Escherichia coli (cid:1) -Glutamylcysteine Synthetase TWO ACTIVE SITE METAL IONS AFFECT SUBSTRATE AND INHIBITOR BINDING*

(cid:1) -Glutamylcysteine synthetase ( (cid:1) -GCS, glutamate-cys-teine ligase), which catalyzes the first and rate-limiting step in glutathione biosynthesis, is present in many pro-karyotes and in virtually all eukaryotes. Although all eukaryotic (cid:1) -GCS isoforms examined to date are rapidly inhibited by buthionine sulfoximine (BSO), most reports indicate that bacterial (cid:1) -GCS is resistant to BSO. We have confirmed the latter finding with Escherichia coli (cid:1) -GCS under standard assay conditions, showing both decreased initial binding affinity for BSO and a reduced rate of BSO-mediated inactivation compared with mammalian isoforms. We also find that substitution of Mn 2 (cid:2) for Mg 2 (cid:2) in assay mixtures increases both the initial binding affinity of BSO and the rate at which BSO causes mechanism-based inactivation. Similarly, the specificity of E. coli (cid:1) -GCS for its amino acid substrates is broadened in the presence of Mn 2 (cid:2) , and the rate of reaction for some very poor substrates is improved. These results suggest that divalent metal ions have a role in amino acid binding to E. coli (cid:1) -GCS. Electron paramagnetic resonance (EPR) studies carried out with Mn 2 (cid:2) show that E. coli (cid:1) -GCS binds two divalent metal ions; K with 30 ml of 50 m M MOPS buffer, pH 7.4, containing 5 m M L -glutamate, 5 m M MgCl 2 , 0.2 M NaCl, and 1 m M ATP. Active fractions were concentrated to (cid:3) 5 ml using Centriprep concen- trators (Millipore, Bedford, MA) and dialyzed thrice against 8 liters of 20 m M Na (cid:1) HEPES buffer, pH 7.8, containing 1 m M EDTA and then thrice against 8 liters of the same buffer without EDTA. We found that this extensive dialysis procedure was necessary to reduce bound metal ions to acceptable levels (typical analysis: 200 (cid:5) M E. coli (cid:1) -GCS pro-tomers contained 1 (cid:5) M Mg 2 (cid:1) , 0.33 (cid:5) M Mn 2 (cid:1) , and 5 (cid:5) M Cu 2 (cid:1) as deter- mined by flame ionization atomic absorption). Purified and dialyzed E. coli (cid:1) -GCS was concentrated to 25–40 mg/ml using Centriprep concen- trators, and glycerol was added to 25% (v/v). The enzyme was stable for a period of several months when stored in 25% glycerol at (cid:4) 20 °C. Human (cid:1) -GCS was expressed in E. coli and purified as described previously (31). By SDS-PAGE the final preparation was homogeneous and showed the expected heavy and light subunits. The specific activity was 1150 units/mg. Protein Determination— For E. coli (cid:1) -GCS, protein was estimated during the purification procedure by the Bradford assay using bovine serum albumin as the standard (32). However, the protein contents of the final preparations were quantitated on the basis of their A 280 ( (cid:6) (cid:6) 0.83 for 1 mg/ml E. coli (cid:1) -GCS). The (cid:6) value was determined by sub-jecting a pure sample of E. coli (cid:1) -GCS to hydrolysis and quantitative amino 10- (cid:5) l portion of the supernatant was fractionated by reverse-phase high pressure liquid chromatography using a Cu- L -proline elution buffer flowing at 0.5 ml/min and post-column o- phthalaldehyde detection as described by E. Gil-Av et al . (35). In all cases, a new peak corresponding to (cid:1) -glutamylamino acid appeared, and substrate peaks corresponding to L -glutamate and the amino acid tested were diminished in size. EPR Studies— X-band (8.9–9.4 GHz) EPR spectra were recorded on a Varian E109 Century Series spectrometer with a Varian TE102 cavity (Varian, Palo Alto, CA). Samples were placed in a flat cell for studies at room temperature and in a 4-mm OD quartz EPR tube in a finger Dewar filled with liquid N 2 for studies at 77 K. S-band spectra (3.4 GHz) were obtained using a spectrometer with a loop-gap resonator cavity and a low frequency microwave bridge built at the National Biomedical ESR Center at the Medical College of Wisconsin (36). Samples for S-band studies were cooled with N 2 passed through an exchange coil immersed in liquid N 2 . S-band EPR samples were placed in quartz tubes of 4.0-mm OD. Microwave frequencies were measured with a frequency counter (EIP model 548), and the field was calibrated with a magnetometer (Rawson-Lush Instrument Co., Acton, MA).

␥-Glutamylcysteine synthetase (␥-GCS, glutamate-cysteine ligase), which catalyzes the first and rate-limiting step in glutathione biosynthesis, is present in many prokaryotes and in virtually all eukaryotes. Although all eukaryotic ␥-GCS isoforms examined to date are rapidly inhibited by buthionine sulfoximine (BSO), most reports indicate that bacterial ␥-GCS is resistant to BSO. We have confirmed the latter finding with Escherichia coli ␥-GCS under standard assay conditions, showing both decreased initial binding affinity for BSO and a reduced rate of BSO-mediated inactivation compared with mammalian isoforms. We also find that substitution of Mn 2؉ for Mg 2؉ in assay mixtures increases both the initial binding affinity of BSO and the rate at which BSO causes mechanism-based inactivation. Similarly, the specificity of E. coli ␥-GCS for its amino acid substrates is broadened in the presence of Mn 2؉ , and the rate of reaction for some very poor substrates is improved. These results suggest that divalent metal ions have a role in amino acid binding to E. coli ␥-GCS. Electron paramagnetic resonance (EPR) studies carried out with Mn 2؉ show that E. coli ␥-GCS binds two divalent metal ions; K d values for Mn 2؉ are 1.1 M and 82 M, respectively. Binding of L-glutamate or L-BSO to the two Mn 2؉ / ␥-GCS species produces additional upfield and downfield X-band EPR hyperfine lines at 45 G intervals, a result indicating that the two Mn 2؉ are spin-coupled and thus apparently separated by 5 Å or less in the active site. Additional EPR studies in which Cu 2؉ replaced Mg 2؉ or Mn 2؉ suggest that Cu 2؉ is bound by one N and three O ligands in the ␥-GCS active site. The results are discussed in the context of the catalytic mechanism of ␥-GCS and its relationship to the more fully characterized glutamine synthetase reaction.
Glutathione (GSH, ␥-glutamylcysteinylglycine) is present in the cells of virtually all eukaryotes ranging from protozoa to plants and mammals, and its physiological role as an enzyme cofactor, antioxidant, and sacrificial nucleophile able to detoxify reactive electrophiles is well established in those systems (1)(2)(3). Glutathione is more narrowly distributed among bacteria (4), and its function in bacteria is less well defined. For example, Escherichia coli contain concentrations of GSH com-parable with those in mammalian cells, but they apparently lack the GSH peroxidase activity necessary for GSH to effectively control levels of H 2 O 2 and organic peroxides (5,6). Nevertheless, several studies comparing wild type bacteria to mutants that either lack or overexpress the enzymes of GSH biosynthesis suggest that GSH does have a role in bacterial defenses against many toxic insults including exposure to thiolreactive agents (7)(8)(9)(10)(11), methylglyoxal (7,8,12,13), certain oxidative (7,8,11) and nitrosative (14) stress molecules, x-rays (15,16), and some antibiotics (7). In contrast, other studies suggest bacterial GSH is not important in controlling toxicity from H 2 O 2 (17), tert-butyl hydroperoxide (18), redox-cycling agents such as paraquat (18), x-rays (7,17) or singlet O 2 (19). In several cases, the discrepancies among these reports remain unresolved in the literature (20).
Glutathione is synthesized in both eukaryotes and prokaryotes by the sequential action of ␥-glutamylcysteine synthetase (␥-GCS, 1 Equation 1) and GSH synthetase (Equation 2) (1,2). In all cases examined, ␥-GCS is rate-limiting and is feedback-inhibited by GSH. In protozoa and mammals, potent inhibitors of ␥-GCS such as buthionine sulfoximine (BSO) have been used to block GSH synthesis and promote GSH depletion (21)(22)(23). Cells depleted of GSH are more susceptible to the toxic effects of radiation (16,24), many cancer chemotherapy drugs (25), and internally generated oxidants (22), and the increased vulnerability of cells pharmacologically depleted of GSH has been exploited therapeutically (23,25). L-glutamate ϩ L-cysteine ϩ ATP 3 L-␥-glutamyl-L-cysteine ϩ ADP ϩ P i (Eq. 1) L-␥-glutamyl-L-cysteine ϩ glycine ϩ ATP 3 GSH ϩ ADP ϩ P i (Eq. 2) We are interested in the possibility that pharmacological depletion of bacterial GSH might also be therapeutically useful and have therefore examined the ability of L-buthionine-Ssulfoximine (L-S-BSO), the active diastereomer of BSO (26), to inhibit E. coli ␥-GCS. In confirmation of most (18), but not all (27), literature reports, we find that BSO is a relatively poor inhibitor of bacterial ␥-GCS, particularly in the presence of physiological levels of L-glutamate. We have therefore investigated the interactions of E. coli ␥-GCS with L-S-BSO and with its amino acid substrates and various substrate analogs in an effort to elucidate binding interactions that might be exploited in designing novel, more potent inhibitors. We find that substrate specificity depends on the divalent metal ion present (i.e. Mg 2ϩ or Mn 2ϩ ), a result suggesting that metal ions have a role in the binding of amino acid substrates and inhibitors in addition to their expected role in ATP binding. Electron paramagnetic resonance (EPR) studies carried out with Mn 2ϩ show that E. coli ␥-GCS contains two active site divalent metal ions. Binding of L-glutamate or L-S-BSO induces additional 45 G coupling within the EPR spectra indicative of Mn 2ϩ -Mn 2ϩ coupling through spin exchange. These results are discussed in terms of the mechanism of action of E. coli ␥-GCS.
Methods of Expression and Purification of E. coli and Human ␥-GCS-JM109 E. coli cells containing the wild type E. coli coding sequence in a pGMS300 -10 plasmid were obtained from K. Murata (Research Institute for Food Science, Kyoto University, Uji Kyoto, Japan). The ␥-GCS coding region was removed and inserted into a pKK223-3 plasmid (Amersham Biosciences, Inc.) using standard techniques, and the resulting plasmid was transfected into JM105 E. coli (Amersham Biosciences, Inc.). DNA sequence analysis confirmed that the ␥-GCS coding region of the plasmid matched that reported for wild type E. coli (SWISS-PROT accession number P 06980).
In a typical preparation, E. coli were grown at 37°C in seven Fernbach flasks each containing 1.5 liters of 2ϫ YT media (24 g of tryptone, 15 g of yeast extract, 7.5 g of NaCl, and 30 ml of glycerol per 1.5 liters) and 100 mg/L of ampicillin; the inoculum was 15 ml of a similar culture grown the night before. Cells were grown 3-4 h until A 600 ϳ0.8, and then isopropyl ␤-D-thiogalactopyranoside was added to a final concentration of 1 mM. After an additional 3-4 h, the cells were harvested by centrifugation and stored at Ϫ20°C; the yield was ϳ12 g cells/L. E. coli ␥-GCS was isolated using a protocol modified slightly from those used previously for E. coli (27,29), rat (30), and human (31) ␥-GCS. Thawed cells (10 -15 g) were suspended in ϳ50 ml of isolation buffer (50 mM Tris HCl, pH 7.4, 5 mM L-glutamate and 5 mM MgCl 2 ) and were broken with a French press. The crude homogenate was clarified by centrifugation, and the supernatant was fractionated with ammonium sulfate. Proteins precipitating between 40 and 80% ammonium sulfate were collected, dissolved in ϳ50 ml of isolation buffer, and dialyzed overnight against 16 liters of the same buffer. The protein solution was then applied to a 2.5 ϫ 20-cm column of Whatman DE-52 anion exchange resin equilibrated with isolation buffer. After washing with ϳ200 ml of isolation buffer, ␥-GCS was eluted with a linear gradient established between 400 ml of isolation buffer and 400 ml of isolation buffer containing 0.2 M NaCl. Fractions containing ␥-GCS activity (ϳ0.1 M NaCl) were pooled, and ␥-GCS was precipitated by addition of ammonium sulfate to 80%. The dissolved pellet (Ͻ10 ml) was applied to a 1.6 ϫ 85-cm Superdex 200 fast protein liquid chromatography column that was equilibrated and eluted with 50 mM MOPS buffer, pH 7.0, containing 5 mM L-glutamate and 1 mM dithiothreitol. Active fractions were pooled, made 5 mM in MnCl 2 , and applied to a 1 ϫ 8-cm column of ATP affinity resin (C8-linked; 9 atom spacer; Sigma A2767) that was equilibrated with 50 mM MOPS buffer, pH 7.4, containing 5 mM L-glutamate and 5 mM MnCl 2 . The column was washed successively with 25 ml of equilibration buffer, 15 ml of equilibration buffer containing 0.7 M NaCl and 10% glycerol, and 20 ml of 50 mM MOPS buffer, pH 7.4, containing 5 mM L-glutamate and 5 mM MgCl 2 . The enzyme was then eluted with 30 ml of 50 mM MOPS buffer, pH 7.4, containing 5 mM L-glutamate, 5 mM MgCl 2 , 0.2 M NaCl, and 1 mM ATP. Active fractions were concentrated to ϳ5 ml using Centriprep concentrators (Millipore, Bedford, MA) and dialyzed thrice against 8 liters of 20 mM Na ϩ HEPES buffer, pH 7.8, containing 1 mM EDTA and then thrice against 8 liters of the same buffer without EDTA. We found that this extensive dialysis procedure was necessary to reduce bound metal ions to acceptable levels (typical analysis: 200 M E. coli ␥-GCS pro-tomers contained 1 M Mg 2ϩ , 0.33 M Mn 2ϩ , and 5 M Cu 2ϩ as determined by flame ionization atomic absorption). Purified and dialyzed E. coli ␥-GCS was concentrated to 25-40 mg/ml using Centriprep concentrators, and glycerol was added to 25% (v/v). The enzyme was stable for a period of several months when stored in 25% glycerol at Ϫ20°C.
Human ␥-GCS was expressed in E. coli and purified as described previously (31). By SDS-PAGE the final preparation was homogeneous and showed the expected heavy and light subunits. The specific activity was 1150 units/mg.
Protein Determination-For E. coli ␥-GCS, protein was estimated during the purification procedure by the Bradford assay using bovine serum albumin as the standard (32). However, the protein contents of the final preparations were quantitated on the basis of their A 280 (⑀ ϭ 0.83 for 1 mg/ml E. coli ␥-GCS). The ⑀ value was determined by subjecting a pure sample of E. coli ␥-GCS to hydrolysis and quantitative amino acid analysis; the concentration of ␥-GCS in that sample was then calculated from the known amino acid composition of the enzyme. Human ␥-GCS was quantitated using the Bradford assay. Concentrations of enzyme are expressed on the basis of the basic catalytic unit: monomers for E. coli ␥-GCS (M r ϭ 58,269) and heterodimer for human ␥-GCS (M r ϭ 103,500).
Assay of ␥-GCS-Activity was routinely determined on the basis of ADP formation using a pyruvate kinase-lactate dehydrogenase-coupled assay (30). Reaction mixtures were equilibrated to 37°C and, in a final volume of 1.0 ml, contained 150 mM Tris HCl buffer, pH 8.2, 25 mM L-glutamate, 25 mM L-␣-aminobutyrate, 10 mM ATP, 10 mM phospho(enol)pyruvate, 30 mM MgCl 2 , 100 mM KCl, 0.2 mM EDTA, 0.42 mM NADH, 7 IU pyruvate kinase, 12 IU lactate dehydrogenase, and ␥-GCS, which was added last to start the reaction. Oxidation of NADH was monitored at 340 nm (⑀ ϭ 6.2 mM Ϫ1 ) and was assumed to equal the rate of ADP formation. Rates were corrected for ␥-GCS-catalyzed ADP formation in the absence of L-glutamate or L-glutamate analog (typically ϳ4 mol/hr/mg of protein, which is 0.14% of the rate with the standard complete reaction mixture). In some studies, 10 mM MnCl 2 or 10 mM CoCl 2 replaced MgCl 2 . Pyruvate kinase is active with both Mn 2ϩ and Co 2ϩ (33) and was shown not to be rate-limiting in those studies. In studies of substrate specificity, the concentrations of the non-varied substrates were ϳ10 times their K m values (see Table II legend). In all assays based on ADP formation or dipeptide formation (see below) small amounts of glycerol were carried into the assay mixtures with ␥-GCS; final concentrations of glycerol were Ͻ 0.1% and had no effect on the observed rates.
Dipeptide formation was quantitated on the basis of L-␣-[ 14 C]aminobutyrate incorporation into dipeptide (28)  For studies with alternative metal ions, Mg 2ϩ was replaced with 10 mM of the alternative metal ion and phospho(enol)pyruvate and pyruvate kinase were omitted from the reaction mixture. Background readings were determined using aliquots removed immediately prior to addition of ␥-GCS. To determine dipeptide formation, 100-l portions were removed at 5, 10, or 15 min intervals and diluted into 1.0 ml of ice-cold 20 mM acetic acid, and 1.0 ml of the resulting solution was applied to small (0.5 ϫ 8-cm) columns of Dowex 1 ϫ 8 (acetate form). The columns were washed free of unreacted L-␣-[ 14 C]aminobutyrate with 7 ml of 20 mM acetic acid, and dipeptide was then eluted with 5 ml of 1.5 M ammonium acetate. Radioactivity in a 2-ml portion of that eluant was quantitated by liquid scintillation counting, and dipeptide formation was calculated based on the specific activity of the L-␣-[ 14 C]aminobutyrate present in the reaction mixture. To determine ADP formation, 20-l portions of the same reaction mixtures were added directly to cuvettes containing 1.0 ml of a reaction mixture similar to that used in the standard ADP formation assay except that ATP was omitted. The amount of ADP present in the reaction mixture aliquot was calculated from the lactate dehydrogenase-mediated NADH oxidation observed within 5 min and was assumed to equal the amount of ADP formed by ␥-GCS.
Similar reaction mixtures were used for qualitative studies in which dipeptide was detected by high pressure liquid chromatography except that L-␣-[ 14 C]aminobutyrate was replaced by 20 mM L-cysteine analog, and the total volume was reduced to 100 l. Aliquots (20 l) were removed before addition of ␥-GCS and were used to determine the elution positions of the substrates. After incubation with ␥-GCS for 2 h at 37°C, the reaction mixtures were quenched by immersion in boiling water for 1 min and were centrifuged to remove precipitated proteins. A 10-l portion of the supernatant was fractionated by reverse-phase high pressure liquid chromatography using a Cu-L-proline elution buffer flowing at 0.5 ml/min and post-column o-phthalaldehyde detection as described by E. Gil-Av et al. (35). In all cases, a new peak corresponding to ␥-glutamylamino acid appeared, and substrate peaks corresponding to L-glutamate and the amino acid tested were diminished in size.
EPR Studies-X-band (8.9 -9.4 GHz) EPR spectra were recorded on a Varian E109 Century Series spectrometer with a Varian TE102 cavity (Varian, Palo Alto, CA). Samples were placed in a flat cell for studies at room temperature and in a 4-mm OD quartz EPR tube in a finger Dewar filled with liquid N 2 for studies at 77 K. S-band spectra (3.4 GHz) were obtained using a spectrometer with a loop-gap resonator cavity and a low frequency microwave bridge built at the National Biomedical ESR Center at the Medical College of Wisconsin (36). Samples for S-band studies were cooled with N 2 passed through an exchange coil immersed in liquid N 2 . S-band EPR samples were placed in quartz tubes of 4.0-mm OD. Microwave frequencies were measured with a frequency counter (EIP model 548), and the field was calibrated with a magnetometer (Rawson-Lush Instrument Co., Acton, MA). Table I, we routinely obtain ϳ40 mg of ␥-GCS from 10 -15 g of cells. The final preparation gives a single protein band with an apparent M r of ϳ58,000 by SDS-PAGE (not shown). These results compare favorably with those obtained using other expression systems, and the specific activity of our final preparations, ϳ3100 units/mg, is 1.7-to 15-fold higher than reported previously for E. coli ␥-GCS purified to apparent homogeneity (27,29,37). For comparison, the best preparations of rat (30) and human (31) (Fig. 1B). In earlier studies, rat kidney ␥-GCS was inhibited with similar efficiency and even more rapidly (38).

Expression and Purification of E. coli ␥-GCS-As shown in
E. coli ␥-GCS can also be fully inactivated by L-S-BSO but only with longer exposures or much higher concentrations of inhibitor. When preincubated with the enzyme in the presence of excess MgATP, 100 M L-S-BSO inhibits E. coli ␥-GCS at about the same rate and to the same extent as 2 M L-S-BSO inhibits human ␥-GCS (Fig. 1, A and B). As expected, L-R-BSO did not inactivate the E. coli enzyme, and 100 M L-S, R-BSO was equipotent with 50 M L-S-BSO, indicating that L-R-BSO does not interfere with inhibition by the active diastereomer (not shown).
Although D-glutamate is a substrate of E. coli ␥-GCS (see later), D-S,R-BSO does not cause significant inactivation (Fig. 1A). There was no evidence of reactivation of E. coli ␥-GCS with time when aliquots of the L-S-BSO-inhibited enzyme were diluted into complete reaction mixtures to assay residual activity. As reported earlier for mammalian ␥-GCS (26,38), BSO-mediated inactivation of E. coli ␥-GCS did not occur in the absence of ATP (not shown).
In the presence of complete reaction mixtures containing physiological levels of L-glutamate (2.5 mM), bacterial and mammalian ␥-GCS show an even greater difference in susceptibility to L-S-BSO-mediated inhibition. As shown in Fig. 1C, 500 M L-S-BSO is required to inhibit E. coli ␥-GCS to an extent comparable with that seen with human ␥-GCS exposed to 1 M L-S-BSO (compare curves 5 and 6). Note that both E. coli and human ␥-GCS show progressive inhibition with time, a result consistent with irreversible binding of L-S-BSO under these conditions.
Studies similar to those in Fig. 1A were carried out at several additional concentrations of L-S-BSO, and the results were analyzed in terms of a model in which L-S-BSO first binds reversibly to ␥-GCS (equilibrium constant ϭ K d ) and is then phosphorylated in a pseudo-first order reaction to become tightly and essentially irreversibly bound (rate constant ϭ k inact ) (38,39). In the presence of MgATP, L-S-BSO is initially bound with a K d of ϳ66 M and then reacts with a k inact of ϳ0.30 min Ϫ1 (t1 ⁄2 ϭ 2.3 min) ( Fig. 2A). As found earlier for rat ␥-GCS (38), inhibition of human ␥-GCS was too fast to analyze by this approach at appropriate inhibitor concentrations (see "Discussion").
Because rat ␥-GCS is active with both Mg 2ϩ and Mn 2ϩ and exhibits a broader substrate specificity in the presence of Mn 2ϩ (40), we considered the possibility that E. coli ␥-GCS would be more effectively inhibited by L-S-BSO in the presence of Mn 2ϩ . As shown in Fig. 1A, inhibition is, in fact, significantly improved in the presence of Mn 2ϩ ; 5 M and 50 M L-S-BSO cause comparable rates and extents of inhibition in the presence of MnATP and MgATP, respectively. Analysis of the MnATP results from Fig. 1A and studies at additional L-S-BSO concentrations indicate that Mn 2ϩ improves both the initial binding affinity (K d ϭ 18.8 M) and the rate of inactivation (k inact ϭ 0.655 min Ϫ1 ; t1 ⁄2 ϭ 1.06 min) over that seen with Mg 2ϩ (Fig. 2B). However, even in the presence of MnATP, the rate of inactivation of E. coli ␥-GCS remains much slower than that seen with human ␥-GCS in the presence of MgATP.
Substrate Specificity of E. coli ␥-GCS-Given the improved activity of L-S-BSO in the presence of MnATP, the metal ion specificity of E. coli ␥-GCS was investigated further. When 25 mM MgCl 2 was replaced in the standard ADP formation assay by 10 mM MnCl 2 or CoCl 2 , the apparent activity was 4 and 16%, respectively, of that seen with Mg 2ϩ . Studies with L-␣-[ 14 C]aminobutyrate showed that ADP formation in the pres- a ␥-GCS activity was quantitated on the basis on L-␣-[ 14 C]aminoburyrate incorporation into dipeptide, as described in "Experimental Procedures." The spectrophotometric assay gives similar results for the DE-52 pool and later fractions but cannot be used for the early fractions due to high non-specific ATPase activity. ence of Mn 2ϩ and Co 2ϩ was coupled to stoichiometric formation of L-␥-glutamyl-L-␣-[ 14 C]aminobutyrate (data not shown). Similar studies of dipeptide formation using 10 mM metal ion concentrations showed Fe 2ϩ and Cd 2ϩ to have activity comparable with Co 2ϩ and Mn 2ϩ , respectively. In contrast, Ca 2ϩ and Zn 2ϩ have detectable but minimal activity (Ͻ1% of the activity with Mg 2ϩ ), and Cu 2ϩ is without activity (data not shown).
Detailed studies of the specificity of E. coli ␥-GCS for Lglutamate and L-cysteine analogs were carried out in the presence of Mg 2ϩ and Mn 2ϩ (Table II). As shown, the specificity for L-glutamate analogs is relatively narrow with only D-glutamate and the N-methyl and ␣-alkyl derivatives of L-glutamate showing significant activity. For each of these active analogs, studies using L-␣-[ 14 C]aminobutyrate showed that dipeptide formation is stoichiometric with ADP formation within a few percent. In general, substitution of Mn 2ϩ for Mg 2ϩ reduces K m values substantially and improves the rate of reaction of poorly reactive glutamate analogs relative to L-glutamate. Analogs inactive with Mg 2ϩ are, however, also inactive with Mn 2ϩ . As with mammalian ␥-GCS (41), L-aspartate is without activity, a finding that suggests L-glutamate binds in an extended conformation that cannot be mimicked by the shorter dicarboxylic amino acid. Note that several L-glutamate analogs including ␤-glutamate, ␤-methylglutamate, ␥-methylglutamate, and L-␣-amino- adipate are essentially inactive with the E. coli ␥-GCS but show significant activity with the rat enzyme (41). Our results are in general agreement with an earlier report that unpurified E. coli ␥-GCS in an immobilized cell matrix exhibited 15 and 5% relative activity with D-glutamate and N-methyl-L-glutamate, respectively, in place of L-glutamate in the presence of MgCl 2 (42).
In contrast to the narrow substrate specificity seen with L-glutamate analogs, several amino acids having side chains approximating the size and hydrophobicity of L-cysteine are good surrogates for that amino acid in the E. coli ␥-GCS reaction. In the presence of Mg 2ϩ , the analogs exhibiting V max values that are ϳ30% or more of the V max seen with L-cysteine include S-methyl-L-cysteine, L-␣-aminobutyrate, ␤-chloro-L-alanine, ␤-cyano-L-alanine, and L-C-allylglycine. These same amino acids show good activity with rat ␥-GCS (Ͼ59% the rate with L-cysteine) (40), and most were active with the immobilized E. coli ␥-GCS preparation (42). Other amino acids such as L-norvaline and ␤-amino-iso-butyrate react well with mammalian ␥-GCS (40, 43) but are much less active or inactive with E. coli ␥-GCS. As with L-glutamate analogs, substitution of Mn 2ϩ for Mg 2ϩ in the assay mixtures decreases the K m values of L-cysteine analogs. For analogs showing activity at least 5% of the activity of L-cysteine, V max values are uniformly higher with Mg 2ϩ than with Mn 2ϩ , but very poor substrates such as L-norleucine, L-leucine, L-valine, and L-isoleucine have higher activity with Mn 2ϩ .
EPR Studies with Mn 2ϩ -The observation that E. coli ␥-GCS has significant activity with Mn 2ϩ suggested that the Mn 2ϩ ion could serve as a useful EPR probe of the active site. In preliminary EPR studies at 77 K, we found that ␥-GCS complexed to one equivalent of Mn 2ϩ in the absence of substrates gives a low intensity but characteristic 6-line X-band EPR spectrum with forbidden transitions. Addition of L-glutamate or L-S-BSO significantly increased the intensity and decreased the peak to peak line width of the Mn 2ϩ signal indicating that Mn 2ϩ is bound near the active site (data not shown).
To determine the number of Mn 2ϩ tightly bound to ␥-GCS and their respective dissociation constants, ␥-GCS was titrated into a solution of 100 M MnCl 2 , and the change in concentration of free Mn 2ϩ was determined from the decrease in the peak to peak amplitude of the low field line of the room temperature free Mn 2ϩ EPR spectra (44). Representative spectra are shown in the Fig. 3A, inset. Scatchard analysis of the titration data indicates that the first and second Mn 2ϩ ions bind to E. coli ␥-GCS with K d values of 1.1 Ϯ 0.2 M and 82 Ϯ 11 M, respectively (Fig. 3A). Similar studies in the presence of 10 mM L-glutamate or 4 mM L-S-BSO give K d values of 0.60 Ϯ 0.14 M or 1.8 Ϯ 0.1 M, respectively, for the first Mn 2ϩ bound and 5.2 Ϯ 0.6 M or 30.9 Ϯ 4.7 M, respectively, for the second Mn 2ϩ bound (Fig. 3, B and C). L-Glutamate and L-S-BSO thus affect the affinity of Mn 2ϩ for the second site more than for the first site.
Addition of L-glutamate or L-S-BSO to ␥-GCS having 1.75 Mn 2ϩ bound per subunit produces a 77-K EPR spectrum having additional hyperfine lines at 45 G intervals both upfield and downfield of the original 6-line spectra (Fig. 4, spectra  1-3). The observed splitting is approximately one-half of the value expected for two magnetically isolated Mn 2ϩ ions and suggests that two bound Mn 2ϩ ions are coupled by an electron spin exchange interaction in the presence of these amino acid ligands (45). Binding of D-glutamate does not produce additional lines (spectra 6) suggesting that D-glutamate is less effective than L-glutamate in inducing Mn 2ϩ -Mn 2ϩ coupling. In contrast to the results with L-glutamate and L-S-BSO alone, addition of ATP in the presence or absence of L-glutamate suppresses both the original 6-line signal and L-glutamateinduced upfield and downfield lines with 45 G coupling (spectra 4 and 5).
Analysis of the changes in signal amplitude during titration of E. coli ␥-GCS with Mn 2ϩ in the presence of L-glutamate or L-S-BSO confirms the Mn 2ϩ -Mn 2ϩ interaction of the bound ions (Fig. 5, A and B). As shown, signal amplitude increases as the initial Mn 2ϩ ion binds but is suppressed by binding of a second  Mn 2ϩ . Interaction of the Mn 2ϩ ions as determined by the presence of the 45 G coupling is not seen with ATP (Fig. 5C) or the absence of additional amino acid ligands (not shown).
EPR Studies with Cu 2ϩ -Although E. coli ␥-GCS has no activity when Mg 2ϩ is replaced with Cu 2ϩ , addition of Cu 2ϩ to a di-Mn 2ϩ -␥-GCS complex causes stoichiometric displacement of Mn 2ϩ with appearance of characteristic Cu 2ϩ X-band EPR signals (not shown). The spectrum of ␥-GCS with two Cu 2ϩ bound gives overlapping EPR signals without evidence of a Cu 2ϩ -Cu 2ϩ interaction, but g ሻ and A ሻ values could be determined from the Cu 2ϩ EPR spectrum obtained from ␥-GCS having one Cu 2ϩ bound (Fig. 6A, upper spectra). Based on Peisach Blumberg plots (46), it is tentatively concluded that Cu 2ϩ is bound to one N and three O ligands in the ␥-GCS active site. 2 Addition of L-S-BSO to the mono Cu 2ϩ -␥-GCS complex alters the EPR spectra in a manner suggesting that Cu 2ϩ then has two N and two O ligands (Fig. 6A, lower spectra).
These conclusions are supported by S-band EPR studies showing that the m1 ϭ Ϫ 1 ⁄2 line in the g ሻ region of the major 2 Although there is little overall sequence homology between glutamine synthetase and ␥-GCS from either E. coli or eukaryotes, J. J. Abbott et al. (34) have recently identified short sequence similarities in the regions of the active sites involved in metal ion binding. Based on modeling and site-directed mutagenesis studies, they conclude that ␥-GCS from Trypanosoma brucei binds two metal ions and that the ligands are three glutamic acid residues for n 1 and two glutamic acid residues and one glutamine residue for n 2 . Their suggested alignment for E. coli ␥-GCS indicates that the n 1 ligands may be two glutamic acids and one aspartic acid, and the n 2 ligands may be two glutamic acids and one histidine. Our mono-Cu 2ϩ EPR studies identify three n 1 O ligands, consistent with the proposal by J.J. Abbott et al. (34), but also find one N ligand. This discrepancy warrants further investigation. Because E. coli ␥-GCS is not active with Cu 2ϩ , it is possible that Cu 2ϩ bridges to a N ligand not involved in binding Mg 2ϩ or Mn 2ϩ . The spectrometer gain was held constant to allow direct comparison of relative peak heights among spectra. Spectrometer conditions: microwave frequency, 9.03 GHz; modulation amplitude, 5 G; modulation frequency, 100 kHz; time constant, 0.128 s; microwave power, 10 mW. species is resolved into three lines attributed to superhyperfine coupling of 13 G from a single N donor atom (Fig. 6B). Upon addition of L-S-BSO, g shifts, suggesting the binding of at least a second N donor atom. The N superhyperfine structure is not resolved well enough to observe a 5-line pattern for two approximately equivalent N donor atoms. DISCUSSION ␥-Glutamylcysteine synthetase has been identified in a wide range of organisms including bacteria, protozoa, fungi, plants, insects, and mammals. Cloning studies show a moderate to high degree of amino acid sequence homology among the mammalian, insect, yeast, and protozoal ␥-GCS isoforms, but little homology between those enzymes and the ␥-GCS of either plants or bacteria (1). Notwithstanding these differences and additional marked differences in quaternary structure, all ␥-GCS, including those of plants and bacteria, are presumed to catalyze a two-step ligase reaction mechanistically similar to that established for rat ␥-GCS (Equations 3 and 4). Supporting that view, most ␥-GCS isoforms are reported to be strongly inhibited by BSO, a mechanism-based inhibitor that initially binds as an analog of the tetrahedral intermediate formed when L-␥-glutamylphosphate and L-␣-aminobutyrate react in the second step of the normal ␥-GCS reaction (1,26,38). Once bound, L-S-BSO is phosphorylated on its sulfoximine N in a reaction that mimics phosphorylation of the ␥-carboxylate of L-glutamate in the first step of the normal ␥-GCS reaction. Studies with rat ␥-GCS show that phosphorylated L-S-BSO is tightly bound to the active site, causing essentially irreversible inhibition (26,38). Mechanism-based inhibition by L-S-BSO thus mimics key aspects of both the first and second steps of the ␥-GCS reaction, and irreversible inhibition by BSO is therefore good evidence for mechanistic similarity between any ␥-GCS isoform inhibited by BSO and the well characterized rat ␥-GCS.
It was in this context that we were interested in the possible discrepancy between a report that BSO inhibits E. coli ␥-GCS comparably to rat ␥-GCS (27) and other reports showing that exposure of intact E. coli to BSO does not cause inhibition of GSH synthesis or GSH depletion (18). As shown in Fig. 1, we find that L-S-BSO can fully inhibit E. coli ␥-GCS, but, at all concentrations tested, the rate of inactivation is much slower than is seen with human (this work) or rat (38) ␥-GCS. Weaker initial binding and a slower rate of sulfoximine phosphorylation both contribute to the slower rate of inactivation. Thus, L-S-BSO binds to E. coli ␥-GCS with a K d value of 66 M whereas human ␥-GCS is inhibited so rapidly by 5 M and higher concentrations of L-S-BSO that meaningful K d values could not be determined. In earlier studies with rat ␥-GCS, it was estimated that the active diastereomer of BSO had an apparent K d of 25 M in the presence of 5 mM L-glutamate, a ligand which competes with the sulfoximine for initial binding (38). Those studies suggest that in the absence of L-glutamate the actual K d for L-S-BSO would be Ͻ8 M, but rapid kinetic studies are necessary to make an accurate determination.
Once bound to E. coli ␥-GCS, L-S-BSO reacts relatively slowly with ATP to form the tightly bound phosphorylated derivative. We find k inact ϭ 0.30 min Ϫ1 (t1 ⁄2 ϳ2.3 min) for E. coli ␥-GCS, a value which is Ͻ8% of the k inact determined earlier for rat ␥-GCS in the presence of L-glutamate (k inact ϭ 3.9 min Ϫ1 ; t1 ⁄2 ϳ11 s) (38). Whether weak binding and slow phosphorylation of L-S-BSO account for its inability to deplete E. coli of GSH requires additional studies. In preliminary experiments, we have confirmed an earlier report that BSO is taken up by E. coli (27), 3 but the intracellular concentration of L-S-BSO relative to L-glutamate and the rate at which E. coli consume or release GSH have not yet been determined.
ATP-dependent ligases such as ␥-GCS require a divalent metal ion, typically Mg 2ϩ , to help bind and orientate the triphosphate portion of the nucleotide in the active site. Such ions may also help to orientate and deprotonate the substrate nucleophile to which phosphoryl transfer will occur (39,45). Although it was earlier reported that E. coli ␥-GCS is not active with Mn 2ϩ , Co 2ϩ , Ca 2ϩ , or Zn 2ϩ (29), we find that highly purified enzyme has substantial activity with Mn 2ϩ , Co 2ϩ , Cd 2ϩ , and Fe 2ϩ . Interestingly, K m values for L-glutamate and glutamate analogs are markedly reduced in the presence of Mn 2ϩ , suggesting that the divalent metal plays a role in Lglutamate binding. This finding is consistent with the general observation that Mn 2ϩ improves ligand binding over that seen with Mg 2ϩ (47). Reaction velocities for most glutamate analogs are reduced in the presence of Mn 2ϩ , but the decrease in V max is less for the analogs than for L-glutamate itself. ␣-Methyl-L-glutamate, which shows very poor activity with Mg 2ϩ , actually has a higher V max with Mn 2ϩ than Mg 2ϩ .
Substitution of Mn 2ϩ for Mg 2ϩ also decreases the K m for L-cysteine and most of its analogs, suggesting that metal ions might also have a role in L-cysteine binding. We do note, however, that the decrease in the K m for L-cysteine itself is small relative to the changes seen for L-glutamate and its analogs. For analogs having Ͼ5% the activity of L-cysteine, V max values were uniformly higher with Mg 2ϩ than Mn 2ϩ , but for very poorly reactive substrates, V max values were higher with Mn 2ϩ . The latter observation is of interest because we find that k inact for L-S-BSO is higher in the presence of Mn 2ϩ . It was earlier proposed that the S-butyl moiety of L-S-BSO occupies the Lcysteine binding site of ␥-GCS (38), and our results are thus consistent with the view that L-S-BSO reacts as a poor ␥-GCS substrate that is phosphorylated relatively slowly.
E. coli glutamine synthetase has no overall amino acid sequence homology with either E. coli or mammalian ␥-GCS, but it catalyzes a mechanistically similar reaction in which L-glutamate is first phosphorylated by ATP and the resulting ␥-glutamylphosphate is then attacked by ammonia to form glutamine (in addition to ADP and P i ) (48). 2 The enzyme is irreversibly inhibited by the L-S-diastereomer of methionine sulfoximine (49), which is the same diastereomer of methionine sulfoximine that potently inhibits rat ␥-GCS (50). Correspondingly, rat (26), human (this work), and E. coli ␥-GCS (this work) are all inhibited by L-S-BSO, suggesting that the relative orientation of substrates is similar in the active sites of glutamine synthetase and mammalian and E. coli ␥-GCS. Extensive EPR (51)(52)(53)(54) and NMR (55,56) studies, confirmed by x-ray crystallography (57,58), have established that each glutamine synthetase active site binds two divalent metal ions. Consistent with the similarity of chemical mechanism and active site geometry between glutamine synthetase and ␥-GCS, our EPR results with Mn 2ϩ indicate that E. coli ␥-GCS also binds two divalent metal ions per active site. The ␥-GCS K d values for Mn 2ϩ , 1.1 M and 82 M, are similar to the Mn 2ϩ K d values of 1.9 M and 200 M reported for glutamine synthetase (59). Furthermore, titration of glutamine synthetase with Mn 2ϩ causes a decrease in EPR signal amplitude from one Mn 2ϩ to two Mn 2ϩ bound (54) that is very similar to that seen with ␥-GCS in the presence of L-glutamate or L-S-BSO (Fig. 5). 4 Interpretation of our ␥-GCS results is thus based in part on previous findings with glutamine synthetase.
In E. coli glutamine synthetase the more tightly bound 3 M.A. Hayward and O.W. Griffith, unpublished data. 4 It should be noted that glutamine synthetase was titrated at 1°C and Mn 2ϩ binding was determined from the peak to peak height of the low field line of the 6-line Q-band EPR spectrum; no resolution of the 45 G coupling was observed (54). In contrast, our titration of ␥-GCS was followed by X-band EPR at 77 K. This approach allowed us to observe the 45 G spin exchange coupling in addition to the decrease in signal amplitude of the low field line of the characteristic 6-line Mn 2ϩ spectrum on going from one to two Mn 2ϩ equivalents bound. However, collection of data at 77 K also raised the possibility that ice crystals that were formed in freezing the samples excluded and therefore concentrated the solutes (see Ref. 60 for discussion). Such problems were minimized or avoided by working at high protein concentrations (e.g. ϳ35 mg/ml in our studies) or by inclusion of solvents that interfere with solvent-solute segregation (e.g. 25% glycerol in our studies). More specifically, the decrease in peak to peak signal amplitude observed with ␥-GCS is attributed to conversion of the strong signal with a 90 G coupling to a weaker signal with a 45 G coupling ( Figure 5). In the case of glutamine synthetase, the decrease in signal amplitude of the line with 90 G coupling was interpreted using Leigh theory, which is based on relaxation effects wherein one paramagnet relaxes much faster than the paramagnet under observation. This is not the situation with ␥-GCS and may not have been the situation with glutamine synthetase. Note that application of Leigh theory in glutamine synthetase predicted a Mn 2ϩ -Mn 2ϩ distance of ϳ10 Å (52-54) whereas the crystal structure shows a distance of 4.5 Å (57). Mn 2ϩ , n 1 , is located near the ␥-carboxylate of substrate Lglutamate; one of the substrate's carboxylate O atoms and O atoms from three enzymatic glutamate residues serve as ligands. The second Mn 2ϩ , n 2 , is located adjacent to the ␥-phosphate of ATP. Its ligands include a ␥-phosphoryl O and the ␤-␥ bridging O of ATP, the N of an enzymatic histidine residue, and O atoms from two enzymatic glutamate residues. Although our current results do not allow us to unambiguously assign the first or second bound metal ions to the site nearer substrate glutamate or ATP in ␥-GCS, our working hypothesis is that a similar arrangement exists in the E. coli ␥-GCS active site. In support of that view, our studies with the mono Cu 2ϩ -␥-GCS complex show that in the absence of substrates or inhibitor Cu 2ϩ has one N ligand, presumably an enzymatic histidine residue. The Cu 2ϩ ion acquires a second N ligand in the presence of L-S-BSO (Fig. 6), and we suggest that the additional ligand may be the sulfoximine N of bound L-S-BSO 2 . We do note that Scatchard analyses show L-glutamate and L-S-BSO affect the K d values of n 2 more than n 1 (Fig. 3), a finding that might argue for a different assignment of n 1 and n 2 . However, Lglutamate binding reportedly increases the K d of both n 1 and n 2 in glutamine synthetase by 20-fold (59), and a major effect on n 2 thus does not rule out the interpretation that L-glutamate and L-S-BSO bind nearer n 1 in ␥-GCS. Binding of ATP apparently increases the rhombicity of both the n 1 and n 2 sites since addition of ATP was found to suppress the low amplitude EPR signal of the mono Mn 2ϩ -␥-GCS complex (not shown) and the stronger signal of the di-Mn 2ϩ -␥-GCS complex (Fig. 4).
Crystallographic studies establish that the metal ions of E. coli glutamine synthetase are separated by 4.5 Å and that substrate binding does not significantly perturb that distance (57). Observation of a 45 G coupling for hyperfine lines in the di-Mn 2ϩ -␥-GCS complex EPR spectra in the presence of Lglutamate or L-S-BSO (Fig. 4) and titration data showing suppression of the EPR signal between 1 and 2 Mn 2ϩ bound (Fig.  5) indicate that the Mn 2ϩ ions bound to E. coli ␥-GCS form a binuclear Mn 2ϩ -Mn 2ϩ site in the presence of those ligands (45). 5 Furthermore, observation of the 45 G coupling suggests that the Mn 2ϩ ions are separated by 5 Å or less (45). Although in the studies reported here 45 G hyperfine lines were not observed in the absence of amino acid ligands, preliminary studies at low temperature suggest a Mn 2ϩ -Mn 2ϩ interaction may also occur in the absence of additional ligands. 6 Further studies are necessary to accurately determine the Mn 2ϩ -Mn 2ϩ distance and to fully identify the enzymatic and substrate ligands of each metal ion.