Multivalent interactions of calcium/calmodulin-dependent protein kinase II with the postsynaptic density proteins NR2B, densin-180, and alpha-actinin-2.

Dendritic calcium/calmodulin-dependent protein kinase II (CaMKII) is dynamically targeted to the synapse. We show that CaMKIIalpha is associated with the CaMKII-binding proteins densin-180, the N-methyl-D-aspartate receptor NR2B subunit, and alpha-actinin in postsynaptic density-enriched rat brain fractions. Residues 819-894 within the C-terminal domain of alpha-actinin-2 constitute the minimal CaMKII-binding domain. Similar amounts of Thr286-autophosphorylated CaMKIIalpha holoenzyme [P-T286]CaMKII bind to alpha-actinin-2 as bind to NR2B (residues 1260-1339) or to densin-180 (residues 1247-1495) in glutathione-agarose cosedimentation assays, even though the CaMKII-binding domains share no amino acid sequence similarity. Like NR2B, alpha-actinin-2 binds to representative splice variants of each CaMKII gene (alpha, beta, gamma, and delta), whereas densin-180 binds selectively to CaMKIIalpha. In addition, C-terminal truncated CaMKIIalpha monomers can interact with NR2B and alpha-actinin-2, but not with densin-180. Soluble alpha-actinin-2 does not compete for [P-T286]CaMKII binding to immobilized densin-180 or NR2B. However, soluble densin-180, but not soluble NR2B, increases CaMKII binding to immobilized alpha-actinin-2 by approximately 10-fold in a PDZ domain-dependent manner. A His6-tagged NR2B fragment associates with GST-densin or GST-actinin but only in the presence of [P-T286]CaMKII. Similarly, His6-tagged densin-180 or alpha-actinin fragments associate with GST-NR2B in a [P-T286]CaMKII-dependent manner. In addition, GST-NR2B and His6-tagged alpha-actinin can bind simultaneously to monomeric CaMKII subunits. In combination, these data support a model in which [P-T286]CaMKIIalpha can simultaneously interact with multiple dendritic spine proteins, possibly stabilizing the synaptic localization of CaMKII and/or nucleating a multiprotein synaptic signaling complex.

dominate in the brain and are involved in normal regulation of synaptic transmission (1)(2)(3)(4). Calcium/calmodulin-dependent autophosphorylation at Thr 286 (numbered as in CaMKII␣) enhances the affinity for calcium/calmodulin and confers autonomous (calcium-independent) activity after calmodulin dissociates. Consequently, CaMKII is capable of integrating information conveyed by the amplitude, frequency, and duration of local calcium transients to which it is exposed. Autophosphorylation at Thr 305/306 occurs only in the absence of bound calcium/calmodulin and blocks subsequent calmodulin binding (reviewed in Ref. 5).
Knock-in mutation of the Thr 286 or Thr 305/306 autophosphorylation sites in murine CaMKII␣ to Ala or Asp drastically alters some forms of hippocampal synaptic plasticity and disrupts spatial learning behaviors (reviewed in Ref. 6). The specificity of changes in synaptic responses to their dedicated inputs implies that postsynaptic actions of CaMKII, such as modulation of the trafficking and activity of AMPA-type glutamate receptors, are exquisitely regulated in a spatial and temporal manner (7)(8)(9).
CaMKII interacts with other proteins that we refer to collectively as CaMKII-associated proteins (CaMKAPs). At postsynaptic sites, these include multiple subunits of the N-methyl-D-aspartate-type glutamate receptor (NMDA receptor) (10 -13), densin-180 (14,15), ␣-actinin (15,16), cyclin-dependent protein kinase 5 (16), synGAP␤ (17), and filamentous actin (18,19). However, specific contributions of each of these interactions to neuronal signaling and the regulation of synaptic plasticity remain unclear (reviewed in Refs. [7][8][9]. To specifically address this issue, it is critical to have a thorough understanding of the molecular bases for these interactions and the factors regulating complex formation. Prior studies have identified dissimilar high affinity CaMKII-binding domains in the NR2B subunit of the NMDA receptor and in densin-180. These interactions are differentially regulated by calcium/calmodulinbinding, autophosphorylation at Thr 286 , and phosphorylation of the binding proteins (10, 11, 13-15, 20, 21). Moreover, densin-180 and NR2B do not compete with each other for binding to CaMKII (14). Interestingly, densin-180 makes an additional direct interaction with ␣-actinin (15), and ␣-actinin can bind to the NR1 and NR2B subunits of the NMDA receptor (21)(22)(23)(24). However, little is known about the molecular determinants for CaMKII binding to ␣-actinin. Here, we report an initial molecular dissection of the CaMKII-binding domain in ␣-actinin-2 and explore the relationship of this interaction to CaMKII binding with densin-180 and NR2B. The data indicate that CaMKII itself may serve as a structural scaffold for the assembly of a postsynaptic signalosome.
GST Cosedimentation Assays-Purified GST fusion proteins or GST alone (Ϸ250 nM full-length protein) were incubated for 2 h at 4°C with [P-T 286 ]CaMKII or Thr 287 -autophosphorylated CaMKII␤/␥/␦ (Ϸ250 nM) and glutathione-agarose beads (Sigma, 25 l of packed resin) in pull-down (PD) buffer (50 mM Tris-HCl, pH 7.5, 200 mM NaCl, 0.5% Triton X-100; 500-l final volume). Where indicated, His 6 -tagged fusion proteins (Ϸ250 nM each) were included with the kinase. Beads were sedimented by centrifugation and washed in 500 l of PD Buffer 6 times for 5 min each. Beads were transferred to new microcentrifuge tubes, and sedimented proteins were eluted with SDS-PAGE sample buffer, resolved by SDS-PAGE, and transferred to nitrocellulose membranes. Membranes were stained using Ponceau S and digitally scanned to quantify protein loading using ImageJ (rsb.info.nih.gov/ij/). Pilot studies showed 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 per 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. Statistical comparisons (one-way analysis of variance with Tukey post-hoc tests) were performed using Prism (GraphPad Software). To confirm the identity of Ponceau-stained proteins, membranes were immunoblotted for CaMKII or His 6 -tagged proteins using affinity purified polyclonal goat anti-CaMKII primary antibodies (25), or monoclonal mouse anti-His 6 antibodies (Amersham Biosciences), respectively, and alkaline-phosphatase-conjugated secondary antibodies (Jackson ImmunoResearch).
CaMKII Overlay Assays-Nitrocellulose membranes containing immobilized GST fusion proteins (1-5 g) were incubated with [ 32 P-T 286 ]CaMKII (50 nM subunit), washed and autoradiographed, essentially as described (25). Where indicated, [ 32 P-T 286 ]CaMKII was incubated on ice for 30 min with soluble GST fusion proteins before addition to the membrane. Films exposed within the linear range were digitally scanned, and band intensities (determined using ImageJ) were normalized to GST protein loading (see above).

CaMKII Is Associated with Multiple CaMKII-binding Proteins in PSD-enriched
Fractions-PSD-enriched synaptosomal fractions isolated from adult rat forebrain were enriched in CaMKII and three CaMKII-binding proteins, ␣-actinin, NR2B, and densin-180. About 50% of the CaMKII␣, but less than 20% of the CaMKAPs, were solubilized from the PSD-enriched fraction using Triton X-100. CaMKII immunoprecipitates from this solubilized extract also specifically con-tained ␣-actinin, densin-180, and occasionally NR2B (Fig. 1). Densin-180 immunoprecipitates also contained CaMKII and NR2B, but no detectable ␣-actinin (Fig. 1). These data confirm previous observations that CaMKII, ␣-actinin, NR2B, and densin-180 are enriched in PSDs. The effects of solubilization conditions and the differential regulation of each interaction 4 preclude definitive conclusions about the nature of specific protein-protein interactions in PSDs. However, the data are consistent with the existence of multiprotein complexes containing CaMKII in vivo.
Role of the C-terminal Association Domain of CaMKII␣-The role of the C-terminal domain in binding to each of the CaMKAPs was investigated using CaMKII␣ truncations at either residue 420 or residue 380. Both proteins are monomeric but can be Thr 286 -autophosphorylated to a similar extent as the wild type kinase using modified reaction conditions, as assessed by immunoblotting using phospho-Thr 286 -specific antibodies and by the generation of autonomous kinase activity (data not shown). We previously showed that CaMKII-(1-420) exhibited reduced binding to GST-NR2B (20), but under the conditions used here, both GST-NR2B and GST-actinin-(819 -894) bound with comparable efficacy to the monomeric kinases and the wild type holoenzyme. However, there was no statistically significant binding of either truncation mutant to GST-densin-(1247-1495) (Fig. 3, A and E). These data show that the NR2B and ␣-actinin-2 binding sites lie within the first 380 amino acids of CaMKII and that formation of a CaMKII holoenzyme is not essential for these interactions. However, CaMKII␣ binding to densin-180 appears to require an intact association domain.
Non-competitive Binding of Densin-180, NR2B, and ␣-Actinin-2 to CaMKII-Residues 819 -894 of ␣-actinin-2 share no obvious amino acid sequence similarity to the CaMKII-binding domains of NR2B or densin-180. To compare mechanisms for interactions of CaMKII with ␣-actinin-2, NR2B, and densin-180, competition overlay assays were performed. Nitrocellulose membranes containing immobilized GST fusion proteins were incubated with [ 32 P-T 286 ]CaMKII (50 nM) in the absence or presence of each of the soluble GST fusion proteins (up to 1.25 M). Binding of kinase to immobilized proteins was determined by autoradiography of the washed membranes.
As shown above, [ 32 P-T 286 ]CaMKII binds very weakly to immobilized ␣-actinin-2 in overlay assays. However, soluble GST-densin-(1247-1495) increased Ϸ8-fold the apparent interaction of CaMKII with immobilized GST-actinin-(819 -894), whereas soluble GST-NR2B and soluble GST-actinin-(819 -894) had no effect (Fig. 4). In combination, these data demonstrate that interactions of NR2B, densin-180, and ␣-actinin-2 with [P-T 286 ]CaMKII are non-competitive and suggest that densin-180 may facilitate binding of CaMKII to ␣-actinin-2. The lower panels show CaMKII immunoblots of proteins isolated on glutathione-agarose (P: 40% of total pellet) and the soluble proteins remaining in the supernatant (S: 2% of total). B, CaMKII overlay assay. GST fusion proteins were immobilized on nitrocellulose membranes and incubated with wild type [ 32 P-T 286 ]CaMKII. Membranes were washed and autoradiographed. C, quantification of wild type (WT) CaMKII␣ binding to the three GST fusion proteins in pull-down (top) and overlay (bottom) assays. Binding to GSTactinin and GST-densin was normalized to binding to GST-NR2B in each experiment and averaged across three to eight experiments; data are shown as mean Ϯ S.E. D and E, quantitative comparison of CaMKII isoform (D) and CaMKII␣ truncation mutant (E) binding to densin, NR2B, and ␣-actinin-2 in glutathione-agarose cosedimentation assays. Binding of each isoform or truncation mutant was normalized to wild type CaMKII␣ for each GST fusion protein, and then averaged across three to eight experiments. Data are expressed as mean Ϯ S.E. *, p Ͻ 0.001; ࡗ, p Ͻ 0.01 versus binding to wild type CaMKII␣.
CaMKII Simultaneously Binds to Multiple CaMKAPs-Because at least three CaMKAPs interact non-competitively with CaMKII, we hypothesized that CaMKII could simultaneously bind multiple CaMKAPs. To test this idea, GST, GST-NR2B, GST-densin-(1247-1542), or GST-actinin-(819 -894) were incubated with either [P-T 286 ]CaMKII, a mixture of His 6 -tagged versions of all three CaMKAPs, or all four proteins. Complexes harvested on glutathioneagarose were analyzed for bound CaMKII and His 6 -tagged proteins. GST alone did not interact with CaMKII or any of the His 6 -CaMKAPs (Fig. 6A).
GST-actinin-(819 -894) isolated comparable amounts of CaMKII in the presence and absence of the mixture of His 6 -tagged CaMKAPs. In addition, His 6 -densin associated with GST-actinin-(819 -894) in the presence and absence of [P-T 286 ]CaMKII, presumably due to CaMKIIindependent interaction of the PDZ domain of densin-180 with the C terminus of actinin. However, His 6 -NR2B and His 6 -actinin were isolated using GST-actinin-(819 -894) only in the presence of CaMKII, showing that the holoenzyme simultaneously bound to GST-actinin-(819 -894) and either His 6 -NR2B or His 6 -actinin.
To determine the role of the interaction between the PDZ domain of densin-180 and the C terminus of ␣-actinin in CaMKII-mediated assembly of this multiprotein complex, we performed similar assays using GST-densin-(1247-1405), which lacks the PDZ domain. GSTdensin-(1247-1405) bound CaMKII equally well in the presence or absence of the His 6 -tagged CaMKAPs, but bound the three His 6 -tagged CaMKAPs in the presence, but not the absence, of CaMKII (Fig. 6B). Therefore, a PDZ-mediated interaction of densin and ␣-actinin is not necessary for their simultaneous binding to CaMKII.
The role of CaMKII holoenzyme structure in assembly of this multiprotein complex was examined using the monomeric truncation mutant CaMKII␣-(1-420). Although CaMKII␣-(1-420) cannot bind densin-180, it is capable of interacting with both NR2B and ␣-actinin- (Fig. 3). CaMKII␣-(1-420) facilitated the association of His 6 -actinin with GST-NR2B on glutathione-agarose (Fig. 7). The amount of His 6actinin isolated was somewhat less than that recovered with the wild type holoenzyme in side-by-side experiments (not shown), but this might be expected, because there is only one available binding site for ␣-actinin when a CaMKII monomer binds to a GST-NR2B molecule, whereas binding of a CaMKII holoenzyme to GST-NR2B immobilizes 12 binding sites for ␣-actinin. Thus, single CaMKII subunits are capable of binding both NR2B and ␣-actinin simultaneously.
CaMKII was previously reported to bind ␣-actinin-1 and ␣-actinin-4 via a poorly defined C-terminal domain (15). Here, we show that CaMKII directly binds to ␣-actinin-2 in vitro and provide an initial characterization of the molecular determinants for interaction. Together, our data show that the interactions of CaMKII with NR2B, densin-180, and ␣-actinin are mechanistically distinct in multiple ways. First, the minimal CaMKII-binding domain identified in ␣-actinin-2 (residues 819 -894, Fig. 2) is substantially larger than, and shares no amino acid sequence similarity with, the minimal CaMKII-binding domains of NR2B and densin-180 (14,20). However, deletion of FIGURE 7. Simultaneous binding of NR2B and ␣-actinin-2 to monomeric CaMKII subunits. GST or GST-NR2B were incubated with [P-T 286 ]CaMKII-(1-420) alone, with His 6 -actinin-(547-894) alone, or with a mixture of both proteins, as indicated below. Complexes isolated on glutathione agarose (P: 40% of the total), as well as the remaining soluble proteins (S: 2% of the total), were resolved on SDS-polyacrylamide gels, transferred to nitrocellulose, and then sequentially stained with Ponceau S to detect GST fusion proteins and CaMKII␣ (top two rows) and immunoblotted for His 6 -tagged proteins (bottom row). Data are representative of three independent experiments. Complexes isolated on glutathione-agarose (P: 40% of the total), as well as the remaining soluble proteins (S: 2% of the total), were resolved on SDSpolyacrylamide gels, transferred to nitrocellulose, and then sequentially stained with Ponceau S to detect GST fusion proteins and CaMKII␣ (top two rows) and immunoblotted for His 6 -tagged proteins (bottom three rows). Data are representative of at least three independent experiments. sequences toward the extreme N and C termini of the CaMKII-binding domain that are highly conserved between three ␣-actinin isoforms abrogates CaMKII binding (Fig. 2). Second, in overlay assays, CaMKII binding to ␣-actinin-2 did not exceed background levels (presumably because ␣-actinin-2 failed to refold correctly on the membrane), whereas CaMKII bound effectively to NR2B and densin-180 (Fig. 3). This observation likely explains why initial overlay analyses of brain subcellular fractions failed to detect significant CaMKII-binding proteins with similar electrophoretic mobility to native ␣-actinin (25). Third, NR2B and ␣-actinin-2 interacted effectively with all four CaMKII isoforms, whereas densin-180 was selective for CaMKII␣ holoenzymes (Fig. 3). Fourth, activation of CaMKII by calcium/calmodulin-binding or autophosphorylation at Thr 286 is essential for the interaction with NR2B and potentiates interaction with densin-180 (10, 11, 13-15, 20, 21), whereas binding of CaMKII to ␣-actinin-2 is independent of Thr 286 autophosphorylation. 4 Last, these CaMKAPs do not compete for binding to CaMKII (Figs. 4 and 6). In combination, the data demonstrate that in contrast to AKAPs, which use a conserved molecular determinant (an amphipathic ␣-helix) to bind cAMP-dependent protein kinase regulatory subunits (34), CaMKAPs use divergent mechanisms to interact with CaMKII, with the potential for unique functional effects beyond simply dictating the subcellular localization of CaMKII.
We also explored the role of CaMKII holoenzyme structure in the interactions with ␣-actinin-2, densin-180, and NR2B. NR2B and ␣-actinin-2 bind monomeric truncation mutants of CaMKII␣. Thus, binding determinants for NR2B and ␣-actinin lie within the first 380 amino acids (Fig. 3), which contain the catalytic and regulatory domains. Consistent with these findings, Bayer et al. (13) have shown that catalytic domain mutations affect binding of CaMKII to NR2B, and yeast twohybrid studies have shown that ␣-actinin interacts with residues 1-316 of CaMKII␣ (15). In contrast, densin-180 does not bind to the monomeric CaMKII␣-(1-420) (Fig. 3). Previous studies using yeast two-hybrid approaches have shown that residues 1-316 of CaMKII␣ are incapable of binding to densin-180, and that residues 317-478 are sufficient for binding (15). These data suggest two hypotheses: 1) essential components of the densin-180 interaction domain lie within residues 421-478 of CaMKII; 2) holoenzyme structure is required for association of densin-180 with a domain within residues 317-420. Additional experiments will be required to resolve this issue.
Interactions of CaMKII with individual CaMKAPs have been suggested to play a role in targeting CaMKII within dendritic spines (reviewed in Refs. 4, 7, and 9). In particular, the dynamic regulation of CaMKII interactions with NR2B seems to parallel the transient translocation of CaMKII to PSDs in neurons stimulated electrically, or by glutamate (35,36). However, under some conditions CaMKII translocation to synapses is associated with the loss of synaptic NMDA receptors (37). Interactions with ␣-actinin-2 and/or densin-180 may further modulate CaMKII targeting to facilitate appropriate downstream signaling to substrates, such as AMPA receptor GluR1 subunits (38) and possibly NMDA receptors themselves (39). The CaMKAPs also may functionally modulate CaMKII in different ways (40). 4 It was previously proposed that CaMKII clusters synaptic proteins via diverse protein-protein interactions (1). The present data provide the first direct biochemical demonstration of this potential in that [P-T 286 ]CaMKII was shown to simultaneously interact with NR2B, densin-180, and ␣-actinin in vitro. GST-NR2B interacts with all three His 6tagged CaMKAPs in a CaMKII-dependent manner (Fig. 6). Similarly, GST-densin and GST-actinin isolated complexes containing CaMKII and the other CaMKAPs. Although close to stoichiometric amounts of CaMKII bound to each GST fusion protein in these experiments, it was difficult to detect His 6 -tagged proteins by protein staining of the membranes (not shown), indicating that sub-stoichiometric amounts were recovered. This may result from difficulties in maintaining multiple interactions throughout our rigorous washing procedure and is thus unlikely to reflect the actual stoichiometries of the complexes. Although NR2B and ␣-actinin can bind simultaneously to monomeric CaMKII subunits (Fig. 7), the ability of CaMKII to assemble the complex may in part depend on the holoenzyme structure, permitting independent interactions of individual subunits with CaMKAPs as well as physiological substrate proteins such as AMPA-type glutamate receptor GluR1 subunits (38). Taken together, our data suggest that CaMKII holoenzymes can nucleate multiprotein complexes in dendritic spines.
CaMKII␣ is highly abundant in neurons (e.g. 1-2% of total hippocampal protein), greater than might be expected for a strictly enzymatic role (47,48). Thus, nucleation of protein complexes by CaMKII holoenzymes may play one or more structural roles in dendritic spines. A variety of scaffolding proteins such as PSD-95 and CASK participate in multivalent interactions with cytoskeletal, receptor, and signaling proteins that are essential for proper targeting of receptor and signal transduction complexes (reviewed in Ref. 49). Recent work shows that isolated PSDs contain similar numbers of PSD-95 molecules and CaMKII holoenzymes (50). Interestingly, CASK contains an inactive kinase domain homologous to the catalytic domain of CaMKII that may have evolved from an ancestral calmodulin-dependent protein kinase to lose catalytic function while retaining the ability participate in protein-protein interactions (51,52). Because CaMKII interactions with CaMKAPs are dynamically modulated, 4 we propose that CaMKII may function as an autoregulated structural component, contributing to the reorganization of postsynaptic protein complexes depending on its activation state. These dynamic processes may play a role in the morphological changes in postsynaptic densities and dendritic spines following induction of synaptic plasticity (43,53,54), the ordered clustering of CaMKII in "towers" within the PSD (55), and/or the synaptic insertion of AMPA receptors (1). Further investigation of these structural roles will require selective manipulation of these interactions while preserving the catalytic function of the kinase.