Rabbit Skeletal Muscle Glycogen Synthase I. RELATIONSHIP BETWEEN PHOSPHORYLATION STATE AND KINETIC PROPERTIES*

samples of purified glycogen to deviate least from Michaelis-Menten kinetics. The Hill about 2 phosphates had been incorporated. In a recent paper slope, m, is an index of the extent of the departure from (5), Soderling contended, on the basis of enzyme activities hyperbolic kinetics, so that the results here indicated little measured in the presence and absence of glucose-6-P, that change in the nature of the kinetic behavior for UDP-glucose in complete conversion of glycogen synthase was effected by the passing from a minimally to a highly phosphorylated enzyme.

Glycogen synthase (UDP-glucose:glucogen 4-a-glucosy.ltransferase, EC 2. 4.1.11) is one of the enzymes that is regulated in the cell by covalent modification of the protein molecule. For glycogen synthase, this modification takes the form of phosphorylation and dephosphorylation (for a review, see Ref. l), leading to enzyme species that differ particularly in their activation by glucose-6-P. The dephosphorylated Z form of the enzyme, at high UDP-glucose concentration, is scarcely activated by this effector, whereas the phosphorylated D form has very low activity in the absence of glucose-6-P. As was originally reported by Smith et al. (2), and as is becoming increasingly clear, both from further work in this laboratory (3,4) and from the work of other investigators (5-7), the subunit of *This work was supported by United States Public Health Service Grants 2 ROl AM15334-07 and 1 P17 AM 17042-01 to the University of Virginia Diabetes-Endocrinology Research Center. glycogen synthase can be multiply phosphorylated. For the rabbit muscle enzyme the number of phosphorylation sites per subunit is between three and six. In the work presented here, the relationship between the phosphorylation state of the enzyme and some of its catalytic properties was determined by studying a series of glycogen synthase samples of different phosphate contents.

EXPERIMENTAL PROCEDURE
Materials-[[U-"C]UDP-glucose was prepared by the enzymic conversion of [U-"C]glucose (New England Nuclear) into UDP-glucose (8). Using the standard assay conditions described below, 99% of the radioactivity in [U-"C]UDP-glucose so prepared could be incorporated into glycogen on incubation with purified glycogen synthase. Contamination by radioactive glucose-l-P and glucose-6-P was estimated by measuring labeled glycogen formation after incubation with glycogen, phosphorylase b, AMP, and phosphoglucomutase.
Because relatively high UDP-glucose concentrations were used, and because of the serious implications of glucose-6-P contamination, UDP-glucose (Sigma) was carefully analyzed for glucose-6-P content using glucose-6-P dehydrogenase (9). Internal glucose-6-P standards were also run. No glucose-6-P was detectable in a 60 mre UDP-glucose solution, indicating less than 0.0025% (mole/mole) of the sugar phosphate in the UDP-glucose.
Rabbit liver glycogen (Sigma) was purified by passage through a column of Amberlite MB-3 ion exchange resin and precipitation with ethanol before use in enzyme assays (10). Purified CAMP-dependent' protein kinase (EC 2.7.1.37) from rabbit muscle was the generous gift of Dr. L. C. Huang (University of Virginia School of Medicine).
Purification of Glycogen Synthase-Glycogen synthase I and D forms were purified by the method of Smith et al. (lo), as modified by Takeda et al. (3). The enzyme samples numbered 1 (I form) and 8 (D form) below were so prepared.' Samples 7 and 9 (both D form) were similarly purified except that, after the usual gel filtration step with Sepharose 4B, a second chromatography with Sepharose 6B was added. To prepare glycogen synthase with intermediate phosphorylation state, the normal purification scheme was adjusted slightly. The procedure for the isolation of I form was followed up to the conversion of the enzyme to the I form, and one portion of the enzyme was continued through this preparation scheme (Sample 2). For the rest of the enzyme, the protocol was switched to that for the purification of D form and the enzyme was incubated at 7" with 5 rnM ATP, 10 +I cyclic AMP, and 12 rnM MgCl,. After various times of incubation, a portion of the enzyme was removed and taken through the purification steps for the D form, thus giving rise to Samples 3 to 6. Purified protein kinase was added before the final sample (Sample 6) was taken. The yield for each of these samples was 5 mg. In this preparation (that giving Samples 2 to 6) only fresh rabbit muscle was used, in contrast to the mixture of frozen and fresh muscle employed by Takeda et al. (3,4 Enzyme was diluted immediately before use into cold buffer containing 50 rnM Tris-HCl, pH 7.8; 5 rnM EDTA; 2 rnM EGTA; 50 mM mercaptoethanol; and 1 mg/ml of rabbit liver glycogen, and stored on ice. It was found in these experiments that incubation of the enzyme at 30" before use caused no increase in activity. The reaction was started by the addition of enzyme, and after an appropriate time (usually between 5 and 10 min), 75 pl of the reaction mixture were placed on a filter paper square which was deposited in 66% (v/v) ethanol. The filter papers were washed, dried, and counted as for the standard assay (10). The reaction temperature was 30". Reaction blanks were run by placing enzyme onto a filter paper before adding the rest of the reaction mixture and rapidly immersing the paper in 66% (v/v) ethanol. Where the concentration or specific activity of the [U-"C]UDP-glucose was varied in an experiment, blanks were run to accommodate these conditions.  Takeda et al. (3,4). The same authors also described a proteolytic breakdown product, of molecular weight 81,000, present in normal preparations of the Z form of glycogen synthase. The 81,000 molecular weight species was clearly visible in Sample 1, but only very faintly so in the other samples.
This difference appeared to depend on whether or not frozen muscle was included in the starting material, the use of frozen muscle increasing the proportion of the 81,000 molecular weight species. by the standard assay procedure in the absence of sulfate and the presence of 7.2 mM glucose-6-P.
Enzyme was from 74 to 175 ng per assay.
c Ratio, expressed as percentage, of enzyme activity in the absence of glucose-6-P to that in its presence at 7.2 rnM using the standard assay. Where indicated, SO,'-was present at 10 mM. Enzyme was between 74 and 300 ng per assay.
d Average and standard error of all nine specific activities.
Analyses for alkali-labile phosphate (Table I) indicated that the enzymes embraced the range 0.27 to 3.5 mol of phosphate/85,000 g, the uppermost values being close to the figure of 3.17 phosphates/subunit recently reported by Takeda and Larner (4). The specific activities of the samples (Table I), as determined by the standard assay with saturating glucose-6-P present, were relatively constant (37.1 * 1.0 ymol/min/mg).
In view of the manipulations involved to isolate the enzymes, this relative constancy of specific activity is suggestive, but not conclusive, evidence that the different samples were of similar purity and that all were capable of similar maximal activities in the presence of saturating UDP-glucose and glucose-6-P. The %I activity was a function of the phosphorylation state of glycogen synthase (Table I), decreasing with increasing phosphate content. The same trend was found whether or not 10 mM S0,2-was included in the reaction mixture, although the presence of SO,*-did lead to different absolute values for the %I activity.
Kinetic Behavior of Glycogen Synthase with Respect to UDP-Glucose-When the UDP-glucose concentration was varied in the range 25 pM to 25 mM, all nine glycogen synthase samples consistently showed deviation from Michaelis-Menten kinetics, resulting in Eadie-Hofstee plots convex to the origin. This is illustrated for representative enzyme samples in Fig. 1. Inspection of the three curves in Fig. 1 gives a qualitative indication of the large effect of phosphorylation on the kinetic properties of glycogen synthase, noting of course the different scales used on the axes in this figure. In order to characterize this behavior better, it is desirable to estimate the maximal rates, and as seen in Fig. 1, extrapolation to determine V,,,,, (the intercept with the abscissa) became more difficult with enzyme of greater phosphate content. The same basic problem remains when the data are plotted differently (as double reciprocal plots, for example). Since UDP-glucose concentrations above 25 mM could have caused serious problems due to salt effects (see below), it was not possible to determine V,,, for Samples 7 to 9. The maximal rates for the other enzyme samples, obtained by extrapolation of Eadie-Hofstee plots, are shown in Table II. There appeared not to be a strong correlation between the phosphorylation state of the enzyme and its maximal rate, at least in the range 0.27 to 2.29 mol of phosphate/85,000 g. In order to characterize the shape of the UDP-glucose saturation curves, the data were analyzed by Hill plots (Fig. 2).   Fig. 1. Values of S,. , the UDP-glucose concentration required for half-maximal rate, were calculated from Hill plots as in Fig. 2. *Kinetic parameters for UDP-glucose in the presence of 5 mM glucose-6-P. Both V,,,.. and S,., wre estimated from Eadie-Hofstee plots such as in Fig. 4.
'The M,., for glucose-6-P, concentration for half-maximal activation, was determined from Hill plots of data such as those in Fig. 5  versus log[UDPglucose] using V.,, as in Table I.' The lines are from unweighted least squares fits with log[UDP-glucose] the independent variable.
UDP-glucose saturation curves, as judged by the Hill plots, appeared not to depend on phosphorylation state. By contrast, the values of S,,, derived from such plots showed a very clear dependence on phosphorylation state, increasing with increasing phosphate content (Table II). For the three most phosphorylated samples, precise knowledge of S,, values was not possible, although they were certainly much higher than 25 mM. If, by analogy to enzyme of lower phosphorylation, the maximal rates were in fact close to those in the presence of glucose-6-P, the high values for S,., in Table II would be calculated. In spite of these difficulties the important point is nonetheless clear, namely that the phosphorylation of glycogen synthas? had a very marked influence on the S,, for UDPglucose, while the Hill coefficient, and the V,,., where measurable, were less influenced.
Effect of NaCl Concentration on Glycogen Synthase Actiuity-Interpretation of the deviations from Michaelis-Menten kinetics is in any event not straightforward, and is further complicated here by the fact that the highest substrate levels used significantly altered the ionic strength of the reaction mixture. Experiments to show the effect of salt concentration on reaction rate are shown for enzyme of high and low phosphorylation in Fig. 3. For both enzymes, increasing NaCl concentration caused a decrease in activity, although the more phosphorylated enzyme was more sensitive. Glucose-6-P, 5 mM, afforded protection to this decrease in activity. The influence of salt concentration was such that, for the more phosphorylated enzymes, the rates measured at high UDPglucose concentration were probably underestimates, and this may have contributed to the shape of the UDP-glucose saturation curves. With enzyme of low phosphorylation, however, this is unlikely, judging from Fig. 3. Thus, we feel that the observed anomalies in the kinetic behavior of glycogen synthase with respect to UDP-glucose did not result primarily from changes in salt concentration.
Another explanation for non-Michaelis-Menten kinetic behavior would be the occurrence of UDP-glucose-dependent changes in the association state of glycogen synthase. Such behavior should be detected as a variation in kinetic properties Effect of NaCl concentration on the activity of glycogen synthase. The reaction rate with enzyme Samples 2 and 7 (as indicated) was determined as described under "Experimental Procedure" with 0.2 rnM [U-"C]UDP-glucose (12,000 cpm/nmol) and the NaCl concentration shown. Glucose-6-P was either absent (0) or 5 mM (0). The rates are normalized so that the velocity in the absence of NaCl is 1. The amount of enzyme per assay was: Sample 2, 75 ng; Sample 7, with glucose-6-P, 79 ng, without glucose-6-P, 2.1 pg.
with enzyme concentration.
For a glycogen synthase sample of somewhat intermediate phosphorylation state (Sample 4), little change in the kinetics with respect to UDP-glucose was observed over an &fold range of enzyme concentration.
We therefore have no evidence for this possibility.

Kinetic
Behavior of Glycogen Synthase with Respect to Glucose-6-P-Glucose-6-P is a well known activator of glycogen synthase from many sources (l), and it was of interest to study this activation as a function of phosphorylation state. In the presence of 5 mM glucose-6-P, the deviations from Michaelis-Menten kinetics for the variation of UDP-glucose concentration were virtually abolished (Fig. 4), leading to Hill slopes close to unity, irrespective of phosphorylation state. Inhibition of the enzyme at high UDP-glucose levels was seen for enzyme of low phosphate content (Fig. 4). For all enzyme samples, the S,., for UDP-glucose was greatly reduced by glucose-6-P, the greatest decrease occurring with enzyme of high phosphate content (Table II). On the other hand, there appeared to be little evidence for such a strong effect of glucose-6-P on the V max, at least for Samples 1 to 6, where comparisons could be made with the V,,, in the absence of activator (Table II). The S,., for UDP-glucose increased by a factor of 3 in passing from enzyme of low to high phosphorylation state, but the variation of S,, with phosphate content of the enzyme was much less than in the absence of glucose-6-P (Table II).
At nonsaturating UDP-glucose concentrations glucose-6-P can activate glycogen synthase of any phosphorylation state. A substrate concentration of 0.2 mM was selected to study the dependence of this activation on glucose-6-P concentration (Fig. 5). Slight inhibition at high glucose-6-P levels was consistently seen for enzyme of low phosphorylation (Samples 1 to 5). As the phosphorylation state of the enzyme increased, there was a tendency for the relation between the velocity increase caused by the sugar phosphate and its concentration to become more sigmoid (Samples 6 to 9), as evidenced by Eadie-Hofstee plots concave to the origin (Fig. 5). The Hill plots in such cases were generally nonlinear, with slopes varying from 2 at low glucose-6-P to unity at high concentration. When UDP-glucose was increased to 1 mM, the M,, for glucose-6-P was reduced (Table II), but the shape of the saturation curves was little altered (not shown). Values of MO., for glucose&P, estimated from Hill plots are listed in Table II state, and, in fact, changed by almost 3 orders of magnitude in passing from the least to the most phosphorylated enzyme. Increasing phosphorylation of glycogen synthase has usually been regarded as increasing the susceptibility of the enzyme to activation by glucose-6-P, which may seem initially rather opposed to the large increase in M,., for glucose-6-P reported here. The parameter M,,, however, relates only to the increase in rate above that in the absence of activator. Thus, for highly phosphorylated enzyme (D form), exhibiting a very low rate without activator, even a very low saturation with respect to glucose-6-P can cause a manyfold increase in velocity. A similar absolute increase in rate for the nonphosphorylated enzyme (I form) could represent only a small fractional increase over the nonactivated rate.

DISCUSSION
The rationale for these studies was to obtain purified samples of rabbit muscle glycogen synthase that were phosphorylated to varying degrees. This allowed comparison of the chemically determined alkali-labile phosphate content with some enzymic properties and enabled a more detailed description of the effect of phosphorylation on glycogen synthase than is possible by comparing only the extreme (I and D) forms of the enzyme. A basic assumption in this investigation has been the validity of comparing the different enzyme samples described. In justification are the following points. The enzyme samples appeared to be almost homogeneous when analyzed by gel electrophoresis, and had similar specific activities when assayed under standard saturating conditions (Table I). The enzymes were prepared by the same basic purification procedure, save for the adjustments to produce samples of different phosphorylation state. The reproducibility of column elution patterns and yields was good. Most importantly perhaps, the conditions for phosphorylation were comparable (with the single exception of Sample 6, to which purified protein kinase had been added). We feel then that these glycogen synthase samples could be compared constructively on the basis of their alkali-labile phosphate content. It was in fact one of the most striking features of this work that enzymic properties varied monotonically with the phosphate content.
The analyses of alkali-labile phosphate for the various enzyme preparations indicated that the Zform could contain as little as 0.27 mol of phosphate/85,000 g. The D form could contain up to 3.5 mol/subunit. We also feel that the determina-0' I6 i 0 vq units mg-' FIG. 5. Activation of glycogen synthase by glucose-6-P. Glycogen synthase activity was measured as described under "Experimental Procedure" with 0.2 mM UDP-glucose (9300 cpm/nmol) and glucose-6-P as indicated: enzyme Samples 3 (O), 5 fiM to 1 mM; 6 (0) and 7 (A), 5 PM to 10 mM; and 9 (O), 50 PM to 10 mM. From 17 to 30 ng per assay were used. The results are shown as Eadie-Hofstee plots using the velocity increase, Y., caused by glucose-6-P (activated rate minus rate in the absence of sugar phosphate) in place of velocity. The rates for the different samples in the absence of glucose-6-P were 8.2 units/mg (Sample 3), 0.24 units/mg (Sample 6), and 0.0 units/mg (Samples 7 and 9).
tion of a whole range of alkali-labile phosphate levels for glycogen synthase lends credence, in an admittedly indirect way, to the uppermost values, especially in view of the continuous variation of enzymic properties with phosphate content. The values reported here for the D form of the enzyme are close to those recently quoted by Takeda and Larner (4) although somewhat lower than the earlier number of 6 found by Smith et al.
(2). The reason for this difference remains unresolved but it now seems certain, from work in this laboratory and others (5-7), that rabbit muscle glycogen synthase consists of subunits capable of multiple phosphorylation.
Glycogen synthase is a polymeric enzyme (13)(14)(15)(16) with the possibility for multiply phosphorylated subunits (2, 5-7) with several known effecters (1) and a macromolecular substrate, glycogen. Overinterpretation of kinetic results is obviously a hazard, and so we attempt to be cautious in the following discussion. In the absence of glucose-6-P, the dependence of the reaction rate on the concentration of the substrate, UDP-glucose, lead to Eadie-Hofstee plots convex to the origin. Deviations from Michaelis-Menten kinetics with respect to UDP-glucose have been observed several times for glycogen synthase from various sources (17)(18)(19)(20)(21)(22). The most usual interpretations of the behavior reported in the present work are negative cooperativity (23), heterogeneity of catalytic sites (24), or effects on the association state of the enzyme (25). In this work, the fact that the non-hyperbolic kinetics did not depend on enzyme concentration is perhaps an argument against the last possibility. Differentiation between negative cooperativity and heterogeneity of catalytic sites is essentially impossible from UDP-glucose saturation curves alone. From a priori considerations of an enzyme that can be multiply phosphorylated and that combines with an ill defined polysaccharide substrate, one might indeed predict considerable heterogeneity of catalytic sites. Further, the extremes of highest and lowest phosphate content ought to display the least heterogeneity with regard to phosphorylation, since both zero and maximal phosphorylation should correspond to single species. Consequently, the extreme forms might be anticipated to deviate least from Michaelis-Menten kinetics. The Hill about 2 phosphates had been incorporated. In a recent paper slope, m, is an index of the extent of the departure from (5), Soderling contended, on the basis of enzyme activities hyperbolic kinetics, so that the results here indicated little measured in the presence and absence of glucose-6-P, that change in the nature of the kinetic behavior for UDP-glucose in complete conversion of glycogen synthase was effected by the passing from a minimally to a highly phosphorylated enzyme. introduction of 2 mol of slp,/90,000 g of enzyme. The reason We also know that in the presence of saturating glucose-6-P, all for the disparity between his results and our alkali-labile the UDP-glucose sites behaved as though they were identical, phosphate determinations is not entirely clear, although differsince hyperbolic kinetics were found. The final observation ences in phosphorylating conditions may be important. We relevant to these considerations is the finding of positive would point out, however, that %I activity is a useful but not cooperativity with respect to glucose-6-P activation, as has necessarily the most sensitive parameter by which to monitor been noted previously (20,26,27). This is difficult to accom-"conversion." For example, at a very low substrate concentramodate on a model of heterogeneous sites. On balance, the tion, a lo-fold increase in K, causes a lo-fold decrease in an results here can be economically explained if glycogen synthase enzymic rate, although both rates may be very close to zero. exhibited subunit interactions, but other explanations cannot In our studies, we have clear evidence for further kinetic be rigorously excluded.
changes at alkali-labile phosphate contents greater than 2. Irrespective of the detailed interpretation of the shapes of the kinetic curves, it is clear that the properties of glycogen synthase with respect to variation of both UDP-glucose and glucose-6-P concentrations varied extensively with the phosphorylation state of the enzyme. The M,, for glucose-6-P, which is a well defined parameter in an operational sense varied by 800-fold over the phosphorylation range studied. The S,., for UDP-glucose, though poorly defined for the three most phosphorylated samples (Samples 7 to 9), nonetheless showed considerable variation as a function of the alkali-labile phosphate content. With both parameters, the greatest change occurred above 2 phosphates/subunit.
For enzyme Samples 1 to 6 neither phosphorylation state nor the presence of 5 mM glucose-6-P greatly altered the V,,,,,. It has generally been suggested that glucose-6-P acts on muscle glycogen synthase by altering the V,,,,, of the phosphorylated D form (22,28,29), although in liver this question is less certain (30-33). It is not obvious why the results here differ from earlier studies with the muscle enzyme, although we point out that the evaluation of V mar may depend on the substrate range used (Fig. 2). For Samples 7 to 9, where evaluation of V,,, in the absence of glucose-6-P was not possible, effects on V,,,., were unknown, but phosphorylation or glucose-6-P binding had large effects on the apparent affinity for UDP-glucose. Thus, we envisage an enzyme in which the pronounced effects of phosphorylation and glucose-6-P on enzymic activity are mediated primarily through correspondingly pronounced effects on the apparent affinity of the enzyme for its substrate, UDP-glucose.
In studies of the phosphorylation of glycogen synthase, we as others, have measured the average number of phosphate residues per enzyme subunit (although in s*P incorporation experiments the starting phosphate content is not usually Average phosphates/molecule As preface to a discussion of the multiple phosphorylation of the glycogen synthase subunit in relation to the enzymic properties reported here, a recent investigation of Cohen et al. (34) is of interest. These workers demonstrated, using purified CAMP-dependent protein kinase from rabbit muscle, that, of 22 proteins tested, only phosphorylase kinase (a and p subunits), glycogen synthase, and histone Fl could be phosphorylated to any significant extent (greater than 0.05% compared with the fi subunit of phosphorylase kinase). Under their conditions, there was a high degree of specificity for phosphorylation.
It therefore seems a reasonable working assumption that the observed multiple phosphorylation of the glycogen synthase subunit does not represent the gratuitous introduction of phosphate residues. It follows then to ask whether the present report contains any evidence for function of the different phosphorylation sites. We can say unequivocally that, with the phosphorylating conditions used, a detectable change in glycogen synthase properties was observed as up to 3.5 phosphates/85,000 molecular weight subunit were introduced, and in fact such changes were most pronounced after Averoge phosphotesknolecule FIG. 6. Two site model for multiple phosphorylation of an enzyme. Two sites of phosphorylation, designated as a circle and a square, are assumed per enzyme molecule. In A, both are assumed to be occupied with equal probability. The curues show the dependence of the fraction of given classes of enzyme species on the average phosphorylation state. 1, all enzyme molecules with at least 1 phosphate per molecule; II, enzyme molecules with one particular site, the circle, occupied regardless of the state of the other site. A similar curve would result if the other site was considered. III, enzyme molecules with both sites phosphorylated. In B, phosphorylation probability is assumed unequal, the circular site being phosphorylated with 10 times the probability of the square site. Once the fast site is saturated, all subsequent increases in average phosphorylation were assumed to fill the slower, square site. Curue I then shows the variation in fractional concentration of enzyme with the faster, circular site filled as a function of average phosphorylation. Curue II is analogous for the slower, square site. known). Clearly, the relationship between the level of phosphorylation at a specific site and the average phosphorylation could be quite complex. From the work presented here, it is evident that the behavior of the enzyme samples of intermediate phosphorylation was not deducible from the appropriate combination, on the basis of phosphate content, of the properties of the extreme forms of the enzyme.' Several explanations for this nonadditivity are possible, some of which are illustrated by the model of Fig. 6. Here, two phosphorylation sites per enzyme molecule have been assumed, and the relative concentrations of various phosphorylated species of enzyme are related to the average phosphorylation state. A property such as the reaction rate under fixed conditions will be proportional to the concentration of the relevant enzyme species. With equal probability of phosphorylation at each site (Fig. 6A), a nonlinear relationship between an enzymic property can only arise if both sites are relevant to changing this property. This could be the interaction of two identical sites (curoe IZfl or the equal but independent effect of the two sites (curue I). With unequal phosphorylation probabilities at the different sites (Fig. 6B), such nonlinearity can result even if only one site is relevant to changing the property of interest. The data presented here cannot distinguish these possibilities, or others not considered, and the model is presented only to emphasize the sort of statistical interpretations necessitated for an enzyme subject to multiple phosphorylation.