The Free Radical of the Anaerobic Ribonucleotide Reductase from Escherichia coli Is at Glycine 681*

The anaerobic ribonucleoside triphosphate reductase of Escherichia coli is an iron-sulfur protein carrying an oxygen-sensitive organic radical, which is essential for catalysis. The radical was tentatively proposed to be on glycine 681, based on a comparison with the glycyl rad- ical-containing enzyme pyruvate formate-lyase. By EPR spectroscopy of selectively 2 H- and 13 C-labeled anaerobic ribonucleotide reductase, the radical was now un- ambiguously assigned to carbon-2 of a glycine residue. The large 1 H hyperfine splitting (1.4 millitesla) was assigned to the (cid:97) -proton. Site-directed mutagenesis was used to change glycine 681 into an alanine residue. In separate experiments, the two adjacent residues, cys-teine 680 and tyrosine 682, were changed into serine and phenylalanine, respectively. All enzyme activity. Our results identify glycine as the stable free radical site in E. coli anaerobic ribonucleotide reductase.

Ribonucleotide reductases are essential enzymes in all living organisms. They catalyze the reduction of ribonucleotides to the corresponding deoxyribonucleotides by radical chemistry. Three different reductases have been found in Escherichia coli, one functional during aerobic growth (for a recent review, see Sjöberg (1994)) and one during anaerobic growth (Reichard, 1993b), whereas the third one is cryptic during normal growth (Jordan et al., 1994). Both aerobic and anaerobic reductases are free radical-containing enzymes, but they utilize different mechanisms for radical generation (Reichard, 1993a;Sjöberg, 1995).
The anaerobic ribonucleotide reductase, which is encoded by the nrdD gene (Sun et al., 1993), is a homodimer of 160 kDa. This active protein contains a stable organic free radical and a poorly defined iron-sulfur cluster (Mulliez et al., 1993). The radical is introduced into an inactive form of the reductase in a reaction requiring a 17.5-kDa protein (Sun et al., 1995), Sadenosylmethionine (AdoMet), NADPH (Eliasson et al., 1990;Harder et al., 1992), flavodoxin, and ferredoxin (flavodoxin) NADP ϩ oxidoreductase (abbreviated flavodoxin reductase below) (Bianchi et al., 1993a(Bianchi et al., , 1993b. The organic radical is oxygen-sensitive, and exposure of the radical-containing enzyme to air leads to truncation at Gly-681 (King and Reichard, 1995), accompanied by inactivation (Sun et al., 1993). There is a striking amino acid sequence similarity between a stretch of 5 residues comprising glycine 681 in the anaerobic reductase (Sun et al., 1993) or glycine 580 in the corresponding bacteriophage T4 anaerobic ribonucleotide reductase (Young et al., 1994) and the sequence surrounding glycine 734 at the active site of the glycyl radical-containing enzyme pyruvate formatelyase (PFL) 1 (Wagner et al., 1992). Formation of an oxygensensitive radical at Gly-734 in PFL and oxygen-dependent truncation of PFL at this position are analogous reactions to the ones described above for the anaerobic reductase. Glycine 681 in the anaerobic reductase was therefore proposed to be the position of the free radical in this system (Sun et al., 1993). In agreement with this proposition, the EPR spectrum of the radical of the anaerobic reductase shares some common features with that of the glycyl radical of PFL, in particular the large dominant doublet hyperfine splitting (Mulliez et al., 1993).
In this study, the chemical nature of the free radical of the anaerobic ribonucleotide reductase has been investigated by isotopic substitution experiments. The effects of isotopically labeling the glycines of the enzyme on the hyperfine structure of the characteristic EPR spectrum were consistent with the organic radical being on a glycine residue. Moreover we used site-directed mutagenesis to identify the location of the glycyl radical. In addition to the mutation G681A, the adjacent mutations C680S and Y682F were also constructed. Mutant proteins were characterized with regard to the presence of ironsulfur cluster, glycyl radical, and enzyme activity. Taken together our results demonstrate that the free radical of anaerobic ribonucleotide reductase is located on glycine 681. In a similar study, the stable glycyl radical of the T4 anaerobic * This work was supported by grants from the Swedish Cancer Society (to B.-M. S.), the Swedish Natural Science Research Council (to A. G.), Centre National de la Recherche Scientifique (to M. F.), European Union (to P. R. and M. F.), and the Swedish Medical Research Council (to P. R.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. *

EXPERIMENTAL PROCEDURES
Materials-E. coli strain JM109(DE3) was obtained from Promega. Restriction enzymes were from New England Biolabs and Promega. The PCR kits were from Perkin-Elmer. The PCR purification spin kit and plasmid preparation kit were from Qiagen. The gene clean kit was from BIO101. The Taq DyeDeoxy terminator cycle sequencing kit was from Applied Biosystems. The oligonucleotides used for mutagenesis were synthesized by Scandinavian Gene Synthesis (Köping, Sweden). The Muta-Gene phagemid in vitro mutagenesis kit was from Bio-Rad.
Construction of Plasmids-Plasmid pDA carries the nrdD operon under T7 promoter control. It was constructed by insertion of a 673-base pair MunI-KpnI fragment, containing the nrdG gene, from pPX41 into MunI-KpnI-linearized pRSS plasmid, containing the reductase gene, nrdD; pPX41, pREH, and pRSS were constructed in previous work (Sun et al., 1995). Plasmid pAE183 was described by Eliasson et al. (1994).
Overproduction of Anaerobic Ribonucleotide Reductase-E. coli strain JM109(DE3) containing wild type or mutant plasmid pDA was grown anaerobically at 37°C in LB medium with 2 g/liter glucose and 200 g/ml ampicillin. When the A 640 of the culture was 0.3, isopropyl-1-thio-␤-D-galactopyranoside was added to a final concentration of 0.2 mM. Cells were harvested by centrifugation 2 h after induction and stored in liquid nitrogen. Isotopic labeling of the reductase was achieved by anaerobically growing E. coli strain JM109(DE3) cells containing plasmid pREH at 37°C in minimal medium supplemented with 6 g/liter glucose, 200 g/ml ampicillin, 7% LB medium, 2 mM MgSO 4 , 4 g/ml thiamin, and 6 mM labeled glycine. After 16 h of growth, cells were harvested and centrifuged at 5000 rpm for 20 min and stored in liquid nitrogen. A small amount of LB medium was required to accelerate the growth rate. This additive provided a negligible amount of unlabeled glycine. A very high level of protein expression was obtained, as shown by SDS-polyacrylamide gel electrophoresis, but the majority of the recombinant protein produced by cells grown in synthetic medium resided in the insoluble portion of the cell extract, presumably as inclusion bodies, from which soluble reductase was extracted.
Purification of the Enzyme and Activity Assays-The anaerobic ribonucleotide reductase was purified as described in Eliasson et al. (1992), with modifications according to Mulliez et al. (1995). Prior to enzyme activity measurements the reductase was preincubated anaerobically at room temperature for 60 min with the components required for the activation step. Typically, the 35-l premixture contained 1-5 g of anaerobic ribonucleotide reductase, 0.4 mM AdoMet, 1 mM NADPH, 5 mM dithiothreitol, 30 mM KCl, 30 mM Tris (pH 8.0), 5 mM sodium formate, 2 g of flavodoxin reductase (Bianchi et al., 1993a), and 1 g of flavodoxin (Bianchi et al., 1993b). The actual reduction reaction was started by adding 15 l of anaerobic substrate mixture (3.3 mM [ 3 H]CTP, 3.3 mM ATP, and 33 mM MgCl 2 ) and was stopped after 20 min by exposure of the reaction mixture to oxygen. The amount of dCTP formed was determined as described previously . One unit of activity corresponds to the reduction of 1 nmol of CTP/min at 25°C. Specific activity is defined as units per mg of protein.
Enzyme Activation for EPR Experiments-In order to measure the glycyl radical content, purified wild type, labeled, or mutant proteins were activated anaerobically. All the components were degassed with argon for 1 h prior to use. The activation was carried out by mixing degassed components in EPR tubes and incubating the mixtures at room temperature for 1 h. Normally, 200 l of activation mixture contained 25-30 M reductase, 1.5 mM AdoMet, 5 mM dithiothreitol, 10 mM sodium formate, 80 g of flavodoxin reductase, 20 g of flavodoxin, 20 g of NrdG protein (Sun et al., 1995), 5 mM NADPH, 45 mM Tris-HCl, pH 7.5, 45 mM KCl. The contents of the EPR tubes were frozen anaerobically in liquid nitrogen after activation. The specific activities of isotopically labeled enzyme preparations were significantly lower than that of the wild type reductase (500 and 200 nmol/mg of protein/min, respectively, compared with 1000 nmol/mg of protein/min for the wild type enzyme), in all probability because of their origin from inclusion bodies.
EPR Spectroscopy-First derivative spectra were recorded at 77 or 100 K with a Bruker ESP 300E spectrometer operating at X band and using the Bruker variable temperature device ER4411 VT or a Bruker ESP 300 spectrometer equipped with an Oxford Instrument helium flow cryostat. Samples of 0.2 ml were introduced into quartz tubes under argon and sealed with a rubber septum. In order to apply EPR spectroscopy to whole cells, E. coli cell cultures were centrifuged, and cell pellets were transferred to an anaerobic box and then resuspended in culture medium to be introduced into an EPR tube. After 20 min of centrifugation at 1000 rpm, the supernatant was eliminated, and packed cells were frozen in liquid nitrogen. Quantitation of the glycyl free radical was based on comparisons of the double integral of the EPR spectra with that of a standard 1 mM Cu 2ϩ (in 1.2 M NaClO 4 , pH 1.8) solution. This was used for quantitation of a standard anaerobic ribonucleotide reductase glycyl radical-containing sample. The standard reductase sample was kept at 77 K and used as a secondary standard for quantitation of other samples with an identical EPR line shape.

RESULTS AND DISCUSSION
The E. coli anaerobic ribonucleotide reductase gene, nrdD, has previously been cloned in different plasmids for overproduction of the enzyme. The downstream nrdG gene encodes a 17.5-kDa protein, which is essential for activation of the anaerobic ribonucleotide reductase (Sun et al., 1995). Only the enzyme purified from constructs containing both nrdD and nrdG genes is active, and plasmid pREH (Sun et al., 1995), used for the isotopic labeling experiments, is one such plasmid. In this study, we also constructed plasmid pDA containing both nrdD and nrdG genes. Bacteria carrying plasmid pDA with a wild type or mutant nrdD gene gave good overexpression; extracts contained 6 -12% anaerobic ribonucleotide reductase (data not shown).
Evidence through Isotopically Labeled Enzyme That Anaerobic Ribonucleotide Reductase Contains a Glycyl Radical-From cells carrying pREH and grown anaerobically in minimal medium in the presence of isotopically labeled glycine we prepared [ 13 C]-or [ 2 H]glycine-substituted enzyme. Fig. 1 shows the EPR spectra of the native and the two isotopically labeled reductases. The main feature of the EPR spectrum of the unlabeled enzyme (spectrum A) is a large doublet splitting (14 -15 G) centered at g ϭ 2.0033. This splitting was originally assigned to hyperfine coupling of the unpaired electron spin to a hydrogen nucleus (Mulliez et al., 1993). The effects of labeling the glycine residues of the reductase on the EPR signal now directly prove that the radical is centered at C-2 of a glycine residue. The EPR features of the [2-13 C]glycine-containing enzyme clearly reflect strong coupling of the unpaired electron to the 13 C (I ϭ 1/2) nucleus (spectrum B). Moreover, the presence of a singlet signal in the [ 2 H]glycine-substituted enzyme unambiguously demonstrates that the 14 G doublet splitting in the wild type enzyme originates from the hyperfine coupling of the unpaired electron spin to the ␣-hydrogen of the glycyl radical center (spectrum C). Spectrum B proved to be difficult to simulate, and further studies are required to find out the correct g and A tensors describing the radical. However, from the outermost lines of the signal, it is possible to obtain a value for the hyperfine 13 C coupling constant A z in the 46 -50 G range. Moreover, the A x and A y components of this tensor seem to be in the 0 -5 G range. Such small values are in good agreement with 13 C A x,y constants obtained for radicals of related model compounds (Sinclair and Codella, 1973) or for other -type radicals. As a consequence, A iso , the isotropic 13 C coupling constant should be in the 15-21 G range.
The glycyl radical of PFL has the same coupling characteristics with a large 1 H hyperfine splitting (15 G) and an estimated A iso hyperfine coupling constant also in the range of 16 -21 G. A theoretical model study of a dipeptide analog of a glycine radical demonstrated that only planar or nearly planar conformations are energetically accessible due to -electron delocalization. The hyperfine couplings computed for these conformations (Barone et al., 1995) are in excellent agreement with the experimental values for both glycyl radicals. On the other hand, it is possible to distinguish between the glycyl radicals of the anaerobic reductase and PFL. First, over a 24-h incubation period (Mulliez et al., 1993) 3 the reductase radical does not exchange its H␣ atom, responsible for the doublet splitting, with the solvent, whereas the PFL radical exchanges rapidly under these conditions. Second, the EPR spectrum of the PFL radical is more complex, with partially resolved subdoublet splitting, arising from two nonexchangeable protons (Wagner et al., 1992).
The EPR signal of the glycyl free radical also can be recorded directly in bacterial pellets from E. coli JM109(DE3)/pREH cells induced for overexpressing the reductase. The EPR spectra of cells grown in minimal medium in the presence of normal or labeled glycine gave EPR spectra identical to those of the corresponding pure enzyme. This shows that the reductase exists in the radical form within anaerobic cells and that its concentration is such that the endogenous PFL signal is not detectable. On the other hand, in plasmid-free cells only the background glycyl radical signal typical of PFL (Knappe and Sawers, 1990) was observed.
Evidence by Site-directed Mutagenesis That the Essential Radical Is Located on Glycine 681-To map the site of the glycyl radical we mutated Gly-681 in the nrdD gene and used mutations in the adjacent residues, Cys-680 and Tyr-682, as controls. All three mutant proteins (C680S, G681A, and Y682F) behaved like wild type anaerobic ribonucleotide reductase during the purification. An important observation was that all three mutant proteins were retained by dATP-Sepharose chromatography, indicating the presence of intact allosteric binding sites. Moreover, the purified mutant proteins migrated like wild type protein during SDS-polyacrylamide gel electrophoresis (see below), indicating that all preparations contained fulllength polypeptides.
The iron-sulfur center of the anaerobic ribonucleotide reductase is essential for enzyme activity, which was found to correlate linearly with the iron content of the protein (Mulliez et al., 1993). The iron content of the protein preparations purified from the mutants varied between 1.1 and 1.3 atoms of iron/ protein dimer (Table I). These values are similar to the iron content of wild type protein preparations from JM109(DE3)/ pDA cells, which is typically between 1.2 and 1.5 iron atoms. The presence of intact iron-sulfur centers in the preparations from the mutants was verified by UV-visible spectroscopy. Fig. 2 shows that mutant and wild type preparations display a spectrum between 300 and 600 nm, characteristic of iron-sulfur proteins. The small difference between the spectra in the region between 300 and 350 nm is not considered to be significant. The absorption coefficients at 420 nm for the mutant proteins are similar and compare well with those published for the active enzyme prepared from plasmid pEH10 (Mulliez et al., 1993).
If the catalytically essential radical of anaerobic ribonucle-3 E. Mulliez, unpublished data.
FIG. 1. EPR spectra of wild type and isotopically labeled anaerobic ribonucleotide reductase. X-band EPR spectra of the unlabeled wild type enzyme (Spectrum A), [2-13 C]glycine-labeled enzyme (Spectrum B), and [ 2 H]glycine-labeled enzyme (Spectrum C) are shown. The enzyme preparations were from JM109(DE3)/pREH cells grown in the presence of 6 mM unlabeled and labeled glycine, respectively. Spectra were recorded at 100 K; microwave power, 50 microwatts; and modulation amplitude, 1.6 G. otide reductase is located at glycine 681, we would expect the mutant G681A to be devoid of both radical and enzyme activity. Fig. 3 shows the EPR signals recorded from the activated proteins. With the exception of the G681A, which completely lacked a radical EPR signal, all the other proteins showed the typical glycyl radical signal. We also quantified the amount of radical (Table I). In the wild type protein the radical content is between 0.43 and 0.49 radical/reductase enzyme. The mutant protein C680S has about 20% radical and Y682F has about 12% radical/reductase, compared with the wild type protein (Table  I). Since all purified enzymes contained substantial amounts of truncated polypeptides (see below), the actual radical content per full-length protein might be significantly higher. The EPR signal of the glycyl radical also can be recorded directly in bacterial pellets from induced JM109(DE3)/pDA cells, thereby avoiding extensive purification protocols, which may affect the recovery of the radical. Whereas cells carrying the wild type plasmid gave the glycyl radical signal typical of FIG. 3. EPR spectra of wild type and mutant anaerobic ribonucleotide reductases. The EPR spectra were recorded from activated proteins at 10 K under nonsaturating microwave power conditions. The protein concentration of the wild type is 31.5 M. C680S is 27.8 M, Y682F is 26.6 M, and G681A is 29.8 M. The EPR spectra correspond to radical concentrations of 15.5, 4, 1.8, and 0 M, respectively. Spectra were recorded with microwave power of 20 microwatts; modulation amplitude is 3.0 G. Relative gains x1 and x4 are indicated. anaerobic ribonucleotide reductase, cells carrying the G681A plasmid only showed the glycyl radical signal from PFL (data not shown), as expected if the induced mutant anaerobic reductase lacks a glycyl radical. In this case, washing of the cell pellet with anaerobic D 2 O resulted in reduced line width of the EPR spectrum (data not shown), in agreement with the solvent exchangeability of the proton responsible for the dominating doublet hyperfine splitting of the PFL radical.
Enzyme activities of the pure wild type and mutant reductases are shown in Table I. The G681A mutant protein completely lacked enzyme activity, whereas the C680S protein had about 14% activity and the Y682F protein about 7% as compared with the wild type protein. Taken together, these results clearly show that the G681A mutant protein lacks both radical and enzyme activity, whereas the C680S and Y682F mutant proteins still retain some radical and some enzyme activity. The difference between radical content and corresponding enzyme activity of the mutants C680S and Y682F might be caused by experimental variation or by conformational changes introduced by the mutations that influence the enzyme activity and the generation of radical differently.
What are the functions of Cys-680 and Tyr-682? Our results indicate that their presence influences the surroundings of the radical-harboring Gly-681 residue but that neither is essential for formation of the glycyl radical or for activity of the anaerobic reductase. The mutation C680S generates the sequence RVSGY between residues 678 and 682, which in fact is identical to the corresponding sequence surrounding the glycyl radical of PFL (Frey et al., 1994;Sun et al., 1993). In reciprocal experiments in the PFL system, substitution of Ser by Cys in the active site oligopeptide RVSGYLG led to loss of substrate efficiency (Frey et al., 1994), whereas substitution of Ser by Ala in PFL proper gave an enzyme with 34% residual activity (Frey et al., 1994). 4 In addition, substitution of Tyr by Phe in the PFL active site oligopeptide only led to partial loss of substrate efficiency. Thus, mutations of the residues adjacent to the essential glycine residue are tolerated to some extent both by the anaerobic reductase and by PFL.
Conclusion-Our results clearly demonstrate that the presence of Gly-681 is essential for generation of a glycyl radical and concomitant enzyme activity in the anaerobic ribonucleotide reductase, whereas mutations of the adjacent residues result in proteins that still can harbor a glycyl radical and retain some enzymatic activity. Together with the isotope substitution experiments these results demonstrate that the location of the glycyl radical-harboring residue in E. coli anaerobic ribonucleotide reductase is at position 681.