Multiple Ca-Calmodulin-dependent Protein Kinase Kinases from Rat Brain PURIFICATION, REGULATION BY Ca-CALMODULIN, AND PARTIAL AMINO ACID SEQUENCE

We have purified to near homogeneity from rat brain two Ca-calmodulin-dependent protein kinase I (CaM kinase I) activating kinases, termed here CaM kinase I kinase-α and CaM kinase I kinase-β (CaMKIKα and CaMKIKβ, respectively). Both CaMKIKα and CaMKIKβ are also capable of activating CaM kinase IV. Activation of CaM kinase I and CaM kinase IV occurs via phosphorylation of an equivalent Thr residue within the “activation loop” region of both kinases, Thr-177 and Thr-196, respectively. The activities of CaMKIKα and CaMKIKβ are themselves strongly stimulated by the presence of Ca-CaM, and both appear to be capable of Ca-CaM-dependent autophosphorylation. Automated microsequence analysis of the purified enzymes established that CaMKIKα and -β are the products of distinct genes. In addition to rat, homologous nucleic acids corresponding to these CaM kinase kinases are present in humans and the nematode, Caenorhabditis elegans. CaMKIKα and CaMKIKβ are thus representatives of a family of enzymes, which may function as key intermediaries in Ca-CaM-driven signal transduction cascades in a wide variety of eukaryotic organisms.

Ca 2ϩ -calmodulin (CaM) 1 -dependent protein kinases (CaM kinases) I and IV are distinguished among members of the CaM kinase subfamily by their dependence upon phosphorylation by distinct protein kinases (CaM kinase kinases) for maximal activity (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11). Activating phosphorylation occurs at an identically positioned Thr residue in both cases: Thr-177 in CaM kinase I (8,13) and  in CaM kinase IV (11), although for the latter, Ser phosphorylation in the NH 2 -terminal region may also contribute to activation (10,14). The activating Thr is three amino acids NH 2 -terminal to a highly conserved GTPXXXAPE sequence present in protein kinase catalytic subdomain VIII (15). Phosphorylation in this region, termed the "phosphorylation lip" or "activation loop," has been shown to be essential for the regulation and activity of a number of protein kinases including protein kinase A (16), protein kinase C ␤ II (17,18), members of the mitogen-activated protein kinase family (19 -21), and the cyclin-dependent protein kinases (22)(23)(24).
In the context of CaM kinase regulation, an important unresolved issue is the number and identity of the CaM kinase kinases. An activating kinase of approximately 52 kDa was purified from pig brain using CaM kinase I as substrate (5,9), while one of 66 -68 kDa was purified from rat brain using CaM kinase IV as substrate (6,7). It was subsequently reported, however, that CaM kinase I kinase purified from pig brain (11), or an enzyme partially purified from rat brain (13), was capable of phosphorylating and activating CaM kinase IV. Conversely, a recently cloned and expressed, rat brain CaM kinase IV kinase was found to be capable of activating CaM kinase I (25). This raised the possibility that a single kinase kinase may be responsible for phosphorylating both CaM kinases I and IV. However, multiple chromatographic peaks of CaM kinase I kinase activity were observed during its purification from pig brain (5), suggesting that multiple CaM kinase kinases with overlapping substrate specificities (11) could be present.
We have now purified CaM kinase kinases from rat brain using CaM kinase I activation as an assay and report that two such kinases are present, both having the ability to activate CaM kinase I and CaM kinase IV. We also show by direct amino acid sequencing that they are isoenzymes encoded by different genes and are likely to be representatives of a family of enzymes widely distributed among eukaryotic organisms.

EXPERIMENTAL PROCEDURES
CaM Kinase Activation Assay-CaM kinase kinases were assayed by their ability to enhance the activity of expressed CaM kinases I and IV or their respective mutants toward a synthetic peptide substrate based on site 1 of synapsin I (5,8,11). A coupled procedure was used in which a standard amount (typically 0.2 ng/l) of CaM kinase I, CaM kinase IV, or their respective mutants was added to CaM kinase kinase-containing fractions in the presence or absence of Ca 2ϩ -CaM as described in the figure legends and incubated at 30°C in a mixture containing 50 mM Tris, pH 7.6, 0.5 mM DTT, 0.5 mg of bovine serum albumin (BSA)/ml, 10 mM MgCl 2 , 200 M [␥-32 P]ATP (DuPont NEN,ϳ 100 cpm/pmol), and 50 M synapsin site 1 peptide (NYLRRRLSDSNF). Quantification of 32 P incorporation into synapsin site 1 peptide was by a phosphocellulose filter paper method (1). Activating activity is defined as CaM kinase activity in the presence of CaM kinase kinase minus activity in its absence and is given in units/l CaM kinase kinase-containing fraction where 1 unit ϭ 1 pmol of 32 P incorporated into synapsin site 1 peptide/ min/ng of CaM kinase I or IV. Synapsin site 1 peptide was custom synthesized and purified by HPLC at the Biomedical Research Core Facilities of the University of Michigan. CaM kinase I (wild type), its activation site mutant (T177A), and Ca 2ϩ -CaM-independent COOHterminal truncation mutant (1-294) were expressed and purified from Escherichia coli as described in Haribabu et al. (8). CaM kinase IV and its respective activation site mutant (T196A) were expressed and purified from E. coli as described in Selbert et al. (11).
Purification of CaM Kinase Kinases-The procedure for purifying CaM kinase kinases from rat brain was a modification of the method for purification of pig brain CaM kinase I kinase (5). CaM kinase kinase activity was monitored during purification by CaM kinase I activation as described above. All procedures were at 0 -4°C. Frozen Sprague-Dawley rat brains (202 g) were homogenized in 4 volumes of buffer A (20 mM MOPS, pH 7.2, 1 mM DTT, 1 mM EDTA, 1 mM EGTA, 0.5 mM phenylmethylsulfonyl fluoride, and 5 mg of leupeptin/liter) and centrifuged at 40,000 ϫ g for 30 min. The supernatant was filtered through a 100% cotton handkerchief and chromatographed at ϳ40 ml/h on a DEAE-Sepharose CL-6B column (2.5 ϫ 36.5 cm) equilibrated in buffer A. After sample application, the column was washed with 1,400 ml of buffer A and eluted with a 1,000-ml linear gradient of 0 -0.35 M NaCl in buffer A. Active fractions were pooled, adjusted with glycerol and Triton X-100 to final concentrations of 10 and 0.025%, respectively, and chromatographed at ϳ25 ml/h on a column of hydroxylapatite (2.5 ϫ 12 cm, BioGel HT, Bio-Rad). The pool was applied to the column equilibrated in buffer B (buffer A, 10% glycerol, 0.025% Triton X-100) containing 0.2 M NaCl and washed with 165 ml of buffer B containing 0.2 M NaCl. Elution was with a 500-ml, linear gradient of 20 -225 mM sodium phosphate, pH 6.8, in buffer B (MOPS omitted). Pooled fractions were dialyzed against buffer C (10 mM MOPS, pH 7.0, 0.1 mM DTT, 0.05 M NaCl, 10% glycerol, 0.025% Triton X-100, 0.5 mM phenylmethylsulfonyl fluoride, and 2 mg leupeptin/liter) and adjusted to contain: 2 mM MgCl 2 , 1.5 mM CaCl 2 , 10 mg of leupeptin/liter, and 5 mg/liter each of calpain inhibitors I and II. The dialyzed pool was then chromatographed at ϳ15 ml/h on a CaM-Sepharose (Pharmacia Biotech Inc.) column (1.5 ϫ 8.5 cm) equilibrated in buffer D (buffer C, 2 mM MgCl 2 , 0.5 mM CaCl 2 , and 5 mg/liter each of calpain inhibitors I and II). After sample application, the column was washed with 100 ml of buffer C containing 0.2 M NaCl, followed by 15 ml of buffer C containing 50 mM NaCl, and eluted with buffer C containing 2 mM EGTA. Active fractions were chromatographed at ϳ12 ml/h on a heparin-Sepharose (Pharmacia) column (1.5 ϫ 17 cm). The pool was applied to the column equilibrated in buffer B, and the column was washed with 750 ml of this buffer. Elution was with a 300-ml, linear gradient of 0 -0.6 M NaCl in buffer B.
Peptide Sequencing-Peptides were derived from stained bands cut from SDS-PAGE gels by a modification of the previously described in situ proteolysis method (26). The excised bands were destained in 50% acetonitrile, 200 mM NH 4 HCO 3 and dried in a vacuum centrifuge. The dried gel pieces were rehydrated in a minimum volume of 50 mM NH 4 HCO 3 in 10% acetonitrile, pH 8, containing 1-4 g (depending on protein load, protease to protein ratio of 1:10, w/w) of modified sequencing grade trypsin (Promega). Digestion was allowed to proceed at 37°C overnight, and peptides were extracted with sequential washes of 500 l each of 20, 40, and 60% acetonitrile in water in a sonicating water bath. Extracts were pooled and dried in a vacuum centrifuge. Peptides were dissolved in 6 M guanidine HCl and chromatographed using reversed phase HPLC as described previously with manual peak collection. Peptides were sequenced on either an Applied Biosystems 471A protein sequencer or a Hewlett Packard G1000A protein sequencer.
Other Methods-SDS-PAGE was performed as described previously (5). M r values were calculated by a plot of log M r versus electrophoretic mobility 2 on a 7.5% acrylamide SDS gel using the following values for standards: Bio-Rad, 97,000, 66,000, 43,000; Life Technologies, Inc., 90,000, 80,000, 70,000, 60,000, 50,000, 40,000. The concentrations of CaM kinases I and IV and their respective mutants were determined by the method of Lowry et al. (27) with some modifications (28) using BSA as standard.

RESULTS AND DISCUSSION
The purification of CaM kinase kinases from rat brain consisted of sequential chromatography on DEAE-Sepharose, hydroxylapatite, calmodulin-Sepharose, and heparin-Sepharose. The activity profile of fractions eluted from the heparin-Sepharose column with a linear salt gradient is shown in Fig. 1. Two clearly separated activity peaks were observed whether the fractions were assayed for the ability to activate CaM kinase I or CaM kinase IV. The virtual superimposition of the profiles obtained using the two kinase substrates indicates that the same enzymatic species are responsible for activation of both CaM kinases. When the activating phosphorylation sites of CaM kinases I and IV, Thr-177 and Thr-196, respectively (8,11,13), were mutated to non-phosphorylatable alanine residues, activation by any of the column fractions was undetectable. This indicates that kinase kinases present in both peaks activate the two CaM kinases by phosphorylating the equivalent activation loop Thr residue. Aliquots of each column fraction were also analyzed by SDS-PAGE and protein staining with Coomassie Blue R-250 (Fig. 1, lower panel). CaM kinase kinase activity throughout the first peak correlated with the relative amounts of a band that electrophoresed slightly faster 2 The plot of log M r versus electrophoretic mobility was best fitted by a linear relationship (r 2 ϭ 0.983). (M r ϳ69,600 2 ) than the BSA standard whereas activity in the second peak was associated with a more slowly migrating band (M r ϳ 73,200 2 ). Note that some of the most active fractions (e.g. fractions 18 -20, 23-25) contain exclusively the ϳ69or ϳ73-kDa protein, indicating that the the two-peak pattern is not explainable by carryover of one or the other protein into the adjacent peak. We have designated the faster migrating band as CaM kinase I kinase-␣ (CaMKIK␣) and the slower migrating band CaM kinase I kinase-␤ (CaMKIK␤). Based on the level of purity obtained, the purification procedure described here is suitable for the preparation of both CaM kinase kinases to near homogeneity.
In Fig. 1 we have expressed CaM kinase kinase activities as units/l column fraction, where 1 unit is defined as 1 pmol of synapsin site 1 peptide phosphorylated/min/ng of CaM kinase I or IV. The differences in scale between CaM kinase I and CaM kinase IV activating activities suggest that both CaMKIK␣ and CaMKIK␤ prefer CaM kinase I as substrate relative to CaM kinase IV. It should be noted, however, that in this coupled assay absolute differences in units can reflect not only the extent of activation of the CaM kinases but also their maximal peptide kinase specific activities achievable under these assay conditions and that the latter may differ between CaM kinases I and IV. Nonetheless there is precedence for the notion that kinase kinases of this class while not absolutely substratespecific are relatively stringent in their substrate preferences. Hawley et al. (29) demonstrated that the kinase kinase responsible for the phosphorylation and activation of 5Ј-AMP-activated protein kinase (AMPK) is capable of phosphorylating and activating CaM kinase I and that conversely pig brain CaM kinase I kinase can phosphorylate and activate AMPK but that in both of these heterologous reactions the rate of phosphorylation is 2-3 orders of magnitude slower than that of the corresponding homologous reactions. Whether a similar selectivity of phosphorylation of CaM kinase I or IV by CaMKIK␣ and CaMKIK␤ exists remains to be established through future detailed kinetic comparisons.
Three lines of evidence indicated that the previously characterized pig brain CaM kinase I kinase is itself a Ca 2ϩ -CaMregulated protein kinase. 1) It demonstrates Ca 2ϩ -dependent binding to CaM-Sepharose (5). 2) It phosphorylates and activates, in a Ca 2ϩ -CaM-stimulated fashion, CaM kinase I (1-294), a form of CaM kinase I that has lost its CaM-binding domain through truncation mutagenesis (8).
3) It phosphorylates and activates, in a Ca 2ϩ -CaM-stimulated fashion, a non-CaM-binding enzyme, AMPK (29). We therefore examined whether either or both CaMKIK␣ and CaMKIK␤ are themselves Ca 2ϩ -CaM-regulated. As shown in Fig. 2 by two criteria, the activities of both enzymes are strongly stimulated by Ca 2ϩ -CaM. First, Ca 2ϩ -CaM enhances the abilities of both CaM kinase kinases to activate the CaM-independent fragment, CaM kinase I (1-294) (Fig. 2, top panel), and second, both CaM kinase kinases appear to autophosphorylate in a Ca 2ϩ -CaMstimulated fashion (Fig. 2, bottom panel). Consistent with previous reports using pig brain CaM kinase I kinase (8,29), there is detectable, albeit slight, activity of both CaMKIK␣ and -␤ in the absence of Ca 2ϩ -CaM. Previous studies have also established that Ca 2ϩ -CaM binding to CaM kinase I promotes its phosphorylation and activation by CaM kinase I kinase by inducing exposure of Thr-177 (or a conformation favorable for phosphorylation) (8,29). Further studies will thus be required to assess the relative importance of these dual effects of Ca 2ϩ -CaM in promoting phosphorylation and activation of CaM kinase I by CaMKIK␣ and -␤.
In order to prove that CaMKIK␣ and CaMKIK␤ are distinct proteins, samples of each were subjected to automated micro-sequencing as described under "Experimental Procedures." We obtained a sequence of 165 residues of CaMKIK␣ and 166 residues of CaMKIK␤, which, based on the sequence of the cDNA of CaM kinase IV kinase (CaMKIVK) reported by Tokumitsu et al. (25), could represent 33% of the amino acid sequences of both ␣ and ␤. The alignment of these amino acid sequences with that of CaMKIVK is presented in the bottom two lines of Fig. 3. Based on a 99% identity over 165 residues, CaMKIK␣ appears to represent a protein product either identical or highly related to CaMKIVK (25). On the other hand, CaMKIK␤ clearly represents a separate gene product relative to CaMKIK␣ or CaMKIVK showing only 76% identity with CaMKIK␣ over 67 overlapping residues and 73% identity with CaMKIVK over 166 overlapping residues. Moreover, these differences are distributed throughout the homologous regions. Taken together, Figs. 1-3 demonstrate the existence of two functionally similar but structurally distinct CaM kinase kinases, CaMKIK␣ and CaMKIK␤.
The peptide sequences were used to search the available protein data bases for previously cloned homologous nucleic acids using BLAST. This search was conducted before the paper by Tokumitsu et al. (25) was published, so only three highly similar sequences were found. These sequences, named CELC05H8 -2, hBRAIN, and R56818, are also presented in Fig. 3. R56818 represents a partial cDNA sequenced as part of the Human Genome Project by the Washington University Expressed Sequence Tag Project and cloned by the IMAGE Consortium at the Lawrence Livermore National Laboratory. hBRAIN is a designation we have given to a sequence assembled from four overlapping partial human cDNAs (R50465, H12132, H19237, and H19394), cloned and sequenced as was R56818, and one additional overlapping clone (F06422). As shown in Fig. 3, hBRAIN aligns with the NH 2 -terminal portion of CaMKIVK, and R56818 aligns with the COOH-terminal part without overlap of the two sequences. We have obtained a human cDNA by screening a brain cDNA library with sequences derived from those present in R56818. Whereas the cDNA is not complete, we have obtained enough sequence to ascertain that hBRAIN and R56818 represent parts of the same mRNA (data not shown). Based on the data shown in Fig.  3, the human sequence is much more similar to CaMKIK␤ (95% identity over 132 overlapping residues) than to either CaMKIK␣ (76% identity over 76 overlapping residues) or CaMKIVK (70% identity over 268 overlapping residues). Thus, the putative protein encoded by the human nucleic acid appears to be a homologue of rat CaMKIK␤.
The other sequence shown in Fig. 3, CELC05H8 -2, is part of a cosmid containing genomic DNA from the nematode, Caenorhabditis elegans. The C05H8 cosmid was cloned and se-quenced by the Genome Sequencing Center at Washington University in collaboration with the Sanger Centre in Cambridge, United Kingdom. The putative C. elegans protein kinase is predicted to be comprised of 357 amino acids and would be 56% identical to CaMKIVK over its entire length. The coding sequences were predicted from computer analysis using the Genefinder program. We have used a fragment of the cosmid containing the nematode gene to screen a C. elegans cDNA library. Sequence analysis of the positive clones reveals a region identical to the predicted sequence from Ile-366 to Arg-465. The likely existence of a protein in the nematode that corresponds to a CaM kinase kinase underscores the idea that this newly discovered family of regulatory protein kinases may be widely distributed among eukaryotic organisms.
It may be inferred from our data that mRNAs for two distinct CaM kinase kinases exist in rat brain. It seems likely that multiple activating kinases are required to regulate the activities of CaM kinases I and IV in different cells and tissues. This idea is consistent with the differential cellular and tissue distributions of CaM kinases I and IV. Whereas CaM kinase IV expression is restricted to brain, thymic lymphocytes, and meiotic male germ cells (30 -35), CaM kinase I is present in most, FIG. 3. Amino acid sequence alignment of CaM kinase kinases. Peptide sequences of rat brain CaMKIK␣ and CaMKIK␤ are aligned with the cDNA-derived amino acid sequence of CaMKIVK (25) and other homologous nucleic acids detected by data base search. hBRAIN (assembled overlapping expressed sequence tags with accession numbers F06422, R50465, H12132, H19394, and H19237) and R56818 represent partial cDNAs that were sequenced as part of the Human Genome Project by the Washington University Expressed Sequence Tags Project. CELC05H8 -2 (accession number U11029) is part of a cosmid containing genomic DNA from the nematode, C. elegans. The sequences, aligned with the Pileup program (GCG, University of Wisconsin and Ref. 42), were formatted with the residues identical to the CaMKIVK cDNA-derived sequence being shaded. Amino acids that could not be confidently identified are represented with an X. A gap inserted into a continuous sequence is indicated by a dash. Dots represent an absence of sequence information.
if not all, cell types (8,12). Even within the brain, the regions in which these two CaM kinases are enriched differ markedly. For example, CaM kinase IV is abundant in the cerebellar granule cells but is largely absent from most brain stem nuclei (36). By contrast, CaM kinase I was not detected in cerebellar granule cells (37) whereas intense immunoreactivity was observed in brain stem nuclei (38). At the subcellular level CaM kinase I is cytosolic (38) whereas CaM kinase IV has a predominantly nuclear localization (39,40). Our results now make it possible to test, in future studies, whether CaMKIK␣ and -␤ demonstrate differential tissue and/or subcellular localizations that correlate with the distribution of either of the CaM kinase targets. Differences in distribution, combined with the possibility of tissue-specific mechanisms for initiation of the CaM kinase signal transduction cascades, suggest that future efforts to understand the regulation and function of the CaM kinase kinase family of enzymes will prove to be very rewarding.
Note Added in Proof-While this paper was under review a paper appeared (Tokumitsu, H., and Soderling, T. R. (1996) J. Biol. Chem. 271, 5617-5622) reporting activation of CaM kinase IV by phosphorylation of Thr-196 by the rat brain CaM kinase kinase (CaMKIVK), which appears to correspond to CaMKIK␣.