Cellular Signaling through Multifunctional Ca 2 1 /Calmodulin-dependent Protein Kinase II*

Ca 2 1 /CaM-dependent protein kinase II (CaM-KII) 1 is a ubiqui-tously expressed protein kinase that transduces elevated Ca 2 1 signals in cells to a number of target proteins ranging from ion channels to transcriptional activators. CaM-KII has a unique ho- loenzyme structure and autoregulatory properties that allow it to give a prolonged response to transient Ca 2 1 signals and to sense cellular Ca 2 1 oscillations. In neurons CaM-KII is highly expressed and localized with certain subcellular structures. Upon activation it can translocate to excitatory synapses where it regulates a num- ber of proteins involved in synaptic transmission and its down-stream signaling pathways. Elevated intracellular free calcium (Ca 2 1 i ), in response to ago-nist stimulation or cell depolarization, is highly regulated and involves influx through voltage- and ligand-gated Ca 2 1 -permeable ion channels, release from intracellular stores through ryanodine-and inositol 1,4,5-trisphosphate-sensitive channels, sequestration by Ca 2 1 pumps and exchangers, and signaling through specific Ca 2 1 transducer proteins (1). Changes in intracellular calcium can display variable responses ranging from highly localized, transient elevations within subcellular structures ( e.g. a dendritic spine of a neuron) to Ca 2 1 waves that spread throughout the cell including the nucleus. The most ubiquitous calcium-sensing protein is cal-modulin (CaM), which contains four “EF” hand motifs with high specificity for binding Ca 2 1 . The Ca 2 1 /CaM complex interacts with and modulates the functionality of a large number of proteins (2) including several Ser/Thr protein kinases (CaM-Ks). This review, which is part of a series on Ca 2 1 /CaM-dependent


Sensor of Cellular Ca 2؉ Oscillations
The efficiency of synaptic transmission between neurons can be modulated, a process known as synaptic plasticity. At excitatory synapses using glutamate as neurotransmitter, synaptic plasticity such as long term potentiation (LTP) is triggered by increased Ca 2ϩ in postsynaptic spines (14) and is dependent on the frequency of afferent stimulation (15). How is the frequency of Ca 2ϩ i oscillations in the postsynaptic spine decoded? CaM-KII has been touted as a decoding mechanism (Fig. 2) because of its unique activation properties as discussed above and its localization in dendritic spines in an organelle called the postsynaptic density (PSD) (16). The magnitude of constitutive CaM-KII activity because of autophosphorylation of Thr-286 on adjacent subunits in the oligomeric holoenzyme should depend on the duration, amplitude, and frequency of elevated Ca 2ϩ i , and a recent in vitro study (17) shows this to be the case. The abilities of CaM-KII to decode the frequency of synaptic stimulation and to give a prolonged readout beyond the initial stimulus are two characteristics required for a molecule involved in generation of synaptic plasticity.

Subcellular Localization and Translocation
In many cells CaM-KII is largely soluble and widely distributed throughout the cell, but discrete subcellular pools of CaM-KII have recently become recognized. Localization of signaling enzymes close to their substrates has, in general, important regulatory consequences (18), especially for broad specificity enzymes such as CaM-KII. Thus, characterization of mechanisms for subcellular localization of CaM-KII and its physiological roles are intense areas of investigation.
Alternative splice variants of ␣, ␦, and ␥ isoforms contain a nuclear localization signal (19,20), and nuclear CaM-KII is likely to play a role in Ca 2ϩ -mediated transcriptional regulation of genes such as brain-derived neurotrophic factor (21) and atrial natriuretic factor (22) through phosphorylation of transcription factors including CCAAT/enhancer-binding protein (C/EBP) (23,24). It is intriguing that CaM kinases I and IV can phosphorylate a Ser adjacent to the nuclear localization signal and prevent nuclear localization of the CaM-KII, but whether this occurs physiologically is uncertain (25). CaM-KI exhibits broad cellular distribution and is largely cytoplasmic, whereas CaM-KIV has a rather restricted tissue distribution and exists as both cytosolic and nuclear isoforms (26). CaM-KII may also exert negative effects on transcription through phosphorylation of the transcription factor CREB. Surprisingly, although the activation site (Ser-133) in CREB can be efficiently phosphorylated by CaM-KII, it simultaneously phosphorylates another site (Ser-142) that exerts a dominant negative role (27). Thus, it is possible that nuclear CaM-KII can inhibit CREBdependent transcription meditated by kinases such as PKA, but this needs to be verified under physiological conditions. Ca 2ϩstimulated gene transcription through CREB is mediated in part by CaM-KIV (28,29).
Anchoring proteins that localize PKA (18) and protein kinase C (30) close to physiological substrates have been well characterized, and recent studies indicate a similar regulatory scheme for CaM-KII. An anchoring protein, ␣KAP, has recently been identified that localizes skeletal muscle CaM-KII to the sarcoplasmic reticulum. This unique protein contains a hydrophobic N terminus fused to the C-terminal association domain of CaM-KII (31). The C-terminal association domain of ␣KAP can form heteromers with the full-length CaM-KII subunit, and the hydrophobic N terminus of ␣KAP directs the resulting kinase complex to the sarcoplasmic reticulum membrane (32). Likely substrates for CaM-KII in the sarcoplasmic reticulum include the ryanodine receptor (33), phospholamban (34), and the Ca 2ϩ -ATPase pump (35).
In brain, there is evidence for colocalization of the CaM-KII ␤ isoform with the cytoskeleton. Upon stimulation of the Ca 2ϩpermeable NMDA-R ion channel in hippocampal neurons, CaM-KII appears to dissociate from F-actin and undergo translocation to membranous fractions including the PSD (36). The PSD, a complex of postsynaptic membrane proteins involved in mediating and modulating synaptic transmission, is held together and to the cytoskeleton through anchoring proteins of the PDZ/SAP family (16,37). CaM-KII is a major constituent of the PSD where it is anchored in part through the protein densin-180 (38). This interaction of CaM-KII with densin-180 does not appear to depend on the activation state of the kinase. In contrast, additional CaM-KII can associate with the PSD through interaction with the NMDA-type glutamate-gated ion channel, but this translocation appears to require activation of the kinase and its autophosphorylation on Thr-286 (39,40). Translocation to the PSD occurs in hippocampal slices upon treatments that activate CaM-KII, and it promotes the phosphorylation of CaM-KII substrates in the PSD such as the AMPA-R (39,41). The anchoring interaction appears to occur between the catalytic domain of CaM-KII and residues 1290 -1309 in the cytosolic tail of the NR2B subunit of the NMDA-R (42). Translocation would localize CaM-KII at a critical site of Ca 2ϩ influx into the dendritic spine because the NMDA channel has considerable Ca 2ϩ permeability, and its activation is required for several types of synaptic plasticity. In addition to localizing CaM-KII to a site of Ca 2ϩ influx, this translocation also situates the activated kinase in close proximity to several very important neuronal substrates (Fig. 3) as discussed below.

Activation and Synthesis of CaM-KII in Neurons
There is considerable evidence that activation of CaM-KII in pyramidal neurons of region CA1 in hippocampus is intimately involved in the phenomena of LTP (43,44). LTP is touted as a cellular model of learning and memory because it represents a neuron-specific mechanism for increasing the efficacy of synaptic transmission (45). Induction of LTP, through activation of the NMDA-R, in hippocampus triggers CaM-KII autophosphorylation on Thr-286 and formation of its constitutively active form (46 -48) (Fig. 3). Maintaining the constitutive activity of CaM-KII for at least 1 h during LTP requires inhibition of protein phosphatase 1, which can dephosphorylate Thr-286 (49), through PKA-mediated phosphorylation of the protein phosphatase 1 inhibitor (50). Induction of LTP by multiple tetanic stimuli results in global activation of CaM-KII throughout apical dendrites and the pyramidal cell somas (48). However, very mild NMDA-R stimulation of cultured neurons can result in very restricted activation of CaM-KII within individual dendritic spines (51).
Dendrites contain the machinery for localized protein synthesis, and one of the more abundant mRNAs in dendrites encodes the ␣ subunit of CaM-KII. The 3Ј-untranslated region (UTR) of CaM-KII mRNA is responsible for its dendritic migration (52). Induction of LTP results in a selective increase in dendritic ␣ CaM-KII protein, which is blocked by anisomycin and detected within 5 min, strongly suggestive of localized synthesis (53). How might dendritic synthesis of proteins such as CaM-KII be regulated (54)? The 3Ј-UTR in the mRNA encoding ␣ CaM-KII contains two cytoplasmic polyadenylation elements (CPEs). These CPEs interact with a CPEbinding protein (CPEB), which is present in hippocampal dendrites and enriched in PSDs. Binding of CPEB to the 3Ј-UTR of CaM-KII mRNA promotes its polyadenylation, thereby enhancing translation (55). Certain forms of synaptic plasticity in the visual cortex increase polyadenylation of ␣ CaM-KII mRNA with a 1.7-fold increase in CaM-KII protein in isolated synaptoneurosomes (55). The signaling pathway(s) between neural activity and enhanced CPEBdependent polyadenylation is not known, but phosphorylation of CPEB may be involved (56).

Neuronal Substrates of CaM-KII
CaM-KII can phosphorylate a large number of proteins in vitro (3), and recently several substrates that may be involved in synaptic plasticity have been identified (Fig. 3). Numerous studies have documented that activated CaM-KII can phosphorylate the GluR1 subunit of the AMPA-R and enhance its current (43). A recent report documents AMPA-R phosphorylation by CaM-KII during LTP in region CA1 of hippocampus (57). Phosphorylation of Ser-831 in GluR1 by CaM-KII potentiates AMPA-R current by increasing single channel conductance (58). Indeed, about 60% of CA1 neurons that exhibit potentiation during LTP show an increase in unitary conductance (59). Experiments with transgenic mice also support a crucial role of CaM-KII phosphorylation of GluR1 as an important component of CA1 LTP in mature animals. For example, mice lacking GluR1, the AMPA-R subunit phosphorylated by CaM-KII, show a specific deficit in LTP (60). Likewise, a single-site mutation in ␣CaM-KII (T286A) results in a mouse that is deficient in CA1 LTP (61). This subtle mutation does not effect activation of CaM-KII through binding of Ca 2ϩ /CaM, but it precludes generation of constitutive activity by autophosphorylation. Thus, CaM-KII phosphorylation of AMPA-Rs with resultant potentiation of current is thought to contribute prominently to LTP at the CA1 synapse of hippocampus (43,44).
Several other PSD proteins can also be phosphorylated by CaM-

Minireview: CaM Kinase II 3720
KII. Indeed, the NMDA-R that acts as an anchor for activated CaM-KII can be phosphorylated by CaM-KII (62). Although phosphorylation of Ser-1303 in the NR2B subunit has not been reported to directly regulate channel properties, it appears to decrease its binding affinity for CaM-KII (42). Another substrate is a novel Ras-GTPase-activating protein (SynGAP) localized at the PSD of hippocampal neurons through its interaction with PDZ domains of the scaffold proteins PSD-95 and SAP102 (63,64). SynGAP is phosphorylated and potently inhibited by CaM-KII. This suggests that activation of the NMDA-R, which is also part of the PSD-95 complex, may result in activation of CaM-KII, which in turn phosphorylates and inhibits SynGAP, thereby potentiating activation of the mitogen-activated protein kinase pathway that appears to be important in some forms of synaptic plasticity. Regulation of mitogen-activated protein kinase through CaM-KII-mediated phosphorylation of SynGAP needs to be verified in intact neurons. Another CaM-KII substrate anchored to PSD-95 is neuronal nitric oxide synthase (nNOS). Phosphorylation of nNOS at Ser-847 results in partial inhibition of its activity (65). Nitric oxide, the product of NOS, appears to be an important signaling molecule in brain (66).

Minireview: CaM Kinase II 3721
In Caenorhabditis elegans wild-type CaM-KII appears necessary for normal trafficking of glutamate receptors from the cell body and its clustering at neuromuscular synapses (67). In both C. elegans (67) and Drosophila (68), expression of constitutively active CaM-KII results in disruption of normal synaptic structure. In Drosophila this appears to be because of phosphorylation of the synaptic clustering protein DLG, a homologue of the mammalian SAP family involved in clustering of glutamate receptors. In mammals CaM-KII may also be involved in trafficking of glutamate receptors through interactions with SAP proteins. When GluR1 is overexpressed in CA1 hippocampal neurons, it translocates to the synapse in response to either activation of CaM-KII or LTP induction (69). This effect of CaM-KII was not because of phosphorylation of Ser-831 in GluR1, but insertion of GluR1 into the synapse was abolished by mutation of its C-terminal PDZ domain interaction site. GluR1 interacts with SAP97 (70), and SAP97 contains the CaM-KII phosphorylation site identified in Drosophila DLG. Thus, prolonged CaM-KII activity might promote recruitment of AMPA-Rs into synapses as predicted by the "silent synapse" hypothesis of LTP (44,71).

Concluding Remarks
CaM-KII has a unique holoenzyme structure that endows it with unusual regulatory properties required for sensing and transducing various types of intracellular Ca 2ϩ signals. Tremendous progress has occurred over the past 5 years in understanding the cellular and subcellular regulation of CaM-KII and in identifying physiological substrates. The latter area has been hampered by the absence of highly specific CaM-KII inhibitors, but recently a naturally occurring CaM-KII inhibitor protein, CaM-KIIN, has been cloned (72). A 27-residue peptide derived from CaM-KIIN retains its high specificity and potency for inhibition of CaM-KII, and this probe should be useful in identifying additional physiological functions of CaM-KII (73).