Structure and function of the glutamine phosphoribosylpyrophosphate amidotransferase glutamine site and communication with the phosphoribosylpyrophosphate site.

Glutamine phosphoribosylpyrophosphate (PRPP) amidotransferase from Escherichia coli exhibits a basal PRPP-independent glutaminase activity having a kcat/Km that is 0.3% of fully active enzyme. Binding of PRPP activates the enzyme by a structural change that lowers the Km for glutamine 100-fold and couples glutamine hydrolysis to synthesis of 5-phosphoribosylamine. By analysis of the x-ray structure of the glutamine site containing bound 6-diazo-5-oxonorleucine, a glutamine affinity analog, and by site-directed mutagenesis we have identified residues important for glutamine binding, catalysis, and coupling with PRPP. Tyr74 is a key residue in the coupling between the sites for glutamine in the NH2-terminal domain and PRPP in the COOH-terminal domain. Arg73 and Asp127 have roles in glutamine binding. The x-ray structure indicates that there are no amino acid side chains sufficiently close to Cys1 to participate as a proton acceptor in formation of the thiolate needed for nucleophilic attack on the carboxamide of glutamine, nor as a general acid for amide nitrogen transfer. Based on the x-ray model of the glutamine site and analysis of a mutant enzyme we propose that the free NH2 terminus of Cys1 functions as the proton acceptor and donor. The results indicate that the side chain of Asn101 and the backbone nitrogen of Gly102 function to stabilize a tetrahedral oxyanion resulting from attack of Cys1 on the glutamine carboxamide. Cys1, Arg73, Asn101, Gly102, and Asp127 are conserved in the NH2-terminal domain of a subfamily of amidotransferases that includes asparagine synthetase, glucosamine 6-phosphate synthase, and glutamate synthase, implying a common function in the four enzymes. Tyr74, on the other hand, is conserved only in glutamine PRPP amidotransferase sequences consistent with a specific role in interdomain coupling. The catalytic framework of key glutamine site residues supports the assignment of glutamine PRPP amidotransferase to a recently described Ntn (NH2-terminal nucleophile) hydrolase family of enzymes.

Glutamine PRPP 1 amidotransferase catalyzes the initial reaction in de novo purine nucleotide biosynthesis and is the key regulatory enzyme in the pathway. Adenine and guanine nucleotides bind to an allosteric A-site and catalytic C-site and inhibit enzyme activity Zhou et al., 1994). X-ray structures have been obtained for the inhibited Bacillus subtilis enzyme  and the homologous enzyme from Escherichia coli. 2 Glutamine PRPP amidotransferase amino acid sequences are available from 18 prokaryotic and eukaryotic organisms. Although pairwise identity is 40% or more, indicative of homology, the sequences fall basically into two groups. Glutamine PRPP amidotransferases from human (Brayton et al., 1994), rat (Iwahana et al., 1993), chicken (Zhou et al., 1990), Drosophila (Clark, 1994), plants (Kim et al., 1995b;Ito et al., 1994), Mycobacterium leprae (accession number MLU15182), cyanobacteria (accession number U33211), and Lactobacillus (Gu et al., 1992) contain propeptide sequences preceding Cys 1 and four conserved cysteine residues which are ligands to a [4Fe-4S] cluster in the B. subtilis enzyme. This group of enzymes thus most closely resembles the glutamine PRPP amidotransferase from B. subtilis. Enzyme sequences from yeast (Mä ntsälä and Zalkin, 1984b;Ludin et al., 1994), Neurospora, 3 and Haemophilus (accession number U32800) are similar to E. coli glutamine PRPP amidotransferase in lacking the propeptide and cysteinyl ligands to an Fe-S cluster. A sequence from Caenorhabditis elegans (accession number CET04A8) lacks a propeptide but contains the four cysteine residues that are Fe-S ligands in the B. subtilis enzyme. Whether the conserved cysteine residues are used as ligands to an Fe-S cluster is not known. Based on the x-ray structures of the enzymes from B. subtilis and E. coli the amino acids thought to be critical for catalysis and feedback regulation are conserved in all of the enzymes. It is therefore our working hypothesis that the mechanisms for catalysis and inhibition by nucleotide end products are highly similar in all of the enzymes.
Glutamine PRPP amidotransferase catalyzes the reaction: glutamine ϩ PRPP 3 PRA ϩ glutamate ϩ PP i . The enzyme also has the capacity to use NH 3 in place of glutamine in vitro (Messenger and Zalkin, 1979) and in vivo (Mä ntsälä and Zalkin, 1984a). In addition, a glutaminase activity was detected that was less than 5% of the biosynthetic rate (Messenger and Zalkin, 1979;Kim et al., 1995a). From the x-ray structure of the nucleotide-inhibited enzyme there appear to be at least two barriers to catalysis . First, the inhibited enzyme is in a conformation described as "open" in which the glutamine and PRPP sites are too far apart for catalysis. Glutamine binds to a site in the NH 2 -terminal domain (N-domain) and PRPP to a catalytic site (C-site) in the CO 2 H-terminal domain (C-domain). Second, a structural change is required in the glutamine site that enables Cys 1 to attack the carboxamide of glutamine and initiate amide transfer. Given the requirement of PRPP for reaction of glutamine (hydrolysis and amide transfer) and for affinity labeling of Cys 1 by glutamine analogs (Messenger and Zalkin, 1979), the binding of PRPP must in some way activate the enzyme permitting Cys 1 to initiate glutamine amide transfer.
In the present work we have determined the x-ray structure of the E. coli glutamine PRPP amidotransferase glutamine site labeled by DON, a glutamine affinity analog. We have investigated the roles of pertinent amino acids in the glutamine site to determine how binding of PRPP to the C-site in the C-domain is communicated to the glutamine site in the N-domain. Residues important for the binding and reaction of glutamine and for amide transfer to PRPP were identified.

EXPERIMENTAL PROCEDURES
Plasmids and Strains-Plasmid pT7F1 containing the E. coli purF gene under the control of the T710 promoter and with the M13 replication origin was constructed for overexpression and mutagenesis. The plasmid was constructed in two steps. First, an NdeI-HindIII fragment containing a 3Ј segment of the gene was ligated into the NdeI and HindIII sites of pT7-SCA (U. S. Biochemical Corp.). Next, an NdeI site was inserted at the ATG initiation codon of purF by polymerase chain reaction. The DNA template was pGZ13  linearized by PstI and the primers were 5Ј-GGAAAAAGACATATGTGCGGTA-3Ј and 5Ј-GCCGTTGTGGGCAAGCGT-3Ј. The resulting polymerase chain reaction product, digested with NdeI, was ligated into the NdeI site of the intermediate plasmid described above containing the 3Ј segment of purF. The intact purF gene was confirmed by DNA sequencing.
Enzyme Purification-Wild type E. coli glutamine PRPP amidotransferase and mutant enzymes were purified from plasmid pT7F1 and its derivatives in E. coli BL21(DE3) as described (Zhou et al., 1993). Plasmid-bearing E. coli cells were grown at 37°C in LB medium (Miller, 1972) containing 140 g/ml ampicillin. The purF gene was induced by 0.2 mM isopropyl-1-thio-␤-D-galactopyranoside at midlog phase and growth was continued for 2-3 h. Enzyme was overproduced to approximately 30 -40% of total soluble cell proteins. The enzyme was purified to approximately 95% homogeneity as judged by sodium dodecyl sulfate-polyacrylamide gel electrophoresis.
Enzyme Assay-Enzyme activity was assayed by measurement of product formation, either glutamate or PRA. The production of glutamate from glutamine reflects a glutaminase activity. Total glutaminase refers to the activity that is dependent upon PRPP plus the activity that is PRPP independent. The standard assay for total glutaminase activity contained 2.5 mM PRPP (Sigma), 20 -220 mM glutamine depending on the K m , 10 mM MgCl 2 , 1 mg/ml bovine serum albumin, 50 mM Tris-HCl (pH 8.0), and approximately 100 ng of enzyme in a total volume of 100 l. Incubation was at 37°C for 10 min. Reactions were quenched in a boiling water bath for 1 min and glutamate was determined by the glutamate dehydrogenase method (Messenger and Zalkin, 1979). A basal PRPP-independent glutaminase activity was determined by the same procedure except for omission of PRPP in the reaction. PRPP-dependent glutaminase activity was calculated by subtracting the PRPPindependent activity from the total glutaminase activity. Control reactions contained all components except enzyme. Glutamine-dependent PRA synthesis was assayed by coupling the formation of PRA to the production of glycinamide ribonucleotide with glycinamide ribonucleotide synthetase (Schendel et al., 1988). Purified glycinamide ribonucleotide synthetase (Shen et al., 1990) was the generous gift of J. Stubbe, Massachusetts Institute of Technology, Cambridge, MA. The reactions contained in a volume of 40 l, 6 mM PRPP, 20 mM glutamine, 2.5 mM ATP, 2 mM [ 14 C]glycine (1800 cpm/nmol), 10 mM MgOAc, 50 mM Tris-HCl (pH 8.0), and approximately 100 ng of enzyme. Incubation was for 6 min at 37°C. NH 3 -dependent PRA synthesis was assayed by the same procedure except 150 mM NH 4 Cl and Tris-HCl (pH 8.5) replaced glutamine and Tris-HCl (pH 8.0), respectively. Glycinamide ribonucleotide was isolated from the reaction mixture by ion exchange chromatography and was counted for radioactivity (Schendel et al., 1988).
To determine whether mutant enzymes retained glutamine-or NH 3dependent activity in vivo, we measured the capacity of the enzyme to support growth of a purF auxotroph in medium containing either 50 mM NH 4 Cl or 1 mM NH 4 Cl as nitrogen source (Mei and Zalkin, 1989). Growth with 1 mM NH 4 Cl requires assimilation into glutamine and use of glutamine for PRA synthesis by glutamine PRPP amidotransferase, whereas 50 mM NH 4 Cl will suffice for the NH 3 -dependent activity of the enzyme. The wild type or mutant purF gene was cloned into the XbaI site of pTrc99A (Egon et al., 1988) and the recombinant plasmid transformed into E. coli strain TX358 (purF recA). The growth rate of plasmid-harboring cells was determined in M9 medium (Miller, 1972) supplemented with 0.5% lactose, 0.01% acid-hydrolyzed casein, 2 g/ml thiamin, 100 g/ml ampicillin, and 1 or 50 mM NH 4 Cl.
Determination of K m and V m -The kinetic constants K m and V m were determined using the standard assays in which the concentration of one substrate was saturating and the concentration of the second substrate was varied. The concentration range for glutamine in the different glutaminase assays was adjusted for specific mutants: 0 -220 mM for PRPP-independent glutaminase of wild type and all mutants, 0 -20 mM for total glutaminase of enzymes in which the K m was Ͻ3 mM (Ala insertion, N101D, N101G, G102A), 0 -220 mM for total glutaminase of mutants in which the K m was Ͼ15 mM (Arg 73 , Tyr 74 , and Asp 127 mutants). For assays with the low range of glutamine concentrations, two concentrations over 100 mM were also tested to verify saturation. For assays of enzymes having a K m of Ͼ100 mM for glutamine, the lowest standard errors were obtained using two glutamine concentrations Ͻ100 mM and about six concentrations between 100 and 220 mM. The maximum concentration of glutamine was limited to 220 mM by its solubility. The substrate concentrations used for other assays were: 0 -130 mM glutamine for glutamine-dependent PRA synthesis, 0 -150 mM NH 4 Cl for NH 3 -dependent PRA synthesis, and 0 -540 M PRPP for total glutaminase. Kinetic data were fit to the Michaelis-Menten equation using Ultrafit software (Biosoft, Cambridge, UK).
Nucleotide Inhibition-Inhibition of glutaminase activity by AMP and GMP was determined by the standard assay except that the MgCl 2 concentration was increased to 20 mM, PRPP was decreased to 1 mM and 0 -15 mM AMP or 0 -3 mM GMP were included. For the wild type enzyme, 20 mM glutamine was used for nucleotide inhibition of the PRPP-dependent glutaminase activity and 200 mM glutamine for the PRPP-independent or total glutaminase. In the case of Arg 73 and Tyr 74 mutants, 200 mM glutamine was used for nucleotide inhibition of all glutaminase assays.
Equilibrium Dialysis-Nucleotide binding was determined by equilibrium dialysis  using chambers of 100 l that were separated by a 12,000 -14,000 molecular weight cut off dialysis membrane (Spectraphore). One chamber contained 200 mM Tris-HCl (pH 7.5), 20 mM MgCl 2 , 35 nM [8-3 H]GMP (0.1 Ci), 0 -2.0 mM unlabeled GMP, and 300 mM AMP in a total volume of 50 l. The other chamber contained 10 mM Tris-HCl (pH 7.5) and approximately 200 M enzyme subunit in a total volume of 50 l. Dialysis was for 20 h at 4°C in a rotating apparatus (Hoefer). Samples of 40 l were retrieved from each chamber and were counted for radioactivity. Equilibrium binding data were fit to the Scatchard equation for nonlinear regression using Ultrafit. Under these conditions GMP is expected to bind to the A-site and AMP to the C-site Kim et al., 1995a).
Affinity Labeling by DON and Crystallization-Approximately 20 mg of enzyme was incubated at room temperature in 2.0 ml of a mixture containing 50 mM potassium phosphate (pH 7.5), 3 mM PRPP, 5 mM MgCl 2 , and 3 mM DON. After reaction for 30 min, enzyme was dialyzed against 10 mM Tris-HCl (pH 7.5). PRPP-dependent glutaminase activity was below the limit of detection whereas the NH 3 -dependent synthesis of PRA was 38.5 units/mg, not significantly different from the activity of the untreated enzyme (43 units/mg). DON-inactivated glutamine PRPP amidotransferase was crystallized at 20°C by sittingdrop vapor diffusion. Drops containing 4 l of enzyme (18.5 mg/ml) and 4 l of well solution were equilibrated against a well solution of 13% polyethylene glycol 3350, 5% 2-propanol, 100 mM PIPES, 4 mM MgCl 2 , and 4 mM MgSO 4 at pH 6.3. Crystals were transferred to an equivalent solution containing 12% polyethylene glycol, 10% 2-propanol, and 15% DL-meso-2,3-butanediol 10 min prior to freezing.
X-ray Structure Determination of DON-inactivated Enzyme-The structure of native E. coli glutamine PRPP amidotransferase has recently been determined with x-ray crystallography. 2 DON-inactivated glutamine PRPP amidotransferase was crystallized under conditions identical to those used in structure determination of the native protein.
The DON-enzyme crystals are isomorphous to those of the native enzyme (space group C222 1 , a ϭ 116.9 Å, b ϭ 157.5 Å, c ϭ 106.3 Å, with two monomers of the tetrameric enzyme per asymmetric unit). X-ray diffraction data to 2.3 Å were collected from a single crystal, flash frozen to 120 K with an Oxford Cryostream, using an R axis II imaging plate system mounted on a Rigaku RU-200 rotating anode (CuK␣) operated at 100 mA and 50 kV. The data were processed and scaled using DENZO and Scalepack (Otwinowski, 1993). Statistics for the resulting data are shown in Table II. An initial difference electron density map was calculated using ԽFo(DON-enzyme)Խ Ϫ ԽFo(native)Խ and phases from the refined native structure. This indicated that differences were confined to the glutamine active site, with significant shifts of protein atoms being small, and limited to the Asp 127 side chain and Cys 1 sulfhydryl. These shifted side chain atoms and water molecules within 10 Å of Cys 1 were removed from the refined native model. The truncated model was then refined against the DON-enzyme data in X-PLOR (Brü nger, 1992) with 10 steps of isotropic temperature factor refinement and 20 steps of positional refinement. An ԽFoԽ Ϫ ԽFcԽ electron density map calculated from the resulting model showed clear density for the DON atoms. The initial model of the DON-Cys 1 adduct fit this unbiased density remarkably well, demonstrating the accuracy of both the density and the model. This electron density is shown with the current refined model of the DON-enzyme active site in Fig. 3, which was drawn with Molscript (Kraulis, 1991) and Raster 3D (Bacon and Anderson, 1988;Merritt and Murphy, 1994). The completed model of two monomers was refined independently in X-PLOR, with positional and restrained individual temperature factor refinement.
Analyses-NH 2 -terminal amino acid sequencing was carried out on an Applied Biosystems gas-phase sequenator model 470 using standard operating procedures. The mass spectrum of the glutamine PRPP amidotransferase wild type and Ala insertion mutant was determined by matrix-assisted laser desorption ionization mass spectrometry. By these methods wild type enzyme was not detected in the preparation of the Ala insertion. However, the limit of detection, although not precisely determined, was greater than 5%.

RESULTS AND DISCUSSION
Effect of PRPP on Glutamine Hydrolysis-Glutamine hydrolysis is markedly dependent upon PRPP and is tightly coupled to synthesis of PRA. Under the assay conditions used previously (10 mM glutamine), the basal rate of glutamine hydrolysis in the absence of PRPP was determined using conditions appropriate for the reaction with PRPP; the rate was between 0.8 and 4% of the rate with PRPP (Messenger and Zalkin, 1979;Kim et al., 1995a). To understand catalysis, it is necessary to determine how PRPP activates glutamine hydrolysis and how the amide of glutamine reacts with PRPP to form PRA. We have now found a higher rate of basal PRPP-independent glutaminase having a V m of approximately 10 units/mg and K m of 190 mM for glutamine ( Fig. 1 and Table I). At high nonphysiological concentrations of glutamine the basal activity was 35% of the PRPP-dependent glutaminase. Values of k cat /K m were 47 and 14,980 M Ϫ1 s Ϫ1 for basal and PRPP-dependent glutaminase, respectively. The data in Table I indicate that the effect of PRPP is to lower the K m for glutamine by over 100-fold and increase the V max approximately 3-fold.
The basal glutaminase activity was inhibited by GMP and AMP (Fig. 2) indicating that the active site for this reaction is the same as that for PRPP-dependent hydrolysis of glutamine. The nucleotide concentration required for 50% inhibition was lower for reactions in which PRPP was omitted. These data support the idea that the competitive relationship between PRPP and nucleotide inhibitors (Messenger and Zalkin, 1979) results from alternative active and inactive enzyme conformations . The basal activity in the absence of PRPP thus reflects enzyme in a partially active conformation that can be further activated by PRPP or inhibited by nucleotides.
In spite of the basal glutaminase activity, production of phosphoribosylamine was still tightly coupled to glutamine hydrolysis because PRPP-independent glutaminase activity was insignificant at glutamine concentrations of 20 mM or lower (Fig.  1). Thus, at concentrations of glutamine less than 20 mM the rates of total glutaminase determined by glutamate production  and glutamine-dependent synthesis of PRA were similar. The rate of PRA synthesis with NH 3 as nitrogen donor was 2.1 times the rate with glutamine as substrate (Table I). This is similar to a ratio of 2.8 determined previously by a different method (Messenger and Zalkin, 1979). The results indicate that the capacity for phosphoribosylamine synthesis exceeds that for glutamine hydrolysis.
Structure of the Glutamine Site-The current refined model contains 992 (98.4%) of the protein residues and 942 water molecules. Statistics for this model are in Table II. The glutamine active site is in a pocket in the N-domain. The arrangement of key residues in the glutamine site labeled with DON is shown in Fig. 3 and is illustrated schematically in Fig. 4. Distances for interacting atoms are given in Table III. DON is an affinity analog of glutamine, known to alkylate Cys 1 and to mimic the putative covalent glutamyl enzyme thioester intermediate resulting after amide transfer (Zalkin, 1992). The overall structure of the glutamine site of DON-inactivated enzyme is nearly identical to that of the native enzyme (not shown). The largest shifts are 0.5 Å for Cys 1 ␥S and up to 0.7 Å for the Asp 127 side chain and may be due primarily to the covalently attached DON which has an additional carbon atom compared to a glutamine thioester intermediate. Overall, the DON structure is quite similar to the predicted structure of glutamine modeled into the site (Smith, 1995).
The DON ␣-NH 2 group demonstrates good hydrogen bonding to the carboxyl of Asp 127 . In addition, the DON ␣-NH 2 forms a strong hydrogen bond to the backbone carbonyl of Gly 102 , demonstrating that this invariant residue also functions in substrate binding. A third hydrogen bond is formed to a water molecule, completing all possible hydrogen bond interactions to the ␣-NH 2 group.
The carboxyl oxygens of DON exhibit good hydrogen bond distances to Thr 76 and three serine side chains at positions 79, 126, and 128. Of these residues Thr 76 is invariant and at positions 126 and 128 there are either serine or threonine residues in the other glutamine PRPP amidotransferase sequences. Position 79 is totally nonconserved on the other hand. These interactions, by themselves, however, may be insufficient for effective binding of glutamine. This is suggested by temperature factors for the DON carboxyl atoms that are significantly higher than neighboring atoms, indicating that they are not tightly bound. It is possible to model Arg 73 in a favorable conformation that displaces only water molecules and results in hydrogen bonding to the carboxyl of DON or to the carboxyl of glutamine modeled in the site. In this configuration Arg 73 would displace two water molecules in contact with the substrate, creating an active site that more fully encloses the bound glutamine. The proximity of Arg 73 to DON is shown in Fig. 3.
The carbonyl oxygen of DON hydrogen bonds to the backbone NH of Gly 102 and to the amide of Asn 101 . The orientation of these residues results in very favorable hydrogen bonds and supports the proposal that they constitute an oxyanion hole, stabilizing a tetrahedral glutaminyl oxyanion intermediate resulting from attack of Cys 1 thiolate on the carboxamide of glutamine (Brannigan et al., 1995).
Mutagenesis of Residues Involved in Glutamine Binding and Stabilization of the Oxyanion Intermediate-Based on the x-ray structure of the DON-labeled enzyme and native enzyme into which glutamine has been modeled, a number of amino acid residues were identified that are in proximity to the glutamine site and are positioned to participate in binding and in reaction of the glutamine amide. These amino acids include Cys 1 , Arg 26 , Arg 73 , Asn 101 , Gly 102 , and Asp 127 , residues that are invariant in the 18 known glutamine PRPP amidotransferase sequences as well as in 20 asparagine synthetase, glucosamine 6-phosphate synthase, and glutamate synthase sequences. These four enzymes are members of an Ntn glutamine amidotransferase FIG. 3. Electron density map. Refined model of the DON-labeled active site overlaid with the original unbiased difference electron density map, contoured at 3. The map was calculated using the native structure with Cys 1 S, Asp 127 carboxyl, and neighboring water molecules removed. Atoms are colored as follows: peptide C, yellow; DON C, white; N, blue; O, red; S, green. subfamily. 4 We have analyzed the effects of mutations of these amino acids as well as Tyr 74 which is conserved in glutamine PRPP amidotransferase but not in the other enzymes.
Cys 1 -The NH 2 -terminal cysteine is an active site residue required for glutamine amide transfer. In earlier work (Mei and Zalkin, 1989), it was proposed that conserved His and Asp residues function in acid-base catalysis to generate the Cys 1 thiolate for nucleophilic catalysis and in the steps of amide transfer and hydrolysis of a ␥-glutamyl thioester intermediate. More recent evidence, however, indicates that the proposed His 101 and Asp 29 side chains are too far removed from Cys 1 to participate in catalysis . Furthermore, these residues are not conserved using improved alignments with the larger group of more recently acquired sequences of Ntn amidotransferases. More importantly, there are no other amino acid side chains within hydrogen bonding distance to the ␥-sulfhydryl of Cys 1 in the x-ray structure of the E. coli enzyme that could function in acid-base catalysis. 2 In the absence of an amino acid side chain to facilitate formation of the thiolate anion needed for nucleophilic catalysis and for the subsequent step of amide transfer and the proximity of the ␣-NH 2 to the ␥-SH group of Cys 1 (shown in Fig. 4), we have investigated whether a free Cys 1 ␣-NH 2 group is required for catalysis. An alanine residue was inserted between Met Ϫ1 and Cys 1 by sitedirected mutagenesis to block the Cys 1 amino group. The determined NH 2 -terminal amino acid sequence of the purified amidotransferase was Ala-Cys-Gly-Val-Phe, indicating cleavage of the initiator Met in accord with the specificity of E. coli methionine aminopeptidase (Hirel et al., 1989). Data in Table  IV summarize the effects of the Ala insertion. Activity with glutamine was reduced about 20-fold compared to the wild type enzyme, whereas the K m for glutamine was unchanged. Al-though wild type enzyme was not detected in the preparation of the Ala insertion we cannot distinguish whether all of the residual activity is due to the Ala insertion species or whether there was 5% or less contamination by enzme with Ala clipped off. C-domain catalysis was also perturbed as indicated by the 10-fold reduction in NH 3 -dependent synthesis of PRA. Nucleotide sites in the C-domain were, however, functional as shown by the stoichiometry and K d for GMP binding. Under these conditions GMP is bound to the nucleotide A-site and AMP to the C-site (Kim et al., 1995a). Reduced glutamine-dependent activity in the Ala-Cys enzyme is consistent with a role of the Cys 1 ␣-NH 2 group as a proton acceptor for thiolate formation and proton donor for amide nitrogen transfer. The basis for the reduced activity with NH 3 as substrate is not certain, but the ␣-NH 2 of Cys 1 is close enough to the C-site that the alanine insertion may interfere sterically with the PRPP reaction. In penicillin acylase a similar role has been proposed for the ␣-NH 2 group of the NH 2 -terminal serine to generate a seryl nucleophile for covalent catalysis .
Arg 26 -Arginine 26 makes a series of H bonds with residues involved in catalysis at the glutamine site (Fig. 4). The ⑀-N and -N of Arg 26 H-bonds to the main chain oxygen of Cys 1 , an interaction that should contribute to positioning Cys 1 for catalysis. The main chain oxygen of Arg 26 hydrogen bonds with the amino group of Cys 1 , thus increasing its basicity. In addition, the 1 -amino group of Arg 26 is within H-bonding distance to the carboxamide oxygen of Asn 101 . To determine the functional consequence of these interactions two replacements of Arg 26 were made, R26H and R26A. In both cases the mutant enzymes were extremely labile and could not be purified for analysis. In vivo analysis of enzyme function was made by examining the capacity of the R26H and R26L enzymes to support growth of a purF mutant in medium containing either 50 mM NH 4 Cl or 1 mM NH 4 Cl as N-source. Cells having the R26L or R26H enzymes grew at the wild type rate (2-h doubling time) with 50 mM NH 4 Cl but growth was severely limited (10-h doubling) in medium with 1 mM NH 4 Cl. This indicates that the glutaminedependent activity was disabled in these mutants but there was sufficient NH 3 -dependent activity for wild type growth. An R26K mutant was reported previously to have little or no 4 Glutamine amidotransferase subfamilies were previously named G-type (Class 1) and F-type (Class II) after prototype E. coli TrpG and PurF enzymes, respectively. With x-ray structures now available, the nomenclature has been revised to reflect the defining structural features of the glutamine binding/amide transfer domain required for catalysis. The key features are an NH 2 -terminal nucleophile (Ntn) (Brannigan et al., 1995) and a catalytic triad (Tesmer et al., 1996). The G-type (Class I) enzymes are in the Triad subfamily and F-type (Class II) enzymes in the Ntn subfamily.  Table III. glutamine-dependent activity and to be subject to rapid proteolysis in vivo (Mei and Zalkin, 1989). Overall, the results indicate that Arg 26 has roles maintaining a functional glutamine site and in structural integrity of the enzyme, in accord with the participation of Arg 26 in an extensive network of salt bridges within the N-domain. Asn 101 and Gly 102 -Nucleophilic attack of Cys 1 on the carboxamide of glutamine is postulated to lead to a transient oxyanion tetrahedral adduct that would be stabilized by interactions with the amide of Asn 101 and the backbone NH of Gly 102 . N101G, N101D, and G102A replacements were made for these residues of the putative oxyanion hole and the results are shown in Table IV. The N101G mutation had only a marginal 4-fold reduction in V m for glutamine-dependent activity while the K m for glutamine, V m and K m for the NH 3 -dependent reaction, and nucleotide binding to regulatory sites were indistinguishable from the wild type. Inspection of the structure of the glutamine site (Fig. 3) suggests that Arg 188 may occupy the space of Asn 101 in the glycine replacement and thus provide surrogate oxyanion stabilization that could account for the residual 23% activity with glutamine. With a N101D mutation the V m for the glutamine reaction was reduced more than 2,500-fold, possibly as a result of repulsion between the carboxylate and the oxyanion species. The G102A replacement shows that Gly 102 is critical for activity with glutamine but not for glutamine binding, NH 3 -dependent PRA synthesis, nor nucleotide binding. The structure of Gly 102 indicates , angles that are highly unfavorable for alanine. Therefore, the G102A mutation is likely to interfere with positioning of the backbone NH needed to stabilize an oxyanion intermediate.
Arg 73 and Asp 127 -Binding of DON in the glutamine site involves hydrogen bonds between the carboxyl of Asp 127 and the backbone oxygen of Gly 102 with the ␣-amino group and by hydrogen bonds between the O␥ of serines 79, 126, and 128 and Thr 76 with the carboxyl group (Fig. 4). The side chain of Arg 73 although not within hydrogen bonding distance to DON in the x-ray structure of the PRPP-free enzyme, nevertheless, is the only positively charged group capable of interacting with the DON carboxyl without backbone reorientations. Replacements were made for Gly 102 , Arg 73 , and Asp 127 . The K m for glutamine was similar to the wild type value for the G102A enzyme (Table  IV). In the absence of an increase in glutamine K m for the G102A mutant there is no biochemical evidence to support a role of the Gly 102 main chain oxygen in glutamine binding. Two effects were noted for Arg 73 and Asp 127 replacements and are seen in Table IV. K m values for the glutamine reaction were increased 137-fold in the Asp 127 mutant and about 50-fold in the two Arg 73 mutants relative to the wild type. These increased K m values are consistent with roles for Asp 127 and Arg 73 in glutamine binding. However, as shown in Table IV there were also significant reductions in V m for the glutamine reaction, 46-fold in the D127A enzyme and 15-fold in the two arginine mutants. The decreased V m is restricted to glutamine site function since the NH 3 -dependent activity and nucleotide binding were similar to the wild type. The basis for the decreased rate of glutamine site catalysis for Arg 73 mutants, as explained in the section on "Model for the Glutamine Site, Activation by PRPP, and Inhibition by Nucleotides" may be a structural perturbation that impacts on catalysis.
Tyr 74 -Binding of PRPP to the C-site is required to activate Cys 1 for reaction with glutamine and amide transfer. Tyr 74 is the closest N-domain residue to PRPP modeled into the C-site of the nucleotide inhibited enzyme (see Fig. 1 in Kim et al. (1995a)) and by virtue of its proximity to Cys 1 could participate in coupling PRPP binding to glutamine hydrolysis and amide transfer. Tyr 74 is conserved in all of the glutamine PRPP amidotransferase sequences, but not in the other Ntn amidotransferases. In order to investigate whether Tyr 74 functions to couple PRPP binding to the reaction of glutamine we examined glutamine site function and PRA synthesis in three Tyr 74 replacements, Y74S, Y74L, and Y74F. Data summarized in Table  V, although complex, establish that Tyr 74 exerts an important role in coupling glutamine site catalysis with synthesis of PRA. The glutamine site was characterized by two assays. First, a basal glutaminase activity was measured. Second, we determined the capacity of PRPP to activate the basal glutaminase by increase in V m and reduction in glutamine K m . The extent of PRPP activation can be evaluated by comparing basal and total glutaminase in Table V. "Total glutaminase" is given in Table  V rather than PRPP-dependent glutaminase in order to avoid the technical difficulty of calculating a small PRPP-dependent activity in the presence of a high background of PRPP-independent basal activity, a particular problem in the Tyr 74 mutants.
There are four important results that were obtained from the Tyr 74 replacements. (i) The basal glutaminase was perturbed. Activity was decreased 3-fold in Y74S and increased nearly 5-fold relative to wild type in the Y74L enzyme. As shown in Tables IV and V, Arg 73 is critical for glutamine site function and the Tyr replacements may interact with Arg 73 differently. The perturbation of basal glutaminase in Tyr 74 mutants was confined to V m ; K m was not affected. (ii) There was little or no stimulation of basal glutaminase by PRPP in the Tyr 74 mutants. This is shown by similar V m values for basal and total glutaminase in the Y74S and Y74L enzymes. (iii) Little, if any, of the glutaminase was coupled to PRA synthesis. There was apparent 2-fold stimulation of glutaminase by PRPP in the Y74F enzyme. However, only one-third of this glutaminase activity was available for PRA synthesis. There is thus clear evidence for defective coupling of glutamine hydrolysis with PRA synthesis. The K m for PRPP, although technically difficult to determine due to the high basal glutaminase, appeared to be  Table V also show the relationship between functions of Tyr 74 and Arg 73 . In the two Arg 73 mutants the primary defect was impaired binding of glutamine as reflected by the unmeasurable K m for basal glutaminase. There was activation by PRPP which led to low but detectable total glutaminase having a high K m . Most importantly, the resulting glutamine hydrolysis was coupled to PRA synthesis and NH 3dependent PRA synthesis was comparable to wild type. When R73L and Y74L mutations were combined there was no stimulation of basal glutaminase by PRPP and no coupling of glutaminase with PRA synthesis. NH 3 -dependent PRA synthesis was decreased marginally from the wild type level. Thus Arg 73 functions mainly in glutamine binding and catalysis and Tyr 74 in modulating glutamine site catalysis with the availability of PRPP and coupling synthesis of PRA to nitrogen from glutamine and from NH 3 .
Nucleotide Inhibition-Since Tyr 74 has a role in communicating the binding of PRPP in the C-site to glutamine site catalysis in the N-domain, we investigated whether Tyr 74 was also involved in communicating binding of nucleotides to the N-domain. Nucleotide inhibition of Arg 73 mutants was also examined because of the importance of this residue on glutamine site function and the possibility of coordinate effects of residues at positions 73 and 74. Inhibition of total glutaminase by nucleotides is shown in Fig. 5 for the wild type and three mutant enzymes. Y74L and R73L enzymes were each somewhat less sensitive to inhibition by AMP. The R73L but not the Y74L mutant was less sensitive than the wild type enzyme to inhibition by GMP. Similar results, not shown, were obtained for the Y74H, Y74F, and R73H enzymes. Most importantly, however, the R73L/Y74L double mutant was essentially insensitive to inhibition by concentrations of nucleotides that gave 95-100% inhibition of the wild type enzyme.
Binding of GMP to the double mutant was measured in order to determine whether loss of inhibition resulted from a binding defect or from loss of communication by bound nucleotide to the glutamine site. GMP binding in the presence of unlabeled AMP is shown in Fig. 6. The GMP binding stoichiometry extrapolated to 1.25 Ϯ 0.19 equivalents with a K d of 221 Ϯ 76 M. This K d for GMP binding to the mutant is similar to that for binding to the wild type enzyme (Kim et al., 1995a) and indicates that nucleotide binding is not communicated to the glutamine site in the feedback-insensitive R73L/Y74L enzyme.
Model for the Glutamine Site, Activation by PRPP, and Inhibition by Nucleotides-E. coli glutamine PRPP amidotrans-ferase is a tetramer with each subunit organized in an Ndomain, amino acids 1-251, and a C-domain, amino acids 252-504. Two half-sites constitute the active site. Glutamine binds to the N-domain and it is thought that Cys 1 attacks the car-  boxamide of glutamine to initiate amide transfer. The amide nitrogen of glutamine is transferred to PRPP bound to the second half-site (C-site) to yield PRA ϩ PP i and a putative ␥-glutamyl thioester enzyme intermediate. Affinity labeling by DON results in a DON-cysteinyl adduct that is a model for the putative ␥-glutamyl thioester intermediate. C-site residues 362-374 constitute a conserved PRPP fingerprint. Although a primary function of the C-site is binding PRPP, it has features in common with mononucleotide sites and is a binding site for AMP or GMP. Four intersubunit allosteric sites (A-sites) in the tetramer also bind nucleotides. Synergistic binding of AMP and GMP to the C-and A-sites, respectively, results in synergistic inhibition .
We can distinguish three activity states of the enzyme, basal, active, and inhibited, that differ by the occupancy of the A-and C-sites. Nucleotide-free enzyme having no PRPP in the C-site can bind glutamine with low affinity and catalyze its hydrolysis at a rate up to about 35% of the fully activated rate. This form of the enzyme having unoccupied A-and C-sites thus defines a basal activity that can either be activated by PRPP or inhibited by nucleotides. The kinetically competitive relationship between PRPP and nucleotides (Messenger and Zalkin, 1979) thus results from alternative enzyme conformations mediated by the ligands. In the presence of PRPP the K m for glutamine is decreased by 100-fold and the V m for reaction increased by 3-fold. In this active state PRPP is the C-site ligand and the A-site is unoccupied.
How does PRPP activate the enzyme? At least two structural changes must take place. First, it is plausible to imagine that binding of PRPP to the C-site repositions Arg 73 for high affinity binding of glutamine. Replacements of Arg 73 and also Asp 127 , which participates in glutamine binding, led to significantly higher K m values for glutamine. Once the substrate is precisely positioned by the Arg 73 interaction, nucleophilic attack by Cys 1 on the carboxamide can initiate amide transfer. Second, it is likely that the 16-Å space between the glutamine half-site and PRPP in the C-site must close to permit nitrogen transfer. Although it is not possible to explain how these changes take place without knowledge of the x-ray structure of the fully active enzyme, Tyr 74 has a role in communicating PRPP binding to the glutamine site. Replacements of Tyr 74 perturb the binding and reaction of glutamine and amide transfer.
Binding of nucleotides to the A-and C-sites converts the enzyme to the inhibited state. A "flag-loop" structure from the C-domain of one subunit is packed tightly against the N-domain of a neighboring subunit and through interactions with nucleotide in the A-site appears to hold apart the subunit's Nand C-domains stabilizing the inactive open conformation . It also appears that Arg 73 may have a role in inhibition in addition to its role in glutamine binding, since a replacement of Arg 73 augmented the effect of a Tyr 74 mutation and desensitized the enzyme to nucleotide inhibition. Tyr 74 and Arg 73 are involved in transmitting the inhibitory nucleotide binding signal to the glutamine site. An x-ray structure is needed for the active enzyme to precisely describe the changes in conformation.
Relationships to Asparagine Synthetase, Glucosamine-6phosphate Synthase, and Glutamate Synthase-Glutamine PRPP amidotransferase, asparagine synthetase (Van Heek and Schuster, 1989), glucosamine 6-phosphate synthase (Badet et al., 1987), and glutamate synthase (Oliver et al., 1987) all use an N-terminal active site cysteine for glutamine amide transfer and contain N-domains having nine conserved amino acids. Boehlein et al. (1994) have analyzed replacements of Arg 30 and Asn 74 in E. coli asparagine synthetase, residues that correspond to Arg 26 and Asn 101 in E. coli glutamine PRPP amido-transferase. A number of roles were suggested for this Arg residue including binding the glutamine substrate, mediating communication between separate half-reactions involved in asparagine formation, and controlling nitrogen release from glutamine. Suggested roles for Asn 74 included stabilization of an oxyanion intermediate and stabilization of a hypothetical hydroxyimine tatuomeric species of glutamine substrate. It is likely that asparagine synthetase Arg 30 and Asn 74 as well as other conserved amino acids in the N-domains of asparagine synthetase, glucosamine 6-phosphate synthase, and glutamate synthase define the glutamine site in these enzymes and have similar functions in the four enzymes.
Ntn Hydrolases-A family of enzymes having an NH 2 -terminal nucleophilic side chain, a common active site organization, and a common fold was identified recently and named the Ntn hydrolase family (Brannigan, 1995). Glutamine PRPP amidotransferase is an Ntn hydrolase and data reported here support the mechanisms proposed for the enzyme family. In addition to an NH 2 -terminal nucleophile, Cys, Ser, or Thr, the catalytic framework includes a free NH 2 terminus and an oxyanion hole. Analysis of the glutamine PRPP amidotransferase Ala insertion mutant supports the proposed role of a free NH 2 terminus as a general acid-base (Brannigan et al., 1995). The structure of glutamine PRPP amidotransferase-DON adduct and the properties of Asn 101 and Gly 102 mutants show that these residues likely constitute the oxyanion hole for glutamine hydrolysis, as predicted. Interestingly, the binding of DON to glutamine PRPP amidotransferase (Fig. 3) is very similar to the binding of aspartate to the Ntn hydrolase, aspartylglucosaminidase (Oinonen et al., 1995). Interactions of Gly 235 , Arg 211 , and Asp 214 with aspartate in aspartylglucosaminidase correspond to Gly 102 , Arg 73 , and Asp 127 interactions with DON in glutamine PRPP amidotransferase. The common ligand interactions in glutamine PRPP amidotransferase and other Ntn hydrolases support the proposed roles of specific glutamine PRPP amidotransferase active site residues and the idea that these are homologous enzymes.