatic reaction includes the transport of a reducing equivalent, i.e. electron or hydrogen radical, across a 35-Å distance between Tyr ⅐ in R2 and the active site in R1, which are connected by a hydrogen-bonded chain of conserved, catalytically essential amino acid residues. Asp 266 and Trp 103 in mouse R2 are part of this radical transfer pathway. The diferric/Tyr ⅐ site in R2 is reconstituted spontaneously by mixing iron-free apoR2 with Fe(II) and O 2 . The reconstitution reaction requires the delivery of an external reducing equivalent to form the diferric/Tyr ⅐ site. Reconstitution kinetics were investigated in mouse apo-wild type R2 and the three mutants D266A, W103Y, and W103F by rapid freeze-quench electron paramagnetic resonance with >4 Fe(II)/R2 at various reaction temperatures. The kinetics of Tyr ⅐ formation in D266A and W103Y is on average 20 times slower than in wild type R2. More strikingly, Tyr ⅐ formation is completely suppressed in W103F. No change in the reconstitution kinetics was found starting from Fe(II)-preloaded proteins, which shows that the mutations do not affect the rate of iron binding. Our results are consistent with a reaction mechanism using Asp 266 and Trp 103 for delivery of the external reducing equivalent. Further, the results with W103F suggest that an intact hydrogen-bonded chain is crucial for the reaction, indicating that the external reducing equivalent is a H ⅐ . Finally, the formation of Tyr ⅐ is not the slowest step of the reaction as it is in Escherichia coli R2, consistent with a stronger interaction between Tyr ⅐ and the iron center in mouse R2. A new electron paramagnetic resonance visible intermediate named mouse X, strikingly similar to species X found in E. coli R2, was detected only in small amounts under certain conditions. We propose that it may be an intermediate in a side reaction leading to a diferric center without forming the neighboring Tyr ⅐ .
The three established classes of ribonucleotide reductases catalyze the de novo synthesis of all four deoxyribonucleotides, the building blocks of DNA (1). The iron-containing class I is the one best characterized (2)(3)(4) and has been found e.g. in mammalian cells, some prokaryotes such as Escherichia coli and Salmonella typhimurium, and is coded for by DNA viruses of the herpes group. The common protein structure of class I ribonucleotide reductase is ␣ 2 ␤ 2 , i.e. it is composed of two homodimeric proteins, R1 and R2. Protein R2 contains a diferric iron center and a stable free radical on a tyrosine residue necessary for enzymatic activity, whereas protein R1 contains the substrate binding site. The sequence homology in both proteins is Ͼ90% between human, mouse, and hamster ribonucleotide reductases, but only about 25% between mouse and E. coli (3). Nevertheless, the general enzymatic properties are almost the same.
The three-dimensional structures of R1 and R2 from E. coli have been determined, and the holoenzyme has been modelbuilt (5)(6)(7). In the model, the radical-forming residue Tyr-122 ⅐ is situated in R2 in a hydrophobic pocket close to the -oxobridged diiron center, about 35 Å distant from the active site in R1. A long range electron transfer pathway consisting of conserved amino acids has been proposed to be essential for enzymatic activity of class I ribonucleotide reductases (2,(5)(6)(7)(8).
Recently, mouse R2 protein was crystallized, and its threedimensional structure was determined (9). Despite the low sequence homology between mouse and E. coli R2 (3), the overall structure and the enzymatically important features are very similar (9). The ligands of the iron center as well as the R2 part of the proposed electron transfer pathway can be superimposed. Whereas the hydrogen bonds of the Fe1 ligand His 173 involved in the proposed electron transfer pathway are the same as in E. coli R2, His 270 , a ligand of Fe2, has fewer hydrogen bonds and is, therefore, less constrained in mouse R2.
The electron transfer pathway was suggested to consist of the conserved amino acids (Fe1) His 173 -Asp 266 -Trp 103 -Tyr 370 in R2 and Tyr 738 -Tyr 737 -Cys 429 in R1 (mouse numbering) forming a hydrogen-bonded chain between the iron center in R2 and the active site in R1. Mutation of D266A and W103(Y/F) in mouse R2 leads to a complete loss of enzymatic activity (10), as is the case for the corresponding mutations in E. coli R2. 1 The mutant mouse R2 proteins have been well characterized (10) and are identical to wild type R2 regarding secondary structure probed by CD spectroscopy (W103F) and the association/dissociation kinetics and binding constant for R1. Additionally, the unchanged Mn 2ϩ binding constants, as a probe for Fe 2ϩ binding, and the UV-visible spectra strongly indicate a stiochiometric iron binding in all mutant R2 proteins, including W103F (10).
Reconstitution of active R2 can easily be achieved by mixing apoR2 with ferrous iron and oxygen (11)(12)(13)(14). Four electrons reduce the molecular oxygen: one is delivered from the tyrosine residue, and two are supplied from the ferrous iron forming the diferric center. An external electron has to come from a third iron under our conditions. In wild type mouse and E. coli R2 (11)(12)(13)(14)(15) as well as in the mouse mutants D266A and W103Y (10), similar yields of about 1 tyrosyl free radical (Tyr ⅐ ) 2 /R2 (Ϯ0.5) were achieved by various similar reconstitution procedures. However, the low amount of about 0.01 Tyr ⅐ /R2 in apoW103F could not be increased by reconstitution with Fe 2ϩ (10).
The kinetics of the reconstitution reaction of E. coli R2 with Ն Fe(II)/R2 has been studied intensively (13)(14)(15). Rapid freezequench (RFQ)-EPR and -Mössbauer studies at 5°C yielded an EPR-and Mössbauer-visible transient, denoted species X, that was kinetically competent to be a direct precursor of Tyr ⅐ (14,15). Starting from aerobic apoR2, the formation rate of species X was 8 s Ϫ1 (14), whereas preloading of the anaerobic R2 with ferrous iron yielded about 60 s Ϫ1 (15), suggesting that iron binding to the aerobic R2 is the second slowest step in R2 reconstitution (Scheme 1). The slowest step in both procedures was found to be the formation of the stable Tyr ⅐ , with an unchanged rate of 1 s Ϫ1 . Mössbauer studies revealed no other iron center intermediates beside species X starting from apoR2 (14), whereas in the studies with preloaded R2, 10 -20% of the iron showed unusual isomer shifts (15), possibly reflecting small amounts of accumulated iron center intermediates appearing before species X (e.g. peroxo or diferryl as proposed in Scheme 1). Our results on the reconstitution kinetics of mouse R2 proteins as well as the earlier findings in E. coli wild type R2 can be summarized in the reaction pathway sketched in Scheme 1.
Here, we have investigated the kinetics of the reconstitution reaction in wild type mouse R2 protein and in three mutants (D266A, W103Y, and W103F) on the R2 part of the proposed electron transfer pathway, thus probing its role in the reconstitution reaction. Our kinetic results show that the external electron is delivered via the proposed electron transfer pathway and give strong evidence for the participation of an intact hydrogen-bonded chain in the reconstitution reaction. To explain our findings, we propose a novel reaction scheme consisting of the delivery of electron/proton pairs, i.e. hydrogen radicals, to the iron site. Theoretical calculations (16) have recently shown that movement of an uncharged H ⅐ in a hydrogenbonded system is energetically favorable compared with electron transfer, which involves charge separation. Thus we will refer to the electron transfer pathway as radical transfer pathway (RTP).

EXPERIMENTAL PROCEDURES
All experiments were done in 50 mM Tris/HCl buffer, pH 7.5, 100 mM KCl with a ratio of ferrous iron to protein R2 dimer of 4 or 6.
Protein Purification-The mouse proteins were overexpressed in logarithmically growing BL21(DE3) pLysS bacteria. They contained pETR2 plasmids encoding wild type mouse R2 protein (11) or the mutant mouse R2 proteins D266A, W103Y and W103F (10). The mu-SCHEME 1. Proposed reaction mechanism for the reconstitution reaction of ribonucleotide reductases. For certain steps, rate constants are known (Refs. 14 and 15 and this work). The binding of the two substrates (Fe 2ϩ and O 2 ) in ribonucleotide reductases is assumed to be random (k 1,1Ј and k 2,2Ј (26). This will be converted to a peroxo species (k 3 ), which in turn forms a diferryl compound R2-[Fe 2 O 2 ] 4ϩ (k 4 , compound Q in methane monooxygenase) reactive enough to perform the proposed H ⅐ transfer along the hydrogen-bonded chain (k 5 , Fig. 7). The source of the external reducing equivalent (a H ⅐ in our model) in step k 5 is proposed to be a (hydrated) Fe 2ϩ (11,13), which is likely to be bound to Trp 103 /Tyr 370 (38). The resulting Fe 3ϩ -Fe 4ϩ intermediate R2-[Fe 2 O 2 H] 4ϩ , species X, eventually abstracts the H ⅐ from the tyrosine (k 6 ). Note that not all ligands to the iron centers are shown in the structures. tant proteins were purified according to Rova et al. (10). Wild type protein was purified as in Mann et al. (11) with slight modifications. Grinding with aluminum oxide could be omitted due to the pLysS strain. Mutant and wild type mouse R2 were mainly obtained as apo protein. If necessary Tyr ⅐ and the iron center in wild type mouse R2 were removed by hydroxyurea/EDTA treatment and subsequent passage through a G-25 column. Fractionation yielded a protein pool with less than 10% N-terminal degraded protein monitored by SDS-polyacrylamide gel electrophoresis, whereas up to 50% N-terminal degraded protein were detected in a second protein pool. Reconstitution kinetics from both protein pools under identical conditions exhibited no significant difference (data not shown). E. coli R2 proteins were prepared as described for wild type in (17). E. coli apoproteins were produced by growing cells in iron-depleted medium (18).
Preparation of the Anaerobic Fe 2ϩ Solution-About 150 ml of buffer was thoroughly degassed in a septum-sealed 250-ml bulb for several h. A second bulb containing 5 to 15 mg of Mohr's salt crystals (NH 4 ) 2 Fe(SO 4 ) 2 was sealed with a septum and evacuated. The bulbs were repeatedly flushed with oxygen-free argon (L'air liquide, Alphagaz, Sorbal 500) and evacuated. The bulbs were left under a slight excess pressure of argon. A 60-ml plastic syringe was made anaerobic by washing twice with argon gas and twice partly filled with degassed buffer and argon gas. Then the appropriate volume of anaerobic buffer to prepare the desired iron concentration was transferred into the bulb containing Mohr's salt.
RFQ-Rapid freeze-quench (19) was performed with a System 1000 apparatus from Update Instruments by mixing 1:1 aerobic protein solution (usually 100 M) and anaerobic Fe 2ϩ solution (400 -600 M). An EPR tube connected with a funnel fixed in a home-built holder was immersed completely into an isopentane bath. The stirred isopentane in the bath was cooled by a steady flow of gaseous N 2 , which in turn was cooled by liquid N 2 . For sample preparation stirring was switched off to avoid movement of the funnel and to allow its proper placement beneath the nozzle. Funnels were made from tips for 5-ml Gilson pipetman by cutting off the small end and widening it.
One syringe of the System 1000 apparatus was washed twice with the anaerobic Fe 2ϩ solution and then filled for experiments. The stability of both anaerobic Fe 2ϩ solutions in the bulb and in the syringe was tested by transferring them to anaerobic cuvets at different resting times and measuring their UV-visible spectra. Comparisons with a spectrum of freshly prepared Fe 2ϩ and oxidized Fe 3ϩ showed that in both solutions the oxidation of Fe 2ϩ after 6 h was less than 10%.
The reaction mixtures were quenched by spraying into isopentane (Ϫ100 to Ϫ120°C), and the crystals were packed to the bottom of an EPR tube with known inner diameter using a packing rod from Teflon. The usually recommended isopentane temperature of Ϫ140°C was difficult to maintain with our equipment. Quenching time determination according to Ballou (19) revealed an increase of quenching time from 8 Ϯ 3 ms at Ϫ140°C to 14 Ϯ 3 ms at Ϫ100°C without change in the accuracy of rate determination. As the determined rates of the EPR visible intermediates did not require a further decrease in quenching time (see Table I), we used for practical reasons the much easier obtained isopentane temperatures between Ϫ100°C and Ϫ120°C. The ram velocity was 1 cm/s because this produced crystals that could be packed easily. The isopentane in the EPR tube was removed, and then the EPR tubes were kept for at least 30 min under high vacuum to get rid of the isopentane trapped between the crystals. During this time the tubes were stored in a separate n-pentane bath cooled to Ϫ120°C. Keeping samples under these conditions did not influence the intensity of any detectable EPR signal. Especially, samples containing species X from E. coli and mouse X were tested for stability of signal intensity.
The reaction time was varied by changing the length of the reaction tube between mixer and spray nozzle. The ram velocity was kept constant because it influences the freezing velocity of the crystals and therefore the dead time of the sample. For every time course, a 15-min sample was prepared by push-push mode RFQ and compared with an infinite value prepared from long-aged samples (e.g. contents of the reaction tubes). This comparison yielded the packing factor of about 0.5 used to calculate concentrations in the RFQ samples. For every protein batch a blank sample containing only apoprotein was measured.
Slow Freeze Quench (SFQ)-Slow freeze-quench samples were obtained by hand-mixing. EPR tubes filled with 80 l of apoprotein and the septum-sealed bulb with the anaerobic Fe 2ϩ solution were thermostated at the reaction temperature. A gas-tight Hamilton syringe with a long needle was washed several times with the anaerobic Fe 2ϩ solution to remove oxygen. Then, 80 l of Fe 2ϩ solution was thermostated inside the syringe at the reaction temperature by immersing the syringe but not the end of the needle into the water bath. After temper-ature equilibration, the iron solution was swiftly injected into the apoprotein solution, and the reaction was stopped by immersing the EPR tube in cold n-pentane (Ϫ120°C). For very short time points like 4 s, the mixing was done quickly above the n-pentane bath. The results confirmed that this was done rapidly enough to avoid changes in temperature.
Reconstitution of Active R2 from Fe(II)-preloaded R2 Protein-To investigate if iron binding is the rate-limiting step in the reconstitution reaction or if it is affected by the mutation, both wild type and W103Y R2 proteins were preincubated with ferrous iron before oxygen addition. An EPR tube fitted with a stopcock was filled with 200 l of protein solution (approximately 23-30 M). Anaerobic protein solution was obtained by equilibration of the EPR tube with argon for 30 min on ice. Then a 3-4-fold molar excess of ferrous iron (a freshly prepared, argonpurged solution of Mohr's salt) was added anaerobically to the protein.
The mixture was then preincubated at 4°C (wild type R2) or 16°C (W103Y R2) for 5 min. Then a 3-fold molar excess of oxygen was added in form of 50 l of oxygen-saturated water equilibrated at the reaction temperature. The reaction was stopped as described above (SFQ).
EPR Measurements-The EPR measurements were made at 9 GHz (X-band) on a Bruker ESP 300 spectrometer equipped with an Oxford Instruments continuous flow cryostat for temperatures below 77 K or a cold finger Dewar for 77 K. Signals were measured at three different microwave powers to ensure that they were not saturated with microwave power. An average spin concentration was calculated by comparison with a 1 mM copper standard (12 mM H 3 ClO 4 , pH 1.8) or a mouse R2 standard with a well known radical content calibrated earlier against the copper standard. Composite spectra containing the signal of the stable tyrosyl radical (Tyr ⅐ ) and the new EPR singlet (mouse X) were evaluated by subtracting fractions of the Tyr ⅐ spectrum of the corresponding 15-min sample using the ESP 300 software. The concentration of mouse X given by the resulting spectra was determined by comparison with the copper standard. The concentration of Tyr ⅐ in these composite spectra was calculated in two ways, from the multiplication factor used and the Tyr ⅐ content of the 15-min samples and by the difference between total radical and mouse X concentration.
Evaluation of Kinetics-The Tyr ⅐ concentrations measured were corrected for the Tyr ⅐ content of the blank. The packing factor was estimated by comparing the Tyr ⅐ concentration in samples frozen directly and those that were produced by RFQ, both aged for long reaction times. The Tyr ⅐ concentrations were divided by the protein concentration corrected for the packing factor. Tyr ⅐ /R2 versus time plots were first evaluated with a first order rate equation including a variable for the dead time, which varied between 13 and 30 ms. For presentation in the figures, the dead time was added to the reaction times calculated from the length of the reaction tube.

Formation of the Stable Tyr ⅐ in Wild Type
Mouse R2-The reconstitution reaction of wild type mouse R2 with 4 or 6 Fe(II)/R2 was investigated at several temperatures between 5 and 32°C by RFQ and SFQ methods. At 5 and 10°C reaction temperatures none of the EPR spectra recorded between 4 and 77 K from RFQ or SFQ samples quenched from 8 ms reaction times up did reveal any EPR visible species beside Tyr ⅐ . As shown in the inset of Fig. 1, all spectra recorded have about the same shape as the final spectrum after 15 min, showing only the stable Tyr ⅐ . Fig. 1 (filled symbols) shows the time course of the reconstitution reaction of apo wild type mouse R2 at 5°C with 4 Fe(II)/R2. The curve obtained by RFQ samples could be fitted to the first order rate equation (1), yielding a formation rate k form ϭ 0.14 s Ϫ1 (Table I, Equation 1).  (Fig. 1, filled triangles). Scheme 1 will yield a very complex kinetic pattern for formation and decay of its intermediates if the rates are such that all intermediates can accumulate to reasonable amounts (Ն0.1 [Tyr ⅐ ] ϱ ). If one or more intermediates do not accumulate, the mathematical description can be simplified.
In the case of mouse R2, the situation seems to be even simpler as we detect no intermediate but only the product Tyr ⅐ . This is the classical pattern of a kinetic reaction including a rate-limiting step, where the time dependence of the product is described by Equation 1. Equation 1 derives from time dependence of C in a consecutive reaction (Reaction 1),   reconstitution reaction of such a mutant protein may be regarded as a hint that this particular hydrogen bond is involved in the delivery of the external electron and may elucidate some aspects of the reaction mechanism. Up to now it was not clear whether W103F forms an unstable Tyr ⅐ or does not at all form Tyr ⅐ upon reconstitution, whereas its light absorption spectra indicates formation of a normal iron center (10). Therefore, we scanned the reconstitution kinetics of W103F completely down to the limit of time resolution by reaction times equally spaced on a logarithmic scale (inset of Fig. 2). Combined RFQ and SFQ experiments between 20 ms and 18 min at 20°C 6Fe(II)/R2 show that W103F does not form any transient Tyr ⅐ under our conditions. An unchanged background amount of about 0.01 Tyr ⅐ /R2 was measured in all samples (10). Slowing Down of the Reconstitution Reaction upon Alteration of the Hydrogen-bonded Chain-The mutant mouse R2 proteins D266A and W103Y have been shown to develop a stable Tyr ⅐ up to a yield of 1 Tyr ⅐ /R2 with an anaerobic reconstitution procedure (10). We measured the kinetics of the reconstitution reaction of these mutant proteins with 6 Fe(II)/R2 at various temperatures. The W103Y and D266A R2 proteins developed the stable tyrosyl radical only to 0.3-0.5 Tyr ⅐ /R2 under our conditions. The time course of the generated stable Tyr ⅐ is shown in Fig. 2 for D266A at 25°C. All time courses for D266A and W103Y at different reaction temperatures exhibit the same kinetic pattern (data not shown) with different formation rates k form (Table I). No EPR visible intermediates could be detected at any of the reaction temperatures investigated, neither in D266A nor in W103Y. The formation kinetics of the stable Tyr ⅐ in these two mutant proteins are still first order without a detectable lag period but on average a factor of 20 slower than in wild type mouse R2 (Table I).
Reconstitution Kinetics of R2 Proteins Preloaded with Ferrous Iron-To elucidate if iron binding is a rate-limiting step in the reconstitution or if the mutations affect the kinetics of iron binding, we performed reconstitution experiments with Fe(II)preloaded R2 proteins. SFQ measurements with wild type R2 at 4°C yielded k form of 0.10 s Ϫ1 (Fig. 1, open squares), which is not significantly different from k form ϭ 0.14 s Ϫ1 obtained without preloading of iron at 5°C (Fig. 1, closed symbols). As in the nonpreloading measurements for wild type R2, no other EPR visible intermediate could be detected. Taken together these findings indicate that Fe(II) binding is not a limiting step in the reconstitution reaction of mouse wild type R2.
Analogous reconstitution experiments with the Fe(II)-preloaded mutant W103Y at 16°C are shown in Fig. 3 (open squares) in comparison with the reconstitution data gained from apo R2 W103Y protein without Fe(II)-preloading at 15°C (filled triangles). The nearly identical kinetics show clearly that the slower formation of Tyr ⅐ in the mutants compared with wild type R2 is not due to a slower iron binding kinetics in the mutant proteins.
Temperature Dependence of the Reconstitution Reactions-The temperature dependence of the Tyr ⅐ formation rates k form of wild type and mutant mouse R2 proteins is shown in Fig. 4. At 5°C the Tyr ⅐ yield in the mutants D266A and W103Y was very low, and therefore the kinetics was not measurable. Above 30°C the mutant apoproteins started to precipitate. However, between 10 and 25°C the reconstitution kinetics of the mutant proteins gave reproducible results and could be evaluated according to the Arrhenius law (Equation 3).
The 1/T dependence of ln(k form ) of wild type mouse R2 is linear between 5 and 32°C and gave similar results for 4 (q) or 6 (OE) Fe(II)/R2. The activation enthalpy ⌬H a of about 140 Ϯ 20 kJ/mol for wild type mouse R2 is only slightly affected by the mutations, leading to 110 Ϯ 20 kJ/mol and 90 Ϯ 20 kJ/mol for D266A and W103Y, respectively.
Appearance of a New EPR Singlet Species, Mouse X-Reconstitution of apo wild type mouse R2 with Fe 2ϩ resulted in the appearance of small amounts of a new EPR active intermediate at reaction temperatures between 18 and 32°C (Fig. 5). Composite EPR spectra were obtained from RFQ samples up to 200 ms of reaction time, e.g. after 61 ms reaction time at 25°C, the second EPR component is clearly detectable (Fig. 5A) compared with the final EPR signal of Tyr ⅐ after 15 min of reaction time (Fig. 5B). After subtraction of a fraction of a pure Tyr ⅐ spectrum, an EPR singlet-like spectrum denoted as mouse X became visible (Fig. 5C). The peak-to-trough line width of mouse X was determined to 18 Ϯ 2 G and was found to be invariant between 4 and 77 K. The line width is apparently identical to the 18 Ϯ 2 G of species X from E. coli R2 Y122F (Fig. 5D) at 8 K (13). Upon reconstitution with 57 Fe and subtraction as above, broadening and hyperfine coupling to the nuclear spin of 57 Fe is observed in both mouse X and species X of E. coli R2 Y122F (13 and data not shown).
The microwave power saturation behavior of mouse X was analyzed according to Hales (20), assuming Gaussian broadening and compared with that of species X from the E. coli R2 , k form ϭ 0.29 s Ϫ1 ) but 24 times slower (Table I) than in mouse wild type R2 under the same conditions (Fig. 6). The inset shows that in W103F no tyrosyl radical above the background is formed under the same conditions as above. A constant contamination of 0.01 Tyr ⅐ /R2 was detected in all samples including the apoprotein.
(data not shown). The P 50 values for mouse X and E. coli X were determined to 34 and 11 W at 4 K, 3.2 and 3.0 mW at 29 K, and 190 and 160 mW at 49 K, respectively. The similar temperature dependence of their power saturation behavior taken together with the spectral similarities indicate that mouse X is also located at the iron center and may have a similar structure as reported for species X, an Fe(III)-Fe(IV) intermediate (21).
Kinetics of Mouse X-The concentration of mouse X after different reaction times was determined after spectra subtraction as described above. In Fig. 6 the time course for mouse X

FIG. 3. Kinetics of Fe(II)-preloaded W103Y mouse R2 at 16°C (Ⅺ) compared with apo-W103Y mouse R2 at 15°C (OE). For the experiments with
Fe(II)-preloaded W103Y mouse R2, error bars were included to visualize the variance in the measurements. Points are averages of at least two experiments beside 12 s (Ⅺ). The time courses are identical in the error range of the measurements, indicating that the 20 times slower formation of Tyr ⅐ (Y ⅐ ) in the mutants is not due to a different iron uptake kinetics. The Tyr ⅐ formation in the wild type R2 at 18°C is included for comparison (q). and the stable Tyr ⅐ in the reconstitution reaction of wild type mouse R2 with 6 Fe(II)/R2 at 25°C is shown. The formation of Tyr ⅐ could still be described by the first order rate Equation 1 yielding k form ϭ 7.1 s Ϫ1 . However, subtracting simulated curves of Equations 1 and 2 from each other using the values A ϱ ϭ 1, k 5 ϭ 7 s Ϫ1 (ϭ k form with respect to k a ), and e.g. k 6 ϭ 10 ⅐ k 5 ϭ 70 s Ϫ1 (ϭ k b in Equation 2) yielded a maximal difference of 0.08 at 35 ms. Such a small difference is not detectable with a point to point kinetic method like RFQ, and 8% is less than the usual error for spin concentration determination in EPR. Thus, a k 6 only 10 times faster than k 5 is already sufficient to not lead to any detectable lag period in the time course of Tyr ⅐ formation. Inserting the parameters above in the time dependence of the intermediate B of a consecutive reaction as shown in Equation 4 allowed an accumulation of B to about 0.08. This is comparable with the amount of mouse X found in the reconstitution of wild type mouse R2 with 6 Fe(II)/R2 at 25°C shown in Fig. 6. to the mouse X data in Fig. 6 yielded k 6 ϭ 112 Ϯ 38 s Ϫ1 , which is much bigger than k 5 , as expected, but the shape of the curve (solid line in the inset of Fig. 6) fits badly to the experimental points.
Assuming that mouse X is only the precursor of a part of Tyr ⅐ allows one to set ([X]/[R2]) ϱ as an additional variable and just keep k 5 fixed to 7 s Ϫ1 . The result of this fit is shown as the long dashed curve in Fig. 6 inset with ([X]/[R2]) ϱ ϭ 0.37 Ϯ 0.33 and k 2 ϭ 36 Ϯ 37 s Ϫ1 . We found that this assumption describes neither the measured data nor does it yield well defined variables. In another trial we fixed ([X]/[R2]) ϱ to 1.1 and k 6 to 7 s Ϫ1 assuming that Tyr ⅐ formation is governed by X decay. The resulting best fitting curve (short dashes in Fig. 6) with k 5 ϭ 0.64 Ϯ 0.35 s Ϫ1 does not give the expected k 5 Ͼ k 6 and lies way outside of the data. The other possible combinations of taking the parameters ([X]/[R2]) ϱ , k 5 , and k 6 as variables and/or constants result in similar bad or worse fits. The data of mouse X found at other temperatures and 4 Fe(II)/R2 (Table I) could similarly not be described by one of the models outlined above.
By including a lag period ⌬t in Equation 4, yielding Equation 5, the measured data for mouse X could be described by manually inserting values for the four parameters ([X]/[R2]) ϱ , k 5 , k 6 , and ⌬t into Equation 5 and fitting the resulting curve to the data by eye (dashed and dotted line in Fig. 6).  Table I), indicating strongly that the amount of mouse X accumulating under this conditions is not competent to be an intermediate of the reaction sequence in Scheme 1 leading to Tyr ⅐ formation. We propose that the visible mouse X is an intermediate in the side reaction sketched in Scheme 2, leading to a diferric center without Tyr ⅐ . Further clarification is expected from ongoing Mossbauer RFQ experiments.

DISCUSSION
In respiratory and photosynthetic processes, unidirectional electron transport occurs between redox centers of different redox potentials (22)(23)(24). Two models for electron transport, one proposing a uniform traveling barrier for electrons unaffected by protein structure (23), the other including guidance of an electron through covalent and hydrogen bonds (24), are commonly discussed for distances up to 20 Å. For the enzymatic reaction in class I ribonucleotide reductases an electron has to be transferred over 35 Å, since it is believed that Tyr ⅐ in R2 abstracts an electron from a cysteine at the active site of R1, and the radical transfer must be reversible (2, 7). These features are clearly different from those usually found in other redox proteins and raise the question what kind of transfer actually takes place during catalysis in ribonucleotide reductases. The presence of a conserved hydrogen-bonded chain between the two sites, Tyr 177 ⅐ in mouse R2 and Cys-429 in mouse R1, the fact that Tyr ⅐ is a neutral radical lacking a H ⅐ , and that the first step of the enzymatic reaction is proposed to be H ⅐ abstraction from the substrate suggests a transport of a proton concomitant with the transfer of an electron, i.e. H ⅐ transfer. Therefore, the hydrogen-bonded chain might serve as a RTP. This is in good agreement with recent theoretical calculations leading to high activation energies for a pure electron transfer from one uncharged amino acid (R1-Cys 429 ) to another (R2-Tyr 177 ⅐ ) due to the charge separation necessary (16).
Reactions at the Iron Center-For the following discussion we assume the reaction pathway for the reconstitution as outlined in Scheme 1 (13)(14)(15)25). Binding of the two substrates Fe 2ϩ (k 1 and k 1Ј ) and O 2 (k 2 and k 2Ј ) is random, since both aerobic and anaerobic reconstitution procedures are possible. A probably very short lived, O 2 -R2-[Fe 2 ] 4ϩ intermediate is proposed in analogy with the early O 2 -[Fe 2 ] 4ϩ complex in the enzymatic cycle of soluble methane monooxygenase (26). The irons are oxidized via a peroxo intermediate (k 3 ) to a diferryl compound (k 4 , Q in methane monooxygenase), which is converted to species X by one external reducing equivalent (k 5 ). The source of the external reducing equivalent was shown to be either Fe 2ϩ , as under our conditions, or other reductants like ascorbic acid (12)(13)(14). The last step is the formation of Tyr ⅐ by abstraction of an H ⅐ (k 6 ) from the tyrosine residue adjacent to the iron center.
In the reconstitution of apo E. coli R2, two of the steps in Scheme 1 (k 1Ј and k 6 ) could be measured. The formation of Tyr ⅐ is the rate-limiting step with k 6 ϭ 1 s Ϫ1 at 5 and 25°C (data not shown). Species X accumulates during the reconstitution reaction of E. coli R2 and is the direct precursor of Tyr ⅐ (13-15). For species X from E. coli we found a formation rate of 7 s Ϫ1 at 25°C (data not shown), whereas 8 s Ϫ1 was reported at 5°C (14). In E. coli R2, Fe(II)-preloading of the protein resulted in a much faster formation of species X, revealing iron binding (k 1,1Ј ) as the rate-determining step for the formation of species X (14,15). Despite the slow formation of Tyr ⅐ in mouse R2 at 5°C and 10°C, no EPR visible precursor did accumulate to detectable amounts (Fig. 1). Therefore, the formation of Tyr ⅐ (k 6 ) is not the rate-limiting step in the reconstitution of mouse R2. Since experiments with Fe(II)-preloading of the mouse R2 protein did not give rise to detectable amounts of the postulated intermediate mouse X, we conclude that iron binding is not rate-limiting for mouse X formation.
Nevertheless, we assume an identical pathway for the reconstitution of mouse as for E. coli R2 regarding the reaction steps and the intermediates formed (Scheme 1). The kinetic differences in reactions between the two proteins might arise from structural differences. In the mouse R2 protein, the channel leading from the iron binding site to the surface of the protein is more open to the solvent compared with the E. coli protein (6,9). This might allow a faster transfer and binding of iron for the mouse protein. Minor amounts of mouse X observed in the present study were kinetically not competent for Tyr ⅐ formation (Scheme 1) and are proposed to be part of a side reaction (Scheme 2). The differences in the ability to accumulate species X might be explained by different environments of Tyr ⅐ . Highfield EPR measurements suggested hydrogen bonding of Tyr ⅐ in mouse R2 (27), whereas no hydrogen bond has been found for E. coli R2 (28,29). Recent electron nuclear double resonance measurements confirm the presence of a D 2 O exchangeable, hydrogen-bonded proton in the vicinity of Tyr ⅐ in mouse R2 (30). The results possibly reflect a shorter distance of the Tyr ⅐ to the iron site in mouse R2. This might facilitate hydrogen abstraction from the tyrosine and speed up the formation of Tyr ⅐ in mouse R2. The hydrogen-bonding partner to Tyr ⅐ was suggested to be a water ligand to Fe1. These results suggest that the hydrogen-bonded chain of the RTP is continued on the radical side of the iron center in mouse R2.
Nature of the External Reduction Equivalent-In contrast to W103Y, the mutant W103F does not even form a transient Tyr ⅐ during reconstitution, although a diferric iron center seems to be formed (10). The only molecular difference between W103Y and W103F is the phenyl hydroxyl group, which is apparently crucial for the reconstitution. This observation cannot be explained by either of the two currently most discussed electron SCHEME 2. Proposed reaction mechanism for the formation of the diferric site without formation of Tyr ⅐ , a side reaction of the reconstitution in vitro. For this reaction, evidenced by the finding of more diferric sites than Tyr ⅐ radicals (14,28), we likely detected mouse X as an intermediate. The difference from Scheme 1 is the source of the second H ⅐ (k 6 ), which here might come from a second external [Fe(H 2 O) 6 ] 2ϩ instead of the tyrosine. transport models (22)(23)(24).
From structural data (6) it is known that Trp 103 is hydrogenbonded to Asp 266 , which in turn is hydrogen-bonded to a ligand of Fe1 (Fig. 7). A tyrosine at position 103 may still preserve a hydrogen bond to the ␦-oxygen of Asp 266 by its phenyl hydroxyl group, but a phenylalanine should certainly interrupt it. From structure modeling 3 it seems possible that a tightly bound water molecule in the position of the carboxylate group of Asp 266 compensates for the interruption of the hydrogenbonded chain in the D266A mutation. The amount of Tyr ⅐ formed in the mutant R2 proteins indicates that radical formation is not an all or none mechanism. Rather it seems that the yield of Tyr ⅐ depends on the quality of the hydrogen-bonded chain: good in wild type, medium in D266A and W103Y, and very bad in W103F.
Based on these results we would like to put forward the model that the external reduction equivalent is not merely an electron but an electron accompanied by a proton, i.e. a hydrogen radical, H ⅐ , that is transported via the hydrogen-bonded chain of the RTP (Fig. 7). This process would require only small shifts of the hydrogen atoms forming the hydrogen-bonded chain sketched in Fig. 7, so that a former hydrogen bond becomes a covalent one and vice versa, in agreement with the recent theoretical model (16). From the slow kinetics of Tyr ⅐ formation in D266A and W103Y, one might expect an accumulation of radical intermediates, e.g. Trp 103 ⅐ in D266A and possibly Tyr 103 ⅐ in W103Y. The formation of a transient radical on Trp 48 in E. coli corresponding to mouse Trp 103 has been proposed (31,32). However, no such EPR visible transients have been detected in the present study.
Effect of Alteration of the Hydrogen-bonded Chain-Trp 103 and Asp 266 have been shown to participate in radical transfer during catalysis (10). Our present result shows that they are also involved in the reconstitution reaction, since the kinetics are on average slowed down by a factor of 20 compared with wild type mouse R2 (Table I). Their activation enthalpies are very similar to that of wild type mouse R2 (Fig. 5). This suggests that the overall reaction mechanism is only little affected by these mutations.
Major structural changes are unlikely as shown earlier (10), although the mutations could possibly induce two kinds of minor structural changes, in the binding of Fe 2ϩ and O 2 or on the delivery pathway for the external reducing equivalent. We consider it very unlikely that the oxidation of ferrous iron by dioxygen (k 3,4 in Scheme 1), involving only the direct environment of the iron-oxygen center, is affected by the mutations. Therefore, they might affect three steps in Scheme 1, k 1(') or k 2(') , or k 5 . To rule out the first possibility we made two kinds of binding experiments under identical conditions, Fe 2ϩ binding of the aerobic protein (k 1' ) and O 2 binding of the Fe(II)-preloaded protein (k 2' ). Both experiments gave the same formation rates for Tyr ⅐ . The results indicate that only the delivery of the external reduction equivalent (k 5 ) can be affected by the mutations on the RTP. The striking feature of k form is that it is affected by mutations on the RTP. Thus, the most likely assignment is k form ϭ k 5 , i.e. the proposed transfer of an external H ⅐ is the slowest step in the reconstitution of mouse R2 proteins.
Consequently, the calculated ⌬H a in Fig. 5 should be the activation enthalpy for the delivery of the proposed external H⅐. It is nearly unchanged in the D266A and W103Y mutants, suggesting that the activated complex is not formed on the hydrogen-bonded chain from residue 103 to the iron ligand His 173 . Rather it seems to be formed at the iron site, probably when the proposed H ⅐ is delivered from the histidine to one of the iron site oxygens. Energy might be necessary to adjust the otherwise flexible iron center of mouse R2 (9) into a rigid structure possibly required for the last step of the proposed hydrogen transfer.
From What Species Is the External Reducing Equivalent Delivered?-It is unlikely that a hydrogen radical is delivered from the (reduced) iron site of another R2 subunit without creation of an EPR visible signal, e.g. the mixed valent state (33,34). It is more likely that some external iron binds to Trp 103 (Fig. 7) or maybe to Tyr 370 and delivers the electron/ proton pair. An almost concomitant hydrogen transfer from a water ligand of an Fe 2ϩ bound to the surface of R2 as well as from the tyrosine to the proposed [Fe 2 O 2 ] 4ϩ intermediate of the iron center is consistent with all findings in the mouse R2 reconstitution reaction so far (Fig. 7).
The  7. Proposed model for the last steps (k 5 and k 6 ) of the reconstitution of mouse R2 proteins under nonlimiting iron conditions. The arrows indicate the shift of a hydrogen radical along the RTP (k 5 ). The highly reactive [Fe 2 O 2 ] 4ϩ species might abstract a H ⅐ from the N of His 173 , which in turn takes a H ⅐ from Asp 266 , which takes a H ⅐ from Trp 103 . Eventually the Trp 103⅐ is quenched by the simultaneous oxidation/deprotonation of hydrated Fe 2ϩ bound to the aromatic system of Trp 103 by cation-interaction (38). This consecutive series of H ⅐ abstractions is supposed to be very fast if not simultaneous in the sense that the abstraction of the initially covalently bound hydrogen is supported by the delivery of the initially hydrogen-bonded hydrogen. Finally, mouse X is thought to abstract a second H ⅐ from Tyr 177 , forming the stable Tyr ⅐ (k 6 ). The latter process is proposed to be faster than the delivery of the external H ⅐ (k 6 Ն k 5 ) so that the paramagnetic iron intermediate mouse X cannot accumulate. bauer investigations on E. coli R2 revealed only little of this intermediate (15), but it may be more favorable for observation in mouse R2 due to its different kinetics.
What Regulates the Functional State of the RTP?-The entire RTP seems to depend on intact hydrogen bonds, since it can be knocked out by mutations similar to W103F on residues in the E. coli R1 protein corresponding to mouse R1 residues Y737F and Y738F (36).
After the formation of Tyr ⅐ (right-hand panel of Fig. 7), the hydrogen-bonded chain is disabled to deliver a second H ⅐ because now the N proton of His 173 is pointing in the opposite direction of the iron center/radical site. As for the substrate reaction, a H ⅐ has to be transported from the active site in R1 to Tyr ⅐ in R2. This conformation (right-hand panel of Fig. 7) of the hydrogen-bonded chain has to flip back into the conformation postulated for the apoR2 (left-hand panel of Fig. 7) upon binding of either R1, effector, and/or substrate. This flip requires a real (reverse to Fig. 7) or virtual (rotation) movement of the hydrogen from one nitrogen to the other in His 173 , likely accompanied and facilitated by small structural changes of the iron center. To bridge the distance between the His 173 ligand and the Tyr ⅐ on the other side of the iron center, the water ligands of the irons might be employed. (Hereby, the iron center might change its function from a Tyr ⅐ protector to a guide for H ⅐ .) In the reconstitution reaction, a H ⅐ transfer is proposed to occur between an easily oxidized Fe 2ϩ at the protein surface and the proposed highly oxidative [Fe 2 O 2 ] 4ϩ intermediate over a distance of 10 Å. The difference in redox potentials between these two sites may still allow the proposed H ⅐ transfer through an altered hydrogen-bonded chain, even if its structure is no longer perfectly adjusted to this reaction. On the other hand, in the enzymatic reaction, the difference in redox potentials between Tyr ⅐ in R2 and the substrate in R1 is presumably much smaller, and the transfer distance is much longer. Thus, any misalignment of the hydrogen-bonded chain will suppress the proposed H ⅐ transfer suggested to occur in the enzymatic reaction.
The present results are consistent with a new type of transfer process in proteins: radical transfer in the form of the shift of a H ⅐ along a hydrogen-bonded chain. Other recent theoretical (16) as well as experimental (37) results support this type of transfer. In all mutants of R2 where the catalytically essential tyrosine is replaced by a phenylalanine, only neutral transient radicals on surrounding amino acids have been detected, which again argues in favor for transport of a neutral electron/proton pair (32,37). The coupling of an electron transport to a simultaneous proton transport allows the protein to guide the way of the radical across long distances inside the protein, minimizing protein damage by the highly reactive radical as well as its loss. This type of transfer may not only be valid for the reconstitution and, highly likely, for the enzymatic reaction in ribonucleotide reductases, but may also occur in other proteins that show transport of reducing equivalents/electrons along a hydrogen-bonded chain.