Molecular Cloning and Characterization of CLICK-III/CaMKIγ, a Novel Membrane-anchored Neuronal Ca2+/Calmodulin-dependent Protein Kinase (CaMK)*

During a screen for novel putative Ca2+/calmodulin-dependent protein kinase (CaMK)-like CREB kinases (CLICKs), we have cloned a full-length cDNA for CLICK-III/CaMKIγ, an isoform of the CaMKI family with an extended C-terminal domain ending with CAAX motif (whereAA is aliphatic acid). As expected from the similarity of its kinase domain with the other CaMKI isoforms, full activation of CLICK-III/CaMKIγ required both Ca2+/CaM and phosphorylation by CaMKK. We also found that Ca2+/cAMP-response element-binding protein (CREB) was a good substrate for CLICK-III/CaMKIγ, at least in vitro. Interestingly enough, CLICK-III/CaMKIγ transcripts were most abundant in neurons, with the highest levels in limited nuclei such as the central nucleus of the amygdala (CeA) and the ventromedial hypothalamus. Consistent with the presence of the CAAXmotif, CLICK-III/CaMKIγ was found to be anchored to various membrane compartments, especially to Golgi and plasma membranes. Both point mutation in the CAAX motif and treatment with compactin, a 3-hydroxy-3-methylglutaryl-coenzyme A reductase inhibitor, disrupted such membrane localization, suggesting that membrane localization of CLICK-III/CaMKIγ occurred in a prenylation-dependent way. These findings provide a novel mechanism by which neuronal CaMK activity could be targeted to specific membrane compartments.

Neuronal Ca 2ϩ is known to play a critical role as an intracellular second messenger, linking neuronal excitability with many kinds of cellular biological events including synaptic plasticity and neuronal cell survival/apoptosis (1)(2)(3)(4). One of the unique features of Ca 2ϩ is that its concentration can be dynam-ically regulated both temporally and spatially (5). Despite a growing knowledge about the critical molecules involved in neuronal Ca 2ϩ influx and mobilization (e.g. N-methyl-D-aspartic acid receptors, voltage-gated Ca 2ϩ channels, inositol 1,4,5trisphosphate receptors, and ryanodine receptors), how these are converted to the specific cellular events remains largely unknown.
A significant part of signaling downstream Ca 2ϩ is thought to be mediated by calmodulin (CaM), 1 a ubiquitous and evolutionary well conserved intracellular Ca 2ϩ receptor (6,7). Although a large number of molecules have been shown to be targeted and activated by the Ca 2ϩ /CaM complex, one subgroup of multifunctional kinases, Ca 2ϩ /calmodulin-dependent protein kinases (CaMKs), has been ascribed a prominent role. This is because several unique characteristics of this group of kinases, such as rapid activation by Ca 2ϩ , steep Ca 2ϩ /CaM dependence, and induction of an autonomous kinase activity following activation, render its members good candidates as molecular devices able to convert a transient burst of synaptic activity into a longer lasting covalent modification of substrate proteins (8 -15). Indeed, among the CaMK family members, CaMKII␣ and ␤ isoforms, which are present postsynaptically and presynaptically, have been implicated in various kinds of synaptic plasticity and homeostasis (8 -11), whereas a CaMKK/ CaMKIV cascade has been shown to couple synaptic stimuli with CREB-dependent gene expression (4,12,16). The ability of the CaMKI isoforms and a related kinase, CKLiK, to phosphorylate neuronal substrates, such as synapsin I and CREB in vitro, has been demonstrated so far (13,17,18). However, little is yet known about the physiological role of CaMKI, although CaMKI␣ and CaMKI␤ have been shown to be expressed both in neural as well as non-neural peripheral tissues (13,(17)(18)(19)(20)(21).
In this study, we report the molecular cloning of mouse full-length CLICK-III/CaMKI␥ (occasionally abbreviated to CLICK-III in this paper), an isoform of CaMKI family, that has a longer C-terminal region terminating with a CAAX motif (where AA is aliphatic acid). This motif has been shown to be conjugated with isoprenoid lipids, thereby allowing proper tar-geting of various signaling molecules to cellular membrane compartments (22,23). CLICK-III was identified during a screen for novel putative CaMK-like CREB kinases (CLICKs), as a kinase homologous to CLICK-I and CLICK-II (the cloning and characterization of CLICK-I and -II will be described elsewhere). 2 The kinase activity of CLICK-III was similar in many respects to CaMKI␣ and related kinases. Thus, full activation of CLICK-III required both Ca 2ϩ /CaM and phosphorylation by a CaMKK (24 -31). In addition, Ca 2ϩ /cAMP-response elementbinding protein (CREB) (4,32) was a good substrate for CLICK-III, at least in vitro. Interestingly, CLICK-III transcripts were most abundant in neurons, with the highest levels in limited nuclei such as the central nucleus of the amygdala (CeA) and the ventromedial hypothalamus (VMH). Furthermore, CLICK-III was found to be anchored to membrane compartments, consistent with the presence of the CAAX motif at its C-terminal end. A point mutation in the CAAX motif and treatment with compactin, a 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitor (33), disrupted such membrane localization, suggesting that membrane localization of CLICK-III occurred in a prenylation-dependent way. These findings provide a novel mechanism by which neuronal CaMK activity could be targeted to specific membrane compartments.

EXPERIMENTAL PROCEDURES
Cloning and Plasmid Constructions-Human hippocampal cDNA was prepared from poly(A) ϩ RNA (Clontech) using Omniscript reverse transcriptase (Qiagen) and oligo(dT) primers. An hCLICK-III cDNA was obtained by nested PCR using two primer pairs (F1, 5Ј-CCA CTC CCT GCA ATA AAG CAT CCT C-3Ј, and R1, 5Ј-CTG CCT ATG AGT GGG AGA GGC CTT T-3Ј as a first primer set; F2, 5Ј-TGG AGG CAA TGG GTC GAA AGG AAG AA-3Ј, and R2, 5Ј-TGT CCA TTT CTT TCA GTC CTG TTG A-3Ј as a nested primer set) and subcloned into pCR-Blunt vector (Invitrogen). ICR mouse hippocampal poly(A) ϩ RNA was purified using Trizol (Invitrogen) and MACS mRNA isolation kit (Miltenyl Biotec), and mCLICK-III was subsequently obtained by the 3Ј-RACE procedure, using SMART RACE cDNA amplification kit (Clontech) following the manufacturer's instructions. A gene-specific 5Ј-primer for 3Ј-RACE was designed based on the cDNA sequence of hCLICK-III and that of rat CaMKI␥ (19) at the 5Ј-untranslated region adjacent to the first methionine (5Ј-GCA GCT TCA ACT CTG GAG G-3Ј). A single RACE-amplified fragment was obtained and subcloned into pCR-Blunt vector, and its nucleotide sequence was determined. Two independent clones were recovered and yielded an identical sequence. To construct pIRES-HAmCL3wt and pIRES-HAmCL3dC, cDNA fragments were amplified by PCR using primers F and R1 or R2 (F, 5Ј-GGG GGA ATT CTG GCG GCC GCT ATG GGG CGT AAG GAG  GAG GAG-3Ј, and R1 for wt, 5Ј-GGG GGA ATT CGT TTG GCG GCC  GCC TCC TGG GAT CAC ATA ACG AG-3Ј, or R2 for dC, 5Ј-GGG GGA  ATT CGT TTG GCG GCC GCA AAG TTC TTC TAA ATC TGG AG-3Ј) and inserted in-frame downstream to an HA tag cassette at the NotI site of pIRES-S-Tag-EGFP vector (a kind gift from Dr. Hirohide Takebayashi). To construct pcDNA3-HAmCL3wt and pcDNA3-HAmCL3dC, EcoRV/EcoRI fragments of each pIRES-S-Tag-EGFP construct were blunted and subcloned into the EcoRV site of pcDNA3 (Invitrogen). To construct pEGFP-mCL3 and pEGFP-mCL3C474S, cDNA fragments amplified with primers F and R1 or R2 (F, 5Ј-GCT  TCG AAT TCA GGC TTC AAC TC-3Ј, and R1 for wt, 5Ј-CCC TCC CGC  GGT CAC ATA ACG AGA GAC ACC CCA GTC-3Ј, or R2 for C474S,  5Ј-CCC TCC CGC GGT CAC ATA ACG AGA CAC ACC CCA GTC-3Ј) were inserted into a pEGFP-C3 vector (Clontech) cut at EcoRI and SacII sites. To construct pEGFP-CLVM, two oligonucleotide pairs (5Ј-CTG TCT CGT TAT GTG AGT GCA CA-3Ј and 5Ј-GAT CTG TGC ACT CAC ATA ACG AGA CAG GGC C-3Ј) were annealed, phosphorylated, and inserted to a pEGFP-C2 vector (Clontech) at the ApaI and BamHI sites. To construct pd2EGFPN1-CaMKKactiveMyc, a 1.3-kb cDNA fragment encoding the N-terminal kinase domain of rat CaMKK␣ was PCR-cloned from a Sprague-Dawley rat hippocampal cDNA pool, fused to a Myc epitope tag using a pair of primers (F, 5Ј-ACG GTA CCA TGG AGC GCA GTC CAG CCG, and R, 5Ј-ATT GGA TCC CTA CAG GTC CTC CTC GCT GAT CAG CTT CTG CTC TCC ATG CTT GGT CAC CCA-3Ј), and inserted into a pd2EGFPN1 vector (Clontech) cut at KpnI and BamHI site. To construct pEGFP-rCaMKI␣, a cDNA fragment corresponding to the full-length coding region of rat CaMKI␣ was obtained by PCR using primers (F, 5Ј-AAC TCGAGT GGG CCA TGC CAG  GGG CAG TG-3Ј, and R, 5Ј-ATT GGA TCC TAG TCC TAG TCC ATG  GCC CTA GAG CT-3Ј), subcloned into a pBluescriptII KS(ϩ) vector cut at XhoI and BamHI sites, and thereafter transferred in-frame to pEGFP-C1 vector (Clontech) at EcoRI and BamHI sites. All inserts in the expression vectors were verified by sequencing.
Northern Blotting-For Northern blot analysis, a total RNA blot filter was purchased from Seegene (Mouse Brain Aging Blot), and poly(A) ϩ RNA blot filters were from Clontech (all other blots). Doublestranded probe templates, corresponding to the unique sequence of CLICK-III at the C-terminal region, were generated by PCR using primers (F, 5Ј-AAG CCT CAG AAA CCT CTA GAC CCA G-3Ј, and R, 5Ј-TTC AGA CCC AAG CTG GGG CTC CAT CT-3Ј for hCLICK-III; F, 5Ј-ATG AAC CTG CAC AGC CCC AGT G-3Ј, and R, 5Ј-TTA TTG GCC TTT CTG AAG AGG-3Ј for mCLICK-III). The probes were labeled with [␣-32 P]dCTP using Ready-to-Go DNA labeling beads (Amersham Biosciences) and hybridized with the filters in ExpressHyb Hybridization Solution (Clontech) at 68°C. The filters were washed four times in 0.05% SDS, 2ϫ SSC for 10 min at 50°C, washed once in 0.1% SDS, 0.1ϫ SSC for 40 min at 50°C, and subjected to x-ray film autoradiography.
Luciferase Assay-COS-7 cells were plated onto 24-well plates at a density of 5 ϫ 10 4 per well, and 12 h later were transfected with pFR-Luc, pFA-CREB, pRL-CMV (200 ng, 50 ng, 4 ng each; Stratagene), and pIRES-HAmCL3dC (100 ng). 24 h after transfection, luciferase activity was measured using Dual-Luciferase Reporter Assay System (Promega) according to the manufacturer's protocol. The firefly and the Renilla luciferase activities were both measured using SIRIUS luminometer (Berthold). All results were normalized using Renilla luciferase (pRL-CMV), which was co-transfected with the other reporter genes (pFR-Luc and pFA-CREB).
Culture and Transient Transfection of CA1/CA3 Hippocampal Neurons-Culture of mouse CA1/CA3 hippocampal neurons was carried out as described previously for rat (16). A cDNA transfection was carried out at 7 days in vitro using a modification of the calcium phosphate method. 3 BODIPY-TR-ceramide Labeling and Fluorescence Microscopy-For BODIPY-TR-ceramide (Molecular Probes) and green fluorescent protein (GFP) imaging, COS-7 were plated on Lab-Tek chambered coverglass (4 well, Nunc) and transfected with expression vectors (0.4 g/ well), pEGFP-rCaMKI␣, pEGFP-mCL3, pEGFP-mCL3C474S, and pEGFP-CLVM using LipofectAMINE 2000 reagent. After 24 h, the cells were rinsed twice in HBSS/HEPES, incubated for 20 min at room temperature with 2.5 M BODIPY-TR ceramide/bovine serum albumin in HBSS/HEPES. The cells were washed twice with HBSS/HEPES and replaced with pre-warmed fresh medium, followed by a 1-h incubation 3 S. Takemoto-Kimura and H. Bito, manuscript in preparation.

FIG. 1. CLICK-III/CaMKI␥ is an isoform of the CaMKI family with an unusually extended C-terminal domain.
A, the deduced amino acid sequence of mCLICK-III was aligned with CKLiK, rCaMKI␣, and rCaMKI␤ by using the ClustalW Multiple Sequence Alignment Program version 1.8. The asterisks indicate positions of identical amino acid residues. The colons and dots indicate stronger and weaker degrees of residue conservation, respectively. B, comparison of amino acid identities between kinase domain sequences of known CaMKI-related kinases. The kinase domain of mCLICK-III shows high identity with all known CaMKI-related kinases, CKLiK (74.8%), CaMKI␣ (71.1%), and rCaMKI␤ (65.3%), respectively, in the order of identity (left column). The exact position of amino acid (aa) residues used for comparison are shown in parentheses. C, similarity and difference between CLICK-III and other CaMKI-related kinases. CLICK-III shares a high homology in its catalytic domain at the N terminus followed by a regulatory domain consisting of an autoinhibitory domain (AID) and a Ca 2ϩ /CaM binding domain (CBD). Unlike the other CaMKI-related kinases, however, mCLICK-III has a distinctively longer C-terminal region ending with a putative CAAX motif (CLVM). A 309-amino acidlong partial sequence of rat of CLICK-III has been reported before as a rat CaMKI␥ (asterisk) (19). GenBank TM accession numbers are as follows: mCLICK-III, AY212936; CKLiK, AF286366; rCaMKI␣, L24907 and L26288; rCaMKI␤, D86556; and rCaMKI␥ (partial cDNA), D86557.

FIG. 2. Abundant expression of CLICK-III transcripts in the CNS.
A, Northern blotting using specific probes for CLICK-III showed that both mCLICK-III and hCLICK-III are most strongly expressed in the brain. Note that mCLICK-III transcripts are expressed from embryonic day (E) 11, and hCLICK-III transcripts are expressed in a forebrain-specific manner in the CNS. Sk, skeletal. B, in situ hybridization was performed with 35 S-labeled riboprobes. Film autoradiographs of mouse brain parasagittal (I and II) and coronal sections (III and IV) showed specific hybridization signals for mCLICK-III antisense probes in restricted brain areas such as the olfactory bulbs (OB), the cerebral cortex (Cx), the hippocampus (Hippo.), the central nucleus of amygdala (CeA), and the ventromedial nucleus of hypothalamus (VMH). These signals disappeared when sense probes were used (II and IV).
in a CO 2 incubator. To simultaneously acquire GFP and BODIPY-TRceramide images under live conditions, a Carl Zeiss LSM 510 system equipped with a Carl Zeiss Axiovert 100TV inverted microscope and a ϫ63 Plan-Apochromat (NA 1.4, oil) objective (Carl Zeiss) was used. For all images, stacks of multiple Z-scan sections were obtained, and projected images were calculated off-line using a projection software on the LSM 510 system. All pseudocolor representations were assembled using Photoshop version 5.5 (Adobe). For compactin pretreatment, the normal medium was replaced with a growth medium containing 40 M compactin (Wako Pure Chemicals) for 12 h. No apparent phototoxicity was observed under our experimental conditions. Subcellular Fractionation-COS-7 cells were grown to 70% confluency on a 10 cm-dish and transfected with each expression vector (4 g per dish) using LipofectAMINE 2000 reagent. 48 h after transfection, the cells were washed twice in ice-cold PBS(Ϫ), suspended in 250 l (per dish) of homogenizing buffer containing 10 mM HEPES-NaOH (pH 7.5), 10 mM KCl, 2 mM MgCl 2 , protease inhibitors, disrupted in a Potter-Elvehjem homogenizer with 40 strokes, and adjusted to 0.25 M sucrose. To remove unbroken cells and nuclei, the homogenate was centrifuged at 600 ϫ g for 10 min. The supernatant was separated by centrifugation at 100,000 ϫ g for 60 min. After collecting the resulting supernatant (cytosolic fraction, S), the membrane pellet was washed once in the homogenizing buffer containing 0.25 M sucrose and again centrifuged at 100,000 ϫ g for 60 min. For the salt wash experiment, the 600 ϫ g supernatant was divided into two microtubes and separated by ultracentrifugation at 100,000 ϫ g for 60 min. The membrane pellet was resuspended in homogenization buffer containing 10 mM HEPES-NaOH (pH 7.5), 10 mM KCl, 2 mM MgCl 2 , 0.25 M sucrose or high salt homogenization buffer containing 1 M NaCl, followed by incubation on ice for 1 h, and ultracentrifuged again at 100,000 ϫ g. For all analyses, the membrane pellets were finally extracted with the elution buffer containing 25 mM HEPES-NaOH (pH 7.5), 2% SDS, 150 mM NaCl, protease inhibitors for 1 h at room temperature and centrifuged for 15 min at 15,000 rpm. This supernatant was collected and designated as crude membrane fraction (P or PЈ). The volume of the elution buffer was adjusted to be equal to the volume of input (identical to the 600 ϫ g supernatant). For Western blot analyses, one-third volume of 4ϫ Laemmli buffer was added and boiled for 4 min, and 20 l of each fractions were subjected to SDS-PAGE. For compactin treatment, the medium was replaced with growth medium containing 40 M compactin for 16 h.
Western Blot Analysis-After SDS-PAGE, the proteins were transferred onto a nitrocellulose membrane (Optitran BAS-85, Schleicher & Schuell), and immunoreactive proteins were detected using ECL-Plus (Amersham Biosciences) with the following concentration of primary antibodies: anti-HA tag monoclonal antibody (1:2500, 12CA5; Roche Applied Science), anti-GFP monoclonal antibody (1:1000, 3E6; Molecular Probes), and anti-phospho-CREB (Ser-133) polyclonal antibody (1: 1000; Cell Signaling). Horseradish peroxidase-linked anti-mouse or anti-rabbit IgG (1:2000; Amersham Biosciences) were used as secondary antibodies. The chemiluminescence image was acquired using a FAS-1000 system (Toyobo) equipped with a 16-bit cooled CCD camera. For quantification of signal intensity of detected immunoreactive bands, a rectangular window, which had a sufficient area to completely surround each band, was defined, and the total pixel intensity of each area was calculated using Photoshop 6.0 (Adobe). The signal intensity was normalized as a percentage of the total intensity, calculated as the sum of the band intensities detected in cytosolic and membrane fractions (S or P or PЈ/(S ϩ P)). Statistical analyses were carried out using Prism 3.0 (GraphPad software). Statistical significance (p Ͻ 0.05) was determined between two groups using unpaired Student's t test and between three groups using one-way analysis of variance (followed by post hoc Bonferroni and Neuman-Keuls tests). All data are given as means Ϯ S.E.

Molecular Cloning of CLICK-III, a Novel Brain-enriched
CaMK-During the course of studying two novel putative CaMK-like CREB kinase-I and -II (CLICK-I and -II) cloned by degenerate PCR strategies, 2 BLAST search revealed the presence of a putative human gene product with high homology to both CLICKs. We named this novel kinase human CLICK-III. A putative open reading frame of human CLICK-III had been deposited as a novel rat Ca 2ϩ /calmodulin-dependent protein kinase-like gene, an assembly of presumed exon sequences obtained from the draft Human Genome Sequence (Gen-Bank TM accession number AL023754). We cloned a human CLICK-III (hCLICK-III) cDNA (GenBank TM accession number AY212935) by PCR from a human hippocampal cDNA pool, using primer sequences corresponding to the deposited sequence, and we confirmed the authenticity of this gene product. We next used a 3Ј-RACE strategy to obtain a full-length mouse hippocampal CLICK-III cDNA (GenBank TM accession number AY212936). Mouse CLICK-III (mCLICK-III) showed a high degree of amino acid identities with CKLiK, a recently reported CaMK-like kinase (18), as well as with rat CaMKI␣ (13) and CaMKI␤ (19) (Fig. 1, A and B). A partial sequence of a rat ortholog of CLICK-III was previously reported as rCaMKI␥ (19) (asterisk, Fig. 1C).
The open reading frame of CLICK-III cDNAs contained an N-terminal kinase domain, a conserved threonine at position 178 in the activation loop, and an overlapping autoinhibitory/ Ca 2ϩ /CaM-binding domain in its middle portion; these domains showed high homology across the CaMKI-related kinases (Fig.  1B). The C-terminal end of CLICK-III was unusually extended in comparison with CKLiK, CaMKI␣, and CaMKI␤, suggesting a unique role for CLICK-III (Fig. 1C). The very C-terminal end of hCLICK-III and mCLICK-III were, however, divergent, indicating the existence of multiple C-terminal variants (data not shown).
As shown in Fig. 2A, Northern blot analyses suggested that in adult tissues, mCLICK-III was highly expressed in the brain, although heart, testis, and kidney also showed detectable amounts of hybridization signals. The mCLICK-III transcript first became detected at embryonic day 11 (E11), in parallel with the onset of the development of the central nervous system (CNS). Its expression level remained constant from E11 onward, throughout development and during adulthood Silver grains were present only on neuronal cell bodies (arrows, B and D). Moderate signals were detected on the layer V pyramidal neurons in the cerebral cortex and the CA1 pyramidal neurons in the hippocampus (E). In the hippocampus, non-principal neurons were also labeled with silver grains (arrows, F). CA1, CA1 area for Ammon's horn; hi, the hilus of the dentate gyrus; Gr, the granule cell layer of the dentate gyrus. Scale bars, A, C, and E, 0.5 mm; B, D, and F, 50 m.
( Fig. 2A, upper panels). In human mRNA blots, CLICK-III expression was almost confined to the brain, with small amount of signals in the skeletal muscles, kidney, spleen, and liver ( Fig. 2A, lower panels). Within the CNS, the strongest signal was detected in the forebrain neocortex (cerebral cortex, occipital pole, frontal lobe, and temporal lobe), the striatum (putamen and caudate nucleus), and the limbic system (amygdala and hippocampus) ( Fig. 2A, lower panels).
CLICK-III Is Neuronally Expressed and Is Abundant in the CeA and the VMH-We next examined CNS expression of CLICK-III mRNA by in situ hybridization using 35 S-labeled cRNA riboprobes. X-ray film autoradiography of hybridized parasagittal and coronal sections showed specific signals in the neuronal cell layers of the cerebral cortex, olfactory bulb, and hippocampus (Fig. 2B, I and III). Most remarkably, intense hybridization signals were shown in the CeA and the VMH (Fig. 2B, III). We obtained identical results using two independent antisense probes, either from the N-terminal or from C-terminal regions (data not shown), whereas sense probes only generated background signals (Fig. 2B, II and IV). Hybridization in the presence of excess cold riboprobes also completely abolished the antisense probe signals (data not shown), confirming the specificity of these obtained signals.
Emulsion autoradiography of individual sections revealed that CLICK-III was particularly expressed in neurons, but not found in glia, in any areas that we examined. Heavy amounts of silver grains associated with most neuronal cell bodies present in the CeA (Fig. 3, A and B) and the VMH (Fig. 3, C and D). Pyramidal cells in the cerebral cortex, especially in the layer V, and those in the area CA1 of the hippocampus were also positive in CLICK-III mRNA signals (Fig. 3E). In contrast, in the dentate gyrus of the hippocampus, only scattered non-principal cells, presumably hilar interneurons, had significant amount of silver grains (Fig. 3F).
Taken together, these sets of data established that CLICK-III was a CaMKI family member specifically expressed in neurons, with a remarkable abundance in the CeA and VMH. Such specificity in its expression profile particularly stands out

FIG. 4. Activation of CLICK-III by Ca 2؉ /CaM and CaMKK. A, mCLICK-III expressed in COS-7 cells is bound to
CaM-Sepharose beads in the presence of Ca 2ϩ . Lysates from COS-7 cells transfected with either an empty vector (mock) or a mCLICK-III cDNA (HAmCL3wt) were incubated with CaM-Sepharose beads. mCLICK-III remained bound to CaM-Sepharose after washing in the presence of 2 mM Ca 2ϩ (Beads, Ca 2ϩ ), whereas most CLICK-III was washed out in 0 mM Ca 2ϩ , 2 mM EGTA (Beads, EGTA). WB, Western blot. B, immunoprecipitate kinase assays of CLICK-III. COS-7 cells were transfected with an empty vector (mock), a wild type mCLICK-III cDNA (HAmCL3wt), or mCLICK-III C-terminal deletion mutant (HAmCL3dC) with (ϩ) or without (Ϫ) a cDNA encoding a constitutive active form of rat CaMKK␣ (CaMKK). After immunoprecipitation with an anti-HA antibody, kinase assays were performed in the presence (ϩ) or the absence (Ϫ) of Ca 2ϩ /CaM using MBP (5 g) or CREM (2.5 g) as substrates. Note that both mCL3wt and mCL3dC require CaMKK activity for their full activation. C-terminal deletion mutant of mCLICK-III showed a clear constitutive kinase activity in the presence of CaMKK, independent of the presence of Ca 2ϩ /CaM, whereas the wild type kinase was only active in the presence of Ca 2ϩ /CaM. C, CREM phosphorylation was also detected in a manner similar to 32 P incorporation, using an anti-phospho-CREB (Ser-133) antibody. D, the C-terminal deletion mutant of CLICK-III can activate CREB-dependent transcription in COS-7. Cells were transfected with CLICK-III C-terminal deletion mutant (pIRES-HAmCL3dC) and reporter genes coding Gal4-CREB and UAS-luciferase and collected, and lysates were extracted to measure the luciferase activity 24 h later. The bars represent luciferase activities relative to that of mock-transfected lysates (pIRES) and are shown as means Ϯ S.E. (n ϭ 3). *** denotes p Ͻ 0.001.
Enzymatic Properties of CLICK-III-We next tested whether CLICK-III was indeed a Ca 2ϩ /CaM-dependent protein kinase, as suggested from its domain structure (Fig. 1B). To this end, an HA-tagged CLICK-III was constructed and transfected to COS-7 cells. Crude lysates of transfected COS-7 cells were collected and incubated with CaM-Sepharose beads in the presence of 2 mM Ca 2ϩ . The beads were centrifuged and separated from the supernatant and then washed several times in the presence or absence of Ca 2ϩ . SDS-PAGE analyses showed that HA-immunoreactive bands were mostly recovered with the beads in the presence of Ca 2ϩ , whereas they were largely washed out in 0 Ca 2ϩ , 2 EGTA (Fig. 4A). Thus, CLICK-III was clearly able to bind CaM in a Ca 2ϩ -dependent manner.
Ca 2ϩ /CaM-dependent regulation of the protein kinase activity of CLICK-III was investigated using an immunoprecipitate kinase assay, with MBP and CREM as kinase substrates (Fig.  4B). Both Ca 2ϩ /CaM and CaMKK cotransfection were required to reach maximal levels of activation of the wild type CLICK-III (HAmCL3wt). In contrast, a C-terminal deletion mutant in which the entire autoinhibitory/Ca 2ϩ /CaM binding domain was removed (HAmCL3dC) showed a constitutive, Ca 2ϩ -independent kinase activity even in the absence of Ca 2ϩ , when active CaMKK was cotransfected (Fig. 4B). CLICK-III activity monitored with [␥-32 P]ATP incorporation into CREM closely paralleled the amount of phospho-CREB Ser-133 immunoreactivity, suggesting that wild type CLICK-III is able to directly phosphorylate CREB at Ser-133, in a Ca 2ϩ /CaM-and CaMKK-dependent manner, at least in vitro (Fig. 4C). We further confirmed that CLICK-III may work as CREB kinase in vivo by using a Gal4-CREB/UAS-luciferase reporter system. Transfection of a constitutively active CLICK-III (pIRES-HAmCL3dC) significantly augmented CREB-dependent gene expression in COS-7 cells (Fig. 4D). The result was normalized using Renilla luciferase (pRL-CMV), which was co-transfected with the other reporter genes (pFR-Luc and pFA-CREB).
Taken together, these data indicated that the enzymatic properties of CLICK-III recapitulated all major features of CaMKI isoforms, such as dual regulation by Ca 2ϩ /CaM and CaMKK, and activation of CREB pathway in vitro and in a heterologous system.
Membrane Localization of CLICK-III in Hippocampal Neurons and in COS-7 Cells-As we failed to raise high titer antibodies against CLICK-III, we expressed GFP-tagged CLICK-III in hippocampal pyramidal neurons 7 days in vitro, and we monitored its subcellular distribution in fixed (Fig. 5) or live samples (data not shown). Under either condition, GFP alone was diffusely distributed throughout the cytoplasm and the nucleus, whereas GFP-rCaMKI␣ remained largely ex-FIG. 6. Localization of GFP-mCL3 with the Golgi complex and plasma membranes in COS-7 cells. A, COS-7 cells transfected with GFP, GFP-rCaMKI␣, and GFP-mCL3 were stained with a Golgi-specific vital dye (BODIPY TR-ceramide) and examined under live conditions. Projected images from stacks of z-planes (Merge and TR-ceramide) and single z-plane images taken near the bottom, middle, and top of the cells showed localization of CLICK-III to the Golgi complex (arrows) and to the plasma membranes (arrowheads). Green, GFP image; red, BODIPY TR-ceramide. Bar, 10 m. B, lysates of COS-7 cells transfected with the indicated constructs were fractionated by ultracentrifugation at 100,000 ϫ g. The supernatants (S) were collected as cytosolic fractions, and the pellets (P) were washed once and centrifuged again. The resulting pellet was recovered as crude membrane fraction (P). Each fraction was blotted and visualized using an anti-GFP antibody. C, chemiluminescent signal intensities, quantified by a CCD camerabased imaging system, were presented in percentages relative to the total signal intensities detected in both fractions (S ϩ P). The bars represent means Ϯ S. E. (n ϭ 4). ***, p Ͻ 0.001. cluded from the nucleus, in keeping with prior reports (Fig. 5) (15,16). A striking difference was noted in the subcellular localization of GFP-tagged CLICK-III (GFP-mCL3); GFP-mCL3 distribution was not diffuse but seemed rather associated with intracellular compartments reminiscent of endomembrane systems such as the Golgi complexes or the endoplasmic reticulum (Fig. 5, arrowheads). Furthermore, CLICK-III was also found to be enriched at the tips of thin processes that were most likely filopodia (Fig. 5, arrows); these structures have been proposed to constitute intermediate membrane protrusions that precede the formation of stable dendritic spines during synaptogenesis (38).
To better analyze in detail the potential mechanisms that might underlie the distinct localization of CLICK-III, we overexpressed CLICK-III in COS-7 cells. Ectopic expression of GFP-mCL3 in COS-7 cells confirmed that subcellular distribution of CLICK-III was indeed distinct from that of GFP or GFP-rCaMKI␣. Unlike the diffuse distribution of GFP (Fig. 6A, left panels) or the largely cytoplasmic distribution GFP-rCaMKI␣ (Fig. 6A, middle panels), GFP-mCL3 localization overlapped with the perinuclear Golgi complex, as demonstrated by a high degree of spatial superimposition between GFP fluorescence of GFP-mCL3 and a Golgi-specific vital dye, BODIPY TR-ceramide (Fig. 6A, right panels, arrows). Such a co-localization was not detected between GFP-rCaMKI␣, or GFP alone, and the BODIPY TR-ceramide stain (Fig. 6A, upper  panels). In addition, a significant amount of GFP-mCL3 signals was seen at the plasma membranes (Fig. 6A, right panels,  arrowheads). Consistent with such significant localization of GFP-mCL3 with the Golgi complex and the plasma membranes, a sizable pool of GFP-mCL3 protein was recovered in the membranes fraction (P), after 100,000 ϫ g ultracentrifugation (Fig. 6, B and C). In contrast, GFP-rCaMKI␣ was predominantly present in the supernatants (S) and little in the membrane pellet fraction (P) (Fig. 6, B and C).
A Critical Role for a CAAX Motif in the Localization of CLICK-III to the Golgi Complex and Plasma Membranes-What might be the cause of such distinct subcellular localization between rCaMKI␣ and CLICK-III? Sequence search identified a putative CAAX membrane-anchoring motif in the very C-terminal end of mCLICK-III ( Fig. 1 and Table I). We thus examined whether the CAAX box in mCLICK-III was necessary for its membrane localization. A single amino acid substitution at the putative prenylation site (GFP-mCL3C474S) completely abolished recruitment of mCL3 to the perinuclear Golgi (Fig. 7A, left panels, arrows) and plasma membranes (Fig. 7A, left panels, arrowheads), and this mutant was now diffusely expressed in the cytoplasm (Fig. 7A, middle panels). Conversely, attachment of the CAAX box (CVLM) of the mCL3 to the C terminus of GFP (GFP-CLVM) was sufficient to confer perinuclear Golgi localization to GFP (Fig. 7A, right panels, arrows). We found, however, that GFP-CLVM did not appear to localize well to the plasma membranes, consistent with the notion that residues upstream of the CAAX motif may also play a role in the correct localization and the proper sorting to the plasma membranes (Fig. 7A, right panels) (22,23). In accordance with these observations, biochemical fractionation experiments showed a significant reduction in the amount of membrane-bound GFP-mCL3C474S especially after high salt wash (PЈ), whereas GFP-CVLM still remained heavily associated with the membranes (Fig. 7, B and C). Together, these data suggested that the CAAX box of CLICK-III is necessary to localize CLICK-III to the Golgi membranes.
To confirm this point further, mCL3-overexpressing cells were pretreated with compactin, an HMG-CoA reductase inhibitor. As blockade of this rate-limiting step of cholesterol synthesis should also significantly reduce prenylation of the CAAX box cysteine (33), we expected that compactin treatment should mimic the effect of Cys-to-Ser mutation at position 474. Consistently, although vehicle (Me 2 SO) treatment had no effect on localization of GFP-mCL3, which was strongly detected in association with Golgi and plasma membranes (Fig. 8A, left  panels, arrow), an overnight compactin treatment dramatically abolished this membrane anchoring (Fig. 8A, right panels). Prevention of membrane recruitment of GFP-mCL3 and its retention to the soluble fractions by compactin treatment was also verified by using biochemical fractionation as well (Fig. 8,  B and C).
Together, these experiments demonstrated that CLICK-III, a neuronally expressed CaMKI isoform, was able to be localized to the Golgi apparatus and plasma membranes, at least in part, via its C-terminal CAAX box in a prenylation-dependent manner. DISCUSSION CaMK represents a class of the multifunctional protein kinase family, with a particular significance for the physiology of excitable cells such as neurons or smooth muscles (1-10). A better understanding of the role of neuronal CaMKs has been of urgency because CaMK activity has been suggested to be crucial for linking neuronal activity with various types of neuronal plasticity (2-4, 8 -11, 39 -42). Recent experiments have further revealed that, in fact, various CaMK isoforms play a critical role in establishing distinct types of memory in mammals (9,12,(43)(44)(45)(46). So far, however, most analyses have focused on the role of CaMKII␣, ␤, and CaMKIV, because these are the most well characterized CaMKs widely expressed in forebrain neurons (35,36). In contrast, studies concerning CaMKI have lagged behind, in part because the originally isolated CaMKI␣ has been expressed ubiquitously and little has been shown so far about its physiological role or substrates in neurons (13,15,17).
In this study, we have identified a neuronally enriched CaMKI isoform that has the potential to be membrane-anchored. This CaMKI isoform, CLICK-III/CaMKI␥, possessed an extended C-terminal domain distinct from known CaMKs. We found that CLICK-III had three particular characteristics that were noteworthy.
First, we confirmed that the activation of CLICK-III required not only Ca 2ϩ /CaM but also a phosphorylation by other kinases, presumably by a CaMKK. Such strong CaMKK dependence for its enzymatic activation was qualitatively similar to what was reported previously for CaMKI␣ (27)(28)(29), indicating that the catalytic core of CLICK-III may resemble CaMKI␣ and may be regulated in a manner identical to CaMKI␣. Thus, maximal activation of CLICK-III might only be triggered in close vicinity to an upstream kinase such as CaMKK.
Second, unlike CaMKI␣, however, CLICK-III expression was strongly enriched in neurons. More interestingly, examination of its mRNA distribution, by in situ hybridization analyses, revealed a very strong expression in the CeA and the VMH. Such peculiar distribution stands out among all known CaMKs and indicates the possibility that CLICK-III may play a role in the proper function of these nuclei. The CeA has been shown to be a relay for most autonomic outputs and, furthermore, has been associated with the expression of the stimulus-specific state of fear (47). The VMH, on the other hand, has been linked with control of the homeostasis of feeding and sexual behaviors (48,49). Future studies are needed to determine whether some of these functions mediated through either CeA or VMH might indeed be controlled by the kinase activity of CLICK-III.
Third, our studies have clarified for the first time a mechanism by which CaMK may be anchored to the Golgi and plasma membranes. A CAAX motif that we identified in the unusually extended C-terminal end of CLICK-III played a critical role in determining the targeting of CLICK-III to the membrane compartments in a heterologous expression system. The high degree of sequence similarity of the CAAX motif of CLICK-III with that of Ras (Table I) and the disruption of membrane anchoring by either a point mutation at the putative prenylation site or by pretreatment with an HMG-CoA reductase inhibitor have suggested that prenylation of the CAAX box may constitute a mechanism through which CLICK-III may be actively sorted to the membranes. The prenyl moiety is derived from the mevalonate/cholesterol biosynthetic pathway. It is interesting to note that this pathway has been shown recently (50,51) to affect various neuronal phenotypes. Furthermore, it is also notable that many neuronal signaling proteins (e.g. Ras, Rho, paralemmin, and PSD-95) have been modified by the addition of long fatty acid chains including prenylation and palmitoylation (52,53). These post-translational modifications are suggested to play a critical role for precise membrane localization especially in highly polarized neuronal cells.
What could be the biological significance of such membrane targeting of a CaMK? In the case of Ras, CAAX box-dependent membrane anchoring was absolutely required to allow proper membrane recruitment of its downstream kinases, c-Raf and B-Raf (23). Similarly, CAAX-mediated localization of CLICK-III to the membranes might represent an advantageous mechanism to help tightly couple upstream signals to local downstream phosphorylation targets. In principle, a direct anchoring of CaMK activity to various membrane compartments is likely to facilitate the synaptic activity-induced local protein phosphorylation at or very close to the site of Ca 2ϩ entry/mobilization. Whether such membrane-delimited excitation-phosphorylation coupling can indeed be triggered by electrical activity of the neurons, either postsynaptically or presynaptically, remains to be studied. Alternatively, or additionally, the presence of a CaMK in the Golgi apparatus might enable an efficient coupling of synaptic activity with intracellular trafficking of neuronal proteins. Experiments are underway to provide answers to these issues.