Autophosphorylation-dependent Targeting of Calcium/ Calmodulin-dependent Protein Kinase II by the NR2B Subunit of theN-Methyl- d-aspartate Receptor*

Activation and Thr286autophosphorylation of calcium/calmodulindependent kinase II (CaMKII) following Ca2+ influx viaN-methyl-d-aspartate (NMDA)-type glutamate receptors is essential for hippocampal long term potentiation (LTP), a widely investigated cellular model of learning and memory. Here, we show that NR2B, but not NR2A or NR1, subunits of NMDA receptors are responsible for autophosphorylation-dependent targeting of CaMKII. CaMKII and NMDA receptors colocalize in neuronal dendritic spines, and a CaMKII·NMDA receptor complex can be isolated from brain extracts. Autophosphorylation induces direct high-affinity binding of CaMKII to a 50 amino acid domain in the NR2B cytoplasmic tail; little or no binding is observed to NR2A and NR1 cytoplasmic tails. Specific colocalization of CaMKII with NR2B-containing NMDA receptors in transfected cells depends on receptor activation, Ca2+influx, and Thr286 autophosphorylation. Translocation of CaMKII because of interaction with the NMDA receptor Ca2+channel may potentiate kinase activity and provide exquisite spatial and temporal control of postsynaptic substrate phosphorylation.

Mechanisms by which CaMKII is targeted to its postsynaptic substrates are poorly understood. Previous gel overlay analyses revealed a candidate PSD-associated CaMKII-anchoring protein, p190, that binds selectively to the Thr 286 -autophosphorylated kinase ([P-T286]CaMKII␣) (16). The NR2A and NR2B subunits of the NMDA receptor share several properties with this CaMKII-binding activity, including apparent size, enrichment in PSDs, and regional and developmental expression profiles 2 (17). Here, we demonstrate a direct and specific interaction between [P-T286]CaMKII␣ and NR2B and show that NR2B targets CaMKII in intact cells.
Generation and Analysis of NMDA Receptor Fusion Proteins-The entire cytoplasmic domains (C terminus starting immediately after transmembrane region IV) of NR1 (splice variant A containing both C1 and C2 exon cassettes), NR2A, and NR2B subunits, as well as shorter NR2B constructs, were subcloned from full-length cDNAs by polymerase chain reaction using Pfu polymerase and primers containing restrictions sites or by restriction digests. Fragments were sequenced and ligated into pRSET-A His 6 tag (Qiagen) or pGEX-2T glutathione Stransferase (GST) (Amersham Pharmacia Biotech) fusion vectors. His 6 tag fusions were expressed, and GST fusions were expressed and puri-* This work was supported by National Institutes of Health Grant GM47973 (to R. J. C.) and American Heart Association Grant-in-aid 96010040 (to R. J. C., an Established Investigator of the American Heart Association). Confocal microscopy was performed using the Vanderbilt University Medical Center Cell Imaging Resource (supported by National Institutes of Health Grants CA68485 and DK20593).The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  fied according to the manufacturers' instructions. His 6 tag fusion protein lysates were subjected to CaMKII overlay (see above) or immunoblotted with anti His 6 tag antibodies (CLONTECH) and 125 I-labeled secondary antibodies for expression levels, followed by PhosphorImager quantification.
GST Pull-down Analysis-GST fusion proteins were incubated (1 h, 4°C) with either purified CaMKII␣ (Fig. 2D, see caption) or with a freshly prepared rat brain cytosolic extract (ϳ3 mg/ml extract protein, 10 g/ml GST fusion protein) containing 2 M microcystin-LR and 0.5% Triton X-100, precipitated with glutathione-agarose, washed extensively, and eluted with SDS sample buffer. CaMKIV antibodies were from Transduction Laboratories.
HEK293 Cell Colocalization-HEK293 cells were seeded on coverslips in 35-mm dishes, transfected with a total of 3 g/dish DNA (1 g of SR␣ promotor-CaMKII␣ expression plasmid, 2 g of cytomegalovirus promotor plasmids with NMDA receptor subunits at a mass ratio of 1:3 NR1a and NR2A/B subunits), and grown for 48 h as described (21). Robust expression of NMDA currents was verified by patch-clamp recording of parallel cultures. 3 Cells were washed and incubated in Mg 2ϩfree Hanks' balanced saline containing 2 mM CaCl 2 and either the NMDA receptor antagonist 2-amino-5-phosphonovaleric acid (APV, 50 M) or NMDA/glycine (100/10 M) for 15 min. Cultures were fixed and processed for immunofluorescence (see above) using 1:500 antibody dilutions of goat anti-CaMKII (16), mouse anti-NR1 (PharMingen), and rabbit anti-NR2A/B (Chemicon). Between 2 and 5% of cells were strongly positive for at least one label; only those cells expressing high levels of each antigen (Ͼ50% of transfected cells) were included in the analyses. Under basal conditions, CaMKII␣ expression was diffusely cytoplasmic. Irrespective of agonist treatment, NR1 and NR2A/B strictly colocalized (mean scores Ͼ3.4, see below) in a patchy or reticular, often perinuclear pattern as seen previously in heterologous cells (22). Cultures were randomized prior to sampling digital images on a confocal microscope to prevent operator bias. Coded images (as in Fig.  3) were assigned a colocalization score by a second, naive observer: 0, mutual exclusion; 1, coincidental overlap; 2 or 3, increasing degrees of colocalization, 4, complete overlap of labels. For reference, the cells in Fig. 3 scored a 0, 1, 2, 2, and a 3 (from left to right, top to bottom).

RESULTS AND DISCUSSION
To determine whether NR2 subunits contribute to the previously characterized "p190" overlay binding activity (16), we analyzed immunoprecipitated NR2A/B by gel overlay with [ 32 P-T286]CaMKII␣ (Fig. 1A). A CaMKII-binding activity comigrating with NR2A and NR2B was immunoprecipitated with NR2A/B antibodies, but not control antibodies, indicating that NR2A and/or NR2B are CaMKII-binding proteins.
This interaction may be physiologically relevant, because triple immunofluorescent labeling of cultured cortical neurons demonstrated that CaMKII colocalizes with NMDA receptors in many punctae along dendritic shafts, identified as synapses by the adjacent or overlapping presence of synaptophysin (Fig.  1B). Higher magnification revealed a mostly postsynaptic localization of CaMKII in dendritic spines (Fig. 1C). Moreover, a complex of CaMKII with NMDA receptor subunits can be immunoprecipitated from PSDs using CaMKII antibodies, but not preimmune IgG (Fig. 1D). NR2B was more efficiently coprecipitated than NR1, likely because association of CaMKII with NR1 is indirect (i.e. via NR2B, see below). Recovery of the receptor-kinase complex required pretreatment of PSDs with a reversible cross-linker prior to essentially complete PSD solu-bilization in 2% SDS, indicating that the interaction of CaMKII with NMDA receptors is not stable in harsh detergents. The specificity of the cross-linking procedure was demonstrated by the absence of other abundant PSD proteins in the immunoprecipitate, including the catalytic subunit of protein phosphatase 1 (Fig. 1D).
NMDA receptor subunits have a common transmembrane topology with three membrane-spanning regions and a C-terminal tail of variable length, which forms the intracellular portion of the receptor ( Fig. 2A, diagram). Bacterial lysates expressing the cytoplasmic domains of the predominant forebrain NMDA receptor subunits, NR1, NR2A, and NR2B, as His 6 tag fusion proteins were screened for [ 32 P]CaMKII␣ binding by overlay (Fig. 2A). The NR2B cytoplasmic domain bound about six times more [ 32 P-T286]CaMKII␣ than the corresponding region of NR2A; neither NR1 nor any endogenous bacterial proteins showed detectable binding. Interactions with NR2A and NR2B were specific for autonomously active CaMKII, as   NR2B, not shown). B, cultured cortical neurons were triple-labeled with antibodies to CaMKII (red), NR1 (green), and the synaptic vesicle marker synaptophysin (blue) and imaged by confocal fluorescence microscopy. Arrowheads point to representative synapses where the three labels overlap (white in the merged image). C, double-immunofluorescence staining of a dendritic branch of a cultured cortical neuron demonstrates localization of CaMKII (red) in dendritic spines adjacent to synaptophysin-positive presynaptic terminals (blue). D, reversibly cross-linked PSDs (input) were solubilized in SDS and immunoprecipitated with either CaMKII antibodies or preimmune IgG, followed by immunoblotting with the indicated antibodies. Data are representative of at least three experiments.
CaMKII␣ phosphorylated in the absence of calcium/calmodulin at Thr 305/306 ([P-T306]CaMKII␣) bound only weakly (Ͻ5%). Because NR2B displayed the most robust interaction with CaMKII, we mapped its CaMKII-binding domain by creating a series of truncation and internal deletion constructs. Only constructs containing NR2B residues 1260 -1309 showed CaMKII binding similar to the full-length cytoplasmic tail. Fusion of NR2B-(1260 -1309) to GST demonstrated that this domain is also sufficient for interaction with autonomous CaMKII (Fig. 2B).
A solution interaction assay was employed to examine binding of CaMKII to NR2B that had not undergone denaturation/ renaturation for gel overlay analysis. [ 32 P-T286]CaMKII␣ bound saturably to a His 6 tag NR2B fusion protein containing residues 1260 -1309, but not to a construct that starts at resi-due 1310, C-terminal of this domain (Fig. 2C). Scatchard analysis indicated that binding involves a simple bimolecular interaction with a K d of 138 Ϯ 60 nM (n ϭ 3) (Fig. 2C, inset). This K d is ϳ100 times lower than the average concentration of CaMKII␣ in forebrain (16,23), suggesting that the interaction can readily occur in neurons.
The CaMKII-binding domain in NR2B contains a high-affinity phosphorylation site, Ser 1303 , which is phosphorylated by CaMKII in vitro and is also phosphorylated in vivo (13). However, three lines of evidence indicate that the binding of CaMKII to NR2B-(1260 -1309) is not dependent on a substrate interaction. First, the model peptide substrate syntide-2 only weakly inhibits CaMKII binding (ϳ30%) at concentrations of ϳ100-fold the K m for phosphorylation (not shown). Second, even though NR2A residues 1255-1298 are 36% identical to NR2B-(1260 -1309), and sequences surrounding the phosphorylation site are almost perfectly conserved (NR2B, LRRQH-SYD; NR2A, INRQHSYD) (13), CaMKII binding to NR2A-(1255-1298) is ϳ10-fold weaker under our overlay conditions (10.7 Ϯ 1.8%, n ϭ 3, Fig. 2B), suggesting that nonconserved residues in NR2B-(1260 -1309) are important for high-affinity CaMKII binding. Third, "pull-down" experiments, in which GST-NR2B fusion protein was purified with glutathione-agarose, showed that calcium/calmodulin alone did not promote CaMKII interaction with NR2B, but that stoichiometric interaction was instead strictly dependent on CaMKII␣ autophosphorylation at Thr 286 (Fig. 2D). On the other hand, calcium/ calmodulin binding is sufficient for full CaMKII activation, and Thr 286 autophosphorylation stabilizes the active conformation of the kinase in the absence of calcium/calmodulin (1,2). Thus, CaMKII residues outside the substrate binding site are involved in the interaction with NR2B.
Further evidence for specific association of CaMKII with NR2B was obtained by performing GST-NR2B pull-downs from brain cytosolic extracts. ␣ and ␤ isoforms of CaMKII were isolated following incubation with GST-NR2B-(1260 -1309), but not GST alone. Affinity-purified CaMKII␣ displayed an upward electrophoretic mobility shift characteristic of autophosphorylation (Fig. 2E). CaM kinase IV, a related kinase with a similar phosphorylation site preference (24), as well as other kinases and phosphatases tested, were not detected in the precipitated material, strongly indicating that NR2B-(1260 -1309) binds selectively to CaMKII.
The NR2B subunit of the NMDA receptor was shown to target Thr 286 autophosphorylated CaMKII in HEK293 cells. CaMKII␣ was coexpressed with various NMDA receptor subunit combinations, and their distributions were compared by immunofluorescence (Fig. 3). Whereas NR1 alone does not form functional NMDA receptors in HEK293 cells, activation of both NR1/NR2A and NR1/NR2B receptors leads to massive calcium influx (25). Coexpression of CaMKII␣ and NR1 alone resulted in low colocalization scores that were unaffected by acute treatment with the receptor coagonists NMDA/glycine (Fig. 3A). Perhaps reflecting the low but detectable CaMKII binding activity of NR2A (Fig. 2, A and B), additional expression of the NR2A subunit led to a small increase in CaMKII␣ and NR1/ NR2A colocalization, which was not significantly increased by NMDA/glycine treatment (Fig. 3B). In cells expressing NR2B with CaMKII␣ and NR1, we observed a similarly modest increase in colocalization in the absence of agonist treatment compared with CaMKII␣ and NR1 alone (Fig. 3, C and D). In contrast to NR2A-containing NMDA receptors, activation of NR1/NR2B receptors with NMDA/glycine caused a highly significant redistribution of CaMKII␣ into receptor-positive patches (Fig. 3, C and D), strongly suggesting that receptor activation induced the formation of a CaMKII⅐NR2B complex. Replacing extracellular calcium with barium, which is receptor-permeable but binds only poorly to calmodulin, completely blocked the effect of NMDA (Fig. 3D). Thus, opening of NMDA receptors is not sufficient for complex formation, but calcium influx is essential, presumably to stimulate calcium/calmodulindependent autophosphorylation of CaMKII. Consistent with this interpretation, an autophosphorylation-incompetent form of the kinase, T286A-CaMKII␣ (26,27), expressed at similar levels of wild-type CaMKII␣ failed to show activity-induced colocalization with NR1/NR2B containing NMDA receptors (Fig. 3D). Thus, NR2B mediates targeting of CaMKII to NMDA receptors in a calcium-and Thr 286 autophosphorylation-dependent manner in intact cells.
Our data support a model in which dendritic calcium influx induced by synaptic activity triggers CaMKII autophosphorylation at Thr 286 and subsequent binding to residues 1260 -1309 in the NR2B subunit of the NMDA receptor. What are the functional consequences of this interaction? Autonomous CaMKII in the PSD is inactivated by PSD-associated serine/ threonine phosphatases (18,28,29). Once dephosphorylated at Thr 286 , CaMKII positioned near the mouth of the NMDA receptor calcium channel is likely to undergo rapid re-autophosphorylation even during periods of low level NMDA receptor activation. Thus, an interaction of CaMKII with NMDA recep-tors is predicted to boosts autonomous kinase activity, leading to enhanced phosphorylation of nearby downstream effectors of synaptic plasticity (15). Furthermore, recruitment of CaMKII into the PSD structure (6), possibly via association with NR2B, may play a role in the rapid ultrastructural changes of synapses that undergo LTP (30,31). The developmental appearance of NR2A and down-regulation of NR2B in the mammalian visual system correlate with the end of the "critical period" of synapse maturation (32,33). Preferential association of CaMKII with NR2B over NR2A may therefore provide a mechanism by which NMDA receptor subunit composition can impact developmental plasticity.