Differential Modulation of Ca2+/Calmodulin-dependent Protein Kinase II Activity by Regulated Interactions with N-Methyl-D-aspartate Receptor NR2B Subunits and α-Actinin*

Neuronal Ca2+/calmodulin-dependent protein kinase II (CaMKII) interacts with several prominent dendritic spine proteins, which have been termed CaMKII-associated proteins. The NR2B subunit of N-methyl-d-aspartate (NMDA)-type glutamate receptor, densin-180, and α-actinin bind comparable, approximately stoichiometric amounts of Thr286-autophosphorylated CaMKIIα, forming a ternary complex (Robison, A. J., Bass, M. A., Jiao, Y., Macmillan, L. B., Carmody, L. C., Bartlett, R. K., and Colbran, R. J. (2005) J. Biol. Chem. 280, 35329-35336), but their impacts on CaMKII function are poorly understood. Here we show that these interactions are differentially regulated and exert distinct effects on CaMKII activity. Nonphosphorylated and Thr286-autophosphorylated CaMKII bind to α-actinin with similar efficacy, but autophosphorylation at Thr305/306 or Ca2+/calmodulin binding significantly reduce this binding. Moreover, α-actinin antagonizes CaMKII activation by Ca2+/calmodulin, as assessed by autophosphorylation and phosphorylation of a peptide substrate. CaMKII binding to densin (1247-1542) is partially independent of Thr286 autophosphorylation and is unaffected by Ca2+-independent autophosphorylation or Ca2+/calmodulin. In addition, the CaMKII binding domain of densin-180 has little effect on CaMKII activity. In contrast, the interaction of CaMKIIα with NR2B requires either Thr286 autophosphorylation or the binding of both Ca2+/calmodulin and adenine nucleotides. NR2B inhibits both the Ca2+/calmodulin-dependent and autonomous activities of CaMKII by a mechanism that is competitive with autocamtide-2 substrate, non-competitive with syntide-2 substrate, and uncompetitive with respect to ATP. In combination, these data suggest that dynamically regulated interactions with CaMKII-associated proteins could play pleiotropic roles in finetuning CaMKII signaling in defined subcellular compartments.

ent interactions of CaMKII with the NMDAR NR2B subunit that are enhanced by Thr 286 autophosphorylation (21,22), but others detect binding of only [P-T 286 ]CaMKII to NR2B (18). Despite this discrepancy, it is clear that CaMKII activation is important for interactions with NR2B in intact cells (18,21,23,24). In this respect, NR2B remains a strong candidate for mediating the activity-dependent synaptic translocation of CaMKII in neurons. However, although autophosphorylation at Thr 305/306 promotes CaMKII dissociation from PSDs, it was reported to have only a modest effect on CaMKII binding to NR2B (22). In contrast, inactive CaMKII␣ binds densin-180 in the absence of Ca 2ϩ /calmodulin; autophosphorylation at Thr 286 potentiates this interaction (17), but the effects of Thr 305/306 autophosphorylation are unknown. In addition, a minimal binding domain for [P-T 286 ]CaMKII was recently identified in ␣-actinin, but the regulation of this interaction has not been explored (25). Thus, current knowledge of the dynamics of any single protein-protein interaction cannot account for PSD targeting in intact cells.
An additional effect of CaMKAPs may be to modulate CaMKII activity in specific subcellular compartments. For example, NR2B can trap an autonomous form of CaMKII in the absence of Thr 286 autophosphorylation in vitro (21). Recent studies in Drosophila have shown that Camguk is a CaMKAP that can modulate CaMKII activity and promote Thr 305/306 autophosphorylation of CaMKII (26). However, the effects of other CaMKAPs on CaMKII activity are unknown.
Here, we compare the effects of autophosphorylation and of Ca 2ϩ / calmodulin binding on CaMKII interactions with NR2B, densin-180, and ␣-actinin, clarifying the mechanisms of CaMKII binding to NR2B and ␣-actinin. In addition, the effects of these CaMKAPs on CaMKII activity are described. The diversity of the interaction mechanisms and their distinct effects on CaMKII activity together suggest that CaMKAPs provide subtle and dynamic modulation of CaMKII activity in discrete subcellular compartments.
GST Cosedimentation Assays-Purified GST fusion proteins or GST alone (Ϸ250 nM full-length protein) were incubated with CaMKII␣ (Ϸ250 nM in the indicated autophosphorylation state) and glutathioneagarose beads (Sigma; 25 l of packed resin) in pull-down buffer (50 mM Tris-HCl pH 7.5, 200 mM NaCl, 0.5% Triton X-100) to a final volume of 500 l for 2 h at 4°C. Free Ca 2ϩ concentrations were calculated using an internet-based tool (32) or were set using a Ca 2ϩ /EGTA buffering system (33). Where indicated, various nucleotides (100 M) were added to pull-down buffer during binding only. Beads were sedimented by centrifugation and washed in 500 l of pull-down buffer 6 times for 5 min each. After transfer of beads to new microcentrifuge tubes, proteins were eluted with SDS-PAGE sample buffer, resolved by SDS-PAGE, and transferred to nitrocellulose membranes. Proteins on the membrane were visualized by staining with Ponceau S and quantified from digitally scanned images using ImageJ (rsb.info.nih.gov/ij). Pilot studies established that pixel densities of individual protein bands are linearly related to the amount of protein loaded on the gel in the range of 0.04 -2.5 g/lane for both GST and CaMKII␣ (0.7-42 pmol). Amounts of CaMKII␣ sedimented on the beads were normalized to the recovered GST fusion protein, and background binding to GST alone was subtracted. To confirm the identity of the Ponceau-stained proteins, membranes were immunoblotted for CaMKII using affinity-purified polyclonal goat anti-CaMKII primary antibodies (29) and alkaline-phosphatase-conjugated secondary antibodies (Jackson Immunoresearch).
Kinetic Analyses of Inhibition by NR2B-Ca 2ϩ /calmodulin-dependent activity of non-phosphorylated CaMKII was assayed in the presence of the indicated concentrations of NR2B (1290 -1309: S1303A) peptide (AQKKNRNKLRRQHAYDTFVD, Macromolecular Resources, Fort Collins, Colorado). Two model peptide substrates were used; syntide-2 is based on a phosphorylation site in glycogen synthase, whereas autocamtide-2 is based on the sequence surrounding the Thr 286 autophosphorylation site in CaMKII. Assays were performed using either fixed [␥-32 P]ATP concentrations (0.4 mM) and variable peptide substrate concentrations or fixed peptide substrate concentrations (0.2 mM syntide-2 or 0.1 mM autocamtide-2) and variable [␥-32 P]ATP concentrations. Raw data were fit to Michaelis-Menten kinetics and inhibitory mechanisms were investigated using double-reciprocal (Lineweaver-Burk) plots (Prism 4.0, GraphPad).
Initial rates of CaMKII Autophosphorylation-CaMKII␣ (1.25 M) was preincubated with GST, GST-NR2B, or GST-actinin (2.5 M each) in 50 mM HEPES, pH 7.5, 1 mg/ml bovine serum albumin, 1 mM dithiothreitol, 0.5 mM CaCl 2 , and 2.5 M calmodulin. Control reactions lacked GST fusion proteins. The reaction was initiated by the addition of [␥-32 P]ATP (2500 cpm/pmol) and magnesium acetate to 20 M and 10 mM, respectively, and incubations were performed at 4°C. Aliquots (10 l) of each reaction were spotted on P81 phosphocellulose papers after 10, 20, 30, and 40 s to determine the initial reaction rates (mol/mol/s), which were calculated from the slope of a line fitted to the data. Additional 10-l aliquots were quenched by the addition of SDS after 50 s and analyzed by autoradiography of SDS-polyacrylamide gels.

RESULTS
Effects of Autophosphorylation on Interactions-Previous studies have shown that CaMKII associates with NR2B, densin-180, and ␣-actinin in brain (see the Introduction). GST-NR2B, GST-densin, and GST-actinin contain the CaMKII binding domains from the parent proteins and bind similar, approximately stoichiometric amounts of [P-T 286 ]CaMKII in glutathione-agarose cosedimentation assays ( Fig. 1) (25). However, non-phosphorylated CaMKII bound equally well to GST-actinin and at a somewhat reduced level to GST-densin (33 Ϯ 5% of that detected using [P-T 286 ]CaMKII) but failed to bind significantly to GST-NR2B under these conditions (but see Fig. 6) (Fig. 1, A and B, compare white and black bars in B). Thus, Thr 286 autophosphorylation alone was required for CaMKII binding to NR2B and potentiated CaMKII binding to densin by about 3-fold but had no effect on CaMKII binding to actinin.
Ca 2ϩ -independent autophosphorylation also selectively regulated these interactions. Basal Ca 2ϩ -independent autophosphorylation alone failed to support CaMKII binding to GST-NR2B and had no significant effect on binding to GST-densin ( Fig. 1A and B, compare light gray and white bars in B). Moreover, Ca 2ϩ -independent autophosphorylation after initial Thr 286 autophosphorylation had no significant effect on CaMKII binding to GST-NR2B or GST-densin (Fig. 1, A and B, compare dark gray and black bars in B). However, CaMKII binding to GSTactinin was substantially reduced (70 -90%) by Ca 2ϩ -independent autophosphorylation regardless of whether Thr 286 was initially autophosphorylated ( Fig. 1, A and B). Thr 305 and Thr 306 have been identified as major sites of Ca 2ϩ -independent autophosphorylation, although additional sites have been identified (e.g. Ser 314 ) (22,25). Mutation of both Thr 305 and Thr 306 to Ala abrogated the effect of Ca 2ϩ -independent autophosphorylation on CaMKII binding to GST-actinin. However, the double-mutated kinase bound normally to GST-NR2B (i.e. in a Thr 286 autophosphorylation-dependent manner) (Fig. 1C). Thus, Ca 2ϩ -independent autophosphorylation at Thr 305 and/or Thr 306 abrogates CaMKII binding to ␣-actinin but has no effect on CaMKII-binding to GST-NR2B or GST-densin.
Regulation of Interactions by Ca 2ϩ /Calmodulin-Because Thr 305/306 autophosphorylation prevents binding of CaMKII to both Ca 2ϩ /calmodulin (28) and ␣-actinin (see Fig. 1), we compared the effects of Ca 2ϩ /calmodulin on binding of non-phosphorylated CaMKII␣ and [P-T 286 ]CaMKII to these CaMKAPs. These assays were performed in the presence of either CaCl 2 (5 mM) or chelator (1.2 mM EDTA) because excess calmodulin (2.5 M) and chelators (1.5 mM EDTA) were always present as a carryover from the kinase preincubation. The addition of Ca 2ϩ had no significant effect on the binding of non-phosphorylated CaMKII to ␣-actinin, NR2B, or densin-180 (Fig. 2, A and B, compare the white and dark gray bars in B). However, Ca 2ϩ reduced binding of [P-T 286 ]CaMKII to ␣-actinin by Ͼ90% without affecting binding to NR2B or densin-180 (Fig. 2, A and B, compare the black and light gray bars in B). The effect of Ca 2ϩ on binding of [P-T 286 ]CaMKII to ␣-actinin was dose-dependent, with an estimated EC 50 of 11 M free Ca 2ϩ (Fig.  3A). The affinity of CaMKII for calmodulin is increased Ϸ1000-fold by Thr 286 autophosphorylation (34), presumably sufficient to allow Ca 2ϩ / calmodulin to displace ␣-actinin from [P-T 286 ]CaMKII but not from nonphosphorylated CaMKII under these conditions.
To determine whether calmodulin and ␣-actinin also compete for nonphosphorylated CaMKII, we examined the effect of ␣-actinin on calmodulin-dependent CaMKII activation. His 6 -actinin (1 M) increased the apparent k a for calmodulin from Ϸ150 nM to Ϸ1.5 M (Fig. 3B, inset); double-reciprocal (Lineweaver-Burk) analysis of these data yielded straight lines intersecting on the y axis (Fig. 3B), indicative of competitive inhibition of CaMKII by ␣-actinin with respect to Ca 2ϩ / CaM. In addition, ␣-actinin inhibited Ca 2ϩ /calmodulin-dependent substrate (autocamtide-2) phosphorylation (Fig. 4A, EC 50 Ϸ1 M) and autophosphorylation ( Fig. 4C) but had no effect on the Ca 2ϩ -independent activity of [P-T 286 ]CaMKII (Fig. 4A). In combination, these data show that ␣-actinin and calmodulin compete for binding to both [P-T 286 ]CaMKII and nonphosphorylated CaMKII.
Effects of CaMKAPs on CaMKII Activity-As controls for inhibition experiments described above, GST and GST-densin (1402-1452) were shown to have no effect on CaMKII activity (not shown). In contrast to a previous report that NR2B can trap an autonomously active form of CaMKII in the absence of Thr 286 autophosphorylation (21), GST-NR2B potently inhibited both Ca 2ϩ /calmodulin-dependent exogenous peptide substrate phosphorylation and Ca 2ϩ /calmodulin-dependent autophosphorylation under our conditions (Fig. 4, B and C). In addition, GST-NR2B inhibited the Ca 2ϩ -independent activity of [P-T 286 ]CaMKII with similar potency (EC 50 s 10 -20 nM) (Fig. 4B). In contrast, ␣-actinin inhibited only Ca 2ϩ /calmodulin-dependent activity of non-phosphorylated CaMKII (see Fig. 4A). Moreover, CaMKII activity was inhibited with similar potency by a synthetic peptide corresponding to the minimal CaMKII binding domain in NR2B (residues 1290 -1309) as well as   by the entire C-terminal tail of NR2B (His 6 -NR2B (965-1482)) (data not shown).
Kinetics of CaMKII Inhibition by NR2B-To gain a better understanding of the interaction between NR2B and CaMKII, we determined the mechanism for inhibition of Ca 2ϩ /calmodulin-dependent CaMKII activity by NR2B. Inhibition kinetics were determined with respect to two model peptide substrates with different modes of interaction at the catalytic site (syntide-2 and autocamtide-2) (35) as well as with respect to ATP. At fixed (saturating) ATP concentrations, inhibition by NR2B was non-competitive with respect to syntide-2 (Fig. 5A, lines intersect on the x axis) and competitive with respect to autocamtide-2 (Fig. 5C,   lines intersect on the y axis). Surprisingly, at fixed saturating concentrations of either peptide substrate the inhibition was uncompetitive with respect to variable ATP concentrations (Fig. 5, B and D, lines essentially  parallel).
Effect of Nucleotides on NR2B-CaMKII Interaction-Uncompetitive inhibition of CaMKII with respect to ATP indicates that nucleotide binding is required for the interaction of non-phosphorylated kinase with NR2B. Therefore, we examined the effect of an adenine nucleotide on CaMKII binding to GST-NR2B under a variety of conditions. ADP (100 M) was used to avoid potential confounding effects of ongoing phosphorylation during incubation. Binding of GST-NR2B to [P-T 286 ]CaMKII was unaffected by the addition of ADP (although it should be noted that 4 M adenine nucleotide was carried into the [P-T 286 ]CaMKII incubations from the autophosphorylation reaction). However, the addition of ADP revealed a Ca 2ϩ -dependent interaction of non-phosphorylated CaMKII with NR2B (Fig. 6) that was not seen in the absence of ADP (cf. Fig. 2). This Ca 2ϩ /calmodulin-dependent interaction was also supported by AMP and by a non-hydrolyzable analog of ATP (AMP-PNP) (data not shown). Thus, the combined effect of binding both nucleotide and Ca 2ϩ /calmodulin induces conformational changes in nonphosphorylated CaMKII that facilitate interaction with NR2B.

DISCUSSION
PSD targeting of CaMKII appears to involve differentially regulated CaMKII interactions with multiple PSD-enriched proteins (see the Introduction). In addition to targeting CaMKII to specific subcellular locations and substrates, CaMKAPs may directly modulate CaMKII activity. For example, Camguk localizes a pool of Drosophila CaMKII that can be readily activated and Thr 286 -autophosphorylated but in the absence of Ca 2ϩ Camguk promotes Thr 305/306 autophosphorylation, creating an inactive pool of kinase (36). However, similar regulation of mammalian CaMKII has not been reported. The present work suggests that NR2B, densin-180, and ␣-actinin may make different contributions to the dynamic subcellular targeting of CaMKII and differentially modulate CaMKII activity.
CaMKII Interaction with NR2B-Much attention has focused on NR2B because of the potential involvement in activity-dependent synaptic translocation of CaMKII (see the Introduction). There are two distinct CaMKII binding sites within the C-terminal intracellular region of NR2B. Thr 286 autophosphorylation is required for binding to a poorly characterized membrane-proximal CaMKII binding site (within residues 839 -1120) (19). Autophosphorylation at Thr 286 is also necessary for interaction with a high affinity binding site in NR2B corresponding to residues 1290 -1309 (Fig. 2), as initially reported (18). We sometimes detect a weak Ca 2ϩ /calmodulin-stimulated CaMKII interaction with GST-NR2B (barely detected in Fig. 2A), but on average this is not sig-  nificantly greater than nonspecific interactions with GST (Fig. 2B). However, we show here that adenine nucleotides stabilize a strong Ca 2ϩ /calmodulin-dependent interaction between CaMKII and NR2B (Fig. 6). Others have previously reported that Ca 2ϩ /calmodulin binding to CaMKII is sufficient to promote interactions with residues 1290 -1309 of NR2B, but it is unclear whether these incubations contained a source of nucleotides (19,21,22).
Preincubation of CaMKII and NR2B with Ca 2ϩ /calmodulin traps an autonomously active form of CaMKII in subsequent kinase assays without autophosphorylation at Thr 286 (21). Although we confirmed this observation (data not shown), we found that NR2B potently inhibits Ca 2ϩ /calmodulin-stimulated and autonomous CaMKII activity when added directly to kinase assays (Fig. 4). Importantly, inhibition was not a function of the specific GST fusion protein constructs used initially, because similar inhibition was detected using a synthetic peptide analog of amino acid residues 1290 -1309 and using a His-tagged protein corresponding to the entire cytosolic C terminus of NR2B (not shown). The apparent inhibitory mechanism depends on the peptide substrate used; inhibition is competitive with autocamtide-2 but non-competitive with syntide-2 (Fig. 5). This difference in NR2B inhibitory mechanisms is likely explained by a previous report that these peptide substrates can bind in different ways at the active site (35). Autocamtide-2 is based on the sequence surrounding Thr 286 in the CaMKII autoinhibitory domain, which is similar to the sequence surrounding Ser 1303 in NR2B, whereas syntide-2 is based on sequences surrounding a phosphorylation site in glycogen synthase (37).
The mechanisms of CaMKII inhibition by NR2B with respect to peptide substrates are reminiscent of CaMKII inhibition by its autoinhibitory domain peptides (35,38,39), perhaps not surprising considering the sequence similarity between NR2B and the autoinhibitory domain. These prior studies suggested a model in which the N-terminal portion of the autoinhibitory domain (surrounding Thr 286 ) occupies a binding pocket on the C-terminal lobe of the catalytic domain termed the T-site, whereas the C-terminal portion acts as a pseudosubstrate, occluding the substrate binding (S) site (39). Interaction of autoinhibitory domain peptides at the T site appears to induce long range conformational changes in the ATP binding site of CaMKII, resulting in apparent competitive inhibition of CaMKII with respect to ATP. Thus, occupation of the T site and the nucleotide binding site are mutually exclusive (35,38,39). However, inhibition by NR2B(1290 -1309) is uncompetitive with respect to ATP irrespective of the peptide substrate used (Fig. 5, B and  D), suggesting that NR2B inhibition requires prior nucleotide binding to CaMKII. This is further demonstrated by the fact that nucleotides support a Ca 2ϩ /calmodulin-dependent interaction with NR2B (Fig. 6). Thus, it is apparent that NR2B and the autoinhibitory domain are differentially coupled to nucleotide binding even though the nucleotide binding site is on the N-terminal lobe of the catalytic domain, and the T site is on the C-terminal lobe. More studies will be required to elucidate mechanisms for this differential communication.
Differences in the interaction of NR2B and the autoinhibitory domain with CaMKII catalytic domains can also be inferred from the fact that CaMKII efficiently phosphorylates Ser 1303 in NR2B within the complex (40), whereas Thr 286 in peptide analogs of the autoinhibitory domain is a poor CaMKII substrate unless Ca 2ϩ /calmodulin is bound to the peptide, abrogating its inhibitory interaction at the T site (38). Moreover, Thr 286 in the autoinhibitory domain is a poor substrate for its cognate catalytic domain even after binding of Ca 2ϩ /calmodulin and is phosphorylated by the catalytic domain of an adjacent subunit in the activated holoenzyme (41)(42)(43). Differences in the interaction mechanisms are further supported by the fact that mutation of Ile 205 in the catalytic domain to Lys severely compromises binding to NR2B (21) but only marginally disrupts the autoinhibitory interaction (39). In combination, these data strongly suggest that NR2B does not directly mimic the autoinhibitory domain in binding to the T site in the catalytic domain.
CaMKII is routinely exposed to millimolar adenine nucleotide concentrations in intact cells. Removal of the autoinhibitory domain from the T site in response to Ca 2ϩ /calmodulin binding facilitates nucleotide binding, which may be sufficient to induce CaMKII binding to NR2B in the absence of Thr 286 autophosphorylation. This may explain activitydependent synaptic translocation of T286A-mutated CaMKII␣ in intact cells (7). However, ATP is the predominant nucleotide under physiological conditions. Thus, it seems likely that wild type CaMKII␣ will autophosphorylate at Thr 286 before it can diffuse to and bind NR2B. Autophosphorylation at Thr 286 may stabilize the interaction. Our data suggest that CaMKII subunits interacting with NR2B will be inhibited, but it is possible that NR2B can trap autonomous kinase activity once Thr 286 gets dephosphorylated. Whichever effect dominates in situ, steric constraints suggest that only a fraction of the subunits in a CaMKII holoenzyme are capable of simultaneously interacting with NR2B(s) anchored in the membrane. Thus, interaction of a few CaMKII subunits from the holoenzyme with NR2B may localize additional CaMKII subunits to the NMDAR and PSDs that do not directly interact with NR2B. Subunits that are not bound to NR2B presumably remain active to phosphorylate nearby proteins, although it is possible that intersubunit cooperativity allows NR2B to affect the activity of the entire holoenzyme. Consistent with this model, interaction of CaMKII with the high affinity binding site in NR2B was recently shown to enhance phosphorylation of a model substrate engineered onto the same membrane-targeted polypeptide (24). However, dexamethasone treatment of rats decreased the amount of PSD-associated NR2B and increased the amount of PSD-associated CaMKII activity (44). Moreover, although levels of Thr 286 autophosphorylation are elevated for 60 min after long term potentiation induction, Ca 2ϩ /calmodulin-independent CaMKII activity is elevated for only 5 min (45). These data are consistent with the idea that NR2B can also inhibit CaMKII activity in vivo.
Despite evidence supporting an important role for CaMKII interaction with NR2B in cells, this interaction cannot fully account for CaMKII targeting. For example, autophosphorylation at Thr 305/306 promotes CaMKII dissociation from PSDs (8,9) but does not significantly affect binding to NR2B (Fig. 1).
CaMKII Interaction with Densin-180-Densin-180 is a PSD-enriched, transmembrane glycoprotein (46) that was recently shown to play a key role in regulating neuronal morphology and branching (47). Initial studies showed that CaMKII interacts with a C-terminal domain in densin-180 that is distinct from CaMKII binding domains in other CaMKAPs and that fragments of densin-180 constitutively target CaMKII in transfected HEK293 cells (17). Here we confirm that CaMKII binding to densin-180 is partially autonomous, although the amount of activation-independent binding appears to depend on the assay conditions ( Fig. 1) (16, 17). We have shown for the first time that CaMKII binding to densin is unaffected by Ca 2ϩ /calmodulin binding or by Ca 2ϩ -independent autophosphorylation at Thr 305/306 and other sites ( Figs. 1 and 2) and does not modulate CaMKII activity (not shown). These observations are not surprising since densin-180 interacts with the C-terminal association domain of CaMKII (16,25). Thus, densin-180 may function to anchor both inactive and active CaMKII holoenzymes in intact cells.
CaMKII Interaction with ␣-Actinin-CaMKII interacts with ␣-actinin, a ubiquitous actin-bundling protein (16). A C-terminal 75-amino acid domain of ␣-actinin is necessary and sufficient for CaMKII binding (25). Here we show that ␣-actinin bound nonphosphorylated and [P-T 286 ]CaMKII with similar efficacy but that Ca 2ϩ /calmodulin binding to CaMKII or Thr 305/306 autophosphorylation strongly abrogated the interaction (Figs. 1-3). Moreover, ␣-actinin competitively inhibited Ca 2ϩ /calmodulin-dependent activation of nonphosphorylated CaMKII ( Fig. 3B and 4). In combination, these data suggest that ␣-actinin may interact near the calmodulin binding domain. Indeed, we have shown that ␣-actinin binds to the C-terminal-truncated monomeric catalytic/ regulatory domain of CaMKII (25). However, synthetic peptide analogs of the calmodulin binding domain cannot effectively compete with ␣-actinin for binding to CaMKII (data not shown). Further studies will be required to precisely identify the ␣-actinin binding site in CaMKII.
The dynamic regulation of CaMKII binding to ␣-actinin suggests that this interaction may target an inactive pool of nonphosphorylated CaMKII, requiring higher local Ca 2ϩ concentrations for activation than a soluble kinase pool. Moreover, cell biological data obtained using CaMKII with mutated calmodulin binding sites may need to be reinterpreted in light of the potential effect of these mutations on binding to ␣-actinin. The fact that Thr 305/306 autophosphorylation decreases CaMKII binding to intact PSDs (8,9,11) and to ␣-actinin suggests that ␣-actinin may play an important role in CaMKII targeting in dendritic spines.
Implications for CaMKII Signaling Complexes-The present studies suggest provocative models for tight control of localized CaMKII activity (Fig. 7). Constitutive interaction of the C-terminal association domain with densin-180 may serve to prevent CaMKII holoenzymes from escaping the PSD, leaving catalytic and regulatory domains available for dynamically modulated interactions with ␣-actinin or NR2B. Antagonism of Ca 2ϩ /calmodulin-dependent CaMKII activation by ␣-actinin may serve to prevent inappropriate kinase activation at low levels of Ca 2ϩ influx. Binding to NR2B could inhibit some subunits in the kinase holoenzyme, whereas others remain active or bound to other CaMKAPs. Indeed, we recently showed that [P-T 286 ]CaMKII␣ can simultaneously interact with densin, ␣-actinin, and NR2B to form a putative PSD-anchored signalosome (25). Additional interactions with NMDAR NR1 or NR2A subunits or other CaMKAPs (see the Introduction) may provide more nuances to the assembly and function of these complexes. Moreover, some CaMKAPs interact with each other independently of CaMKII (16,48) creating additional possibilities that are not illustrated in Fig. 7. For example, ␣-actinin associates with NMDAR NR1 and NR2B subunits under basal conditions (48), possibly localizing inactive CaMKII to functional NMDARs. Local NMDAR-mediated Ca 2ϩ influx might promote sequential CaMKII dissociation from ␣-actinin, activation/autophosphorylation, and binding to NR2B. Such events may play a role in the NR2 subunit-selective modulation of NMDAR desensitization (49). Additional experiments must be designed to test various aspects of these models.