Identification and Characterization of a Critical Region in the Glycogen Synthase from Escherichia coli*

The cysteine-specific reagent 5,5′-dithiobis(2-nitrobenzoic acid) inactivates the Escherichia coli glycogen synthase (Holmes, E., and Preiss, J. (1982) Arch. Biochem. Biophys. 216, 736-740). To find the responsible residue, all cysteines, Cys7, Cys379, and Cys408, were substituted combinatorially by Ser. 5,5′-Dithiobis(2-nitrobenzoic acid) modified and inactivated the enzyme if and only if Cys379 was present and it was prevented by the substrate ADP-glucose (ADP-Glc). Mutations C379S and C379A increased the S0.5 for ADP-Glc 40- and 77-fold, whereas the specific activity was decreased 5.8- and 4.3-fold, respectively. Studies of inhibition by glucose 1-phosphate and AMP indicated that Cys379 was involved in the interaction of the enzyme with the phosphoglucose moiety of ADP-Glc. Other mutations, C379T, C379D, and C379L, indicated that this site is intolerant for bulkier side chains. Because Cys379 is in a conserved region, other residues were scanned by mutagenesis. Replacement of Glu377 by Ala and Gln decreased Vmax more than 10,000-fold without affecting the apparent affinity for ADP-Glc and glycogen binding. Mutation of Glu377 by Asp decreased Vmax only 57-fold indicating that the negative charge of Glu377 is essential for catalysis. The activity of the mutation E377C, on an enzyme form without other Cys, was chemically restored by carboxymethylation. Other conserved residues in the region, Ser374 and Gln383, were analyzed by mutagenesis but found not essential. Comparison with the crystal structure of other glycosyltransferases suggests that this conserved region is a loop that is part of the active site. The results of this work indicate that this region is critical for catalysis and substrate binding.

The reaction was described in bacteria in 1964 (1) and its equilibrium is strongly shifted toward the formation of ADP (2). The glycogen synthase from E. coli has been purified to homo-geneity (2), and some of its properties have been characterized (3)(4)(5)(6). Bacterial glycogen synthases and plant starch synthases catalyze the same chemical reaction and share between 30 and 36% identity, suggesting that they could have a common structure and catalytic mechanism. Both of them use ADP-Glc as a glucosyl donor and have molecular masses of 48 -55 kDa, whereas the yeast and mammalian glycogen synthases have molecular masses of 70 -85 kDa and prefer UDP-glucose. Another feature that differentiates the latter enzymes from bacterial and plant glycogen/starch synthases is their regulation by phosphorylation and allosteric activation by glucose 6-phosphate (7,8). Malfunction of human glycogen synthase has been associated with several metabolic diseases, such as diabetes mellitus and glycogen storage disease (9,10). Although there is no obvious sequence similarity between bacterial and mammalian glycogen synthases, it has been proposed that they share a common structure and catalytic mechanism. This was based on prediction of the secondary structure, threading, and hydrophobic cluster analysis (11,12). Despite the importance of these enzymes, little is known about the structure of the catalytic site. More knowledge of the structure-function relationship in bacterial glycogen synthases could be instrumental in understanding the molecular basis of those disorders.
Few studies have been made to elucidate possible substratebinding residues, in both bacterial glycogen synthases and plant starch synthases, but the nature of the substrate-binding site remains unclear. Furukawa et al. (13) showed that ADP-Glc protects Lys 15 in the E. coli glycogen synthase from reaction with pyridoxal phosphate. Replacement of Lys 15 by Arg, Gln, or Glu, increases the S 0.5 7-, 32-, and 46-fold, respectively (13). In the starch synthase IIa from maize endosperm, Argspecific modification experiments were performed with phenylglyoxal (14). However, the studied Arg residues are not 100% conserved, and, when mutated, no significant shifts in S 0.5 for ADP-Glc were shown. It has been shown that cysteine-reactive reagents inactivate E. coli glycogen synthase, and that this effect could be prevented by the substrates (15). However, the residues in E. coli glycogen synthase that are involved in this inactivation have not been identified. DTNB (5,5Ј-dithiobis(2nitrobenzoic acid)) has the ability to form mixed disulfide bridges with cysteines, and it has been used to identify the ones involved in catalysis or substrate binding in different enzymes (16 -21). In the present study, a combinatorial approach was used to mutate the three cysteines of the E. coli glycogen synthase. This strategy enabled us to determine which of those cysteines were responsible for DTNB modification and concurrent loss of activity. In addition, the kinetic properties of the mutants were analyzed to ascertain the functional role of the Cys involved. Because this Cys is in a putative loop with conserved residues that could be interacting with the substrate, we characterized the region by site-directed mutagenesis and chemical modification studies. 14 C]Glucose 1-phosphate was obtained from ICN Pharmaceuticals. ADP-[ 14 C]Glc was synthesized as previously described (22). Oligonucleotides were synthesized and purified by the Macromolecular Facility at Michigan State University. Pfu DNA polymerase was purchased from Stratagene. Rabbit liver glycogen, DTNB, iodoacetic acid (IAA), iodoacetamide, and iodopropionic acid were purchased from Sigma. Alkaline phosphatase-linked mouse anti-rabbit IgG and BM purple AP-substrate precipitating reagent were from Roche (Indianapolis, IN). All other reagents were purchased at the highest quality available. ADP-glucose pyrophosphorylase from E. coli was purified as described (23).

Materials-[
Construction of the Plasmids-Plasmid pAY1 was generated by subcloning the GS-encoding gene (glgA) from E. coli into the NdeI-SacI sites of the pET-24a vector (Novagen). For that purpose, a PCR product was obtained using as primers 5Ј-GGAGCGCATATGCAGGTTTTA-CATG-3Ј and 5Ј-GTCAGTTATTGCTCAGCGG-3Ј, to introduce an NdeI site at the N-terminal. The insert was checked by DNA sequencing and was consistent to the GS gene present in data bases for the E. coli genome project (24). Plasmid pAY3 was generated by subcloning the glgA gene from E. coli into the NdeI-XhoI sites of pET-24a. This construct attached a His Tag® sequence (6 consecutive histidine residues, Novagen) to the 3Ј end of glgA.
Glycogen Synthase Assay-The assay was conducted as previously described (1) in the presence of 1 mM ADP-[ 14 C]glucose (ϳ500 dpm/nmol or ϳ1500 dpm/nmol), 0.01 M magnesium acetate, 2.5 mg/ml rabbit liver glycogen, 50 mM Bicine-NaOH (pH 8.0), and 0.5 mg/ml bovine serum albumin in a total volume of 200 l, unless otherwise stated. The reaction was stopped by boiling 1 min and the formed glycogen was precipitated by adding 2 ml of a 75% methanol, 1% KCl solution, washed twice before dissolving in water, and the radioactivity was measured in a scintillation counter. One unit of the enzyme is defined as to 1 mol of glucosyl units transferred per minute at 37°C.
Expression and Purification of the Enzymes-E. coli BL21(DE3) cells harboring pAY1, either wild-type or with different mutations, were grown at 37°C up to A 600 ϭ ϳ0.6. At that point, 1 mM isopropyl-1-thio-␤-D-galactopyranoside was added and the induction was carried out at room temperature (23-25°C) for 16 h. The cells were collected by centrifugation and resuspended in 50 mM potassium phosphate (pH 6.8), 5 mM dithiothreitol, and 0.1 M NaCl. The resuspension was sonicated, centrifuged at 10,000 ϫ g for 15 min, and the supernatant (crude extract) was precipitated with 25% ammonium sulfate. The pellet was resuspended in 20 mM triethanolamine-HCl (pH 7.5) and desalted with Econo-Pac® Bio-Rad 10 DG columns equilibrated with the same buffer. Dithiothreitol was excluded from this buffer to avoid interference with the DTNB modification experiments. The protein sample was loaded onto a Mono Q HR 5/5 (FPLC, Pharmacia) column and eluted with a linear KCl gradient (20 column volumes, 0 -0.6 M). The active fractions of the Mono Q were monitored by SDS-PAGE to pool the purest fractions. The sample was concentrated, brought to 20 mM triethanolamine-HCl (pH 7.5) using the Ultrafree®-15 centrifugal filter device (Millipore) with a Biomax-10 membrane, and stored at Ϫ80°C. BL21(DE3) cells harboring pAY3 or their mutated derivatives were grown following the same procedure. The cell pellet was resuspended in 50 mM sodium phosphate buffer, 300 mM NaCl (pH 7.8), and sonicated. After centrifugation, the crude extract was loaded into a nickel-charged NTA-agarose column (nickel-nitrilotriacetic acid HisBind® Resin, Novagen) and washed with 20 column volumes of the same buffer plus 10 mM imidazole. His-tagged glycogen synthase was eluted with 8 column volumes of the same buffer plus 250 mM imidazole. The eluted sample was desalted with Econo-Pac 10 DG columns, concentrated, and stored at Ϫ80°C. Protein Electrophoresis and Immunoblotting-SDS-PAGE was performed according to Laemmli (25) using 4 -15% Tris-HCl pre-cast gradient gels from Bio-Rad. After electrophoresis, proteins were visualized by staining with Coomassie Brilliant Blue R-250 or electroblotted onto a nitrocellulose membrane. The nitrocellulose membrane was then treated with anti-E. coli glycogen synthase rabbit IgG (3) and the antigen-antibody complex was visualized via treatment with alkaline phosphatase-linked mouse anti-rabbit IgG and subsequent staining with BM purple AP-substrate precipitating reagent.
Inactivation by DTNB-Inactivation of the E. coli glycogen synthase by DTNB was carried out at 0°C in a reaction mixture that contained 0.5-1 g of enzyme dissolved in 50 mM Bicine buffer (pH 8.0), in a total volume of 100 l. The reaction was started by adding 5 l of DTNB (pH 7.8) to obtain a final concentration of 10 M. At different time points, aliquots were removed from the reaction mixture and assayed for remaining activity. Protection by substrates was studied in the same way, but the reaction mixture contained 1 mM ADP-Glc, 2.5 mg/ml rabbit liver glycogen, or the indicated ligand.
DTNB Modification-Extent of modification of the sulfhydryl groups of the glycogen synthase mutants was analyzed spectrophotometrically. Each purified enzyme (10 M) was added into a cuvette containing 50 mM Bicine/NaOH buffer (pH 8.0) with or without the indicated ligand in a total volume of 200 l. The reaction was started with the addition of 4 l of DTNB (final concentration of 40 M) and the visible spectrum was recorded after 10 s at room temperature (23-25°C). Spectra were taken 10 s after the addition of DTNB because the reaction was instantaneous at room temperature. Incubations of more than 1 min sometimes led to an increase in turbidity at such high concentrations of glycogen synthase. Spectra were taken every 0.5 nm from 340 to 600 nm in a Beckman DU650 spectrophotometer at a scan speed of 600 nm/min. The first derivative was calculated with a window of 2 nm and smoothed by the Savistky-Golay (26) algorithm with a window of 41 points and coefficients of second degree.
Kinetic Characterization-The kinetic data were plotted as initial velocity (nmol/min) versus substrate or effector concentration. The kinetic constants were acquired by fitting the data to the Hill equation (27) with a nonlinear least square formula using the program Origin TM 5.0. The constant n H is the Hill coefficient. S 0.5 is the concentration of substrate giving 50% of the maximal velocity. I 0.5 is the concentration of DTNB that inactivated 50% of the glycogen synthase after 10 min of incubation. K I is the dissociation constant of the enzyme-inhibitor complex, and was determined by means of a replot of the data taken from the reciprocal plot (28).
Chemical Modification with Iodoacetic Acid-Chemical modification of the glycogen synthase quadruple mutant C7S,E377C,C379S,C408S (TM-E377C) was performed at 0°C in the dark, in a reaction mixture containing 0.1 mg of enzyme in 50 mM Bicine-NaOH (pH 8.0) in a total volume of 100 l. The reaction was started with the addition of 20 mM iodoacetic acid reaching a final concentration of 1 mM. At different time points, 2-l aliquots were withdrawn and assayed for activity. To prepare modified enzyme for kinetic studies, the reaction was stopped after 10 min by adding dithiothreitol to a final concentration of 20 mM.
Co-sedimentation with Glycogen-The procedure is a modification of a previously published protocol (29). Each purified recombinant enzyme (10 g) was incubated in 1 ml of 50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 0.1% (v/v) 2-mercaptoethanol, 0.1 mg/ml bovine serum albumin containing 10 mg/ml rabbit liver glycogen at 4°C for 30 min. After centrifugation at 100,000 ϫ g for 90 min, the supernatant and the pellet fractions were collected and subjected to SDS-PAGE and activity determination.
Structure Prediction and Homology Modeling-Secondary structures were predicted with the PHD program (30) and the PSI-PRED method (31). Those programs are available on the PredictProtein server 2 and on the PSI-PRED server (32), 3 respectively. Homology modeling of domain II of the E. coli glycogen synthase was performed using the Swiss-Model protein-modeling server (33) (Glaxo Wellcome Experimental Research, Geneva, Switzerland). 4 The known atomic coordinates of domain II of the E. coli trehalose-6-phosphate synthase complexed with UDP (34) (Protein Data Bank code 1gz5) were used as a template. The reliability of the model was evaluated using the Verify3D program (35) and WHATCHECK (36) (WHAT IF program (37)).

Expression and Purification of Wild Type and Mutant
Glycogen Synthases-Overexpression of the wild-type recombinant glycogen synthase was performed as described under "Experimental Procedures." In this system, the recombinant enzyme was typically obtained in crude extracts with a specific activity of 70 -100 units/mg. This was at least 6000-fold higher than the endogenous activity from BL21(DE3) cells transformed with pET-24a (data not shown). Western blot analysis showed that the expression of the mutant proteins was similar to the wildtype enzyme and all of them were purified to apparent homogeneity determined by SDS-PAGE (data not shown). The mutants with low specific activity were purified using a histidine tag to ensure that no endogenous protein was being co-purified. The expression level of the wild-type enzyme with a C-terminal histidine tag was similar to the wild-type enzyme as determined by Western blotting (data not shown). Controls were performed during the purification procedure to assure that no endogenous glycogen synthase was being co-purified. No glycogen synthesis activity (Ͻ0.002 units/mg) was detected in the eluates of the nickel-nitrilotriacetic acid resin when we loaded crude extracts of BL21(DE3) cells transformed with pET-24a. The enzymes purified using a histidine tag were homogeneous when subjected to SDS-PAGE (data not shown).
DTNB Inhibition of the E. coli Glycogen Synthase-We replaced Cys 7 , Cys 379 , and Cys 408 by Ser and measured the activity in the presence of different concentrations of DTNB. The wild-type enzyme was strongly affected by DTNB, with an I 0.5 of 1.2 Ϯ 0.1 M. The I 0.5 for mutants C7S and C408S were 7.8 Ϯ 2.8 and 7.0 Ϯ 1.1 M, whereas the I 0.5 for C379S and the triple mutant (TM) increased to 130 Ϯ 30 and 1200 Ϯ 130 M, respectively. The replacement of Cys 379 was the one that made the enzyme most resistant to inactivation by DTNB. However, these results also suggested that DTNB at high concentrations interfere with the glycogen synthesis reaction because the triple mutant, with no cysteines present, was inhibited at high concentrations of DTNB.
The modification by DTNB on single cysteines was studied on double mutants with only one cysteine present (DM1, DM2, and DM3). To reduce nonspecific effects as the ones mentioned above, these studies were performed in two steps, separating the chemical modification by DTNB and the analysis of activity. The enzymes were incubated with DTNB and aliquots were withdrawn at different times to assay the activity, diluting 100-fold the DTNB carried into the assay mixture. With this approach, the wild-type enzyme was rapidly inactivated by DTNB, whereas TM was not affected. The wild-type enzyme had a t 0.5 of inactivation of less than 1 min at 0°C (Fig. 1) and TM remained active even after 20 min (data not shown). DM2 was inactivated as the wild-type enzyme but DM1 and DM3 were insensitive to DTNB treatment (Fig. 1). Because DM2 has only Cys 379 present, this must be the residue responsible for the inactivation by DTNB. The presence of 1 mM ADP-Glc in the incubation mixture prevented the inactivation by DTNB of both the wild-type and DM2. On the other hand, 2.5 mg/ml glycogen did not significantly protect glycogen synthase (Fig.  1). Inactivation by DTNB seems to be reversible. After a 20-min incubation, treatment with 5 mM dithiothreitol recovered the activity of DM2 and the wild-type enzyme, 95 and 70%, respec-tively (data not shown). This indicates that in the conditions of this experiment, DTNB only affects the activity through its ability to covalently modify sulfhydryl groups.
Protection of the DTNB Modification-The above mentioned experiments showed that DTNB inactivated the enzyme only if Cys 379 was present. To verify that the inactivation proceeded through a chemical modification of this particular residue, the cleavage of DTNB was followed spectrophotometrically in the presence of the different glycogen synthase mutants. First the derivative spectra of DTNB were recorded to increase the sensitivity. As a positive control, 10 M Cys solution altered the DTNB spectra and a negative peak appeared at 445 nm (Fig. 2). A similar change in the spectrum was obtained by the addition of 10 M wild-type glycogen synthase, showing that ϳ1 sulfhydryl/molecule of enzyme reacted with DTNB. ADP-Glc prevented the spectral change, indicating that it protects the enzyme from the modification by DTNB (Fig. 2). A negative control in the presence of the mutant with no Cys (TM) showed that DTNB did not change its spectra, either in the presence or absence of ADP-Glc. Double mutants were analyzed to confirm which Cys was responsible for the spectral changes of DTNB. DM2, which has only Cys 379 in its sequence, was the only double mutant that reacted with DTNB as the wild-type (Fig.  2). ADP-Glc also prevented the reaction of DTNB with DM2 (Fig. 2).
Kinetic Characterization of the Cys Mutant Enzymes-The purified wild-type and mutant glycogen synthases were assayed to compare their kinetic parameters. The S 0.5 for ADP-Glc determined for the wild-type enzyme was 11 M, which agrees with previously reported data (2) ( Table I). The S 0.5 for C7S and the C408S mutants were not significantly different, whereas the values for C379S and TM increased 38-and 56fold, respectively. The specific activities of mutants C7S and C408S were 40 -60% of that of the wild-type, whereas mutant C379S and TM decreased 6-and 12.5-fold, respectively. DM2 showed a S 0.5 for ADP-Glc of 18 M and the specific activity was not significantly different from the wild-type. Conversely, DM1 and DM3 had an S 0.5 in the same order as the TM, nearly 60-fold higher than the wild-type. DM2 had about 90% of the wild-type enzymatic activity, whereas DM1 and DM3 were 22-24% active. When Cys 379 was substituted by Ala or Thr, the S 0.5 increased 77-and 123-fold, respectively. The substitution of Cys 379 by Asp yielded an enzyme with no detectable glycogen 4 www.expasy.ch/swissmod. synthesis activity. The S 0.5 for glycogen was also determined for the wild-type and the mutant enzymes. The S 0.5 of glycogen for different mutants were in a narrow range, from 5-fold lower to 2-fold higher than the wild type.
Inhibition by Glc-1P and AMP-Both Glc-1P and AMP were inhibitors of the wild-type glycogen synthase competing with the substrate ADP-Glc with a K I of 1.15 and 0.17 mM, respectively. When the same parameter was calculated for the mutant C379A, the K I for AMP showed a 2.8-fold increase (Table  II), whereas the K I for Glc-1P increased 75-fold with respect to that of the wild type.
Scanning of Essential Residues-As Cys 379 is located in a conserved region that is predicted to be a loop (Table III), we decided to study its importance for enzymatic activity. The side chain of each of the 100% conserved residues in that putative loop was replaced by Ala and the mutant enzymes obtained were characterized. The residues studied were Ser 374 , Glu 377 , Cys 379 , and Gln 383 .
Kinetic Characterization of Wild-type and Glu 377 Mutant Enzymes-The kinetic parameters of glycogen synthase with a histidine tag were very similar to the wild-type enzyme (Table  2. First derivative of the visible spectra of DTNB in the presence of different glycogen synthase mutants. The spectra and the reaction were carried out as described under "Experimental Procedures." All spectra were recorded in the presence of 40 M DTNB with the indicated enzyme either in the absence (control) or presence of 1 mM ADP-Glc. IV). The mutant enzymes were assayed as described under "Experimental Procedures" to compare their kinetic parameters with the wild type. The specific activity of E377A was 4 orders of magnitude lower than the wild-type enzyme and an even bigger effect was observed for the E377Q mutant (Table  IV). Despite this dramatic effect in the catalytic efficiency, the apparent substrate affinity for ADP-Glc did not change significantly for E377A and E377Q. The specific activity of E377D decreased 57-fold but this mutant showed an alteration in the apparent affinity for ADP-Glc. The apparent affinity for glycogen seems to be higher for these mutants (Table IV). Restoration of Activity by Chemical Modification at Position 377-To assess if the presence of a negative charge in position 377 could restore the enzymatic activity of an inactive mutant, we replaced Glu 377 with Cys, which is reactive to carboxymethylation by IAA. To avoid interference with other Cys, this experiment was performed with the triple mutant. The purified mutant TM-E377C (C7S,E377C,C379S,C408S) was incubated with 1 mM IAA, and aliquots were withdrawn and activity was measured at different times as described under "Experimental Procedures." The replacement of Glu 377 by Cys generated an enzyme with extremely low enzymatic activity, 2,300-fold less than the TM (Table V). Incubation with 1 mM IAA reactivated the TM-E377C enzyme to the same activity level of the TM (Fig. 3). On the other hand, incubation with iodoacetamide did not increase the activity of TM-E377C. This control indicates that a carboxyl group but not a carboxyamido group in this position restores the activity. ADP-Glc and ADP, but not glycogen, prevented the restoration of activity by IAA (Fig. 3). Incubation with iodopropionic acid did not affect the activity of the TM-E377C mutant (data not shown). The kinetic parameters of the carboxymethylated TM-E377C enzyme were in the same range of the TM enzyme (Table V).
Co-sedimentation with Glycogen-To determine whether the mutant enzymes with extremely low activity (4 order of magnitude less active than the wild type) retained its ability to bind to the substrate glycogen, we ultracentrifuged the enzymes in the presence of 10 mg/ml rabbit liver glycogen. In these conditions, the wild-type enzyme was found mainly in the pellet as determined by SDS-PAGE (Fig. 4). After ultracentrifugation, 98% of the activity was found in the pellet and only 2% in the supernatant (data not shown). As a control, when centrifuged in the absence of glycogen, the wild-type enzyme was found in the supernatant (Fig. 4). The inactive mutants E377A, E377Q, and TM-E377C followed the same pattern of co-sedimentation as the wild type. As a negative control, ADP-glucose pyrophosphorylase from E. coli does not co-sediment with glycogen ( Fig. 4).
Kinetic Characterization of Mutants S374A and Q383A-Purified mutants S374A and Q383A were analyzed kinetically (Table VI). Mutant S374A showed no significant differences in the kinetic parameters compared with the wild-type enzyme. The specific activity was comparable, and the apparent affinities for both substrates were in the same range as the wild-type enzyme. However, the specific activity of mutant Q383A decreased 23-fold. The interaction with the substrates of the Q383A mutant did not seem to be impaired. The apparent affinity for ADP-Glc remained unchanged, whereas that for glycogen was 2 orders of magnitude higher (Table VI). DISCUSSION In this work, we report the results of a chemical modification and mutational analysis of a putative loop that plays a critical role in the glycogen synthase from E. coli. We first found that Cys 379 is relevant for the interaction with the substrate ADP-Glc and the analysis was extended to other conserved residues of that area.
It has been reported that DTNB inactivates the enzyme, but it had not been determined which cysteine was involved in the process (15). We replaced each of the three Cys by Ser and analyzed the activity in the presence of different concentrations of DTNB to find out the responsible residue. Mutants C379S and TM were clearly the most resistant to inhibition, but DTNB seemed to interfere per se with the catalytic reaction at high concentrations. For that reason, further experiments were performed in two steps, separating the modification by DTNB from the assay reaction. Double mutants with only one cysteine were used to investigate which sulfhydryl group was responsible for the DTNB-mediated inactivation. When Ser replaced Cys 379 , in any combination, the enzyme was no longer inactivated by DTNB. On the other hand, ADP-Glc prevented inactivation of the wild-type enzyme or the DM2, both of which contain Cys 379 . The inactivation by DTNB correlates with the modification of one sulfhydryl group per molecule of wild-type glycogen synthase (Fig. 2). The other two cysteines only reacted when the enzyme was previously denatured with 6 M guanidine chloride (data not shown). This suggests that Cys 7 and Cys 408 are most probably buried in the structure of the enzyme. The only double mutant with a sulfhydryl group reactive to DTNB  is DM2 (Fig. 2). Therefore, Cys 379 must be the only residue in the wild-type glycogen synthase that is reactive to DTNB modification. ADP-Glc prevented the modification of Cys 379 and further inactivation of the enzyme, strongly suggesting that this residue is located in or near the substrate site. Kinetic analysis of the Ser mutant enzymes revealed the importance of Cys 379 , compared with the other cysteines. Replacement of Cys 379 by Ser decreased the apparent affinity for ADP-Glc 38-fold, whereas mutations on Cys 7 and Cys 408 did not change it significantly (Table I). This effect was specific for the substrate ADP-Glc because the S 0.5 for glycogen was not altered significantly in any of the single mutants (Table I). All the single mutations, such as C7S, C379S, and C408S, slightly affected the catalytic efficiency. They had, respectively, a 2.4-, 5.8-, and 1.6-fold lower specific activity than the wild type. We replaced Cys 379 by other amino acids to study the effect of different groups at this position on the kinetic properties of the enzyme. We found that C379A and C379T had a 77-and 123fold lower apparent affinity for ADP-Glc than the wild type (Table I). These significant shifts in the S 0.5 indicate that Cys 379 is involved in the interaction with the ADP-Glc or contributing to the proper architecture of its binding site. There does not seem to be an obvious role of this residue in glycogen binding because single mutations did not affect the S 0.5 for glycogen more than 1.8-fold. Moreover, double mutant DM1 and the triple mutant had a lower S 0.5 for glycogen than the wild type, whereas the individual mutations showed the opposite effect. It is possible that these mutations caused minor perturbations in the structure that were reflected in those slight kinetic changes.
The V max of the C379S mutant is only six times lower than the wild-type, and is not sufficient to assign a specific catalytic role to Cys 379 . Mutation of Cys 379 for Ala mainly affected the affinity for ADP-Glc, as the S 0.5 increased 77-fold and the V max decreased only 4-fold. However, other mutations at this site showed a greater effect on the catalytic efficiency. A mutation to Thr had an even higher S 0.5 for ADP-Glc (123-fold) but the V max was only 0.3% of the wild-type. It is possible that a steric hindrance is responsible for this effect because the difference between Thr and Ser is only a methyl group. In fact, when a bulkier group such as the Asp group replaced Cys 379 the activity of the enzyme decreased more than 4 orders of magnitude ( Table I). Mutation of Cys 379 by Leu generated an inactive unstable form that was proteolyzed in the cell, yielding a polypeptide of 41 kDa (data not shown). Replacement of Cys 379 by residues of smaller or similar size (Ala, Ser) decreased the affinity for ADP-Glc and only when bigger groups were used (Thr, Asp, and Leu) was catalysis severely affected.
Substrate saturation analysis of the purified mutants indicates that Cys 379 is involved directly or indirectly in the interaction of the enzyme with ADP-Glc. To further investigate what portion of the ADP-Glc molecule is engaged in this interaction, we studied the affinity of both the enzyme and the mutant C379A for Glc-1P and AMP. Both AMP and Glc-1P can be considered as portions of the molecule of the substrate ADP-Glc. A differential effect on the affinity for these mole-   cules would indicate what part of ADP-Glc is involved in the studied interaction. In this case, AMP and Glc-1P worked as inhibitors of glycogen synthase competing with ADP-Glc. This behavior is expected for compounds that bind into the substrate pocket but do not participate in the catalytic reaction. For a competitive inhibitor, the inhibition constant (K I ) is indicative of the affinity for the enzyme. Replacing Cys 379 for Ala slightly changed the K I of AMP indicating that this residue is not critical for the binding of the nucleotide monophosphate. On the other hand, mutation C379A severely affected the affinity for Glc-1P in the same magnitude as it did to the apparent affinity for ADP-Glc (75-and 77-fold, respectively; Tables I and  II). This indicates that Cys 379 is important for the interaction of the enzyme with the phosphoglucose moiety of the ADP-Glc and not with the nucleoside portion.
Role of the Putative Loop Bearing Cys 379 -The prediction of the secondary structure of the glycogen synthase from E. coli classifies the enzyme as an ␣/␤ protein (38). It has been observed that in enzymes with ␣/␤ structures the active sites are constructed with residues in the loops between each strand of a ␤-sheet and its following helix (39). Particularly, this has been confirmed among the superfamily of glycosyltransferases (38). The region comprised by residues 373-380 is predicted as a loop between a ␤-sheet and an ␣-helix (Table III). Taking into account that Cys 379 was found to be important for the interaction with the substrate, and that other amino acids in this region are highly conserved, our hypothesis was that this putative loop is near the substrate site and could play other important roles for the enzyme. We decided to characterize the functional role of other residues of this area by site-directed mutagenesis.
Replacement of Glu 377 for Ala or Gln rendered an enzyme with a 10,000-or 25,000-fold decrease in specific activity, respectively. We studied in detail mutants E377A and E377Q and found that, despite being essentially inactive, retained the ability to bind both substrates. This suggests that Glu 377 is not involved in binding of the substrates, but rather plays a major role in catalysis. When we replaced Glu 377 by Asp, we obtained a mutant enzyme with a 57-fold decrease in catalytic activity. Although this specific activity was significantly lower, it was still considerable, pointing out the importance of a negative charge in position 377. This was further confirmed when the catalytic activity of mutant TM-E377C was restored by carboxymethylation. These chemical reactivation experiments also indicate that it is extremely unlikely that mutants of the Glu 377 lost activity because of the generation of misfolded inactive forms.
Ser 374 is not only a conserved residue among bacterial glycogen synthases and plant starch synthases, but also among mammalian and fungal glycogen synthases. However, replacement of Ser 374 by Ala did not significantly change the kinetic properties of the enzyme. Despite its conservation, this residue does not play an essential role in the reaction mechanism. On the other hand, mutagenesis of Gln 383 generated an enzyme with no decrement in the apparent affinities for both substrates, but with a 23-fold decrease in V max . This Gln seems to contribute to the catalytic efficiency but it cannot be assigned as essential. It is also possible that this residue serves a local structural purpose.
Relationship with Other Glycosyltransferases-Glycosyltransferases have been divided into 65 families based on sequence similarity to founding members with experimentally demonstrated glycosyltransferase activity (40). 5 However, upon analysis of the growing number of crystallographic data, it has been suggested that this separation into families is concealing fundamental structural relationships (41). All these families have been recently grouped into three different folds, 5 afmb.cnrs-mrs.fr/CAZY. GT-A, GT-B, and GT-C (42). Bacterial glycogen synthases belong to glycosyltransferase family 5 and would have a GT-B fold by several structure prediction analyses (38,(41)(42)(43). This family 5 seems to share more structural similarity with the glycosyltransferases from families 3 (eukaryotic glycogen synthases) and 4 (sucrose synthase, trehalose phosphorylase, and mannosyltransferases). From these three families, no threedimensional structures are available. Family 5 is also structurally related, although more remotely, to family 35 (glycogen, starch, and maltodextrin phosphorylase) and family 20 (trehalose-6-phosphate synthase). Crystal structures from members of these two families are available (34,44,45) and have a GT-B fold. This fold comprises two Rossmann-like domains, each composed of a "sandwich" of parallel ␤-sheets between ␣-helices (38,43). All members of families 3, 4, 5, 20, and 35 retain the anomeric configuration of the monosaccharide transferred. The sequence similarities among these families have been discussed, and a conserved negatively charged residue has been localized in a loop that would correspond to the region studied in this work (38,41).
In this work, we described the first essential catalytic residue found in bacterial glycogen synthases. Despite the extremely low similarity in sequence, prediction analysis of the secondary structure and further alignment can identify this homologous residue among members of other glycosyltransferases that use sugar-nucleotide donors (38). Studies in enzymes of those other families provided evidence of the importance of a negative charge in this region. Mutation E510A of a recombinant fusion of the human muscle glycogen synthase (family 3) with green fluorescent protein reduced the enzymatic activity in crude extracts from 97 to 12 milliunits/mg (12). In that work, it was not possible to accurately determine the decrease of catalytic efficiency because of the very high endogenous activity present in COS-1 cells (12 milliunits/mg). In starch synthases (family 5), this putative loop has also conserved homologous residues to Ser 374 , Glu 377 , Cys 379 , and Gln 383 . A scanning of acidic residues detected that Glu 391 (homologous of Glu 377 ) mutated to Gln was inactive in crude extracts, but no further studies were performed on this mutant (46). The bacterial ␣-mannosyltransferase AceA (family 4) was mutated in Glu 287 . The mutant E287A, in a partially purified fusion protein with an S-tag, lost the ability to transfer mannose from GDP-mannose to the acceptor Glc␤(1-4)Glc␣-P-Ppolyprenol. However, the extent of this reduction was not quantified (47). To assess a residue as catalytic by site-directed mutagenesis is necessary to meet some requirements. It is expected that the reduction in specific activity or k cat should be several orders of magnitude. For that reason, the mutant should be purified in a system without interfering endogenous activity. In addition, there should be evidence that the ability to bind each of the substrates has not been severely altered. Finally, the possibility should be ruled out that the mutant is expressed as a misfolded form; i.e. the original residue plays a structural role in the enzyme rather than a functional role. Despite this negatively charged residue was found important among members of other families, this is the first time that the above mentioned requirements were met to assign a catalytic role in glycosyltransferases that use sugar-nucleotides as donors.
Our results with Glu 377 mutants are in good agreement with mutagenesis studies on the maltodextrin phosphorylase. In this family of oligo-or polysaccharide phosphorylases (family 35), a negative charge in this region has been postulated to play a role in catalysis. Despite that the substrates of the phosphorylases are not a sugar-nucleotide, there are many similarities. Not only that they have the GT-B fold (38,(41)(42)(43), but also the reverse reaction is chemically equivalent to the bacterial glycogen synthases forward reaction. An ␣1,4-glucosyl bond is formed upon breakage of a phosphoester bond with the OH1 of ␣-glucose. The residue with the negative charge, Glu 672 in the rabbit muscle glycogen phosphorylase b, interacts with OH2 and OH3 of the glucose of the non-reducing end. Glu 672 has been postulated, based on the crystal structure, to be involved in binding (44). However, biochemical evidence indicated that the main contribution is catalytic. This conserved residue has been mutated in the E. coli maltodextrin phosphorylase (Glu 637 ) by Asp and Gln and it showed a reduction in k cat of 600-and 900-fold, respectively (48).
FIG. 6. Comparison of conserved residues of loops after ␤-sheets IIE1 and IIE4 in retaining glycosyltransferases. The crystal structure of maltodextrin phosphorylase (Protein Data Bank code 1l5v) is depicted with Arg 534 , Lys 539 , and Glu 637 residues that are postulated to be involved in the catalytic reaction (48,50,52). Residues in the same position are shown for the crystal structure of trehalose-6-phosphate synthase (Protein Data Bank code 1gz5) and a homology model of the E. coli glycogen synthase built as described under "Experimental Procedures." In all cases only ␤-sheets IIE1 and IIE4 and helices IIH1 and IIH4 are shown. This nomenclature corresponds to Fig. 5 (38).
Inhibition studies by Glc-1P and AMP on the wild type and mutant C379A indicated that the conserved Cys 379 interacts with the phosphoglucose moiety of the ADP-Glc rather than the nucleoside. These results correlate with the prediction of the structure of the E. coli glycogen synthase (38). The Cys 379 is in a putative loop between ␤-sheet 4 and ␣-helix 4 of domain II (38). Based on the crystal structures of enzymes with the GT-B fold (34,49), that loop has been observed to be in close contact with the monosaccharide donor (Fig. 5).
The catalytic mechanism of retaining glycosyltransferases has been a challenging problem in enzymology. The strongest mechanistic candidate probably involves a late oxonium-like transition state (34,50). An important role in the catalytic mechanism of the maltodextrin phosphorylase has been given to Arg 534 and Lys 539 . The former stabilizes the substrate phosphate and the latter positions the phosphate group of the cofactor pyridoxal 5Ј-phosphate (34,48,50). According to crystallographic data, Glu 637 partially neutralizes the charge of Lys 539 . It has been postulated that Glu 637 may participate in catalysis through a dual role. First, it would lock the substrate in a more favorable position in the transition state through interaction with 3Ј-OH of the glucosyl pyranose. Second, it would be involved in the charge network with Lys 539 to keep a balanced protonation of the phosphates (48). There is a striking similarity with the structure of the retaining glycosyltransferase trehalose-6-phosphate synthase. It also has two positively charged residues, Arg 263 and Lys 268 , which interact with the ␣and ␤-phosphates of UDP-glucose. The complete catalytic center is almost identical to that of the maltodextrin phosphorylase and those interactions are in a very similar arrangement (34). The difference is that rather than Glu 637 , trehalose-6-phosphate has Asp 362 . Because the glycogen synthase is predicted to have the same fold, it is tempting to compare it with maltodextrin phosphorylase and trehalose-6phosphate synthase. According to a homology model, Glu 377 is interacting with Lys 305 , in a neighbor loop that also has Arg 300 (Fig. 6). The arrangement of these three residues is remarkably similar to both the maltodextrin phosphorylase and the trehalose-6-phosphate synthase. If the catalytic mechanism and the role of Glu 377 were the same as the maltodextrin phosphorylase, Arg 300 and Lys 305 should also be important residues for catalysis. These residues are conserved in not only all bacterial glycogen synthases, but also in members of families 3-5 (38). Experiments in the E. coli glycogen synthase showed that mutations to Ala yielded enzymes at least 3 orders of magnitude less active. 6 To reveal whether Glu 377 works interacting with Lys 305 further experiments and crystallographic data will be needed.
Because of the numerous and very important reactions that glycosyltransferases catalyze, members of this superfamily of enzymes are very good candidates for protein design. In this regard, determination of the essential features, defined as a structural rule for the class of proteins that perform a particular function by means of a particular fold, are particularly relevant (51). Of all the conserved residues of this putative loop, Glu 377 seems to be the only one that plays an essential function.