Function of Conserved Residues of Human Glutathione Synthetase

Glutathione synthetase is an enzyme that belongs to the glutathione synthetase ATP-binding domain-like superfamily. It catalyzes the second step in the biosynthesis of glutathione from γ-glutamylcysteine and glycine in an ATP-dependent manner. Glutathione synthetase has been purified and sequenced from a variety of biological sources; still, its exact mechanism is not fully understood. A variety of structural alignment methods were applied and four highly conserved residues of human glutathione synthetase (Glu-144, Asn-146, Lys-305, and Lys-364) were identified in the binding site. The function of these was studied by experimental and computational site-directed mutagenesis. The three-dimensional coordinates for several human glutathione synthetase mutant enzymes were obtained using molecular mechanics and molecular dynamics simulation techniques, starting from the reported crystal structure of human glutathione synthetase. Consistent with circular dichroism spectroscopy, our results showed no major changes to overall enzyme structure upon residue mutation. However, semiempirical calculations revealed that ligand binding is affected by these mutations. The key interactions between conserved residues and ligands were detected and found to be essential for enzymatic activity. Particularly, the negatively charged Glu-144 residue plays a major role in catalysis.

Glutathione synthetase (1)(2)(3)(4) catalyzes the second and final step in the biosynthesis of glutathione (GSH) 1 from ␥-glutamylcysteine and glycine in an ATP-dependent manner. This process involves formation of an enzyme-bound acyl phosphate (␥-glutamylcysteinyl phosphate), followed by attack of the glycine and formation of an enzyme-product complex, which finally dissociates with the release of GSH, ADP, and phosphate (P i ), as shown in Reaction 1. Glutathione is present in the majority of living cells and is also the most abundant intracellular thiol. It has a number of vital functions: it protects cells against oxidative damage, facilitates the formation of deoxyribonucleotides, reacts with toxic compounds, participates as a coenzyme for enzymes such as glyoxalase (5) and glutathione-dependent formaldehyde dehydrogenase (6). Glutathione is also involved in amino acid transport, in metabolism of therapeutic drugs, mutagens, and carcinogens, and in the maintenance of protein thiol groups and ascorbic acid in its reduced form (7). Lowered levels of glutathione have been associated with some diseases, for example, human immunodeficiency, hepatitis C, type II diabetes, ulcerative colitis, idiopathic pulmonary fibrosis, adult respiratory distress syndrome, and cataracts (7). Substantial attention has been given to human glutathione synthetase because of the biological implications for human patients with hereditary glutathione synthetase deficiency (8). In generalized glutathione synthetase deficiency, lowered levels of GSH induce an overproduction of ␥-glutamylcysteine due to the lack of feedback inhibition of ␥-glutamylcysteine synthetase by GSH. Even though ␥-glutamylcysteine can compensate for GSH in many aspects of cellular defense against oxidative stress (9), the increased amounts of ␥-glutamylcysteine lead to accumulation of 5-oxoproline (8,10,11). On the basis of clinical symptoms, patients with glutathione synthetase deficiency can be classified into three phenotypes: mild, moderate, and severe (or generalized). Patients with mild glutathione synthetase deficiency have hemolytic anemia as their only clinical symptom. Those with moderate glutathione synthetase deficiency usually display symptoms starting from the neonatal period, i.e. metabolic acidosis, 5-oxoprolinuria, and hemolytic anemia. Those with severe glutathione synthetase deficiency also develop progressive neurological symptoms such as seizures and psychomotor retardation (12). The severe form of glutathione synthetase deficiency is caused by mutations in the coding sequence of human glutathione synthetase that lead to a reduction of enzyme activity (13). However, studies on patients affected by this genetic disorder indicate a residual activity of glutathione synthetase, suggesting that a complete loss of its function is probably lethal (14,15).
Glutathione synthetase has been purified and sequenced from a variety of sources (2, 16 -23). The first highly purified mammalian glutathione synthetase was isolated from rat kidney in 1979 (24) and then was cloned and sequenced in 1995 (25). In the same year 1995, Gali et al. (26) reported the amino acid sequence for human glutathione synthetase. Currently, * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. the rat and human forms are the most studied mammalian glutathione synthetases. The sequence analysis of the glutathione synthetase enzymes showed that human glutathione synthetase is very similar to rat glutathione synthetase, sharing 88.6% identity, whereas the amino acid sequence comparison to other eukaryotic enzymes revealed a lower identity (18.2-68.8%) (23,27).
The most studied glutathione synthetase enzyme is from Escherichia coli (28 -36), which only has 10% sequence identity with the human enzyme (37). Furthermore, both human and rat glutathione synthetase enzymes are homodimers with 474 amino acids in each unit, whereas E. coli glutathione synthetase is 158 residues shorter and exists as a tetramer. Despite sharing the same function, the lack of sequence similarity between the two enzymes makes the previous studies of E. coli glutathione synthetase unsuitable for establishing the structure-function relationship of the human enzyme. Extensive kinetic studies of mammalian glutathione synthetase were carried out on the rat and recombinant human enzymes (38,39). The results suggested that there is a close catalytic dependence between the two substrates of the homodimer, generating a negative cooperativity for binding of ␥-glutamylcysteine substrate.
Little was known about the structure of human glutathione synthetase until recently, when Polekhina et al. (37) reported its crystal structure. Human glutathione synthetase was cocrystallized with the products: glutathione, ADP, one sulfate ion, which mimics the cleaved ␥-phosphate from the ATP, and two magnesium ions (Fig. 1). The three-dimensional structure of human glutathione synthetase shows that it belongs to the glutathione synthetase ATP-binding domain-like superfamily, which consists of enzymes with ATP-dependent carboxylateamine ligase activity and whose catalytic mechanisms are likely to involve acyl-phosphate intermediates (40). The ATPgrasp members display a unique nucleotide-binding fold, referred to as a palmate, or ATP-grasp fold.
Even though the catalytic mechanism of human glutathione synthetase is not fully understood, it appears to resemble those of the other ATP-grasp members. Therefore, structural comparison and analysis of the binding sites of the members, especially the ATP-binding site, can provide an insight into the catalytic mechanism of human glutathione synthetase. Additionally, structural and molecular properties investigations of human glutathione synthetase (wild type) compared with human glutathione synthetase (mutants) should reveal significant information about the function of the involved residues in the ATP-grasp superfamily.

EXPERIMENTAL PROCEDURES
Materials-Oligonucleotide primers for mutagenesis and sequencing were synthesized by Integrated DNA Technologies, Inc. Restriction enzymes were obtained from New England Biolabs. The QuikChange TM site-directed mutagenesis kit was obtained from Stratagene. ␥-Glutamyl-␣-aminobutyrate (gluABA) was synthesized as described previously (24,41). Iisopropyl-1-thio-␤-D-galactopyranoside and lactate dehydrogenase were obtained from Amresco. All other reagents were from Sigma (St. Louis, MO).
Recombinant DNA Methods-N-terminal His tag (His 6 ) was added to wild type human glutathione synthetase by subcloning human glutathione synthetase from plasmid pT7-7 to pET-15b vector (Novagen) at BamHI and NdeI sites. Mutants were generated using the QuikChange TM kit (Stratagene). The internal primers used for mutants are shown in Table I. The mutations were confirmed by partial sequencing.
Glutathione Synthetase Growth and Purification-The growth and purification protocol used for the recombinant wild type glutathione synthetase and mutant glutathione synthetase was the same. E. coli BL21(DE3) was transformed with pET-15b vector containing glutathione synthetase and grown to an A 600 of 1.0 at 37°C in Luria broth containing ampicillin (100 g/ml). After the A 600 reached 1.0 (isopropyl-1-thio-␤-D-galactopyranoside; 0.8 mM) was added. After induction (4 -6 h), cells were harvested by centrifugation (4225 ϫ g, 10 min, 4°C) and washed with cold 0.85% NaCl. All purification steps were performed at 4°C. The cells were resuspended in metal chelate affinity chromatography-0 buffer (MCAC-0 buffer; 20 mM Tris-Cl, 500 mM NaCl, 10% glycerol, pH 8.0) and disrupted by sonication (three pulses of 2 min with cooling between pulses). Cell debris was removed by centrifugation (11,950 ϫ g, 20 min). The clear lysate (ϳ25 ml) was applied to a nickel-nitrilotriacetic acid (Novagen) column previously equilibrated with MCAC-0 buffer. After loading, the column was washed with MCAC-0 buffer. Nonspecific bound proteins were removed by adding MCAC-50 buffer (MCAC-0 buffer plus 50 mM imidazole). Glutathione synthetase enzymes were eluted with MCAC-100 buffer (MCAC-0 buffer plus 100 mM imidazole). The fractions containing human glutathione synthetase were pooled and dialyzed twice against 20 mM Tris-Cl buffer (pH 8.0 containing 1 mM EDTA, 4 liters). The purified human glutathione synthetase proteins were deemed pure by SDS-PAGE standards (at least 99%). Enzyme Assays and Kinetic Analysis-All kinetic analyses were done in duplicates using purified recombinant glutathione synthetase. The enzyme activity was measured at 37°C using a spectrophotometric assay, which couples ADP production to NADH oxidation and is monitored at 340 nm (38,39). In brief, the standard assay contained buffer (100 mM Tris-Cl, pH 8.2, 50 mM KCl, 20 mM MgCl 2 , 5 mM sodium phospho(enol) pyruvate, 0.2 mM NADH), 10 units of pyruvate kinase (Type III rabbit muscle), 10 units of lactic acid dehydrogenase (Type II rabbit muscle), and glutathione synthetase substrates (see below) (final volume of 0.2 ml) and was initiated by the addition of human glutathione synthetase. For V max and k cat determinations, the concentration of ATP, gluABA, and glycine were 10, 20, and 10 mM, respectively. Instead of ␥-glutamylcysteine, gluABA was used to avoid the complication of thiol oxidation. The apparent K m values were determined using the standard assay where two substrates were held at saturating levels, whereas the third was varied by about 10-fold around the putative K m value. Control reactions contained the standard mix minus gluABA. The Michaelis-Menten kinetic equation was used for glycine and ATP data analysis. For gluABA kinetic analysis, the initial velocity (v), concentration (S), and V max are substituted into the Adair equation (Equation 1) for negative cooperativity, and the first K m and alpha (␣), or interaction factor are calculated. The second "estimated" K m is obtained by multiplying the first K m by the interaction factor ␣. The level of negative cooperativity is assessed by use of the Hill plot (V/V max Ϫ V) versus S, to obtain a Hill coefficient, h. A unit of enzyme is defined as the amount that catalyzes 1 mol of product per minute at 37°C. Protein concentration was determined by the Lowry method (42) using bovine serum albumin as the standard. Kinetic data were plotted and nonlinear regression analysis was performed using SigmaPlot software (SPSS Science Inc.).
Computational Methods-Crystallographic coordinates of the x-ray structure of human glutathione synthetase (glutathione synthetase, human form), reported to a resolution of 2.10 Å, and those of the other proteins were obtained from the Protein Data Bank (www.rcsb.org/ pdb/) (43).
The K2 program (available at zlab.bu.edu/k2/), an automated method that uses a genetic algorithm for aligning the three-dimensional structures of proteins, was employed to study the similarities between human glutathione synthetase and all the other members of the ATPgrasp superfamily. The K2 software first aligns the secondary structure elements of the proteins, starting with the more conserved ones, and then the alignment is extended to include any equivalent positions in loops and turns until the best fit is obtained (44).
Identification of the ATP-grasp members was established using the (Eq. 1)

REACTION 1
Structural Classification of Proteins (SCOP) data base (available at scop.mrc-lmb.cam.ac.uk/scop/index.html) where proteins are clustered into classes, folds, superfamilies, families, and domains, which represent hierarchy levels based on their evolutionary and structural relationships (45). The FSSP (Fold classification based on Structure-Structure alignment of Proteins) data base (www2.ebi.ac.uk/dali/fssp/ fssp.html) (46), another classification tool, was also employed to obtain structural neighbors of human glutathione synthetase. The active sites of the K2 aligned structures were visualized and analyzed with Swiss Pdb-Viewer (us.expasy.org/spdbv/) (47), a graphics program for studying macromolecular structure, in connection with the LPC software (Ligand-Protein Contacts) (sgedg.weizmann.ac.il/lpc/) (48).
Wild type and mutant 2HGS were modeled with the MOE 2002.03 software (Molecular Operating Environment) 2 using the Amber'94 force field (50). Starting with the crystal structure coordinates of human glutathione synthetase, the water molecules were removed, hydrogen atoms were added, and then the resulting structure was minimized with molecular mechanics. Mutants were built by altering distinct residues of the wild type enzyme, followed by energy minimization, before submitting them to the same molecular dynamics protocol (vide infra) employed for wild type human glutathione synthetase. All energy minimization procedures comprised three phases: first the steepest descent algorithm was employed (until r.m.s. gradient Ͻ 1000), then the conjugate gradient technique (until r.m.s. gradient Ͻ 100), and finally truncated Newton method (until r.m.s. gradient Ͻ 0.01).
Molecular dynamics simulations with constant NVT (thermodynamic ensemble where the number of particles, the volume, and the temperature are held fixed) and a time step of 0.001 ps were employed to find the lowest energy conformation of each structure (wild type and mutants). Initially, the temperature was raised progressively from 0 to 300 K in 1 ps. The system was then equilibrated at 300 K for 1000 ps during which the atomic coordinates were saved every 100 ps. Tests were carried out with different time length simulations before arriving at 1000 ps as the optimal combination of computational efficiency and accuracy. Finally, the 10 saved geometries were energy minimized and the lowest energy conformation among these was stored for further analysis.
Investigations of the active site were carried out for both wild type and mutant human glutathione synthetase enzymes. Hydrogen bonds and atomic distances were computed with the MOE 2002.03 program while interaction enthalpies between the considered constituents of the active site (residues Glu-144, Asn-146, Lys-305, and Lys-364 and ligands ADP 2Ϫ , SO 4 2Ϫ , and 2Mg 2ϩ ) were determined at the semiempirical quantum mechanics level using the Spartan'02 software. 3 The single point enthalpies (heats of formation) of the active site residues and cofactors were computed with the PM3 Hamiltonian (52). The interaction enthalpy between the two active site elements, denoted in Equation 2 as X and Y, was then estimated. (45) analysis shows that human glutathione synthetase (PDB code ϭ 2HGS) belongs to the Pre-ATP-grasp superfamily, which is the only member of the Pre-ATP-grasp fold. All members of the Pre-ATP-grasp fold are also members of the ATP-grasp fold. Consequently, human glutathione synthetase is also a member of the glutathione synthetase ATP-binding domain-like superfamily that is one of the two superfamilies included in the ATP-grasp fold. The glutathione synthetase ATP-binding domain-like superfamily consists of 51 proteins that are clustered into six families. A fold, which has one or more superfamilies, consists of proteins with similar major secondary structures in identical arrangement and topological connections. Superfamilies are subsets of a fold and consist of proteins with low sequence identities but whose structural and functional features suggest that a common evolutionary origin is probable. Thus, in the SCOP hierarchy, a common fold indicates a major structural similarity, whereas a common superfamily indicates a probable common evolutionary origin. A common family indicates a clear evolutionary relationship, proteins from the same family share Ϸ30% or greater sequence identity (45). However, two families (succinyl-CoA synthetase and pyruvate phosphate dikinase) from the glutathione synthetase ATP-binding domain-like superfamily are slightly more different than the other families, 2 Chemical Computing Group Inc., www.chemcomp.com. 3 Wavefunction Inc., www.wavefun.com.

Similarities between Glutathione Synthetase and Members of ATP-grasp Superfamily-SCOP
FIG. 1. Two views of the four conserved residues involved in the ATP binding site. Front (left) and side (right) view of human glutathione synthetase binding site. The modeled product-complex structure was obtained using the MD technique followed by optimization. Only the ligands and the four conserved residues are presented. The SO 4 ion mimics the cleaved ␥-phosphate from the ATP. Note that GSH is distant from the conserved residues.

TABLE I Primers used for site-directed mutagenesis of glutathione synthetase
a The underlined bases indicate the nucleotide positions that were changed. because their members are not included in the Pre-ATP-grasp domain superfamily. The members of the two aforementioned families display the same manner of binding the nucleotide inside the active site, but their similarity to the members of the Pre-ATP-grasp is limited to only two domains (pyruvate phosphate dikinase) or one domain (succinyl-CoA synthetase), whereas the members of the other families (eukaryotic and prokaryotic glutathione synthetases, D-alanine-D-alanine ligase, biotin carboxylase, and synapsin Ia) have three common domains.
The six most similar proteins to 2HGS were chosen from the Pre-ATP-grasp fold using FSSP data base (Table II). Only those proteins with an FSSP Z score higher than 4 (level of similarity is analogous to second cousins (46)) are desired, because all structural neighbors display a Z score of at least 2, which means that they belong to the same fold as 2HGS. The PDB codes for the six selected proteins are 1M0W (53), 1GSA (36), 1AUV (54), 1IOW (55), 1B6R (56), and 1EZ1 (57). Conserved residues in the active site of human glutathione synthetase were identified by using LPC analysis and K2 structural alignment for 2HGS and selected proteins. The LPC analysis listed 40 residues in the active site of 2HGS and each one was verified against the alignment output generated by K2 with each of the six proteins just mentioned.
The K2 structural alignments results of 2HGS with the six most structurally similar Pre-ATP-grasp proteins showed that there are only four highly conserved residues involved in the ligands binding site: Glu-144, Asn-146, Lys-305, and Lys-364 (Fig. 2). The K2 alignment of 2HGS with other proteins was extended to all members of the Pre-ATP-grasp domain superfamily, and the results confirmed that these four residues are conserved. Glu-144 is fully conserved all along the series (for both the most similar Pre-ATP-grasp domain superfamily members shown above, as well as for even more structurally disparate members of this superfamily). Lys-364 is almost fully conserved in all families, being replaced by Arg only in carbamoyl phosphate synthetase, whereas Lys-305 is replaced by Arg in N 5 -carboxyaminoimidazole ribonucleotide synthetase, Purt-   a FSSP Z score describes the level of structural similarity relative to 2HGS. Only those with Z score greater than 4 are shown. b The K2 final score is computed by using the number and the root mean square distance of the of aligned residues. c The FSSP Z score for 1M0W is not available yet, but this protein displays the highest level of similarity to 2HGS, based on K2 structural alignment. encoded glycinamide ribonucleotide transformylase, and carbamoyl phosphate synthetase. Asn-146 is replaced by Val in synapsin Ia, Ala in N 5 -carboxyaminoimidazole ribonucleotide synthetase, or Ser in Purt-encoded glycinamide ribonucleotide transformylase. Additional results for the alignments of the more structurally disparate members belonging to the Pre-ATP-grasp superfamily and SCOP classification of the Pre-ATP-grasp superfamily and the glutathione synthetase ATP-binding domain-like superfamily are given in the Supplemental Material.
Comparison of human glutathione synthetase with all other members of the Pre-ATP-grasp superfamily shows that Glu-144 residue is fully conserved, Lys-364 and Lys-305 are conserved as regards the charge, and Asn-146 is backbone-conserved. Our attempt to extend this conclusion to members of the succinyl-CoA synthetase and pyruvate phosphate dikinase families, which are classified by SCOP (45) as ATP-grasp members, was not successful. Thus, the active site of 2HGS is highly similar only to members of the Pre-ATP-grasp fold.
Comparison of Wild Type and Mutant Glutathione Synthetase-Only four residues (Glu-144, Asn-146, Lys-305, and Lys-364) were found to be highly conserved in the active site domain of 2HGS. Thus, it is reasonable to conclude that these four residues are crucial for the biological activity of human glutathione synthetase. This hypothesis was tested experimentally and computationally using site-directed mutagenesis of the conserved residues. The Glu-144, Asn-146, Lys-305, and Lys-364 residues were replaced with amino acids that cannot interact favorably with the ligands; the structure and catalytic activity of the corresponding mutant enzymes were then compared with wild type.
The functions of the four conserved residues in glutathione synthetase were examined by preparing the human glutathione synthetase mutants using site-directed mutagenesis and His tag purification. The results (Table III) show that two mutants E144K and N146K had undetectable activity. Five other mutants had very low activities (E144A and N146D, 0.05%; N146A and K364A, 0.1%; and K364E, 0.2% active as wild type human glutathione synthetase). The only human glutathione synthetase mutants with significant activity were K305A and K305E, with 6.5 and 5% activity of the wild type, respectively. Two mutant enzymes with sufficient activity (K305A and K305E) were subjected for further kinetic analysis (Table  IV). The apparent K m values for glycine were found the same for the K305E mutant and decreased (ϳ7-fold) for the K305A mutant. The apparent K m values for gluABA for both K305A and K305E were about the same as the wild type human glutathione synthetase, and the mutant enzymes no longer display negative cooperativity. The apparent K m values for ATP increased dramatically for K305A (10-fold) and K305E (40-fold).
The conformations of the mutants were computed using molecular mechanics and molecular dynamics (MD) simulations and then superimposed on wild type human glutathione synthetase using Swiss Pdb-Viewer. Experimental and calculated results revealed that none of the nine (E144A, E144K, N146A, N146D, N146K, K305A, K305E, K364A, and K364E) mutant enzymes adopts a markedly different tertiary structure than that of the wild type human glutathione synthetase. The backbone r.m.s.d. calculation applied to structural alignments of mutants with wild type human glutathione synthetase ranges from 1.7 to 2.3 Å, suggesting a high degree of tertiary structure similarity and is consistent with the results of circular dichroism spectroscopy. Because the overall tertiary structure of human glutathione synthetase is not altered by mutation, the reduced activity of the mutant enzymes is probably due to the nature of ligand binding. This hypothesis was explored by studying the interaction between the structural units (the cofactor ligands and the four conserved residues involved in the binding pocket of ATP).
The atomic distance analysis of the human glutathione synthetase structure obtained from MD simulation showed that cofactor ligands (ADP 2Ϫ , SO 4 2Ϫ , and 2Mg 2ϩ ) interact with all four conserved residues through a network of contacts (Table  V). The two-dimensional representation of these contacts is depicted for wild type human glutathione synthetase in Fig. 3. The ⑀-amino group of Lys-364 interacts with ADP through three hydrogen bonds: 2.71 Å to N (amino group of ADP), 3.44 Å to N7 (adenine ring of ADP), and 2.12 Å to O ␦Ϫ (␣-phosphate of ADP). The ⑀-amino group of Lys-305 hydrogen bonds to the ␤-phosphate group of ADP and to the ␥-carboxylate group of Glu-144. There is one hydrogen bond (2.72 Å) between ⑀-amino group (Lys-305) and O ␦Ϫ (␤-phosphate of ADP) and two hydrogen bonds (1.72 and 2.15 Å) between ⑀-amino group (Lys-305) and both O ␦Ϫ atoms of ␥-carboxylate group (Glu-144). The ␥-carboxylate group of Glu-144 also interacts with Mg 2ϩ ion (number 501) using both oxygen atoms, although one is significantly shorter than the other (Mg-O distances are 2.10 and 3.83 Å). The ␤-carbonyl group of Asn-146 coordinates the Mg 2ϩ ion (501) within a calculated distance of 2.34 Å. Additionally, the SO 4 2Ϫ ion (which mimics the cleaved ␥-phosphate) forms a bridge between the two Mg 2ϩ ions. The distances linking the negative oxygen atoms of sulfate, and the Mg 2ϩ ions are 2.12, 2.48, and 3.51 Å for Mg 2ϩ 501 and 2.08, 2.14, and 3.45 Å for Mg 2ϩ 502.
In addition we examined the calculated atomic distances (Table V) for human glutathione synthetase mutant enzymes using the MD structures. When Glu-144 is mutated (Fig. 4, A  and B), Lys-305 and Asn-146 side chains are reoriented. The Lys-305 side chain moves closer to ADP, forming an additional hydrogen bond (N ϩ -H-O ␦Ϫ ), whereas Asn-146 rearranges its side chain and forms a hydrogen bond with sulfate ion (N-H-O ␦Ϫ ), which replaces the cleaved ␥-phosphate of ATP. Similarly, when Lys-364 is mutated (Fig. 4, C and D), the side chain of Lys-305 moves closer to ADP and forms one additional ionic hydrogen bond of the type N ϩ -H-O ␦Ϫ . Furthermore, a change in the position of the Glu-144 side chain occurs so that the ␥-carboxylate group coordinates more effectively the Mg 2ϩ 501 ion in K364A or coordinates both Mg 2ϩ 501 and 502 ions in K364E structure. Mutation of Lys-305 (Fig. 4, E and F) shows that Glu-144 and Lys-364 side chains are affected. The ␥-carboxylate group of Glu-144 is closer to both magnesium ions resulting in a stronger bond between ␥-carboxylate and Mg 2ϩ 501 (K305E), as well as additional bonds between the ␥-carboxylate and Mg 2ϩ 502 (K305A and K305E). On the other hand, the ⑀-amino group of Lys-364 is further away from the adenine ring, which causes the loss of two hydrogen bonds relative to the wild type enzyme. For the K305E mutant, the favorable contact between the side-chain carbonyl group of Asn-146 and Mg 2ϩ (501) ion is lost, and an ionic hydrogen bond (N-H-O ␦Ϫ ) between the amide group and the sulfate ion is gained. Replacement of Asn-146 with Ala or Asp (Fig. 4, G and H) gives a weaker hydrogen bond between Lys-305 and ADP. However, for the N146A there is a stronger interaction between Glu-144 and both Mg 2ϩ ions. In contrast, the N146K mutant ( Fig. 4 I) forms a second ionic hydrogen bond between Lys-305 and ADP, and the interaction becomes stronger. Also, the ␥-carboxylate group of Glu-144 is altered so that there are changes with respect to both Mg 2ϩ (501 and 502), resulting in a significant decrease in the distance between the oxygen atoms and magnesium ions. For the N146K both Mg ions are coordinated by ␥-carboxylate group of Glu-144. To summarize, this analysis revealed that a single mutation of these conserved residues induces changes in the orientation of the other residues side chains causing deviations in ligand binding.
Semiempirical PM3 calculations were applied to MD structures for both wild type and mutant enzymes to study the interaction enthalpies between conserved residues and cofactor ligands. The cofactor ligands (ADP 2Ϫ , SO 4 2Ϫ , and 2Mg 2ϩ ) were treated computationally as a single unit, because the atomic distance calculations (Table V, part C) show that the interactions between magnesium ions (501 and 502) and the sulfate ion, which mimics the cleaved ␥-phosphate, remain as it is seen in wild type structure. The results illustrate that interactions between conserved residues and cofactor are sensitive to selected mutations (Fig. 5). For example, in the case of interaction enthalpy between cofactor and residue 144, eight out of nine mutants (E144A, E144K, K364A, K364E, K305A, K305E, N146A, and N146K) display a significant change (Ͼ15 kcal/ mol) compared with wild type.
The interaction enthalpy between cofactor and residue 146 shows five mutants (E144A, K305E, N146A, N146D, and N146K) with a deviation in energy greater than 15 kcal/mol relative to the wild type enzyme. Correspondingly, for two mutants (K364A and K364E) a substantial variation occurs in the interaction enthalpy between cofactor and residue 364 and finally, interaction enthalpy between cofactor and residue 305 show that only one mutant, which is E144K and not K305A or K305E, displays a greater change in energy than 15 kcal/mol. Consistent with the previous analysis, the PM3 calculations substantiate that each mutated residue alters the interaction enthalpy between itself and the cofactor, and interestingly the mutation also affects the interaction for the other three conserved residues.

DISCUSSION
Previous studies (49,51,58) of the crystal structure of proteins belonging to the Pre-ATP-grasp superfamily have identified several important residues situated near ATP; some of these residues are conserved and some are involved in the ATP binding. However, the structural alignments searches for conserved residues in the Pre-ATP-grasp family included mostly the prokaryotic proteins. The current work is the first study that examines all members of the Pre-ATP-grasp superfamily, including human glutathione synthetase (a eukaryotic member), on the subject of highly conserved residues and their function. We found four residues in the ATP binding site of human glutathione synthetase that are highly conserved in all of the Pre-ATP-grasp superfamily: Glu-144, Asn-146, Lys-305, and Lys-364; and thus, the function of these was expected to be essential for catalytic activity. The current modeling and experimental results suggest that the mutations of the four conserved residues of human glutathione synthetase yield enzymes whose activity is not affected by a major change in the tertiary structure of the enzyme, but rather by alteration of the ATP binding site. In a protein, replacement of a charged residue with a neutral or oppositely charged residue is expected to disturb the local electrochemical stability. Our MD simulations of human glutathione synthetase showed that, in addition to the mutated residue, the other three conserved residues are also affected with regard to their interaction with the ligands (ADP 2Ϫ , SO 4 2Ϫ , and 2Mg 2ϩ ). Hence, a single point mutation of human glutathione synthetase causes a cascading chain of events that influences the overall binding of the ATP.
Only K305A and K305E human glutathione synthetase mutants show any significant enzyme activity, suggesting that all four conserved residues are critical for enzyme activity. The higher apparent K m values for ATP for the Lys-305 mutants imply that Lys-305 is important for ATP binding. These findings suggest that mutation of any of the other three conserved residues in human glutathione synthetase is also likely to affect ATP-Mg 2ϩ binding. Further bonding studies, such as iso-thermal microcalorimetry will be useful.
The calculations of interaction enthalpies indicate that the interaction between Lys-305 and cofactor is the least sensitive to mutation because Lys-305 residue is closer to Glu-144 than it is to ADP. Thus, whether Lys-305 or the other conserved residues are mutated there is no significant change in interaction enthalpy; however, for the E144K mutation, the interaction enthalpy between Lys-305 and ADP increases noticeably as a result of electrostatic repulsion between the two lysine residues (144 and 305). In contrast, the interaction between Glu-144 and cofactor is the most sensitive to mutation. The large variations of interaction enthalpies between Glu-144 and cofactor ligands for human glutathione synthetase mutants are most likely due to differences in coordination of magnesium (501 and 502) ions by the negatively charged ␥-carboxylate group of Glu-144.
Our findings suggest that Glu-144 is probably very important for the stabilization of the ␥-glutamylcysteinyl-phosphate intermediate. Since both magnesium ions are strongly coordinated by the ␥-phosphate in the product-complex structure of human glutathione synthetase, it is likely that the acyl-phosphate intermediate coordinates the magnesium ions in a similar manner. The role of Glu-144 then, could be to compensate the positive charge around this intermediate and possibly to assist in the ligation of glycine to ␥-glutamylcysteine. A reason why both E144K and N146K mutant enzymes exhibit an extremely low catalytic activity is that the Asn-146 residue may also help to compensate for the positive charge in the environs of the reaction intermediate.
All four conserved residues are essential for the human glutathione synthetase enzyme, and the counterparts of these conserved residues are presumably essential for the other members of the Pre-ATP-grasp superfamily. This is especially likely for Glu-144 since it is fully conserved in the Pre-ATP-grasp superfamily and participates in the formation of the acyl-phosphate intermediate. Thus, we propose that this residue is necessary for the catalytic function of glutathione synthetase and also for the other enzymes in the superfamily. The charged conserved resi- dues, Lys-364 and Lys-305, balance the negative charge of ATP and therefore, these residues and their homologous in the Pre-ATP-grasp enzymes are expected to be involved in the optimal orientation of ATP in relation to the carboxyl group of the substrate. The backbone-conserved residue, Asn-146, coordinates one of the Mg 2ϩ ions and our results suggest that it takes part in catalysis as well. Since few proteins in the Pre-ATP-grasp superfamily uses Ala, Ser or Val residues instead of Asn, it appears that these proteins have a slightly different mechanism for ATP activation than human glutathione synthetase. FIG. 5. Calculated interaction enthalpies between ligands and conserved human glutathione synthetase residues for wild type and each mutant enzyme. The interaction enthalpies between ligands (ADP 2Ϫ , SO 4 2Ϫ , and 2 Mg 2ϩ ) and individual conserved residues were computed with the PM3 Hamiltonian available in Spartan'02 software, using the single point enthalpies (heats of formation). All calculations are based on the modeled structures that resulted after 1000 ps of molecular dynamics simulations.