Regulatory Domain of Calcium/Calmodulin-dependent Protein Kinase I1 MECHANISM OF INHIBITION AND REGULATION BY PHOSPHORYLATION*

Regulatory mechanisms of rat brain Ca2+/calmodu-lin-dependent protein kinase I1 (CaM-kinase 11) were probed using a synthetic peptide (CaMK-(281-309)) corresponding to residues 281-309 (a-subunit) which contained the calmodulin (CaM)-binding and inhibi- tory domains and also the initial autophosphorylation site (ThrZs6). Kinetic analyses indicated that inhibition of a completely Ca"/Ca"independent form of CaM- kinase I1 by CaMK-(281-309) was noncompetitive with respect to peptide substrate (syntide-2) but was competitive with respect to ATP. Interaction of CaMK-(281-309) with the ATP-binding site was independ- ently confirmed since inactivation of proteolyzed CaM-kinase I1 by phenylglyoxal (t% = 7 min) was blocked by ATP analog plus M e + or by CaMK-(281-309). In the presence of Ca2+/CaM, CaMK-(281-309) no longer protected against phenylglyoxal inactivation, consist- ent with our previous observations (Colbran, R. J., Fong, Y.-L., Schworer, C. M., and Soderling, T. R. (1988) J. Biol. Chem. 263,18146-18161) that binding of Ca2+/CaM to CaMK-(281-309) 1) blocks its inhibi- tory property,

portion of each subunit, with a centrally located consensus CaM-binding domain and a carboxyl-terminal portion which may be involved in the assembly of the dodecameric holoenzyme or subcellular localization of the kinase. Synthetic peptide analogs of various regions of the a-subunit of the kinase have been used to define more precisely the CaM-binding domain to residues 296-309 in this subunit (9,10). Payne et al. (9) also investigated whether synthetic peptide analogs of this region could inhibit Ca2'-independent forms of CaMkinase I1 and might therefore represent an inhibitory domain.
The analog of residues 290-309 was shown to inhibit (ICso = 24 PM) the autophosphorylated, partially Ca'+-independent form of the kinase in a partially competitive manner with respect to synthetic peptide substrate (9). Subsequent studies have demonstrated that extension of this peptide at the amino terminus to give the peptide 281-309 results in a 20-fold more potent inhibition (11,12). However, a kinetic mechanism for the inhibition was not determined in these latter studies.
Investigation of the inhibitory domain of CaM-kinase I1 is particularly interesting because autophosphorylation has a profound effect on its regulatory properties. Purified CaMkinase I1 activity is totally dependent on the addition of Ca'+/ CaM to the assay. However, following Ca'+/CaM-dependent autophosphorylation the kinase becomes partially Ca2'/CaMindependent, exhibiting 50430% of total activity in the presence of EGTA (13)(14)(15)(16). Recent results have demonstrated that autophosphorylation at threonine 286 in the a-subunit (threonine 287 in p-subunit) correlates with the generation of Ca'+/CaM-independent kinase activity (17)(18)(19). Thus, this regulatory autophosphorylation site is within the inhibitory domain of the kinase, as defined using synthetic peptides (see above).
The present work suggests that the mechanism for interaction between the inhibitory and catalytic domains of CaMkinase I1 in the absence of Ca2'/CaM does not appear to be strictly pseudosubstrate in nature as has been suggested for several other protein kinases (reviewed in Ref. 20). In addition, data are presented indicating that autophosphorylation at threonine 286 results in partial Ca'+-independence of CaMkinase I1 by reducing the effectiveness of the inhibitory domain.
Lines of best fit (f > 0.98) were computed from the experimental data of the kinetic analyses by the method of least squares using the RS/1 program (BBN Software Products Corp.).
Limited Proteolysis of CUM Kinase ZZ-CaM-kinase I1 (100-500 pg/ml) was autophosphorylated and then subjected to limited proteolysis with chymotrypsin as described previously (12) except that 50 PM ATP was used in the autophosphorylation reaction.
Stoichiometric Phosphorylation ofCaMK-(281-309/-CaMK-(281-309) (40 p~) was incubated for 30 min at 30 "C with 50 mM HEPES, pH 7.5, containing 10 mM magnesium acetate, 0.5 mM calcium chloride, 0.5 mM [y3*P]ATP (50-200 cpm/pmol), 0.01% Triton X-100, 67 p~ CaM, and 130 nM native CaM-kinase 11. The reaction was stopped by the addition of trichloroacetic acid (10% final concentration). The trichloroacetic acid pellet contained all the CaMK-(281-309) and was washed with 10% trichloroacetic acid and then redissolved in 70% (v/v) formic acid. The peptide was applied to a C-18 cartridge (Baxter Healthcare Corp.) equilibrated in 0.1% trifluoroacetic acid, washed with equilibration buffer, and eluted with 40% acetonitrile in 0.1% trifluoroacetic acid. Following evaporation of the solvent in a Speed-Vac (Savant), the peptide was again redissolved in 70% formic acid and further purified by reverse phase HPLC on a C-18 column (Beckman) developed with a gradient of 20-40% acetonitrile in 0.1% trifluoroacetic acid over 50 min. The concentration of CaMK-(281-309) was determined by amino acid analysis, and the stoichiometry of phosphorylation was calculated from the radioactivity in the sample. A control (non-phosphorylated) sample of CaMK-(281-309) was treated similarly except that calcium chloride was replaced with 0.5 mM EGTA and CaM was omitted from the phosphorylation incubation.
Materials-[y-32P]ATP was obtained from Du Pont-New England Nuclear. Triton X-100 was from Bio-Rad. Phenylglyoxal and AMP-PCP were obtained from Sigma and Boehringer Mannheim, respectively. The sources of other materials were described previously (12).

RESULTS
Kinetic Analyses of Inhibition-The synthetic peptide corresponding to residues 281-309 of the n-subunit of rat brain CaM-kinase I1 (CaMK-(281-309)) has been shown to inhibit a proteolyzed Ca"/CaM-independent form of CaM-kinase I1 with an ICso = 0.9 pM (12) using either syntide-2 or glycogen synthase as substrate. In addition, this peptide has been shown to inhibit Ca2+/CaM-independent autophosphorylation and Ca'+/CaM-independent phosphorylation of synapsin I by an autophosphorylated Ca'+-independent form of CaMkinase I1 (11). However, the kinetic mechanism of inhibition was not investigated in either of these studies.
Proteolyzed CaM-kinase I1 was assayed using fixed [ATP] (0.4 mM) and variable concentrations of CaMK-(281-309) (0-2 p M ) and syntide-2 (5-60 pM). When the data were analyzed by double-reciprocal (Lineweaver-Burk) plots, the computed lines of best fit intersected at the r axis ( Fig. 1, top) indicating that the mechanism of inhibition is noncompetitive with respect t o syntide-2. When the slopes of the fitted lines were replotted against the concentration of CaMK-(281-309), a straight line was obtained and a mean Ki of 0.21 & 0.04 p~ was calculated from three similar experiments (data not shown). CaMK-(281-309) was also shown to inhibit proteolyzed CaM-kinase I1 by a noncompetitive mechanism with respect to glycogen synthase substrate (data not shown).
In order to further investigate the mechanism of inhibition by CaMK-(281-309) proteolyzed kinase was also assayed at no significant phosphorylation of CaMK-(281-309) compared to synthat the mechanism of inhibition is competitive with respect to ATP. Slope replots again gave a linear relationship, and a mean Ki of 0.16 * 0.01 pM was calculated from three similar experiments (data not shown). Similar results were obtained when glycogen synthase was used as the phosphate acceptor substrate rather than syntide-2 (data not shown). Similar analyses to those described above were performed with CaMK-(290-309) and indicated that CaMK-(290-309) inhibited proteolyzed CaM-kinase I1 by a mechanism which was noncompetitive with respect to ATP and competitive with respect t o syntide-2 (data not shown). These results confirm and extend previous data (9) obtained using autophosphorylated, Ca'+-independent CaM-kinase 11.

Regulatory Domain of Ca2+/Calmodulin-Kinase II
Since the inhibitory mechanism of CaMK-(281-309) is unusual in that it is competitive with ATP rather than the protein/peptide substrate, we wanted to further probe the interaction of CaMK-(281-309) with the ATP-binding site using an independent approach. Phenylglyoxal is a reagent which modifies primarily arginine residues and has been used to modify and inactivate CaM-kinase I1 (22). Incubation of native CaM-kinase I1 with phenylglyoxal rapidly inactivated the kinase (85% inactive after 15 min) only in the presence of M P , Ca2+, and CaM. Inclusion of ADP in the incubations protected against the inactivation, suggesting that modification occurs at the adenine nucleotide binding site. Inactivation was also shown to be associated with the incorporation of 1 mol of phenylglyoxal/mol kinase subunit (22). Similar experiments were performed in this laboratory using native CaMkinase I1 yielding essentially identical results (not shown).
When proteolyzed CaM-kinase I1 was incubated with phenylglyoxal, rapid inactivation occurred with only 10-15% of original activity remaining after 15 min (Fig. 2). Inactivation was unaffected by the presence or absence of M g + (not shown) or Ca2+/CaM (Fig. 2, bottom). In the presence of M P and absence of phenylglyoxal approximately 25% inactivation was observed (Fig. 2 tivation by phenylglyoxal. In the presence of M P , 1 mM AMP-PCP protected approximately 60% of the inactivation by phenylglyoxal (Fig. 2). 4 mM AMP-PCP completely protected the kinase from phenylglyoxal inactivation in the presence of M P , but was without effect on inactivation in the absence of MgZ+ (not shown). Thus, inactivation by phenylglyoxal appeared to be caused by reaction at the ATP-binding site of the proteolyzed kinase.

DISCUSSION
The present data indicate that the mechanism involved in the suppression of CaM-kinase I1 activity in the absence of Ca2+/CaM is somewhat different than for other protein kinases. Three observations suggest that the inhibitory domain contained within residues 281-309 (a-subunit rat brain kinase) interacts with the catalytic domain and restricts access to the ATP-binding site. Firstly, the inhibition of proteolyzed CaM-kinase I1 by a synthetic peptide analog of this region (CaMK-(281-309)) is by a mechanism which is noncompetitive with respect to syntide-2 (synthetic peptide substrate) but is competitive with respect to ATP (Fig. 1). Secondly, a residue (presumably an arginine) in, or close to, the ATP-binding site of the native kinase reacts with phenylglyoxal producing inactivation only when Ca2+/CaM (and Mg+) are added (22). Limited proteolysis generates a Ca2+/ CaM-independent kinase and removes the CaM-binding and inhibitory domains of the kinase (12,23) and also exposes this residue in the ATP-binding site to phenylglyoxal even in the presence of EGTA (Fig. 2). Thirdly, this reactive residue in the proteolyzed kinase can be protected by the addition of CaMK-(281-309), and this protection is reversed by Ca2+/ CaM (Fig. 2). Binding of CaZ+/CaM to CaMK-(281-309) has previously been shown to block its inhibitory properties (12).
Several published reports suggest that binding of Ca2+/ CaM to native CaM-kinase I1 affects the interaction with adenine nucleotides. Shields et al. (33) reported that the kinase in synaptic junctions could only be photoaffinity labeled with [~u-~~P]8-azido-ATP in the presence of Ca2'/CaM. In addition, ADP has been used to protect the kinase from inactivation by phenylglyoxal (22) and 5'-p-fluorosulfonylbenzoyl adenosine (34). In both cases addition of Ca2'/CaM decreased by 10-fold the concentration of ADP required to give half-maximal protection from inactivation (22,34). These data can be explained by our observations that the inhibitory domain of the kinase restricts access to the ATP-binding site of the kinase in the absence of Ca2'/CaM.
It is interesting that the inhibitory domain of CaM-kinase I1 appears to act by a different kinetic mechanism than that of other protein kinases. It was first suggested that the regulatory subunit/domain of cyclic AMP-and cyclic GMP-de-pendent protein kinases suppresses the activity of constitutively active catalytic subunit/domain by interacting with the catalytic site, mimicking the interaction with phosphate acceptor substrates (24, 25). Such a pseudosubstrate inhibitory domain is also thought to be responsible for suppressing the activity of myosin light chain kinase (26)(27)(28) (12). Furthermore, the synthetic peptide analog of the inhibitory domain (CaMK-(281-309)) appears to interfere with the ATP-binding site of the kinase (see above). The reasons for this difference in mechanism of action of the inhibitory domain are not completely clear. It is possible that the 290-309 region interacts with the protein substrate binding site, which would explain the competitive mechanism for inhibition by CaMK-(290-309) (9). Extension at the amino terminus to CaMK-(281-309) might then also lead to interaction with the ATP-binding site, which must be in close proximity to the protein substrate binding site, resulting in a higher affinity for the interaction. Recent data obtained using the peptide CaMK-(281-289) have indicated that some inhibitory determinants are contained within this peptide which cannot be explained by it acting as an alternative substrate (12). The relatively weak inhibition by CaMK-(281-289) observed using 1 mM syntide-2 as substrate (12) may be due to interaction of the peptide at the ATP-binding site. Thus, the inhibitory domain of CaM-kinase I1 may contain elements that interact with both substrate-and ATP-binding sites. If CaM-kinase I1 shows an ordered reaction mechanism, with binding of Mg. ATP preceeding binding of protein substrate, this could account for the competitive inhibition with ATP.
Such an ordered mechanism has been described for cyclic AMP-dependent protein kinase (32). The inhibitory domain of CaM-kinase I1 is located in close proximity to two previously identified functional domains of the kinase. The CaM-binding domain has been defined to residues 296-309 (9, lo), and threonine 286 has been identified as the initial site of Ca'+/CaM-dependent autophosphorylation which correlates with the generation of the partially Ca2+-independent form of the kinase (17)(18)(19). Therefore, it is likely that there are extensive interactions between these three functional domains. It has been shown that the binding of Ca2+/CaM to CaMK-(281-309), which may result in a conformational change such as induction of an a-helix (35,361, relieves the inhibition of proteolyzed CaM-kinase I1 (12). CaMK-(281-309) has also been shown to be a poor substrate for the proteolyzed kinase in the presence of EGTA, but phosphorylation at threonine 286 is enhanced approximately 10-fold by binding of Ca2+/CaM to the peptide (12). In the present study, the phosphorylated form of CaMK-(281-309)

Regulatory Domain
of Ca2+/Calmodulin-Kine 11 was isolated and shown to be a much less potent inhibitor of proteolyzed CaM-kinase I1 than the non-phosphorylated peptide, although the two forms were equally potent in their ability to bind CaM (Fig. 3). The 10-fold decrease in inhibitory potency that occurs following phosphorylation may be an underestimation since the stoichiometry of phosphorylation was only approximately 0.8.
These data provide additional support to a recently presented model for the regulation of CaM-kinase I1 by Ca2+ /  CaM and autophosphorylation (4, 12). Additional synthetic peptide analogs will be used to test further aspects of this model. However, perhaps the best evidence for or against the model will be obtained using site-directed mutagenesis approaches.