Mechanism and Regulation of Calcium/Calmodulin-dependent Protein Kinase II Targeting to the NR2B Subunit of the N-Methyl-d-aspartate Receptor*

Calcium influx through theN-methyl-d-aspartate (NMDA)-type glutamate receptor and activation of calcium/calmodulin-dependent kinase II (CaMKII) are critical events in certain forms of synaptic plasticity. We have previously shown that autophosphorylation of CaMKII induces high-affinity binding to the NR2B subunit of the NMDA receptor (Strack, S., and Colbran, R. J. (1998) J. Biol. Chem. 273, 20689–20692). Here, we show that residues 1290–1309 in the cytosolic tail of NR2B are critical for CaMKII binding and identify by site-directed mutagenesis several key residues (Lys1292, Leu1298, Arg1299, Arg1300, Gln1301, and Ser1303). Phosphorylation of NR2B at Ser1303 by CaMKII inhibits binding and promotes slow dissociation of preformed CaMKII·NR2B complexes. Peptide competition studies imply a role for the CaMKII catalytic domain, but not the substrate-binding pocket, in the association with NR2B. However, analysis of monomeric CaMKII mutants indicates that the holoenzyme structure may also be important for stable association with NR2B. Residues 1260–1316 of NR2B are sufficient to direct the subcellular localization of CaMKII in intact cells and to confer dynamic regulation by calcium influx. Furthermore, mutation of residues in the CaMKII-binding domain in full-length NR2B bidirectionally modulates colocalization with CaMKII after NMDA receptor activation, suggesting a dynamic model for the translocation of CaMKII to postsynaptic targets.

NR2B cDNA (in a cytomegalovirus promoter-driven mammalian expression vector). Point mutants were generated by PCR using Pfu Turbo polymerase (Stratagene) and sense and antisense primers harboring the mutation as well as diagnostic silent restriction sites according to instructions supplied with the QuikChange kit (Stratagene). A cassette-based approach was used for the construction of internal deletion mutants and the B2A mutant. After disruption of the 3Ј-EcoRI site used for subcloning the NR2B-(1260 -1339) fragment, unique silent restriction sites were introduced by PCR (see above) at the following NR2B amino acids: 1286 -1288 (EcoRI), 1297-1298 (HindIII), and 1308 -1310 (BglII). Mutagenic sense and antisense primers with compatible overhangs were ligated into cassettes generated by cutting the NR2B construct with two of the three restriction enzymes. Since NR2B-(1310 -1339) includes at least one in vitro phosphorylation site for CaMKII in addition to Ser 1303 (data not shown), GST-NR2B mutants analyzed for Ser 1303 phosphorylation were truncated to NR2B-(1260 -1316) by digestion with SfiI and NdeI and fill in/religation. GST-NR2B fusion proteins were bacterially expressed and purified on glutathioneagarose according to standard protocols.
CaMKII Mutants-The CaMKII-(1-420) truncation mutant was PCR-amplified from the murine CaMKII␣ cDNA with a sense primer incorporating a BamHI site and an antisense primer containing an EcoRI site and in-frame stop codon. The ⌬380 -420 internal deletion mutant was generated by three-step "loop-out" PCR utilizing primers spanning the deletion. Mutagenized cDNAs were ligated into the pVL1393 baculovirus transfer vector (Invitrogen). Sf9 cells were infected with recombinant baculovirus, and protein was expressed and purified by calmodulin-agarose affinity chromatography as described (11). Point mutants of murine CaMKII␣ were generated by PCR as described for NR2B mutants and subcloned into the pME18S mammalian expression vector (chimeric simian virus 40/retrovirus (SR␣) promoter-driven; DNAX).
Mitochondrion-targeting Protein (MTP)-The basis of this multidomain fusion protein is the mammalian green fluorescent protein (GFP) expression vector pEGFP-N1 (CLONTECH). The GST coding sequence including the C-terminal multiple cloning site, but excluding the stop codons from pGEX-2T, was PCR-amplified using a sense primer with an XhoI adaptor and an antisense primer with a HindIII adaptor. The GST fragment was ligated into pEGFP-N1, whose EcoRI and BamHI sites had previously been removed by fill in/religation, to create a GST-GFP fusion plasmid.
Oligonucleotides encoding a mitochondrion-targeting sequence, the 15 amino-terminal amino acids of hexokinase I (16), and a Myc epitope tag were ligated into N-terminal NheI and BglII sites, resulting in a hexokinase-Myc-GST-GFP fusion cDNA. Wild-type or mutant CaMKIIbinding domains in the context of NR2B-(1260 -1316) were ligated into BamHI and EcoRI sites between GST and GFP coding sequences. The resulting hexokinase-Myc-GST-NR2B-GFP fusion plasmid resulted in the expression of a 60-kDa protein and mitochondrion-localized GFP fluorescence (see Fig. 7A), demonstrating that the protein was expressed intact in cells. Transfection of plasmids encoding CaMKII and NR2B mutants led to expression of proteins of the correct size at levels similar to those of the wild type. Sequences of all constructs were verified using an ABI 310 fluorescence sequencer at Center for Molecular Neuroscience, Vanderbilt University Medical Center.

Microtiter Plate Solution Binding
This solution binding assay is a modification of the Ni 2ϩ -coated microtiter plate assay previously described (12) using GST fusion proteins adsorbed to glutathione-coated 96-well plates (Pierce) as the binding surface. Briefly, plates were adsorbed for 2-16 h with GST fusion proteins at room temperature or 4°C (25 g/ml, 200 l/well, ϳ50% of binding capacity) in wash buffer (5 mg/ml bovine serum albumin, 200 mM NaCl, 50 mM Tris, pH 7.5, 0.1% Tween 20, 5 mM ␤-mercaptoethanol, and 0.1 mM EDTA). After extensive washes, 200 l/well [ 32 P-T286]CaMKII␣ diluted in wash buffer was allowed to bind to the tethered fusion protein for 2 h at room temperature, followed by 10 -12 more washes. Bound [ 32 P-T286]CaMKII was solubilized in 1% SDS, 0.2 N NaOH, and 10 mM EDTA and quantified by liquid scintillation counting. Nonspecific binding to GST alone (5-20% of the total, same as wash buffer without GST) was subtracted from total binding to obtain specific binding.

CaMKII⅐NR2B Dissociation Assays
[ 32 P-T286]CaMKII␣ and GST-NR2B-(1260 -1316) wild-type or S1303A fusion protein were incubated (30 min, 4°C) at 1-2 M each in binding buffer (200 mM NaCl, 50 mM Tris, pH 7.5, 0.25 mg/ml bovine serum albumin, 0.1% Triton X-100, 1 mM dithiothreitol, and 1 mM EDTA). After addition of 0.1 volume of a 50% glutathione-agarose slurry and 15 min of continued incubation, CaMKII⅐NR2B complexes were recovered by brief centrifugation and washing in binding buffer. The agarose slurry was resuspended in dissociation buffer (0.5 ml of binding buffer with NaCl concentration reduced to 100 mM to permit efficient phosphorylation or dephosphorylation) supplemented with Mg⅐ATP or protein phosphatase as described in the figure legends and rotated at 25 or 30°C. Aliquots were removed at the indicated time points and analyzed for soluble and glutathione-agarose-bound CaMKII by immunoblotting and/or autoradiography. [ 32 P-T286]CaMKII␣ dephosphorylation was quantified by adjusting aliquots to 20% (w/v) trichloroacetic acid and liquid scintillation counting the supernatant after high-speed microcentrifugation.

HEK293 Cell Colocalization
HEK293 cells were seeded on coverslips (no. 1) in 35-mm dishes, transfected at 40 -70% confluency with a total of 4 -6 g/dish DNA (2 g of CaMKII␣ expression plasmid plus either 2 g of MTP plasmid or 2 g of each NR1a and NR2B cytomegalovirus promoter plasmids) using TransIT-LT1 transfection reagent (Panvera) according to the manufacturer's instructions, and grown for 48 h in minimal essential medium with 10% fetal bovine serum and 1 mM glutamine. In experiments with MTP, dishes were either immediately fixed for immunofluorescence analysis or first incubated for variable amounts of time with 2 M calcium ionophore A23187 (Sigma) in growth medium. When NMDA receptor subunits were transfected, the growth medium was supplemented with the NMDA receptor antagonist 2-amino-5-phosphonovaleric acid (APV; 1 mM), and cells were washed and incubated in Mg 2ϩ -free Hanks' balanced saline buffered with 20 mM Hepes, pH 7.5, containing 2 mM CaCl 2 and either 50 M APV or NMDA/glycine (100/10 M) for 15 min prior to fixation. Cultures were fixed in 1:1 acetone/ methanol for 10 -15 min at 25°C and processed for immunofluorescence using 1:500 antibody dilutions of goat anti-CaMKII antibody (11) and either mouse anti-Myc tag or rabbit anti-NR2A/B (Chemicon International, Inc.) antibody as described (12,17). Cultures were randomized and coded prior to sampling digital images on a confocal microscope, so the degree of colocalization between signals for CaMKII and MTP or NR2B could be estimated blindly. Cells (15-25/dish) were assigned a colocalization score from 0 to 4: 0, mutual exclusion; 1, coincidental overlap; 2 and 3, increasing degrees of colocalization; and 4, complete overlap of labels (12).

RESULTS
CaMKII binding to NR2B is mediated by 50 amino acids (positions 1260 -1309) in the NR2B C terminus (12). To investigate whether CaMKII binding to intact PSDs and to NR2B occur by similar mechanisms, co-sedimentation binding experiments (10) were performed in which [ 32 P-T286]CaMKII␣ was allowed to interact with isolated native PSDs in the presence of increasing concentrations of NR2B-(1260 -1309) fused to GST. GST-NR2B-(1260 -1309) potently (IC 50 ϭ 50 nM) inhibited cosedimentation of CaMKII with PSDs ( Fig. 1). Inhibition was specific to NR2B because a GST fusion protein with the corresponding region of NR2A, which does not bind appreciably to CaMKII (12), had no effect. Although it is possible that NR2B allosterically interferes with the CaMKII/PSD association, we consider it more likely that a single domain in CaMKII interacts with both PSDs and NR2B since the two binding events share a dependence on CaMKII autophosphorylation and have similar affinities (10,12). The 30% residual binding observed at the highest NR2B concentrations could either be nonspecific (there is no meaningful blank for this assay) or reflect CaMKII associating with PSDs via a separate mechanism.
Analysis of the CaMKII-binding Domain in NR2B-To further elucidate the molecular determinants for CaMKII binding in NR2B, we initially narrowed down the CaMKII-binding domain by constructing a series of overlapping GST fusion proteins. Binding of [ 32 P-T286]CaMKII␣ by overlay depended on the presence of NR2B residues 1290 -1309 ( Fig. 2A), which thus constitute the "core" CaMKII-binding domain. However, we cannot formally rule out redundant stabilizing effects of NR2B residues 1260 -1289 and 1310 -1339, even though deleting either flanking region by itself had little effect on CaMKII binding. To further define the core domain, small internal deletion mutants were analyzed in the context of NR2B-(1260 -1339). Whereas deletion of residues C-terminal of Ser 1303 (⌬1304 -1307 and ⌬1306 -1309) had a modest to no effect, incremental deletions of N-terminal amino acids from positions 1291 to 1296 reduced CaMKII binding by up to 75% (Fig. 2B).
NR2B residues conserved in NR2A were also subjected to mutational analysis. Mutation of the phosphorylation site Ser 1303 to Ala had little effect, whereas introduction of a negative charge (S1303D) or a hydrophobic side chain (S1303L) severely interfered with the CaMKII interaction (75-80% reduction). NR2B Leu 1298 and Arg 1300 align with the predicted CaMKII substrate recognition motif (I/L)XRXX(S/T) (19). Replacing Arg 1300 with Glu or Gln diminished binding by Ͼ85%, and the L1298A mutation almost completely obliterated the interaction with CaMKII, similar to the R1300Q/S1303D double mutant. As expected, the conservative substitution mutant L1298I displayed CaMKII binding indistinguishable from the wild-type. In agreement with White et al. (19), who noted a preference of CaMKII for substrates containing Gln at position Ϫ2, replacing Gln 1301 with Ala in NR2B reduced CaMKII binding by 50%. Binding to GST-NR2A-(1255-1298) wild-type (w.t.) fusion protein, which contains the region homologous to the core CaMKII-binding domain in NR2B (identical residues are shown in gray), is shown for comparison. Mutations are grouped into internal (Int.) deletion mutants, mutants in which NR2B amino acids were replaced with corresponding residues in the NR2A sequence (reversal mutants), and mutations of the CaMKII phosphorylation site Ser 1303 (18) and more N-terminal residues that are part of the consensus CaMKII phosphorylation site (Arg 1300 and Leu 1298 ) (19). Plotted are means Ϯ S.E. of three to six experiments expressed as percent binding to GST-NR2B-(1260 -1339) wild-type fusion protein. C, the indicated GST-NR2B-(1260 -1316) fusion proteins were adsorbed to glutathionecoated 96-well plates (5 g/well) and incubated with the indicated concentrations of [ 32 P-T286]CaMKII␣; bound ligand was quantified by liquid scintillation counting. Data are representative of four similar experiments.
To verify that critical mutations affect CaMKII binding, as opposed to the ability of fusion proteins to renature on the blot prior to CaMKII overlay, solution-phase binding assays were performed with native GST-NR2B fusion proteins. Whereas the S1303A mutant bound [ 32 P-T286]CaMKII␣ similarly to the wild type, the R1300Q, R1300E, and B2A mutants were severely binding-impaired ( Fig. 2C and data not shown), confirming results from overlay experiments. None of the NR2B mutants displayed specific binding to CaMKII autophosphorylated at Thr 305/306 in the absence of Ca 2ϩ /calmodulin (data not shown), in agreement with previous results demonstrating that interaction with NR2B requires autophosphorylation at the autonomy site, Thr 286 (12).
The finding that NR2B amino acids important for the interaction with CaMKII include the substrate recognition motif (Ser 1303 , Arg 1300 , and Leu 1298 ) prompted us to examine whether peptide substrates are effective competitors for the CaMKII/NR2B interaction. Syntide-2, derived from a CaMKII phosphorylation site in glycogen synthase, at concentrations of up to 400 M (20 times the K m for phosphorylation (20)) did not compete for CaMKII binding to NR2B (Fig. 3B), in agreement with previous, more qualitative data (12). In contrast, autocamtide-2, a peptide substrate modeled after the autoregulatory domain surrounding Thr 286 in CaMKII␣, was an effective competitor, with an IC 50 of 10 M similar to the K m for its phosphorylation (21). A possible explanation for this dramatic difference in inhibitory potency is provided by an alignment of the two peptides with the corresponding sequence in NR2B, revealing more extensive homology of NR2B to autocamtide-2 compared with syntide-2 N-terminal of the phosphorylated residue (Fig. 3A). In particular, autocamtide-2 contains residues corresponding to the mutation-sensitive NR2B residues Gln 1301 and Arg 1299 but syntide-2 does not.
To investigate whether NR2B Ser 1303 phosphorylation not only inhibited initial association of CaMKII, but also could dissociate CaMKII previously bound to NR2B, release of CaMKII from CaMKII⅐GST-NR2B complexes was monitored with or without ATP. In the absence of ATP, CaMKII remained stably associated with GST-NR2B for Ͼ1 h under these conditions (Fig. 5). Addition of ATP led to stoichiometric phosphorylation of NR2B Ser 1303 by 5 min, revealed by the appearance of a lower mobility band (Figs. 4 and 5A, compare asterisks). Phosphorylation of NR2B was accompanied by dissociation of CaMKII, albeit incomplete and with a protracted time course (10% released after 75 min). No ATP-dependent release of CaMKII from complexes with the NR2B S1303A mutant was detected (Fig. 5B), demonstrating that this dissociation is a consequence of NR2B Ser 1303 phosphorylation, as opposed to continued, calcium/calmodulin-independent autophosphorylation of CaMKII.
Autophosphorylation of CaMKII at Thr 286 is required for high-affinity binding to PSDs and to NR2B (10,12). We therefore investigated the reversibility of the CaMKII/NR2B interaction by dephosphorylation of Thr 286 . Incubation of kinase⅐NR2B complexes with a 2 g/ml concentration of the catalytic subunit of protein phosphatase 1 resulted in Ͼ70% dephosphorylation of [ 32 P-T286]CaMKII␣ in 2 h, measured as release of trichloroacetic acid-soluble radioactivity (Fig. 5C), as well as a decrease in immunoreactivity with a phospho-Thr 286specific CaMKII antibody (data not shown). Paralleling the time course of dephosphorylation, ϳ20% of the CaMKII⅐NR2B complexes dissociated during this time period. Higher protein phosphatase 1 concentrations (10 g/ml) led to release of up 40% CaMKII under otherwise identical conditions (data not shown). Both dephosphorylation and dissociation were blocked by inhibiting protein phosphatase 1 with microcystin, demonstrating that Thr 286 dephosphorylation promotes release of CaMKII from NR2B.
Monomeric CaMKII Mutants Do Not Bind to NR2B-In an effort to delineate domains in CaMKII important for the interaction with NR2B, we constructed deletion mutants of CaMKII␣. Two mutations in the C-terminal domain required for formation of a holoenzyme consisting of 10 -12 subunits (22,23) were expressed in insect cells. The 1-420 mutant lacks the C-terminal 58 amino acids, whereas the ⌬380 -420 mutant lacks 41 residues in the middle of the oligomerization domain (Fig. 6A). As expected, both deletions disrupt holoenzyme formation since the mutants migrated as monomers by gel filtration chromatography and sucrose gradient centrifugation (data not shown). Both mutants underwent calcium/calmodulin-dependent autophosphorylation at Thr 286 and attained levels of autonomous activity similar to those of the wild type (30 -50% of calcium/calmodulin-dependent activity). However, whereas autophosphorylation of wild-type CaMKII is rapid (seconds), occurring between adjacent subunits of the same holoenzyme (24), maximal autophosphorylation of the monomeric mutants required high enzyme concentration and prolonged incubation (1-3 min) at 30°C, consistent with an intermolecular reaction. Both mutants displayed normal catalytic activity toward syntide-2 peptide substrate (wild-type: K m ϭ 12.3 M, K cat ϭ 324 min Ϫ1 ; 1-420: K m ϭ 13.2 M, K cat ϭ 236 min Ϫ1 ; and ⌬380 -420, K m ϭ 14.8 M, K cat ϭ 348 min Ϫ1 ; n ϭ 2-3) and phosphorylated GST-NR2B-(1260 -1316) with comparable efficiency (Fig. 6B). Moreover, a detailed comparison of GST-NR2B-(1260 -1316) phosphorylation kinetics failed to reveal significant differences between the wild type and the ⌬380 -420 mutant (wild-type: S 0.5 ϭ 0.80 Ϯ 0.20 M, K cat ϭ 9.0 Ϯ 3.0 min Ϫ1 , K cat /S 0.5 ϭ 11.3 min Ϫ1 M Ϫ1 ; and ⌬380 -420: S 0.5 ϭ 1.28 Ϯ 0.25 M, K cat ϭ 10.8 Ϯ 2.4 min Ϫ1 , K cat /S 0.5 ϭ 8.4 min Ϫ1 M Ϫ1 ; n ϭ 3). 2 Although CaMKII oligomerization mutants were catalytically normal, their ability to bind to NR2B was severely compromised. In qualitative GST co-sedimentation assays, GST-NR2B-(1260 -1316) pulled down stoichiometric amounts of wild-type CaMKII in an autophosphorylation-dependent manner (Fig. 6C, inset), in agreement with previous data (12). In contrast, only small amounts of the 1-420 mutant were detected in the glutathione-agarose pellet, even though both kinases displayed similar levels of autophosphorylation detected with a phospho-Thr 286 -specific CaMKII antibody. Similar data were obtained with the ⌬380 -420 mutant; neither kinase bound to GST alone (data not shown). Glutathione microtiter plate assays confirmed these results, showing little or no binding of either monomeric [ 32 P-T286]CaMKII␣ mutant to GST-NR2B-coated wells (Fig. 6C).
Regulation of CaMKII Targeting to NR2B in Cells-In previous cell transfection studies, CaMKII was shown to colocalize with NR2B, but not with NR2A-containing NMDA receptors. This colocalization depended on receptor agonist treatment, calcium influx, and Thr 286 autophosphorylation (12). In the present study, we set out to investigate whether the CaMKIIbinding domain in NR2B identified in in vitro assays is sufficient to localize CaMKII in cells. To this end, a mitochondriontargeted multidomain fusion protein (MTP) was constructed, containing 15 N-terminal amino acids of hexokinase I that bind to the outer mitochondrial membrane protein porin (16), other domains useful in detecting expression, and a multiple cloning site for insertion of additional sequences (Fig. 7A). MTPs with or without various NR2B inserts were coexpressed with CaMKII in HEK293 cells, and colocalization was assayed by immunofluorescence confocal microscopy (see "Experimental Procedures"). Expression of MTP without an NR2B insert did not affect CaMKII localization; the kinase remained diffusely cytosolic, as evidenced by low colocalization scores (Fig. 7B). Likewise, insertion of NR2B-(1260 -1316) into MTP did not cause a redistribution of CaMKII to mitochondria. However, when a MTP containing the NR2B S1303A mutant fragment was expressed, CaMKII assumed a discrete mitochondrial localization, reflected in near-perfect colocalization scores.
To investigate the calcium dependence of translocation, we examined the colocalization of CaMKII to the MTP-NR2B wildtype fusion protein at various times after addition of the calcium ionophore A23187 to the medium. Low but significant colocalization was observed after 2 and 5 min of ionophore treatment, returning to base line after 15 min (Fig. 7B). The R1300Q mutant of NR2B-(1260 -1316), which is binding-defective in vitro (Fig. 2, B and C), was not able to target CaMKII to mitochondria at any time point.
These data are consistent with the following interpretation: NR2B Ser 1303 is substantially phosphorylated in HEK293 cells under basal conditions, and CaMKII autophosphorylation levels are insufficient for targeting to Ser 1303 -phosphorylated NR2B, but sufficient for promoting colocalization with the non-phosphorylatable S1303A mutant. Calcium mobilization increases CaMKII autophosphorylation, allowing it to transiently overcome the inhibitory effect of Ser 1303 phosphorylation. Alternative, more complex scenarios involving, for exam-ple, calcium-activated phosphatases can also be invoked to explain these results.
The question of whether amino acids identified as important in binding assays with NR2B fusion proteins are also important for the agonist-dependent translocation of CaMKII to NMDA receptor ion channels was addressed next. HEK293 cells were cotransfected with NR1, NR2B, and CaMKII, challenged for 15 min with co-agonists NMDA and glycine or the receptor antagonist APV, fixed, and analyzed for colocalization of CaMKII and NR2B. In agreement with previous results (12), NMDA/glycine treatment led to a 2-fold increase in the CaMKII/NR2B colocalization score compared with APV treatment (Fig. 7D). Changing six amino acids in full-length NR2B to the corresponding residues in NR2A (B2A mutant) completely abolishes the activity-induced increase in CaMKII colocalization, providing a molecular explanation for the inability of NR1/NR2A receptor activation to recruit CaMKII (12). Likewise, the single point mutant R1300E abrogated activity-dependent recruitment of CaMKII into NR2B patches. Intriguingly, but in complete agreement with the MTP data (Fig. 7B), significant colocalization of CaMKII with NR2B S1303A was apparent even in the absence of receptor activation, and optical overlap was further increased by NMDA/glycine treatment. These data further support the notion that NR2B Ser 1303 phosphorylation is a negative feedback regulator of CaMKII binding. In apparent contrast to the in vitro binding experiments showing a 75% reduction of [ 32 P-T286]CaMKII binding to the NR2B S1303D mutant (Fig. 2B), this mutation transferred into the full-length subunit did not appreciably affect colocalization scores compared with wild-type NR2B. A possible explanation for this discrepancy consistent with the data on mitochondriontargeted NR2B is that Ser 1303 is highly phosphorylated in cells.
A previous report suggested that NMDA-dependent translocation of GFP-CaMKII to PSDs in cultured hippocampal neurons does not require autophosphorylation, but that calcium/ calmodulin binding to the kinase is sufficient (25). For this reason, we reexamined the autophosphorylation dependence of the CaMKII/NR2B colocalization in cells. A calmodulin binding-defective CaMKII mutant (A302R) (26), an ATP hydrolysisdefective, "kinase-dead" CaMKII mutant (K42R), and a nonphosphorylatable CaMKII mutant (T286A) were not able to translocate to NMDA receptor patches after receptor activation (Fig. 7D), confirming our previous result with the T286A mutant that translocation to NMDA receptors requires Thr 286 autophosphorylation (12).

DISCUSSION
The CaMKII-binding Domain in NR2B-This report presents the first detailed characterization of a targeting domain for CaMKII, an abundant kinase important in learning and memory. The NR2B subunit of the NMDA receptor is unique in its CaMKII-targeting function since NR1 and NR2A alone cannot specify the subcellular localization of CaMKII (12), although other laboratories have previously reported CaMKII interaction with NR1 and NR2A in vitro (13)(14)(15).
A relatively small region in NR2B including amino acids between positions 1290 and 1303 was shown to be critical for CaMKII targeting in intact cells. Data reported by Leonard et al. (15) imply the existence of two separate CaMKII-binding domains in the cytosolic tail of NR2B, one between residues 839 and 1120 and the other C-terminal of residue 1120 (presumably corresponding to the domain characterized herein). Our previous domain mapping experiments (12) are not consistent with the existence of the domain between residues 839 and 1120, and the transfection studies in this report demonstrate that residues surrounding the phosphorylation site Ser 1303 in NR2B are both necessary and sufficient to direct CaMKII localization. . CaMKII colocalizes with NR2B S1303A, but not with NR2B R1300E, in discrete puncta (arrowheads, yellow in merged image), as shown in images of representative cells that received NMDA/glycine treatment on the right. Scale bars ϭ 10 m. *, significant increase (p Ͻ 0.0001) by two-tailed Student's t test compared with control (wild-type NR2B under control conditions). Furthermore, we have not been able to demonstrate colocalization of CaMKII with MTPs containing residues 839 -1120 of NR2B (data not shown).
Although the degree to which NR2B itself contributes to the interaction of CaMKII with PSDs is unclear, competition experiments suggest that the mechanisms of CaMKII binding to NR2B and to intact PSDs are similar (Fig. 1). An understanding of the targeting determinants in NR2B may thus aid in identifying additional CaMKII-targeting proteins involved in physiological and pathophysiological translocation of CaMKII to PSDs and other cytoskeletal structures (10,25,(27)(28)(29).
Mechanism of CaMKII Interaction with NR2B-The mechanism of NR2B interaction with CaMKII has not been entirely resolved. Residues N-terminal to Ser 1303 are clearly important to the interaction, including several that are not conserved in the corresponding region of NR2A (e.g. Lys 1292 and Arg 1299 ). However, the residues most sensitive to single point mutations (Ser 1303 , Arg 1300 , and Leu 1298 ) are key determinants of a consensus CaMKII phosphorylation site (I/L)XRXX(S/T) (19). These data might be interpreted to suggest that CaMKII and NR2B interact via an enzyme/substrate mechanism. However, several pieces of data suggest that the interaction is more complex. For example, autophosphorylation at Thr 286 is essential for interaction with NR2B, but substrate phosphorylation can be stimulated by calcium/calmodulin binding in the absence of autophosphorylation (e.g. in a T286A mutant (21,30)). Furthermore, monomeric CaMKII mutants that phosphorylate Ser 1303 comparably to the wild-type kinase are unable to form stable complexes with GST-NR2B (Fig. 6).
Our studies using synthetic peptides to compete for the interaction provide some insight into the mechanism. Syntide-2, a "pure" peptide substrate based on a CaMKII phosphorylation site in glycogen synthase, is unable to compete for [P-T286]CaMKII binding to NR2B at 20 times its K m concentration, arguing that a typical enzyme/substrate interaction is not critical for stable binding. In contrast, and somewhat surprisingly, autocamtide-2, another peptide substrate for CaMKII based on the sequence surrounding the Thr 286 autophosphorylation site, potently competes for binding (Fig. 3B). Our interpretation of this observation rests on the previous demonstration that autocamtide-2 interacts with two distinct sites on the Thr 286 -autophosphorylated catalytic domain of CaMKII (31). The first site (site A) appears to be a conventional substrate-binding site. The second site (site B) is thought to be occupied by the Thr 286 autophosphorylation site within the CaMKII autoinhibitory domain in its non-phosphorylated state. Interestingly, a peptide corresponding to the autoinhibitory domain of CaMKII (residues 281-302) also potently inhibits the CaMKII/NR2B interaction (data not shown). Thus, one hypothesis is that autophosphorylation of CaMKII exposes site B for binding to NR2B. Although the high degree of identity between NR2B residues surrounding Ser 1303 and autocamtide-2 supports this idea, an indirect mechanism is also possible, in which site B occupancy (by autocamtide-2 or the autoinhibitory domain) allosterically occludes a third interaction site for NR2B. A recent mutagenesis study provides insights into the identity of residues that constitute site B and its location in the CaMKII catalytic domain (32), and it will be important to test whether site B mutations disrupt the CaMKII/NR2B interaction.
Although these peptide competitor studies implicate the catalytic domain in the interaction with NR2B, the monomeric CaMKII mutants containing fully functional catalytic domains are unable to form stable stoichiometric complexes with GST-NR2B, although weak interactions can be detected (Fig. 6). It is possible that monomeric mutants are missing amino acids that are making direct contacts with NR2B, in addition to intersubunit contacts in the dodecameric holoenzyme. However, we currently favor the interpretation that CaMKII oligomerization has a permissive role, increasing binding affinities by allowing for concerted binding of multiple catalytic domains in the CaMKII holoenzyme to multiple NR2B subunits. This model suggests an intriguing mechanism by which the CaMKII holoenzyme may be preferentially attracted to membranes with a threshold density of NR2B subunits, present in either the same tetrameric or pentameric NMDA receptor complex or in separate adjacent receptors.
A Dynamic Model for CaMKII/NR2B Association-Regardless of the mechanism of CaMKII/NR2B association, the cell transfection experiments demonstrate that NR2B sequences can efficiently redistribute CaMKII in cells. Moreover, the enhanced targeting of the NR2B S1303A mutant combined with the biochemical data provides strong evidence for a modulatory role of this serine residue. Thus, the CaMKII/NR2B interaction is inversely controlled by two phosphorylation events, i.e. enhanced by CaMKII Thr 286 autophosphorylation and inhibited by NR2B Ser 1303 phosphorylation. However, NR2B phosphorylated at Ser 1303 appears to retain residual affinity for CaMKII, as evidenced by overlay assays (Fig. 4) and by slow dissociation of CaMKII from NR2B after phosphorylation (Fig. 5) and as suggested by the activity-dependent targeting of CaMKII by the NR2B S1303D mutant (Fig. 7D). Furthermore, the extremely low turnover rate (K cat ) of NR2B Ser 1303 phosphorylation by CaMKII (ϳ10 min -1 compared with 300 min Ϫ1 for syntide-2) implies that CaMKII⅐NR2B complexes exist for several seconds before phosphorylated NR2B can dissociate, even in the presence of a large excess of non-phosphorylated NR2B.
Phosphatases add another layer of complexity to the regulation of the CaMKII/NR2B interaction. Although in vitro data suggest a negative role for protein phosphatases by returning CaMKII to its non-phosphorylated state (Fig. 5C), the in vivo situation is undoubtedly more complex since dephosphorylation of NR2B Ser 1303 is expected to increase the affinity for CaMKII. Thus, the lifetime of the kinase⅐channel complex appears to be controlled by the balance of phosphatase activities toward CaMKII Thr 286 and NR2B Ser 1303 . Whereas CaMKII Thr 286 is dephosphorylated by protein phosphatase type 1 or 2A depending on its subcellular localization (33), the identity of the NR2B Ser 1303 phosphatase is unknown.
Induction of LTP triggers persistent CaMKII autophosphorylation that is dependent on NMDA receptor activation (34 -36). Synaptic activity also induces translocation of CaMKII to PSDs (10,25) and increases co-immunoprecipitation of NMDA receptor subunits with CaMKII (15). Data presented here provide important insights into the molecular mechanism for the association of CaMKII with NMDA receptors. These studies also have implications for our understanding of CaMKII association with the PSD, although other proteins may be involved in addition to NR2B. The reversibility of the CaMKII/NR2B interaction by phosphorylation/dephosphorylation suggests that CaMKII targeted to the PSD structure may not remain permanently anchored to NR2B. Instead, NR2B could act to increase the local concentration of the kinase in the cytoskeletal lattice of the PSD, where it is more likely to be activated by subsequent calcium entry through the NMDA receptor. In addition, CaMKII released from NR2B may diffuse to other important postsynaptic substrates, such as the GluR1 glutamate receptor subunit (10,36), to bring about long-term changes in synaptic efficacy (37). Future studies will also address whether CaMKII affects NMDA receptor activity, through either binding to the NR2B C terminus or phosphorylation of Ser 1303 .