Calmodulin-dependent multifunctional protein kinase Evidence for isoenzyme forms in mammalian tissues

Transcription

1 Eur. J. Biochem. 161, (1986) 0 FEBS 1986 Calmodulin-dependent multifunctional protein kinase Evidence for isoenzyme forms in mammalian tissues Shirish SHENOLIKAR ', Ron LICKTEIG', D. Grahame HARDIE4, Thomas R. SODERLING', Rochelle M. HANLEY and Paul T. KELLY' Departments of Pharmacology, Neurobiology and Anatomy and Internal Medicine, University of Texas Medical School at Houston Department of Biochemistry, University of Dundee Department of Molecular Physiology and Biophysics, Vanderbilt University School of Medicine, Nashville, Tennessee (Received July 22/September 8, 1986) - EJB Calcium/calmodulin-dependent multifunctional protein kinases, extensively purified from rat brain (with apparent molecular mass 640 kda), rabbit liver (300 kda) and rabbit skeletal muscle (700 kda), were analysed for their structural, immunological, and enzymatic properties. The immunological cross-reactivity with affinitypurified polyclonal antibodies to the 50-kDa catalytic subunit of the brain calmodulin-dependent protein kinase confirmed the presence of common antigenic determinants in all subunits of the protein kinases. One-dimensional phosphopeptide patterns, obtained by digestion of the autophosphorylated protein kinases with S. aureus V8 protease, and two-dimensional fingerprints of the '251-labelled proteins digested with a combination of trypsin and chymotrypsin, revealed a close similarity between the two subunits (51 kda and 53 kda) of the liver enzyme. Similar identity was observed between the 56-kDa and/or 58-kDa polypeptides of the skeletal muscle calmodulindependent protein kinase. The data suggest that the subunits of the liver and muscle protein kinases may be derived by partial proteolysis or by autophosphorylation. The peptide patterns for the 50-kDa and 60-kDa subunits of the brain enzyme confirmed that the two catalytic subunits represented distinct protein products. The comparison of the phosphopeptide maps and the two-dimensional peptide fingerprints, indicated considerable structural homology among the 50-kDa and 60-kDa subunits of the brain calmodulin-dependent protein kinase and the liver and muscle polypeptides. However, a significant number of unique peptides in the liver 51-kDa subunit, skeletal muscle 56-kDa, and the brain 50-kDa and 60-kDa polypeptides were observed and suggest the existence of isoenzyme forms. All calmodulin-dependent protein kinases rapidly phosphorylated synapsin I with a stoichiometry of 3-5 mol phosphate/mol protein. The two-dimensional separation of phosphopeptides obtained by tryptic/chymotryptic digestion of 32P-labelled synapsin I indicated that the same peptides were phosphorylated by all the calmodulin-dependent protein kinases. Such data represent the first structural and immunological comparison of the liver calmodulin-dependent protein kinase with the enzymes isolated from brain and skeletal muscle. The findings indicate the presence of a family of highly conserved calmodulin-dependent multifunctional protein kinases, with similar structural, immunological and enzymatic properties. The individual catalytic subunits appear to represent the expression of distinct protein products or isoenzymes which are selectively expressed in mammalian tissues. Protein phosphorylation is widely recognised as a major mechanism by which hormones elicit control of cellular processes leading to a variety of physiological responses [l]. The modification of cellular proteins is mediated by specific protein kinases following changes in the intracellular concentration of selected effector molecules or 'second messengers'. In this regard the ubiquitous presence of cyclic-amp-dependent protein kinase in mammalian tissues has been viewed as support for the thesis that this enzyme mediates most, if not Correspondence to S. Shenolikar, Department of Pharmacology, University of Texas Medical School at Houston, 6431, Fannin, P. 0. Box 20708, Houston, Texas, USA Abbreviations. CaM, calmodulin; SDS-PAGE, polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulphate. Enzymes. Glycogen synthase kinase (Ca2 +/CaM-dependent protein kinase) (EC ), Staphylococnrs aureus V8 proteinase (EC ); trypsin (EC ); chymotrypsin (EC ). all, of the actions of hormones which increase cellular cyclic AMP levels [2]. In comparison, less is known of the role of protein kinases whose activity is dependent on calcium ions. It is predicted that many hormones and neurotransmitters, which elicit increases in intracellular calcium, would also modulate the activity of protein kinases regulated by calmodulin, a high-affinity calcium-binding protein. Two such calmodulin-dependent protein kinases, namely phosphorylase kinase and myosin light chain kinase, have been extensively studied. Phosphorylase kinase has been implicated in the coordinated hormonal and neural control of glycogenolysis in skeletal muscle [3]. Myosin light chain kinase plays a central role in the control of contraction in smooth muscle [4]. Both of these Ca''/calmodulin-dependent protein kinases demonstrate a rather restricted substrate specificity. In contrast, a calmodulin-dependent multifunctional protein kinase has

2 740 been identified in a variety of tissues, including brain [5 - lo], liver [ll- 131, skeletal muscle [14, 151, pancreas [16], mammary glands [17] and adrenal chromaffin cells [18]. The broad substrate specificity demonstrated by this group of protein kinases has led to the suggestion that calmodulindependent protein kinases may play important roles in the control of such diverse processes as microtubule assembly [19], neurotransmitter synthesis [20], neuronal function [21, 221 and glycogenesis [134, 14, 231. Recent studies have suggested the presence of tissuespecific forms of calmodulin-dependent protein kinase, which differ in their subunit composition but show a remarkable similarity in their relative specificity towards a variety of protein substrates [ Although some structural (i. e. autophosphorylated peptide patterns) and immunological comparison between the skeletal muscle and brain enzymes has been undertaken, data on the relationship of these protein kinases to calmodulin-dependent protein kinases from other tissues (e. g. liver) has not been available. Therefore, we have undertaken the structural, immunological and enzymatic analysis of calmodulin-dependent protein kinases, extensively purified from rat forebrain, rabbit liver and rabbit skeletal muscle, to characterise the nature of the differences in enzymes from different tissue sources. The data presented supports the concept of isoenzyme forms of calmodulin-dependent protein kinases, which though closely related in structure and function, may result from the selective expression of distinct gene products in mammalian tissues. MATERIALS AND METHODS Proteins Calmodulin-dependent protein kinases were purified from rat brain [27], rabbit liver [13] and rabbit skeletal muscle [15], omitting the final chromatography step on DEAEcellulose which primarily eliminated contaminating phosphorylase kinase. Calmodulin was isolated from bovine brain [28] or from rabbit skeletal muscle [29]. Bovine brain calmodulin was also purchased from Calbiochem. Synapsin I from bovine brain was gift from Drs Teresa McGuiness and Paul Greengard, Rockefeller University. Enzyme assay Autophosphorylation of the protein kinases was undertaken at 37 C in 10mM Hepes buffer ph 7.2 containing calmodulin (50 nm), 1.O mm CaCl,, 5 mm MgCl,, 0.02 mm [y-32p]atp(2.0 x lo6 cpm/pmol) and the respective protein kinase (0.5 pg/loo pl). Calcium-independent protein phosphorylation was examined in companion assays replacing CaC1, with EGTA (2 mm), and omitting calmodulin. The time-dependent phosphorylation of calmodulin-dependent protein kinases was followed by the incorporation of [32P]phosphate into subunits of the enzymes separated by 7-16% polyacrylamide gradient gel electrophoresis in the presence of sodium dodecyl sulphate (SDS-PAGE), according to Laemmli [30]. Gels were stained either by the silver stain procedure of Oakley et al. [31] or with Coomassie blue, dried and autoradiographed on Kodak XAR-5 film. Phosphorylase b (Mr 94000), bovine serum albumin (MI 66500), ovalbumin (Mr 40000), carbonic anhydrase (MI 31000), soy bean trypsin inhibitor (Mr ) and lysozyme (MI 14000) were used as marker proteins of standard molecular mass. Phosphorylation of synapsin I was carried out in a similar manner using 50 pg/ml substrate, calmodulin-dependent protein kinase (approximately 5 pg/ml), calmodulin (50 nm), [y-32p]atp (0.2 mm), and MgCl, (2.0 mm) in the presence and absence of CaCl, (1 mm) (assays were also carried out in the presence and absence of calmodulin to establish the calmodulin dependance of the enzyme). Enzyme reactions, incubated at 30 C for 1,2 and 5 min, were terminated by the addition of 4 x SDS-sample buffer and subsequent heating at 70 "C for 5 min. The phosphorylated proteins were analysed by SDS-PAGE. Phosphorylation of synapsin I was linear for the initial 60 s of the reaction. Detection of calmodulin-binding proteins Calmodulin was iodinated using 251-labelled Bolton and Hunter reagent (Amersham) according to the procedure described by Bolton and Hunter [32]. The binding of radiolabelled calmodulin to proteins separated by SDS-PAGE was carried out by the gel overlay procedure according to Glenney and Weber [33]. The dried gels were autoradiographed using Kodak XAR-5 film in cassettes with tungsten phosphate intensifier screens. Immunological analysis Polyclonal antibodies against the purified rat forebrain calmodulin-dependent protein kinase (CaM kinase 11) were generated in rabbits (Kelly, P. T., unpublished results). The antibodies were affinity-purified by a modification of the procedure of Renart et al. [34] using the 50-kDa subunit of CdM kinase I1 electrophoretically transferred on to nitrocellulose. Western immunoblot analysis of the isolated skeletal muscle, liver, and brain calmodulin-dependent protein kinases was carried out by the method of Towbin et al. [35] using the affinity-purified antibodies. Immune complexes obtained on nitrocellulose were detected using 251-labelled affinitypurified protein A (specific activity > 30 Ci/g protein) and subsequent autoradiography. [32P]peptide analysis The phosphopeptide maps of the autophosphorylated CaM kinase I1 were obtained by digestion with Staphylococcus aureus V8 protease prior to separation on 10-23% gradient polyacrylamide gels in the presence of SDS [36]. Two-dimensional phosphopeptide maps of synapsin I digested with trypsin and chymotrypsin were undertaken in the following manner. Phosphorylated synapsin I (2 pg protein) was subjected to SDS-PAGE on 7-16% gradient slab gels. The protein bands corresponding to synapsin I (both I a and Ib) were excised. The gel slice was incubated with trypsin (50 pg/ml) and chymotrypsin (50 pg/ml) at 30 C for 10 h. A second aliquot of trypsin/chymotrypsin was added and the incubation continued overnight. 32P-labelled peptide fragments thus extracted were subjected to two-dimensional highvoltage electrophoresis on thin-layer cellulose plates at ph 3.5 in the first dimension (1 kv for 3.5 h) and in the second dimension at ph 1.9 (1 kv for 1 h) prior to autoradiography [37l. Two-dimensional peptide fingerprints of calmodulin-dependen t protein kinases The purified protein kinases (2 pg protein) were separated by SDS-PAGE on 7-16% gradient slab gels. Protein bands

3 741 Fig. 1. Comparison of calmodulin-dependent protein kinases from brain, liver andskeletal muscle. (A) The purified protein kinases (approximately 0.5pg total protein) were separated on a 7-16% polyacrylamide slab gel as described in Materials and Methods, and stained for protein with silver stain: lane 1, brain CaM kinase 11; lane 2, liver enzyme; lane 3, skeletal muscle enzyme. (B) Autophosphorylation of the calmodulindependent protein kinases was carried out as described. Phosphoproteins were separated on SDS-PAGE and visualised by autoradiography. The autophosphorylation of brain (lanes 4 and 5), liver (lanes 6 and 7) and skeletal muscle (lanes 8 and 9) calmodulin-dependent protein kinases was carried out in the absence and presence of + CaZ /calmodulin respectively. (C) Calmodulin-binding proteins in the protein kinase preparations were detected by the gel overlay procedure (see Materials and Methods) using * 1-labelled calmodulin, and subsequent autoradiography: lane 10, brain; lane 11, liver; lane 12, skeletal muscle stained with Coomassie blue were excised, and iodinated with radiolabelled iodine using the chloramine-t reaction. The iodinated gel slice was incubated with trypsin (50 pg/ml) and chymotrypsin (50 pg/ml) overnight at 30 C. 251-labelled peptides were subjected to two-dimensional electrophoresis and chromatography on thin-layer cellulose plates [38], followed by autoradiography. RESULTS Purification of calmodulin-dependent protein kinases from brain, liver and skeletal muscle The calmodulin-dependent protein kinases were purified to near homogeneity from rat forebrain, rabbit liver and rabbit skeletal muscle (Fig. 1 A). Comparison of the three protein kinases by SDS-PAGE indicated distinct subunit compositions. The rat brain enzyme, which was purified to a purity greater than 85% and demonstrated a native molecular mass of 740 kda, was composed of two distinct subunits of apparent molecular mass 50 kda and 60 kda (lane 1). The 60-kDa polypeptide was frequently resolved into a doublet of 58 kda and 60 kda. These have been shown to be closely related to each other, as judged by two-dimensional peptide fingerprints, and may represent covalently modified forms of the same protein (results not shown). The 50-kDa and 60- kda subunits of the brain CaM kinase I1 were present at an approximate ratio of 4: 1. The liver calmodulin-dependent protein kinase with a native molecular mass of 300 kda [26] was judged to be approximately 70% pure and was composed of equimolar amounts of two subunits of 51 kda and 53 kda (lane 2). The calmodulin-dependent protein kinase from rabbit skeletal muscle was somewhat less pure (approximately 55%) as judged by SDS-PAGE and comprised a major protein band(@ of 56/58 kda (lane 3). The skeletal muscle calmodulin-dependent protein kinase displayed an apparent native molecular mass of 700 kda by sedimentation equilibrium analysis. The calmodulin-dependent protein kinases from all three tissues, like most other protein kinases, demonstrated the capacity to autophosphorylate. In this respect each protein kinase preparation was rapidly phosphorylated in the presence of [p3 P]ATP and Ca +/calmodulin (Fig. 1 B). Both subunits of the brain and the liver enzymes incorporated 32P label at approximately equal rates, supporting the view that both subunits possessed catalytic properties. A polypeptide (M, 15 kda), migrating slightly faster than calmodulin, frequently contaminated the liver protein kinase preparation. This peptide was phosphorylated in a CaZ +/calmodulin-dependent manner but was not recognised by the antibody to brain calmodulin-dependent protein kinase nor bound calmodulin and may be unrelated to the catalytic subunits. The skeletal muscle preparation incorporated [32P]phosphate primarily into the 56/58-kDa protein band indicating this to be the catalytic subunit. Ca2+/calmodulin-dependent phosphorylation of two proteins of 140 kda and 125 kda in the skeletal muscle preparation was also noted. These may represent minor contamination with phosphorylase kinase. No

4 742 Fig. 2. Phosphopeptide mapping of autophosphorylated calmodulin-dependent protein kinases. Autophosphorylated enzymes from brain, liver and skeletal muscle (2 pg protein each) were separated by SDS- PAGE. The 32P-labelled bands (with relative molecular mass indicated) were excised and digested with S. aureus V8 protease (5 pg/ slice), prior to separation on a 10-23% polyacrylamide SDS-PAGE as outlined in Materials and Methods Fig. 3. Immunological comparison of calmodulin-dependent protein kinases from brain, liver and skeletal muscle. The purified protein kinases were separated by SDS-PAGE and electrophoretically transferred to nitrocellulose (0.22 pm) sheets. The transferred proteins were reacted with affinity-purified polyclonal antibodies (0.05 pg IgG/ml) to 50-kDa subunit of the brain CaM kinase 11, followed by lz5ilabelled protein A (see Materials and Methods). Lane 1 represents brain (0.4 pg protein); lane 2 is skeletal muscle (4 pg); lane 3 is liver (2 pg); lane 4 is brain (4 pg); lane 5 is the Ca2+/calmodulin-dependent cyclic nucleotide phosphodiesterase purified from bovine brain (2 pg protein); and lane 6 represents calcineurin from bovine brain (2 pg protein). significant phosphorylation of phosphorylase b by the skeletal muscle protein kinase preparation was measured, suggesting either trace contamination or the presence of relatively inactive phosphorylase kinase. The major polypeptides in each kinase preparation, which exhibited prominent autophosphorylation, also bound lz5ilabelled calmodulin in a calcium-dependent manner when analysed by the gel overlay procedure (Fig. 1 C). None of the proteins bound 'Z51-labelled calmodulin in the presence of 5 mm EGTA (data not shown). The skeletal muscle enzyme preparation occasionally possessed a minor '2sI-calmodulinbinding component of 52 kda, which was neither autophosphorylated nor recognised by the antibody to the brain CaM kinase 11, implying that it was unrelated to the catalytic species. The subunits of the putative phosphorylase kinase, which contaminated the muscle preparation, failed to bind '251-calmodulin in the presence or absence of calcium. Phosphopeptide analysis of autophosphorylated calmodulin-dependent protein kinases The protein kinases from brain, liver and skeletal muscle were rapidly autophosphorylated in the presence of calcium (1 mm) and calmodulin (50 nm). No incorporation of [32P]phosphate was observed with the inclusion of 2 mm EGTA. Partial digestions of the autophosphorylated subunits of the three calmodulin-dependent protein kinases were undertaken using S. aureus V8 protease (Fig. 2). The 50-kDa and 60-kDa subunits of brain CaM kinase I1 revealed distinct phosphopeptide patterns, as has been previously reported [21, 221. In contrast, the 51-kDa and 53-kDa subunits of the liver protein kinase produced indistinguishable phosphopeptide maps, suggesting a close relationship between these two polypeptides (data not shown). More importantly, the phosphopeptide pattern obtained from the 51-kDa liver enzyme closely resembled that obtained with the 50-kDa subunit of the brain enzyme (Fig. 2). The autophosphorylated muscle enzyme was composed of a broad 32P-labelled band in the 56/58-kDa region (lane 9, Fig. 1). Phosphopeptide maps obtained with gel slices containing either the 56-kDa or 58- kda muscle phosphoproteins were very similar (Fig. 2). Minor differences in the phosphopeptides observed could represent partial proteolysis or the presence of endogenous phosphate in the purified protein kinase. The 52-kDa protein, present in the muscle preparation that bound '2sI-calmodulin (Fig. 1, lane 12), was not phosphorylated under basal or Ca2 +/calmodulin-stimulated conditions and also failed to bind the affinity-purified antibody (Fig. 3, lane 2) thus appeared unrelated to the calmodulin-dependent protein kinase polypeptides. Immunoblot analysis of calmodulin-dependent protein kinases The calmodulin-dependent protein kinases isolated from different tissues were examined by Western immunoblot analysis using affinity-purified antibodies against the 50-kDa subunit of rat forebrain CaM kinase 11. These antibodies have been extensively utilised in the analysis of brain CaM kinase I1 (Kelly et al., unpublished results). The affinity-purified antibodies recognised both 50-kDa and 60-kDa subunits of the brain enzyme, confirming previous findings that these

5 143 Fig. 4. Two-dimensionalpeptide maps of calmodulin-dependent protein kinases. The purified protein kinases were first subjected to SDS-PAGE, and stained with Coomassie blue to visualise the subunits. Protein bands corresponding to individual subunits of the protein kinases (identified in kda) were excised and iodinated by the chloramine T procedure using "'1. The labelled slices were digested with trypsin/chymotrypsin (see Materials and Methods) before separation of the '251-peptides by two-dimensional electrophoresis/chromatography. The figure demonstrates autoradiographs of the peptides obtained with the 50-kDa (C) and 60-kDa (D) subunits of the rat forebrain CaM kinase 11, the 51-kDa subunit of rabbit liver enzyme (A) and the 56-kDa subunit of rabbit skeletal muscle enzyme (B). Landmarks (designated by +, -, 00, and *) represent some of the common determinants shared by all subunits. Unique peptides are designated by open arrowheads (A). Electrophoresis (first dimension) was carried out from left to right and chromatography (second dimension) was from bottom to top subunits demonstrated considerable molecular and immunological similarities [21, 24, 251. A variety of Western blot analyses were carried out in which the protein loading ( pg protein/lane) or antibody concentration ( pg IgG/ml) was varied. A representative Western blot that contained each kinase at approximately equivalent protein loading is shown in Fig. 3. Muscle 56/58-kDa proteins and the liver 51-kDa and 53-kDa subunits reacted equally well with affinity-purified antibodies when compared to the two subunits of brain CaM kinase 11. No other immunoreactive polypeptides were observed in the kDa molecular mass range. Using different preparations of the affhitypurified anti-(50-kda) antibody, the brain CaM kinase I1 could be detected at 5-10-ng levels (data not shown). Two other calmodulin-dependent enzymes, namely calcineurin and the high-k, cyclic AMP phosphodiesterase, both of which possess catalytic subunits in the 60-kDa range and which could potentially contaminate the calmodulin-dependent protein kinase preparations, did not cross-react with affinitypurified anti-(cam kinase 11) even at high protein loadings (e, g. 2 pg protein). Additional studies with two-dimensional peptide mapping of purified calcineurin or cyclic AMP phosphodiesterase (provided by Dr. Randall Kincaid, National Institutes of Health) confirmed the absence of these calmodulin-dependent enzymes in any of the calmodulin-dependent protein kinase preparations (data not shown). Two-dimensional pep tide mapping of calmodulin-dependent protein kinases Comparison of the primary structures of 1251-labelled subunits of different calmodulin-dependent protein kinases was achieved by digestion of electrophoretically purified polypeptides with a combination of trypsin and chymotrypsin, followed by separation of '251-peptides by twodimensional electrophoresis and chromatography. Two-di- mensional fingerprints of the immunoreactive and autophosphorylated kinase polypeptides were prepared and included: (a) 50-kDa versus 58/60-kDa subunits of the brain CaM kinase 11, (b) 51-kDa versus 53-kDa polypeptides of the liver kinase, and (c) kDa polypeptides of skeletal muscle calmodulin-dependent protein kinase. Individual l2 51- peptide fingerprints of most of these polypeptides are shown in Fig. 4. The '251-labelled peptides (i. e. predominantly tyrosine-containing peptides), resolved from each kinase subunit, numbered between 24 and 28. Approximately half of these lz5i-peptides produced prominent autoradiographic spots and most likely correspond to the predicted 13 tyrosinecontaining peptides of CaM kinase I1 based on reported amino acid composition The 58-kDa and 60-kDa polypeptides that are frequently resolved from brain CaM kinase I1 [8,9,21,27l, when mapped individually or as a mixture, produced identical two-dimensional fingerprints (data not shown). Likewise, peptide maps of either 51-kDa or 53-kDa subunits of the liver enzyme were identical (data not shown) and suggested that they may be derived by a covalent modification of each other. The skeletal muscle protein kinase, which was often resolved as two or three closely spaced bands on SDS-PAGE (Fig. 1, lane 3), was divided into three molecular mass regions (i. e. 56/57, 57/ 58, 58/59 kda) and fingerprinted separately. Identical twodimensional maps were obtained from each gel band and suggested that partial proteolysis (and/or autophosphorylation) may have occurred during purification of the skeletal muscle enzyme (results not shown). Similar protein staining patterns for the muscle kinase with kDa polypeptides were reported by McGuiness et al. [25]. As a further check of purity, the subunits of brain CaM kinase I1 were separated by two-dimensional PAGE combining isoelectrofocusing followed by SDS-PAGE. Subsequent peptide mapping of the 1261-labelled 50-kDa and 60-kDa subunits revealed that fingerprints obtained by this

6 744 Fig. 5. Phosphopeptide maps of synapsin Iphosphorylated by calmodulin-dependent protein kinases from brain, liver and skeletal muscle. Synapsin I (2 pg protein) was phosphorylated by purified protein kinase (0.2 pg) for 2 min at 30 C in the presence of [p3 P]ATP and Ca +/calmodulin (see Materials and Methods). Protein bands containing electrophoretically purified synapsin I (Ia and Ib) were exised and digested with trypsin/chymotrypsin. Resulting phosphopeptides were separated by two-dimensional electrophoresis on thin-layer cellulose plates. The figure represents autoradiographs of 32P-labelled synapsin I phosphopeptides obtained using the liver (A), skeletal muscle (B) and brain (C) calmodulin-dependent protein kinases. The inset demonstrates the apparent microheterogeneity of phosphopeptide 4 (marked 4a, 4 b, and 4c) achieved by extended electrophoresis (1.0 kv h) in the second dimension at ph 1.9. The origin is marked by the symbol Or procedure were virtually identical to those obtained with polypeptides separated by one-dimensional SDS-PAGE (data not shown). Although at this level of analysis the individual maps of each protein kinase polypeptide appeared similar, we undertook a systematic analysis of the common 251-peptides by mixing experiments using combinations of peptide digests from selected subunits to establish unambiguously the comigration of individual peptides; approximately 30 such mixing experiments were carried out (data not shown). Results from these analyses demonstrated that each kinase subunit contained 1 or 2 distinct peptides that distinguished it from the other kinases (designated by open triangles, Fig. 4). More importantly, by a peptide-to-peptide comparison all kinase subunits (i. e. brain 50-kDa and 60-kDa, liver 51-kDa or 53-kDa, and muscle 56/58-kDa) displayed homology greater than 85% (21 peptides) based on peptide comigration. By rank order there was (a) greatest homology between liver and muscle proteins, (b) intermediate homology between the brain 60-kDa polypeptide and the liver or muscle enzymes, and (c) lowest homology between brain 50-kDa and 60-kDa polypeptides. Substrate specificity of calmodulin-dependent protein kinases Calmodulin-dependent protein kinases from brain, liver and skeletal muscle were used to investigate site-specific phosphorylation using synapsin I as substrate. All protein kinases phosphorylated synapsin I to achieve a stoichiometry of 3-5 mol phosphate/mol synapsin I during 2-5 rnin incubation. When subjected to limited proteolysis by S. aureus V8 protease (2.5 pg protease per [32P]synapsin I band), more than 90% of the [32P]phosphate incorporated into synapsin I by any of the three protein kinases was associated with a 28-kDa phosphopeptide or the upper molecular mass phosphopeptide (results not shown; see [39] for nomenclature). To obtain more detailed analysis of phosphorylation sites, complete digestion of 32P-labelled synapsin I was undertaken by the simultaneous action of trypsin and chymotrypsin followed by separation of phosphopeptides by two-dimensional electrophoresis. Synapsin I a and I b were analysed together as well as individually, but showed no significant differences in either the stoichiometry of [32P]phosphate incorporation or the phosphopeptides labelled (characterised by their electrophoretic mobility at two different ph values). Two-dimensional separations identified the phosphorylation of six prominent sites on synapsin I by each of the calmodulin-dependent protein kinases (Fig. 5). Huttner et al. [39] have used similar proteolytic conditions and a different two-dimensional separation scheme to resolve six major phosphopeptides in rat brain synapsin I following phosphorylation by CaM kinase 11. Tryptic peptide mapping of synapsin I either from rat brain [40] or from bovine brain [25] has demonstrated differences in the migration of 32P-labelled peptides by electrophoresis (PH 3.5)/chromatography, suggesting a difference in the primary structure of synapsin I from the two species. However, these studies identified between five and eight sites of phosphorylation on either rat or bovine synapsin I after phosphorylation with rat brain CaM kinase 11. During the current investigation, when synapsin I [32P]phosphopeptides were subject to extended electrophoresis in the second dimension at ph 1.9 (i. e. 2 h at 1 kv), apparent microheterogeneity of peptide 4 was observed (Fig. 5, inset), This heterogeneity was present in synapsin I phosphorylated by each calmodulin-dependent kinase, and no significant difference in the pattern produced with any of the kinases was observed. With the exception of phos-

7 745 phopeptide 6, the two-dimensional phosphopeptide fingerprints of synapsin I were virtually identical with the three protein kinases. Although peptide 6 only incorporated detectable levels of [32P]phosphate when phosphorylated for 5 min, its extent ofphosphorylation was greatest with the liver enzyme and the lowest with the muscle calmodulin-dependent protein kinase. Some minor differences in the relative rates of phosphorylation of specific sites on synapsin Ia and Ib were observed with individual protein kinases, but no differences in the number or the mobility of phosphopeptides were noted. Such small differences in the phosphorylation of synapsin I a and I b were also reported by Huttner et al. [38]. DISCUSSION Changes in cytoplasmic calcium concentrations resulting from influx of extracellular calcium ions or mobilisation of intracellular stores in response to physiological stimuli have been demonstrated in a variety of biological settings (review [41]). The calcium-dependent increase in cellular protein phosphorylation, which follows, could play an important role in mediating a variety of cellular responses [42]. Initially two calmodulin-dependent protein kinases, namely myosin light chain kinase and phosphorylase kinase, were identified, but were shown to display a restricted substrate specificity. The presence of other calcium-dependent protein kinases was postulated to account for the diversity of cellular protein phosphorylation. In this respect, the discovery of 'calmodulindependent multifunctional protein kinases' in a variety of tissues provided potential candidates capable of mediating the actions of Ca2 + on cellular protein phosphorylation. A number of studies have isolated calmodulin-dependent protein kinases from mammalian tissues. However, widely differing molecular sizes for the holoenzymes from brain [9, 25,431, skeletal muscle [15] and liver [ll] were demonstrated. Differences in composition and size of the catalytic subunits of these protein kinases were also noted. Despite this, the calmodulin-dependent protein kinases showed very similar rates of phosphorylation in vitro using such substrates as glycogen synthase, synapsin I, smooth muscle myosin light chain, MAP-2, phospholamban, phenylalanine hydroxylase and tyrosine/tryptophan 5-monooxygenase [ Recent work comparing the calmodulin-dependent glycogen synthase kinase from skeletal muscle with calmodulin-dependent protein kinase isolated from brain, using either synapsin I [25] or tryptophan 5-monooxygenase as substrate [24], provided evidence of close similarity between the enzymes isolated from different tissues (and from different species). The ability of monoclonal antibodies, derived against the rat brain enzyme, to immunotitrate or immunoblot the rabbit skeletal muscle 'glycogen synthase kinase' was viewed as additional evidence of similarity between these enzymes. In this context we have examined the structural, immunological and enzymatic (with special emphasis on the specific sites modified using a multiply phosphorylated protein substrate, such as synapsin I) properties of calmodulin-dependent protein kinases purified from rabbit skeletal muscle, rabbit liver and rat brain. The isolation of these protein kinases reaffirmed the molecular properties observed by other laboratories, namely that the brain enzyme displayed a molecular mass of kda, composed of two subunits of 50 kda and 60 kda at a ratio of 4: 1. The liver enzyme, which appeared somewhat smaller (i. e. 300 kda), displayed an equimolar content of 51-kDa and 53-kDa polypeptides. The calmodulin-dependent protein kinase from skeletal muscle (700 kda) was composed of a single class of subunits, which varied in apparent size from 56 kda to 58 kda. The kinase 'subunits' all bound '251-calmodulin in the presence of calcium ions and rapidly autophosphorylated in a Ca2+/ calmodulin-dependent manner. The inability to demonstrate 32P incorporation into these components in the absence of Ca2 +/calmodulin provided confirmatory evidence of their identity as calmodulin-dependent protein kinases. Comparison of phosphopeptide maps of the individual subunits suggested that the multiple polypeptides in skeletal muscle or liver preparations represent closely related proteins. Even in the brain, the 58/60-kDa doublet demonstrated similar phosphopeptide maps, indicating that the two polypeptides may arise from minor proteolytic cleavage. Alteration in electrophoretic mobility of all the autophosphorylated catalytic subunits was also correlated with the incorporation of phosphate. To what extent the multiple bands were generated by autophosphorylation or proteolysis remains unanswered. However, the 50-kDa and 60-kDa subunits of the brain CaM kinase I1 clearly represent related but distinct protein products. The stoichiometry of catalytic components of the rat forebrain enzyme remained constant throughout purification of the protein kinase by a variety of chromatographic procedures (data not shown) and argues against the copurification of two distinct enzyme species. The unusual ratio of 4: 1 of the 50-kDa and 60-kDa proteins has been observed by many investigators and remains unexplained. In contrast to findings of McGuiness et al. [25], the phosphopeptide maps of the autophosphorylated subunits, digested with S. aureus V8 protease with a specificity for glutamic acid residues, demonstrated a close similarity between both the muscle and liver enzymes with the major 50-kDa polypeptide of the brain enzyme. This structural similarity between the three calmodulin-dependent protein kinases was further strengthened by the immunological crossreactivity of all the enzymes with affinity-purified polyclonal antibodies against the 50-kDa subunit of brain CaM kinase 11. (Detailed characterisation of the antibody indicated the recognition of multiple epitopes on the 50-kDa and 60-kDa subunits of the brain enzyme; Paul T. Kelly, unpublished observation.) The immunological data provide somewhat different information from the phosphopeptide fingerprints of the autophosphorylated subunits, suggesting considerable homology around the antibody-recognition sites in all polypeptide constituents of these calmodulin-dependent protein kinases. The two-dimensional peptide fingerprints of '251-labelled polypeptides, obtained by the simultaneous digestion by trypsin and chymotrypsin (with a specificity for basic and aromatic residues respectively) were first utilised to demonstrate that the two subunits of the brain CaM kinase I1 represented distinct proteins [21]. The comparison of up to 28 peptide fragments provided a detailed assessment of the degree of similarity between the two catalytic subunits. The presence of approximately 23 peptides common to the 50- kda and 60-kDa subunits provided an explanation for their immunological cross-reactivity. At the same time, unique peptides were observed in both CaM kinase I1 subunits and suggested distinct regions of primary structure as indicated by the very different phosphopeptide maps of the autophosphorylated 50-kDa and 60-kDa subunits [44]. The two-dimensional lz5i-peptide patterns for the liver (51/ 53 kda), skeletal muscle (56/58 kda) and brain (58/60 kda) catalytic subunits were more closely related, lacking all the peptides unique to the 50-kDa subunit of brain CaM kinase

8 However, a few differences between these polypeptides were also observed, strongly suggesting discrete isoenzyme forms. Although structural dissimilarities, arising from species differences, could not be ruled out, preliminary studies comparing forebrain CaM kinases from rat, rabbit, sheep, bovine and human sources have suggested that CaM kinase I1 is highly conserved across species [45] (P. T. Kelly, unpublished observations). The data obtained led to four major conclusions. (a) The multiple polypeptides present in the skeletal muscle and liver calmodulindependent protein kinase preparations probably arise through proteolysis and/ or partial phosphorylation. (Extensive studies using a wide variety of protease inhibitors during purification of the liver enzyme failed to alter the stoichiometry of the 51-kDa and 53-kDa subunits.) (b) The brain enzyme was unique among the three tissues in possessing an oligomeric complex of two distinct subunits, 50 kda and 60 kda. (c) Liver calmodulindependent protein kinase represented a hexamer of 51/53-kDa subunits and the skeletal muscle enzyme was a dodecamer of 56/58-kDa subunits. In contrast, the brain CaM kinase I1 possessed a complex of eleven or twelve 50-kDa and three 60- kda polypeptides. (d) Considerable homology in the primary structure existed between all the polypeptide constituents of the three calmodulin-dependent protein kinases. Our overall conclusion from the peptide mapping suggested that the calmodulin-dependent protein kinases from each tissue source represent homologous but distinct gene products and that the major brain enzyme species (50-kDa subunit) differs more than the 60-kDa subunit when compared to the kinase subunits from other tissues. Finally, comparison of specific sites phosphorylated in synapsin I, the best substrate available for calmodulin-dependent multifunctional protein kinases, established the close similarity in substrate recognition by the three calmodulindependent protein kinases and is consistent with previous findings using glycogen synthase and other protein substrates [25, 261. These results lend support to previous findings comparing the relative rates of phosphorylation of a variety of protein substrates in vitro, which suggested a similar substrate specificity for the calmodulin-dependent protein kinases from brain, liver [26] and skeletal muscle [24]. In addition, the current data indicate that the same serine residues in synapsin I, one of the best in vitro substrates known for the calmodulindependent multifunctional protein base, were phosphorylated by all the calmodulin-dependent protein kinases. In summary, glycogen synthase kinases, isolated from skeletal muscle and liver, and the CaM kinase I1 from forebrain represent a family of closely related protein kinases with highly conserved immunological, enzymatic and molecular properties. These findings warrant their classification as isoenzyme forms of a common Ca +(CaM)-dependent protein kinase 11. CaM kinase I1 has been identified as the major protein of postsynaptic density in mammalian brain [8, 10, 211, where it may play a key role in synapse structure and function. Control of liver microsomal Ca2 +/ATPase activity by a membranebound calmodulin-dependent protein kinase has been reported [46]. The phosphorylation of the sarcolemmal protein, phospholamban, by a membrane-bound calmodulin-dependent protein kinase also plays an important role in regulating Caz+ flux across cardiac membranes [47]. Ca2+/ATPase in the skeletal muscle sarcoplasmic reticulum is also regulated by an endogenous Ca +/calrnodulin-dependent protein kinase [48]. The structural relationship between these soluble and membrane-bound calmodulin-dependent protein kinases awaits future investigation. In brain the structural properties of the CaM kinase I1 associated with cytosolic and postsynaptic density are virtually identical [8, 211. However, the stoichiometry of the 50-kDa and 60-kDa subunits of CaM kinase I1 appears to vary in different regions of the brain, e. g. forebrain versus cerebellum [49, 501. Recent studies also suggest that the subunit composition of CaM kinase I1 is altered during neuronal differentiation from predominantly 60-kDa subunit in neonates to a holoenzyme with an increased content of the 50-kDa subunit in the adult forebrain [51] (Kelly et al., unpublished results). Taken together, such data point to the presence of isoenzymic forms, which are differentially expressed in mammalian tissues. The broad substrate specificity of this group of protein kinases, together with their specific subcellular distribution further argues for an important role for calmodulin-dependent multifunctional protein kinases in the control of a wide variety of calciumregulated processes. The authors would like to express their appreciation for the expert technical assistance provided by Jeffery Langston, and to Kitty Goldstein for typing the manuscript. Samples of the calmodulindependent protein kinase purified from rabbit liver were provided by Charles M. Schworer, Department of Physiology, Vanderbilt University School of Medicine, and from rabbit skeletal muscle by David Carling, Department of Biochemistry, University of Dundee. This work was supported by a feasibility grant from the American Diabetes Association and a Biomedical Research Support Grant from the University oftexas School (S. S.); United States Public Health Service grant NS and National Science Foundation grant BNS (P. T. K.); National Institutes of Health grant AM17808 (T. R. S.); Paul T. Kelly is also a recipient of Research Career Developement Award NS-00605; Rochelle M. Hanley is a recipient of a National Institutes of Health Physician-Scientist Award AM01374; D. Grahame Hardie is supported by research group support and project grants from the Medical Research Council (UK) and by British Heart Association. REFERENCES 1. Cohen, P. (1982) Nature (Lond.) 296, Krebs, E. G. (1972) Curr. Top. Cell Regul. 5, Cohen, P., Klee, C. B., Picton, C. & Shenolikar, S. (1980) Ann. N. Y. Acad. Sci. 356, Kamm, K. E. & Stull, J. T. (1984) Annu. Rev. Pharmacol. Toxicol. 25, Yamauchi, T. & Fujisawa, H. (1980) FEBSLett. 116, Miyamoto, E., Fukunaga, K., Matsui, K. & Iwasa, Y. (1981) J. Neurochem. 37, Fukunaga, K., Yamamoto, H., Matsui, K., Higashi, K. & Miyamoto, E. (1982) J. Neurochem. 39, Kennedy, M. D., McGuiness, T. & Greengard, P. (1983) J. Neurosci. 3, Bennett, M. K., Erondu, N. E. & Kennedy, M. B. (1983) J. Biol. Chem. 258, Goldenring, J. R., Gonzales, B., McGuire, J. S., & DeLorenzo, R. J. (1983) J. Biol. Chem. 258, Payne, M. E. & Soderling, T. R. (1980) J. Biol. Chem. 255, Ahmad, Z., De Paoli-Roach, A. A. & Roach, P. J. (1982) J. Biol. Chem. 257, Payue, M. E., Schworer, C. M. & Soderling, T. R. (1983) J. Biol. Chem. 258, Woodgett, J. R., Tonks, N. K. & Cohen, P. (1982) FEBS Lett. 148, Woodgett, J. R., Davison, M. T. & Cohen, P. (1983) Eur. J. Biochem. 136, Gorelick, F. S., Cob, J. A., Freedman, S. D., Delahunt, N., Gershoni, J. M. & Jamieson, J. D. (1983) J. Cell Biol. 97,

9 Brooks, C. L. & Landt, M. (1985) Arch. Biochem. Biophys. 240, Waymire, J. C., Haycock, J. W. & George, R. J. (1984) Neurology and neurobiology, vol. 8 A Catecholamines: Basic andperipheral mechanisms (Usdin, E., Carlsson, A., Dahlstrom, A. & Engel, J., eds) pp , Alan R. Liss, New York. 19. Goldenring, J. R., Vallano, M. L. & DeLorenzo, R. L. (1985) J. Neurochem. 45, Yamauchi, T. & Fujisawa, H. (1983) Eur. J. Biochem. 132, Kelly, P. T., McGuiness, T. L. & Greengard, P. (1984) Proc. Nut1 Acad. Sci. USA 81, Kelly, P. T., Yip, R., Shields, S. & Hay, M. (1985) J. Neurochem. 45, Camici, M., Ahmad, Z., DePaoli-Roach, A. A. & Roach, P. J. (1984) J. Biol. Chem. 259, Woodgett, J. R., Cohen, P., Yamauchi, T. & Fujisawa, H. (1984) FEBS Lett. 170, McGuiness, T. L., Lai, Y., Greengard, P., Woodgett, - J. R. & Cohen, P. (1983) FEBS Lett. 163y Schworer, C. M., McClure, R. W. & Soderling, T. R. (1985) Arch. Biochem. Biophys. 242, Kelly, P. T. & Shenolikar, S. (1986) Methods Enzymol. 139, Wang, J. & Desai, R. (1977) J. Biol. Chem. 252, Shenolikar, S., Cohen, P. T. K., Cohen, P., Nairn, A. C. & Perry, S. V. (1979) Eur. J. Biochem. 100, Laemmli, U. K. (1970) Nature (Lond.) 227, Oakley, B. R., Kirsch, D. R. & Moms, N. R. (1980) Anal. Biochem. 105, Bolton, A. E. & Hunter, W. M. (1973) Biochem. J. 133, Glenney, J. R. Jr. & Weber, K. (1980) J. Biol. Chem. 255, Renart, J., Reiser, J. & Stark, G. R. (1979) Proc. Natl Acad. Sci. USA 76, Towbin, H., Staehelin, T. & Gordon, J. (1979) Proc. Natl Acad. Sci. USA 76, Cleveland, D. W., Fischer, S. G., Kirschner, M. W. & Laemmli U. K. (1977) J. Biol. Chem.252, Shenolikar, S. & Cohen, P. (1978) FEBS Lett. 86, Elder, J. H., Pickett, R. A., Hampton, J. & Lerner, R. A. (1977) J. Biol. Chem Huttner, W. B., DeGennaro, L. J. & Greengard, P. (1981) J. Biol. Chem. 256, Huttner, W. B. & Greengard, P. (1979) Proc. Natl Acad. Sci. USA 76, Exton, J. H. (1985) Am. J. Physiol. 248, Garrison, J. C., Johnson, D. E. &Campanile, C. P. (1984) J. Biol. Chem. 259, Kuret, J. & Schulman, H. (1984) Biochemistry 23, Shields, S. M., Vernon, P. J. & Kelly, P. T. (1984) J. Neurochem. 43, Rostas, J., Kelly, P., Pesin, R. & Cotman, C. (1979) Brain Res. 168, Famulski, K. S. & Carafoli, E. (1984) Eur. J. Biochem. 141, Lindemann, J. P. & Watanabe, A. M. (1985) J. Biol. Chem. 260, Tuna, B. S. & MacLennan, D. H. (1984) J. Biol. Chem. 259, Miller, S. G. & Kennedy, M. B. (1985) J. Biol. Chem. 260, McGuiness, T. L., Lai, Y. & Greengard, P. (1985) J. Biol. Chem. 260, Kelly, P. T. & Vernon, P. (1985) Dev. Brain Res. 18,