INTRODUCTION

Glycogen, a polymer of glucose, is a major storage form of carbohydrate in many cell types. In mammals, three enzymes are involved in the biosynthesis of this polysaccharide, glycogen synthase (EC 2.4.1.11), branching enzyme (EC 2.4.1.18), and the self-glucosylating initiator protein glycogenin (1, 2). Rabbit muscle glycogenin self-glucosylates at a single residue, Tyr¹⁹⁴, via a glucose-1-O-tyrosyl linkage in a priming reaction that forms an attached oligosaccharide up to about 10 residues long (3, 4, 5, 6). Once fully primed, the glycogenin can serve as substrate for glycogen synthase (5, 6, 7), which, together with branching enzyme, catalyzes the formation of mature glycogen molecules.

Glycogen synthesis in the yeast Saccharomyces cerevisiae appears to be mediated by a similar set of enzymes. Two genes, GSY1 and GSY2, encode glycogen synthase (8, 9), and GLC3/GHA1 codes for a branching enzyme (10, 11). Recently, Cheng et al. (12) used a yeast two-hybrid screen to search for proteins interacting with Gsy2p and identified a small COOH-terminal segment of a protein encoded by a gene that was ultimately designated GLG2. A second related gene, GLG1, was also identified. GLG1 and GLG2 encode proteins that, like mammalian glycogenin, self-glucosylate in vitro and can be elongated by purified glycogen synthase in vitro (12). Using genetic techniques, we showed that yeast cells lacking both GLG genes were unable to accumulate glycogen. Loss of a single GLG gene did not affect glycogen deposition, and the rationale for the existence of two GLG genes is at this time unclear. In glg1 glg2 cells, we also observed inactivation of glycogen synthase, suggesting that the presence of the Glg proteins affects the control of glycogen synthase via phosphorylation.

Glg1p and Glg2p display 55% sequence identity over an NH₂-terminal segment of some 250 residues (Fig. 1), in a region that is about 33% identical to glycogenin. Within this region is Tyr¹⁹⁴ of glycogenin, the site of self-glucosylation. Alignment of the three proteins indicates that both yeast proteins have a Tyr residue, Tyr²³², possibly in correspondence to glycogenin Tyr¹⁹⁴ although the degree of sequence similarity in this localized segment is weak between the mammalian and the yeast proteins. In Glg2p, but not Glg1p, there is a second Tyr in this region, at residue 230. The two yeast proteins also share a short COOH-terminal domain, with 13 out of 19 identities, that was present in the initial Glg2p two-hybrid clone and could be a site of interaction with Gsy2p. In this region, there is a Tyr residue conserved between Glg1p and Glg2p.

In the present study we have extended our functional characterization of the Glg proteins, in vivo and in vitro. We have demonstrated association of Glg protein with glycogen in yeast cells and have analyzed the role of potential glucosylation sites in glycogen synthesis. We conclude that, in contrast to the mammalian protein, multiple Tyr residues are important for the function of the yeast Glg proteins in vivo.

EXPERIMENTAL PROCEDURES

Wild type S. cerevisiae strain EG328-1A (MAT trp1 leu2 ura3-52) was provided by Kelly Tatchell (Louisiana State University). Derived isogenic strains DH3 (MAT trp1 leu2 ura3-52 gsy1::LEU2 gsy2::URA3), CC6 (MAT trp1 leu2 ura3-52 glg2::URA3). CC8 (MAT trp1 leu2 ura3-52 glg1-2::LEU2), and CC9 (MAT trp1 leu2 ura3-52 glg1-2::LEU2 glg2::URA3) were produced in our laboratory. Plasmids pRS314-GLG1 and pRS314-GLG2 (12) were used to express Glg1p or Glg2p in CC9 cells. In addition, a 3.7-kb¹ EcoRI-BamHI fragment containing the GLG2 sequence was cloned between the EcoRI and BamHI sites of the high copy number yeast episomal shuttle vector pRS424. The plasmids were transformed into yeast cells by the method of Elble (13) and selected by growth on synthetic medium lacking tryptophan (SD-TRP).

Two step PCR (14) was used to mutate the Tyr²³² of Glg2p to Phe. Using pET28-GLG2 (12) as a template, the initial PCR primers were CTGTTGAAAAGGCCTGAGCTATC (GLG2-StuIs) with GTGAAGACTGGATCCGTAGTTG (GLG2-232a) and CAACTACGGATCCAGTCTTCAC (GLG2-232s) with CAATATTGCCCAATTGCAAGTGTG (GLG2-MunIa). The mutated bases are underlined. The secondary PCR utilized GLG2-StuIs and GLG2-MunIa, and the resulting 0.6-kb product was digested with StuI and MunI and ligated into pET28-GLG2 to generate the plasmid pET28-GLG2 Y232F. To mutate Tyr²³² of Glg1p and Tyr²³⁰ or both Tyr²³⁰ and Tyr²³² of Glg2p to Phe, we did oligonucleotide-directed in vitro mutagenesis using Altered Sites^TM Systems (Promega). The 0.7-kb NdeI-NcoI fragments from pET28-GLG1 (12), pET28-GLG2, and pET28-GLG2 Y232F, respectively, were ligated into the pALTER-Ex1 vector. The mutagenic oligonucleotides used were as follows, with the mutations underlined and sites indicated in the primer names: GGTGAAGATTGGAGCCTAAATTAGG (GLG1-Y232F), GACTGGTATCCGAGTTGGGCATGG (GLG2-Y230F), and GACTGGATCCGAAGTTGGGCATGG (GLG2-Y230F,Y232F). Mutated DNA was then replaced in the appropriate pET28-GLG plasmid. The COOH-terminal mutations of GLG2 were introduced by one step PCR by putting an AccI site in the mutagenic primer. Using primers GGGAGTCTACGGATTTTTAGACCGCGTCC (GLG2-Y367F) and CCATATCATCGAACTTGTAGACGACCTTGTC (GLG2-AccI), we obtained a 0.4-kb PCR product and religated the resulting AccI-AccI fragment into pET28-GLG2. Primers to construct the deletion mutation at Trp³⁶² (362) were CACACACTTGCAATTGGGCAATATTG (GLG2-MunIs) and ATAATCCGTAGACTCTCAACAAATTT (GLG2-362a), where GLG2-362a changes the TGG to TGA stop codon. The resulting 0.2-kb MunI-AccI PCR fragment was religated into pET28-GLG2 using incomplete MunI and AccI digestion to give full-length GLG2 362 cDNA. COOH-terminal mutation of Tyr⁶⁰⁰ to Phe was introduced into GLG1 by two step PCR, using primers GAAGGAGAAACGAAGACGAGTGCAGTTGC (GLG1-BbsIs) with CTCTACTTTGGATAGAAATCAGAGTCCTCC (GLG1-Y600Fa) and GGAGGACTCTGATTTCTATCCAAAGTAGAG (GLG1-Y600Fs) with GTATATAGAATAGTTATGCATTTTTGTATAGTCG (GLG1-NsiIa). The resulting 0.2-kb fragment from the secondary PCR was digested with NsiI and BbsI and ligated into pET28-GLG1. From the above mentioned original mutations, cutting and pasting allowed production of further combined mutants including GLG1 Y232F,Y600F, GLG2 Y232F,Y367F, GLG2 Y230F,Y232F,Y367F, and GLG2 Y230F,Y232F 362. All DNA fragments that were manipulated in the course of the mutagenesis were sequenced to confirm the mutation (15).

To express the mutant GLG genes in yeast, a 1.7-kb NheI-NsiI fragment was excised from mutated GLG1 genes and transferred into pRS314-GLG1. Similarly, a 1.4-kb StuI-XhoI fragment was cut from mutated pET28-GLG2 plasmids and ligated into pRS314-GLG2.

Protein samples were boiled for 5 min in sample buffer, 62 mM Tris-HCl, pH 6.8, 2% (w/v) SDS, 37.9 mM dithiothreitol, 10% (v/v) glycerol, and 0.01% (w/v) bromphenol blue before being separated by SDS-PAGE with 4% acrylamide in the stacking gel and 12% in the separating gel according to the method of Laemmli (16). Samples labeled with ¹⁴C were dried on Whatman paper for autoradiography. For immunoblotting, proteins were first electrophoretically transferred for 1.5 h at 100 V to a nitrocellulose membrane (Schleicher & Schuell) using a Mini Trans-Blot module (Bio-Rad) in buffer containing 25 mM Tris, 192 mM glycine, 20% methanol, pH 8.3. The membrane was blocked with PBST (phosphate-buffered saline containing 0.1% Tween 20) plus 5% powdered milk and then probed with polyclonal anti-Glg2p antibody. The antibody was developed by immunizing a rabbit with purified recombinant His₆-Glg2p protein (Cocalico Biologicals, Inc.). Antisera were affinity-purified by the method of Gu et al. (17), and 2 µg of purified antibody/10-ml hybridization was used for immunoblotting. Signals were either visualized by incubation with ¹²⁵I-protein A (5 µCi/10-ml hybridization) followed by autoradiography or visualized by the ECL system (Amersham Corp.).

All of the mutated Glg1p and Glg2p proteins were expressed in Escherichia coli as His-tagged proteins using plasmids derived from pET28-GLG1 or pET28-GLG2. Cell culture conditions and protein purification procedures were as described previously for wild type proteins (12). Self-glucosylation was carried out with 1.8 µM protein (80 µg/ml) in 50 mM HEPES, pH 7.5, 5 mM MnCl₂, 2 mM dithiothreitol, and 2 mM UDP-[U-¹⁴C]glucose (9 Ci/mol) at 30 °C for the indicated time. An aliquot (5 µl) removed from the reaction was spotted onto P81 chromatography paper followed by three washes (30 min each) in 0.5% phosphoric acid and once in ethanol. The dried paper was counted in a liquid scintillation counter. To analyze the ability of Glg proteins to serve as substrates for elongation by glycogen synthase, proteins (40 µg/ml) were first allowed to self-glucosylate in the presence of 15 µM UDP-[U-¹⁴C]glucose (263 Ci/mol) at 30 °C for 30 min, at which time Gsy2p and unlabeled UDP-glucose were added to final concentrations of 0.12 mg/ml and 27 mM, respectively, and incubation continued for another 120 min. To assay transglucosylating activity, Glg proteins (10 µg/ml) were incubated for 5 min at 30 °C in 50 mM HEPES, pH 7.5, 5 mM MnCl₂, 2 mM dithiothreitol, and 150 µM UDP-[U-¹⁴C]glucose (263 Ci/mol) with 10 mM n-dodecyl -D-maltoside (Sigma) as the glucose acceptor in a total volume of 10 µl. An aliquot (3 µl) of the reaction mixture was then spotted on silica gel thin layer plates (G-25, Aldrich) and subjected to ascending chromatography. The plate was dried and subjected to autoradiography; the migration of oligosaccharide standards was visualized as described by Cao et al. (18). The silica gel corresponding to radiolabeled spots was removed from the glass plate and quantitated by scintillation counting.

To detect glycogen by iodine staining, yeast was grown in an appropriate medium (SD-TRP or SD-complete) to stationary phase, and 5 µl was spotted on SD plates. Patches were grown for 2 days at 30 °C and then inverted over iodine crystals in order to expose the colonies to iodine vapor for visualization of glycogen. For biochemical measurements, 5-ml SD cultures were grown for 24 h before being harvested by centrifugation. Cells were broken by glass beads and centrifuged for 90 s at 3000 × g, essentially as described previously (19). The supernatant was used for measurement of glycogen and glycogen synthase activity. The method for enzymatic determination of glycogen levels was described previously (20). Glycogen synthase activity was measured in the absence and presence of 7.2 mM glucose 6-phosphate, as described by Thomas et al. (21). The /+ glucose 6-phosphate activity ratio is often used to index the activation state of the enzyme, a lower activity ratio corresponding to higher phosphorylation levels.

Purified Glg2p protein was allowed to self-glucosylate in the presence of 15 µM UDP-[U-¹⁴C]glucose at 30 °C for 30 min and subjected to purification on a nickel chelate column to remove unreacted UDP-glucose and other small molecules. The ¹⁴C-labeled Glg2p was cleaved overnight with 5 mg/ml cyanogen bromide (Pierce) in 70% formic acid, lyophilized, and taken through two cycles of dissolving in water and drying. 100 µg of peptide (28,000 cpm) was loaded onto a C18 reverse phase microbore HPLC column (OD-300, Applied Biosystems) and eluted with a linear gradient of acetonitrile in 0.1% trifluoroacetic acid (0-100% acetonitrile in 45 min). Each peptide peak was collected, and the radioactivity was measured by scintillation counting. A major peak of radioactivity, designated peak 1, eluted at approximately 20% acetonitrile and contained about 75% of the total radioactivity. The second peak eluted at approximately 47% acetonitrile and accounted for 8% of the total radioactivity. Peaks 1 and 2 were sequenced with a Porton Instruments (Tarzana, CA) model 2090 integrated microsequencing system.

For some experiments, prior to immunoblot analysis, samples were digested (30 min for cell extracts or 2 h for purified glycogen) at 37 °C with 80 µg/ml -amylase (Worthington) in buffer containing 150 mM sodium acetate and 5 mM CaCl₂, pH 4.6. Purified Gsy2p was produced as a His-tagged protein from pET28a and purified by nickel-chelate chromatography to give protein that was more than 90% pure.² Protein was measured by the method of Bradford (22). Immunoblots and gels were sometimes analyzed and quantitated with a GS-250 Molecular Imager (Bio-Rad). Standard methods for the manipulation and analysis of yeast, including definition of media, were as described previously (23).

RESULTS

Our initial study of the GLG genes (12) had demonstrated, using purified recombinant protein, several biochemical properties expected of yeast glycogenin homologs and had provided genetic evidence for their involvement in glycogen accumulation in vivo. However, we had not directly demonstrated the presence of the Glg proteins in yeast cells, an analysis now made possible, at least for Glg2p, by the availability of antibodies (Fig. 2). Endogenous, free Glg2p was detected in lysates of wild type yeast as a species of M_r 46,000, close to the calculated molecular weight of 44,500. This species was not present in strains in which GLG2 is disrupted. A prominent nonspecific immunoreactive species of apparent M_r 66,000 was present in all samples, and its identity is unknown. Also noticeable from the immunoblotting is the ``trailing'' of the Glg2p signal to higher M_r in extracts from wild type cells.³ We interpret this behavior to reflect the covalent association of Glg2p with carbohydrate that retards electrophoretic mobility. In a gsy1 gsy2 strain, trailing is essentially eliminated, and the amount of free Glg2p protein increases, consistent with a lack of glycogen synthesis (Fig. 2). Similarly, -amylase treatment of extracts containing Glg2p caused a significant increase in the level of free Glg2p, consistent with it being attached to carbohydrate (see Fig. 8).

Proteins functioning like glycogenin would be expected to be associated covalently with high M_r glycogen. Therefore, the glycogen fraction was isolated from yeast extracts, digested with -amylase, and analyzed by Western immunoblotting using anti-Glg2p antibodies. Glg2p was detected in glycogen from wild type yeast (Fig. 3). No signal was seen with an undigested glycogen sample because high M_r glycogen scarcely enters the gel and is poorly transferred (data not shown). Glg2p was also observed in the digest of glycogen derived from glg1 cells, though at a somewhat reduced level. No Glg2p was detectable in the corresponding fraction of gsy1 gsy2 mutants that do not accumulate glycogen. We conclude that Glg2p can be covalently associated with carbohydrate in cells and is distributed between high M_r glycogen and free glycogenin.

In efforts to identify the site or sites of glucosylation of the Glg proteins, we initially allowed purified recombinant Glg2p to self-glucosylate and then cleaved it with CNBr. Two major ¹⁴C-peptides, representing ~80% of the incorporated radioactivity, were separated by reverse phase HPLC. Sequence analysis of the major species (75% of the radioactivity) yielded PNXGXQS (where X indicates no recognizable signal). Thus, the peptide began at Pro²²⁸, consistent with cleavage after Met²²⁷. Limit digestion with CNBr would generate an 11-residue CNBr fragment terminating in Met²³⁸. No phenylthiohydantoin-derivative signal was observed at the cycles corresponding to Tyr²³⁰ and Tyr²³² although Tyr normally gives a good signal on the system used. The sequence of the second peptide (8% of the radioactivity) was NFQQH, consistent with it beginning at Asn²³⁹ due to cleavage at Met²³⁸.

If CNBr cleavage were complete, the majority of the glucose incorporation would be localized to the region between Pro²²⁸ and Met²³⁸. Given also the sequence alignment of Glg1p, Glg2p and glycogenin (Fig. 1), we considered Tyr²³⁰ and Tyr²³² of Glg2p and Tyr²³² of Glg1p as candidate sites for glucosylation, analogous to Tyr¹⁹⁴ of glycogenin, and mutated them to Phe. Most work in vitro focussed on Glg2p, because recombinant protein can be produced easily in relatively pure form. Active recombinant Glg1p can also be expressed, but it is very sensitive to proteolytic degradation and contains several fragments, all of which are catalytically active (12). Purified Glg2p and several mutant forms of the protein were tested for their ability to self-glucosylate (Fig. 4). Wild type Glg2p incorporated in excess of 20 glucose residues/protein molecule, a significantly higher value than that observed for glycogenin. In a control reaction, glycogenin self-glucosylated to a stoichiometry of about 10 mol/mol. The rate of self-glucosylation, ~30 nmol/min/mg, is somewhat higher than that observed with glycogenin, 13 nmol/min/mg in this experiment. Mutation of Glg2p at either Tyr²³⁰ or Tyr²³² caused some reduction in the rate and extent of glucosylation, but both the Y230F and Y232F mutants retained substantial self-glucosylating activity. Mutation of both Tyr residues in the same molecule, however, virtually abolished the ability of Glg2p to self-glucosylate. For Glg1p, which has a single Tyr at position 232, mutation of Tyr²³² to Phe was sufficient to cause a similar loss of self-glucosylating activity (data not shown).

Elongation of glucosylated Glg2p by glycogen synthase (Gsy2p) was also examined (Fig. 5). In these assays, the Glg proteins were allowed to self-glucosylate at low UDP-glucose concentration prior to the addition of purified Gsy2p and unlabeled UDP-glucose. In control reactions, to which Gsy2p was not added, a dominant labeled species of molecular mass 46-60 kDa is seen, depending on the particular mutant. The ability to self-glucosylate essentially matched what was reported in Fig. 4, but since the UDP-glucose was limiting and consumed in the reaction, the labeling seen in Fig. 5A does not provide an accurate quantitation of rate and is essentially constant for all of the active Glg2p samples. Elongation is seen as the accumulation of higher M_r radiolabeled species, up to the top of the separating gel and even at the top of the stacking gel; concomitantly, the low M_r Glg2p disappears as is most clearly seen in Fig. 5 with the wild type Glg2p. The Y232F or Y230F single mutations did not prevent elongation by Gsy2p (Fig. 5B) but reduced the proportion of Glg2p converted to high M_r species. The Y232F,Y230F double mutant of Glg2p had severely impaired self-glucosylation under similar conditions, and consequently, no elongation was detectable (Fig. 5). The same was true for a Glg1p in which Tyr²³² was mutated to Phe (data not shown).

Physiological data presented below indicated that mutation of Tyr²³⁰ and Tyr²³² still did not totally abolish the ability of cells to synthesize glycogen. We considered the possibility that another site of self-glucosylation might exist that, although minor in vitro, might be functional in vivo. Because of the observation of a small amount of self-glucosylation in a fragment of Glg2p starting at Asn²³⁹, we examined the COOH terminus of the proteins. Tyr³⁶⁷ of Glg2p is the only position COOH-terminal to Asn²³⁹ where there is a conserved Tyr with respect to Glg1p (Tyr⁶⁰⁰ in Glg1p). Furthermore, this lies in a small section of high sequence identity (Fig. 1) that, from the original identification of Glg2p by two-hybrid screening, is precisely the region likely to interact with Gsy2p. Thus, we mutated Tyr³⁶⁷ or Tyr⁶⁰⁰ to Phe, and for Glg2p also made a COOH-terminally truncated form ending at residue 362. In Glg2p, both the COOH-terminal point mutation and truncation caused a reduction in self-glucosylation (Fig. 4) and elongation by Gsy2p (Fig. 5B). The latter defect is strongest for the 362 mutant, the vast majority of which remained at its starting mobility even after Gsy2p action. Comparable results, lack of effect on self-glucosylation and impaired elongation by Gsy2p, were observed with the Y600F mutant of Glg1p (data not shown). Adding either of the COOH-terminal mutations of Glg2p to a Y230F,Y232F double mutant did not alter the inability to self-glucosylate (Fig. 4); the same was true for a Glg1p Y232F,Y600F mutant (not shown).

Mammalian glycogenin can transfer glucose to exogenous acceptors such as maltose (18) or n-dodecyl -D-maltoside (24, 25), even after mutation of the single site of glucose attachment (18, 26). We observed that both Glg1p and Glg2p have transglucosylase activity and can transfer glucose to n-dodecyl -D-maltoside (Fig. 6). Compared with maltose, n-dodecyl -D-maltoside was a much better acceptor (data not shown). Note the faint species with the chromatographic mobility of glucose, corresponding to no more than 1% of the UDP-glucose substrate and indicative of a minor side reaction that results in UDP-glucose hydrolysis. Another trace species migrated similarly to maltose, but this was present also in controls lacking Glg proteins and is probably a minor contaminant of the UDP-[U-¹⁴C]glucose. In Glg1p, mutation of Tyr²³² caused a 70% decrease in transglucosylation, whereas modification of Tyr⁶⁰⁰ had a much lesser effect. The Y232F,Y600F double mutant had similar activity to the Y232F mutant. In Glg2p, mutation at Tyr²³⁰ or Tyr²³² significantly increased transglucosylating activity, as has been seen with glycogenin (18, 26). However, the Y230F,Y232F double mutant had transglucosylating activity that was reduced by ~5-fold; the extent was not further affected by additional COOH-terminal mutations. Although some of the Tyr mutations described here decreased transglucosylase activity, catalytic activity was by no means abolished.

Transglucosylation of Glg1p and Glg2p to n-dodecyl -D-maltoside.

The physiological relevance of the mutations described above were examined by expressing the mutant proteins in yeast cells. A glg1 glg2 strain (CC9), totally lacking Glg1p and Glg2p, cannot accumulate glycogen, but this capacity is restored by expression of either GLG1 or GLG2 (Ref. 12; Fig. 7). Wild type and mutant GLG genes were expressed from pRS314 under their own promoters. In GLG2, single site mutations like Y230F or 362 still allowed glycogen accumulation, albeit to a somewhat reduced level. Interestingly, cells expressing the double mutant Y230F,Y232F, which has undetectable self-glucosylating activity, did accumulate glycogen. Although representing only one-tenth of the glycogen of wild type cells, this value is statistically significant (p < 0.005) compared with glg1 glg2 controls containing empty pRS314 vector. Triply mutated GLG2, in which the double mutation Y230F,Y232F was combined with Y367F or COOH-terminal truncation, did not support glycogen storage. Glycogen accumulation was not simply determined by the levels of expression of the different mutant proteins, which did not vary hugely as judged by Western analysis of cell extracts (Fig. 8). For the wild type protein, -amylase treatment of the extract caused a clear increase in the amount of free Glg2p. Some increase was also seen with the Y230F mutant, whereas with the Y230F,Y232F double mutant or the Y230F,Y232F,Y367F triple mutant no obvious change in the protein level was visible after -amylase treatment.

Mutational analysis of GLG1 expressed in glg1 glg2 cells gave results comparable with those found for GLG2 (Fig. 7). In this case, the Y232F mutation, which virtually eliminates self-glucosylation, supported about half as much glycogen accumulation as the wild type. However, the double mutant Y232F,Y600F was unable to store glycogen. Lacking antibodies, we could not analyze Glg1p expression, but the results seem consistent with what was found for GLG2. We conclude that multiple tyrosine residues of both Glg proteins are involved in the initiation of glycogen synthesis.

Cheng et al. (12) showed that disruption of GLG1 and GLG2 did not affect the level of expression of glycogen synthase but decreased the /+ glucose 6-phosphate ratio in stationary phase. We sought here to determine whether this effect on the /+ glucose 6-phosphate activity ratio was linked simply to the presence of the Glg proteins or rather to their functionality. By monitoring the activity ratio associated with expression of the various GLG1 and GLG2 mutants, we observed that the effect on glycogen synthase activation closely paralleled the ability to support glycogen accumulation (compare Fig. 9 with Fig. 7). Thus, the presence of Glg proteins with impaired function correlated with inactivation of glycogen synthase by phosphorylation.

To test whether Glg protein might be rate-determining for glycogen storage, we overexpressed GLG2 from a high copy number pRS424 vector (27) in glg1 glg2 cells. With this vector, we obtained a significant overexpression as compared with that achieved from pRS314 as judged by Western blotting (Fig. 10). However, the glycogen level was no different in low and high expressing cells as determined by iodine staining (Fig. 10) or enzymatic determination (data not shown). Changing the medium glucose concentration did not alter this result (data not shown). The amount of Glg2p was therefore not limiting for glycogen accumulation.

DISCUSSION

One of the main outcomes of this study is to demonstrate that, unlike the mammalian protein glycogenin, multiple Tyr residues in the yeast self-glucosylating Glg proteins are implicated in sustaining glycogen accumulation in yeast cells. These tyrosine residues are localized to two different regions, one in correspondence with Tyr¹⁹⁴ of glycogenin and the other at the extreme COOH terminus in a segment conserved between Glg1p and Glg2p but with no counterpart in glycogenin.⁴ Glg2p, which for technical reasons we could analyze more completely, is the more complex. Glg2p has two tyrosines, Tyr²³⁰ and Tyr²³², that are most likely glucosylated. Evidence supporting this conclusion includes the observation that mutation of either one to Phe did not abolish self-glucosylation; however, mutation of both residues essentially eliminated self-glucosylation at the same time that transglucosylation, although reduced by 5-fold, was still readily detectable. Also consistent with these two sites being glucosylated is the observation from peptide mapping that 75% of the label introduced by Glg2p in a self-glucosylation reaction was localized to a CNBr-peptide containing Tyr²³⁰ and Tyr²³². The mechanistic or physiological rationale for having two almost adjacent residues involved independently in glucose attachment in Glg2p is not clear.

The results strongly suggest a functional role also for the COOH terminus, a region potentially involved in interacting with glycogen synthase. However, it is less clear mechanistically how the COOH terminus is involved. First, we are not certain whether there is glucosylation in this region. A small amount of glucose label (8%) was associated with a peptide starting at Asn²³⁹. On the other hand, mutation of Tyr²³⁰ and Tyr²³² essentially abolished self-glucosylation under conditions where we should have easily seen a 10% level compared with wild type. Could mutations at Tyr²³⁰ and Tyr²³² indirectly affect the ability of Tyr³⁶⁷ to be a glucose acceptor? This hypothesis would not explain the fact that in vivo the Y230F,Y232F double mutant supported glycogen accumulation, albeit to a level only 10% that of wild type. The analogous result with Glg1p was even stronger, since the Y232F mutant accumulated half as much glycogen as wild type. Another possibility is that the Glg COOH terminus, by binding to glycogen synthase, targets glycogen synthase to other sites of glycogen initiation. When the Glg protein does not have oligosaccharide primers attached, perhaps it is more likely to transglucosylate proteins promiscuously and, with glycogen synthase tethered to it, enable some ``illegitimate'' synthesis of glycogen. Alternatively, such targets for Glg transglucosylation might not be spurious but legitimate alternative primers. In either case, disruption of the interaction with glycogen synthase by COOH-terminal mutation could then overcome this occurrence.

In several other respects, a number of properties of the Glg proteins are what might be expected of a glycogenin homolog. Self-glucosylation by Glg2p proceeds to a significantly higher stoichiometry than glycogenin, but if one considers the presence of at least two sites of attachment then the difference is not so large. Both Glg proteins need Mn²⁺ for maximum activity, and known glycogenin inhibitors (5, 28, 29) such as CDP, UDP-xylose, UDP, UTP, and pyrophosphate inhibit Glg1p and Glg2p self-glucosylation.⁵ In assays of transglucosylation, maltose is a significantly worse glucose acceptor for the Glg proteins as compared with glycogenin, whereas n-dodecyl -D-maltoside works equally well for all three wild type proteins. As found with glycogenin, the abolition of self-glucosylation was not strictly linked to loss of transglucosylating activity, but unlike glycogenin, combined mutation at Tyr²³⁰ and Tyr²³² did significantly reduce the basic catalytic activity as judged by transglucosylation. A weak UDP-glucose hydrolyase activity was detected in both Glg1p and Glg2p, as has been observed also for glycogenin (Ref. 30).⁶ This side reaction is so slow in our hands compared with transglucosylation that it is unlikely to be physiologically relevant. For Glg2p, we have demonstrated that the protein is expressed in yeast cells and is attached to carbohydrate. This can be in the form of relatively small M_r species that can be separated by SDS-PAGE or in the high M_r glycogen fraction.

We are also interested in understanding whether there are circumstances in which the Glg protein could be regulatory. First, we showed that overexpression of Glg2p did not accelerate the accumulation of glycogen, indicating that, under these conditions, the amount of Glg2p is not limiting. This does not disprove a regulatory role for Glg2p under other conditions. Furthermore, if the Glg2p were under some negative control, simple overexpression of protein may be insufficient. For example, overexpression of wild type glycogen synthase does not affect the accumulation of glycogen unless activating mutations are introduced (20). There is in fact a potential regulatory role for the Glg proteins. We had already shown that, in the absence of the GLG genes, the activity ratio of glycogen synthase was suppressed in stationary phase, suggesting an effect on glycogen synthase phosphorylation (12). This was not due simply to a lack of glycogen, since the activity ratio of glycogen synthase in a glc3 strain, which also could not make glycogen, was not changed as compared with a control strain. From the present work, we can say that the presence of the Glg proteins alone is not sufficient to reestablish the activation state of the glycogen synthase. Rather, the ability of mutant Glg proteins to restore glycogen synthase activity is quite well correlated with their ability to sustain glycogen accumulation. However, the mechanistic basis for this correlation remains obscure.