A Novel Reaction Catalyzed by Unadenylylated Glutamine Synthetase from Escherichia coli AMP-DEPENDENT SYNTHESIS OF PYROPHOSPHATE AND L-GLUTAMATE FROM ORTHOPHOSPHATE AND L-GLUTAMINE*

The unadenylylated, manganese form of glutamine synthetase (L-glutamate: ammonia ligase (ADP form- ing), EC 6.3.1.2 from Escherichia coli catalyzes a novel, AMP-dependent (reversible) synthesis of pyrophos- phate and L-glutamate from orthophosphate and L-glu-tamine: Gln + 2 Pi The hydrolysis of the L-glutamine amide bond is coupled to the stoichiometric synthesis of pyrophosphate, although as PPi accumulates, additional hydrolysis of L-glutamine occurs in a secondary reaction catalyzed by the enzyme *AMP PPi] complex. The synthesis of PPi probably occurs at the subunit cata- lytic site in the positions normally occupied by the &y-phosphates of ATP. promote PPi synthesis, AMP to the subunit catalytic site rather than to the allosteric inhibitor site; equilibrium binding of the active site. In reaction, M&+ will not substitute for Mnz’, and adenylylated glutamine synthetase is inactive. Pyrophosphate is synthesized by the unadenylylated, manganese enzyme at - 2 8 of the rate of of in the reverse biosynthetic reaction. If Pi is the enzymatic of the AMP- hydrolysis

AMP is a feedback inhibitor of the biosynthetic reaction catalyzed by E. coli glutamine synthetase (3,4): Unadenylylated glutamine synthetase requires Mg2+ in the forward reaction, but can catalyze the reverse reaction in the presence of either Mg2+ or Mn2+ (4, 12,13). AMP binds independently to 12 sites of the dodecamer with K'A = 8 X lo3 M" at 4°C and pH 7.3 in the absence of other ligands (14). An allosteric subunit site for AMP (distinct from that binding the nucleotide substrate) has been inferred from equilibrium binding (6, 14), NMR (8), and calorimetric studies (6). Separate subunit sites for ADP and AMP also are shown by the measurements of the simultaneous binding of [I4C]ADP and [32P]AMP to unadenylylated glutamine synthetase reported here. Nevertheless, AMP (substituted for the nonconsumable substrate ADP) supports the y-glutamyl transfer reaction catalyzed by unadenylylated glutamine synthetase in the presence of Mn2+ or Cd2+ (1,4): L-Glutamine + NH20H ADP or AMP, MnZ+ Arsenate or P, ' y-glutamylhydroxamate + NH:, Adenylylation of the subunit decreases the affinity >2000-fold for ADPMn in Reaction 2 (1). The binding of ADPMn and arsenate or of ADPMn and P, ( K , = 0.03 to 0.08 PM for ADPMn) is strongly synergistic (1,13). The unadenylylated, manganese enzyme has a >3000-fold greater affinity for ADPMn than for AMP in Reaction 2. This fact and the report of Rhee et al. (13) that the unadenylylated, manganese enzyme in the presence of L-glutamine and Pi can phosphorylate AMP to yield ATP (without detecting intermediate ADP formation) suggested that commercial preparations of AMP might contain small amounts of ADP which were actually responsible for both reactions. We have found that chromatographically purified AMP does support the y-glutamyl transfer Reaction 2 catalyzed by unadenylylated, manganese enzyme. However, the product of the reaction studied by Rhee et al. (13) has been discovered to be pyrophosphate rather than ATP. This novel, AMP-dependent reaction is described in the present paper and can be written: L-Glutamine + 2 P, , -L-glutamate + PPi + NH3 (3)

Mn2+, AMP
Mn2+ but not Mg'+ supports this activity of the unadenylylated enzyme; fully adenylylated glutamine synthetase does not catalyze Reaction 3. Reaction 3 is analogous to the reverse 10663 biosynthetic Reaction 1 in which the P,y-phosphate bond of ATP is synthesized, although Reaction 3 is catalyzed at only -2% of the rate of ATP synthesis. Reaction 3 also is related to the irreversible arsenolysis of L-glutamine observed by Levintow and Meister (15) and Purich and Huang:' L-Glutamine + Ha0 Me2+, ADP or AMP Arsenate b L-glutamate + NHB (4) AMP may bind exclusively to the nucleotide substrate site of the enzyme subunit under the conditions of Reactions 2, 3, and 4.
MATERIALS AND METHODS Equine muscle AMP, yeast AMP, CMP, ADP, and ATP were purchased from Sigma Chemical Co. Also, AMP, GMP, and UMP (P-L Biochemical Inc.), AMP (Calbiochem), and IMP (Boehringer Mannheim Biochemical) were used. Commercial AMP was purified on DEAE-Sephadex A-25 employing a gradient elution with 0.05 to 0.15 M (NH4)2COa, followed by 8 to 10 repeated lyophilizations of the isolated AMP to remove (NH4)2C03; ["'PIAMP (1.4 mCi/nmol, Amersham Radiochemical Centre) was diluted to about lo00 cpm/nmol with the purified AMP and relyophilized before use. For the experiments in Figs. 1, 2A, and 3 to 5, 5 to lox lyophilized muscle or yeast AMP from Sigma Chemical Co. was used without further purification. Carrier-free ["2P]orthophosphate at pH 2 to 3 (Amersham Radiochemical Centre) was purified before use by chromatography on Dowex 1-X2 (C1-form, Bio-Rad Laboratories) using the conditions of Fig. 3 (below). Similarly, sodium v2P]pyrophosphate (5.5 mCi/nmol, New England Nuclear) also was purified on Dowex I-C1-before use. The [I4C]ADP (stored frozen at pH 6.8) was from a previous study (1). Chromatography on PEI-cellulose' thin layers by the procedure of Cashel et al. (16) indicated that radioactively labeled nucleotides, P,, and PP, were pure (Figs. 1 Fig. 3 (below). Anhydrous L-glutamine (Sigma Chemical Co., grade 111) was added to solutions of ~-[I~C]glutamine just prior to experiments. Yeast inorganic pyrophosphatase (Sigma Chemical Co.), which was obtained as a lyophilized powder containing about 10% protein and 9 0 % buffer salts (Tris-HC1, sodium citrate, and MgCld, was dissolved in water at a concentration of 0.25 mg/ml (-162 units/ ml). Magnesium titriplex (KzMgEDTA) was from EM Laboratories; Hepes and Trizma base were from Sigma Chemical Co. All other reagents were as described previously (1).
Thin layer chromatography was on precoated plastic sheets of PEIcellulose (Polygram gel 300 PEI from Brinkmann Instruments) in an Eastman chromatogram (No. 13259) developing apparatus. Development was ascending with 0.25 to 1.0 M KH2P04 (pH 3.4) as described by Cashel et al. (16). Autoradiograms were prepared on Kodak RP/ S 54 x-ray film. Liquid scintillation counting was in Aquasol (New England Nuclear), using a Mark I Nuclear Chicago instrument.
Unadenylylated glutamine synthetase (GST) was purified by the procedure of Woolfolk et al. (18,19) and stored as before (19). Unadenylylated enzyme preparations (CST) were used throughout, except for a few experiments with fully adenylylated enzyme (GSr;) which was kindly supplied by R. J. Hohman of this laboratory. Glutamine synthetase preparations were routinely assayed by the pH 7.57 y-glutamyl transfer assay (1,20); the state of adenylylation was determined by assay (20) and spectrophotometric (21) methods. Protein determinations were from AZW nm and A290 " , , , measurements (21). D. L. Purich and C. Y. Huang, unpublished results with glutamine synthetase from Escherichia coli quoted in review reference (3).
The preparation of ~-['~C]glutamine by this method was suggested to us by Dr. S. G. Rhee of this laboratory.
Enzyme solutions were dialyzed against several changes of lo00 X volume of buffer (pH 7.2) containing 20 m~ Hepes/KOH, 100 mM KCl, and 1.0 mM MnC12 before use in the different reactions described in the legends to figures and tables. Ammonia was not removed from the enzyme or buffers in these studies. RESULTS AMP-supported y-Glutamyl Transfer Activity-With AMP obtained from different commercial sources (P-L Biochemicals, Calbiochem, and Sigma Chemical Co.), the maximum activity of the unadenylylated, manganese enzyme in Reaction 2 with arsenate present varied from 7 to 43 units/mg at 37°C. Therefore, a contamination of commercial preparations of AMP with small amounts of ADP was suspected. Two preparations of AMP obtained from Sigma Chemical Co. (yeast AMP, Lot 44C-7290, and muscle AMP, Lot 87C-7090), which varied 6-fold in their ability to support Reaction 2, were purified by chromatography on a DEAE-Sephadex A-25 column using a gradient elution with 0.04 to 0.5 M (NH4)&03. With either lot of AMP, a minor component (-0.05% of the AMP) eluted at the position of ADP, absorbed at 259 nm, and supported the y-glutamyl transfer activity of glutamine synthetase. However, the chromatographed AMP from both lots (after lyophilization to remove ammonium carbonate) supported the activity of the unadenylylated, manganese enzyme in Reaction 2 to the same extent (-50 units/mg at 37"C, pH 7.2) with K,,, 1 mM for AMP. Subsequently, it was found that apparent differences between commercial batches of AMP could be removed by repeated lyophilizations.
Thus, AMP supports the y-glutamyl transfer reaction catalyzed by the unadenylylated, manganese enzyme as previously reported (1,4). However, the amplitude and shape of the enzyme activity versus AMP concentration curve is affected by volatile contaminant(s) present in commercial preparations of AMP, one of which could be (NH4)2C03 (an inhibitor of Reaction 2). AMP-and ADP-supported Reactions with L -Glutamine and Orthophosphate- Fig. 1 shows an autoradiogram of PEIcellulose thin layers of ADP-and AMP-supported reactions catalyzed by unadenylylated glutamine synthetase in the presence of Mn2+, L-glutamine, and 32Pi. With ADP as the only nucleotide present, [32P]ATP is synthesized in the reverse biosynthetic Reaction 1 (Fig. 1, lane 1 ). With AMP as the only nucleotide present, a 32P-labeled compound with a migration corresponding to that of PPi is synthesized in a previously unreported reaction (Fig. 1, lane 8). With 12 mM AMP and varying concentrations of ADP (1 to 100 p~ in lanes 7 to 2 of Fig. l), [32P]ATP synthesis decreases with decreasing ADP, while that of a slower migrating 32P-labeled compound increases. With 5 and 7 PM ADP added to 12 mM AMP, radioactivity at both positions is visible (lanes 5 and 6, Fig. 1). With 1 p~ ADP added to 12 mM AMP (lane 7, Fig. l), only the new 32P-labeled product is synthesized. Fig. 1 (lane 9) also shows that 80 mM NKC1 completely blocks the AMP-supported reaction. A control sample in lane 10 without nucleotide (ADP or AMP) shows only 32Pi at the solvent front.
In the reactions of Fig. 1, 1 PM ADPMn in the absence of AMP is sufficient to saturate 98% of the nucleotide substrate sites of the enzyme with ADPMn (1, 13). With 12 mM AMP present, there are decreasing amounts of [32P]ATP synthesized as the ratio of ADP/AMP is decreased from about 0.88 to 0.06% in Fig. 1. Since the rate of 32PPi synthesis in Reaction 3 is only about 2% of that of ATP synthesis in the reverse biosynthetic Reaction 1 (Table I11 below), AMP predominantly competes with ADP binding in the experiments of Fig.  1. The apparent inhibition of pyrophosphate synthesis by NKC1 in Fig. 1 could be an effect on Keg of Reaction 3 (see below). ; lanes 2 to 7,12 mM AMP and 100,50,10,7,5, and 1 p~ ADP, respectively (arrow indicates direction of increasing ADP concentration); lanes 8 and 9, 12 mM AMP (without ADP) with 80 mM NH4Cl also present in the reaction of lane 9; lane 10, without ADP or AMP present. After terminating reactions by the addition of EDTA (4 mM), aliquots (-2 pl) were spotted on PEI-cellulose (origin); chromatograms were developed (ascending) with 0.85 M KH2PO4 at pH 3.4 (16) and then exposed for 6 days to x-ray film. The mobilities of standard PPi, ATP, and Pi are given on the right.
The chromatograms of Fig. 1 were developed with 0.85 M KP04 buffer (pH 3.4). Under these conditions, ADP, ATP, PPi, and adenosine tetraphosphate (in order of decreasing mobility) are well separated (16). Lower concentrations of this developing buffer gave poorer resolutions of ATP and PPi, which may account for the error of Rhee et al. (13) in identifying the "P-labeled product of Reaction 3 as r3'P]ATP, rather than correctly as "'PPi.
Identification of Reaction 3 Products-Inorganic pyrophosphate was identified as a product of Reaction 3 by chromatography, hydrolysis by inorganic pyrophosphatase, and charcoal adsorption (Figs. 1 to 3). The product, L-glutamate, was identified chromatographically (Fig. 3). The supporting nucleotide in Reaction 3 was shown to be unaltered chemically by using ["'PIAMP ( Fig. 2B).
After the synthesis of the 32P-labeled product of the AMPsupported reaction ( Fig. 2A, lane I ) a portion of the reaction mixture was treated with yeast inorganic pyrophosphatase with Mg' ' present. Subsequent chromatography (lane 2, Fig.  2A) revealed that the "P-labeled product was hydrolyzed by the yeast inorganic pyrophosphatase which is absolutely specific for PPi in the presence of Mg' (22). In a control experiment, authentic "'PPi was added to a reaction mixture without glutamine synthetase present (lane 3, Fig. 2A) and then was treated with inorganic pyrophosphatase (lane 4, Fig. 2A) in the same way as was the enzyme reaction product. After inorganic pyrophosphatase treatment (lanes 2 and 4, Fig. 2A), all of the radioactivity appears at the solvent front as "Pi. Further identification of products of the AMP-supported reaction catalyzed by unadenylylated glutamine synthetase was from ion exchange chromatography on Dowex l-Cl- (Fig.  3). Elution positions are shown by arrows in Fig. 3. Starting with ~-['~C]glutamine and 32Pi in the AMP-supported reac-

TABLE I Charcoal adsorption of reaction products
Reaction conditions as in Fig. 1 (lanes 10, 8, or I ) with 1000 cpm/ nmol of '"P,. After 60 min at 30°C. reactions were diluted 10-fold into M KPO, (pH 7) and 20 mg of charcoal (Norite A, Fisher Chemical Co., prewashed with H:,PO., and water), equilibrated a t 0°C for 1 h, and filtered through Millipore (0.45 p ) filters; the collected charcoal then was washed with water and counted. In AMP-supported reactions with L-glutamine and 32Pi, radioactivity was not adsorbed to any significant extent by charcoal (Table I). Under the same conditions, r3'P]ATP formed from L-glutamine, ADP, and 32Pi (Table I) ( 1 2 ) .
The isolated "'P-labeled product of Fig. 3 represented by the hatched area also was not adsorbed to charcoal; 97 f 2% of the radioactivity was recovered in the eluate after charcoal treatment. Therefore, a "P-labeled nucleotide is not synthesued in Reaction 3; the reaction product "PPi is not adsorbed  Table I and 97 f 2% of the radioactivity was recovered in the filtrate.  After 150 min at 30°C. reaction products were separated at 4°C on a Dowex l-Clcolumn as in Fig. 3 and counted. The micromolar quantity of each product is an average; the number of independent determinations is given in parentheses. Results were the same using 3.8 mg of GST for 10 min at 30°C. Different components of the reaction were omitted or replaced as indicated. 'Separation on Dowex l-Cl-was not performed; instead, "PP, from '"P, (-1300 cpm/nmol) was not be detected in autoradiograms of PEI-cellulose thin layers (e.g. Fig. 1).
Fully adenylylated glutamine synthetase (0.75 mg of GSz) was substituted for GS i.  Table 11. experiments of Fig. 2B, the addition of NH20H produces Reaction 2 in which AMP and Pi are not consumed.) Table 11, ~-['~C]glutamine was used as soon as possible after preparation (see "Materials and Methods") and reactions were stopped by immediate chromatography on Dowex l-Cl-at 4°C (Fig. 3) Table I1 are corrected for the nonenzymatic hydrolysis of ~-['~C]glutamine. Table I1 shows that the AMP-supported synthesis of PPi and L-glutamate from Pi and L-glutamine catalyzed by unadenylylated glutamine synthetase has absolute requirements for Pi, L-glutamine, AMP, and Mn" (50 m~ MgC12 substituted for Mn2+ produced no reaction). The reaction also is specific for the unadenylylated form of glutamine synthetase, since the substitution of the fully adenylylated enzyme at an even higher concentration did not produce detectable 32PPi synthesis (Table 11). The concentration of KC1 (28 to 130 m~) in Reaction 3 appeared to be unimportant.

Reaction 3 Requirements and Apparent Stoichiomern-For the data of
Isolation of the products of the AMP-supported reaction with both substrates radioactively labeled (Table 11) consistently gave a higher value for ~-['~C]glutamate than for 32PPi formation. This deviation from the anticipated exact coupling between L-glutamine hydrolysis and pyrophosphate synthesis  Fig. 4 (Curues 1 and 2, respectively). At early times (t20 min), there is an exact correspondence between the amounts of L-glutamate and PPi formed. Initially, therefore, there is a coupling between hydrolysis of the amide bond of L-glutamine and synthesis of the pyrophosphate bond. After about 20 min at 30°C, Curves 1 and 2 of Fig. 4 show different time courses. Extrapolations from about 90 min to the data in Table I1 for 150 min are shown by the dashed portions of Curves 1 and 2 (Fig. 4). The data are internally consistent and indicate that more L-glutamate than PPi is formed at -30 to 150 min. These data suggested that in addition to Reaction 3, a secondary reaction occurred at the later times which involved L-glutamine hydrolysis independent of PPi synthesis. Such a secondary reaction could be initiated by an accumulation of synthesized PPi if PPi binds to unadenylylated, manganese glutamine synthetase (with Kk  (Table 111). The reverse Reaction 3 required AMP and Mn"; no hydrolysis of "PPi was observed in this reaction when 10 mM MgClz was substituted for 0.5 mM MnC12 (Table  111).
Nucleotide-supported Arsenolysis Reaction 4- Fig.  5 shows the rates of ADP-and AMP-supported arsenolysis reactions catalyzed by the unadenylylated, manganese enzyme. Note that L-glutamine hydrolysis is linear with time in Reaction 4, which is not the case for the reverse biosynthetic Reaction 1 (13) or for PPi synthesis in Reaction 3 (Fig. 4). The relative rate of the AMP-to the ADP-supported Reaction 4 is -0.5 which is about the same ratio as observed in Reaction 2 (Table 111). GMP and IMP (at 12 mM concentration) are -30% as effective as AMP in supporting Reaction 4 (Fig. 5). GMP and IMP also support the y-glutamyl transfer Reaction 2, although K,,, values are higher for GMP and IMP than that for AMP.4 In addition, GDP and IDP are substrates in the reverse biosynthetic reaction (13). Table I11 summarizes activity data for unadenylylated glutamine synthetase in Reactions 1 to 4 so that the activities in AMP-and ADP-supported reactions with either Mn2+ or M e present may be compared. Note that the AMP-supported PPi synthesis in Reaction 3 with Mn2+ present is only 1.5% of the rate of ATP synthesis in the reverse biosynthetic Reaction 1. The hydrolysis of PPi in the reverse Reaction 3 at pH 7.2 occurs at -36% of the rate of the forward direction, with no attempt being made here to optimize the concentrations of reactants in either direction of Reaction 3. When Pi in Reaction 3 is replaced by arsenate, the AMP-supported, irreversible arsenolysis Reaction 4 is -100-fold faster than PPi synthesis in Reaction 3 and is one-half the rate of the ADPsupported arsenolysis of L-glutamine. In forming the product y-glutamylhydroxamate in Reaction 2 with arsenate present, NHzOH increases the rates (relative to rates in Reaction 4) of both the AMP-and ADP-supported reactions 100-fold.

AMP, ADP, and Divalent Cation Support in Reactions 1 to 4-
It is of interest that the unadenylylated enzyme can not catalyze either direction of Reaction 3 with Mg2+ substituted for Mn2+ (Table 111). Moreover, M$+ is unable to support Reaction 2 (and insigmtkantly Reaction 4) when AMP is the E. R. Stadtman and P. Z. Smyrniotis, unpublished data.

TABLE I11
Unadenylylated glutamine synthetase activity Activity data are expressed as initial rates of nanomoles of product formed at pH 7. " y-Glu.NHOH, y-glutamylhydroxamate.

Simultaneous binding of [32P]AMP and [14C]ADP to
unadenylylated glutamine synthetase Equilibrium dialysis was performed as described previously (14).   1 (3, 4, 13) and catalyzes the reverse biosynthetic reaction 4-fold faster with M e than with Mn2+ present (Table 111). NucZeotide Binding Sites-The equilibrium dialysis results given in Table IV show that there are separate binding sites for ADP and for AMP on each subunit of the unadenylylated, manganese enzyme. In the absence of other substrates, -0.7 equivalent/subunit of each [32P]AMP and [I4C]ADP can be simultaneously bound. In obtaining these data, it was observed that the binding of AMP was somewhat slower in the presence than in the absence of ADP; longer equilibration times were required in these cases. Nevertheless, the data of Table IV are consistent with each subunit binding ADP at the catalytic site and binding AMP at an allosteric inhibitor site. A site for AMP distinct from the catalytic site has been suggested previously from other data (6, 8, 14). In addition, AMP binds with equal affinity to the inactive, divalent cationfree enzyme and to the active, manganese or magnesium enzyme (14), whereas Mn2+ (or M F ) is required for tight binding of ADP (1) or ATP (12) to the subunit catalytic site.
Since AMP supports Reactions 2 to 5, it is possible that AMP binds to the ADP site of the subunit under these reaction conditions in which L-glutamine and Pi, arsenate, or PPi are present. Alternatively, it is possible that AMP SUPports these reactions by binding to a noncatalytic site, since AMP is unchanged (Fig. 2B). When 10 mM L-glutamine was present in the binding experiments of Table IV, both AMP and ADP were bound simultaneously to the enzyme subunit. However, orthophosphate, which is synergistic to the binding of ADP (13), is antagonistic to the binding of AMP when ADP is bound also to the subunit catalytic site. These results indicate that AMP is almost totally excluded from binding at the allosteric site of the subunit when ADP and Pi are bound to the catalytic site. In the absence of ADP, orthophosphate apparently has little effect on AMP binding (Table IV). In this case, however, Pi may direct the binding of AMP to the ADP site. Whatever binding sites are involved, the stoichiometry of AMP binding to glutamine synthetase does not exceed 1 equivalent of AMP bound/enzyme subunit (Table  IV, Ref. 14) which could be due to mutually exclusive binding at two sites. The nucleotide binding site at the subunit catalytic site has a rather broad specificity for the purine base, since ADP, GDP, and IDP are phosphorylated in the reverse biosynthetic Reaction 1 (13). Since AMP, GMP, and IMP (substituted for ADP with the unadenylylated, manganese enzyme) support Reactions 24 and 4 (Fig. 5), it appears likely that these 5'nucleotide monophosphates bind to the ADP site when Lglutamine and orthophosphate or L-glutamine and arsenate are present. Recall also (Fig. 1) that with varying amounts of ADP and AMP in mixtures, ATP synthesis in the reverse Reaction 1 was apparently competitive with PPi synthesis in Reaction 3. With low ratios of ADP to AMP, both products ([32P]ATP and 32PPi) were seen (Fig. l), whereas high proportions of ADP only produced [32P]ATP.

DISCUSSION
Unadenylylated glutamine synthetase from E. coli has been discovered to catalyze a reversible, AMP-dependent synthesis of pyrophosphate and L-glutamate from orthophosphate and L-glutamine in the presence of Mn2+. In this reaction, Mg2' cannot substitute for Mn2+ and the fuUy adenylylated enzyme is inactive. Reactions 2 to 5 relate directly to the mechanism of enzyme-catalyzed synthesis of L-glutamine in Reaction 1 and will be discussed from this viewpoint.
If we label potential phosphate binding sites at the catalytic site of each subunit according to the a#, and y-phosphate positions occupied by ATP, pyrophosphate synthesis presumably occurs at the P,y-phosphate positions. The Mn2+ ion at the n2 site which is involved in chelating ADP + Pi or ATP at the catalytic site (1, 9, 23) would complex phosphates at the P,y positions (24). We conclude that the nucleotide site of the enzyme subunit must be occupied either by ADP (GDP or IDP) or by AMP (GMP or IMP) for synthesis of the p,yphosphate bond to occur. The energy of pyrophosphate bond synthesis is coupled to that released by the hydrolysis of Lglutamine. However, Reaction 3 is not energetically favored since the free energy of formation of the PPi bound is -6.4 kcal at pH 7.0, 0.5 mM Mg2+ (25) and the hydrolysis of Lglutamine only releases 3.4 kcal (26). The reversal of L-glutamine synthesis in Reaction 1 is even less favorable (15) with the comparable free energy of formation of the P,y-phosphate bond of ATP = 7.9 kcal (25).
In pyrophosphate synthesis with AMP bound to the nucleotide site, there is no &phosphate bond. If ADP occupies this site (with an intact a,P-phosphate bond), the synthesis of the P,y phosphate bond to form ATP in the reversal of Lglutamine synthesis in Reaction 1 is -70-fold faster than when the a,S-phosphate bond is missing. Apparently, ADP + Pi are better able than are AMP + 2 Pi to promote the correct conformation for coupling P,y-phosphate bond synthesis to L-glutamine hydrolysis. The structure of the nucleotide binding site of the enzyme subunit would be expected to be more flexible when AMP and Pi fi this site, recalling also that an additional oxygen and accompanying negative charges must be accommodated in this case.
The a$-phosphate bond of ADP also is not required for the arsenolysis Reaction 4 or for the y-glutamyl transfer Reaction 2 (l), since AMP can replace ADP in either Mn2+supported reaction catalyzed by unadenylylated glutamine synthetase. When arsenate is substituted for Pi in the AMPsupported Reaction 3, the reaction rate is increased 100-fold (Reaction 4) and then is approximately one-half of the reaction rate of the ADP-supported irreversible arsenolysis of Lglutamine (Table 111). In studies with the pea seedling enzyme, Levintow and Meister (15) noted the close relationship of the arsenolysis reaction to 7-glutamyl transfer and reversal of Lglutamine synthesis. They suggested that when Pi is replaced by arsenate in the reversal of L-glutamine synthesis, the reaction is displaced toward L-glutamate formation as a result of forming the unstable arsenate intermediate, which undergoes spontaneous hydrolysis. The rate of the transfer Reaction 2 with either AMP or ADP is 100-fold faster than that of arsenolysis (Reaction 4) because hydroxylamine is a much better nucleophilic attacking agent than is water (Table 111).
The "0 studies of Varner et al. (27) and Rhee et al. (13) show that the oxygen of either orthophosphate (in the reversal of Reaction 1) or arsenate (in the ADP-supported arsenolysis Reaction 4 ) is transfered to L-glutamate. (This oxygen transfer from P, agrees with that determined earlier (28, 29) for the enzymatic synthesis of L-glutamine in which the y-carboxyl oxygen of L-glutamate is transfered to Pi.) Thus, Pi and arsenate apparently have identical roles in the y-glutamyl transfer Reaction 2 (as suggested earlier (15)) with both anions binding to the same site and forming enzyme-bound intermediates with L-glutamate (13,30). It is reasonable to assume that a similar transfer of oxygen from Pi or arsenate to Lglutamate occurs in the AMP-supported Reactions 2 through

4.
In Reactions 2 and 4, the nucleotides AMP and ADP probably occupy the same site on the enzyme subunit. This site has a low specificity for the purine base (13)4 which is true also for the brain enzyme (31). Substrates bind randomly to the E. coli enzyme (1, 13).' However, a marked synergism in binding has been observed between Mn2+ and substrates (1,19,23,32,33), between ATP and L-glutamate (34), between ADP and arsenate (l), and between ADP and Pi (13). In contrast, the binding of AMP appears not to be influenced by the presence of Pi (Table IV), although it seems likely that Pi directs the binding of AMP to the ADP site while blocking the binding of AMP to the subunit allosteric site.
The requirements for Mn2+ and for the unadenylylated form of glutamine synthetase in Reactions 2 through 4 are interesting. In the biosynthetic Reaction 1, the unadenylylated enzyme is very active with M$+ (but inactive with Mn2+) as the supporting divalent cation (3). Adenylylation of the protein converts glutamine synthetase to a form that requires Mn2+ in Reactions 1 and 2 (3) and that has a lowered affinity for ADP and arsenate in Reaction 2 Rhee et al. (13) proposed that the inability of Mn2+ to support the biosynthetic activity of the unadenylylated enzyme in Reaction 1 is, in fact, due to the very high affinity (with corresponding slow off rates) that this enzyme form has for ADP (K'u = 20 nM) and Pi (KO = 80 p~) .
The unadenylylated, manganese enzyme does catalyze the reversal of L-glutamine synthesis in Reaction 1, although this reaction rate at pH 7 is 4-fold slower than with Mg2+ present (  (4, 13).
The adenylylated, manganese enzyme catalyzes Reaction 1 in both directions and also the ADP-supported y-glutamyl transfer Reaction 2 (3). However, this enzyme form does not catalyze the AMP-dependent synthesis of PPi in Reaction 3 or the AMP-supported transfer Reaction 2 (1, 4). Thus, the Mn2+-promoted conformations of the unadenylylated and adenylylated enzymes must differ when substrates are present. Recent evidence indicates that this is the case (35).
The role of divalent cations in the catalytic mechanism of glutamine synthetase from E. coli has been the subject of much study (1, 3, 5, 19, 23, 32, 33, 36-38). It has been shown that two Mn2' sites per enzyme subunit must be filled for activity expression in the y-glutamyl transfer Reaction 2 (1).
As already mentioned, Mn2+ at the lower affinity (n2) site complexes with the nucleotide at the catalytic site (1). The role of Mn2+ at the high affinity (nl) site is structural as well as serving a possible catalytic function. The binding of Mn2+ to nl sites of the dodecamer promotes a conformational change from an inactive (relaxed) to an active (tightened) form (3-5). NMR (23,36),EPR (33,36-38),and equilibriumbinding (32) data are consistent with the Mn2' ion at the subunit nl site also binding the y-carboxyl group of L-glutamate, possibly with a highly immobilized water molecule intervening between Mn2+ and the y-carboxyl group. A Mn2+ interaction with L-glutamate could serve to orient and polarize the ycarboxyl group in Reaction 1 for its attack on the y-phosphorus of ATP (28, 29), which is followed by nucleophilic attack by ammonia on the y-glutamyl phosphate intermediate (30).
It is noteworthy that Mn2+ at the subunit nl site does not interact with L-glutamine (1,19,32). With the transition state analog L-methionine-S-sulfoximine phosphate and ADP irreversibly bound to the enzyme subunit (2,30), both the nl and nz metal ions are locked tightly into the subunit structure (32).
The major reaction pathway of L-glutamine synthesis in Reaction 1 (reviewed by Meister (30)) catalyzed by glutamine synthetase from E. coli (13, 39) is strikingly similar to that catalyzed by the brain enzyme (30). This pathway involves the formation and utilization of enzyme-bound y-glutamyl phosphate intermediate whose structure is analogous to that of L-methionine-S-sdfoximine phosphate (2,30). Meister (30) has summarized the evidence from his laboratory for formation of the y-glutamyl phosphate intermediate in L-glutamine synthesis. In addition, Todhunter and Purich (40) have obtained chemical evidence that y-glutamyl phosphate is formed during the synthesis of L-glutamine catalyzed by glutamine synthetase from E. coli. The similarity in the partial reactions catalyzed by glutamine synthetases from different sources (2, 3, 13, 30) and the interaction of the different glutamine synthetases with the transition state analog L-methionine-S-sulfoximine (2, 30, 32, 35-38) suggests that the catalytic sites of these enzymes are quite ~imilar.~ Despite the similarities in catalytic mechanism between glutamine synthetase from E. coli and glutamine synthetases from other sources, there are distinct differences. The enzyme from E. coli is a dodecamer (3,5) rather than an octamer (30) and is regulated by adenylylation-deadenylylation (3, 4, 10) which alters the divalent cation requirement and the affinity for nucleotide in Reaction 1. Nevertheless, the AMP-supported reactions catalyzed by the unadenylylated, manganese enzyme from E. coli, which are described here, are analogous to reactions catalyzed by glutamine synthetases from other sources (30) if AMP rather than ADP occupies the substrate nucleotide site.