A New Mechanism-based Radical Intermediate in a Mutant R1 Protein Affecting the Catalytically Essential Glu441 inEscherichia coli Ribonucleotide Reductase*

The invariant active site residue Glu441 in protein R1 of ribonucleotide reductase fromEscherichia coli has been engineered to alanine, aspartic acid, and glutamic acid. Each mutant protein was structurally and enzymatically characterized. Glu441 contributes to substrate binding, and a carboxylate side chain at position 441 is essential for catalysis. The most intriguing results are the suicidal mechanism-based reaction intermediates observed when R1 E441Q is incubated with protein R2 and natural substrates (CDP and GDP). In a consecutive reaction sequence, we observe at least three clearly discernible steps: (i) a rapid decay (k 1 ≥ 1.2 s−1) of the catalytically essential tyrosyl radical of protein R2 concomitant with formation of an early transient radical intermediate species, (ii) a slower decay (k 2 = 0.03 s−1) of the early intermediate concomitant with formation of another intermediate with a triplet EPR signal, and (iii) decay (k 3 = 0.004 s−1) of the latter concomitant with formation of a characteristic substrate degradation product. The characteristics of the triplet EPR signal are compatible with a substrate radical intermediate (most likely localized at the 3′-position of the ribose moiety of the substrate nucleotide) postulated to occur in the wild type reaction mechanism as well.

Ribonucleotide reductase is an essential enzyme of all living cells and catalyzes the reduction of ribonucleotides to the corresponding deoxyribonucleotides. Several classes of ribonucleotide reductases with different subunit composition and cofactor requirements are known, but they all share a radical-based reaction mechanism (1).
The aerobic class Ia ribonucleotide reductase from Escherichia coli is the best characterized enzyme. It consists of two components denoted protein R1 and protein R2, each of which is a homodimer. Protein R1 contains redox-active cysteines essential for catalysis. Cysteines 225, 439, and 462 are located at the active site, where all four physiological substrates (CDP, UDP, GDP, or ADP) can bind. R1 also contains two different allosteric sites that bind nucleoside triphosphate effector molecules. One site regulates the overall enzyme activity, and the other site determines the substrate specificity (2,3). Protein R2 contains a stable tyrosyl free radical at position 122 and an adjacent dinuclear iron center (4 -6). The tyrosyl radical is essential for catalysis.
The separate three-dimensional structures of protein R1 and of protein R2 are known (6 -9). A model-built holoenzyme complex of the R1 and R2 structures indicates that the distance between the active site in R1 and Tyr 122 in R2 is about 30 -40 Å (8). Chains of conserved hydrogen-bonded residues leading from the active site of R1 in the direction of Tyr 122 in R2, and vice versa, have been identified and are believed to be part of a radical transfer pathway between the two sites (1, 6 -9). Mutational analysis of the residues postulated to be involved in radical transfer between R1 and R2 during catalysis supports this hypothesis (4, 10 -14). 1 Similar studies have also been performed in mouse ribonucleotide reductase (15).
Recently, the three-dimensional structure of protein R1 in complex with substrate was determined (16). It shows that Cys 439 is at hydrogen bonding distance to the 3Ј hydrogen of the substrate (Fig. 1A). The redox-active Cys 225 and Cys 462 are on the other side of the ribose moiety, and the thiol of Cys 225 hydrogen-bonds to the 2Ј-hydroxyl group. Conserved residues Asn 437 and Glu 441 form hydrogen bonds to the 2Ј-and 3Јhydroxyl groups of the substrate, respectively. In addition, the side chains of Asn 437 and Glu 441 are connected via a hydrogen bond.
Based on the three-dimensional structure of the R1-substrate complex (16) and numerous biochemical studies on engineered R1 proteins (10,11,17,18), specifically isotopically labeled substrates (19,20) and 2Ј-substituted substrate analogues (21)(22)(23)(24)(25)(26), a detailed reaction mechanism has been proposed (Fig. 2). Initially, a transient protein radical (1) is generated at Cys 439 in R1 by radical transfer to Tyr 122 in R2. The thiyl radical abstracts a hydrogen from the 3Ј-position of the substrate generating an oxidized substrate radical (2). Glu 441 may act as a base to facilitate the leaving of the 2Ј-hydroxyl, giving a 3Ј-keto radical intermediate (3). This intermediate can either be reduced by hydrogen atom transfer from the Cys 225 -Cys 462 redox couple (step a), or as suggested from the recent crystal structure of the R1-substrate complex (16) by electron transfer via the hydrogen-bonded Cys 225 , Asn 437 , Glu 441 and 3Ј-O (steps b1-b2) to give the one-electron reduced intermediate (4). The second electron would be transferred via the same hydrogen-bonded pathway to give the deoxynucleotide radical intermediate (5). The thiyl radical at position 439 is then transiently regenerated by radical transfer to the 3Ј-position (6), and the radical is propagated to Tyr 122 in protein R2. Prior to a new catalytic turnover of protein R1, the redox-active cystine at the active site must be reduced. Reduction of R1 involves yet another redox-active cysteine pair in the C-terminal part of R1 (10,27), which interacts with the physiological reductants thioredoxin or glutaredoxin (28,29).
The role of Glu 441 is to participate in binding of the substrate, and plausibly as a general base in the reaction mechanism (8,16) and as part of an electron transfer pathway during the reduction sequence (16). A recent study with substituted nucleosides as models for ribonucleotide reduction supports the theory that a carboxylate may act as a base in the deoxygenation at the 2Ј-position (30). It was also suggested that the same protonated carboxylate would act as an acid in the reduction sequence of the reaction mechanism (30).
All current evidence for substrate radical intermediates in the reaction mechanism are indirect and inferred from the observations that there is an absolute need for a stable radical and a radical transfer pathway (4, 10 -14), 1 that there is an isotope effect on the 3Ј-hydrogen (19,20), that protein radicals are formed during reaction with suicidal k cat inhibitors (22,24,26), and that the polypeptide chain of an engineered R1 protein is suicidally truncated in a mechanism-based reaction sequence (18). In this study we characterize the reaction between engineered R1 proteins E441A, E441D, and E441Q and normal FIG. 2. Proposed reaction mechanism for reduction of ribonucleotides by ribonucleotide reductase. In 3, two alternative reduction pathways have been indicated, a and b1-b2, respectively.

FIG. 1.
A, structure of substrate binding at the active site of reduced protein R1. The 3Ј-oxygen is hydrogen-bonded to the side chain of Glu 441 , the 2Ј-oxygen is hydrogen-bonded to the side chain of Asn 437 and Cys 225 , and Glu 441 and Asn 437 are hydrogen-bonded. These interactions are indicated by thin lines. The postulated hydrogen-bonded radical transfer pathway is indicated by dashed lines. Adapted from Eriksson et al. (16). B, stereopairs of the structure of the active site environment of the three mutant R1 proteins and wild type R1. The mutant residues at position 441, Asn 437 , Cys 225 , Cys 439 , Cys 462 and Met 620 are indicated. Wild type, green; E441A, blue; E441D, red; E441Q, yellow.
substrates in presence of wild type R2. The reaction of E441D/Q and 2Ј-substrate analogues has been reported previously (25).
The most intriguing result of the current study is the formation of a new transient radical in a substrate-dependent reaction between E441Q and wild type protein R2. It is suggested that the new species is a substrate radical intermediate in the reaction sequence. Our current results also show that the carboxylic functionality at position 441 is absolutely essential as the E441D protein has 6 -10 times lower activity than wild type R1, whereas E441A lacks enzyme activity and the amide of E441Q gives rise to a suicidal enzyme.
Restriction enzymes used were SfuI from Boehringer Mannheim and MluI from Promega.
The 2Ј-azido-2Ј-deoxy-CDP (CzDP) 2 was obtained by cleavage of its CTP derivative by incubation with myosin to complete cleavage. The CzDP was separated from myosin by centrifugation using a Centricon filter from Amicon with a 10,000 M r cut-off, freeze-dried, and dissolved in 50 mM Tris-Cl, pH 7.6. The 2Ј-azido-2Ј-deoxy-CTP from U. S. Biochemical Corp. was purchased from Amersham. The myosin was purchased from Sigma. [ E. coli thioredoxin and thioredoxin reductase were expressed and purified as described by Lunn et al. and Russel et al. (31,32).
E. coli SK3981 used to produce thioredoxin and E. coli A237/pPMR14 used to produce thioredoxin reductase were obtained from A. Holmgren.
Plasmids-Plasmid pTB1 (10) containing the gene coding for protein R1 was used in combination with pGP1-2 (33) for overexpression of the mutant R1 proteins using heat induction of the T7 RNA polymerase system.
Oligonucleotide-directed Mutagenesis-Construction of the site-directed mutations E441A, E441D, and E441Q of pTB1 was done with the uracil-DNA method described by Kunkel et al. (34,35). The Muta-Gene phagemid in vitro mutagenesis kit from Bio-Rad was used.
To verify the absence of secondary mutations, a 532-base pair SfuI/ MluI fragment of the mutants was sequenced and cloned into wild type pTB1 plasmid.
Expression of Mutant R1 Proteins-E. coli MC1009/pGP1-2 containing one of the mutant pTB1 plasmids (E441A, E441D, or E441Q) was grown in five flasks with each 1.5 liters of LB medium (total 7.5 liters of medium) with kanamycin (50 g/ml) and carbenicillin (50 g/ml). The cultures were grown at 30°C and shaken vigorously (260 rpm). When the cultures had grown in logarithmic phase for three generations to an absorbance of A 640 ϭ 0.5-0.7, the temperature was raised to 42°C to induce overproduction of the cloned R1 gene. When the cultures reached stationary phase at A 640 ϭ 1.8 -2.0 after Ϸ4 h of induction, the cells were quickly chilled on ice and harvested by centrifugation. Pellets were frozen on dry ice and stored at Ϫ80°C.
Protein Purification-Frozen cells were disintegrated in a BIOX X-press and resuspended in extraction buffer containing 50 mM Tris-Cl, pH 7.6, 10 mM MgCl 2 , 20% glycerol, 2 mM DTT, and 10 M phenylmethylsulfonyl fluoride. Purification was done as described by Sjöberg   The pH dependence of enzyme activity was measured for E441D and wild type R1 in parallel at pH values from 6.1 to 8.6. Buffers used were PIPES (pH 6.1, 6.6, and 7.1), HEPES (pH 7.1, 7.6, and 8.1), and Tris-Cl (pH 8.1 and 8.6). Reaction conditions were as above but with 33 mM of the specific buffer and with 2 M R2. The same concentrations of R1 E441D and wild type R1 were used and activity of wild type was measured during 3 min and E441D during 5 or 10 min.
The activities of E441D and wild type R1 were also measured by spectrophotometrically determining NADPH oxidation at 340 nm (38) with GDP or CDP as substrate and dTTP as effector. Reaction conditions were as above with the following modifications: 2 mM GDP or CDP, 15 mM Mg(CH 3 COO) 2 , 40 M dTTP, and 3 M R2.
One unit of ribonucleotide reductase activity is defined as the amount of protein R1 that catalyzes the formation of 1 nmol of product/ min in presence of excess R2 protein at 25°C. Specific activity is expressed in units/mg of protein R1.
R1-R2 Complex Formation-The strength of the interaction between protein R2 and the different mutant R1 proteins was determined with activity measurements using the [ 3 H]CDP assay described above.
The R2-E441D interaction was determined directly considering R2 as a substrate. The concentration of R1 was constant on 0.035 M and R2 concentrations varied between 0.04 and 0.7 M. Measurement of a sample without R2 in the reaction mix was used to correct for background activity. The R2 binding to wild type R1 was measured in parallel using the same reaction solutions. V m and K m for R2 were obtained from direct curve fitting correcting for the free R2 concentration using the program Enzfit and Equation 1, where R1 and R2 are the total concentrations used.
The interaction between R2 and the E441A protein was determined in a series of experiments measuring R2 wild type activity and using the inactive E441A protein as competitive inhibitor for wild type R1 (12,13). R2 concentration was constant at 0.04 M, R1 wild type concentration was varied between 0.04 and 0.62 M, and E441A concentrations were 0, 0.052, 0.196, and 0.400 M. At each inhibitor concentration, one sample was measured without R1 wild type added to the reaction mix and this background activity was used to correct the activity. The apparent binding constants, K app , were obtained from double-reciprocal plots of wild type R2 activity versus wild type R1 concentration. The inhibition constant, K i , was obtained from a plot of K app versus inhibitor concentration.
Assays of Nucleotide Binding-Binding of the substrate GDP and the effector dTTP to the mutant R1 proteins was determined using the method of direct partition through ultrafiltration developed by Ormö et al. (39). For each experiment determining a dissociation constant, K d , for a nucleotide to a mutant R1 protein, the K d to wild type R1 protein was measured with the same solutions as a control. The dissociation constants, K d , and the number of binding sites, n, were determined with Scatchard analysis. The regression line of V on V/L was fitted by the method of least squares to directly obtain the K d and n values. L is the concentration of free ligand, and V is the amount in moles of bound ligand/mole of protein R1.
Binding experiments with dTTP were carried out at 25°C as described by Ormö et al. (39), using 0.5-8.0 M tritium-labeled dTTP and R1 concentration constant at 3.2 M or 2.8 M. The K d values were obtained from one experiment using seven different GDP concentrations.
GDP binding experiments were performed at 4°C in presence of 40 M dTTP in 50 mM Tris-Cl, pH 7.6, 10 mM Mg(CH 3 COO) 2 , and 2 mM DTT. Tritium-labeled GDP concentrations of 6.6 -300 M and constant R1 concentration ranging from 6.3 to 18.3 M were used in two to eight different experiment series. Higher nucleotide concentrations were used for the mutants with higher K d values.
EPR Samples and Measurements-The reactions were performed at 25°C by rapidly mixing equal volumes of the protein solution, 150 M R1, 100 M R2 in 50 mM Tris-Cl, pH 7.6, 15 mM Mg(CH 3 COO) 2 , 0.25 mM dTTP, 5 mM DTT and the substrate solution of 3.34 mM substrate (CDP, GDP, or CzDP) in the same buffer. Samples containing protein solution and buffer without substrate were used to detect the initial amount of tyrosyl radical and as a control of unspecific tyrosyl radical decay. Reactions were started by adding the substrate solution to the protein solution and stopped by freezing in n-pentane cooled with liquid nitrogen to Ϫ110°C. Incubation times of 2 s or longer were obtained by this method.
EPR spectra at 9 GHz measured at 77 K were recorded on a Bruker ESP 300 or Bruker 200D-SRC spectrometer using a cold finger Dewar flask for liquid nitrogen. Spin quantitation was obtained with a Cu 2ϩ -EDTA sample (1 mM Cu 2ϩ , 10 mM EDTA) and a secondary standard of active wild type E. coli R2 protein (0.98 mM tyrosyl radical) by comparing the double integrals. Subtractions were performed using the ESP 300 software. The CzDP-derived signal in R1 E441D and the substratedependent signal in E441Q were obtained by partial subtraction of the EPR spectrum of the wild type tyrosyl radical. Evaluation of the power of the half saturation, P1 ⁄2 , from microwave power saturation curves was performed as described by Sahlin et al. (40).
For kinetics at room temperature of E441D in reaction with CzDP, the EPR spectrometer was coupled to a stopped flow accessory as described by Lassmann et al. (41). Syringe A contained 100 M R2 and 150 M R1, and syringe B contained 3.36 mM CzDP. Both syringes contained 500 M dTTP, 15 mM MgCl 2 , 5 mM DTT in 50 mM Tris-Cl, pH 7.6. The formation of the CzDP-derived signal in R1 and the tyrosyl radical decay in protein R2 were determined at a field corresponding to the maximum of the EPR first derivative line of the two studied species. The kinetic scan and the field scan were triggered by the stopped flow accessory.
Time-dependent UV-visible Absorption Spectroscopy-The time dependence of tyrosyl radical decay and formation of a 316-nm chromophore was monitored in a Perkin-Elmer 2 scanning spectrophotometer. The enzyme mixture contained 20 M of E441Q protein, 15 M wild type R2 (with a radical concentration of 20 M), 0.25 mM dTTP, 15 mM Mg(CH 3 COO) 2 , in 50 mM Tris-Cl, pH 7.6. The reaction was started by addition of an aliquot of CDP or GDP to a final concentration of 2 mM, and 300 -450-nm spectra were recorded at 25°C for 90 min. All solutions were deoxygenated by flushing with argon prior to mixing. E441Q was prereduced with 10 mM DTT for 5 min at room temperature and desalted on a NAP-5 column equilibrated with argon-flushed buffer (50 mM Tris-Cl, pH 7.6, 15 mM Mg(CH 3 COO) 2 ).
Crystallization and Data Collection of Glu 441 Mutant Proteins-All mutants were crystallized in the space group R32 (42). The crystals were obtained by hanging drops of 10 l (5 l of protein mixture and 5 l of reservoir solution). The protein mixture contained 96 M (17 mg/ml) R1 protein and a 20-fold excess of a 20-residue peptide corresponding to the C terminus of the R2 subunit. The reservoir contained 17% lithium sulfate and 10 mM magnesium sulfate in 25 mM citrate buffer at pH 6.0.
Cryogenic cooling of the crystals in liquid nitrogen was achieved by rapidly transferring them from the hanging drop through a solution containing 11-15% ethylene glycol and 17% lithium sulfate in 25 mM citrate buffer at pH 6.0. The frozen crystals have smaller cell dimensions: a ϭ b ϭ 224 Å and c ϭ 334 Å compared with non-frozen R1 protein crystals with cell axes a ϭ b ϭ 227 Å and c ϭ 341 Å. Diffraction data for mutants were collected to 2.9-Å resolution using Rigaku rotating anode CuK␣ 1.54 Å and RAXIS2c image plate. The data sets were indexed with Denzo and scaled and reduced with Scalepack (43) and truncated with programs from the CCP4 suite (44). Statistics of data collection are presented in Table I.
Structural Refinement-The wild type coordinates of E. coli R1 ribonucleotide reductase (8) was used as a starting model for determining the phases of the mutant E441A. TNT (45) was first used for rigid body refinement initially with all three subunits together as one rigid body leading to a decrease of R from 52% to 47%. Then, each of the three subunits was treated as one rigid body leading to a further decrease of R to 30%. This worked despite differences in cell dimensions up to 2%. The refined E441A mutant was used for phasing the other two data sets using the same procedure. The models were initially refined with TNT using strict non-crystallographic symmetry. During the final refinement, REFMAC (44) was used with restrained non-crystallographic symmetry. After refinement, SIGMAA weighted 2F o Ϫ F c and F o Ϫ F c maps were calculated. The refined models were evaluated and corrected using O (46) and then further refined with REFMAC. Statistics of the refined models are presented in Table I.

Yield and Purity of Mutant R1 Proteins with Substitutions
for Position Glu 441 -The three mutant R1 proteins E441A, E441D, and E441Q behaved as wild type R1 throughout the protein purification procedure, and the final yields of mutant proteins were 5-10 mg of pure R1 protein/g of wet cells. From an overall yield throughout the purification procedure of 30 -50% and the known amount of chromosomally encoded wild type protein in crude extract (47), we estimate the contaminating wild type protein in these preparations to be 0.5-1%. To verify that the mutant proteins had intact overall structure and functionality, their three-dimensional structures and their capacity to bind protein R2, as well as substrate and effector nucleotides, were determined.
Three-dimensional Structures of the Mutant R1 Proteins-The structures of the mutant R1 proteins were solved by using the wild type structure as the initial model (8). Differences in structures compared with the wild type structure are only at the active site. The electron density at the mutated positions corresponds to the new residues. The E441Q and E441D mutations do not lead to any significant structural changes. The active site topography of the three mutations compared with wild type is seen in Fig. 1B. The glutamine side chain in E441Q has the same orientation as the glutamate side chain in the wild type structure, and differences in enzymatic properties should be due to changes in the chemical nature of that residue. Comparing the E441D structure with the substrate containing structure suggests that the interaction between Asp 441 and the 3Ј hydroxyl would be significantly weaker in the mutant. However, the carbonyl of the side chain of Asp 441 in the mutant makes the same hydrogen bond to Asn 437 as the glutamate in the wild type, maintaining the proposed electron transfer pathway between residues Cys 225 , Asn 437 , and Glu/Asp 441 (16). The mutation of Glu 441 to Ala leads to dramatic changes at the active site. The active site has collapsed. Met 620 has moved to the wild type position of Glu 441 and blocks the active site. The side chain of Cys 439 has also moved toward Met 620 .
Interaction of the Mutant R1 Proteins with Protein R2-The interaction between R1 and R2 can be measured directly in activity assays, as described previously (48). We found that the binding of E441D to R2, with a K m of about 0.04 M, was about the same as that obtained for the wild type R1 protein (Table II), indicating a similar interaction strength. For E441A, which lacks activity (see below), the interaction with R2 was measured as inhibition of the interaction between wild type R1 and R2 protein (12). The K i of 0.22 M for E441A showed that it is able to bind protein R2 almost as well as the wild type R1 control, which in this set of experiments had an apparent K m of 0.09 M (Table II). An attempted K i determination for E441Q indicated that also this protein was able to bind R2. However, the suicidal character of the E441Q protein, as will be discussed later, precludes an accurate determination of the interaction constant.
Interaction of the Mutant R1 Proteins with Effector and Substrate Nucleotides-The ultrafiltration assay (39) was used to measure binding of the allosteric effector dTTP to the mutant R1 proteins (Table II). All three mutant proteins bound the effector with K d values similar to the wild type protein, indicating that the effector binding sites are intact.
To test substrate binding to the mutant R1 proteins, the ultrafiltration assay by Ormö et al. (39) was used with the substrate GDP in presence of the effector dTTP at 4°C. This substrate-effector pair gives the strongest substrate binding to wild type protein (39,49). The K d values in Table II show that all the mutant R1 proteins can bind substrate but with weaker binding than the wild type protein. Higher dissociation constants were expected in the active site mutant proteins, as Glu 441 in the wild type protein forms a hydrogen bond to the 3Ј-hydroxyl group of the substrate. The weakest binding is seen in E441D with a K d of about 100 M. The mutant proteins E441Q, and E441A bind substrate about 2 times more strongly than does E441D. The wild type controls gave an average K d of 15 M, in reasonable agreement with previously reported wild type values (39). The number of binding sites obtained were close to two for all proteins, except for the E441A mutant, which showed approximately one binding site. The low number of binding sites seen with E441A was not further investigated, but may be explained by the observation in the crystal structure that Met 620 occupies part of the substrate binding site of this mutant protein (Fig.  1B). We conclude that the residue Glu 441 as proposed (8,16) contributes to substrate binding. Probably, a carboxylic acid residue of correct side chain length is needed for an optimal interaction with the substrate.
Enzyme Activity of Mutant R1 Proteins as Compared with Wild Type R1-The enzyme activity of the mutant R1 proteins in the presence of CDP as substrate and ATP as effector are compared with that of wild type R1 protein in Table III. The low activities of E441A and E441Q (ϳ1% of wild type activity) can be explained by the small amount of chromosomally encoded wild type R1 protein present in these extracts (10,47). The significantly higher specific activity, approximately 8% of wild type activity, found for the E441D mutant is, on the other hand, most likely intrinsic to the mutant protein. The activity of the E441D protein was also measured in presence of CDP or GDP as substrate and dTTP as effector and compared with the corresponding activity of the wild type protein. The CDP-dependent activity of the mutant protein was 10% of the wild type activity, and the GDP-dependent activity was 18% of the wild type activity.
The pH dependence of catalysis of E441D R1 was compared with that of wild type R1 activity between pH 6.1 and 8.6. The ratios between the two activities were constant over the entire pH range measured with the highest activity values around pH 8.0 (data not shown). A similar pH optimum was observed previously for the wild type protein (50). If the pH dependence of ribonucleotide reductase activity is contributed by the position 441 residue, our results indicates that a glutamic and an aspartic side chain contribute similarly.
Only the Glu 3 Asp Substituted Protein Has Intrinsic Enzyme Activity as Revealed in Reaction with the Substrate Analogue CzDP-To distinguish between low intrinsic activity and contaminating chromosomally encoded wild type protein in the mutant R1 proteins, we utilized the 2Ј-azido-substituted substrate analogue CzDP, which is a mechanism-based inhibitor. CzDP-dependent reactions are capable of a half-turnover reaction involving loss of the tyrosyl radical and appearance of a consecutive intermediate protein radical (cf. Fig. 3e) located at the active site residue Cys 225 (22,25,51). Fig. 3 (a-d) shows the time course of the CzDP-dependent reaction of the E441D mutant protein. Traces of the diagnostic CzDP-derived radical signal are apparent as early as 15 s after the start of incubation, and the signal increases during the first 2 min, after which it slowly decays. At 10 min, a major fraction of the total remaining EPR signal is the CzDP-dependent signal. An EPR spectrum subtracted for the remainder of the tyrosyl radical signal is shown in Fig. 3d. The decay rate of the tyrosyl radical signal and the formation rate of the CzDP-dependent radical are of the same magnitude, and about 20 times slower than corresponding values for the wild type enzyme (Table III). The calculated decay constant of the CzDP-dependent radical signal was 0.08 min Ϫ1 , and similar to the corre-  sponding decay constant (0.1 min Ϫ1 ) in a wild type reaction (22). 1 Similar formation and decay rates for the CzDP-derived species were also obtained by stopped flow EPR measurements at room temperature. These results clearly show that the E441D protein can promote mechanism-based radical transfer, reflecting that it possesses intrinsic catalytic activity. It also indicates that the CzDP-dependent radical decay is independent of the Glu 441 residue.
The mutant E441A protein did not promote any CzDP-dependent inactivation of the tyrosyl radical for at least 1 h, and no traces of other EPR signals were seen (Table III). This result shows that the E441A protein is catalytically inert and incapable of radical transfer between its active site and the tyrosyl radical in wild type R2.
Incubation of the E441Q protein in presence of wild type R2 and CzDP resulted in a very slow decay of the tyrosyl radical signal (ϳ70% radical remaining after 1 h, corresponding to a decay rate of less than 0.02% of that of the wild type mixture; Table III) and no traces of a CzDP-derived signal. One may ask why no CzDP-derived EPR signal was observed in the E441Q containing incubations. Assuming that the decay rate constant of a CzDP-derived radical is 0.1 min Ϫ1 , i.e. independent of the residue in position 441, the expected time of appearance and relative concentration of the CzDP radical can be calculated for any given formation rate. If, however, the formation rate constant in E441Q is only 0.02% of the wild type rate, the decay rate would be 20 times faster than the formation rate, and no CzDP radical would accumulate. Although the low enzyme activity of the E441Q protein as stated above most likely relates to contaminating wild type R1, the slow decay of the tyrosyl radical in the CzDP-dependent reaction indicates that this mutant protein is capable of 3Ј hydrogen atom abstraction. Another possibility is that the slow decay of the tyrosyl radical in the E441Q-dependent reaction follows another reaction pathway than the CzDP-dependent inactivation of the wild type complex.
A Transient Radical Intermediate in a Mechanism-based Reaction with E441Q and Natural Substrates-A rapid decay of the tyrosyl radical signal and appearance of a new transient EPR signal occurred when E441Q protein was incubated with wild type R2 and the natural substrate CDP (Fig. 4). The new EPR signal has its highest concentration between 1 and 2 min and dominates the EPR spectrum from there on (Fig. 5). Similar results were also obtained with the substrate GDP, although a slower formation of the intermediate was seen (data not shown). The transient intermediate obtained in presence of CDP has a formation rate constant of 0.03 s Ϫ1 and a decay rate constant of 0.004 s Ϫ1 . The decay rate of the tyrosyl radical is biphasic, with a fraction decaying with a fast rate constant (Ն1.2 s Ϫ1 ) and the remainder with a rate constant of approximately 0.01 s Ϫ1 (Fig. 5). Yet another transient EPR signal appears in these incubations and is resolvable during the first few seconds of incubation. In CDP-dependent reactions, it has a formation constant faster than 1.2 s Ϫ1 and a decay constant of 0.03 s Ϫ1 (cf. Fig. 5). This very early transient radical intermediate has not been characterized further.
Subtraction of the EPR spectrum of the tyrosyl radical from the composite EPR spectra obtained beyond 1 min of incubation reveals the EPR characteristics of the second transient radical intermediate (Fig. 6). It is an EPR hyperfine triplet, most likely due to coupling to two equivalent protons, with hyperfine coupling constants of 1.1 mT. The g value of 2.005 (5) is indicative of a free radical species. A saturation curve for the intermediate observed after 2 min of incubation time was constructed after subtraction of an appropriate portion of the wild type tyrosyl radical at the same microwave power. The evaluated power of half saturation (P1 ⁄2 ) for the new EPR signal is 75 W at 77 K (b ϭ 1.3). This is significantly different from the tyrosyl radical of protein R2 with a P1 ⁄2 of 13 mW at 77 K (b ϭ 1.0), indicating that the new signal does not interact with a metal center.
The Mechanism-based Reaction of E441Q and Natural Substrate Involves Cleavage of the 3Ј Carbon-Hydrogen Bond-Incubation of E441Q with CDP or GDP in presence of wild type R2 led to formation of a prominent chromophore in the near UV region subsequent to rapid decay of the tyrosyl radical. The formation rate constant of the new chromophore was 0.002 s Ϫ1 in presence of CDP (Fig. 7A), and 0.001 s Ϫ1 in presence of GDP. The subtracted spectrum of the chromophore shows a broad band centered at 316 nm, with a weak shoulder at about 360 nm (Fig. 7B). Similar UV-visible spectra were observed previously in incubations of wild type enzyme with 2Ј-substituted substrate analogues, and of suicidal C225S R1 protein in presence of R2 and substrate (11,18). The substrate-derived chro- mophore relates to R1 adducts of a highly reactive 2-methylene-3(2H)furanone intermediate (52), and is diagnostic for reactions involving suicidal decay of substrate radical intermediates formed by 3Ј carbon-hydrogen bond cleavage. DISCUSSION Before the structure of protein R1 was solved, the only known conserved active site residues determined essential for catalysis were the redox-active cysteines 225, 439, and 462. These were identified by systematic biochemical analysis of mutations of conserved cysteines in protein R1 (10,11,17,18). By solving the structure of protein R1, the conserved active site residues including the Cys 225 , Cys 439 , and Cys 462 were identified (8). Questions could now be asked about functions of other conserved residues seen in the active site and their role in the mechanism of action of ribonucleotide reductase. The substrate binding residues were initially proposed from a model building of substrate into the active site structure (8), and has recently been confirmed in the crystal structure of R1 in complex with substrate (16). One of the proposed substrate binding residues is Glu 441 . This residue is also proposed to participate as a base in the reaction mechanism (8, 25) (Fig. 2). Several lines of indirect evidence (reviewed in Ref. 1) support a radical-based reaction mechanism. However, the key substrate radical intermediate has not yet been identified.
In this study, site-directed mutagenesis was used to investigate the role of the conserved glutamic acid 441 in catalysis. The Glu 441 was converted to alanine, aspartic acid and glutamine, respectively, to (i) abort the carboxylic group, (ii) slightly modify the access of the side chain to the substrate, and (iii) change the chemical property but not the size of the side chain. Each mutant protein contributes information to unravel the role of Glu 441 in the reaction mechanism of ribonucleotide reductase.
The Glu 3 Ala mutation conferred an altered topography to the active site, most likely reflected in a higher binding constant. The lack of enzyme activity in the E441A protein strongly implies that the Glu 441 side chain is important in catalysis, but the possibility that an altered orientation of the substrate in the active site of E441A is the immediate cause cannot be excluded. A correct orientation of substrate may be critical for initiation of the radical transfer between R1 and R2, as was first suggested by Karlsson et al. (53).
The Glu 3 Asp mutant protein retains a substantial enzyme activity, despite the fact that the one carbon shorter aspartate side chain causes about 6 times weaker binding of substrate compared with the wild type protein. This shows that Glu 441 is an important substrate binding residue. The activity of E441D and its identical pH dependence compared with wild type R1 were expected since glutamate and aspartate are known to have similar pK a values in model peptides as well as protein (54), and are strong indications that a carboxylate function is needed at position 441 in protein R1 for ribonucleotide reductase activity.
The Glu 3 Gln mutation has the most illuminating characteristics. Even though the glutamine side chain occupies the identical geometric position in the active site as the glutamate in the wild type protein, the E441Q protein is incapable of catalytic turnover. This underscores the essentiality of a carboxylate function at position 441 for catalysis. However, our observations, that the E441Q protein in presence of natural substrates promotes tyrosyl radical decay and formation of transient radical intermediates and a furanone degradation product, show that it can bind substrate and undergo substrate initiated radical transfer between its active site and the tyrosyl radical in R2. Thus, the carboxylate function is not needed for initiation of the mechanism-based long range radical transfer reaction or for 3Ј-hydrogen abstraction.
The characteristics of the new transient mechanism-based EPR triplet signal in E441Q suggest that it is a free radical and that the unpaired spin has couplings to two equivalent protons. At least three possible locations may be considered for the new radical. (i) It could be located on a side chain in R2 that participates in the long range radical transfer during catalysis. However, the microwave power saturation behavior of the radical indicates that it is not in magnetic interaction with the diiron center of R2 making this possibility unlikely. (ii) It could be on an amino acid side chain in protein R1. Radicals seen in R1 are the CzDP-derived radical (22,24,55) and a perthiyl radical (26), both derivatives of Cys 225 in the active site. The former EPR signal is composed of a nitrogen-related triplet and a proton related 0.6 mT doublet splitting with a g av value of 2.008 (24), and the latter is composed of three discernible doublets with a splitting of 0.6 mT and a g av value of 2.03 (26). Our radical is clearly different from these two. In addition, as a preceding radical intermediate was observed, the intermediate discussed here plausibly occurs further along the reaction mechanism than the radical transfer pathway. (iii) It could be on the substrate. If the new signal is a substrate radical, it could be equivalent to a natural reaction intermediate in the wild type reaction, which in the mutant protein is observable because of a slower reaction rate in E441Q. A less likely possibility is that a new reaction course could be occurring in the mutant protein, as the suicidal reaction promoted by E441Q is mechanism-based.
For a carbon-centered substrate radical, hyperfine couplings would relate to protons in ␤-position to the carbon with the unpaired spin (C␣), i.e. a proton on an adjacent carbon (C␤). Such hyperfine interactions are highly dependent on geometry. The interaction is strongest when the substituent is eclipse with the half-filled p-orbital on C␣, and displays a strong angular dependence. The hyperfine splitting from a ␤-proton neighboring the radical can be calculated from the empirical relation A H ϭ B 1 ϩ B 2 cos 2 (56), where is the spin density and is the angle between the plane containing the p-orbital and the plane defined by the ␤ proton and the C␣-C␤ carbons. B 1 is close to 0 mT and B 2 Ϸ 5 mT (57,58). If all spin resides in the p-orbital and maximum overlap occurs ( ϭ 0°), we would observe a hyperfine splitting constant of ϳ5 mT, which is much larger than the observed 1.1 mT. However, ϭ 60°would give a hyperfine splitting of ϳ1.25 mT, which is close to the observed value. With Ͻ 1 and 0°, the observed EPR pattern could be achieved. A plausible substrate radical candidate would be the 3Ј-radical intermediate (or a 2Ј-radical intermediate, cf. 2-3 in Fig. 2). The crystal structure of the bound substrate in reduced R1 is compatible with 3Ј-endo or 2Ј-endo puckering (16), in which the relevant angles would be Յ40°. It is therefore quite possible that the 4Ј and 2Ј hydrogens (or 3Ј and 1Ј) could give rise to a 1:2:1 triplet with ϳ1.1 mT hyperfine splitting. The observed triplet could also arise from an unpaired electron on a nitrogen, giving a 1:1:1 triplet with a distorted line shape due to immobilization in the protein. We consider this less likely since the EPR lines are too symmetric to originate from an immobilized nitrogen species and the hyperfine splitting constant does not really fit. In addition, if the coupling were due to the nitrogen introduced in the active site by the E441Q mutation, it is unlikely that a similar radical would appear in the reaction with wild type protein and the substrate 2Ј-fluoromethylene-CDP (cf. Fig. 3 in Ref. 25). If, on the other hand, the radical occurred at N1 in the cytosine moiety, it would most likely couple to the proton at position C5 in analogy with what has been found for irradiated crystals of cytosine and uracil (59), but there is no other equivalent carbon linked proton in the cytosine base. Instead, since an almost identical intermediate EPR signal to the one presented in Fig.  6 is obtained also with the GDP substrate, it is most likely that the radical resides in the ribose moiety of the substrate. Definite identification has to await specific isotopic labeling experiments.
Is it plausible that a substrate radical intermediate would form and be stabilized in the E441Q-dependent reaction? Formation of the furanone product is diagnostic for 3Ј-hydrogen abstraction in combination with a defective reduction of the 2Ј-position (18,23,60). The furanone derivative is a degradation product of the 2Ј-deoxygenated nucleotide, suggesting that the E441Q-dependent reaction proceeds at least to 3 in Fig. 2. Lenz and Giese (30) showed that 2Ј-deoxygenation of a model compound was subject to general base catalysis and suggested that the rate-limiting step is deprotonation of the 3Ј-hydroxyl group, which would be needed for efficient leaving of the 2Јhydroxyl group. In the mutant enzyme, the Gln 441 side chain would still be expected to form a hydrogen bond with the 3Ј-hydroxyl group (cf. Fig. 1, A and B), but not to act as a proton acceptor, which means that such a rate-limiting step would be slowed down in the mutant enzyme to allow retention of a 3Ј-radical intermediate (2 in Fig. 2). A plausible scenario is therefore that the E441Q-dependent reaction proceeds through 1 and 2, and with a slower rate than in the wild type reaction to 3, after which degradation to a highly reactive 2-methylene-3(2H)furanone occurs that generates the furanone:protein adduct. If this scenario is correct, the E441Q reaction would be defective in forming 4, suggesting that the carboxylate at position 441 is important for efficient deprotonation of the 3Јhydroxyl group but absolutely essential for the reduction of the 2Ј-position. This is in good agreement with its postulated role in the transfer of redox equivalents from the redox-active Cys 225/462 to the substrate (16).
The different rate constants observed in the reaction of E441Q protein and substrate in presence of protein R2 are indicative of a consecutive reaction sequence involving at least three discernible steps. First, the very rapid decay of a major fraction of the tyrosyl radical of protein R2 (Ն1.2 s Ϫ1 ) matches the very rapid appearance of an (in this report uncharacter-ized) early radical intermediate (Ն1.2 s Ϫ1 ). Second, the decay of the first radical intermediate (0.03 s Ϫ1 ) matches the appearance of the triplet radical intermediate (0.03 s Ϫ1 ) characterized in this report. Third, the decay of the triplet radical intermediate (0.004 s Ϫ1 ) matches the formation of the stable furanone derivative (0.002 s Ϫ1 ). In addition, in evaluating the formation of the furanone derivative, we observed a distinctive lag phase, indicating that the furanone derivative is an end product in a consecutive reaction sequence. The obtained formation rate of the intermediate in this consecutive reaction (0.03-0.05 s Ϫ1 ) matches well with the formation rate of the triplet radical intermediate from above. These results are summarized in Scheme 1, where a plausible sequence of observed radical species and the furanone end product are depicted. Taken together, the kinetic data in combination with the preliminary characterization of the transient EPR signal, the formation of a furanone derivative, and the similarity of our transient radical with the substrate analogue radical of van der Donk et al. (25) strongly suggest that the transient radical intermediate reported here corresponds to one of the substrate radical intermediates (most likely 2 in Fig. 2) postulated for the reaction mechanism of ribonucleotide reductase.
In summary, the worse binding of substrate to all three mutant R1 proteins presented here and the modified active site structure in E441A show that Glu 441 is an important substrate binding residue. The lack of activity in E441A, the low activity of E441D, and the suicidal reaction of E441Q show that Glu 441 is essential for catalysis and that a carboxylate function at position 441 is a minimum requirement for catalysis. The results suggest that it is playing an essential role in the reduction of substrate. They also support the proposed function of Glu 441 as a general base in the reaction mechanism. The most intriguing finding is the observation of a possible transient substrate radical intermediate in the suicidal mechanism-based reaction of E441Q. It is suggested that the radical intermediate is a naturally occurring intermediate during turnover of the wild type enzyme, plausibly a 3Ј-radical intermediate. Characterization of this new species is under way.