Activation of the Anaerobic Ribonucleotide Reductase fromEscherichia coli

The anaerobic ribonucleotide reductase ofEscherichia coli catalyzes the synthesis of the deoxyribonucleotides required for anaerobic DNA synthesis. The enzyme is an α2β2 heterotetramer. In its active form, the large α2 subunit contains an oxygen-sensitive glycyl radical, whereas the β2 small protein harbors a [4Fe-4S] cluster that joins its two polypeptide chains. Formation of the glycyl radical in the inactive enzyme requiresS-adenosylmethionine (AdoMet), dithiothreitol, K+, and either an enzymatic (reduced flavodoxin) or chemical (dithionite or 5-deazaflavin plus light) reducing system. Here, we demonstrate that AdoMet is directly reduced by the Fe-S center of β2 during the activation of the enzyme, resulting in methionine and glycyl radical formation. Direct binding experiments showed that AdoMet binds to β2 with aK d of 10 μm and a 1:1 stoichiometry. Binding was confirmed by EPR spectroscopy that demonstrated the formation of a complex between AdoMet and the [4Fe-4S] center of β2. Dithiothreitol triggered the cleavage of AdoMet, leading to an EPR-silent form of β2 and, in the case of α2β2, to glycyl radical formation. In both instances, 3 methionines were formed per mol of protein. Our results indicate that the Fe-S center of β2 is directly involved in the reductive cleavage of AdoMet and suggest a new biological function for an iron-sulfur center, i.e redox catalysis, as recently proposed by others (Staples, R. C., Ameyibor, E., Fu, W., Gardet-Salvi, L., Stritt-Etter, A. L., Schürmann, P., Knaff, D. B., and Johnson, M. K. (1996) Biochemistry 35, 11425–11434).

Escherichia coli uses different enzymes for the de novo synthesis of deoxyribonucleoside triphosphates (dNTPs) during aerobic and anaerobic growth (1). The aerobic ribonucleotide reductase, coded by nrdAB genes, has an ␣ 2 ␤ 2 structure, with the large ␣ 2 protein carrying allosteric and catalytic sites and the small ␤ 2 protein harboring two oxo-bridged diferric centers capable of forming and stabilizing a tyrosyl radical in their vicinity (2). During catalysis, the stable tyrosyl radical is thought to generate by long range electron transfer a transient thiyl radical on the large protein which, together with two redox-active thiols, catalyzes the reduction of the ribose group to deoxyribose (3). This aerobic enzyme was described already in 1960 and has become the prototype of a whole class of ribonucleoside diphosphate reductases (class I) that produce dNDPs in some bacteria and all higher organisms.
The anaerobic ribonucleoside triphosphate reductase of E. coli was discovered only in 1989 (4). It is the prototype for a group of class III anaerobic enzymes also found in other anaerobically growing microorganisms both in archaebacteria and eubacteria (5). It was first believed to be a single large homodimeric ␣ 2 protein, but subsequently, a second smaller homodimeric ␤ 2 protein was found to be required for catalytic activity (6). The two proteins, coded by the nrdD and nrdG genes, respectively, are tightly bound to each other and form an ␣ 2 ␤ 2 complex as shown by sucrose gradient centrifugation and kinetic experiments (7). The ␤ 2 protein contains a [4Fe-4S] cluster that links its two polypeptide chains, and the ␣ 2 protein harbors an oxygen-sensitive glycyl radical but no metal center (8 -10).
It is likely that both aerobic and anaerobic enzymes use redox-active cysteines as a source of electrons and catalyze the same radical chemistry as judged from studies with mechanism-based inhibitors and from stereochemical analysis of the reduction reaction (11)(12)(13). Both generate a stable free amino acid radical, as part of their protein structure, essential for activity: a tyrosyl radical in the case of the aerobic enzyme and a glycyl radical in the anaerobic one.
The activation of the aerobic reductase is reasonably well understood; generation of the tyrosyl radical is an oxygen-dependent process in which the iron center is involved (14 -16). On the other hand, our knowledge concerning the formation of the glycyl radical of the anaerobic enzyme is limited. As prepared, the pure enzyme lacks the glycyl radical and is inactive. The radical is generated, and the enzyme is activated during anaerobic incubation with S-adenosylmethionine (AdoMet) 1 together with NADPH, flavodoxin, flavodoxin reductase, K ϩ , and dithiothreitol (DTT) (8,(17)(18)(19)(20). Photochemically reduced deazaflavin can substitute for NADPH and the flavodoxin system (19). During the reaction, AdoMet is reduced and cleaved to 5Ј-deoxyadenosine and methionine (17). A similar mechanism seems to operate during generation of catalytic free radicals in pyruvate-formate-lyase from E. coli and lysine 2,3-aminomutase from Clostridium pasteurianum (21,22). Both contain a metal center and require AdoMet, which is converted to 5Ј-deoxyadenosine and methionine during the reaction.
Against this background, we suggested in 1993 a speculative scheme for the generation of the glycyl radical of the anaerobic ribonucleotide reductase, involving four steps (9). In the first, the iron-sulfur cluster of the enzyme is reduced by the flavodoxin system or by photoreduced deazaflavin. In the second step, AdoMet is bound and reduced to a radical intermediate, which in the third step is cleaved to the 5Ј-deoxyadenosyl radical and methionine. Finally, in the fourth step, this radical generates the glycyl radical by abstracting a hydrogen atom of the ␣-carbon of glycine 681 (10).
In the present work, we present a series of experiments involving the separate ␤ 2 protein and the complete ␣ 2 ␤ 2 enzyme that provide evidence for the proposed scheme described above. It is thus shown that AdoMet is reduced by the reduced iron-sulfur center, resulting in methionine and glycyl radical formation. This suggests a new biological function for an ironsulfur cluster.

Methods
Measurements of Enzyme Activity-In the first step, the enzyme (0.32 M), in a total volume of 20 l, was deaerated with moist argon on a manifold during 30 min at room temperature. In parallel, the activation mixture was deaerated separately in the dark, and 15 l were added to the protein solution so that the final concentrations were 5 mM sodium formate, 5 mM DTT, 30 mM potassium chloride, 500 M AdoMet, 38 M 5-DAF, and 30 mM Tris-HCl, pH 8. Irradiation with a slide projector, 20 cm away from the reaction mixture, was then applied for 30 min under argon. When sodium dithionite was substituted for 5-DAF, it was added separately from a deaerated stock solution in 30 mM Tris-HCl, pH 8. In this case, the protein solution received first dithionite (final concentration 4 M), followed by the activation mixture. The solution was then incubated for 30 min. A third activating mixture was also used. It is similar to the one described first except that flavodoxin, flavodoxin reductase and NADPH (final concentrations of 20 g/ml, 40 g/ml, and 1.25 mM, respectively) replaced 5-DAF.
In the second step, 15 l of the substrate mixture (giving a final concentration of 1.4 mM [ 3 H]CTP (10 cpm/pmol), 1 mM ATP, 10 mM MgCl 2 ) was added to initiate the reduction of the substrate. The reaction was stopped by the addition of 0.5 ml of 1 M HClO 4 , and the solution was worked up as described earlier (24). One unit of enzyme activity represents the formation of 1 nmol of dCTP/min.
Catalytic Reduction of AdoMet-␤ 2 (40 pmol) or ␣ 2 ␤ 2 (16 -40 pmol) and [methyl-3 H]AdoMet (15 M) were incubated for 1 h at room temperature under a stream of humidified argon in 0.1 M Tris-HCl, pH 8, containing 40 mM KCl, 8 mM DTT and either 2 mM NADPH, flavodoxin (0.4 g), and flavodoxin reductase (0.1 g) or 5 M 5-DAF in a total volume of 35 l. With 5-DAF, irradiation with a slide projector was required. The reaction was stopped by adding 665 l of cold water, and then 600 l were loaded onto a 2-ml Bio-Rad Dowex 50WX8 (Na ϩ form) column. Methionine was measured from the radioactivity eluted with 3 ml of water. Control experiments indicated that, under these conditions, more than 98.5% of AdoMet was bound and more than 95% of methionine ([ 35 S]methionine) was recovered. Appropriate controls without proteins gave the corresponding blank values.
EPR Spectroscopy-Low temperature EPR spectra were recorded on a Varian E 109 (9.5 GHz) EPR spectrometer or on a Bruker ESP 300 spectrometer, both equipped with an Oxford Instruments ESR 900 Helium Flow Cryostat. Double integrals of EPR signals were evaluated by using a computer online with the spectrometer. All of the 0.2-ml samples were introduced into quartz EPR tubes under argon and sealed with a rubber septum.
1-EPR Spectrum of the Glycyl Radical-1 mg of ␣ 2 ␤ 2 was incubated anaerobically in an EPR tube with 0.5 mM AdoMet, 5 mM DTT, 5 mM sodium formate, 30 mM KCl, 20 g of flavodoxin, 80 g of flavodoxin reductase, 1.25 mM NADPH in 0.2 ml of 30 mM Tris-HCl, pH 7.5. After 10 min reaction under anaerobic conditions, the EPR tube was frozen in liquid nitrogen and analyzed by EPR spectroscopy for its glycyl radical content.
2-EPR Spectrum of the Reduced Iron-Sulfur Cluster-Proteins ␤ 2 or ␣ 2 ␤ 2 (20 -140 M) in 30 mM Tris-HCl, pH 8, were deaerated in EPR tubes inside an anaerobic box at 4°C for 1 h. Then the EPR tubes were transferred to an anaerobic manifold. Some of them received 5-10 mM dithionite exclusively, while others received first a deaerated solution of AdoMet (from 0 to 5 mM) immediately followed by dithionite. After a 40-min incubation at room temperature, the tubes were frozen in liquid nitrogen and analyzed by EPR spectroscopy.
For the power saturation experiment (Fig. 1D) the amplitude of the central component of the EPR signal (g ϭ 1.92) was measured as a function of the incident microwave power in the 40 -4 db (0.02-82 milliwatts) range at 10 K. The amplitude at the lowest power value was taken as 100%. At each power increment, the amplifier gain was changed to keep the signal amplitude constant under nonsaturating conditions.
Methionine Determination-Methionine content was determined with a Pharmacia-LKB4151 Alpha Plus amino acid analyzer.
Single Turnover Experiments-All of the following procedures were done under strict anaerobic conditions. ␤ 2 or ␣ 2 ␤ 2 in 30 mM Tris-HCl, pH 8, 30 mM KCl was deaerated inside an anaerobic box during 1 h at 4°C in an EPR tube. Reduction was started on the manifold by irradiation of the protein solution in the presence of 5-DAF (38 M) at room temperature. After 40 min, the solution was put in the dark using an aluminum foil, and AdoMet (0.5 mM) was added anaerobically. At time intervals (from 0 to 120 min), an aliquot (1 nmol of protein) was removed and assayed for methionine formation. The rest of the solution was immediately frozen in liquid nitrogen for EPR spectroscopy analyses of the iron-sulfur center for ␤ 2 or of the glycyl radical for ␣ 2 ␤ 2 . Recording conditions were as follows for the iron-sulfur center: microwave power, 0.1 milliwatt; modulation amplitude, 1 millitesla; frequency, 9.62 GHz; temperature, 10 K. For the glycyl radical, conditions were as follows: microwave power, 1 microwatt; frequency, 9.62 GHz; modulation amplitude, 1 millitesla; temperature, 20 K.
Binding of AdoMet to ␤ 2 -A filter-binding assay used earlier to measure binding of allosteric effectors to ribonucleotide reductase (25) was adapted to the anaerobic conditions required for the present experiments. In the standard procedure, the enzyme in a volume of 90 l of 50 mM Tris-HCl, pH 7.5, 5 mM DTT, 20 mM KCl was flushed at room temperature with moistened argon for 60 min on an anaerobic manifold, followed by the addition of 10 l of an anaerobic solution of 25 mM dithionite to reduce the iron-sulfur center. Also, reduction by 5-DAF plus light was used, but with less reproducible results. After an additional 20 min, the closed tube was shifted to an anaerobic hood kept at ϩ4°C, received 0.9 ml of 50 mM Tris-HCl, pH 7.5, 20 mM KCl, 1 mM dithionite, and was transferred to the chamber of a syringe attached to a filter device containing a GSWP 01300 Millipore filter (25). [methyl-3 H]AdoMet in 20 l was added, and the mixture was immediately pushed through the filter. The filter was removed, blotted from underneath with Kleenex, and transferred to a scintillation vial. After the addition of 1 ml of 0.1 M HCl, the radioactivity was determined and, after substraction of an appropriate blank lacking the enzyme, used to calculate the amount of AdoMet on the filter.
In some experiments, the volume during the filtration step was decreased to 0.15 ml, and the filtrate was used to analyze for the formation of methionine during the binding process. The smaller volume provided a higher sensitivity for the assays. In these experiments, we could also determine the amount of AdoMet bound to the filter from the loss of radioactivity in the filtrate.
All binding experiments were done as duplicates. With some practice, variation between duplicates was usually below 10%. The limitations of the method arise from a combination of two problems: the limited capacity of the filter to bind protein and the increasing background radioactivity at increasing concentrations of the labeled ligand.

RESULTS
Spectroscopic Properties of the ␣ 2 ␤ 2 Complex-In the following, ␣ 2 ␤ 2 defines solutions obtained by incubation of stoichiometric 1:1 amounts of each pure protein: the iron-free ␣ 2 large protein and the ␤ 2 small protein (containing two irons and two sulfides per polypeptide chain) prepared separately from overproducing E. coli strains (7). Fig. 1 shows the light absorption spectrum of the oxidized enzyme (Fig. 1A) and the EPR spectrum after anaerobic reduction with either dithionite or photoreduced 5-DAF (Fig. 1B). Both spectra of ␣ 2 ␤ 2 were almost identical to those of ␤ 2 , reported previously (7). The power saturation properties (Fig. 1C) and the temperature dependence (not shown) of this EPR signal are characteristic of a [4Fe-4S] ϩ center and close to those of ␤ 2 . Quantification of the EPR signal showed that the reduction was quantitative with 5-DAF as the reducing agent. With dithionite, the reduction yield did not exceed 50 -70%. This then indicates that the properties of ␣ 2 ␤ 2 are exclusively due to the [4Fe-4S] center of ␤ 2 and that binding of ␣ 2 to ␤ 2 has only minor effects on the symmetry and the electronic properties of that center. During affinity chromatography of ␣ 2 ␤ 2 on dATP-Sepharose, the ironsulfur center was partially converted into an oxidized [3Fe-4S] ϩ center, as shown from the appearance of an intense characteristic isotropic EPR signal at g ϭ 2.01 (Fig. 1D). This explains why we earlier concluded that the reductase, when purified by dATP-Sepharose chromatography, contains a [3Fe-4S] ϩ cluster (9).
Activation of ␣ 2 ␤ 2 : Requirement of an Electron Source-Activation of ␣ 2 ␤ 2 involves the generation of a glycyl radical on ␣ 2 during anaerobic incubation with AdoMet, K ϩ , DTT, and a reducing system. Photoreduced 5-DAF (19) could generate active enzymes with specific activities of up to 1000 nmol of dCTP/min/mg of protein. As shown in Fig. 2 (A and B), dithionite was also able to activate ␣ 2 ␤ 2 . The maximal specific activity, obtained with a 12-fold molar excess of dithionite after a 30-min incubation, was 400. As shown above, both dithionite and 5-DAF converted the iron-sulfur center of ␣ 2 ␤ 2 to the EPR-active reduced [4Fe-4S] ϩ form.
NADPH plus flavodoxin plus flavodoxin reductase also generates an active ␣ 2 ␤ 2 protein, with a specific activity of about 400 (19). Quantitation of the EPR signal of the glycyl radical (Fig. 3) showed formation of up to one radical per ␣ 2 ␤ 2 after a 10-min reaction. However, this reducing system did not give rise to any discernible iron-based EPR signal during incubation with ␣ 2 ␤ 2 in the absence of AdoMet. This showed that the enzyme was not reduced to the [4Fe-4S] ϩ form by flavodoxin.
The Cleavage of AdoMet to Methionine Is Catalyzed by the Iron-Sulfur Center-During activation of the enzyme, AdoMet is converted to 5Ј-deoxyadenosine and methionine (17). This reaction can be monitored by using [methyl-3 H]AdoMet and measuring the formation of [methyl-3 H] methionine. Fig. 4 shows that ␣ 2 ␤ 2 catalyzes the reduction of AdoMet to methionine either by NADPH in the presence of flavodoxin and flavodoxin reductase or by photoreduced 5-DAF. DTT greatly stimulated the reaction (data not shown). No methionine could be detected when ␤ 2 or the reducing agent was omitted from the activation mixture. As shown in Fig. 4, ␤ 2 alone could support the reaction with 5-DAF as the reducing agent, with DTT again required for maximal activity. However, no methionine formation was observed with the flavodoxin system as a reducing reagent (data not shown). Thus, ␣ 2 had a clear stimulating effect. Since apo-␤ 2 , the iron-depleted form, was totally inactive (data not shown) with either reducing system even in the presence of ␣ 2 , the iron-sulfur center appeared to be the key component of the enzyme.

Binding of AdoMet to Reduced ␤ 2 : Stoichiometry and Binding Constant-Binding experiments with [methyl-3 H]AdoMet
were done under anaerobic conditions by a filter binding assay (25), as described under "Experimental Procedures." The protein was first reduced on the anaerobic manifold with either dithionite or 5-DAF, transferred to the filtration device in the anaerobic hood, mixed with AdoMet, and filtered. Bound AdoMet was retained on the filter together with protein, and its amount was determined either from the radioactivity on the filter or from the decrease of radioactivity in the filtrate.
Binding strictly depended on the presence of K ϩ , whereas Mg 2ϩ was not required (Table I). The addition of DTT improved the binding capacity of ␤ 2 significantly. When air was admitted to the reduced enzyme, no AdoMet was bound. Methionine was not produced during the short time period it took to mix and filter the solution. However, when enzyme plus AdoMet were incubated at ϩ4°C, a slow formation of methionine, found exclusively in the filtrate, ensued without loss of bound AdoMet (Fig. 5). No methionine was bound to the enzyme at any time point (data not shown). These results show that enzyme-bound AdoMet, at ϩ4°C, was slowly reduced to methionine that was immediately released from the protein. In the presence of an excess of dithionite, the iron-sulfur center remained reduced and could continuously bind new molecules of AdoMet. Fig. 6A shows a Scatchard plot in which the binding stoichiometry ( ϭ mol of bound AdoMet/mol of ␤ 2 ) was calculated at varying AdoMet concentrations. The slope of the linear curve gives a binding constant of 10 M. The curve extrapolates at the abscissa to a value of 0.6 mole/mole of ␤ 2 . Fig. 6B shows results from a separate experiment in which AdoMet binding was determined at increasing protein concentrations. A linear relation is found up to 40 g (1.1 M) of ␤ 2 , but then the curve levels off rapidly. In the experiment depicted in Fig. 6, we used 51 g (1.5 M), slightly outside the proportional range, to give better signal:noise ratios. Correcting for this fact, we calculate that 1 mol of our preparation of ␤ 2 could bind 0.7-0.75 mol of AdoMet and suggest a 1:1 stoichiometry for the binding of AdoMet to ␤ 2 .
The binding behavior of ␤ 2 did not change noticeably when ␣ 2 was added in the experiment (data not shown). However, minor changes would not have been detected, since this experiment was made with only 10 g of ␤ 2 (together with 40 g of ␣ 2 ) in view of the limited protein binding capacity of the membrane.
Binding of AdoMet to ␤ 2 : the Effect on the Iron-Sulfur Center-Characterization of a reduced ␤ 2 -AdoMet complex by EPR spectroscopy can only be achieved if electron transfer to and subsequent cleavage of AdoMet is significantly slowed down. From the observation that DTT greatly stimulated the formation of methionine during the ␤ 2 -catalyzed reduction of AdoMet, a simple way to produce this condition was to omit DTT from the reaction mixture. Anaerobic incubation of ␤ 2 in the absence of DTT with a chemical reductant (dithionite or photoreduced 5-DAF) for 1 h generated the previously reported EPR spectrum, indicating that the iron-sulfur center had been reduced to a [4Fe-4S] ϩ form. When AdoMet was added to the reaction mixture, the EPR spectrum was instantaneously modified with the low field feature high field-shifted and the high field one greatly broadened but with no change of the intensity of the signal. This drastic change was dependent on AdoMet concentration; increasing the AdoMet concentration resulted in the disappearance of the initial EPR spectrum and the concomitant appearance of the new one; conversion was complete at [AdoMet] ϭ 1 mM (Fig. 7). When the same experiment was carried out with ␣ 2 ␤ 2 , the EPR signal was also changed, albeit to a smaller extent (data not shown).
Reaction of the Reduced Iron-Sulfur Cluster with AdoMet-With photoreduced 5-DAF it was possible to dissect the electron transfer from the reduced cluster to AdoMet into two steps. During illumination of a mixture of ␤ 2 , 5-DAF, K ϩ , and DTT, a flux of electrons first generates the reduced form of the cluster in the absence of AdoMet. After transfer to the dark, AdoMet was added, and the electron transfer from reduced ␤ 2 to AdoMet was monitored in the absence of a continued reduction of ␤ 2 . Changes in the [4Fe-4S] cluster were observed by EPR spectroscopy, and the formation of methionine served to measure the one-electron reduction of AdoMet. Fig. 8 shows the time-dependent decay of the EPR signal of reduced ␤ 2 and the parallel formation of methionine. Unexpectedly, the reaction occurred in two steps. In the first, loss of the EPR signal and formation of one methionine per ␤ 2 followed first order kinetics with the same rate constant (0.12 min Ϫ1 ). In the second step, and with the same rate constant, an additional 2 mol of methionine were formed per mol of ␤ 2 with no change of the ironsulfur cluster detectable by EPR spectroscopy, since the solution then remained EPR-silent. In the absence of AdoMet, the EPR signal did not decay during 1 h of incubation in the dark. In the absence of DTT, little methionine (0.1 methionine/␤ 2 ) was formed during the 60-min incubation.
The same experiment was also carried out with ␣ 2 ␤ 2 . In this case, EPR spectroscopy was also used to monitor the generation of the glycyl radical on ␣ 2 . In less than 3 min, glycyl radical formation was complete. Methionine continued to be formed up to 10 min, when 3 methionines/glycyl radical had been generated, but then stopped (Fig. 9). At the earliest time point, the EPR signal of the the iron-sulfur center was no more detecta-  The complete system was as described under "Experimental Procedures" with 51 g of ␤ 2 and 40 M [methyl-3 H]AdoMet in a filtration volume of 1.0 ml. When added, Mg 2ϩ was 10 mM. Where indicated, air was admitted after reduction of the iron-sulfur cluster, and filtration was made aerobically ( ϭ mol of bound AdoMet/mol of ␤ 2 ).

DISCUSSION
The ribonucleotide reductase isolated from anaerobically grown E. coli cells is the prototype of class III ribonucleotide reductases (1). It consists of an ␣ 2 ␤ 2 complex, with ␤ 2 carrying an essential iron-sulfur center (7). Since ␤ 2 after reconstitution binds only 2Fe 2ϩ and 2S 2Ϫ per polypeptide chain and in the reduced form exhibits an EPR signal characteristic of a [4Fe-4S] ϩ center, our present model is that the homodimer ␤ 2 harbors a [4Fe-4S] cluster at the interface of its two polypeptide chains. However, this assignment mainly rests on EPR data and awaits confirmation from other spectroscopic studies. Tight binding of ␣ 2 to ␤ 2 is shown here to have only slight effects on the light absorption and the EPR properties of the iron-sulfur cluster of ␤ 2 .
In fact, this iron-sulfur center seems to be unique in terms of its spectroscopic properties, which are currently under investigation. It is quite labile, especially in the reduced form, but can be significantly stabilized by the addition of DTT (7). This may, at least partly, explain the strong stimulating effect of DTT on the enzyme reaction. All of these unique features might be due to the fact that the [4Fe-4S] center is shared by two polypeptide chains, a situation that has been previously observed only in two cases: the iron protein of the nitrogenase (26) and the cluster F X of the photosystem I (27).
The anaerobic ribonucleotide reductase is a radical enzyme. This is a rather trivial statement, since all ribonucleotide re- Then illumination was stopped, 500 M AdoMet was added anaerobically, and the solution was kept in the dark. At the indicated times, the glycyl radical (q) and methionine (E) were assayed as described under "Experimental Procedures." ductases known so far contain a free radical on one of their polypeptide chains (1). Generation of the radical at glycine 681 of the large component of the enzyme requires AdoMet, K ϩ , DTT, and a source of electrons. The latter can either be a chemical reagent such as photoreduced 5-DAF or dithionite (Fig. 2) or an enzyme system consisting of NADPH plus flavodoxin plus flavodoxin reductase, with comparable activities (19).
In Scheme 1, we summarize our view of the mechanism of the reaction. A first version of this mechanism was published in 1993 but had at that time no experimental support. From the present work, we confirm this earlier speculation with experimental evidence for a direct electron transfer from the reduced cluster to AdoMet.
Since enzyme activation and methionine formation absolutely require the presence of a reducing agent, we suggest that the active form of ␤ 2 is the reduced one in which the cluster is in the EPR-active [4Fe-4S] ϩ state. This form accumulates when ␤ 2 or ␣ 2 ␤ 2 is incubated with dithionite or photoreduced 5-DAF in the absence of AdoMet (Scheme 1, reaction 1). Enzymatically reduced flavodoxin, the physiological electron donor, is unable to reduce the iron-sulfur center to an EPR-detectable level in the absence of AdoMet. Nevertheless, it supports methionine and glycyl radical formation in the presence of AdoMet. A likely explanation for this apparent paradox is that, in the absence of AdoMet, the redox potential of the metal center is below Ϫ350mV, high enough to allow electron capture from dithionite or 5-DAF but not from reduced flavodoxin. Binding of AdoMet and coupling of the reduction of the cluster to its reoxidation by AdoMet might result in an increase of the apparent redox potential of the cluster and drive the whole process. Whether AdoMet affects the redox potential of the iron-sulfur center is a fascinating hypothesis that remains to be studied.
The drastic and instantaneous change of the shape, but not the intensity, of the EPR signal of the reduced cluster upon the addition of AdoMet strongly suggests that, in the second step, AdoMet binds to ␤ 2 (also in the presence of ␣ 2 ) to form an enzyme-AdoMet complex (Scheme 1, reaction 2). The large modification of the EPR signal, reflecting a change of the symmetry of the metal center, can be explained if AdoMet comes to a site in close proximity to that center or if AdoMet binding causes a significant conformational change of the polypeptide chain affecting the center.
The results from direct binding experiments support this interpretation. One mol of ␤ 2 bound close to 1 mol of AdoMet with a K d value of 10 M. Binding depended completely on the generation and maintenance of the reduced cluster, thus explaining why early efforts to demonstrate binding of AdoMet to ␤ 2 by methods involving equilibrium dialysis under aerobic conditions met with no success. DTT and K ϩ were also required, pinpointing the earlier described requirements of enzyme activation for K ϩ and DTT to the binding step. At first sight, it may seem unexpected that the homodimeric ␤ 2 protein binds only one molecule of AdoMet and not two. This instead is in line with the proposal of a unique active center containing both the cluster and the AdoMet binding site at the interface of the two polypeptide chains.
For the third step, we have now evidence from single turnover experiments that the reduced iron-sulfur center of ␣ 2 ␤ 2 in the presence of DTT transfers electrons to AdoMet and induces its cleavage. Methionine was formed at the expense of the reduced cluster which, in parallel, was oxidized to an EPRsilent form. Cleavage of AdoMet is proposed to generate transiently one equivalent of a 5Ј-deoxyadenosyl radical and subsequently one stable glycyl radical (Scheme 1, Reaction 4). Since this reaction is fast, only the final glycyl radical and not the intermediate 5Ј-deoxyadenosyl radical could be detected by EPR spectroscopy. A puzzling observation was that three methionines were reproducibly formed per glycyl radical. That the reduced iron cluster is able to transfer electrons to AdoMet is further supported by the observation that electrons from the iron-sulfur cluster of ␤ 2 were efficiently transferred to AdoMet with no need for ␣ 2 ; again, methionine was formed, while the EPR-active reduced cluster was oxidized to an EPR-silent form. Also in this case, as with ␣ 2 ␤ 2 , 3 methionines/␤ 2 were generated. However, the reaction was much slower, showing that ␣ 2 binding to ␤ 2 had a strong stimulatory effect. The rate of formation of the first equivalent of methionine was identical to the rate of oxidation of the cluster (Fig. 8), in agreement with reaction 3 of Scheme 1. Subsequently, two additional methionines were formed at the same rate, whereas the iron center remained EPR-silent. These results lead to the unusual conclusion that during reduction of both ␤ 2 and ␣ 2 ␤ 2 the ironsulfur cluster has the ability to store as many as three electron equivalents, which subsequently can be delivered at a sufficiently low redox potential to reduce three molecules of AdoMet. In the case of ␣ 2 ␤ 2 , only the first electron serves to generate the glycyl radical, in agreement with reactions 3 and 4 of Scheme 1. This then explains why, as shown in Fig. 9, methionine formation continues, whereas the glycyl radical has reached its maximum.
We have also shown that ␣ 2 ␤ 2 can achieve multiple turnovers during reduction of AdoMet to methionine and thus behaves as an S-adenosylmethionine reductase. The catalytic properties of ␣ 2 ␤ 2 are due to the presence of the iron-sulfur cluster as apo␤ 2 , in the presence of ␣ 2 , or ␣ 2 alone was totally inactive. ␤ 2 alone was by itself able to support the reaction, but less efficiently (Fig. 4). It would be interesting to understand the molecular basis for the potentiating effect of ␣ 2 on the reducing properties of the iron-sulfur center, since it may provide new insights into how the chemical reactivity of ironsulfur centers can be modulated.
Reduction of AdoMet is a new biological function for an iron-sulfur protein, since there is no precedent for an ironsulfur center that catalyzes such a redox reaction. The list of functions for this class of metal clusters has expanded recently and includes electron transport (photosynthesis, respiration, etc.); catalysis, but only in nonredox reactions such as dehydration reactions (e.g. aconitase and related systems); regulation of transcription (28); stabilization of protein structure (29); and stabilization of reactive intermediates (30). We now show that iron-sulfur proteins may also be involved in catalysis of redox reactions. Such a metal center might be required especially when the electron acceptor has to be reduced at a very low potential, as is the case with AdoMet. Sulfonium compounds are known to be very stable molecules that can be reduced at a redox potential of about Ϫ1 V in organic solvents, as shown by electrochemistry (31).
Conclusion-There are still a number of questions to solve and a number of unexpected observations to interpret, in par-ticular the 3:1 stoichiometry for methionine formation that was observed for both ␤ 2 and ␣ 2 ␤ 2 . We believe that many answers and the understanding of the precise mechanism of the reaction will come from further investigation of the iron-sulfur cluster of the anaerobic ribonucleotide reductase. In fact, from many spectroscopic and functional observations, it seems to be unusual. It is also important to investigate the role of DTT, which appears to be a key component of the reaction. DTT serves to stabilize the iron-sulfur center but probably also has functional roles that remain to be identified. Nevertheless, this work provides the first demonstration that cleavage of AdoMet, release of methionine, and generation of the glycyl radical derive directly from the electron transfer from the reduced iron-sulfur cluster of the enzyme to AdoMet. This indicates that iron-sulfur clusters can also have redox catalytic functions, a concept that is currently emerging (32).