Molecular Architecture of the Glucose 1-Phosphate Site in ADP-glucose Pyrophosphorylases*

ADP-Glc pyrophosphorylase (PPase), a key regulatory enzyme in the biosynthetic pathway of starch and bacterial glycogen, catalyzes the synthesis of ADP-Glc from Glc-1-P and ATP. A homology model of the three-dimensional structure of the Escherichia coli enzyme complexed with ADP-Glc has been generated to study the substrate-binding site in detail. A set of amino acids in the model has been identified to be in close proximity to the glucose moiety of the ADP-Glc ligand. The role of these amino acids (Glu194, Ser212, Tyr216, Asp239, Phe240, Trp274, and Asp276) was studied by site-directed mutagenesis through the characterization of the kinetic properties and thermal stability of the designed mutants. All purified alanine mutants had 1 or 2 orders of magnitude lower apparent affinity for Glc-1-P compared with the wild type, indicating that the selected set of amino acids plays an important role in their interaction with the substrate. These amino acids, which are conserved within the ADP-Glc PPase family, were replaced with other residues to investigate the effect of size, hydrophobicity, polarity, aromaticity, or charge on the affinity for Glc-1-P. In this study, the architecture of the Glc-1-P-binding site is characterized. The model overlaps with the Glc-1-P site of other PPases such as Pseudomonas aeruginosa dTDP-Glc PPase and Salmonella typhi CDP-Glc PPase. Therefore, the data reported here may have implications for other members of the nucleotide-diphosphoglucose PPase family.

Most ADP-Glc PPases are allosterically regulated by small effector molecules. Although these vary according to the source, they are all intermediates of the principal carbon assimilation pathway in the respective cell (2)(3)(4)(5)(6). Thus, bacterial glycogen and plant starch syntheses are not modulated only by the availability of ATP but also by the accumulation of key metabolites that represent the carbon and energy balance within the cell. For instance, the enzymes from heterotrophic bacteria such as Escherichia coli are regulated by intermediates of the glycolytic pathway, with Fru-1,6-P 2 as the main activator and AMP as the main inhibitor. On another hand, the ADP-Glc PPases from cells performing oxygenic photosynthesis and assimilating atmospheric CO 2 through the reductive pentose phosphate pathway or the Calvin cycle (specifically cyanobacteria, green algae, and photosynthetic tissues from higher plants) are activated by 3-phosphoglycerate and inhibited by P i (6).
The first ADP-Glc PPase crystal structure became recently available when Jin et al. (14) solved that of the homotetrameric Solanum tuberosum (potato tuber) small subunit in its allosterically inhibited form at a resolution of 2.1 Å. They also reported the structural determination of the enzyme complexed with either ATP or ADP-Glc at 2.6 and 2.2 Å, respectively. Attempts to obtain information on the E. coli enzyme structure through x-ray crystallography were unsuccessful. The potato tuber small subunit has only ϳ33% sequence identity to the E. coli enzyme, but the similar predicted secondary structure profile, together with available biochemical data, suggests that they share a common three-dimensional fold (5).
Previous chemical modification (15) and site-directed mutagenesis (16) studies on E. coli ADP-Glc PPase identified Lys 195 as an important residue for Glc-1-P interaction. Replace-ment with other amino acids generated 100 -10,000-fold increases in the S 0.5 for this substrate, with all other kinetic constants at wild-type levels. Later, Fu et al. (17) reported similar results from analysis of the homologous Lys 198 in the potato tuber catalytic subunit. The proposed role of this amino acid is to form an ionic bond between the ⑀-amino group and the negatively charged phosphate of Glc-1-P. Results with Hex-1-P analogs, which differ from Glc-1-P in their hydroxyl groups, suggest that other residues in the active site participate in substrate binding.
Our aim was to obtain structural information on E. coli ADP-Glc PPase by building a homology model and to probe a set of highly conserved residues in the N-terminal domain possibly involved in Glc-1-P binding. We studied the role of Glu 194 , Ser 212 , Tyr 216 , Asp 239 , Phe 240 , Trp 274 , and Asp 276 by site-directed mutagenesis and kinetic characterization of the mutant enzymes and their thermal stability. All residues were replaced with alanine and other amino acids to evaluate the importance of size, charge, or hydrophobicity on the effects observed in substrate interaction.
Because these residues are highly conserved among ADP-Glc PPases, it was of interest to investigate whether they are present in other PPases that use Glc-1-P as a substrate. The observations made by comparison of the putative Glc-1-P site from our E. coli ADP-Glc PPase model and the reported crystal structures of two pyrophosphorylases, the Pseudomonas aeruginosa dTDP-Glc PPase Rm1A (18) and Salmonella typhi CDP-Glc PPase (19), have led us to propose that the results presented here have implications beyond the family of ADP-Glc PPases.

Materials
Oligonucleotides were synthesized and purified at the Macromolecular Facility of Michigan State University. [ 32 P]PP i was purchased from PerkinElmer Life Sciences, and [ 14 C]Glc-1-P from ICN Pharmaceuticals, Inc. Sodium PP i , ATP, ADP-Glc, AMP, and inorganic pyrophosphatase were purchased from Sigma. Pfu DNA polymerase was purchased from Stratagene (La Jolla, CA). All other reagents were of the highest quality available.

Site-directed Mutagenesis
Site-directed mutagenesis was performed by overlap extension PCR (38). Plasmid pMAB3 containing the E. coli ADP-Glc PPase gene between NdeI and SacI sites, previously obtained in our laboratory, 5 was used as a template. The flanking primers, which annealed with the T7 promoter and the SacI site (underlined) were 5Ј-TAATACGACTCACTATAGGG-3Ј and 5Ј-GATATCTGAATTCGAGCTC-3Ј, respectively. The overlapping primers for each mutant are depicted in supplemental Table 1. The final PCR products were gel-purified, digested with NdeI and SacI, and subcloned to obtain the different pMAB3-single mutant plasmids. Plasmid pETEC-N⌬15-D276N was obtained using pETEC-N⌬15 (39) as a template, with the T7 promoter and T7 terminator as flanking primers and the same mutated overlapping primers used for pMAB3-D276N (supplemental Table 1). All plasmids were sequenced at the Genomics Facility of Michigan State University to confirm incorporation of only the desired mutation.

Purification of pMAB3-Single Mutant Plasmids
One-liter cultures of AC70R1-504 cells transformed with the pMAB3-single mutant plasmids or BL21(DE3) cells transformed with pETEC-N⌬15-D276N were grown in 25 g/ml kanamycin/Luria broth (1 liter) at 37°C up to A 600 ϭ 0.8. Induction was initiated by the addition of isopropyl ␤-D-thiogalactopyranoside (1 mM final concentration), with subsequent incubation at 25°C for 16 h. After induction, cells were harvested, and crude extracts were obtained as described previously (35). After centrifugation, the precipitate was resuspended in buffer A (50 mM Hepes (pH 8.0), 5 mM MgCl 2 , 0.1 mM EDTA, and 10% sucrose). The samples were individually applied to a DEAE-Fractogel column (EMD Biosciences) and eluted with a linear gradient of 0 -0.5 M NaCl. The active fractions were pooled and desalted. After this step, samples were 60 -70% pure and suitable for performing kinetic analysis. Mutants E194A/Q/D, D276A/N, W274A, Y216F, and D239N were resuspended in buffer B (buffer A plus 1.2 M ammonium sulfate); applied to a phenyl-Superose fast protein liquid chromatography column (GE Healthcare) equilibrated with buffer B; and eluted with a linear gradient of 1.2 to 0.001 M ammonium sulfate. Further purification of the rest of the mutants and the wild type was performed by applying the DEAE pool samples to a Matrex TM gel green A affinity chromatography column (Amicon Corp.) and eluting with a linear gradient of 0 -2 M NaCl. The purest fractions of each enzyme were pooled, desalted, and concentrated; and after these steps, the proteins were Ͼ95% pure as assessed by SDS-PAGE (data not shown).
Assay B: Synthesis Direction-Formation of ADP-[ 14 C]Glc from [ 14 C]Glc-1-P was determined by the method of Yep et al. (42). The reaction was carried out for 10 min at 37°C in a mixture containing [ 14 C]Glc-1-P (ϳ400 dpm/nmol), ATP, MgCl 2 , and Fru-1,6-P 2 at varying concentrations according to the mutant enzyme assay; 50 mM Hepes (pH 8.0); 1.5 units/ml pyrophosphatase; and 0.2 mg/ml bovine serum albumin plus enzyme in a total volume of 0.20 ml. One unit of enzyme activity is 1 mol of product (either [ 32 P]ATP or ADP-[ 14 C]Glc) formed per min at 37°C.

Kinetic Characterization
Kinetic data were plotted as specific activity (units/mg) versus substrate or effector concentration. Kinetic constants were acquired by fitting the data to the Hill equation with a nonlinear least-square formula using Origin TM Version 5.0. Hill plots were used to calculate the Hill coefficient and the kinetic constants corresponding to the substrate or activator concentrations giving 50% of the maximal velocity (S 0.5 ) or activation (A 0.5 ).

Thermal Stability
Enzyme samples were in buffer A supplemented with bovine serum albumin to 1 mg/ml in a final volume of 100 l. Half of the sample (50 l) was incubated in a water bath equilibrated at 60°C for 5 min and placed on ice immediately after. The remaining half (50 l) was kept on ice as a control. The enzyme activities for both the heat-treated and control samples were determined in the ADP-Glc synthesis direction as described for Assay B.

RESULTS
Homology Modeling-We obtained a three-dimensional model of E. coli ADP-Glc PPase by comparative modeling using the coordinates of the recently solved crystal structure of the potato tuber ADP-Glc PPase small subunit (Protein Data Bank code 1YP2) as a template as described under "Experimental Procedures" (Fig. 1A). Although modeling is generally guaranteed to be successful if residue identity is Ͼ40%, for lower percentages, errors can be reduced employing an accurate sequence alignment (43)(44)(45). Our two enzymes shared only 33% residue identity; therefore, the alignment was manually edited, incorporating information such as conservation of functional residues and prediction of secondary structures.
Using MODELLER6 Version 1, we generated 143 models after several iterative refinements of the alignment to accommodate gaps, deletions, and insertions of the query sequence with respect to the template in the best possible way. We assessed the models with the program VERIFY_3D (23, 24) as described under "Experimental Procedures," which evaluates the compatibility of a given residue (1D) in a certain environment (3D). A score below zero for a given residue means that the conformation adopted by that residue in the model is not compatible with its surrounding environment. In our study, we considered only those models with all 1D-3D averaged scores above zero; and among them, we chose the one with a profile most similar to that generated by the template crystal structure (Fig. 1B). The two profiles followed the same general trend except for two specific regions, both corresponding to residues located in or adjacent to loops not present in the template structure (indicated by arrows in Fig. 1A). The first, encompassing Phe 90 -Glu 97 in the E. coli enzyme, aligns with a region in the potato tuber enzyme that is disordered in the crystal structure. The second loop, containing Lys 259 -Pro 271 , is an insertion in the bacterial enzyme. Therefore, the final conformation of these two loops in the model, which might also affect immediately adjacent secondary structures, accounted for the differences from the template structure profile. According to the model, these loops are not part of the active site, and they do not contain important conserved residues.
In agreement with our previous biochemical results (46), the modeled monomer shows a two-domain structural organization (Fig. 1A). The N terminus of ϳ300 residues presents a ␤-␣-␤ motif arranged in an open twisted ␤-sheet surrounded by ␣-helices. It resembles the Rossmann fold, typically present in nucleotide-binding domains (47). Residues important for catalysis, Asp 142 (40), and for substrate binding, Tyr 114 for ATP (48) and Lys 195 for Glc-1-P (16), are located in the active-site pocket in close proximity to the ADP-Glc molecule (Fig. 1C), observations that further validate the quality of our model. The C terminus is a separate domain folded as a ␤-helix and linked to the N terminus by a long loop. The two domains are in intimate interaction through extensive hydrophobic contacts, supporting the requirement of a full-length polypeptide to obtain normal enzyme activity and regulation (46).
Selection of Residues for Analysis-The three-dimensional model complexed with ADP-Glc shows the ligand placed in a well defined pocket in the active site (Fig. 1A), and several direct interactions between the ligand and the enzyme are evident (Fig. 2). Lys 195 makes a salt bridge with the glucose phosphate, an interaction that has been biochemically probed by Hill et al. (16) in E. coli ADP-Glc PPase and by Fu et al. (17) in their analysis of the homologous residue (Lys 198 ) in the potato tuber enzyme. Additionally, the hydroxyl groups of the glucosyl moiety of the ligand are involved in a complex network of hydrogen bonds with the enzyme. The side chains of Glu 194 , Asp 276 , and Ser 212 and the backbone of the latter participate in such interactions.
We performed a multiple sequence alignment using the catalytic subunits of 15 ADP-Glc PPases from several sources, each of them representative of a different taxonomic group. Fig. 3 depicts part of the aligned sequences, comprising residues located in and around the putative Glc-1-P-binding domain in the N terminus of the protein. The residues that, in the model, appear to interact through hydrogen bonds with the glucosyl moiety of the ligand are absolutely conserved among all ADP-Glc PPases analyzed, suggesting that they are involved in a conserved role such as substrate binding. According to our structural model, other conserved residues in this region are also located in the substrate-binding pocket. Based on our observations, we selected Tyr 216 , Asp 239 , Phe 240 , and Trp 274 to be characterized together with Glu 194 , Ser 212 , and Asp 276 .
Expression and Purification of pMAB3-Single Mutant Plasmids-All selected amino acids were mutated to alanine to analyze their potential role in Glc-1-P interaction. We created additional mutations to investigate whether the observed effect  on the affinity for Glc-1-P is due to their shape, size, charge, or aromaticity. E. coli wild-type and mutant ADP-Glc PPases were expressed and purified as described under "Experimental Procedures." They had the expected molecular masses upon SDS-PAGE, and they were recognized by the anti-E. coli AC70R1 ADP-Glc PPase antibody in immunoblots (data not shown). Y216A either failed to be expressed or rendered the protein completely susceptible to proteolysis because no band Ն10 kDa was detected by immunoblotting in either the soluble or insoluble fractions of the expression cell lysates. After the first chromatographic step, all enzymes were 60 -70% pure and suitable for kinetic characterization assays. An additional chromatographic step yielded Ͼ95% pure enzymes, which allowed for the proper determination of their specific activities.
Kinetic Characterization-The kinetic characteristics of the mutant enzymes were compared with those of the wild type. All alanine mutations decreased the apparent affinity of the enzyme for the substrate Glc-1-P, as S 0.5 values for all the mutants were 1 or 2 orders of magnitude larger than that for the wild type ( Table 1). The most important increments in this   DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 kinetic parameter were observed with mutations at Glu 194 and Ser 212 . Glc-1-P saturation curves obtained for the Glu 194 mutants are shown in Fig. 4 as an example to illustrate the shift in the S 0.5 between the wild type and Glu 194 mutants. The E194A mutant showed a 165-fold increase (Table 1) compared with the wild type; therefore, we made substitutions to aspartic acid and glutamine to evaluate the importance of the charge and side chain size in such an effect. Mutation to glutamine increased the S 0.5 for Glc-1-P by 85-fold, pointing out the importance of the negative charge for substrate binding. However, mutation to aspartic acid, which also bears a negative charge, caused a larger negative effect on this kinetic parameter (Table 1), highlighting the significance of the side chain size. These two mutations, E194D and E194Q, caused 4-and 5-fold reduced V max values with respect to the wild type, whereas the E194A mutation decreased the V max by 24-fold. The apparent affinities for ATP, Mg 2ϩ , and the activator Fru-1,6-P 2 were not significantly affected by any of these mutations of Glu 194 (Table 2). Our results validate the hydrogen bonds observed in the structural model (Fig. 2) and strongly suggest that Glu 194 plays a role in Glc-1-P binding.

Glucose 1-Phosphate Site from ADP-glucose Pyrophosphorylases
Our structural model proposes that Ser 212 binds O-3 and O-4 of the sugar moiety of the ligand through hydrogen bonds with the side chain and backbone, respectively (Figs. 2 and 6). Here, we probed the role of the side chain in Glc-1-P binding. All Ser 212 mutations maintained apparent affinity properties for ATP, Mg 2ϩ , and Fru-1,6-P 2 at wild-type levels ( Table 2). S212A also showed similar k cat values compared with the wild type, but it displayed a 14-fold increased S 0.5 for Glc-1-P (Table 1). S212V and S212T caused dramatic effects on the apparent affinity for Glc-1-P, with 377-and 274-fold increased S 0.5 values, respectively (Table 1). Mutation to valine decreased the k cat by ϳ16fold, whereas mutation to threonine did so by ϳ2-fold compared with the wild type. Surprisingly, replacement of Ser 212 with tyrosine increased the apparent affinity for Glc-1-P by only 5-fold. The k cat for this mutant was, however, 231-fold lower compared with that for the wild type. These results strongly suggest that Ser 212 is located in the Glc-1-P-binding pocket and that it contributes to the affinity of the enzyme for this substrate.
Asp 276 was replaced with alanine, asparagine, and glutamic acid. The three mutations decreased the apparent affinity for Glc-1-P by 100-, 85-, and 24-fold, respectively (Table 1). Our results point out the importance of Asp 276 for Glc-1-P binding and the significance of both the negative charge and size of its side chain on such effect. The analyses of these mutants suggest an additional role for Asp 276 besides Glc-1-P interaction given that other kinetic parameters were also affected. D276A and D276N had ϳ1000-fold lower V max values compared with the

TABLE 2 Kinetic parameters of E. coli wild-type and mutant ADP-Glc PPases
Reactions were performed in the synthesis direction (Assay B) as described under "Experimental Procedures." Data represent the mean of two or three identical experiments Ϯ the mean difference of the duplicates or triplicates.

Glucose 1-Phosphate Site from ADP-glucose Pyrophosphorylases
wild type (Table 1) and 3.4-and ϳ4-fold higher S 0.5 values for ATP, respectively (Table 2). D276E displayed a 3-fold decreased V max with respect to the wild type (Table 1) but a bigger change in the apparent affinity for ATP, characterized by an 8-fold increased S 0.5 for this substrate ( Table 2). On the other hand, all three mutations decreased the apparent affinity for Mg 2ϩ by ϳ4 -6-fold (Table 2). These results correlate with the role of the Mg 2ϩ ion chelator proposed for the homologous residue (Asp 280 ) in the S. tuberosum enzyme (14). Furthermore, the three Asp 276 mutants had 5-15-fold higher A 0.5 values for Fru-1,6-P 2 compared with the wild type ( Table  2). To investigate whether this residue is involved in the activator site, we studied another mutant. Previous reports showed that deletions of 11 and 15 residues from the N terminus of E. coli ADP-Glc PPase render activated enzymes even in the absence of Fru-1,6-P 2 , with all other kinetic parameters similar to those of the wild type (39,49). On the basis of these results, we combined both the N-terminal deletion and the single mutation D276N to create EcN⌬15-D276N. The activity of the partially purified double mutant was 0.027 Ϯ 0.003 units/mg, similar to that of the partially purified D276N single mutant (data not shown), whereas the A 0.5 for Fru-1,6-P 2 was 51 M, similar to the that of the wild type (Table 2). This strongly suggests that Asp 276 is not directly involved in activator binding but is a pivotal residue for the correct interaction of the substrates with the enzyme, influencing the resulting conformational changes upon their binding.
The role of the size and aromaticity of Trp 274 was studied by substituting it with alanine, leucine, and phenylalanine. Mutation to alanine was characterized by a 22-fold decrease in the apparent affinity for Glc-1-P and did not have significant effect on the V max of the enzyme (Table 1) or on the apparent affinities for ATP, Mg 2ϩ , and Fru-1,6-P 2 ( Table 2). We obtained similar results when leucine was placed in this position. In contrast, all parameters remained almost unchanged compared with wild-type levels when Trp 274 was replaced with phenylalanine, indicating that aromaticity is required at this position for proper interaction of Glc-1-P with the enzyme.
Tyr 216 is conserved not only among ADP-Glc PPases (Fig. 3) but also in RmlA (Tyr 176 ) (Fig. 5, A and  B) (18). Mutation to phenylalanine allowed us to study the role, if any, of the side chain hydroxyl group in this position. The Y216F mutant displayed a 46-fold lower apparent affinity for Glc-1-P (Table 1) and showed small variations in the apparent affinities for ATP, Mg 2ϩ , and Fru-1,6-P 2 , with 1-3fold increases in the respective kinetic constants (Table 2). However, this substitution, in which OH was removed, caused a 10-fold decrease in the V max of the mutant enzyme. Our structural model does not show any direct interaction between this residue and bound ADP-Glc (Fig. 2). Our biochemical data suggest, however, that the side chain OH group plays a role in Glc-1-P interaction, possibly by driving the correct positioning of the substrate in the pocket, which also affects the concomitant catalytic reaction.
Asp 239 and Phe 240 are also conserved residues that are located in close proximity to the ligand and that do not show any evident interaction with it in the three-dimensional model. However, the D239A mutation decreased the apparent affinity for Glc-1-P by 31-fold and the V max by 11-fold without significant changes in the other kinetic constants. Likewise, D239N and D239E increased the S 0.5 for Glc-1-P by 16-and 10-fold, respectively, compared with the wild type. In contrast, the V max of the D239N mutant was 2-fold lower compared with that of the wild type and was not affected by the D239E substitution (Tables 1 and 2). Replacement of Phe 240 with alanine and methionine affected the apparent affinity for Glc-1-P, with S 0.5 values 12-and 7-fold higher, respectively, compared with the wild type. No significant changes in the V max and all other kinetic parameters analyzed here were observed with F240A and F240M (Tables 1 and 2). Together, these results suggest that Asp 239 and Phe 240 are important residues for Glc-1-P interaction. They also point out the significance of the Asp 239 negatively charged side chain for proper catalytic activity.
Thermal Stability-The enzymes were also studied with respect to their thermal stability as described under "Experimental Procedures." The wild-type enzyme and all Glu 194 , Tyr 216 , and Asp 276 mutants, as well as S212A, S212V, F240M, and D239E, showed ϳ80 -85% activity after heat treatment (Table 3).
It is interesting to note that mutation of Phe 240 to alanine caused the thermal stability of the protein to decrease by 50% under the assayed conditions. This result suggests that, at position 240 of E. coli ADP-Glc PPase, not only the hydrophobicity but also the size of the side chain is important for the enzyme to adopt proper and heat-stable folding. A similar situation was observed with Trp 274 . Replacement of this residue with alanine and leucine, two hydrophobic but small side chain amino acids, rendered enzymes with Ͻ1% residual activity after heat treatment, whereas phenylalanine in that position allowed the mutant enzyme to retain at least 50% of the activity (Table 3).
Mutants S212T and S212Y retained 60 and 70% of their initial activities, respectively. These mutations affected not only the apparent affinity for Glc-1-P but also the k cat , suggesting a structural distortion of the active site. It is possible that these side chains also misplace significant structural determinants or disrupt important stabilizing interactions in the protein. Mutations of Asp 239 rendered enzymes with 2, 59, and 84% residual activities after heat treatment when replaced with alanine, asparagine, and glutamic acid, respectively. A negative charge at position 239 is necessary to guarantee the stability of the enzyme at temperatures higher than the optimum for activity.

DISCUSSION
In this work, we have reported the first detailed characterization of the sugar phosphate site and the three-dimensional structure of E. coli ADP-Glc PPase, the Glc-1-P site. We selected a set of residues implicated in shaping this substrate pocket by examination of the primary sequences of several ADP-Glc PPases and the three-dimensional structural of the E. coli enzyme complexed with ADP-Glc obtained by homology modeling. The role of the selected residues in binding Glc-1-P was probed by site-directed mutagenesis and steady-state kinetics. The kinetic characterization of the individual mutants revealed the importance of the replaced amino acids.
Knowledge of the three-dimensional structure of E. coli ADP-Glc PPase is essential to understand the complex network of interactions established between the protein and the substrate for proper binding. The first published ADP-Glc PPase crystal structure is that of the homotetrameric potato tuber small subunit solved by Jin et al. (14), which we used as a template to build a model of the E. coli enzyme. The sequence identity between these two proteins is 33%, which is close to the lowest range of accepted homology for performing modeling (45). However, the functional similarity between our query and template proteins and a careful inspection of the sequence alignment, which included information on predicted secondary structures and functional conserved residues, increased the probabilities of obtaining a reliable model. E. coli ADP-Glc PPase has been the subject of numerous structure-function relationship studies, including those aimed to elucidate the functional role of individual amino acids. Previously, Lys 195 was studied (16) and showed a very specific effect on Glc-1-P interaction. The reported mutations of this residue increased by 100 -10,000-fold the S 0.5 for Glc-1-P without affecting other kinetic constants. To illustrate this, data reported for mutant K195Q have been included here in Tables  1 and 2. These results are consistent with a very specific role of Lys 195 in the binding of Glc-1-P, probably by ionic interaction between the positively charged side chain ⑀-amino group and the negative phosphate group of Glc-1-P (16,17). It is possible that the rest of the amino acids in the substrate pocket, which are the subject of this work, interact with the sugar hydroxyls to increase the affinity of the binding and to provide the correct positioning of the ligand for catalysis.
The three-dimensional model of ADP-Glc PPase that we obtained here allowed us to visualize the spatial arrangement of a set of conserved residues potentially involved in the interaction between the enzyme and the substrate Glc-1-P. It has been reported that, although proteins can bind carbohydrates in many different ways, certain amino acids show high propensity to be in a sugar-binding site (50,51). Some examples are aromatic rings that can pack against the hydrophobic face of a sugar (52) and carboxylates that can form bidentate hydrogen bonds with two adjacent hydroxyls of a saccharide (50). In our model, we identified Trp 274 , Tyr 216 , and Phe 240 , as well as Glu 194 , Asp 239 , and Asp 276 , some of which show direct contacts with the modeled ligand ADP-Glc. We also performed a close examination of the reported three-dimensional structures of enzymes that catalyze reactions very similar to those catalyzed by ADP-Glc PPase. These enzymes are P. aeruginosa RmlA (Protein Data Bank code 1G23) (18) and S. typhi CDP-Glc PPase (Protein Data Bank code 1TZF) (19). We inspected closely their active sites and identified residues homologous to Glu 194 , Lys 195 , Asp 276 , and Trp 274 , as well as the catalytic Asp 142 (40), of ADP-Glc PPase in their active sites (Fig. 5A). Interestingly, CDP-Glc PPase is a trimeric enzyme with three active sites formed in the interface of adjacent monomers (19). Most of the residues contributing to the architecture of the Glc-1-P site belong to one of the subunits, except for Glu 178 and Lys 179 , homologous to Glu 194 and Lys 195 in E. coli ADP-Glc PPase, which are provided by the neighboring subunit (19).
The ADP-Glc PPase structural model shows the hexose moiety of ADP-Glc largely engaged in hydrogen bonds to surrounding residues (side chains of Lys 195 , Glu 194 , Ser 212 , and Asp 276 ) (Fig. 2) and the protein backbone (Ser 212 and Gly 179 ) (Fig. 6). Lys 195 interacts with the ␤-phosphate of the ADP-Glc molecule. This observation is validated by the biochemical characterization reported by Hill et al. (16) and discussed above.
Glu 194 is proposed to interact with O-2 and O-3 of the sugar ring by a bidentate hydrogen bond. The Glu 194 mutants dis-played the greatest changes in Glc-1-P apparent affinity when substituted with other residues ( Table 1). Removal of the negative charge, as observed with the glutamine mutant, caused a large decrease in this kinetic parameter (85-fold), suggesting its importance for substrate interaction. Still, the size of the side chain is also essential given that substitution with aspartic acid decreased the apparent affinity for Glc-1-P by Ͼ380-fold. Given that the distance between two atoms engaged in a hydrogen bond is crucial for the establishment of such an interaction, the effect observed with a shorter side chain at position 194 supports the existence of a hydrogen bond between the ligand and Glu 194 . In addition, the enzyme activity seems to be affected by modifications at this position. It is possible that Glu 194 plays a key role in positioning the substrate in the correct orientation for catalysis, which also agrees with a critical contribution of size to the functionality of this residue. Our results support the central role of Glu 194 in Glc-1-P binding and explain the absolute conservation of this amino acid in the ADP-Glc PPase family (Fig. 3) and other NDP-Glc PPases such as RmlA and CDP-Glc PPase (Fig. 5A).
Ser 212 may bind Glc-1-P through hydrogen bonds between the side chain and backbone and O-3 and O-4 of the sugar ring, respectively (Figs. 2 and 6). We probed the role of the side chain OH group in this interaction by making conservative and nonconservative mutations. Although to various degrees, all Ser 212 mutants affected the apparent affinity for Glc-1-P. Homology modeling of the Ser 212 mutant active-site residues complexed with ADP-Glc showed that the interaction predicted in the wild-type enzyme model between the Lys 195 ⑀-amino group and the phosphate of the ligand is disrupted when Ser 212 is replaced with other amino acid (supplemental Fig. 1). The 14-fold increase in the S 0.5 for Glc-1-P caused by the S212A mutant might be explained by the disruption of one hydrogen bond between the side chain and O-3 of the glucose moiety of the ligand. Surprisingly, the effect of the side chain OH group provided by threonine is counteracted by the presence of an additional methyl group in comparison with serine. A similar situation is observed with valine in position 212. This extra methyl group largely disrupts the proper conformation of the binding pocket. The model predicts that Ser 212 is spatially close to secondary structures containing Glu 194 and Lys 195 which, as indicated previously, are important in Glc-1-P interaction. Ser 212 is also largely engaged in a hydrogen bond network with these structures (Fig. 6). These observations might explain how some of the mutations of Ser 212 affected the apparent affinity for Glc-1-P, as mutations of Glu 194 and Lys 195 did. Surprisingly, substitution of Ser 212 with a bulky side chain amino acid, tyrosine, caused a slight change in the apparent affinity for this substrate specifically. It is possible that, as the homology model predicts, the preferred rotamer for a tyrosine in this position is one directing the phenyl group away from the Glc-1-P pocket, burying the side chain against other hydrophobic side chains and stabilizing this position by a hydrogen bond between the phenyl OH group and an adjacent backbone (data not shown). It is possible that the burying of the phenyl group causes structural arrangements, which probably extend to other parts of the active site, affecting specifically an important catalytic residue. This would be explained by the dramatic reduction in the k cat  DECEMBER 29, 2006 • VOLUME 281 • NUMBER 52 displayed by mutant S212Y. Therefore, the side chain of Ser 212 might contribute to the overall affinity for Glc-1-P by making direct interactions with O-3 of the sugar ring and with adjacent backbones containing important residues for the positioning of this substrate. On the other hand, the model shows the Ser 212 peptide carbonyl group binding O-4 of the hexose through a hydrogen bond (Figs. 2 and 6). This interaction can also be observed in the crystal structures of the NDP-Glc PPases RmlA (18) and CDP-Glc PPase (19). The peptide carbonyl groups of Val 172 in RmlA and Asn 188 in CDP-Glc PPase, homologous to Ser 212 in ADP-Glc PPase (Fig. 6), also make hydrogen bonds with the substrate, implying that this interaction is important for the correct geometry of Glc-1-P in the binding pocket. Apart from the specific interactions, the size of the side chain is important for the proper architecture of the Glc-1-P-binding site.

Glucose 1-Phosphate Site from ADP-glucose Pyrophosphorylases
Asp 276 is important for the enzyme interaction with Glc-1-P, and it may bind O-6 of the hexose through a hydrogen bond (Fig. 2). Substitutions with other residues affected the apparent affinity for this substrate by ϳ25-100-fold, supporting this hypothesis. However, Asp 276 might have a broader role rather than exclusively interacting with the Glc-1-P molecule because the V max and the apparent affinity for the other substrates were also affected by the mutations studied (Tables 1 and 2). Asp 276 is spatially close to the catalytic Asp 142 (40), and its homologous residue in potato tuber ADP-Glc PPase (Asp 280 ) has been proposed as an Mg 2ϩ chelator (14). These observations explain why the different substitutions of Asp 276 also affected other kinetic parameters besides the Glc-1-P apparent affinity. In contrast to the mutations of other residues in the Glc-1-P site, the activation by Fru-1,6-P 2 was also altered in the Asp 276 mutants ( Table 2). The results obtained with EcN⌬15-D276N strongly suggest that this amino acid does not direct participate in activator binding.
Asp 276 may be located in a hinge-like region of the active site between the ATP and Glc-1-P subdomains. Apart from interacting with the sugar ring and Mg 2ϩ , it may also contact other residues from adjacent secondary structures, establishing a network of interactions driving the conformational changes experienced upon binding the substrates. Comparison of the potato tuber ADP-Glc PPase crystal structures complexed with ATP or ADP-Glc illustrates such subdomain movement (14). The observations of Haugen and Preiss (53) also contribute to explaining the negative effects on all the kinetic properties of the enzyme when Asp 276 was mutated. They demonstrate that (a) ATP alone displays half-site occupancy in the homotetrameric enzyme; (b) Glc-1-P does not bind to the enzyme unless MgCl 2 and ATP are present; and (c) ATP displays full-site occupancy in the presence of Glc-1-P. A synergistic effect on the binding of Fru-1,6-P 2 and ATP was also reported. Thus, the cooperative properties and the heterotrophic interactions between substrates and effectors (53) also explain the broad effect on the kinetic properties of the enzyme when the physiochemical properties of this strategically located residue are modified.
Aromatic residues, typically Trp and Phe, are key components of several saccharide-binding sites (52). Usually, these aromatic rings have been found to be involved in stacking inter-actions against the face of a sugar (50). However, in our structural model, none of the three aromatic residues in close proximity to the glucosyl moiety of the ligand orients its side chain parallel to the sugar ring. The great conservation of Trp 274 observed among ADP-Glc PPases (Fig. 3) and other pyrophosphorylases (Fig. 5A) might be explained by its structural role within the Glc-1-P site. Substitution with short aliphatic side chain amino acids such as alanine and leucine not only affected the apparent affinity for Glc-1-P (Table 1) but also greatly decreased the thermal stability of the enzyme (Table 3). These effects were less when Trp 274 was mutated to phenylalanine, suggesting that aromaticity is important at this position. This amino acid might provide the necessary stacking interactions to shape the Glc-1-P site correctly while establishing the proper hydrophobic interactions that increase the thermal stability of the protein.
Tyr 216 is also located close to the ligand, but no evident interaction is observed between the sugar ring and the side chain OH group. We evaluated the role of such a group in Glc-1-P binding with the Y216F mutation, which lower the V max by 10-fold and the apparent affinity for this substrate by 46-fold (Table 1). Tyr 216 is conserved in all ADP-Glc PPases studied so far (Fig. 3) and is present in RmlA (Tyr 176 ) (Fig. 5, A and B). In contrast CDP-Glc PPase bears a phenylalanine (Phe 192 ) in the homologous position, but Tyr 129 , located in an adjacent ␤-strand, orients its side chain so that the OH group overlaps with those of Tyr 216 in ADP-Glc PPase and of Tyr 176 in RmlA (Fig. 5B). Given the conservation of the aromatic ring at this position, it is possible that Tyr 216 plays a structural role in the Glc-1-P site architecture. On the other hand, the OH group could also make a hydrogen bond with a water molecule in direct contact with the substrate, as observed with Tyr 176 in RmlA (18). This interaction might be crucial to drive the correct positioning of the Glc-1-P molecule for the enzymatic reaction because not only the apparent affinity for this substrate but also the catalytic activity was affected by the removal of the side chain OH group.
We also analyzed the possible roles of Asp 239 and Phe 240 as part of the Glc-1-P site. Our results with mutant D239E showed that the change in size significantly affected the apparent affinity for Glc-1-P and that a negative charge at position 239 is necessary to maintain significant enzyme activity and thermal stability (Tables 1 and 3). On the other hand, substitutions with asparagine and alanine caused the greatest alteration in apparent affinity for Glc-1-P, catalytic activity (Table 1), and thermal stability ( Table 3). The structures of RmlA and CDP-Glc PPase show other hydrogen bond donors in the position homologous to Asp 239 : Glu 198 and Thr 208 , respectively (Fig. 5C). Moreover, in the RmlA structure, Glu 198 interacts with O-2 of the dTDP-Glc molecule through a bridging water molecule. Hydrogen bonds and ion pairs with ordered water molecules are considered important interactions that increase the thermal stability of the protein (54) and the binding affinity and specificity for the substrate (50). We cannot rule out the possibility that Asp 239 also interacts with the solvent, which is critical for the correct positioning of Glc-1-P and may also affect the enzyme activity.
The data obtained with Phe 240 mutants demonstrate that a hydrophobic bulky residue is needed to maintain the properties of the enzyme at wild-type levels. The role of Phe 240 might be merely structural, and the effects on Glc-1-P apparent affinity may be a consequence of the close proximity to Asp 239 . In the three-dimensional model, Phe 240 is surrounded by a hydrophobic environment, and it is probably necessary to anchor the loop containing Asp 239 in the correct position. Phe 240 is conserved in most of the ADP-Glc PPases (Fig. 3), except in those from the Mycobacterium sp. taxonomic group, which bear methionine in the homologous position. Similarly, CDP-Glc PPase has Trp 209 , whereas RmlA has a smaller hydrophobic residue (Ile 199 ) (Fig. 5C). Together with our biochemical results, these observations support the role of Phe 240 as an important structural component of the Glc-1-P site.
In this work, we have presented data supporting that key amino acids in ADP-Glc PPase have a role in the affinity of the enzyme for Glc-1-P. Whether establishing direct hydrogen bonds with the hydroxyls in the sugar ring or solvent molecules or properly shaping the substrate pocket, they all have an important role in determining the architecture of the Glc-1-P site. This is the first thorough biochemical characterization performed on ADP-Glc PPases. We combined biochemical data with information from the three-dimensional model, which allowed us to hypothesize the structural basis of substrate binding. Comparison of our model with other NDP-Glc PPases reveals remarkable similarities, suggesting that the architecture of the Glc-1-P site is conserved. Biochemical data involving the amino acids examined have not been reported on other PPases to date. We believe that the results reported in this work can be extended to other members of the NDP-Glc PPase family, providing new insights toward the understanding of the evolution of these enzymes.