The regulatory region of calcium/calmodulin-dependent protein kinase I contains closely associated autoinhibitory and calmodulin-binding domains.

The mechanism for the regulation of Ca2+/calmodulin-dependent protein kinase I (CaM kinase I) was investigated using a series of COOH-terminal truncated mutants. These mutants were expressed in bacteria as fusion proteins with glutathione S-transferase and purified by affinity chromatography using glutathione Sepharose 4B. A mutant (residues 1-332) showed complete Ca2+/CaM-dependent activity. Truncation mutants (residues 1-321, 1-314, and 1-309) exhibited decreasing affinities for Ca2+/CaM and also exhibited decreasing Ca2+/CaM-dependent activities. Truncation mutants (residues 1-305 or 1-299) were unable to bind Ca2+/CaM and were inactive. In contrast, truncation mutants (residues 1-293 or 1-277) were constitutively active at a slightly higher level (2-fold) than fully active CaM kinase I. These results indicate the location of the Ca2+/CaM-binding domain on CaM kinase I (residues 294-321) and predict the existence of an autoinhibitory domain near, or overlapping, the Ca2+/CaM-binding domain. These conclusions were supported by studies which showed that a synthetic peptide (CaM kinase I(294-321)) corresponding to residues 294-321 of CaM kinase I inhibited the fully active kinase in a manner that was competitive with Ca2+/CaM and also inhibited the constitutively active mutant (residues 1-293) in a manner that was competitive with Syntide-2, a peptide substrate, (Ki = 1.2 μM) but was non-competitive with ATP. Thus, these results suggest that CaM kinase I is regulated through an intrasteric mechanism common to other members of the family of Ca2+/CaM-dependent protein kinases.

tributed Ca 2ϩ receptor, calmodulin (CaM) 1 (for reviews, see Manalan and Klee, 1984). Strong evidence, in turn, indicates that the effects of Ca 2ϩ /CaM are often achieved through the regulation of protein phosphorylation (for reviews, see Nairn et al., 1985;Hanson and Schulman, 1992;Nairn and Picciotto, 1994). A family of Ca 2ϩ /CaM-dependent protein kinases has been identified: phosphorylase kinase, myosin light chain kinase, and EF-2 kinase (CaM kinase III) that are highly specific enzymes, while CaM kinases II and IV are multifunctional enzymes. CaM kinase I was first purified from bovine brain based on its ability to phosphorylate site 1 of the neuronal protein synapsin I (Nairn and Greengard, 1987). Since then the enzyme has been found to be expressed in both neuronal and non-neuronal tissues (Ito et al., 1994a;Picciotto et al., 1995), thus necessitating a re-evaluation of the kinase as a potential multifunctional kinase. In this respect, two other good substrates for CaM kinase I have been identified in vitro: CREB, a cAMP-response element-binding protein (Sheng et al., 1991), and CF-2, a portion of the R-domain of the cystic fibrosis transmembrane conductance regulator (Picciotto et al., 1992).
CaM kinase I is active as a monomer (Nairn and Greengard, 1987), although various molecular masses of 37-43 kDa have been assigned to the enzyme. A preparation from bovine brain, purified using synapsin I as a substrate, was found to consist of two major polypeptides of M r 37,000 and 39,000 and a minor polypeptide of M r 42,000 (Nairn and Greengard, 1987). Two other Ca 2ϩ /CaM-dependent protein kinases, termed CaM kinases Ia (M r 43,000) and Ib (M r 39,000), have been purified from rat brain using a synthetic peptide based on site 1 of synapsin I as a substrate (DeRemer et al., 1992a(DeRemer et al., , 1992b. Another preparation, termed CaM kinase V (M r 41,000), has been purified from rat brain using the synthetic peptide, Syntide-2, as a substrate and its partial amino acid sequences determined (Mochizuki et al., 1993). The full-length cDNA for rat CaM kinase I has been cloned and a M r of 41,643 deduced from the amino acid sequence (Picciotto et al., 1993;Cho et al., 1994). Based on a high identity of amino acid sequences, CaM kinase V appears to be closely related to, though distinct from * This work was supported by a grant-in-aid for the encouragement of young scientists and by a grant-in-aid for scientific research from the Ministry of Education, Science, and Culture, Japan (1993,1994). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
‡ Recipient of a long term fellowship from Japan Society for the Promotion of Science.
CaM kinase I (Picciotto et al., 1993). These results, together with recent immunological studies of CaM kinase I and V (Ito et al., 1994a(Ito et al., , 1994bPicciotto et al., 1995), suggest the possibility that CaM kinase I, Ia, Ib, and V form a family of related isoforms.
Comparison of the deduced amino acid sequence of CaM kinase I with that of the ␣ subunit of rat brain CaM kinase II revealed a high identity (approximately 42%) throughout the catalytic domain (Picciotto et al., 1993). In addition, a second domain was identified near the COOH terminus of CaM kinase I that showed limited identity (approximately 32%) to the Ca 2ϩ /CaM-binding domain of the CaM kinase II subunit. In the present study, the structural basis for the Ca 2ϩ /CaM-dependent regulation of the enzyme has been established. The results obtained indicate that CaM kinase I is regulated through an intrasteric mechanism common to other members of the family of Ca 2ϩ /CaM-dependent protein kinases.
CaM was purified from bovine brain as described (Endo et al., 1981). Smooth muscle myosin light chain kinase (smMLCK) was purified from chicken gizzard as described (Adelstein and Klee, 1981). The catalytic subunit of cAMP-dependent protein kinase (PKA) was purified from bovine heart as described (Beavo et al., 1974). Ca 2ϩ -activated phospholipid-dependent protein kinase was partially purified from rabbit brain essentially as described (Inagaki et al., 1987).
The peptides, Syntide-2, and smooth muscle myosin light chain (smMLC) peptide were purchased from Peptide Institute Inc. (Osaka, Japan). The amino acid sequence of the smMLC peptide, KKRAAR-ATSNVFA, corresponds to that surrounding the site phosphorylated in smMLC by smMLCK, except that the residues, AA, replaced for the original residues, PQ (Pearson et al., 1982). The peptide, CaM kinase I(294 -321), based on the amino acid sequence of residues 294 -321 of rat CaM kinase I (IKKNFAKSKWKQAFNRTAVVRHMRKLQL) was synthesized with a model 431A peptide synthesizer (Applied Biosystems Inc.). CaM kinase I(294 -321) was purified by preparative reversed-phase HPLC (C-18). Aliquots of CaM kinase I(294 -321) were subjected to analytical reversed-phase HPLC (C-18) and Pico-tag amino acid analysis (Waters Associates). The peptide was further characterized by amino acid sequencing with a model 473A protein sequencer (Applied Biosystems Inc.) (data not shown).
Protein Kinase Inhibitors,and Ca 2ϩ /CaM Antagonist, were synthesized by the method of Hidaka et al. (1984).
Oligo-1 s. was made with a BamHI site added upstream of the position 1 ("start codon") of a cDNA encoding rat CaM kinase I. Oligo-277a. s., Oligo-293a. s., Oligo-299a. s., Oligo-305a. s., Oligo-309a. s., Oligo-314a. s., Oligo-321a. s., and Oligo-332a. s. correspond to complements of the coding strand with an EcoRI site included upstream of the positions 831, 879,897,915,927,942,963, and 998, respectively. Oligo-1 s. and Oligo-277a. s., Oligo-1 s. and Oligo-293a. s., Oligo-1 s. and Oligo-299a. s., Oligo-1 s. and Oligo-305a. s., Oligo-1 s. and Oligo-309a. s., Oligo-1 s. and Oligo-314a. s., Oligo-1 s. and Oligo-321a. s., and Oligo-1 s. and Oligo-332a. s. were used to amplify 855-, 903-, 921-, 939-, 951-, 966-, 987-, and 1022-base pair DNA fragments, respectively, from 5 ng of the cDNA using Taq DNA polymerase (Promega). The conditions for polymerase chain reaction (PCR) were 25 cycles of the following: 94°C for 1 min, 65°C for 2 min, and 72°C for 2 min. The amplified DNAs were extracted with phenol-chloroform, ethanol-precipitated, and digested with BamHI and EcoRI. The purified fragments were subcloned into the BamHI and EcoRI sites of pGex-2T (Pharmacia P-L Biochemicals Inc.) using Ligation Pack (NIPPON GENE Ltd., Tokyo, Japan) according to the manufacturer's instructions. The sequences of the DNA products encoding mutant 1-332 and mutant 1-293 were determined on both strands using Sequenase Ver. 2.0 (U. S. Biochemical Corp.) and found to be correct. The complete sequences of the other DNAs produced by PCR were not determined. While it is possible that errors in DNA sequences were introduced based on the error rate of Taq DNA polymerase (2 ϫ 10 Ϫ4 /base pair, Saiki et al., 1988), the fact that the molecular weights of the protein products of the other mutants were as expected, and the fact that many of the mutants expressed activity suggested that no critical nonsense mutation was introduced by the PCR technique. Recombinants carrying the insert of interest were isolated using JM109 as a bacterial host and induced essentially as described (Maruta et al., 1991). Briefly, 500 ml of LB plus 50 g/ml ampicillin was inoculated with a 50-ml overnight culture of the clone of interest and allowed to grow to OD 650 of 0.8 at 25°C (approximately 2 h necessary for the growth). IPTG was added to a final concentration of 0.1 mM and the culture allowed to grow overnight. Cells were harvested by centrifugation at 4,000 ϫ g for 10 min and resuspended in 50 ml of lysis buffer containing 50 mM Tris (pH 7.5), 25% saccharose, 2 mM dithiothreitol, and 50 g/ml leupeptin. All further steps were carried out at 0 -4°C. Nonidet P-40 and MgCl 2 were added to a final concentration of 0.5% and 5 mM, respectively, and cells lysed by sonication using first a 1-min step at mid-power followed by four 1-min steps at high power. Debris was pelleted by centrifugation at 12,000 ϫ g for 20 min. The supernatant was recovered and applied onto a column of glutathione Sepharose 4B (10 ml wet volume) pre-equilibrated with wash buffer containing 20 mM Tris (pH 7.5), 2 mM MgCl 2 , 1 mM ␤-mercaptoethanol, and 100 M phenylmethylsulfonyl fluoride. The column was washed extensively with wash buffer and eluted with elution buffer containing 50 mM Tris (pH 8.8), and 5 mM glutathione. 3-ml fractions were collected in tubes containing 1 ml of 200 mM Tris (pH 7.5) for pH neutralization. Protein concentrations were determined by the Bradford assay (1976) and fractions containing peak protein values were pooled (approximately 12 ml in all cases), dialyzed against wash buffer, adjusted with glycerol to a final concentration of 20% and stored in aliquots at Ϫ70°C. CaM kinase I fusion proteins were essentially homogeneous as judged by staining of SDS-polyacrylamide gels with Coomassie Brilliant Blue and the protein concentration of the pure protein was used to determine the specific activity (see below).
Radioiodination of CaM and 125 I-CaM Gel Overlay-CaM (200 g) was treated with 125 I-Bolton-Hunter reagent (18.5 MBq) in 10 l of 0.1 M sodium borate (pH 8.5) on ice for 30 min. The reaction was stopped by addition of 500 l of 0.1 M sodium borate (pH 8.5), and 0.2 M glycine and placed on ice for 10 min. 125 I-CaM was separated from unreacted 125 I by passing through a column of Sephadex G-25M (Pharmacia LKB Biotechnology Inc.) pre-equilibrated with 25 mM HEPES (pH 7.5), and 0.25% gelatin. Fractions of 500 l were collected. The first peak fractions of the radioactivity were pooled, adjusted with BSA to a final concentration of 0.1%, and stored in aliquots at Ϫ20°C. 125 I-CaM gel overlay was performed essentially as described (Glenney and Weber, 1980). Briefly, the gels were fixed with 10% acetic acid and 40% methanol for 30 min and rinsed with distilled water three times each for 10 min. The gels were renatured with the following: 10% ethanol overnight, 0.1 M imidazole-HCl (pH 7.0) for 2 h, and 20 mM imidazole-HCl (pH 7.0), 0.2 M KCl, 0.1% BSA, 0.02% sodium azide, plus either 1 mM EGTA or 1 mM CaCl 2 for 2 h. 125 I-CaM was added to the buffers (5 ϫ 10 4 cpm/ml) and the incubation carried out at 4°C for 24 h. The gels were washed with the corresponding EGTA-and CaCl 2 -containing buffer at 4°C for 2 days and exposed, without drying, to x-ray film with an intensifying screen at Ϫ70°C overnight. The film was developed with a FPM800A film processor (Fuji medical systems Ltd., Tokyo, Japan).
CaM-coupled Affinity Chromatography-CaM was coupled to a column of NHS-activated Sepharose HP according to the manufacturer's instructions. The column was attached to a fast protein liquid chromatography system (Pharmacia LKB Biotechnology Inc.) driven at a flow rate of 1 ml/min and equilibrated with 25 mM Tris (pH 7.5) and 0.5 mM CaCl 2 . Aliquots (0.3-0.4 mg) of recombinant kinases were adjusted to 1 mM CaCl 2 , applied onto the column, and washed with the 0.5 mM CaCl 2 -containing buffer for 20 min. The proteins were eluted with 25 mM Tris (pH 7.5) and 1 mM EGTA. Protein in the eluate was monitored by absorbance at 280 nm.
Protein Kinase Assays-Recombinant kinases were assayed in a final volume of 50 l, containing 35 mM HEPES (pH 7.5), 10 mM MgCl 2 , 1 mM dithiothreitol, 0.01% Tween-20, 50 M Syntide-2, plus either 1 mM EGTA or a set of 1 mM EGTA, 1.5 mM CaCl 2 , and 1.5 M CaM. smMLCK was assayed in a final volume of 50 l, containing 35 mM HEPES (pH 7.5), 10 mM MgCl 2 , 1 mg/ml BSA, 0.1% Tween-80, 50 M smMLC peptide, plus either 1 mM EGTA or a set of 1 mM EGTA, 1.5 mM CaCl 2 , and 1 M CaM as described (Kemp et al., 1987). Reactions were initiated by addition of [␥-32 P]ATP to a final concentration of 100 M (20 cpm/ pmol) and the mixtures incubated at 30°C for 10 min. Aliquots of recombinant kinases and smMLCK were added so that the assay demonstrated a linear phosphate incorporation into the substrate over 10 min. 32 P incorporation into the substrate was determined by pipetteting a 25-l aliquot of the mixture onto P-81 paper (23 mm in diameter) (Roskowski, 1985). The paper was washed three times each for 10 min with 75 mM phosphoric acid, dried, and subjected to quantification of Cerenkov radiation using a model LS5000 liquid scintillation counter (Beckman Instruments Inc.). PKA was routinely assayed in a standard mixture, in a final volume of 200 l, containing 25 mM Tris (pH 7.0), 10 mM MgCl 2 , 2 mM EGTA, and 0.2 mg/ml histone VII-S as described (Hidaka et al., 1984). Protein kinase C was routinely assayed in a standard mixture, in a final volume of 200 l, containing 25 mM Tris (pH 7.0), 10 mM MgCl 2 , 0.2 mg/ml histone III-S, either 1 mM EGTA or a set of 1 mM EGTA, 1.5 mM CaCl 2 , and 50 g/ml PS as described (Inagaki et al., 1987). Reactions were initiated by addition of [␥-32 P]ATP to a final concentration of 10 M (150 cpm/pmol) and the mixtures incubated at 37°C for 5 min. The reactions were stopped by addition of 1 ml of 20% trichloroacetic acid followed by addition of 100 l of 1 mg/ml BSA as a carrier protein. The mixtures were centrifuged at 3,000 revolutions/min for 10 min. The pellets were washed with 5% trichloroacetic acid, the centrifugation repeated, and the pellet dissolved in 1 ml of 1 N NaOH prior to quantification of Cerenkov radiation.
Other Procedures-SDS-PAGE was performed according to the method of Laemmli (1970). Gels were stained with Coomassie Brilliant Blue (0.25% in 45% methanol, and 10% acetic acid) followed by destaining in 10% methanol and 10% acetic acid. Gels were dried onto cellophane paper in vacuo. Apparent molecular weights on SDS-PAGE were determined using SDS-PAGE molecular standards (Bio-Rad). Protein was determined by the Bradford assay (1976) using BSA as a standard.

Construction and Expression of a Series of COOH-terminal
Truncated Mutants-Analysis of the deduced amino acid sequence of CaM kinase I has suggested that the catalytic domain is included within residues 1-270 (Picciotto et al., 1993). In addition, a second domain showing moderate identity to the Ca 2ϩ /CaM-binding domains of other Ca 2ϩ /CaM-binding proteins was found near the COOH terminus of CaM kinase I (for example, Ϸ32% identity to CaM kinase II). To directly identify the location of the regulatory region containing the Ca 2ϩ /CaMbinding domain, a series of COOH-terminal truncated mutants were generated. The sense oligonucleotide, Oligo-1 s., corresponding to the 5Ј end of the coding region and one of several antisense oligonucleotides corresponding to different parts of the 3Ј end of the coding region were used to amplify DNA using PCR. The DNAs encoding a series of COOH-terminal truncated mutants were subcloned into pGex-2T. The clones were induced by addition of IPTG and expressed in Escherichia coli as GST-fusion proteins (Fig. 1A). The resulting products were purified by affinity chromatography on a column of glutathione Sepharose 4B. Transformants with plasmids carrying the mutants 1-293 or 1-277 displayed inhibited growth following induction with IPTG (data not shown). Thus, the yield of mutants 1-293 and 1-277 were significantly decreased.
All preparations were purified to apparent homogeneity as demonstrated by staining of SDS-polyacrylamide gels using

FIG. 1. Construction and expression of a series of COOH-terminal truncated mutants of CaM kinase I as GST-fusion proteins.
A, a series of DNAs was amplified from a cDNA encoding rat CaM kinase I using PCR as described under "Experimental Procedures." The DNA fragments were subcloned into pGEX-2T and expressed in E. coli as GST-fusion proteins. Mutants 1-314 and 1-309 have an Arg residue resulting from a sequencing error instead of the correct Ala residue at position 309. A highly conserved catalytic domain (open box residues 1-270) and a non-conserved regulatory domain (shadowed box residues 271-332) are shown. B, aliquots (0.7-2.4 g of protein/lane) of recombinant kinases purified by affinity chromatography on a column of glutathione Sepharose 4B as described under "Experimental Procedures" were analyzed by SDS-PAGE (6% acrylamide) and staining with Coomassie Brilliant Blue. The top and bottom of the gel were removed to highlight the differences in the mobility of the mutants.
Coomassie Brilliant Blue (Fig. 1B). Mutant 1-332 exhibited a molecular mass of 62 kDa, consistent with the predicted value of 37 kDa of rat CaM kinase I plus 26 kDa of GST. A proportional increase in mobility on SDS-PAGE was observed with removal of increasing amounts of the COOH-terminal region (i.e. mutants 1-321, 1-314, 1-309, 1-305, and 1-299, Fig. 1B). In contrast, mutant 1-293 migrated slower than mutant 1-299 although residues 294 -299 had been deleted. In addition, mutant 1-277, which was expected to migrate with the fastest mobility, co-migrated with mutant 1-299. The basis for the anomalous migrations of mutant 1-293 and 1-277 is not known. Previous studies of mitogen-activated protein kinase (Matsuda et al., 1992) and cdc25-C (Hoffmann et al., 1993) have indicated that phosphorylation by their respective activating kinases resulted in a decreased mobility on SDS-PAGE. Recent studies have indicated that CaM kinase I is also phosphorylated at an active site Thr residue by one or more kinases (Sugita et al., 1994), and that this results in a slower mobility on SDS-PAGE. 2 However, there is no evidence that any of the recombinant proteins were phosphorylated during preparation.
Ca 2ϩ -dependent CaM Binding and Ca 2ϩ /CaM Independent and Dependent Activities of CaM Kinase I Mutants-125 I-CaM gel overlay showed that mutant 1-332 bound CaM in a Ca 2ϩdependent manner ( Fig. 2A). Mutants 1-321 and 1-309 displayed significantly lower Ca 2ϩ -dependent CaM binding. Mutant 1-314 and all other shorter mutants failed to bind Ca 2ϩ / CaM using the overlay technique. The abilities of recombinant kinases to bind Ca 2ϩ /CaM were also determined using CaM-Sepharose affinity chromatography (Fig. 2B). Mutants 1-332 and 1-321 bound to the column in the presence of Ca 2ϩ and were eluted with EGTA, a Ca 2ϩ -chelating agent. In contrast to the gel overlay method, mutant 1-314 bound to CaM-Sepharose though a small amount of protein was not absorbed to the column in the presence of Ca 2ϩ . The majority of mutant 1-309 protein was not absorbed to the column though a small amount of the protein did bind and was eluted with EGTA. Mutants 1-305 exhibited properties essentially the same as mutant 1-309 (data not shown). Three other shorter mutants did not bind to the column, indicating that they did not bind Ca 2ϩ /CaM.
Wild-type recombinant CaM kinase I was completely dependent on Ca 2ϩ /CaM (Fig. 2C), as previously demonstrated (Picciotto et al., 1993). Using Syntide-2 as a substrate, its specific activity was 73.8 Ϯ 7.6 nmol/min/mg. The activity of recombinant CaM kinase I is significantly lower than the fully active form of the enzyme that requires phosphorylation by an activating kinase (Sugita et al., 1994). However, recent studies have indicated that the substrate specificity and CaM dependence of recombinant CaM kinase I are the same as the fully activated enzyme. 3 Mutants 1-332 and 1-321 displayed similar properties to wild-type kinase (Fig. 2C). However, the Ca 2ϩ / CaM-dependent activities of mutants 1-314 and 1-309 were significantly decreased, and mutants 1-305 and 1-299 were not activated at all by Ca 2ϩ /CaM. Mutants 1-293 and 1-277 were fully active in the absence of Ca 2ϩ /CaM. Notably, they were 2-fold more active than the Ca 2ϩ /CaM-activated mutant 1-332. In fact, these two mutants were slightly inhibited by Ca 2ϩ in the absence or presence of CaM (data not shown).
Taken together, these results suggest that residues T 310 AVVR 314 play a pivotal role in the binding of CaM kinase I to Ca 2ϩ /CaM. The Ca 2ϩ /CaM-binding domain of the enzyme contains at least residues 310 -314 and may extend further toward the NH 2 terminus of the enzyme. Mutant 1-293 acquired Ca 2ϩ /CaM-independent activity when residues I 294 KKNFA 299 were deleted, suggesting the existence of an autoinhibitory domain, which may extend toward the COOH terminus from residues 294 -299.
Effect of CaM Kinase I(294 -321) on CaM kinase I and smMLCK-The peptide, CaM kinase I(294 -321), was expected from the studies described above to contain both the Ca 2ϩ / CaM-binding domain and the autoinhibitory domain of CaM kinase I. Mutant 1-332 was activated by Ca 2ϩ /CaM in a dosedependent manner with its activation curve showing a sigmoidal relationship (Fig. 3A). The concentration of CaM producing 50% of the maximal activity (Kact.) was 51.0 Ϯ 4.5 nM (mean Ϯ S.E.), comparable to that of the wild-type enzyme (K act. of 42.8 Ϯ 3.3 nM, data not shown). The enzyme was inhibited by CaM kinase I(294 -321) in a dose-dependent manner as illustrated by the right shift in the K act. for Ca 2ϩ /CaM. However, maximal activity was measured upon addition of excess Ca 2ϩ /CaM, suggesting that peptide inhibition is competitive with respect to Ca 2ϩ /CaM. We further examined the effect of CaM kinase I(294 -321) on smMLCK, another Ca 2ϩ /CaM-dependent protein kinase (Fig. 3B). smMLCK was activated by Ca 2ϩ /CaM in a dose-dependent manner with a K act. of 5.4 nM. The enzyme was also inhibited by CaM kinase I(294 -321) competitively with respect to Ca 2ϩ /CaM. These results suggest that CaM kinase I(294 -321) directly binds Ca 2ϩ /CaM, preventing it from activating the two kinases.

Effects of CaM Kinase I(294 -321) or Other Kinase Inhibitors on Ca 2ϩ /CaM-independent Protein Kinases-The ability of
CaM kinase I(293-321) to inhibit the constitutively active mutant 1-293 as well as other Ca 2ϩ /CaM-independent protein kinases was examined (Table I) et al., 1990) and a relatively selective inhibitor of protein kinase C (Hidaka et al., 1984), respectively. Mutant 1-293 was not inhibited by either H-89 or H-7 with concentrations of at least 100 M and was not inhibited by the Ca 2ϩ /CaM antagonist, W-7 (Hidaka and Tanaka, 1983). PKA was inhibited by H-89 with an IC 50 value of 0.33 M whereas it was not inhibited by CaM kinase I(294 -321) at all. Protein kinase C was inhibited by H-7 with an IC 50 value of 24 M, though not inhibited by CaM kinase I(294 -321). Thus, CaM kinase I(294 -321) inhibited only mutant 1-293 among the Ca 2ϩ /CaM-independent protein kinases tested, suggesting that the peptide inhibits the catalytic mechanism of CaM kinase I in a manner relatively specific to that enzyme.
Kinetic Analyses of the Inhibition of Mutant 1-293 by CaM Kinase I(294 -321)-Kinetic analyses were carried out to fur-ther characterize the inhibition of mutant 1-293 by CaM kinase I(294 -321). The enzyme was inhibited by the peptide in a manner that was competitive with respect to Syntide-2 with a K i of 1.2 M (Fig. 4A) and non-competitive with respect to ATP (Fig. 4B). These results suggested that CaM kinase I(294 -321) associated with the substrate-binding domain of mutant 1-293 and blocked access of the enzyme to Syntide-2.
Corrected DNA Sequence of Rat CaM kinase I-During preparation of this article, a cDNA was isolated that encodes a predicted rat CaM kinase I (Cho et al., 1994) and that was 3 M. R. Picciotto and A. C. Nairn, manuscript in preparation.   (294 -321) or inhibitors showed "no effect" on the activity in concentrations up to 100 M. distinct from that previously reported (Picciotto et al., 1993). Analysis of the two cDNAs suggested that the original DNA sequence was incorrect. We have carried out further DNA sequencing of our original cDNA and have found that indeed there are a number of sequencing errors (Picciotto et al, 1995). Notably, Arg 309 is replaced by Ala 309 and a frameshift occurred at residue 322 resulting in an extended COOH terminus (Fig.  5A). Wild-type kinase (residues 1-374) and a truncation mutant corresponding to that predicted by the original cDNA sequence (mutant 1-332) were expressed, digested with thrombin, and compared by SDS-PAGE (Fig. 5B). Wild-type kinase migrated with a mobility of M r 42,000, consistent with the calculated M r of 41,643 predicted from the corrected sequence. By comparison, mutant 1-332 migrated with a M r of 37,000 (Fig. 5B). The Ala to Arg change that resulted from the original sequencing error, and that is included in mutants 1-314, 1-309, and CaM kinase I(294 -321), does not appear to influence the binding of either mutant, or inhibitory peptide to Ca 2ϩ /CaM (Figs. 2 and 3). DISCUSSION In the present study, we have directly analyzed the structural basis for the regulation of CaM kinase I by analyzing the activity and Ca 2ϩ /CaM-binding of a series of COOH-terminal truncated mutants. Analysis of the binding of Ca 2ϩ /CaM to the mutants indicated that the Ca 2ϩ /CaM-binding domain of the kinase is located between residues 294 -321. This conclusion was supported by the demonstration that a synthetic peptide encompassing these residues competed with Ca 2ϩ /CaM in the activation of either CaM kinase I or smMLCK. Deletion of residues 322-332 reduced the binding of Ca 2ϩ /CaM as measured by 125 I-CaM overlay, but had little effect on binding of Ca 2ϩ /CaM as determined by affinity chromatography on CaM-Sepharose or activity measurements. Truncation of residues T 310 AVVR 314 had the greatest effect on binding of Ca 2ϩ /CaM, particularly as measured by CaM-Sepharose chromatography. Further truncation of residues K 300 SKWKQ 305 completely abolished binding of Ca 2ϩ /CaM. Notably, analysis of binding of Ca 2ϩ /CaM by CaM overlay, CaM-Sepharose chromatography and activation of kinase activity indicated that there is not an absolute correlation between all measurements of binding of Ca 2ϩ /CaM and stimulation of kinase activity.
The Ca 2ϩ /CaM-binding domains of numerous proteins have been shown to have a propensity for the formation of amphipathic ␣-helices (O'Neil and DeGrado, 1990). Helical wheel analysis of residues 299 -316 indicated that these residues of CaM kinase I are predicted to contain a hydrophobic domain and a basic, hydrophilic domain (Fig. 6B). In addition, analysis of residues 303-316 by the method of Chou and Fasman (1978) predicted an ␣-helical conformation (data not shown). Direct structural analysis of the binding of Ca 2ϩ /CaM to synthetic peptides encompassing Ca 2ϩ /CaM-binding domains have indi- In the correct sequence, Arg 309 is replaced by Ala 309 , and the deletion of one nucleotide, G, in the third position of the codon GGG encoding Gly 322 causes a frameshift resulting in an extension of the COOH-terminal amino acid sequence. B, wild-type kinase was expressed with the cDNA subcloned into pGex-2T by Picciotto et al. (1993) and purified as described under "Experimental Procedures." Aliquots (10.4 g of protein/lane) of mutant 1-332 and wild-type recombinant protein were treated with or without 0.1 unit of thrombin at 37°C for 30 min prior to analysis by SDS-PAGE (10% acrylamide). Mutant 1-332 and wild-type protein are indicated by the arrow and arrowhead, respectively. cated several common features that are critical for high affinity binding (Ikura et al., 1992;Meador et al., 1992Meador et al., , 1993 (Fig. 6A). Upon binding of Ca 2ϩ to the four binding sites in CaM, the two lobes of the protein fold around the helical structure making contact between the NH 2 -terminal domain of Ca 2ϩ /CaM and the COOH-terminal part of the peptide, and between the COOH-terminal domain of Ca 2ϩ /CaM and the NH 2 -terminal part of the peptide. Two hydrophobic amino acids, 8 -12 residues apart (Trp 580 and Phe 593 in skMLCK), are necessary to bind to hydrophobic pockets in Ca 2ϩ /CaM, although extensive contacts between many other hydrophobic residues (Phe 584 and Val 587 in skMLCK) are apparent. In addition, specific basic residues (Arg 812 in smMLCK) are believed to act as "fulcrums" to initiate the interaction between the central helix of Ca 2ϩ / CaM and the target enzymes (Meador et al., 1992). The variability in the exact location of the principal hydrophobic amino acids in the different structure is tolerated by the fact that the central helix of Ca 2ϩ /CaM is highly flexible and extends to accommodate the particular binding peptide (Meador et al., 1993).
Comparison of the proposed Ca 2ϩ /CaM-binding domain of CaM kinase I with that of other Ca 2ϩ /CaM-binding proteins indicated several conserved features (Fig. 6A). Residues T 310 AVV 313 of CaM Kinase I are likely to represent the COOHterminal end of the proposed Ca 2ϩ /CaM-binding domain. These residues are essential for binding to Ca 2ϩ /CaM and are predicted from the method of Kyte and Doolittle (1982) to be the most hydrophobic region of the proposed Ca 2ϩ /CaM-binding domain. Residues K 300 SKWKQ 305 are also critical for binding Ca 2ϩ /CaM and likely represent the NH 2 -terminal end of the Ca 2ϩ /CaM-binding domain. Notably, Trp 303 of CaM kinase I aligns with conserved tryptophan residues in several other Ca 2ϩ /CaM-binding proteins (Fig. 6A). Thus, the results from the present study, together with the results obtained from the crystallographic analyses (Meador et al., 1993), suggest that the Ca 2ϩ /CaM-binding domain of CaM kinase I contains at least residues 302-314. However, the results from the CaM binding and activity studies (Fig. 2, A and C) suggest that the presence of some part or all of residues 315-321 (HMRKLQL) contribute to binding of CaM, possibly by stabilization of the structure of the CaM-binding domain. In addition, the precise NH 2 -terminal limit of the CaM-binding domain cannot be assigned based on the present experimental results.
Full-length wild-type CaM kinase I is completely dependent on Ca 2ϩ /CaM for activity. Deletion of residues 294 -332 resulted in a fully active enzyme. In contrast, a slightly longer CaM kinase I mutant (residues 1-299) was basally inactive and could not be activated by addition of Ca 2ϩ /CaM. In addition, the constitutively active kinase was inhibited by a synthetic peptide corresponding to residues 294 -321 in a manner competitive with respect to peptide substrate. Together, these results indicate that an autoinhibitory domain exists within residues 294 -321 and that the presence of residues 294 -299 is sufficient to inactivate CaM kinase I catalytic activity.
The elucidation of the molecular basis for the regulation of protein kinases has been the subject of a number of recent studies (Knighton et al., 1991(Knighton et al., , 1992Cruzalegui et al., 1992Cruzalegui et al., , 1993Ito et al., 1991;Fitzsimons et al., 1992). The structure of the catalytic subunit of PKA in a complex with the high affinity pseudosubstrate inhibitor (PKI (5-24)) has been determined by x-ray crystallography (Knighton et al., 1991(Knighton et al., , 1992. Arg 18 in PKI (the P(Ϫ3) position) interacts with several carboxyl side chains of the enzyme in a manner that places a non-phosphorylatable alanine residue in the place of the normally phosphorylated serine or threonine residue at the P(0) position (Fig. 7). Furthermore, a phenylalanine residue (Phe 10 ) that interacts with a hydrophobic pocket formed by residues Y 235 PPHH 239 of PKA is critical for the high affinity of the inhibitor.
The best characterized substrates for CaM kinase I are the members of the synapsin family. Synapsin I and II are both phosphorylated at NH 2 -terminal serine residues (Fig. 7) that are also phosphorylated by PKA (Czernik et al., 1987). The substrate specificity of CaM kinase I has recently been investigated and compared with that of PKA using a series of peptide substrates based on the sequence of site 1 of synapsin I (Lee et al., 1994). These studies indicate that the second of three consecutive arginine residues corresponds to the basic amino acid residue found in substrates for PKA at the P(Ϫ3) position (Fig. 7). In addition, the studies by Lee et al. indicated that hydrophobic amino acids are required at the P(Ϫ5) and FIG. 6. Comparison of Ca 2؉ /CaM-binding domains and ␣-helical wheel analysis. A, amino acid sequences of Ca 2ϩ /CaM-binding domains were compared. Trp 580 , Phe 584 , Val 587 , and Phe 593 of skMLCK as well as Arg 812 of smMLCK are believed to be important for binding to Ca 2ϩ /CaM (positions indicated by asterisks). Conserved amino acid residues of CaM kinase I (residues 294 -321) and other Ca 2ϩ /CaMbinding domains were aligned with these key positions. Amino acid sequences known to be directly involved in binding Ca 2ϩ /CaM with structural studies (Meador et al., 1992(Meador et al., , 1993Ikura et al., 1992) are underlined. Ca 2ϩ /CaM-binding domains: skMLCK (Ikura et al., 1992), smMLCK (Meador et al., 1992), CaM kinase II (Meador et al., 1993), CaM kinase IV (Sikera et al., 1987), calcineurin (Kincaid et al., 1988), plasma membrane Ca 2ϩ pump (Kataoka et al., 1991), and constitutive nitric oxide synthase (cNOS) (Zhang et al., 1994). B, the amino acid residues 299 -316 of CaM kinase I were plotted as an ␣-helical wheel. The diagram represents an end-on view of this region being in a righthanded ␣-helical conformation viewed from the NH 2 end. A basic, hydrophilic face and a hydrophobic face are underlined with a corresponding note.  5-24)), and peptide substrates for CaM kinase I. The amino acid sequences surrounding the phosphorylation sites of site 1 of synapsin I (Czernik et al., 1987), the cyclic AMP-response element-binding protein (CREB) (Sheng et al., 1991), and Syntide-2 were aligned with that of PKI (5-24). P(0) represents the position in PKI (5-24) that is occupied by a phosphorylatable amino acid residue in a substrate for PKA (Knighton et al., 1991). P(Ϫ11), P(Ϫ5), P(Ϫ3), and P(ϩ4) are the positions relative to P(0). Conserved amino acid residues at these positions are boxed. Three alternative comparisons of CaM kinase I (residues 286 -321) are shown in which potentially important basic residues (at P(Ϫ3)) and hydrophobic residues (at P(Ϫ11), (Ϫ5), and (ϩ4)) are aligned. P(ϩ4) positions (Fig. 7). With the exception of the synapsins, other physiological substrates for CaM kinase I remain to be identified. In the present study, we have found that Syntide-2 is an effective substrate for CaM kinase I. In addition, we have found that a peptide (LSRRPSYRKILNDL) corresponding to the amino acid sequence surrounding the PKA site phosphorylated in the cAMP-response element-binding protein (CREB) is also a very good substrate for CaM kinase I. 2 Comparison of the amino acid sequence of residues 294 -321 of CaM kinase I with that of PKI (5-24) and the phosphorylation sites of synapsin I, syntide-2, and CREB (Fig. 7) revealed several common features that support the possibility that the autoinhibitory domain of CaM kinase I interacts with the catalytic domain through a pseudosubstrate mechanism. Three alignments of the autoinhibitory domain that support a pseudosubstrate mechanism were possible. In the first (which required insertion of two gaps in the sequence), Lys 304 of CaM kinase I at the P(Ϫ3) position would place Trp 303 and Ala 311 at the P(Ϫ5) and P(ϩ4) positions, respectively. These two positions would correspond to the critical hydrophobic amino acids present in each of the known substrates for the enzyme; however, alanine is not normally considered a hydrophobic residue. In this alignment, Phe 307 would replace the phosphorylatable residue at P(0). Notably, in this alignment Phe 298 is placed at the P(Ϫ11) position in an analogous position to that of the critical phenylalanine found in PKI (5-24). In PKI, Phe 10 interacts with the hydrophobic domain Y 235 PPFF 239 in PKA catalytic subunit (Knighton et al., 1991). By analogy, Phe 298 may interact with the corresponding hydrophobic domain Y 215 PPFY 219 in CaM kinase I. However, this model is weakened by the fact that the alignment required insertion of gaps, the presence of an alanine residue at the P(ϩ4) position, and the fact that the pseudosubstrate domain would correspond to W 303 KQAFNATA 311 , which is not supported by the observation that residues I 294 KKNFA 299 are sufficient to maintain the enzyme inactive form. An alternative alignment that maintains the necessary features of the pseudosubstrate places residues I 294 KKNFAKSKW 303 in the P(Ϫ5) to P(ϩ4) positions. Important features of this alignment are the lack of insertions of gaps in the sequence, the presence of alanine instead of phenylalanine at the P(0) position, and the excellent alignment of the hydrophobic amino acids at P(Ϫ5) and P(ϩ4) positions. Finally, a third alignment that also maintains all the necessary features of a pseudosubstrate places Phe 298 and Phe 307 in the P(Ϫ5) and P(ϩ4) positions. In either alignment two or three, elements of the CaM-binding domain overlap the proposed pseudosubstrate domains, suggesting that while distinct regions of the regulatory domain of CaM kinase I are primarily responsible for mediating autoinhibition and CaM-binding, the two domains are closely associated and probably overlap.
Comparison of the results obtained in this study with that of other CaM kinases reveals a number of similarities, as well as a number of differences. There is general agreement concerning the position of the Ca 2ϩ /CaM-binding domain of the various enzymes (Fig. 6A), although the exact features of the mechanism of autoinhibition vary. A number of studies of MLCK, including detailed molecular modeling (Ito et al., 1991;Knighton et al., 1992), have suggested that this enzyme may be regulated by a strict pseudosubstrate mechanism. However, other studies using site-directed mutagenesis and/or preparations of truncated MLCK mutants have suggested that the original pseudosubstrate model requires modification (Fitzsimons et al., 1992;Yano et al., 1993). Recently, the crystal structure of a fragment of twitchin kinase, which shares a high level of identity with smMLCK, has been elucidated (Hu et al., 1994). While the structure clearly establishes the regulation of twitchin kinase by an intrasteric autoinhibitory mechanism, the lack of amino acid sequence identity of the regulatory domain of twitchin kinase with MLCK and the absence of information from site-directed and deletion mutagenesis make it difficult to evaluate the existence of a strict pseudosubstrate mechanism. Studies of CaM kinase II (Cruzalegui et al., 1992) and CaM kinase IV (Cruzalegui et al., 1993) have also suggested regulation by an autoinhibitory mechanism. In the case of CaM kinase II, the presence of residues K 291 KFN 294 (Fig. 6A) was, like CaM kinase I, sufficient to maintain the enzyme in an inactive state. However, replacement of KKFN by AAAL had little effect on Ca 2ϩ /CaM-dependent activation of the enzyme, although deletion of the amino acids generated a partially Ca 2ϩ /CaM-independent kinase. Based on these results, it was therefore concluded that autoinhibition of CaM kinase II required determinants in addition to residues 291-294. In the case of CaM kinase IV, truncation at Leu 313 (K 311 KL 313 , Fig.  6A) generated a fully active kinase that did not require Ca 2ϩ / CaM. Notably, truncation of CaM kinase I, II, and IV at an analogous hydrophobic residue (Fig. 6A) resulted in an active enzyme; however, the position of the truncation placed a pair of lysine residues either as part (CaM kinase I and II) of the autoinhibitory domain or not (CaM kinase IV). These results, together with the observations that the substrate specificity of CaM kinase I depends on hydrophobic amino acids, suggest that at least in the case of CaM kinase I, the mechanism of autoinhibition is highly dependent on hydrophobic amino acids.
As discussed above, there was not an absolute correlation between binding of Ca 2ϩ /CaM and stimulation of kinase activity. These results suggest, therefore, that although Ca 2ϩ /CaM may bind efficiently to a limited domain in the regulatory domain of the kinase, the complete regulatory domain is necessary (including COOH-terminal amino acid residues) to allow the removal of the autoinhibitory domain from the active site of the enzyme upon binding of Ca 2ϩ /CaM. The models presented above for the position of the proposed pseudosubstrate domain also raise additional questions concerning the exact role of the binding of Ca 2ϩ /CaM in the activation of the kinase. In the case of the second or third models, Ca 2ϩ /CaM would be predicted to bind directly to the pseudosubstrate domain and release it from the active site of the kinase. In contrast, in the first model binding of Ca 2ϩ /CaM would be accompanied by both removal of the pseudosubstrate domain from the active site and an alteration in the interaction of residues 294 -299 with the catalytic domain of the kinase.
In summary, these various studies suggest that the CaM kinases are not all regulated by exactly the same structural mechanism. CaM kinase I appears to be the smallest of the autoinhibited kinases, shows an overlapping substrate specificity with PKA, and appears to be regulated by a pseudosubstrate mechanism that shares more features with PKA than the other CaM kinases. In contrast, the autoinhibitory mechanism used by MLCK and CaM kinase II may have evolved from the pseudosubstrate mechanism, perhaps as a reflection of the changes in substrate specificity of these enzymes. The present studies, together with future elucidation of the structure of CaM kinase I and other CaM kinases, should help to resolve these issues.