Purification and Characterization of a Rabbit Liver Calmodulin-dependent Protein Kinase Able to Phosphorylate Glycogen Synthase*

A calmodulin-dependent protein kinase has been ex- tensively purified from rabbit liver by the criterion of its ability to phosphorylate muscle glycogen synthase. The enzyme bound to phosphocellulose, DEAE-cellu-lose, and blue dextran-agarose. The enzyme also bound, in a Caz+-dependent manner, to a calmodulin-agarose affinity column. Expression of activity required the presence of both calmodulin and Ca”, with half-maxi-mal activation occurring at approximately 80 n M calmodulin. Trifluoperazine, 50 p ~ , completely inhibited the enzyme. Enzyme activity was associated with two polypeptide species of apparent molecular weights 53,000 and 51,000. The molecular weight of the native enzyme was approximately 5O(r,OOO, as judged from gel filtration, and 275,000 as determined from sucrose density gradient sedimentation, suggesting an oligomeric structure. Incubation of the enzyme with ATP and Mg2+ led to phosphorylation of the constituent polypeptides and an accompanying decrease in electrophoretic mobilities in the presence o€ sodium dodecyl sulfate. The protein kinase had an apparent K,,, for ATP of 27 PM and had half-maximal activity at 0.75 mM Mg2+. The protein kinase phosphorylated muscle glycogen synthase to a stoichiometry of greater than 1 phosphate/ subunit with significant inactivation. Phosphate was introduced into at least two sites on the glycogen syn- thase. Another effective substrate for the protein kinase was the smooth muscle 20,000-dalton myosin

glycogen synthase kinase' have been identified: 1) cyclic AMPdependent protein kinase, 2 ) phosphorylase kinase, and 3) a cyclic nucleotide and Ca"-independent category that includes (5,6), PCo.4 (5), and the glycogen synthase kinase-3 (GSK-3) of Cohen and collaborators (7,8). The enzymology of liver glycogen synthase and its protein kinases is much more poorly understood. Since the regulation of liver glycogen metabolism displays both important similarities to and differences from that of muscle, it is of considerable significance to understand to what extent potential regulatory elements such as protein kinases are analogous in the two tissues. Of the cyclic nucleotide and Ca"-independent category, liver does contain (9), probably an activity analogous to PCM,~ and most likely an enzyme corresponding to GSK-3 (10). Both cyclic AMP-dependent protein kinase and phosphorylase kinase are known to occur in liver. It is not completely clear to what extent the liver phosphorylase kinase is a glycogen synthase kinase though one report does document such activity (11). Rabbit liver contains another Ca2'-dependent glycogen synthase kinase, first described by Payne and Soderling (12,13), that is a calmodulin-requiring enzyme. No clearly analogous enzyme has so far been reported in muscle. Special interest in Ca"-activated glycogen synthase kinases, especially from liver tissue, rests in the suggestion of Ca2+-mediated inactivation of glycogen synthase in isolated hepatocytes stimulated by a-adrenergic agonists, vasopressin or angiotensin (14)(15)(16). We have succeeded in confirming the existence of the liver calmodulin-dependent protein kinase and in purifying the enzyme close to apparent homogeneity. This report provides an initial characterization of the enzyme, in particular showing that the enzyme consists of two polypeptides of apparent molecular weights 51,000 and 53,000 arranged in an oligomeric structure. mg/ml, and other substrates were as indicated. When included, calmodulin and CaC12 were present at final concentrations of 0.57 pM and 0.4 mM (in excess of EGTA), respectively. Because of the very low yield of protein in the purified enzyme, determination of exact concentration was difficult. We estimate, however, that the calmodulin-dependent protein kinase was normally in the range of 0.05-3.5 pg/ml. Reaction was for 20 min at 30 "C. Variations from the above are noted. ."P-incorporation into protein was measured either by a modification (5) of the chromatographic method of Huang and Robinson (18) (assay 1) or by the following procedure (assay 2). At the end of the incubation time 10 pl of 10 mg/ml of bovine serum albumin were added, and duplicate 20-4 aliquots were spotted onto squares of Whatman 31-ET filter paper to which had previously been applied 30 pl of 20% (w/v) trichloroacetic acid containing I mM ATP and 2 mM potassium pyrophosphate. The filter paper squares were washed first for 15 min in 10% (w/v) trichloroacetic acid containing 5 mM potassium pyrophosphate and 0.25 mM ATP and then three times for 20 min each in 5% (w/v) trichloroacetic acid. After a final wash, 5 min, in diethyl ether, the papers were dried, placed in a scintillant of 0.5% (w/v) diphenyloxazole in toluene, and counted in a Beckman LS7500 scintillation counter. Protein kinase activity expressed as picomoles/ assay is referred to the whole 50-pl assay volume. We found assay 2 convenient for processing large numbers of reactions hut preferred assay 1 when the measurement of phosphorylation stoichiometry was of primary importance. Proteins-Glycogen synthase was purified from rabbit skeletal muscle as described previously (19). Phosphorylase from the same source was purified by a modification of the method of Fischer and Krebs (20) and was treated before use as described by DePaoli-Roach et RZ. (21). Phosphorylase kinase was purified from rabbit skeletal muscle, by the method of Cohen (22), as described previously (21). Calmodulin was purified to homogeneity from rabbit skeletal muscle by a variation of the method of Dedman et aZ. (23). A sample of rabbit liver phosphorylase was generously supplied by Dr. D. J. Graves, Iowa State University. Whole casein was kindly provided by Dr. E. W. Bingham of the Eastern Regional Research Center, Philadelphia. The 20,000-dalton smooth muscle myosin light chain, cardiac myosin light chains, and smooth muscle myosin light chain kinase were kindly supplied by Dr. David Hathaway, Indiana University. Phosvitin and histone (type IIA) were obtained from Sigma.
Purification of Calmodulin-stimulated Glycogen Synthase Kinase-Rabbit liver (300 g) was homogenized with 900 ml of a solution containing 4 mM EIITA, 250 mM sucrose, 1 mM (NHI),SO,, 0.1 mM TLCK, 0.5 mM phenylmethylsulfonyl fluoride, 250 pug/liter of leupeptin, and 1 mM dithiothreitol, at pH 7.0, and centrifuged a t 13,000 X g for 40 min. The supernatant was adjusted to pH 6.0 with 3.5 N acetic acid and centrifuged at 13,000 X g for 40 min. The precipitate was resuspended in buffer A (20 mM 1,4-piperazinediethanesulfonic acid, I mM EGTA, 1 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 0.1 KIM TLCK, 250 pg/liter of leupeptin, and 10% (v/v) glycerol, adjusted to pH 6.8) and centrifuged at 32,500 rpm in a Beckman 45Ti rotor for 90 min. The supernatant was applied to a column (2 X 20 cm) of phosphocellulose (Whatman P-11) and eluted with a gradient, 400 ml, formed of equal volumes of buffer A and buffer A plus 0.8 M NaCI. A peak of glycogen synthase kinase activity partially stimulated by calmodulin, eluting around 0.36 M NaC1, was pooled and made GO' % saturated with ammonium sulfate. The precipitate was dissolved in buffer A, dialyzed against 1 liter of buffer A for 2 h with one change, and applied to a column (2.6 X 82 cm) of Bio-Gel A-1.5m (Fig. 1) equilibrated with buffer A plus 0.2 M NaCI. The eluted calmodulin-stimulated glycogen synthase kinase fraction was made 6 mM in CaCL before being applied to a column (4 ml) of calmodulin-agarose. After washing with buffer A plus 0.2 M NaCl and 6 mM CaCL, the enzyme was eluted with buffer A plus 7 mM EGTA (Fig. 2). Further elution of the column with buffer A plus 7 mM EGTA a n d 0.2 M NaCl recovered a smaller proportion of protein kinase activity of much lower specific activity (Fig. 2). In fact, in earlier preparations, the calmodulin-agarose column was eluted directly with buffer A plus 7 mM EGTA and 0.2 M KC1,4 but the new procedure yielded greater purification at this stage. The smaller second peak of protein kinase activity (see Fig. 2) has not been extensively studied, and we do not know if it represents a distinct enzyme form with different calmodulin-binding properties. If taken through the purifi-' In our initial work, we utilized KC1 for the various chromatographic elutions hut have switched to NaCl because of the lower solubility of potassium dodecyl sulfate, a problem for polyacrylamide gel electrophoretic analysis in the presence of SI)fj. cation, however, enzyme of' similar polypeptide composition to the main fraction resulted. Active fractions from the calmodulin-agarose column were pooled and applied to a column (4 ml) of blue dextranagarose which was eluted with buffer A plus 0.5 M NaCl (Fig. 3). Pooled calmodulin-stimulated activity from the gel filtration step was applied in the presence of 6 mM Ca2+ to a column (4 ml) of calmodulin-agarose. The first arrow indicates the start of elution with buffer A plus 7 mM EGTA, and the second arrow indicates elution with buffer A plus 7 mM EGTA and 0.2 mM NaCI. T h e flow rate was 5 ml/h, and fractions of 1 ml were collected. Glycogen synthase kinase activity, in the presence of Ca2+ and calmodulin, was determined (assay 2) and is shown by the solid line. The AI,, is indicated by the dashed line. activity from the calmodulin-agarose column (Fig. 2) was applied to a column (4 ml) of blue dextran-agarose equilibrated with buffer A. After washing with 70 ml of buffer A, elution was initiated with buffer A plus 0.4 M NaCl (starting at fraction 1). The flow rate was 5 ml/h, and fractions of 1 ml were collected. Glycogen synthase kinase activity, in the presence of CaY+ and calmodulin, was determined using assay 2 and is shown hy the solid line. The A?,, is indicated by the dashed line.
A purification table is shown in Table   I. We must note that determination of protein at. the latter stages of the purification was very difficult because of the low amounts and concentrations of protein involved, and our values are not of the highest accuracy. Specific activities of the purified enzyme were estimated to he in the range of 100-900 nmol/min/mg. Attempts to concentrate the dilute fractions of the purified enzyme have so far led to significant losses of activity. Storage a t -70 "C or, with 50% (v/v) glycerol a t -20 "C, has been applied with some success although loss of activity with storage has been observed.
We note that earlier purification schemes had employed DEAEcellulose chromatography with which the enzyme eluted at 0.12 M KCl. The blue dextran-aragose chromatography has been substituted for this step.
Gel Electrophoresis-Polyacrylamide gel electrophoresis in the presence of SDS followed the method of Laemmli (24) with 0.75-mm thick slab gels and the indicated percentage of acrylamide. Gels were stained to visualize protein by the ultrasensitive silver method (25). Molecular weight standards were phosphorylase, bovine serum albumin, ovalbumin, and trypsinogen which were assigned subunit molecular weights of 95,000, 68,000, 45,000, and 24,000, respectively. Where necessary, autoradiograms were made by placing dried gels in contact with Cronex 4 x-ray film in Quanta 111 intensifying cassettes (DuPont) at -70 "C. Scanning of either dried gels or autoradiograms was achieved using a Beckman DU-8 spectrophotometer with a white light source.
Autophosphorylation of Calmodulin-dependent Protein Kinase-The protein kinase was incubated in a reaction mixture containing 15 mM 1,4-piperazinediethanesulfonic acid, 0.75 mM EGTA, and 30 mM Tris, pH 7.2, and, where indicated, 0.11 mM [y-'"P]ATP (906 cpm/ pmol) and 6 mM Mg'+. When present, Ca'+ and calmodulin were a t 1.1 mM and 10 pg/ml, respectively. At the desired times, aliquots of reaction mixture were removed and mixed with electrophoresis sample buffer to give a final concentration of 1% (w/v) SDS and heated for 4 min at 100 "C. These samples were later analyzed by polyacrylamide gel electrophoresis in the presence of SDS (see Fig. 7).

Other Materials and
Methods-Protein determination for the calmodulin-dependent protein kinase was achieved with Coomassiebinding assay (26). CNBr fragmentation and partial tryptic degradation of glycogen synthase was as described previously (5, 6). The concentration of [ y -'"PIATP was routinely determined by UV absorbance. Calmodulin-agarose was prepared as reported previously (27).
[y-12P]ATP was obtained from Amersham Corp., and trifluoperazine was from Smith, Kline and French.

RESULTS
Purification of Calmodulin-dependent Protein Kinase-The methods detailed under "Experimental Procedures" have allowed the extensive purification of the protein kinase, close to apparent homogeneity. The exact degree of purification is difficult to estimate because of the multiplicity of glycogen synthase kinases present at the early stages. In addition, calmodulin-stimulated activity was not readily detectable until partial purification had been achieved. Some degree of stimulation was usually observed in the 100,000 X g supernatant but only after phosphocellulose chromatography was a significant and reproducible stimulation found. Referring pu-rification to the calmodulin-dependent component of the phosphocellulose eluate gives a 500-fold further purification with 16% yield (Table I). If instead purification is referred simply to glycogen synthase kinase specific activity (in the presence of Ca'+ and calmodulin) of the crude extract, 20,000fold purification with 1.1% yield is obtained. The true values probably lie somewhere between these extremes.
The phosphocellulose eluate, only partially calmodulin-dependent, was resolved into two fractions of glycogen synthase kinase activity by the subsequent gel filtration step (Fig. I).
One of these fractions was totally dependent on Ca2' and calmodulin for activity and retained this property through further purification steps. The second protein kinase fraction was unaffected by Ca2+ and calmodulin. Work continues on this calmodulin-independent enzyme, but initial evidence suggests that it is a distinct protein kinase. The purification scheme of Table I has evolved from earlier trials, differing in two main ways. First, we initially eluted the enzyme from the calmodulin-agarose affinity column with an excess of EGTA in the presence of 0.2 M KCl.4 It was found though that the enzyme could be more specifically eluted with EGTA alone (Fig. 2 ) . Secondly, our first purifications used DEAE-cellulose chromatography instead of the blue dextran agarose step. It should be noted that the purification of the calmodulin-dependent protein kinase diverges from that of a set of liver Ca'+ and cyclic nucleotide independent glycogen synthase kinases ("PC" series enzymes; see Refs. 5 , 6 , and 9) at the stage of the pH 6.0 precipitation, the calmodulin-stimulated enzyme being found in the precipitate and the other protein kinases in the supernatant.
Molecular Weight and Polypeptide Composition-From calibration of the gel filtration column (Fig. 1) used in the purification (with thyroglobulin, aldolase, bovine serum albumin, and ovalbumin) an apparent molecular weight of approximately 500,000 was obtained. Sucrose density gradient sedimentation of the purified enzyme gave an apparent molecular weight of 275,000 (Fig. 4).
Analysis, by polyacrylamide gel electrophoresis in the presence of SDS, of enzyme eluted from the calmodulin-agarose column (Fig. 2 ) or the blue dextran-agarose column (Fig. 3) indicated the presence of two major polypeptide species visualized with the silver staining technique (Fig. 5). These polypeptides had apparent molecular weights of 51,000 and 53,000 and will be designated a and / 3, respectively. Other polypeptides are scarcely visible in Fig. 5 except for low molecular weight species running close to the tracking dye. This material has been variable in amount and was partially removed by the blue dextran column. The a and p polypeptides, in contrast, correlated strictly with enzyme activity in several chromatographic separations, besides those shown, and in several different enzyme purifications. Furthermore,  The main fraction, eluting with buffer A + 7 m EGTA (see Fig. 2). Gel electrophoretic analysis of calmodulin-dependent protein kinase. Fractions recovered from the calmodulin-agarose chromatography (Fig. 2) and blue dextran-agarose chromatography (Fig. 3 ) were analyzed by polyacrylamide gel electrophoresis in the presence of SDS and stained by the silver technique. A, shows the analysis (97 acrvlamide) of the equivalent of 15 p1 of fractions 5 to 13 (tracks I to 7, respectively) from the elution shown in Fig. 2. B, shows the analogous analysis (but with IOG: acrylamide) of the equivalent of 15 pI of fractions 6 to 12 (tracks I to 7. respectively) from the elution shown in Fig. 3. electrophoretic analysis of fractions from sucrose density gradient sedimentation (Fig. 4) showed a similar correspondence between enzyme activity and the (Y and /3 pobypeptides (Fig.   6). Densitometric scanning of gels stained by the silver technique gave a ratio of intensities (Y$ ranging from 1:1.08 to 1:1.2 or, normalizing for apparent molecular weights, molar ratios a:/3 of 1:1.04 to 1:1.13." It should be pointed out that it is not clear how valid such determinations with silver stained polypeptides are for estimating stoichiometry. Nonetheless, the a$ ratio so determined was constant through the last stages of the purification ( e . g Fig. 5) and between different preparations. We suggest that enzyme activity is associated with the (Y and /3 polypeptides.

Autophosphorylation-When the protein kinase was incu-
' ' Curiously, casual inspection of gels such as shown in Fig. 4 gave us the impression that the lower band (a-pol?rpeptide) was more intense but, upon closer scrutiny, this may be offset by the fact that the upper band (/3-polypeptide) is in fact slightly thicker. bated with [y-:"2P]ATP and Mg", in the presence of Ca" and calmodulin, phosphorylation of constituent polypeptides occurred (Fig. 7 ) . From gel electrophoretic analysis, "'I' was first observed with mobility close to that of the /I-polypeptide (i.e. apparent M , = 53,000) and later in a more slowly migrating species (apparent M , = 54,500). Radioactivity also appeared, somewhat more slowly, close to the dye front (see Fig. 7) and represented the phosphorylation of the low molecular weight contaminating material discussed above. Indeed, the extent of this latter phosphorylation varied for different preparations. By cutting gels, such as shown in Fig. 7, in the regions corresponding to the molecular weight range 53,000-54,500, and counting the associated radioactivity, it was estimated that up to approximately four phosphates per original (Y plus

Calmodulin-stimulated Glycogen
Synthase Phosphorylation p were incorporated (data not shown), a rather high level of phosphorylation. Concomitant with the phosphorylation reaction(s) was a significant alteration in the pattern of the silver staining material on the gel (Fig. 7). Already after 30 s, no polypeptide was seen with the mobility of the a-polypeptide and at longer times, the two main species visible had apparent molecular weights of 53,000 and 54,500, in correspondence with the ""Plabeling. AS shown in Fig. 7, no phosphorylation or alterations in electrophoretic mobilities occurred either when Ca'+ and calmodulin were omitted or when the enzyme was incubated without Ca", calmodulin, ATP, and Mg2'. In addition, incubation with Ca2+ and calmodulin, but without ATP and Mg", had no effect on electrophoretic mobilities (not shown). Using the standard assay, no significant change in activity accompanied the phosphorylation reaction (not shown). The changes in electrophoretic behavior correlated with the incorporation of "' P, but the exact sequence of events has yet to be worked out. For instance, we cannot distinguish whether only a or both LY and acquire altered mobility upon phosphorylation.
The reduction of the electrophoretic mobility of a polypeptide by phosphorylation may not be a general occurrence but neither is it a novel observation. A prominent example is the regulatory subunit of type I1 cyclic AMP-dependent protein kinase (28,29). The same phenomenon was recently described for the GSK-3 enzyme of Hemmings et al. (8) which also self-incorporated as many as four phosphates per subunit. Activation by Calmodulin-The enzyme was totally dependent on the presence of Ca2+ and calmodulin for activity also toward exogenous substrates and was completely inhibited by 50 PM trifluoperazine (Table 11) which is known to inhibit calmodulin-sensitive enzymes (30). The dependence of enzyme activity on calmodulin concentration (Fig. 8) indicated half-maximal stimulation at approximately 80 nM. Interaction with calmodulin was further indicated by the fact that, during the purification, the enzyme bound to a calmodulin-agarose affinity column in the presence of Ca2+ and was eluted with an excess of EGTA (Fig. 2).
Substrate Specifzcicity-The activity of the protein kinase toward a number of protein substrates is shown in Table 11.
The most significant substrate, other than skeletal muscle glycogen synthase, was the smooth muscle 20,000-dalton light chain. Glycogen synthase and myosin light chain kinase activities were also seen to co-elute from a DEAE-cellulose column and also from the calmodulin-agarose column (not shown). The calmodulin-stimulated glycogen synthase kinase was not, however, effective in phosphorylating cardiac myosin light chains (Table 11). This behavior contrasts with legitimate smooth muscle myosin light chain kinase which did not phosphorylate glycogen synthase but did phosphorylate cardiac light chain (Table  11). The glycogen synthase kinase also phosphorylated, at moderate initial rates, phosvitin and casein in a Ca')+-and calmodulin-dependent reaction (Table 11).
No phosphorylation of histone, either in the presence or absence of 10 PM cyclic AMP, was observed with the protein kinase. Heparin (0.25 pg/ml), a potent inhibitor of certain protein kinases (31, 32), had no effect on this calmodulindependent enzyme. An important negative result was the inability of the protein kinase to phosphorylate rabbit liver or skeletal muscle phosphorylase, under conditions where both substrates could be phosphorylated by muscle phosphorylase    (Table 111).
Nucleotides a n d M p -T h e depende,lce of protein kinase activity on ATP concentration (Fig. 9) was hyperbolic and was characterized by an apparent X, of 27 PM. With [y-"*P] ATP at 25 p~, 200 PM G T P (unlabeled) had no effect on activity and 1 mM G T P caused 33% inhibition. These results suggest that GTP is unlikely to be an effective substrate. Variation of the Mg" concentration indicated that approximately 0.75 mM was necessary for half-maximal activity (Fig.  9).
Phosphorylation of Glycogen Synthase-The calmodulindependent protein kinase phosphorylated glycogen synthase to a stoichiometry of greater than 1 phosphate/subunit (Figs. 10 and 11). The phosphorylation inactivated glycogen synthase. In Fig. 10, where a maximum level of approximately 1 phosphate/subunit was achieved, the percentage of I activity was decreased from 95% I to 45% I activity. From analysis of CNBr fragments of phosphorylated glycogen synthase, phosphate was shown to be introduced simultaneously into two CNBr fragments (Fig. 11) corresponding to the 12,000-dalton and 21,000-dalton fragments that have been described previously (5). Partial proteolysis with trypsin (as in Ref. 6) re- In an incubation similar to that of Fig. 5, aliquots of glycogen synthase were removed to determine total phosphorylation (assay 1; 0 ) or for fragmentation with CNBr, as described under "Experimental Procedures." CNBr fragments were separated by gel electrophoresis with 18% acrylamide. Autoradiograms were scanned with a Beckman DU-8 spectrophotometer to quantitate the relative amounts of radioactivity in the 21,000-dalton fragment (A) and the 12,000-dalton fragment (A) of the glycogen synthase. The phosphorylation, in moles of "P/mol of fragment, was then calculated and is plotted in the figure. A control reaction without added protein kinase was analyzed for total phosphorylation (0). moved approximately half of the phosphate introduced by the protein kinase and led to the loss, upon subsequent CNBr analysis, of the 21,000-dalton CNBr fragment (data not shown). Though further work on the phosphoryiation of glycogen synthase by this protein kinase is required, it is clear that the enzyme phosphorylated at least two distinct sites.

DISCUSSION
The present paper documents procedures for the extensive purification of a calmodulin-stimulated protein kinase from rabbit liver and some initial characterization of that enzyme (summarized in Table IV). The results confirm and extend the work of Payne and Soderling (12,13) which, at the time of writing, appears to be the only study of this protein kinase. Our data largely concur with those of the earlier reports (12, 13), suggesting that the same enzyme was studied. One difference is that, in our studies, the protein kinase did not display the extremely restricted substrate specificity reported previously. Even though probably not of physiological significance, the calmodulin-dependent phosphorylation of smooth muscle 20,000-dalton light chain, phosvitin and casein, is indicative of a somewhat broader substrate specificity for the protein ki-

nase.
The results presented provide a first insight into the subunit structure of the enzyme and indicate that activity is associated with two polypeptides of apparent molecular weights 51,000 and 53,000. Their relative abundance is probably not far from equimolar although our analyses, based on densitometric scanning of stained polyacrylamide gels, cannot be considered definitive. We can say that the relative proportions of the polypeptides did not vary in different preparations. Such stoichiometric information would argue that the two polypeptides are distinct subunits of the enzyme. We hesitate from fully endorsing this hypothesis, however, because we know that the electrophoretic mobility of one or of both of the polypeptides is diminished upon phosphorylation. Could a and , 8 be different phosphorylation states of the same poly- Inhibition by -Based on densitometric scanning of silver stained gels (see also text).
' Of course, this is based simply on those proteins so far tested. " M I 5 , concentration required for half-maximal activity.

Calmodulin-stimulated Glycogen
Synthase Phosphorylation peptide? Initial attempts to dephosphorylate the native enzyme have not altered the electrophoretic pattern (data not shown). Nonetheless, we believe that it is prudent to reserve final judgment on whether or not the enzyme consists of two chemically unrelated polypeptides but, from the determinations of native molecular weight, we can assert with some confidence that the enzyme is an oligomer composed of probably six or more subunits. Although a number of calmodulin-dependent phosphorylation reactions have been described (see, for example, Refs. [33][34][35][36][37][38], only two other calmodulin-stimulated protein kinases are well characterized enzymes. One of these, myosin light chain kinase, by its ability to phosphorylate cardiac light chains and by its inability to phosphorylate glycogen synthase, is clearly distinguishable from the enzyme of this report. Likewise, the liver calmodulin-dependent glycogen synthase kinase, from its inability to phosphorylate muscle or liver glycogen phosphorylase, appears distinct from phosphorylase kinase. The information on polypeptide composition reinforces the above conclusions. We would suggest, then, that the enzyme described here is a representative of a third class of calmodulin-stimulated protein kinase that is approaching definition at the molecular level. Phosphorylation of rabbit muscle glycogen synthase by the calmodulin-dependent kinase involves the introduction of phosphate, at roughly equivalent rates, into both an NH2terminal 12,000-dalton CNBr fragment (containing serine-7, the site of action of phosphorylase kinase (11,39,40) and PCo4 (1,5)) and a COOH-terminal 21,000-dalton CNBr fragment (containing the primary sites of action of cyclic AMP-dependent protein kinase (1, 5) and PCo.; (1,6)). At least two sites on glycogen synthase are, therefore, phosphorylated. Such site specificity is different from that of other glycogen synthase kinases that we have studied. Indeed, a variety of properties distinguishes the enzyme of this report from other known glycogen synthase kinases (reviewed in Refs. 1-4), the most prominent characteristic being its stimulation by Ca'+ and calmodulin. Nonetheless, there is one intriguing property shared with the GSK-3 enzyme of Hemmings et al. (8), an enzyme that appears identical with the protein-activating factor, FA, of the ATP-Mg"-dependent protein phosphatase system of Merlevede's group (41, 42). Hemmings et al. (8) noted that GSK-3 was a monomer with subunit molecular weight 51,000 as judged by gel electrophoresis. However, GSK-3 underwent autophosphorylation to at least four phosphates/subunit with an accompanying decrease in electrophoretic mobility of the subunit, a property very reminiscent of the enzyme of this report. Whether these common properties reflect a deeper similarity between the two protein kinases is an interesting possibility that will be tested.
While progress is being made on the molecular definition of the calmodulin-dependent glycogen synthase kinase, its physiological role has yet to be established. Of course, the possibility that the enzyme could be an element in Ca"-mediated controls of glycogen synthase activity in liver has been an important motivation for this and the earlier study of the enzyme. The criticism may be raised, however, that muscle glycogen synthase has been used as a substrate in all the work so far, as actually has been the case in most studies of nonmuscle glycogen synthase kinases. Though a degree of similarity in properties almost certainly exists between liver and muscle glycogen synthases, one cannot at present predict how far the two enzymes compare, in detail, as protein kinase substrates. Work is under way to address this broader question as well as to ascertain specifically whether the calmodulindependent protein kinase is able to phosphorylate liver gly-cogen synthase. Such study may help assess the possible role of the enzyme in the regulation of glycogen metabolism. For now, the present report provides an important first view of some structural features that will aid comparison of this liver calmodulin-dependent enzyme both with other calmodulindependent protein kinases and with other glycogen synthase kinases.