αCaMKII and PSD-95 differentially regulate synaptic expression of NR2A and NR2B-containing NMDA receptors in hippocampus

Abstract

NMDA receptors (NMDARs) are critical determinants of bidirectional synaptic plasticity, however, studies of NMDAR function have been based primarily on pharmacological and electrophysiological manipulations, and it is still debated whether there are subunit-selective forms of long-term potentiation (LTP) and long-term depression (LTD). Here we provide ultrastructural analyses of axospinous synapses in CA1 stratum radiatum of transgenic mice with mutations to two key underlying postsynaptic density (PSD) proteins, PSD-95 and αCaMKII. Distribution profiles of synaptic proteins in these mice reveal very different patterns of subunit-specific NMDAR localization, which may be related to the divergent phenotypes of the two mutants. In PSD-95 PDZ3 truncated mice in which LTD could not be induced but LTP was found to be enhanced, we found a subtle, yet preferential displacement of synaptic NR2B subunits in lateral regions of the synapse without affecting changes in the localization of NR2A subunits. In persistent inhibitory αCaMKII T305D mutant mice with severely impaired LTP but stable LTD expression, we found a selective reduction of NR2A subunits at both the synapse and throughout the cytoplasm of the spine without any effect on the NR2B subunit. In an experiment of mutual exclusivity, neither PSD-95 nor αCaMKII localization was found to be affected by mutations to the corresponding PSD protein suggesting that they are functionally independent of the other in the regulation of NR2A and NR2B-containing NMDARs preceding synaptic activity. Consequently, there may exist at least two distinct PSD-95 and αCaMKII-specific NMDAR complexes involved in mediating LTP and LTD through opposing signal transduction pathways in synapses of the hippocampus. The contrasting phenotypes of the PSD-95 and αCaMKII mutant mice further establish the prospect of an independent and, possibly, competing mechanism for the regulation of NMDAR-dependent bidirectional synaptic plasticity.

In the hippocampus, LTP and LTD are two related, but physiologically opposing forms of synaptic plasticity proposed to modulate the acquisition and storage of information in synapses (Stanton and Sejnowski, 1989, Bliss and Collingridge, 1993). The cellular and molecular changes leading to LTP and LTD are initiated by ionotropic glutamate receptors of the N-methyl-D-aspartate (NMDA) receptor subtype, the major gateway for calcium influx in the spine required to modify the structure, function, and efficacy of synaptic connections (Monyer et al., 1992, Linden, 1999, Lisman, 2003). Importantly, the subunit stoichiometry of the tetrameric receptor determines the gating and channel conductance properties of the assembled complex, the best studied examples being the NR2A and NR2B-containing NMDARs which exhibit fast and slow gating of the receptor, respectively (Hollmann and Heinemann, 1994, Cull-Candy et al., 2001). Targeted disruptions and C-terminal truncations of the NR2A and NR2B subunits had initially suggested distinct functional properties and intracellular signaling pathways, however, recent studies show that LTP, in particular, can be mediated by both subunits (Sakimura et al., 1995, Kutsuwada et al., 1996, Mori et al., 1998, Sprengel et al., 1998, Kohr et al., 2003). Still, there have been a number of conflicting studies on the subunit-specific roles of NMDARs in bidirectional synaptic plasticity while studies investigating the regulation of subunit composition of receptors at the synapse have remained sparse (Liu et al., 2004, Massey et al., 2004, Berberich et al., 2005).

Receptor targeting and trafficking by PSD-95-like scaffolding proteins belonging to the family of membrane-associated guanylate kinases (MAGUKs) is generally recognized as one of the primary mechanisms responsible for regulating the expression of NMDA receptors at the synapse (Losi et al., 2003, Sans et al., 2003). PSD-95 is a well-characterized MAGUK which has been shown to interact directly and indirectly with membrane-associated glutamate receptors, a multitude of proteins in the PSD as well as corresponding anchoring proteins associated with the presynaptic structure (Husi et al., 2000, Sheng and Lee, 2000, Valtschanoff and Weinberg, 2001). In the absence of PSD-95 at the PSD in PSD-95^PDZ12 mutants, LTP was found to be enhanced while LTD was completely abolished, suggesting that PSD-95 is required to modify synaptic transmission through its interactions at the synapse (Migaud et al., 1998). Other studies have shown that overexpression of PSD-95 prevents LTP while enhancing LTD, presumably through the trafficking of α-amino-3-hydroxy-5-methyl-4-isoxazole proprionic acid (AMPA) receptors (AMPARs) to the synapse (Beique and Andrade, 2003, Stein et al., 2003).

The other purely superficial mechanism of NMDAR regulation is through receptor modification and interaction with kinases and phosphatases associated with calcium signaling, in which αCaMKII plays a dominant role (Hudmon and Schulman, 2002, Colbran, 2004). αCaMKII is a serine/threonine kinase thought to be an important regulator of NMDAR function resulting in synaptic potentiation (Lisman et al., 2002, Chen et al., 2005). The currently accepted model of αCaMKII activation is by synaptic activation of NMDARs, indirectly through calcium influx and through a direct interaction with the NMDAR subunits (Strack et al., 2000, Bayer et al., 2006). Autophosphorylation and subsequent translocation to the PSD allows αCaMKII to directly interact with NMDAR subunits; this not only enables modifications to the receptor to modulate channel properties and signaling pathways, but also prolongs calcium-independent activity of the αCaMKII holoenzyme (Bayer et al., 2001, Bayer et al., 2006). In CaMKII^T305D mutant mice with reduced autophosphorylation and reduced PSD-association, LTP was severely impaired but LTD was stable (Elgersma et al., 2002). Other studies with αCaMKII mutants either preventing or mimicking autophosphorylation resulted in blocking LTP or shifting the system more favorably towards LTD at lower frequencies, respectively (Mayford et al., 1995, Giese et al., 1998). In addition to its molecular role in LTP, αCaMKII is an ideal candidate to compare to PSD-95 because of its proposed competitive interactions with PSD-95 (Gardoni et al., 2001).

The contrasting phenotypes of the PSD-95^PDZ12 and CaMKII^T305D mutants, therefore, lead us to suspect that there exists a dimorphic regulation of subunit-specific NMDARs by these two PSD proteins. Through these mutations, we establish functionally distinct interactions of the NR2A and NR2B subunits with αCaMKII and PSD-95, respectively, which may comprise two separable NMDAR complexes involved in synaptic plasticity.

Experimental Procedures

Animals

We used six C57BL/6 mice comprising three wild type mice (n=3) and three mice homozygous (n=3) for the targeted mutation in the PSD-95 gene, and referred to as PSD-95^PDZ12 for the intact portion of the three PDZ domains, previously described in (Migaud et al., 1998). The ages of the mice ranged from 7-10 months and all were male. Additionally, thirteen C57BL/6 mice comprising six wild type mice (n=6) and seven mice homozygous (n=7) for the T305 substitution in the αCaMKII gene, referred to as CaMKII^T305D and previously described in (Elgersma et al., 2002), were used. The ages of these mice ranged from 3-8 months and all were female. Neither set of mice had been stimulated in electrophysiological experiments or put through training paradigms to assess learning prior to this study. All procedures and experiments were conducted in accordance with the Institutional Animal Care and Use Committee of Mount Sinai School of Medicine and University of California, Los Angeles, as well as the Wellcome Trust Sanger Institute Ethical Review Board under the approval of the Home Office regulations (PPL 80/1867 & 60/2662) and the Animals (Scientific Procedures) Act, 1986.

Tissue Preparation

Animals were perfused transcardially in 1% paraformaldehyde for 1 min., followed by 4% paraformaldehyde with 0.125% glutaraldehyde for 12 min. To avoid ischemic conditions resulting in PSD thickening (Dosemeci et al., 2001), the time of incision to perfusion was kept at less than 1 minute. Sections (500 μm-thick) were cut using a Leica vibratome and subsequently cut into hippocampal CA1 blocks, 1 × 1 mm in dimension. This was followed by freeze substitution and low-temperature embedding of the specimens, previously described by (Adams et al., 2001). Ultrathin sections were cut using a diamond knife on a Reichert-Jung ultramicrotome and mounted on nickel mesh grids using an adhesive pen. Grids with ultrathin sections were then pretreated with a saturated solution of NaOH in absolute ethanol followed by 6 min. in 0.1% sodium borohydride and 50 mM glycine. Following a 30 min. incubation in blocking buffer (0.005 M Tris-buffered saline (TBS) containing 2% HSA), sections were then incubated overnight in appropriate antibodies in blocking buffer with polyethylene glycol (5 mg/ml).

Antibody Specificity and Selection

The following primary antibodies used were affinity purified and have been extensively characterized in prior studies. Polyclonal anti-NMDA receptor 2A antibodies (Upstate Biotechnology, Lake Placid, NY, catalog # 07-632) were raised in rabbits to a hexahistidine-tagged fusion protein to the C-terminus of mouse NR2A, amino acids 1265-1464 (Lau and Huganir, 1995). Immunoblot analysis of rat brain microsomal preparations probed with anti-NR2A detected the appropriate 170 kDa band. Polyclonal anti-NMDA receptor 2B antibodies (Novus Biologicals, Littleton, CO, catalog # NB300-106) were raised in rabbits to a fusion protein to amino acids 1253-1391of the C-terminal region of rat NR2B (Moriyoshi et al., 1991). Western blot of rat brain hippocampus probed with anti-NR2B detected a 180 kDa band with no cross-reactivity with NR2A. Specificity of the anti-NR2A and anti-NR2B antibodies was additionally verified in both western blot analysis across a panel of HEK293 cells transfected with all NMDA receptor cDNAs (Adams et al., 2004), and in blocking experiments in which immunolabeling for anti-NR2A and anti-NR2B was blocked by pre-adsorption with the immunogens used to generate the respective antibodies.

Polyclonal αCaMKII antibodies (Chemicon, Temecula, CA, now Millipore, catalog # AB3111) were raised in rabbits to a synthetic rat αCaMKII peptide corresponding to amino acids 7-20 conjugated to KLH (Erondu and Kennedy, 1985). SDS-PAGE immunoblots detected CaMKII isoforms of 50-60 kDa and recombinant rat αCaMKII was detected in sf9 insect cells. Cross-reactivity with β isoforms is not expected but overlap with the δ and γ isoforms is predicted to occur due to conservation over this N-terminus region. δ and γ isoforms, however, are weakly expressed in the hippocampus and there have been no evidence of any functional roles in excitatory synapses of pyramidal neurons in the hippocampus. Polyclonal anti-PSD-95 antibodies (Zymed Laboratories Inc., San Francisco, CA, now Invitrogen, catalog # 51-6700) were raised in sheep to a fusion protein to amino acids 1-386 of the N-terminus region of human PSD-95 (Cho et al., 1992). Western blots of rat brain homogenates or lysates of PSD-95 transfected HEK293 cells confirmed positive reactivity of PSD-95 to the 95 kDa band. Cross-reactivity with other MAGUKs and endogenous proteins was absent.

Postembedding Immunogold Immunocytochemistry

For PSD-95^PDZ12 mutant and control groups, sections were incubated with either NR2B (13.3 μg/ml) or NR2A antibody (33.3 μg/ml) and αCaMKII (1 μg/ml) in blocking buffer for 25 hours, followed by washing with TBS. Sections were then incubated with goat-anti-rabbit IgG Fab2 10 nm gold-tagged secondary antibody (NR2B 1:20; NR2A 1:50; αCaMKII 1:40; Electron Microscopy Sciences) in blocking buffer (TBS containing 2% HSA and polyethylene glycol; 5 mg/ml) for 90 min. and washed with distilled water. Sections were then counterstained with 1% uranyl acetate for 45 min. in the dark and 0.3% lead citrate for 3 min.

For αCaMKII^T305D mutant and control groups, sections were incubated with either NR2B (6.7 μg/ml) or NR2A antibody (25 μg/ml) and PSD-95 (12.5 μg/ml) in blocking buffer for 25 hours, followed by washing with TBS. Sections were then incubated with goat-anti-rabbit IgG Fab2 10 nm gold-tagged secondary antibody (NR2B 1:40; NR2A 1:40) or donkey-anti-sheep IgG 10 nm gold-tagged secondary antibody (PSD-95 1:40) in blocking buffer (TBS containing 2% HSA and polyethylene glycol; 5 mg/ml) for 90 min. and washed with distilled water. These sections were also counterstained with 1% uranyl acetate for 45 min. in the dark and 0.3% lead citrate for 3 min.

Image Analysis and Quantitation

Images were obtained using a Jeol 1200EX electron microscope (Tokyo, Japan) and imaged with an Advantage CCD camera (Advanced Microscopy Techniques Corp., Danvers, MA). For each labeling experiment, conditions were met such that labeling was uniform and predominantly synaptic. All synapses collected were isolated to stratum radiatum of hippocampus CA1 where perforated synapses are typically minimal (Nicholson et al., 2006). Unbiased selection of approximately the first 75-100 synapses were collected per animal and per condition at 30,000x across two separate, non-serial sections, and only those synapses that fit a predesigned criteria were included in the analyses. The criteria for this study were: 1) synapses must be axospinous asymmetric synapses; 2) there must be a clear delineation between the pre- and postsynaptic elements as well as synaptic membranes separating the two; 3) the cleft and PSD must be clearly defined, lacking any tangential cuts; and 4) there must be two or more gold particles (GPs) in the PSD or cleft. EM images were then analyzed using a software program called SynBin to quantify the number and distribution of GPs into designated bins assigned on the basis of proximity to pre- and postsynaptic membranes, the structure of the PSD, and other theoretical constraints of resolution and synaptic topography (Adams et al., 2001, Moga et al., 2006).

Bin Designations

Although we were primarily concerned with compartments belonging to the postsynaptic spine, bins were created to account for all compartments of the synapse, both presynaptic and postsynaptic. Presynaptic bins included the 0-30 nm and 30-60 nm active zone bins, corresponding to the PSD, as well as the remainder of the presynaptic terminal (not shown). The spine was partitioned into cleft, 0-30 nm synaptic, 30-60 nm PSD, and nonsynaptic cytoplasmic bins in order to establish association of proteins to the synaptic membrane and the PSD (Fig. 1A and 1B). Since immunogold localization with 10 nm GPs have a theoretical resolution of 15-25 nm in radius, we have designated the 0-30 nm synaptic bin as definitive synaptic labeling.

Figure 1. SynBin and synaptic bin designations.

A representative EM image of a synapse at 30,000x reveals the presynaptic terminal, presynaptic vesicles of the active zone, the electron-dense postsynaptic density, and the postsynaptic spine (A). Outlines of the PSD and pre and postsynaptic membranes illustrate bin assignments using SynBin image analysis software (B and C). The postsynaptic spine is divided into 0-30 nm synaptic, 30-60 nm PSD, and nonsynaptic compartments perpendicular to the postsynaptic membrane (B). This analysis reflects the laminar localization of proteins at the synapse. Bins are redesigned to reflect the lateral arrangement of receptor subunits and associated proteins across the synaptic surface (C). The PSD remains divided into 0-30 nm synaptic and 30-60 nm PSD bins, however, the 0-30 nm synaptic bin has been subdivided into middle and lateral 50% bins and an extrasynaptic bin extending 125 nm from either edge of the PSD has also been included. Scale bar = 100 nm.

Bins directly adjacent to the 0-30 nm synaptic bin may also be construed as synaptic, however, due to the resolution of the GPs these bins were considered buffers to true synaptic localization. Thus, labeling in the cleft and 0-30 nm active zone of the presynaptic terminal, directly abutting the postsynaptic membrane, may be attributed to cross-over from synaptic labeling. The 30-60 nm PSD bin, conversely, may serve as a transitional area where pools of synaptic protein lay in reserve to be inserted or have been internalized from the synapse, or where higher order, downstream signaling and structural protein interactions take place. Consequently, GPs falling into this versatile bin can belong to either the synapse, PSD or the nonsynaptic pool and are, therefore, not regarded as purely synaptic. In our analyses then, segregation of these three compartments represent different states of expression, activity, and trafficking.

In re-designating bin assignments from the laminar distribution of the synapse to a lateral distribution, parallel to the surface of the synaptic membrane, we were able to look at the lateral profiles of proteins at the synapse. To establish lateral distribution profiles across the synapse, we subdivided the 0-30 nm synaptic bin into increments of non-cumulative and non-overlapping 10% bins, from the lateral 10% to the medial 40-50% bin from either edge of the PSD (Fig. 3A inset and 4D inset). However, because SynBin does not distinguish laterality, the lateral 0-20%, 20-40%, 40-60%, 60-80% and 80-100% bins were divided by two to illustrate the 10% subdivisions across the synapse; probability of labeling should be the same on either side of the synaptic midline. From average PSD lengths, each 10% bin, therefore, represents roughly 25-30 nm of the synapse or the resolution of a single GP. Since the absence or presence of an effect in a single 10% bin may lack any meaning, we also divided the synapse equally into the middle 50% and the lateral 50%, accounting for 25% on either edge of the PSD (Fig. 1C). Logically, this would reveal any obvious lateral reorganization at the synapse; biologically, this divides the synapse into two areas which have been proposed to affect differentially synaptic plasticity (Massey et al., 2004, van Zundert et al., 2004). As a measure of extrasynaptic changes, we have also included an extrasynaptic (ES) bin extending out 125 nm from either edge of the PSD. Given that the average PSD length of synapses analyzed across all groups was approximately 250-300 nm (data not shown), the combined extrasynaptic 250 nm bin approximates the same length and area as the 0-30 nm synaptic or 30-60 nm PSD bins.

Figure 3. Synaptic reductions of NR2B in PSD-95^PDZ12 mutant mice and NR2A in CaMKII^T305D mutant mice are laterally distinct.

(A) Electron micrographs show representative NR2A labeling in the synapse (outlined) of a PSD-95^PDZ12 mutant and a wild type littermate. Lateral distribution of NR2A across the synapse is quantified in the adjacent graph where the 0-30 nm synaptic bin is subdivided into 10% bins (inset). For comparison, all bins are normalized to the bin with the highest label. The lateral profile of NR2A in PSD-95^PDZ12 mutants (white diamonds) and wild type littermates (black squares) display no genotype differences, but reveals a fairly uniform distribution across the synapse.

(B) Lateral distribution is further grouped into middle 50% and lateral 50% bins, with a proportional extrasynaptic (ES) 250 nm bin for comparison. All bins are normalized to wild type middle 50% synaptic bins where labeling is highest within experiments and across groups. In PSD-95^PDZ12 mutant mice (white bars), there is no change in NR2A localization to middle, lateral, or extrasynaptic bins compared to wild type littermates (black bars). There is, however, significantly greater localization in both middle and lateral bins compared to extrasynaptic bins (Middle:ES; ***p < 0.0001 for both (−/−) and (wt); Lateral:ES; ***p < 0.0001 (−/−) and **p = 0.003 (wt); ANOVA with post hoc bonferroni test), but no difference between the two bins.

(C) Comparative EM images depict lateral displacement of NR2B in the synapse (outlined) of a PSD-95^PDZ12 mutant compared to a wild type littermate. Quantification of NR2B distribution reveals a significant 16% decrease specifically in the lateral 10-40% of the synapse (*p < 0.05) in PSD-95^PDZ12 mutants (grey diamonds) compared to wild type littermates (black squares). Individually, however, only the lateral 10-20% bin reaches significance (*p < 0.05), with a 15% decrease. NR2B distribution at the synapse is typically dumbbell-shaped, with two peaks on either side of the centerline, but in the absence of PSD-95 this shape is lost.

(D) NR2B is reduced 15% in the lateral 50% of the synapse (**p = 0.012) in PSD-95^PDZ12 mutants (grey bars) compared to wild type littermates (black bars). Comparisons across regions show significantly greater NR2B localization to the middle 50% than to the lateral 50% (Middle:Lateral; **p = 0.003 (−/−) and ***p < 0.0001 (wt)) or the extrasynaptic 250nm bin (Middle:ES; ***p < 0.0001 for both (−/−) and (wt); Lateral:ES; ***p < 0.0001 for both (−/−) and (wt); ANOVA with post hoc bonferroni test).

(E) Comparative EM images illustrate synaptic reduction of NR2A (outlined) in a CaMKII^T305D mutant compared to a wild type littermate. Although significance is only attained in the lateral 0-20% bin (p < 0.05), as a 23% reduction, lateral NR2A distribution reveals a separation in all bins across the synapse between CaMKII^T305D mutant mice (white diamonds) and wild type littermates (black squares). As in PSD-95^PDZ12 mutants, NR2A is evenly distributed across the synapse.

(F) Quantification of NR2A distribution shows a cumulative 16% reduction in the middle 50% of the synapse (*p < 0.05) in CaMKII^T305D mutants (white bars) compared to wild type littermates (black bars). Bin comparisons show NR2A localization is significantly greater in both middle and lateral bins compared to extrasynaptic bins (Middle:ES; ***p < 0.0001 for both (−/−) and (wt); Lateral:ES; ***p < 0.0001 for both (−/−) and (wt); ANOVA with post hoc bonferroni test), however, NR2A is reduced to a greater extent in the lateral 50% compared to the middle 50% bin in CaMKII^T305D mutants (Middle:Lateral; **p = 0.005 (−/−); ANOVA with post hoc bonferroni test).

(G) Representative electron micrographs show similar NR2B localization in the synapse (outlined) of a CaMKII^T305D mutant and wild type littermate that is clustered laterally. Lateral NR2B distributions in synapses of CaMKII^T305D mutant (grey diamonds) and wild type littermates (black squares) exhibit no genotype differences, however, NR2B is highly localized to medial regions of the synapse, rapidly declining laterally, but still retaining the dumbbell-shaped curve characteristic of NR2B distribution. Scale bar = 100 nm.

(H) There is no difference in NR2B localization to middle, lateral, or extrasynaptic bins between CaMKII^T305D mutants (grey bars) and wild type littermates (black bars). There is, however, significant differences in localization between middle, lateral, and extrasynaptic bins, decreasing in that order (Middle:Lateral; ***p < 0.0001 for both (−/−) and (wt); Middle:ES; ***p < 0.0001 for both (−/−) and (wt); Lateral:ES; ***p < 0.0001 for both (−/−) and (wt); ANOVA with post hoc bonferroni test). Error bars represent means ± SEM.

Figure 4. PSD-95 and αCaMKII are functionally and synaptically expressed independently of the other.

(A) Total αCaMKII (white bars) and PSD-95 (grey bars) localization in presynaptic terminals and postsynaptic spines averaged per group, in PSD-95^PDZ12 mutant and CaMKII^T305D mutant compared to their respective wild type littermates. Labeling is normalized to respective wild type postsynaptic bins for comparison between αCaMKII and PSD-95. There are no genotype differences in αCaMKII or PSD-95 labeling between mutant and wild type littermates, however, postsynaptic labeling is consistently higher than presynaptic labeling (**p = 0.003, ***p < 0.0001; ANOVA with post hoc bonferroni test).

(B) Synaptic bins are normalized to the wild type 0-30 nm synaptic bin, consistent with synaptic NR2 subunit localization. In PSD-95^PDZ12 mutant mice (white diamonds), no change is associated with αCaMKII localization in synaptic bins compared to wild type littermates (black squares), however, αCaMKII labeling reveals a preferentially nonsynaptic arrangement (SYN:NONSYN; **p = 0.007 (wt); ANOVA with post hoc bonferroni test). A reference figure illustrating superficial 0-30 nm synaptic PSD, deeper layer 30-60 nm PSD, and nonsynaptic, cytoplasmic bins (inset).

(C) In CaMKII^T305D mutant mice (grey diamonds), no changes are observed in PSD-95 localization in synaptic bins compared to wild type littermates (black squares). The highly synaptic distribution of PSD-95 resembles that of NR2A and NR2B (SYN:PSD; ***p < 0.0001 for both (−/−) and (wt); SYN:NONSYN; ***p < 0.0001 for both (−/−) and (wt); ANOVA with post hoc bonferroni test).

(D) Electron micrographs show representative αCaMKII labeling in the synapse of a PSD-95^PDZ12 mutant and a wild type littermate that is equally distributed between synaptic (outlined) and nonsynaptic bins of the spine. The lateral profile of αCaMKII reveals a fairly homogeneous, cosine-function-like distribution across the synapse, and there are no genotype differences between PSD-95^PDZ12 mutant mice (white diamonds) and wild type littermates (black squares) in any of the 10% bins. For comparison, all bins are normalized to the bin with highest label. A reference figure corresponding to 10% bins within the 0-30 nm synaptic bin (inset). Scale bar = 100 nm.

(E) There is no difference in αCaMKII localization to middle, lateral, or extrasynaptic bins in PSD-95^PDZ12 mutant mice (white bars) compared to wild type littermates (black bars). Localization to middle and lateral synaptic bins is significantly higher than to extrasynaptic bins (Middle:ES; *p < 0.05 (−/−) and ***p < 0.0001 (wt); Lateral:ES; **p = 0.005 (−/−) and ***p = 0.0001 (wt); ANOVA with post hoc bonferroni test), but αCaMKII is relatively evenly distributed between middle and lateral bins. Bins are normalized to the middle 50% bin, consistent with lateral NR2 subunit localization.

(F) Electron micrographs show representative PSD-95 labeling in the synapse of a CaMKII^T305D mutant and a wild type littermate that is primarily synaptic and closely associated to the synaptic membrane(outlined). The lateral profile of PSD-95 reveals highly synaptic labeling in medial regions of the synapse, in close alignment to that of NR2B. No genotype differences were found for PSD-95 localization between CaMKII^T305D mutant mice (grey diamonds) and wild type littermates (black squares). Scale bar = 100 nm.

(G) No difference in PSD-95 localization to middle, lateral, or extrasynaptic bins in CaMKII^T305D mutants (grey bars) compared to wild type littermates (black bars) is evident. There is, however, significantly higher localization in middle compared to lateral bins and between middle and lateral synaptic bins compared to extrasynaptic bins (Middle:ES; ***p < 0.0001 for both (−/−) and (wt); Lateral:ES; ***p < 0.0001 for both (−/−) and (wt); ANOVA with post hoc bonferroni test). Error bars represent means ± SEM.

Statistics

Six separate experiments were performed over a period of two years, with each labeling experiment of each antibody in each set of mice accounting for a single experiment. Comparisons were only made within individual experiments since immunolabeling conditions varied between tissue from PSD-95^PDZ12 and CaMKII^T305D groups and between experimental sets for the same antibody. Thus, multiple comparisons across experiments were not practical and all data reflect genotype differences for their respective immunolabeling experiments. Additionally, the two sets of mice were of different genders; the PSD-95^PDZ12 group were all males and the CaMKII^T305D group all females. However, comparisons of NR2A and NR2B lateral distributions with a third set of mice, SAP102 hemizygous mice which consisted of all males, showed no consistent gender differences between males and females with respect to control groups (data not shown).

All genotype comparisons were analyzed using a two-tailed independent samples t-test in SPSS 12.0 (SPSS Inc., Chicago, IL). Where distributions were not normal, as was the case in certain lateral bin distributions across the synaptic surface, we used the nonparametric Mann-Whitney U test as the test of comparison between the two groups. Normal distributions were determined using the Shapiro-Wilks' test and equality of variance was determined using Levene's test, both found in SPSS 12.0. Multiple bin comparisons between 0-30 nm synaptic, 30-60 nm PSD, and nonsynaptic bins and between Middle 50%, Lateral 50%, and ES 250 nm bins were analyzed using a one-way ANOVA with post hoc bonferroni test in SPSS 12.0. 0-30 nm synaptic : 30-60 nm PSD and 0-30 nm synaptic : nonsynaptic comparisons are denoted SYN:PSD and SYN:NONSYN, respectively. Middle : Lateral, Middle : ES, and Lateral : ES comparisons are denoted as such. Differences were considered to be significant at an α level of 0.05 and all values are given in mean ± SEM.

Results

Synaptic distribution of NR2B-containing NMDARs in wild type and PSD-95^PDZ12 mice

Initial analysis of NR2A and NR2B labeling shows that total synaptic levels, measured either presynaptically or postsynaptically, are unchanged between PSD-95^PDZ12 mutant and wild type littermates (Fig. 2A). Bin comparisons, however, reveal that NR2A and NR2B are preferentially localized postsynaptically, with minimal presynaptic labeling (***p < 0.0001; Fig. 2A), and both are localized primarily to the plasma membrane-associated 0-30 nm synaptic bin as compared to the deeper laminations of the 30-60 nm PSD and nonsynaptic bins (*p < 0.05, **p < 0.01, ***p < 0.0001; Fig. 2C and 2E). While general abundance and distribution patterns are similar across groups, in PSD-95^PDZ12 mutant mice, NR2B is reduced 12% from the synapse compared to wild type littermates (PSD-95^PDZ12, 1.88 ± 0.05; WT, 2.14 ± 0.06; *p < 0.05; Fig. 2E). In corroboration of this partial effect, we find a nearly identical synaptic redistribution of NR2B in SAP102 knockout mice (Cuthbert et al., 2007), albeit with a significant increase in the 30-60 nm PSD bin (unpublished observation). In contrast to NR2B, there are no genotype-dependent changes in the synaptic profile of NR2A (Fig. 2A and 2C).

Figure 2. PSD-95 and αCaMKII mutations differentially alter synaptic expression of NR2B and NR2A subunits, respectively.

(A) Total NR2A (white bars) and NR2B (grey bars) labeling in presynaptic terminals and postsynaptic spines averaged per group, in PSD-95^PDZ12 mutants compared to wild type littermates (black bars). Labeling is normalized to respective wild type postsynaptic bins to demonstrate similarity in labeling between NR2A and NR2B. No gross changes are evident in NR2A or NR2B labeling between mutant and wild type littermates in axons or spines, but there is significantly higher postsynaptic as compared to presynaptic NR2A and NR2B labeling for both mutant and wild type mice (***p < 0.0001, ANOVA with post hoc bonferroni test).

(B) Total NR2A and NR2B labeling in presynaptic terminals and postsynaptic spines averaged per group, in CaMKII^T305D mutants compared to wild type littermates. No genotype differences are observed for NR2B labeling, however, NR2A labeling is reduced 18% in postsynaptic spines of CaMKII^T305D mutant mice compared to wild type littermates (**p = 0.006). Presynaptic labeling is consistently lower for both NR2A and NR2B labeling in both mutant and wild type littermates (***p < 0.0001, ANOVA with post hoc bonferroni test).

(C) The postsynaptic spine is partitioned into a membrane-associated 0-30 nm synaptic PSD bin, a deeper layer, cytoskeletal-associated 30-60 nm PSD bin, and a nonsynaptic, cytoplasmic bin to reflect the functional localization and stability of proteins at the synapse. All bins are normalized to the wild type 0-30 nm synaptic bin where NMDA receptors should be primarily localized. Synaptic localization of NR2A in PSD-95^PDZ12 mutants (white diamonds) is no different than in wild type littermates (315 synapses in n = 3 PSD-95^PDZ12 mice, and 332 synapses in n = 3 wild type littermates). 0-30 nm synaptic : 30-60 nm PSD (SYN:PSD) and 0-30 nm synaptic : nonsynaptic (SYN:NONSYN) bin comparisons show that NR2A subunits are typically localized at or near the synaptic membrane (SYN:PSD; ***p < 0.0001 (−/−) and **p = 0.005 (wt); SYN:NONSYN; **p = 0.004 (−/−) and *p < 0.05 (wt); ANOVA with post hoc bonferroni test). A reference figure illustrating 0-30 nm synaptic, 30-60 nm PSD, and nonsynaptic bins (inset).

(D) NR2A is reduced 16% in the 0-30 nm synaptic bin (***p < 0.001) and 27% in the nonsynaptic, cytoplasmic bin of the spine (**p = 0.006) in CaMKII^T305D mutants (white diamonds) compared to wild type littermates (337 synapses in n = 7 CaMKII^T305D mice, and 286 synapses in n = 6 wild type littermates). NR2A labeling is significantly greater at the synapse than at the PSD (SYN:PSD; ***p < 0.0001 for both (−/−) and (wt); ANOVA with post hoc bonferroni test) but no different than nonsynaptic labeling for both CaMKII^T305D mutants and wild type littermates.

(E) NR2B is reduced 12% in the 0-30 nm synaptic bin (*p < 0.05) in PSD-95^PDZ12 mutants (grey diamonds) compared to wild type littermates (241 synapses in n = 3 PSD-95^PDZ12 mice, and 244 synapses in n = 3 wild type littermates). NR2B distribution is predominantly synaptic (SYN:PSD; ***p < 0.0001 for both (−/−) and (wt); SYN:NONSYN; ***p < 0.0001 for both (−/−) and (wt); ANOVA with post hoc bonferroni test).

(F) There is no change in NR2B localization in synaptic bins between CaMKII^T305D mutant (grey diamonds) and wild type littermates (334 synapses in n = 7 CaMKII^T305D mice, and 278 synapses in n = 6 wild type littermates). NR2B distribution, however, is also highly synaptic in this group (SYN:PSD; ***p < 0.0001 for both (−/−) and (wt); SYN:NONSYN; ***p < 0.0001 for both (−/−) and (wt); ANOVA with post hoc bonferroni test). Error bars represent means ± SEM.

Reduction of NR2A subunits in CaMKII^T305D mutant mice

We next looked at transgenic mice exhibiting a phenotype in contrast to the PSD-95^PDZ12 mutant mice to determine whether the subunit-specific regulation of NR2A and NR2B can be further correlated with effects on synaptic plasticity. Total synaptic labeling reveals an 18% subunit-specific reduction of NR2A in the spines of αCaMKII^T305D mutant mice (CaMKII^T305D, 3.56 ± 0.18; WT, 4.32 ± 0.14; **p = 0.006; Fig. 2B). Preferential postsynaptic localization is again demonstrated by both NR2A and NR2B antibodies in these mice (***p < 0.0001; Fig. 2B). Specifically, NR2A is reduced at both the 0-30 nm synaptic bin (CaMKII^T305D, 1.56 ± 0.04; WT, 1.86 ± 0.03; ***p < 0.0001) as well as the nonsynaptic pool of the spine (CaMKII^T305D, 1.29 ± 0.10; WT, 1.76 ± 0.10; **p = 0.006), by 16% and 27%, respectively (Fig. 2D). Total NR2B levels in the spine and its synaptic association are unaffected by the αCaMKII mutation (Fig. 2B and 2F). It is also interesting to note that NR2A and NR2B share similar synaptic profiles, regardless of mutation or genotype; they are highly localized to the 0-30 nm synaptic bin while being relatively poorly represented in the inner 30-60 nm PSD bin (∼20-40% of synaptic levels; **p = 0.005, ***p < 0.0001; Fig. 2C-2F) where downstream signaling proteins and cytoskeletal elements are prevalent. NR2B expression, in particular, is predominantly synaptic with roughly 2 to 4-fold higher levels associated with the 0-30 nm synaptic bin compared to the deeper layers of the 30-60 nm PSD or nonsynaptic bins (***p < 0.0001; Fig. 2E and 2F). The following rise in NR2A labeling in nonsynaptic bins may be attributed to pools available to be inserted at the synapse or pools of NR2A-containing NMDARs internalized from the synapse (*p < 0.05, **p = 0.004; Fig. 2C and 2D).

Discrete lateralization of the NR2 subunits with respect to PSD-95 and αCaMKII mutations

Insertion and internalization of functional NMDARs from the synapse is a crucial mechanism for modifying synaptic transmission, however, mobility of receptors across the membrane surface has also been shown to modify synaptic efficacy (Tovar and Westbrook, 2002). From studies suggesting extrasynaptic endocytosis of NMDARs as well as an extrasynaptic component to NR2B-mediated LTD, we extended our synaptic analysis of NR2A and NR2B to bins tangential to the PSD. (Roche et al., 2001, Massey et al., 2004).

The lateral distribution of both NR2 subunits across the synapse reveal that they are synaptically targeted with subtle variations in their profile (Fig. 3A, 3C, 3E, 3G). Both NR2 subunits fall off dramatically past the edge of the PSD as evident in the extrasynaptic 250 nm bins (**p = 0.003, ***p < 0.0001; Fig. 3B, 3D, 3F, 3H). This synapse-specific localization of NR2A and NR2B supports the methodological techniques employed and the specificity of the antibodies used. Minor differences, however, mark the lateral profiles of NR2A and NR2B in PSD-95^PDZ12 and CaMKII^T305D mutant mice and their respective wild type littermates. NR2A is relatively equally distributed across the synapse (Fig. 3A and 3E), between middle 50% and lateral 50% bins (Fig. 3B and 3F), but is reduced to a greater extent in the lateral 50% of the synapse in CaMKII^T305D mutant mice (**p = 0.005; Fig. 3F). In contrast to NR2A, NR2B-association to the synapse declines rapidly at the lateral edges of the PSD (Fig. 3C and 3G) and is consistently expressed at lower levels compared to the middle 50% of the synapse across all groups (**p = 0.003, ***p < 0.0001; Fig. 3D and 3H). Although belonging to the same strain as the CaMKII^T305D group, NR2B distribution in wild type littermates of PSD-95^PDZ12 mutant mice appear to be more lateral than wild type littermates of CaMKII^T305D mutant mice (Fig. 3C and 3G); however, comparisons of middle and lateral 50% bins show that NR2B localization in middle regions is significantly higher in both groups regardless of mutation, genotype, or gender (Fig. 3D, 3H and Experimental Procedures).

Quantification of NR2 subunits into 10% bins across the synapse reveal NR2B-specific reductions in distribution of PSD-95^PDZ12 mutant mice compared to wild type littermates (Fig. 3A and 3C). In individual bins, we find a 21% reduction of NR2B in the lateral 20% bin of PSD-95^PDZ12 mutant mice (PSD-95^PDZ12, 0.72 ± 0.03; WT, 0.91 ± 0.06; *p < 0.05; Fig. 3C). Although the 30% and 40% bins do not reach significance on their own, there is a coinciding, combined 16% reduction in the lateral 10-40% of the synapse (PSD-95^PDZ12, 0.6 ± 0.03; WT, 0.71 ± 0.01; *p < 0.05; Fig. 3C). Subdivision of the synapse into evenly proportioned and biologically meaningful middle 50% and lateral 50% bins reveal that the decreases observed in NR2B distribution is significantly lateral (PSD-95^PDZ12, 0.78 ± 0.02; WT, 0.92 ± 0.02; **p = 0.012; Fig. 3D), although there is slight overlap with the middle 50% bin. Together, these findings suggest that in the absence of PSD-95, NR2B instability at the synapse occurs laterally. No changes are observed in the lateral distribution of NR2A in the PSD-95^PDZ12 mutant mice, as expected (Fig. 3A and 3B).

In CaMKII^T305D mutant mice, NR2A is preferentially reduced 16% from the middle 50% of the synapse compared to wild type littermates (CaMKII^T305D, 0.86 ± 0.02; WT, 1.02 ± 0.06; *p < 0.05; Fig. 3F). The detailed distribution of NR2A across the synapse, however, reveals a definitive separation in localization between CaMKII^T305D mutant and wild type littermates which overlaps both middle and lateral 50% bins (Fig. 3E). Accordingly, NR2A is significantly reduced by 23% in the lateral 0-20% of the synapse (CaMKII^T305D, 0.27 ± 0.01; WT, 0.35 ± 0.04; *p < 0.05; Fig. 3E) and although not statistically significant as distributed, the significant decrease in the middle 50% bin (Fig. 3F) gives evidence to the reduction of NR2A that occurs in medial regions of the synapse (Fig. 3E). Cumulatively, these data show that NR2A is reduced across the synapse, in both middle and lateral regions of the synapse. Again, there is an absence of corresponding changes in the lateral distribution of NR2B in the CaMKII^T305D mutant mice (Fig. 3G and 3H).

An NR2B-PSD-95 complex may antagonize an NR2A-αCaMKII complex

In the original studies, the truncated PSD-95 protein was not found in PSD, synaptosomal, or synaptic membrane fractions in PSD-95^PDZ12 mutant mice, and αCaMKII was severely reduced in hippocampal PSD fractions of CaMKII^T305D mutant mice (Migaud et al., 1998, Elgersma et al., 2002). To determine if there is a functional interaction between PSD-95 and αCaMKII which could influence the differential regulation of the NR2 subunits, we looked at αCaMKII localization in PSD-95^PDZ12 mutants and PSD-95 localization in CaMKII^T305D mutants compared to their respective wild type littermates.

Although we expected a compensatory increase or altered distribution of the other PSD protein to account for subunit-specific NMDAR effects, we found no change in total αCaMKII or PSD-95 localization in spines between either mutant mice and their respective wild type littermates (Fig. 4A). There is, however, approximately two-fold higher levels of both proteins in the postsynaptic spine as opposed to the presynaptic terminal (**p = 0.003, ***p < 0.0001; Fig. 4A). No significant genotype effects are manifest in the synaptic-association of αCaMKII or PSD-95 in either mutant mice with respect to their wild type littermates (Fig. 4B and 4C). The synaptic profile of PSD-95, however, tightly overlaps with the profiles seen for NR2A and NR2B (Fig. 2C-F), suggesting that PSD-95 is primarily associated with the synaptic membrane, in the 0-30 nm synaptic bin, in close proximity to the NMDAR subunits (***p < 0.0001; Fig. 4C). In contrast to PSD-95 and the NR2 subunits, αCaMKII is highly nonsynaptic (**p = 0.007) and exhibits a uniform laminar distribution at the synapse, between the 0-30 nm synaptic PSD bin and the deeper, cytoskeleton-associated 30-60 nm PSD bin (Fig. 4B). αCaMKII organization in the spine reflects its spatial and functional potential to translocate to the PSD following activation. This is in agreement with studies showing that synaptic activation is required to target αCaMKII to the PSD where it can interact with NMDAR subunits (Shen and Meyer, 1999, Strack et al., 2000, Hudmon and Schulman, 2002).

The lateral synaptic distributions of αCaMKII and PSD-95 further typify the contrasting patterns of localization between the two NMDAR-interacting proteins (Fig. 4D and 4F). The lateral profile of αCaMKII in PSD-95^PDZ12 mutant and wild type littermates reveals a homogeneous distribution of αCaMKII across the synapse (Fig. 4D), even when separated into middle 50% and lateral 50% synaptic bins (Fig. 4E). Interestingly, the lateral profile of αCaMKII (Fig. 4D) bears close resemblance, if superimposed, to the lateral profile of NR2A (Fig. 3E) in wild type littermates of CaMKII^T305D mutant mice, suggesting that they are closely aligned to one another in regulation or function. PSD-95, in contrast, is highly synaptic and declines sharply towards the lateral edges of the PSD (***p < 0.0001; Fig. 4F and 4G); characteristic of the lateral distributions of NR2B (Fig. 3C, 3D, 3G, 3H). Thus, the lateral profiles of PSD-95 (Fig. 4F and 4G) and NR2B (Fig. 3G and 3H) in CaMKII^T305D mutant and wild type mice are also very tightly aligned when superimposed. Aside from contrasting distributions at the synapse, both αCaMKII and PSD-95 are primarily synaptic with 4 to 5-fold lower localization at extrasynaptic sites extending out 125 nm on either side (*p < 0.05, **p = 0.005, ***p < 0.0001; Fig. 4E and 4G). More importantly, no genotype-dependent effects in lateral distribution are observed for either αCaMKII or PSD-95 localization (Fig. 4D-G).

Discussion

Subunit-specific NMDAR regulation by PSD-95 and αCaMKII may predict synaptic plasticity

Our study reveals that NR2A and NR2B distributions are fairly resilient to change and stable in both mutants, with subtle disruptions in the expression of NR2A and NR2B subunits in the synapses of mice with selective mutations to either αCaMKII or PSD-95. The notion of subunit-specificity in LTP and LTD has been a point of contention, but our findings suggest that the determinant may not be the subunits themselves so much as the interactions with PSD proteins to form complexes, which will not only initiate signaling pathways required for different forms of synaptic plasticity, but also moderate the composition and distribution of subunit-specific NMDARs at the synapse (Fig. 5). In support of our observations, it has been shown that NR2A and NR2B differentially regulate the surface expression of GuR1-containing AMPARs at the synapse through the Ras-ERK pathway, a mechanism essential for modifying synaptic strength (Kim et al., 2005). Synaptic AMPAR expression then, is determined by the subunit composition of NMDARs at the synapse, necessitating the need first to understand the interactions that govern NR2A and NR2B expression at the synapse. Here, we provide evidence for two such interactions, with αCaMKII and PSD-95.

Figure 5. Synaptic model illustrating regulation of subunit-specific NMDARs by αCaMKII and PSD-95.

NMDARs are composed of an obligatory NR1 subunit in combination with NR2A, NR2B, or other NR subunit in a heterodimeric, heterotrimeric, or possibly heteromultimeric receptor. NR2A and NR2B-containing NMDARs are interspersed across the synaptic membrane, interacting with a multitude of proteins in the PSD that can stabilize the receptors at the synapse or signal a cascade of events. The PSD is comprised primarily of αCaMKII, PSD-95, and, to a lesser degree, other MAGUKs, depending on developmental stage and synaptic activity (Sans et al., 2000, Chen et al., 2005). They are, therefore, considered important regulators of NMDAR function and expression at the synapse. We propose that αCaMKII is important for synaptic stability of NR2A-containing NMDARs in the absence of activity but that with activity, αCaMKII may increase targeting of NR2A-containing NMDARs to the synapse. This potential interaction may explain the reduced LTP in CaMKII^T305D mutant mice in the absence of αCaMKII (outlined in Red). PSD-95, on the other hand, may be important for the synaptic stability of a subset of NR2B-containing NMDARs laterally at the synapse where PSD-95-dependent regulation is thought to mediate LTD (outlined in Blue). The absence of PSD-95 from the PSD may, therefore, be able to explain the inability of PSD-95^PDZ12 mutant mice to induce LTD. In addition to PSD-95, there exists overlapping, functionally redundant MAGUKs which may be important in stabilizing different populations of NMDARs at the synapse.

PSD-95-dependent expression of synaptic NR2B-containing NMDARs

How might PSD-95 and αCaMKII selectively regulate the NR2 subunits? Preferential interaction of PSD-95 for the NR2B C-terminal tail, although similar in sequence to that of NR2A, may provide a separable mechanism for the regulation of one subunit over another and, accordingly, it has been shown that internalization of NR2B-constructs was prevented by coexpression with PSD-95, whereas a deletion of the PDZ-binding domain on NR2B increased its internalization (Niethammer et al., 1996, Roche et al., 2001, Lim et al., 2002). Absence of the PDZ-binding domain on NR2A, however, did not affect its expression at the synapse suggesting that NR2A stability does not require MAGUK interactions (Prybylowski et al., 2005). Our data are consistent with studies suggesting PSD-95 preference for NR2B and the mounting evidence that NR2B-containing NMDARs, at least, are regulated by internalization at lateral edges of the PSD where PSD-95 is no longer present to provide stability to the receptors (Lavezzari et al., 2003, Prybylowski et al., 2005). This is clearly evident in the lateral profiles of NR2B and PSD-95 in CaMKII^T305D mutant and wild type mice which have nearly perfect overlap (Fig. 3G and 3H; 4F and 4G). Even so, heterotrimeric NMDARs, containing both NR2A and NR2B subunits, may still be internalized through its association with NR2B subunits.

Additionally, SAP97 and PSD-93 have also been shown to inhibit PDZ domain-mediated internalization of NR2B-containing NMDARs and may, therefore, be important for the stabilization of specific subpopulations of NR2B-containing NMDARs at the synapse either in the presence or absence of PSD-95 (Lavezzari et al., 2003). Detailed experiments have shown that PSD-95 and PSD-93 equally contribute to the normal synaptic expression of both AMPARs and NMDARs and, in the event of a knockdown of both MAGUKs, SAP102 is able to functionally compensate for their loss (Elias et al., 2006). In our study, the partial dissociation of NR2B in the absence of PSD-95 also indicates there are functionally redundant, compensatory MAGUKs present to stabilize NR2B-containing NMDARs at the synapse (Fig. 5). This is clearly evident in increased levels of SAP102 in PSD-95^PDZ12 mutant mice and increased PSD-95 coimmunoprecipitation with NMDA receptors in SAP102 hemizygous mice (Cuthbert et al., 2007). As further proof of functional overlap, we find a nearly indistinguishable redistribution of NR2B in CA1 synapses of SAP102 hemizygous mice compared to their wild type littermates (unpublished observation), albeit with even smaller significant differences than PSD-95^PDZ12 mutant mice.

Using quantitative immunogold analysis, we have found a subtle redistribution of NR2B in synapses which could not be measured either through optical imaging techniques or standard biochemistry. A redistribution would explain the absence of change in total synaptic levels of NR2B quantified by immunogold EM or synaptic fractions in western blots (Migaud et al., 1998). More importantly, differences in subunit distribution, although relatively small, may constitute enough change to lower the threshold for the induction of LTP or prevent LTD altogether, as characterized in PSD-95^PDZ12 mutant mice. Thus, modest differences in physiology underscored by the partial redistribution of NR2B in MAGUK mutants support the notion of compensation by related scaffolding proteins.

Functional correlates of αCaMKII-dependent regulation of NR2A-containing NMDARs

Reduction of NR2A expression in the spines of CaMKII^T305D mutants alludes to αCaMKII inactivity or weak PSD-association as the cause of NR2A instability at the synapse (Fig. 5). The partial reduction of NR2A, however, suggests that the remaining NR2A may either be stabilized in triheteromers at the synapse or, alternatively, bound to SAP97 (Gardoni et al., 2003). The former possibility is exemplified in the lateral profile of NR2A in CaMKII^T305D mutant mice (Fig. 3E), which adopt a more NR2B-like lateral profile (Fig. 3G). Thus, the reduction of NR2A, but not NR2B, suggests that αCaMKII may indeed confer partial synaptic stability specific to NR2A-containing NMDARs, either through direct binding, modifications to the NR2A subunit, or signaling mechanisms required for protein synthesis and trafficking of NR2A-containing NMDARs to the synapse. But this is surprising since it is thought that αCaMKII associates preferentially and with greater affinity to NR2B subunits, and that this interaction is responsible for enhancing synaptic plasticity (Mayadevi et al., 2002, Barria and Malinow, 2005).

Both NR2A and NR2B C-terminal tails contain binding sites as well as phosphorylation sites for αCaMKII, however, the role of the two interactions and the preference for interaction with αCaMKII is still uncertain (Gardoni et al., 1999, Mayadevi et al., 2002). Our findings in LTP-reduced CaMKII^T305D mutant mice support previous work with both NR2A knockouts and C-terminally truncated NR2A subunits displaying severely deficient LTP in the presence of full-length NR2B (Sakimura et al., 1995, Sprengel et al., 1998). Additionally, blockade of NR2A, but not NR2B-containing NMDARs, has recently been found to inhibit protein synthesis in dendrites, providing an intriguing explanation for reductions of LTP in NR2A-deficient mice (Tran et al., 2007). These findings are inconsistent with a theory of an NR2B-specific role in αCaMKII-mediated LTP and suggests that αCaMKII dynamics is not limited to one NMDAR subunit. However, since our animals were not stimulated or trained in learning paradigms, another explanation may be that in basal states of activity interactions with αCaMKII are neither required for nor affect the stability or distribution of NR2B-containing NMDARs at the synapse. αCaMKII interactions with NR2A and NR2B may, consequently, depend on the functional state of αCaMKII and its localization at the synapse with activity and development. This implies two roles of αCaMKII; as a novel structural scaffold and as an enzymatic kinase, and suggests unique interactions with NR2A and NR2B subunits depending on synaptic activity.

PSD-95 and αCaMKII antagonism governs NMDAR-mediated bidirectional synaptic plasticity

Subunit-specific changes in NR2B and NR2A localization implicate two separate complexes, NR2B-PSD-95 and NR2A-αCaMKII, which may act in opposition to each other in moderating synaptic plasticity. Although PSD-95 and αCaMKII occupy different structural constraints as well as functional roles at the synapse, they are both closely associated with NMDARs at the synapse (Valtschanoff and Weinberg, 2001) and as such, both have dominant roles in regulating NMDAR function. With translocation of αCaMKII to the PSD, both PSD-95 and αCaMKII are in close proximity to the NMDARs and the binding sites for both reside within a close framework on the C-terminal tails of the NR2 subunits (Garner et al., 2000, Bayer and Schulman, 2001). This may result in steric hindrance which may favor one interaction over the other and, consequently, theories for the competitive binding of PSD-95 and αCaMKII for the NR2A and NR2B subunits may provide a possible mechanism to explain changes associated with NMDAR-dependent synaptic plasticity (Gardoni et al., 2001, Meng et al., 2004). This competitive interaction between PSD-95 and αCaMKII may extend to antagonistic modifications of the NMDAR subunits as well as antagonistic signaling mechanisms which could further explain the contrasting phenotypes of the mice.

In our study then, the independent regulation of the NR2 subunits by PSD-95 and αCaMKII also support a competitive or antagonistic interaction of PSD-95 and αCaMKII for the NR2 subunits. PSD-95 may serve to stabilize heterodimeric and heterotrimeric NR2B-containing NMDARs at the synapse in the absence of activity. With synaptic activation, however, the static network of proteins scaffolded by PSD-95 and other MAGUKs may become destabilized by high calcium levels and increases in kinase activity. αCaMKII and NMDAR-dependent increases in casein kinase 2 activity, for instance, has been shown to phosphorylate the PDZ-binding domain on NR2B to disrupt NR2B-PSD-95 interactions (Chung et al., 2004). In contrast, activity-dependent αCaMKII phosphorylation of SAP97 disrupts its interaction with NR2A, providing an αCaMKII-dependent mechanism for targeting NR2A-containing NMDARs to the synapse (Gardoni et al., 2003). These findings point to overlapping, yet partially dissociable, mechanisms for the independent regulation of NR2A and NR2B to modulate synaptic plasticity to different effects and suggests that the regulation of NMDARs and its associated complexes is much more complicated than a single set of associations but, rather, involves multiple competing systems of interaction.

List of Abbreviations

AMPA

α-amino-3-hydroxy-5-methyl-4-isoxazole proprionic acid

AMPAR

AMPA receptor

CA1

cornu ammonis field 1 of hippocampus

C-terminal

carboxyl (COOH) terminal

EM

electron microscopy

ES

extrasynaptic

GluR1

AMPA receptor subunit 1/A

GP

gold particle

HSA

human serum albumin

LTD

long term depression

LTP

long term potentiation

MAGUK

membrane-associated guanylate kinase

NMDA

N-methyl-D-aspartate

NMDAR

NMDA receptor

NONSYN

nonsynaptic

NR2A

NMDA receptor subunit 2A

NR2B NMDA

receptor subunit 2B

PDZ

PSD-95, Dlg, ZO-1/Dlg-homologous region

PDZ12

remaining PDZ 1,2 domain of N-terminally truncated PSD-95

PSD

postsynaptic density

PSD-93

postsynaptic density protein 93/ chapsyn-110

PSD-95

postsynaptic density protein 95/ SAP90

Ras-ERK

Ras-activated extracellular signal-regulated kinase pathway

SAP102

synapse-associated protein 102

SAP97

synapse-associated protein 97

SEM

standard error mean

SYN

synaptic

T305D

Thr305 substituted with Asp in αCaMKII

TBS

tris-buffered saline

αCaMKII

α-isoform of calcium-calmodulin kinase II