Module 4.7.4

Glycogen: Biosynthesis and Regulation

JACK PREISS

[SECTION EDITOR: LYNN SILVER]

Posted September 26, 2009

Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, MI 48824

Mailing address: Department of Biochemistry and Molecular Biology, Michigan State University, East Lansing, MI 48824. Phone: (517) 353–3137, Fax: (517) 353–9334, E-mail: This e-mail address is being protected from spambots. You need JavaScript enabled to view it .

The accumulation of glycogen occurs in Escherichia coli and Salmonella enterica serovar Typhimurium as well as in many other bacteria. Glycogen will be formed when there is an excess of carbon under conditions in which growth is limited due to the lack of a growth nutrient, e.g., a nitrogen source (reviewed in references 88, 89, 91, and 93). Glycogen is a homopolysaccharide of glucosyl residues consisting of about 95% α-1,4 linkages and about 5% α-1,6-branch linkages. The average length of the chains is usually about 8 to 12 glucose units. Figure 1 shows the linkages present in two oligosaccharide branches in the glycogen molecule. The molecular size of bacterial glycogen has been estimated to be about 10⁷ to 10⁸ Daltons.

Fig. 1Structure of a branched oligosaccharide showing the α-1,4 and α-1,6 glucosyl linkages seen in glycogen. Also shown are the nonreducing ends, which are in violet, with the branched glucosyl residue in the α-1,6 linkage in aqua. R represents the rest of the glycogen molecule.

The precise function of glycogen in bacteria is unknown. Published results suggest that it may play a role in prolonging the viability of microorganisms in stationary phase by providing a source of carbon and energy (reviewed in references 89, 93, and 95). It may also be utilized as source of energy for sporogenesis in bacilli (117) and in clostridia (119).

ENZYMES INVOLVED IN THE SYNTHESIS OF GLYCOGEN

The reactions leading to glycogen synthesis in bacteria were first described in 1964 (reviewed in references 10, 88, 91, 93, and 95). ADP-glucose was shown to be the glucosyl donor for bacterial glycogen synthesis (41) and is synthesized in a reaction catalyzed by ADP-glucose pyrophosphorylase (ADP-Glc PPase; EC 2.7.7.27), as follows:

ATP + α-glucose-1-P ↔ ADP-glucose + PP_i (1)

The glucosyl unit of ADP-glucose is then transferred, in a reaction catalyzed by an ADP-glucose-specific glycogen synthase (EC 2.4.1.21; reaction 2), to either a maltodextrin or glycogen primer to form a new α-l,4-glucosidic linkage.

ADP-glucose + α-glucan → α-l,4-glucosyl-glucan + ADP (2)

After chain elongation by glycogen synthase catalysis, branching enzyme (BE; EC 2.4.118) catalyzes the formation of branched oligosaccharide chains having α-l,6-glucosidic linkages as seen in glycogen (reaction 3):

Linear α-l, 4-polyglucosyl chain → branched α-l,4-α-l,6-glucan → (i.e., glycogen) (3)

The BE reaction occurs in two phases. First, the BE catalyzes the cleavage of an oligosaccharide, anywhere from 6 to 9 glucosyl units long, from a nonreducing end of a glycogen molecule. It then transfers the cleaved oligosaccharide to a C-6 hydroxyl group of a glucose residue in the interior part of the glycogen molecule or to another glycogen molecule.

These enzyme-catalyzed reactions have been observed in extracts from over 50 bacterial species (10, 88, 89, 91, 96).

Genetic evidence that bacterial glycogen synthesis occurs solely by the ADP-glucose pathway has been obtained with mutants of E. coli and Salmonella serovar Typhimurium devoid of or deficient in glycogen (10, 23, 39, 40, 56, 88, 89, 91, 93, 95, 100, 101). Compared to the wild-type strains, these mutants are defective in either ADP-Glc PPase or glycogen synthase activity or both. Moreover, mutants containing glycogen in amounts in excess of those observed in wild-type strains have also been isolated. These mutants overexpress ADP-Glc PPase, glycogen synthase, and BE activities. Thus, at least for the bacteria E. coli and Salmonella serovar Typhimurium, the data strongly indicate the importance of the ADP-glucose pathway for the synthesis of bacterial glycogen in vivo.

Information has been obtained with respect to genetic regulation of the expression of the biosynthetic enzymes and will be reviewed herein. The structural genes of the glycogen biosynthetic enzymes of E. coli and Salmonella serovar Typhimurium (72, 84) have been cloned previously, and that has provided insights in the genetic regulation of glycogen synthesis. It also has permitted oligonucleotide-directed mutagenesis of the E. coli glgC gene, the structural gene encoding ADP-Glc PPase, of the glgA gene, the structural gene encoding glycogen synthase, and of the glgB gene, the structural gene encoding BE. These mutagenic studies have revealed various aspects of structure-function relationships concerning the ADP-Glc PPase, glycogen synthase, and BE catalytic activities. An important aspect of the regulation of glycogen synthesis is the allosteric regulation of the ADP-Glc PPase. The current information, views, and concepts regarding the regulation of enzyme activity and the expression of the glycogen biosynthetic enzymes are thus presented herein. The recent information on the amino acid residues critical for the activity of both glycogen synthase and BE is also presented.

ADP-GLC PPase SUBUNIT AND QUARTENARY STRUCTURE

ADP-Glc PPase has been purified from both E. coli (31, 47) and Salmonella serovar Typhimurium (68, 118), and as seen for almost all bacteria, it is homotetrameric in structure, with a subunit molecular size of about 50 kDa (10, 89). Exceptions to this pattern are observed in the ADP-Glc PPases of Bacillus stearothermophilus and B. subtilis (123). These bacteria contain two genes, glgC and glgD, encoding two ADP-Glc PPases homologous in amino acid sequences to other prokaryotic ADP-Glc PPases. The GlgC protein from B. stearothermophilus has 387 amino acids, with a predicted molecular mass of 43.3 kDa, and shows 42 to 70% identity to other bacterial ADP-Glc PPases. The GlgD product (343 amino acids, with a predicted molecular mass of 38.9 kDa) has a lower degree of homology (20 to 30% identity) (123). B. stearothermophilus glgC expression renders an active recombinant enzyme, but GlgD exhibits negligible activity. When glgC and glgD genes are expressed together, however, the resulting GlgCD protein exhibits higher affinity for substrates and a twofold-higher V_max in catalyzing ADP-Glc synthesis than GlgC alone.

ALLOSTERIC REGULATION OF ADP-GLC PPase

An important regulatory aspect of bacterial glycogen synthesis is exerted through the allosteric modulation of ADP-Glc PPase activity (10, 56, 88, 89, 91, 93, 95, 96, 100, 101).

Over 50 ADP-Glc PPases, mainly from bacteria (10, 89, 93, 95, 100, 101) but also from plants (11, 90, 92, 99, 100), have been studied with respect to regulatory properties. In almost all cases, glycolytic intermediates activate ADP-Glc synthesis while AMP, ADP, and/or P_i are inhibitors. Glycolytic intermediates in the cell can be considered to be signals for carbon excess. Therefore, under conditions of limited growth with excess carbon in the medium, the accumulation of glycolytic intermediates may occur and would be signals for the activation of ADP-Glc synthesis. For many of the ADP-Glc PPases studied, the glycolytic intermediate activator increases the enzyme's apparent affinity for the substrates ATP and glucose-l-P. Also, increasing concentrations of the activator reverse the inhibition caused by the inhibitor AMP, ADP, or P_i.

The bacterial and plant ADP-Glc PPases can be catalogued into nine groups on the basis of their specificity of activation as well as the mode of activation by the various glycolytic intermediates (10, 11, 56, 88, 89, 90, 91, 92, 93, 99). The variation of activator specificity observed has been postulated to correlate with the nature of carbon assimilation dominant in the bacterium or plant tissue. This issue has been discussed in detail in a number of reviews (56, 88, 89, 90, 91, 92, 93, 95). E. coli or Salmonella serovar Typhimurium obtains its energy mainly through glycolysis. The primary activator for the ADP-Glc PPases of these bacteria is fructose l,6-bis-P, and 5′-adenylate is the major inhibitor. Thus, their ADP-Glc synthetic activity is regulated by the [fructose l,6-bis-P]/[AMP] ratio in the cell.

Evidence has accumulated to strongly suggest that the kinetic allosteric activation and inhibitor effects observed in vitro also occur in vivo in bacterial cells. There is a class of mutants of E. coli and of Salmonella serovar Typhimurium LT2 affected in their abilities to accumulate glycogen. Members of this mutant class have ADP-Glc PPases with altered regulatory properties. Generally, mutants that have ADP-Glc PPases with higher affinities for the activator fructose l, 6-bisphosphate and/or lower affinities for the allosteric inhibitor AMP accumulate glycogen at a higher rate than the wild-type parent strain. Mutants that have enzymes with lower affinities for the activator accumulate glycogen at a lower rate than the parent strain.

Table 1 summarizes the allosteric properties of the mutant ADP-Glc PPases that have been studied and their abilities to accumulate glycogen in the stationary phase. With respect to E. coli B or K-12, there is a direct relationship between the affinity of the enzyme for the activator and the ability of the mutant to accumulate glycogen. If the apparent affinity for the activator fructose 1,6-bis-P is higher than that of the wild-type enzyme, the level of glycogen accumulation by the mutant is higher. If the apparent affinity for the activator is lower, as seen for mutant SG14 ADP-Glc PPase (97), the level of glycogen accumulation in the mutant is lower than that in the parent strain. The two Salmonella serovar Typhimurium mutant strains, JP23 and JP51, have ADP-Glc PPases that are more affected in the affinity for the inhibitor than in that for the activator (118). Both JP23 and JP51 mutant enzymes have lower affinity for the inhibitor than the wild-type enzyme, and these mutants accumulate larger amounts of glycogen than the parent strain (118). An E. coli K-12 allosteric ADP-Glc PPase mutant has also been isolated and characterized previously (23, 66, 73). As seen in Table 1, it accumulates two times more glycogen than its parent strain, having an ADP-Glc PPase with 4-fold-higher affinity for its activator, fructose 1,6-bisphosphate, and 13-fold-lower affinity for the inhibitor AMP.

The findings of the above-described studies and others showing a direct relationship between the fructose l, 6-bis-P concentration in the E. coli cell and the rate of glycogen accumulation (26, 27) clearly point out that fructose l,6-bis-P is an allosteric activator of ADP-Glc PPase and a physiological activator of glycogen synthesis in E. coli and in Salmonella serovar Typhimurium.

STUDIES INDICATING THE CRITICAL AMINO ACIDS INVOLVED IN ACTIVATOR/INHIBITOR BINDING

Chemical modification and site-directed mutagenesis studies of the E. coli ADP-Glc PPase have provided evidence for the location of the activator binding site (86, 87), the inhibitor binding site (61, 62), and the substrate binding sites (15, 47, 61, 62, 87). These experiments have used pyridoxal P as an analog either for the activator fructose 1,6-bis-P (86, 87) or, as subsequently demonstrated, for the substrate glucose-1-P (15, 67, 87). For an ATP analog, the photoaffinity reagent 8-azido-ATP proved to be a substrate for the E. coli enzyme (67), whereas 8-azido-AMP is an effective inhibitor analog (61, 62). Since the E. coli ADP-Glc PPase gene, glgC, has been cloned previously and its sequence has been determined (6, 84), the identification of the amino acid sequence around the modified residue enables researchers to determine the location of the modified residue in the primary structure of the enzyme. The amino acid residue involved in binding the activator is Lys39 (86), and the amino acid involved in binding the adenine portion of the substrates (ADP-Glc and ATP) is Tyr114 (67). Tyr114 is also the major binding site for the adenine ring of the inhibitor AMP (61, 62).

Results from experiments with the chimeric enzymes also indicate that the C terminus of the E. coli enzyme contributes to determining the selectivity/specificity of the activator fructose 1,6-bisphosphate. As indicated above, the E. coli ADP-glucose PPase is activated mainly by fructose 1,6-bisphosphate (98). The Agrobacterium tumefaciens enzyme is activated by fructose 6-phosphate and pyruvate (30). To determine the functions of different domains in activator binding, two chimeric enzymes were constructed (12). Chimeric AE contained the N terminus (271 amino acids) of the A. tumefaciens ADP-glucose PPase and the C terminus (153 residues) of the E. coli enzyme, and EA comprised the inverse construction (12). The characterization of the purified chimeras showed that the C terminus of the E. coli enzyme is relevant for the selectivity of fructose 1,6-bisphosphate. Chimeric AE is activated by both fructose 1,6-bisphosphate and fructose 6-phosphate, neither of which affects EA. Pyruvate activates EA with higher affinity than AE, suggesting that the C terminus of the A. tumefaciens enzyme plays a role in the binding of this effector.

Since it has been found previously that Lys39 in the E. coli enzyme interacts with the allosteric activator (36, 86, 87), it is very possible that the regulation is determined by a combined arrangement between the N- and C-terminal regions.

The findings from other experiments have indicated the N-terminal region of the E. coli ADP-glucose to be very important for allosteric function (14, 128, 129). Proteolysis of the E. coli enzyme with proteinase K inactivates the enzyme and generates two polypeptides of 21 and 28 kDa from the 49.7-kDa native subunit. In the presence of the substrates ADP-glucose and Mg²⁺ and the activator fructose 1,6-bis-P, proteinase K removes only 10 to 13 amino acids from the N terminus and 2 amino acids from the C terminus. When purified, the truncated enzyme is fully active and no longer requires an activator for maximal activity (128). When the abridged form of the recombinant E. coli ADP-Glc PPase lacking the 11 amino acids of the N terminus is expressed, the same results are obtained (129). Also shown was that removing only the two amino acids at the C terminus has no effect on the allosteric activation (129). The N-terminally truncated enzyme is also insensitive to AMP inhibition (128, 129). The results of further studies showed that the truncation of the N terminus by 3 or 7 amino acids has no effect on the activation by fructose 1,6-bis-P (14). The deletion of 11 or 15 amino acids via recombinant methods does produce an enzyme no longer requiring an activator for full activity (14). However, the truncation of 19 or 22 amino acids of the N terminus produces a virtually inactive enzyme (14). Thus, the N-terminal portion of the enzyme may play a role as an allosteric “switch” to regulate enzyme activity.

With regard to the involvement of the N and C termini being important for allosteric activation, computational analysis predicts that the E. coli ADP-Glc PPase has a central core including the substrate and catalytic residues and that the fold of the enzyme has two domains (13, 31). This prediction is supported by the results of pentapeptide (15-bp) scanning mutagenesis of the E. coli ADP-glucose PPase, generating an enzyme with the pentapeptide inserted between residues Asp321 and Arg322 (13). The corresponding 15-bp insertion in the gene results in a stop codon, producing a mutant enzyme containing a peptide of residues 1 to 323 and residues 328 to 431. Residue 328 is a methionine residue, and expression probably results in the synthesis of the remaining 12-kDa polypeptide encompassing the ADP-Glc PPase C-terminal region. Purification of the insertion mutant ADP-Glc PPase results in an enzyme having 37- and 12-kDa polypeptides (13). The pure enzyme is as active as the wild-type enzyme but has 5.7-fold-lower affinity for the activator fructose 1,6-bisphosphate. All other kinetic constants are similar to those of the wild-type enzyme. These findings supported the hypothesis that the E. coli ADP-Glc PPase is organized into at least two distinct domains and that in the insertion mutant gene, the stop codon is inserted at a site corresponding to the peptide loop that separates the two putative domains (13). This view is also supported by expressing the two polypeptides comprising residues 1 to 323 and 328 to 431 of the E. coli ADP-Glc PPase and showing that the enzyme formed is as active as the wild-type enzyme (13). This active enzyme was purified, and it was shown that both peptides were still present and could not be separated by chromatography. The enzyme made up of the two peptides is as active as the wild-type enzyme in the pyrophosphorolysis direction and has 80% of the activity of the wild-type enzyme in the synthesis direction (13). The two-peptide enzyme, however, has 4.7-fold-lower affinity for ATP, 3.6-fold-lower affinity for the activator, fructose1,6-bis-P, and 2.6-fold-lower affinity for the inhibitor, AMP.

Pentapeptide-scanning mutagenesis has also produced another interesting mutant protein (9). The insertion of the 15-bp sequence into a plasmid with the E. coli ADP-Glc PPase gene produced a mutant enzyme with the pentapeptide insertion between Leu102 and Pro103. This mutant enzyme cannot be activated by fructose 1,6-bis-P, and moreover, fructose 1,6-bis-P cannot increase the affinity for ATP (9). In the model of the structure around this insertion, the site is at the end of a β-sheet and just before the loop of Gln105 to Gly116 that is proposed to interact with ATP (9). The results of analysis of the insertion mutant enzyme suggest that Leu102 is critical for allosteric activation and that the Gln105-Gly116 loop is important for ATP binding. As indicated previously and further detailed below, Tyr114 is involved in the binding of ATP, and the residue at position 114 also affects the apparent affinity for the activator, fructose 1,6-bis-P, when the amino acid at this site is chemically modified or mutated.

In summary, although Lys39 has been shown to be the residue binding the activator, other regions of the enzyme participate in the allosteric activation. First, as will be discussed below, point mutations at different areas of the enzyme lead to enzymes with different affinities for the inhibitor and activator. The consequences of modifying the C terminus by forming a hybrid enzyme from E. coli and A. tumefaciens enzymes with different activator specificities indicate that the C-terminal region participates with the N terminus in the allosteric activation. Also, fructose 1,6-phosphate activation causes an increase in affinity for ATP binding, and it is believed that the region near Leu102 and the Gln105-Gly116 loop is critical for the allosteric regulation. In addition, the N-terminal portion (residues 1 through 15) is required for the allosteric function of the enzyme and plays a role as an allosteric switch to regulate enzyme activity.

AMINO ACIDS INVOLVED IN BINDING OF THE SUBSTRATES

Glucose-1-Phosphate

Site-directed mutagenesis at position 195 of the E. coli ADP-Glc PPase to change lysine to either glutamine, arginine, histidine, isoleucine, or glutamate provides enzymes whose K_m values for glucose-1-phophate are 100- to 10,000-fold greater than that of the wild type (47). Kinetic constants for other substrates, ATP and Mg²⁺, and the activator, fructose 1,6-bisphosphate are similar to those of the wild-type enzyme, indicating that Lys195 is involved in the binding solely of the substrate glucose-1-phosphate. Moreover, the catalytic rate constant, k_cat, for the glutamine mutant enzyme is similar to that of the wild type, ruling out the role of this residue in the catalytic reaction (47). It was proposed that Lys195 interacts with the phosphate residue of glucose-1-P.

The equivalent amino acid in plant ADP-Glc PPase was also found to be important for glucose-1-P binding. Site-directed mutagenesis was done at the site of the equivalent and conserved residue, Lys198, in the catalytic subunit of potato tuber ADP-Glc PPase (32). The mutation of Lys198 to Arg, Ala, or Glu had little or no effect on the kinetic constants for ATP, Mg²⁺, the activator, 3-phosphoglycerate (3-PGA), and the inhibitor (P_i), but the apparent affinity for Glc-1−P decreased 135- to 550-fold.

Recently, a homology model of the three-dimensional structure of the E. coli enzyme in a complex with ADP-Glc was generated to examine the substrate binding site in greater 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 substrate (15). The amino acids identified were Glu194, Ser212, Tyr216, Asp239, Phe240, Trp274, and Asp276, and they were mutated to alanine by site-directed mutagenesis. The kinetic properties of the Ala mutant ADP-Glc PPases were determined, and all the purified alanine mutant enzymes had apparent affinities for glucose-1-phosphate that were 1 or 2 orders of magnitude lower than that of the wild type, indicating that these amino acids play an important role in the interaction with the substrate (15). Those amino acids are conserved within the ADP-Glc PPase family and were replaced by other amino acids to investigate the effect of size, hydrophobicity, polarity, aromaticity, or charge on the apparent affinity for glucose-1-P, giving us a full design and characterization of the glucose-1-phosphate binding site.

Table 2 shows the effects of the mutations of the amino acids in the model. As described above, the largest effect is seen with Lys195. Large effects are also seen with mutations of Glu194 and Ser212, indicating that the hydrogen bonds proposed in the model (15) are valid. The apparent affinities for the cosubstrates ATP and Mg²⁺ and the activator fructose 1,6-bisphosphate are not appreciably affected (15). Notable decreases in apparent affinity for glucose-1-P or in k_cat values are also seen with mutations of amino acids Tyr216, Asp239, Phe240, Trp274, and Asp276.

In summary, the study of the selected set of amino acid residues of the model via mutagenesis and kinetic characterization of the mutant enzymes reveals the importance of those amino acids in shaping the glucose-1-phosphate substrate pocket, with Glu194, Ser212, and Asp276 interacting with the oxygen molecules of the sugar ring by hydrogen binding (15). Of particular significance is the observation that the Glc-1-P site in the model overlaps not only with the Glc-1-P sites of other bacterial and higher-order plant ADP-Glc PPases but also with those of other bacterial sugar-nucleotide PPases, such as the Pseudomonas aeruginosa dTDP-glucose PPase (18) and the Salmonella serovar Typhi CDP-glucose PPase (54), whose crystal structures have been determined.

ATP

For an ATP analog, the photoaffinity reagent 8-azido-ATP proved to be a substrate for the E. coli enzyme (67), and thus, 8-azido-ADP-Glc could be synthesized. Upon UV irradiation, the E. coli ADP-Glc PPase is inactivated in the presence of both photoaffinity reagents. As noted above, the amino acid involved in binding the adenine portion of the substrates (ADP-Glc and ATP) is Tyr114 (67). Tyr114 is also the major binding site for the adenine ring of the inhibitor, AMP. Chemical modification with 8-azido-AMP, an effective inhibitor analog (61, 62), and site-directed mutagenesis studies of the E. coli ADP-Glc PPase have provided good suggestive evidence for the location of the inhibitor binding site (61, 62).

Figure 2 shows the amino acid sequence of the E. coli ADP-Glc PPase and the substrate and allosteric sites identified via chemical modification. Figure 2 also shows the deduced amino acid sequence of the Salmonella serovar Typhimurium enzyme (71, 72). There is about 80% identity between the nucleotide sequences of the E. coli and Salmonella serovar Typhimurium glgC genes and 90% identity between the corresponding amino acid sequences. Most of the changes are conservative. However, the amino acids that have been shown to be involved in substrate and allosteric effector binding and catalysis, as well as those involved in maintaining allosteric function for the E. coli enzyme, are all conserved. Thus, the information obtained with the E. coli enzyme most certainly applies to the Salmonella serovar Typhimurium enzyme.

Fig. 2Comparison of amino acid sequences of ADP-Glc PPases of E. coli (EC; first row) and Salmonella serovar typhimurium (ST; second row). Amino acids are represented by the one-letter amino acid code. Only those amino acids that are different in the Salmonella serovar typhimurium enzyme are shown. Various amino acids in the E. coli sequence are underlined and have been shown to be involved in either substrate or allosteric effector binding or are replaced by other amino acids in allosteric mutant E. coli enzymes. K39 is involved in the binding of the activator fructose 1,6-bis-P, A44 is replaced by Thr in the allosteric mutant SG14 enzyme (82), R67 is replaced by Cys in the allosteric mutant CL1136 enzyme (37), and Y114 has been shown to be involved in the binding of the substrate ATP (67) and the inhibitor AMP (61,62). D142 has been identified as being involved in catalysis (31). K195 has been shown to be involved in the binding of the substrate glucose-1-P (47), P295 is replaced by Ser in the allosteric mutant SG5 enzyme (81), and G336 is replaced by Asp in the allosteric mutant 618 enzyme (58, 66, 73). All these amino acids are conserved in the Salmonella serovar Typhimurium ADP-Glc PPase.

DETERMINATION OF THE ADP-GLC PPase CATALYTIC RESIDUE

The residue involved in catalysis in the E. coli ADP-Glc PPase was determined by comparing a predicted structure of the enzyme with the known three-dimensional structures of sugar-nucleotide PPase domains (31). This comparison allowed the identification of highly conserved residues throughout the sequences of the PPase superfamily enzymes in spite of their low levels of homology. In E. coli ADP-GlcPPase, Asp142 was predicted to be close to the substrate site, and site-directed mutagenesis replacing Asp142 with Ala and Asn indicated that the major role for Asp142 is catalytic (31). Kinetic analysis showed a decrease in the specific activity of 4 orders of magnitude compared to that of the wild-type enzyme, whereas other kinetic parameters, the K_m values for the substrates and the affinities for the activator fructose 1,6-bisphosphate and the inhibitor AMP, showed no significant changes (Table 3). A conservative substitution of Glu for Asp resulted in only a 100-fold decrease in the k_cat, indicating that the negative charge of the residue is important for the catalytic reaction. The Glu142 mutant enzyme is also altered in having 11.5- and 47-fold-lower affinities for ATP and glucose-1-phosphate, respectively, than the wild-type enzyme. Moreover, the concentration of the activator required for 50% maximal activation of the mutant enzyme is 17-fold higher than that of the wild-type enzyme. Due to the lower affinity for the substrates, the D142E mutant enzyme would not be functional under in vivo conditions.

It should be pointed out that the E. coli UDP-glucosamine PPase/N-acetyltransferase Arg18 was proposed previously to be the catalytic residue (20). This residue, although conserved in the ADP-GlcPPases, may not be directly involved in the catalytic reaction, as the mutation of the homologous Arg25 in the enzyme from A. tumefaciens yields an enzyme with a reduction in activity of only 2 orders of magnitude (38). It is expected that a more dramatic effect, a reduction closer to 4 orders of magnitude, would occur if the mutation affected a catalytic residue. Thus, in A. tumefaciens the ADP-Glc PPase catalytic residue equivalent to the E. coli ADP-Glc PPase Asp142 would be Asp135. In the Salmonella serovar Typhimurium enzyme, Asp142 is conserved and most probably also acts as the catalytic residue, as in the E. coli enzyme.

CLONING OF ADP-GLC PPases WITH ALTERED ALLOSTERIC PROPERTIES FROM E. COLI MUTANTS AFFECTED IN GLYCOGEN SYNTHESIS

As indicated in Table 1, a class of mutants of E. coli and of Salmonella serovar Typhimurium with altered rates of glycogen accumulation were found to have ADP-Glc PPases that were affected in their allosteric properties. To gain insight with respect to amino acid residues or domains involved in maintaining allosteric function, the allosteric mutant ADP-Glc PPases were cloned (37, 72, 81, 82).

Table 4 shows the various amino acid substitutions in the allosteric mutant enzymes that have been cloned and analyzed. Of interest is that the mutations causing large changes in the allosteric properties of the enzyme occur throughout the sequence of the ADP-Glc PPase. These changes of amino acids in the enzyme affect not only the affinity for the allosteric effectors (Table 1) but also the apparent affinities for the substrates ATP and Mg²⁺ (36, 37, 73, 81, 82, 94, 97). The mutations are spread throughout the ADP-Glc PPase sequence, suggesting once again that a number of domains in the ADP-Glc PPase interact in the allosteric function.

SIMILARITY OF THE ADP-GLUCOSE PPase STRUCTURE TO THOSE OF OTHER SUGAR-NUCLEOTIDE PPases

Site-directed mutagenesis in the ADP-Glc PPase gene and analysis of various allosteric mutant genes have provided much information about the structure-function relationships of the substrate and catalytic sites. What is needed for greater clarification is knowledge of the three-dimensional structure of the enzyme.

Information on the three-dimensional structure of an ADP-Glc PPase has been obtained by using the crystallization of the potato tuber enzyme (51). The E. coli enzyme has been crystallized previously (83), but the crystals are of poor diffraction quality and are sensitive to X-ray exposure damage. Nevertheless, a variety of methods have been used to predict the ADP-Glc PPase structure (31, 111). The hydrophobic cluster analyses (69) of ADP-Glc PPases from bacterial and plant sources and from the different classes suggest that ADP-Glc PPases of bacteria and plants are extremely similar in both the distribution and the pattern of the clusters. Thus, it is very likely that the ADP-Glc PPases have a common folding pattern, despite the different quaternary structures seen for α₂β₂ in plants and α₄ in bacteria and the different specificities of the nine groups of enzymes for the activator.

The alignment of representative bacterial ADP-Glc PPases from each class has been done, and from these analyses, a general structure that fits all of the ADP-Glc PPases was postulated (31). Biochemical data that supported the model were also presented (31).

A number of sugar-nucleotide PPases have been crystallized, and their structures have been determined. The first resolved PPase structure was that of the expression product of the gene glmU from E. coli (20). The enzyme is bifunctional. One domain of the GlmU protein is a UDP−N−acetylglucosamine PPase, and another is an acetyltransferase. Others have also crystallized the GlmU PPase (55, 85), and another PPase structure to be resolved was that of TDP-glucose PPase (18, 116). A residue equivalent to Lys195, the Glc-1−P binding site in ADP-Glc PPases, is also present in the N-acetylglucosamine uridyltransferase but is shifted one position.

All these structures verify the predicted secondary structure model of the ADP-Glc PPase (31). That sugar-nucleotide PPases have similar catalytic domains is supported by the observation that the homologous glucose-1-phosphate site is also present in the GDP-mannose PPase from Pseudomonas aeruginosa (79). The GDP-mannose PPase is specific for mannose-1-phosphate and not glucose-1-phosphate. Thus, another portion or domain of the sequence in the PPases must also be important for the sugar-1-phosphate specificity. This scenario would also be true for the N-acetylglucosamine-1-phosphate substrate site of the E. coli UDP−N−acetylglucosamine PPase.

GLYCOGEN SYNTHASE

The E. coli and Salmonella serovar Typhimurium glycogen synthases are specific for the sugar nucleotide ADP-glucose. An affinity analog of ADP-Glc, adenosine diphosphopyridoxal (ADP-pyridoxal), was used previously to identify the ADP-Glc binding site (33). The incubation of the enzyme with the analog plus sodium borohydride led to an inactivated enzyme. The degree of inactivation correlated with the incorporation of about 1 mol of analog per mol of enzyme subunit required for 100% inactivation. After tryptic hydrolysis, one labeled peptide was isolated and the modified Lys residue was identified as Lys15 (33). The sequence Lys-X-Gly-Gly, where lysine is the amino acid modified by ADP-pyridoxal, has been found to be conserved in the mammalian glycogen synthase (78, 121) and the plant starch synthases (115).

The structural gene for glycogen synthase, glgA, has been cloned from both E. coli and Salmonella serovar Typhimurium (72, 84), and the nucleotide sequence of the E. coli glgA gene has been determined (57). It consists of 1,431 bp specifying a protein of 477 amino acids with a molecular weight (MW) of 52,412. The Salmonella serovar Typhimurium glycogen synthase gene nucleotide sequence has been determined up to 135 bp, starting at the point corresponding to the N-terminal amino acid, which is methionine (P. Leung and J. Preiss, unpublished results). The deduced 45-amino-acid sequence is the same as that of the E. coli glycogen synthase except at residues 37 and 43. In E. coli, residue 37 is alanine, while in Salmonella serovar Typhimurium, it is valine. E. coli residue 43 is also Ala, and in Salmonella serovar Typhimurium, it is glycine.

The availability of the E. coli glycogen synthase gene has enabled Furukawa et al. (33, 34) to perform site-directed mutagenesis experiments to determine structure-function relationships for a number of amino acids in the E. coli glycogen synthase. The results of substituting other amino acids for Lys at residue 15 suggest that the Lys residue is involved mainly in the binding of the phosphate residue adjacent to the glycosidic linkage of the ADP-Glc and not in catalysis. The major effect on the kinetics of the mutant enzymes with changes at residue 15 is the elevation of the K_m for ADP-Glc, by about 30- to 50-fold, when either Gln or Glu is the substituting amino acid. The substitution of Ala for Gly at residue 17 decreases the k_cat about 3 orders of magnitude compared to that of the wild-type enzyme. The substitution of Ala for Gly at residue 18 decreases the rate constant only 3.2-fold. The effects on the K_m values for the substrates glycogen and ADP-Glc are minimal. It was postulated previously (33) that the two glycyl residues in the conserved Lys-X-Gly-Gly sequence participate in the catalysis by assisting in maintaining the correct conformational change of the active site or by stabilizing the transition state. The Salmonella serovar Typhimurium glycogen synthase also has the Lys-Thr-Gly-Gly sequence, starting at residue 15 (Leung and Preiss, unpublished data).

Since the Lys15Gln mutant enzyme still binds the ADP-Glc and shows appreciable catalytic activity, the ADP-pyridoxal modification was repeated, and in this instance, a concentration about 30 times higher was needed for inactivation of the enzyme (35). The enzyme was maximally inhibited by about 80%, and tryptic analysis of the modified enzyme yielded one peptide containing the affinity analog and the sequence Ala-Glu-Asn-modified Lys-Arg. The modified Lys was identified as Lys277. Site-directed mutagenesis to replace Lys277 with Gln was done, and the K_m for ADP-Glc was essentially unchanged but the k_cat was decreased 140-fold. It was concluded that Lys residue 277 is more involved in the catalytic reaction than in substrate binding.

The findings of chemical modification studies (48) have shown that two distinct sulfhydryl groups are important for enzyme activity and that they are protected by the primer glycogen and the substrate ADP-Glc, respectively. The reactive sulfhydryl residue is located probably at or near the binding sites for the substrates glycogen and ADP-Glc. To determine the responsible residue, all cysteines present in the enzyme, Cys7, Cys379, and Cys408, were replaced with Ser. 5,5′-Dithiobis(2-nitrobenzoic acid) modifies and inactivates the enzyme only when Cys379 is present (133). The inactivation is prevented by the substrate ADP-Glc. Mutations C379S and C379A decrease the specific activity 5.8- and 4.3-fold, respectively, and increase the S_0.5 (substrate concentration required for 50% of V_max) for ADP-Glc 40- and 77-fold. The results of inhibition studies with glucose-1-phosphate and AMP suggested that Cys379 is involved in the interaction of the enzyme with the phosphoglucose moiety of ADP-Glc. Other mutations, C379T, C379D, and C379L, showed that the site is intolerant of bulkier side chains. Cys379 is in a conserved region. Thus, other residues were scanned by mutagenesis. The replacement of Glu377 by Ala and Gln decreases the V_max more than 10,000-fold but does not affect the apparent affinity for ADP-Glc and glycogen binding (Table 5). The mutation of Glu377 to Asp decreases the V_max by only 57-fold, showing that the carboxyl group of Glu377 is essential for catalysis. The E377C mutation, in an enzyme form having no other Cys residues, yields an inactive protein (Table 5). When the enzyme is carboxymethylated by iodoacetate, however, the glycogen synthase activity is restored (Table 5). This result also indicates that a carboxyl residue is necessary for activity. No appreciable effect on activity is observed when another conserved residue in the region, Ser374, is mutated. However, the mutation of another conserved amino acid, Gln383, to alanine causes a 23-fold decrease in the V_max but no decrease in the affinity for the substrates (133).

Glycogen synthases are glucosyl transferases that retain the anomeric configuration of the glucosyl linkage of ADP-glucose to the nonreducing glucosyl linkage of glycogen. The E. coli glycogen synthase was therefore modeled based on three other crystallized glycosyl linkage-retaining glycosyltransferases having a GT-B fold (131). Comparison of the model with the active-site structures of glycosyl linkage-retaining GT-B glycosyltransferases showed conserved residues in glycogen synthase with the same geometrical orientation. These conserved residues in the E. coli glycogen synthase were mutated to confirm the importance of these residues as predicted by the model. Mutations D137A, R300A, K305A, and H161A decreased the specific activity 8,100-, 2,600-, 1,200-, and 710-fold, respectively. These mutations did not increase the K_m for glycogen, and only H161A and R300A mutant enzymes had higher K_m values for ADP−glucose, by 11- and 8-fold, respectively (Table 5). The conserved residues were essential, validating the model that shows strong similarity between the active site of E. coli glycogen synthase and those of the other glycosyl linkage-retaining GT-B glycosyltransferases known to date (131, 132, 133).

Recently, the glycogen synthase crystal structures of A. tumefaciens and of E. coli have been elucidated (21, 114a) and have confirmed the correctness of our proposed model (131, 132, 133). The structures consist of two domains, one at the N terminus and one at the C terminus, and these domains interact with respect to the catalysis of the transglycosylation reaction. Of particular interest is the recently determined active closed crystal structure containing bound ADP-glucose and HEPPSO [N-(2-hydroxyethyl)piperazine-N′-(2-hydroxypropanesulfonic acid)] buffer (Sheng et al., submitted). The ADP-glucose in the structure has been split into ADP and glucose or a glucose-like structure. The carbonyl group of His161 and residues Arg300 and Lys305 are suggested by the structure to be critical for the transglycosylation reaction. Glu377 interacts with the 3-OH group of the glucose of ADP-glucose, and its important function is the proper positioning of the glucose moiety for the reaction (114a). Figure 3 presents a diagrammatic scheme of the proposed binding site for ADP-glucose in the glycogen synthase.

Fig. 3Schematic representation of the proposed ADP-Glc binding domain of the glycogen synthase (132). The amino acids interacting with the AMP portion of the sugar nucleotide are in red. Those binding with the glucose-1-P portion are in blue.

Thus, recent research efforts have extensively increased our knowledge with respect to the structure and function of the E. coli glycogen synthase. Indeed, the acquisition of a closed structure confirms and supports the postulated roles that various amino acid residues play with regard to substrate binding and enzyme catalysis. These amino acids are even highly conserved in the starch synthases that have the same glycosyl donor, ADP-glucose, and essentially the same glycosyl acceptors, amylopectin and amylose, component polysaccharides in the starch granule. Almost all residues present at the E. coli glycogen synthase active site have identical counterparts in the starch synthases. These residues are those in the K15TGGL motif, Glu377, Lys305, Arg300, His161, Tyr95, Asp137, and Trp138. The detailed study of the bacterial glycogen synthase has therefore contributed greatly to the study of the starch synthases. Figure 4 shows the alignments of the conserved residues in some of the bacterial glycogen synthases and starch synthases.

Fig. 4Alignment of residues found to be critical for activity in E. coli glycogen synthase and conserved in other bacterial glycogen synthases and plant starch synthases. Conserved critical residues are shown in bold lettering. The sequences were obtained online from NCBI Sequence Viewer version 2.0. C. reinhardtii, Chlamydomonas reinhardtii; SSIII, starch synthase III.

BE

The structural gene of the E. coli BE, glgB, has been cloned (7), and its open reading frame (ORF) is about 200 bp downstream from the asd gene. The complete nucleotide and corresponding deduced amino acid sequences were determined to be consistent with the amino acid analysis of the pure protein, the MW determined by sodium dodecyl sulfate gel electrophoresis, and the amino acid sequence analyses of the amino terminal and of the isolated cyanogen bromide-derived peptides (7). The gene consists of 2,181 bp specifying a protein of 727 amino acids with a MW of 84,231.

BE Belongs to the α-Amylase Family

The relationship among the amino acid sequences of BE and amylolytic enzymes such as α-amylase, pullulanase, glucosyltransferase, and cyclodextrin glucanotransferase, particularly in those regions believed to be sites of contact between the substrate and the enzyme, was first reported by Romeo et al. (107). Baba et al. (4) then showed that there was marked conservation in the amino acid sequences of the four catalytic regions of amylolytic enzymes and four corresponding regions of maize endosperm BE I. As seen in Table 6, four regions that putatively constitute the catalytic regions of the amylolytic enzymes are conserved in the starch branching isoenzymes of maize endosperms, rice seeds, and potato tubers and the glycogen BEs of E. coli. This high-level conservation in the α-amylase family has been pointed out and the analysis has been greatly expanded by Svensson (120) and by Jesperson et al. (50) with respect to sequence homology but also the prediction of the α¤β eight-barrel structural domains, which have a highly symmetrical fold of eight inner, parallel β-strands surrounded by eight helices, in the various groups of enzymes in the family. The α¤β eight-barrel structural domain was determined from the crystal structure of some α-amylases and cyclodextrin glucanotransferases.

The conservation of the putative catalytic sites of the α-amylase family in the glycogen and starch BEs would be expected, as the BE catalyzes two consecutive reactions in synthesizing α-1,6-glucosidic linkages by the cleavage of an α-1,4-glucosidic linkage in an 1,4-α-d-glucan to form a nonreducing-end oligosaccharide chain that is transferred to a C-6 hydroxyl group of the same or another 1,4-α-d-glucan.

The eight highly conserved amino acid residues of the α-amylase family are also functional in BE catalysis (74), and these are identified in Table 6. Also, the crystal structure of a truncated form of the E. coli BE at 2.3-Å resolution has been determined (1). The BE structure has a central α/β barrel that contains the presumed eight functional amino acids, and this domain is similar to the α/β barrels of the other members of the α-amylase family. The C- and N-terminal portions of BEs from various bacteria, as well as those from higher plants, are dissimilar in sequence and in size. The findings of studies by Kuriki et al. (59) show that the C- and N-terminal regions are functional with respect to BE specificity and substrate branching preference (amylose or amylopectin), as well as preferences for the size of the oligosaccharide chain transferred and the extent of branching. The truncation of 113 amino acids of the N terminus causes the E. coli BE to transfer longer branch chains than the wild-type enzyme, with amylose as the substrate (16, 17). This result indicates that the N-terminal region is involved in specifying the size of the chain transferred. Indeed, Fig. 5 shows the chain transfer patterns of the wild-type BE and BEs truncated by 63, 83, and 112 amino acids at the N terminus. The chain transfer patterns show that the greater the number of amino acids truncated from the N terminus, the more pronounced is the shift of the BE toward forming longer chains in the branched product (25).

Fig. 5High-performance anion-exchange chromatography analysis of debranched α-glucans formed by the actions of wild-type BE (gray), a 112-amino-acid-truncated BE (blue), an 83-amino-acid-truncated BE (red), and a 63-amino-acid-truncated BE (green) on reduced amylose AS-320 after 2 h (top panel) and 16 h (bottom panel) of branching (data are from reference 25). The deviation is calculated for the results from at least two experiments. The data given by high-performance anion-exchange chromatography were processed in the following way: the sum of the integration of peaks 6 to 40 from malto-oligosaccharide corresponds to 100%, and the integration of each peak was calculated as a percentage of that total. D.P is degree of polymerization; i.e., the malto-oligosaccharide size (e.g., maltose has a D.P. of 2, and maltopentaose has a D.P. of 5). Reproduced from reference 25 with permission.

GENETIC REGULATION OF GLYCOGEN SYNTHESIS IN E. COLI

The enzymes of the glycogen biosynthesis pathway are induced in the stationary phase. The rate of glycogen synthesis is inversely related to the growth rate when growth is limited by the availability of certain nutrients, e.g., nitrogen. Consistent with this relationship is the finding that the levels of glycogen biosynthetic enzymes in E. coli increase as cultures enter the stationary phase (56, 88, 89, 93, 100, 101). This is true for E. coli B as well as for E. coli K-12 strains and is also true for Salmonella serovar Typhimurium (118).

When cells of E. coli B are grown in an enriched medium containing yeast extract and 1% glucose, the specific activities of ADP-Glc PPase and glycogen synthase increase 11- to 12-fold, while glycogen BE increases fivefold as the cultures enter stationary phase (reviewed in references 56, 88, 89, and 95). However, when the organism is grown in a defined medium, the ADP-Glc PPase and glycogen synthase activities are elevated in the exponential phase. BE in defined medium is fully induced in the exponential phase, with only about a twofold increase in the specific activities of the ADP-Glc PPase and glycogen synthase when cells grown in a defined medium reach the stationary phase. The same phenomena are also seen with the glycogen biosynthetic enzyme levels in Salmonella serovar Typhimurium (118).

The results of these experiments suggest that the gene encoding the BE is regulated differently from the genes for ADP-Glc PPase and glycogen synthase. As would be expected for a pathway that is under transcriptional control, the addition of inhibitors of RNA or protein synthesis to pre-stationary-phase cultures prevents the enhancement of glycogen synthesis in the stationary phase (22).

LOCATION OF THE STRUCTURAL GENES FOR GLYCOGEN BIOSYNTHESIS AND CATABOLISM

The structural genes for glycogen biosynthesis are clustered in two adjacent operons, which also contain genes for glycogen catabolism. The structural genes for glycogen synthesis were shown to be located at approximately 75 min on the E. coli K-12 chromosome, and the gene order at this location was subsequently established by transduction to be glgA-glgC-glgB-asd (63). glgA, glgC, and glgB encode the enzymes glycogen synthase, ADP-Glc PPase, and glycogen BE, respectively, and are close to asd, the structural gene for the enzyme aspartate semialdehyde dehydrogenase (EC 1.2.1.11).

The molecular cloning of the E. coli glg K-12 structural genes (84) greatly facilitated the subsequent study of the genetic regulation of bacterial glycogen biosynthesis. The glg genes are dispensable for growth and are not amenable to direct selection. They were cloned into pBR322 via selection with the closely linked essential gene asd. Among several asd⁺ plasmid clones that were isolated, pOP12 was found to contain a 10.5-kb PstI fragment which included the structural genes glgC, glgA, and glgB. A generally applicable method for cloning α-1,4-glucan biosynthetic genes based upon the screening of clones with iodine vapor was developed (39, 109). This approach does not require coselection via an essential gene and in principle should allow the direct cloning of structural glg genes from any bacterium into E. coli, as well as the cloning of regulatory genes that affect glycogen synthesis.

The arrangement of genes carried by pOP12 has also been determined by deletion-mapping experiments (84), and the nucleotide sequence of the entire glg gene cluster was determined (6, 7, 57, 58, 98, 104, 107, 130). The genetic and physical map of the E. coli K-12 glg gene cluster is shown in Fig. 6. The continuous nucleotide sequence of over 15 kb of this region of the genome has been determined and includes the sequences of the flanking genes asd (44) and glpD (encoding glycerol phosphate dehydrogenase; EC 1.1.99.5 and 1.1.1.8) (3). This region of the E. coli K-12 chromosome is centered at kb 4140 on the physical map by Kohara et al. (53) and is encompassed within kb 3584 to 3594 on version 6 of the physical map by Rudd et al. (112). Nucleotide sequence analysis indicates that in addition to the glgC, glgA, and glgB genes, pOP12 contains an ORF, glgX, located between glgB and glgC and a second ORF, originally termed glgY, located downstream from glgA (107).

Fig. 6E. coli K-12 glycogen gene cluster. This map was constructed from the sequences obtained from references 3, 6, 7, 44, 57, 107, and 134. asd corresponds to aspartate semialdehyde dehydrogenase, glgB encodes the BE, glgX encodes isoamylase, glgC encodes ADP-glucose PPase, glgA encodes glycogen synthase, glgP encodes glycogen phosphorylase, and glpD encodes glycerophosphate dehydrogenase. All the genes are transcribed from left to right (counterclockwise on the genome), except for the glpD gene.

The glgY gene, now designated glgP, was identified by homology to a gene for rabbit muscle glycogen phosphorylase (104, 134). The glgP gene encodes glycogen phosphorylase, as indicated by its expression and the characterization of its gene product (134). A recent study (2) showed that a 20-fold increase of glycogen phosphorylase activity by the overexpression of the glgP gene can decrease glycogen amounts to undetectable levels. The lowering of glycogen levels can be directly correlated with various increases in the expression of glycogen phosphorylase activity (2). In glgP deletion mutants, the chain lengths of E. coli glycogen increase into the range of 13 to 30 glucose units at the expense of chains of 6 to 8 residues (2). Thus, the regulation of glgP expression would be able to dictate the levels of glycogen in the bacterium.

The function of glgX is not known, but the deduced amino acid sequence of the glgX gene product is significantly related to the sequences of α-glucanases and transferases, including α-amylases, pullulanase, cyclodextrin glucanotransferase, the glycogen BE, and others. The homologous regions include residues that have been reported to be involved in substrate binding and cleavage by α-amylases and other members of the amylase family (50, 107, 120).

The catabolism of glycogen in mammals and plants requires debranching activity for the cleavage of α-1,6 linkages of glycogen (56, 107). Evidence that a debranching enzyme is present in E. coli was published in 1976 (49). Yang et al. (130) showed that mutants lacking the chromosome region containing glgX do not express detectable levels of glycogen debranching activity. However, the expression of glgX in those mutants increases debranching activity to high levels (130). The favored substrate is the phosphorylase limit dextrin derived from glycogen, and the debranching activity with glycogen as the substrate is less than 6% of that observed with the phosphorylase limit dextrin. The determination that the debranching activity corresponds to an isoamylase was made by Dauvillée et al. (24). The glgX product was purified and shown to have specificity similar to those previously reported by Jeanningros et al. (49) and by Yang et al. (130). The debranching activity on the phosphorylase limit dextrin is about 5.3 times greater than that on glycogen (24). The debranching enzyme has greater specificity for chains of only 3 to 4 glucose units than for longer chains and is essentially inactive with chains of greater than 6 glucose residues. Of interest are the findings of studies on the effects of glycogen levels in the GlgX mutants. Iodine staining of glgX-deficient cells is darker than that of cells containing glgX activity, suggesting that the glycogen in the mutants is at a greater level. The glgX-deficient cells have two to three times more glycogen than the wild-type cells, indicating that GlgX plays a role in glycogen degradation.

Neither glgX nor glgP is needed for glycogen or α-1,4 glucan synthesis, suggesting that both may be more involved in glycogen catabolism (107). It has also been suggested that the strict specificity of GlgX for only short chains of 3 to 4 glucose residues ensures that it does not start a possible futile cycle during glycogen synthesis (24). BE generates oligosaccharide chains of greater than 6 glucose residues. Thus, during glycogen synthesis, GlgX would not be active. It would be active only when glycogen synthesis is limited due to the lack of carbon, conditions under which phosphorylase activity commences to produce a glycogen molecule with chains of 3 or 4 glucose residues. The anabolic and catabolic genes of glycogen metabolism can therefore exist at the same time during the bacterial growth cycle.

It should additionally be pointed out that E. coli also produces a cytoplasmic α-amylase, AmyA (102). Its precise role in glycogen metabolism remains to be explored.

Inspection of the organization of the gene cluster suggests that the glg genes may be transcribed as two tandomly arranged operons, glgBX and glgCAP (Fig. 6). The coding regions of the glgB and glgX ORFs overlap by 1 bp; glgC and glgA are separated by 2 bp, and the genes glgA and glgP are separated by 18 bp. The close proximity of these genes suggests translational coupling within the two proposed operons. However, a noncoding region of approximately 500 bp separates glgX and glgC. As will be discussed below, the results obtained from studies of the regulation of the glg structural genes, with ′lacZ translational fusions, are consistent with a two-operon arrangement for the glg gene cluster, in which the glgCAP and glgBX operons may be preceded by growth phase-regulated promoters. Transcription processes initiating upstream of glgC have been analyzed by S1 nuclease mapping (104) and will be discussed below.

GENETIC REGULATION OF GLYCOGEN SYNTHESIS BY cAMP AND GUANOSINE 5'-DIPHOSPHATE 3'-DIPHOSPHATE

The first evidence that cyclic AMP (cAMP) affects bacterial glycogen synthesis was the finding that the addition of exogenous cAMP to E. coli W4597(K) results in a modest enhancement in the rate of in vivo glycogen biosynthesis (27, 28). It was later observed that the genes cya, encoding adenylate cyclase (EC 4.6.1.1), and crp, encoding cAMP receptor protein (CRP), are required for the optimal synthesis of glycogen and that exogenous cAMP can restore glycogen synthesis in a cya mutant but not in a crp mutant (64).

It was later shown that cAMP and CRP are strong positive regulators of the expression of the glgC and glgA genes but do not affect glgB expression (110). The addition of cAMP and CRP to E. coli strain S30 extracts having in vitro coupled transcription-translation reactions and containing pOP12 as the genetic template resulted in up to 25- and 10-fold increases in the expression of glgC and glgA, respectively, but did not affect glgB expression (110). cAMP and CRP also enhance the expression of glgC and glgA carried by either plasmids or restriction fragments in reaction mixtures with completely defined compositions in the dipeptide synthesis assay (124). In the reactions, the formation of the first dipeptide of a specified gene product, directed by a DNA template, is quantified (110, 124). A restriction fragment that contains glgC and 0.5 kb of DNA from the upstream noncoding region of glgC is sufficient to permit cAMP-CRP-regulated expression in the dipeptide synthesis assay (124), suggesting that the glgC gene contains its own cAMP-regulated promoter(s). Evidence for a CRP binding site on a 243-bp restriction fragment from the upstream region of glgC was obtained using gel retardation analysis (110). There are also potential consensus CRP binding sequences within the upstream region preceding the glgC gene in both E. coli (110) and Salmonella serovar Typhimurium (108). These sequences are shown in Fig. 7.

Fig. 7cAMP-CRP binding sites and sequences upstream of the glgC genes of Salmonella serovar typhimurium and E. coli. The E. coli site was detected by DNase I protection analysis (110), and that of Salmonella serovar typhimurium was derived on the basis of homology to the E. coli sequence (108). The asterisks indicate the conservation in the sequences of the transcripts with the consensus CRP binding site.

Experiments in which proteins encoded by the glg structural genes were expressed from plasmid pOP12 in maxicells showed that exogenously added cAMP stimulates the expression of glgC and glgA but does not affect glgB expression (104). Evidence that glgC is regulated by cAMP in vivo was obtained by constructing an in-frame plasmid-carried glgC′-′lacZ translational fusion, designated pCZ3–3, which contained 0.5 kb of the upstream noncoding region of glgC (104). This gene fusion is expressed approximately fivefold better in a cya⁺ strain than in an isogenic Δcya strain, and expression in both strains is stimulated by exogenous cAMP.

The first suggestion that glycogen biosynthesis in E. coli is positively regulated by ppGpp was the observation that relA strains are glycogen deficient (19, 65, 122). It is now well-established that the expression of the glgC and glgA genes is stimulated by ppGpp (104, 110). The expression of glgC in transcription-translation reactions is increased three- to fourfold in the presence of ppGpp; glgA expression exhibits approximately a twofold enhancement. The expression of glgB is not affected by ppGpp.

cAMP and ppGpp also have independent effects on glgCA expression in in vitro transcription-translation experiments (110). Actually, their combined effects on glgC expression in transcription-translation experiments can be synergistic (110). The addition of cAMP-CRP or ppGpp results in an increase of 6.3- or 1.6-fold, respectively, of the expression of glgC over the basal or unactivated level of expression, while their addition together leads to 18.8-fold stimulation.

Evidence for the positive regulation of glgC expression in vivo by ppGpp was obtained using the glgC′-′lacZ translational fusion in pCZ3–3 (104). This gene fusion was introduced into strains in an isogenic series that varied in basal levels of ppGpp due to increasingly severe mutations in spoT (113). The spoT gene affects the levels of ppGpp in the cell. The expression of the glgC′-′lacZ gene fusion is exponentially correlated with ppGpp levels in this series of strains (104).

NEGATIVE GENETIC REGULATION OF GLYCOGEN SYNTHESIS

Results from studies of glycogen excess E. coli B mutants SG3 and AC70R1, which exhibit enhanced levels of the enzymes in the glycogen synthesis pathway (i.e., they are derepressed mutants), suggested that glycogen synthesis is under negative genetic regulation (56, 84, 89, 100). The mutations in these strains, glgR and glgQ mutations, respectively, affect glg transcription (110), although these mutations have not been isolated and sequenced.

The 5′ termini of four in vivo transcripts from the region upstream of glgC, within 0.5 kb of the glgC coding region, were identified by S1 nuclease protection analyses and designated transcripts A to D (110) (Fig. 8 and Fig. 9).

Fig. 8The glgC transcripts A, B, and C were mapped using S1 nuclease protection analysis (110) and are located at −60, −130, and −245 relative to the glgC initiation codon, respectively. The 5′ termini of glgC transcripts are indicated by asterisks, and the best −10 and −35 regions are underlined.

Fig. 9Regulatory sites for glgC transcription. The CRP binding region was located by mobility shift analysis (110). Transcripts from wild-type strains of E. coli B and E. coli K-12 and isogenic glycogen synthesis mutants were examined by S1 nuclease protection analysis (110).

The glgR mutation is closely linked to the glycogen structural genes by P1 transduction analysis, and the mutation results in 8- to 10-fold-higher levels of ADP-Glc PPase and 3- to 4-fold-higher levels of glycogen synthase in exponential phase but does not alter the level of BE in minimal medium (91). Analyses of RNA transcripts for glgC in strain SG3, having the glgR mutation, reveal an increase in transcript B only (110) (Fig. 8 and Fig. 9). Therefore, it appears that the glgR mutation may alter a cis-acting site involved in the regulation of transcript B. This effect may be mediated via a negative regulatory site, but the current experimental evidence is also consistent with an overexpression phenotype or a higher-affinity CRP binding site.

The glgQ mutation is not linked to the glycogen gene cluster by P1 transduction analysis and results in 11-, 5.5-, and 2-fold increases in ADP-Glc PPase, glycogen synthase, and glycogen BE, respectively (96). Therefore, glgQ appears to affect one or more trans-acting factors for the expression of the genes in the two glycogen operons. Levels of the four transcripts for the glgC gene are elevated in the glgQ mutant AC70R1 (110) (Fig. 8 and Fig. 9). Transcript A was affected the most dramatically, with approximately 25-fold-higher levels being present in AC70R1 than in wild-type E. coli B or the SG3 strain (110). Since the levels of BE are also elevated in AC70R1, it was not considered likely that glgQ was a mutation in the cAMP-CRP or ppGpp regulatory systems, which do not affect glgB expression. The patterns of expression of the chromosomal lacZ gene in AC70R1 (the glgQ mutant) and in the wild-type E. coli B strain were also similar, providing further evidence for the idea that glgQ affects a different regulatory system for the glg genes (108). In summary, glgQ affects mainly transcript A and glgR affects mainly transcript B.

THE csrA GENE PRODUCT IS A NEGATIVE EFFECTOR OF THE glg OPERONS

The identification and characterization of a gene from E. coli K-12, csrA, that encodes a negative factor for glg transcription have been reported previously (106). The relationship of csrA to the E. coli B glgR and glgQ mutations remains unknown.

Transposon mutations that affect glycogen biosynthesis in E. coli K-12 were isolated previously (106). The transposon approach used facilitated the identification, molecular cloning, and mapping of trans-acting regulatory genes for glycogen biosynthesis. Mutations were introduced into a strain that contained pCZ3–3 (a plasmid having a glgC′-′lacZ translational fusion), and the resulting mutants were stained with I₂ vapor to detect cell colony glycogen. The plasmid-encoded β-galactosidase (EC 3.2.1.23) was also detected in glycogen excess mutants (106), and several glycogen excess mutants overexpressing the plasmid-borne glgC′-′lacZ fusion were isolated. One mutant, TR1–5, accumulated about 24-fold more glycogen than an isogenic wild-type strain (106). The gene affected by the TR1–5 mutation, csrA (for carbon storage regulator), has been cloned, sequenced, and mapped on the E. coli genome, and its regulatory effects have been studied (105, 106).

The TR1–5 mutation was also shown to affect glycogen levels by causing elevated expression of genes representative of both glycogen operons, glgC and glgB. Levels of ADP-Glc PPase expressed from the chromosome were approximately 10-fold higher in the TR1–5 mutant than in an isogenic csrA⁺ strain in the stationary phase. The β-galactosidase activities expressed from the glgC′-′lacZ and glgB′-′lacZ translational fusions were approximately seven- and two- to threefold higher in the TR1–5 mutant, respectively, than in an isogenic csrA⁺ strain. The TR1–5 mutation affects glycogen levels and the expression of the glgB and glgC genes in the exponential as well as the stationary phase.

The csrA gene also appears to regulate the expression of the gluconeogenic enzyme phosphoenolpyruvate carboxykinase (EC 4.1.1.38). The expression of a phosphoenolpyruvate carboxykinase operon fusion (pckA′-′lacZ) in the TR1–5 mutant in exponential and stationary phases was increased about twofold compared to that in an isogenic csrA⁺ strain, suggesting that gluconeogenesis may also be under negative control by csrA (106). When several isogenic strains were grown on synthetic medium, it was found that csrA⁺ and csrA::kanR strains (transductants with the TR1–5 mutation) were capable of growth on a wide variety of carbon sources. However, a strain that contained the functional csrA gene carried on a multicopy (pUC19-based) plasmid, pCSR10, could grow on glucose and fructose but not on any of the gluconeogenic substrates, succinate, glycerol, pyruvate, and l-lactate (106). When strains were plated onto a richer defined medium, the growth of a pCSR10-containing strain using some gluconeogenic substrates, including acetate, as a major carbon source did occur. However, the pCSR10-containing strain formed only pinpoint colonies on succinate, whereas each of the other strains grew well. Perhaps the csrA gene affects succinate utilization independently of its effect on gluconeogenesis, possibly at the level of succinate transport into the cell.

The csrA gene, mapped by genetic and physical approaches (105), is located at 58 min or at kb 2830 on the physical map of the E. coli K-12 genome (53). The csrA gene is between the gene alaS, which encodes alanyl-tRNA synthetase (EC 6.1.1.7), and the serV operon of tRNA genes and is transcribed in the counterclockwise direction on the chromosome (105).

Nucleotide sequence analysis revealed that the largest ORF, located between alaS and serV and present in pCSR10, exhibits an upstream sequence typical of E. coli ribosome binding sites (96, 105). The sequencing of the csrA mutant allele in strain TR1–5 showed that this ORF is disrupted by the kanR marker in this mutant (96, 105).

The csrA ORF encodes a 61-amino-acid polypeptide, which was strongly expressed from the plasmid pCSR10 in previous S30 transcription-translation experiments (106). Deletion-mapping experiments with the plasmid-borne csrA gene demonstrated that the ORF is required to mediate the inhibitory effects on the glycogen synthesis phenotype in vivo.

Analysis of glgC transcripts by S1 nuclease protection mapping showed that the steady-state levels of all four glgC transcripts are elevated in the TR1–5 mutant compared to those in the wild type and are severely depressed in a pCsr10-containing strain, indicating that csrA affects the transcriptional regulation of glgC (96, 106).

The csrA gene product, CsrA, is an RNA binding protein and negatively modulates by facilitating the decay of glgCA mRNAs (76, 77). Thus, it modulates the stability of mRNAs. It binds to the small RNAs CsrB (369 nucleotides [nt]) and CsrC (242 nt) (75, 125, 127). The binding of CsrA by CsrB RNA inhibits the repression of the glgCA operon and the glgB gene. CsrB and CsrC have multiple binding sites for CsrA, and the binding of CsrA is cooperative (5, 125, 127). There are a great number of GGA sequence motifs in CsrB, and it is believed that they are important for the binding of CsrB (5, 75). Results from biochemical studies show that CsrA binds to CsrB in a ribonucleoprotein complex, with CsrB containing several repeated sequence elements that perform as CsrA binding sites (75). Thus, CsrB acts as an antagonist to CsrA function by forming a complex with CsrA. Similarly, another small RNA, CsrC, also can bind CsrA and acts similarly to CsrB in inhibiting csrA gene function (125). It also has multiple binding sites for CsrA (75).

As CsrB and CsrC levels increase in stationary phase (42, 127), it is believed that the repression of the glycogen biosynthetic genes is reversed by the binding of the csrA gene product by CsrB and CsrC. A detailed review of the CsrB family in bacteria has recently been published (5). CsrA is a global regulator, as it mediates the repression of glgC, cystA, and pga. These are the initial genes of operons for glycogen synthesis (8), peptide transport (29), and biofilm adhesin synthesis (125), respectively.

SIGMA FACTOR E·σ70 TRANSCRIBES glgCAP

Analyses of transcripts originating upstream from the glgCAP operon by S1 nuclease protection mapping revealed four transcripts. Three of these transcripts were mapped at high resolution (110). The DNA sequences immediately preceding the 5′ ends of these transcripts were weakly related to consensus sequences for E. coli promoters (110) (Fig. 8). Although positively regulated promoters typically show weak similarity to the consensus sequence (43), it is also possible that one or more of the glg promoters is recognized via an alternative sigma factor. Therefore, the dependence of glgC expression on three sigma factors was tested previously in coupled transcription-translation experiments using monoclonal antibodies that selectively inhibit transcription by specific and selective recognition of the sigma factors (52, 70, 103).

A monoclonal antibody directed against E. coli σ70 inhibited up to 85% of the glgC expression (96), while antibody directed against σ⁵⁴ or σ³² did not inhibit the expression of glgC, in contrast to that of glnAP2 (σ⁵⁴-dependent) and dnaK (σ³²-dependent) controls. Therefore, even though nitrogen limitation enhances glycogen synthesis in vivo, the expression of glgC is not regulated by the nitrogen starvation, σ⁵⁴-dependent transcriptional controls. This result agrees with the finding that NtrC and NtrA (σ⁵⁴) did not enhance the expression of glg genes in experiments with strain S30 (110). The heat shock regulatory system (σ³² dependent) also appears to have no involvement in the control of glgC expression. Therefore, the major active form of RNA polymerase (E. coli σ70) is apparently utilized for glg expression in the S30 transcription-translation system. The S30 extracts used in the experiments were prepared from exponential-phase cells to obtain optimal translational activity and would not have contained endogenous activity from a sigma factor or an accessory factor synthesized in stationary phase. Therefore, the expression of one or more glgC transcripts may have gone undetected in these experiments.

POSSIBLE REGULATION OF GLYCOGEN SYNTHESIS BY A STATIONARY-PHASE SIGMA FACTOR

The findings of studies from three laboratories provide evidence for a class of stationary-phase-induced genes depending on an alternative sigma factor for expression (45, 80, 114). The gene for this sigma factor is referred to as katF or rpoS. Iodine staining of katF-positive and -negative strains suggest that a functional katF allele is required for the optimum accumulation of glycogen in E. coli strain MC4100 (60). Hengge-Aronis and Fischer (46) have isolated, cloned, and sequenced a gene in E. coli K-12 that stimulates glycogen synthesis, glgS, requiring katF for optimal expression. The glgS gene appears to be transcribed via a cAMP-dependent promoter and a katF-requiring promoter. Based on the iodine staining of colonies, a glgS null mutant can accumulate more glycogen than a katF mutant, indicating that katF has additional effects on glycogen synthesis besides the induction of glgS. A glgS mutation, however, does not affect the expression of glgA or glgC in vivo. Therefore, the effect of katF on glycogen synthesis does not function via the regulation of the structural genes for the enzymes in the glycogen pathway.

Although several mechanisms by which glgS may affect glycogen biosynthesis have been suggested, the function of glgS and the role of katF in glycogen synthesis remain to be defined. The possibility remains that these factors regulate a process occurring prior to the glycogen biosynthesis pathway. The actual mechanism by which katF and glgS affect the ultimate level of glycogen accumulation has not been elucidated.

Therefore, E σ70 is the primary form of RNA polymerase responsible for the expression of glgC, although there is a formal possibility that one or more of the proposed glgC promoters utilizes a yet-uncharacterized alternative sigma factor.

PROPOSED INTEGRATED MODEL FOR THE GENETIC REGULATION OF THE GLYCOGEN BIOSYNTHESIS PATHWAY IN E. COLI

The regulation of glycogen metabolism involves a multitude of factors coordinating the glycogen synthetic rate with the physiology of the cell. The genetic regulation of the glycogen biosynthesis pathway by cAMP and ppGpp allows E. coli to adjust its metabolic capacity for converting the available carbon substrate into glycogen in response to the availability of carbon or energy or amino acids. When cells are rapidly multiplying, the levels of the enzymes are repressed, and although the energy and carbon are available for glycogen synthesis, the rate of glycogen synthesis is low. Upon nutrient deficiency, the synthesis of ADP-glucose PPase and glycogen synthase is induced and the capacity for glycogen synthesis is greater. The level of glycogen that is ultimately accumulated will be dependent upon the substrate availability and is subject to the allosteric regulation of the ADP-glucose PPase activity. Fructose 1,6-bis-P, the allosteric activator, is a signal of carbon excess in the glycolytic pathway and thus an appropriate signal to commence glycogen synthesis. With carbon excess, the energy level in the form of ATP is high and is another positive signal for glycogen synthesis, since the levels of other adenosine nucleotides, such as AMP, the inhibitor of ADP-Glc PPase, are low. Genetic regulation determines the capacity for glycogen synthesis, and this process should be distinguished from the regulation of the absolute glycogen levels. As an example, the glycogen biosynthetic enzymes are induced in stationary phase when cells are grown on enriched medium lacking glucose. Yet glycogen synthesis does not occur because of the insufficient carbon source. However, in medium with excess glucose, where nitrogen is limiting, the expression of the genes for the biosynthetic enzymes is somewhat weaker, as cAMP is at a low level. In contrast, the glycogen synthetic rate is relatively greater because of the carbon availability, and the conditions are such that allosteric activation occurs.

Mutants that are affected in either negative (glgR, glgQ, or csrA) or positive (cya, crp, relA, or spoT) control systems for glgCA gene expression clearly demonstrate that the genetic regulation of the levels of glycogen biosynthetic enzymes is the most important factor in determining the ultimate level of glycogen synthesized and accumulated under any given physiological condition.

The structural and regulatory genes involved in glycogen metabolism in E. coli are listed in Table 7, and the effects of both positive and negative regulatory factors which control the expression of the glg genes of the glycogen biosynthesis pathway are indicated. Many important questions remain to be solved, particularly the regulatory role of the global regulator csrA in at least three processes—glycogen synthesis, biofilm synthesis, and peptide transport—and the physiological relationships (5, 8, 29, 42, 75, 76, 77, 106, 108, 109). Moreover, the actual functions of the glgS, katF, and many carbon starvation induced genes (45, 126) in glycogen synthesis at the biochemical and molecular levels remain to be established.