CaMKII translocation requires local NMDA receptor‐mediated Ca2+ signaling

Agnes Thalhammer, York Rudhard, Cezar M Tigaret, Kirill E Volynski, Dmitri A Rusakov, Ralf Schoepfer

Author Affiliations

  1. Agnes Thalhammer1,,
  2. York Rudhard1,,
  3. Cezar M Tigaret1,
  4. Kirill E Volynski2,
  5. Dmitri A Rusakov2 and
  6. Ralf Schoepfer*,1
  1. 1 Department of Pharmacology, Laboratory for Molecular Pharmacology, UCL, London, UK
  2. 2 Institute of Neurology, UCL, London, UK
  1. *Corresponding author. Department of Pharmacology, Laboratory for Molecular Pharmacology, UCL, Gower Street, London WC1E 6BT, UK. Tel.: +44 20 76797242; Fax: +44 20 76797245; E‐mail: r.schoepfer{at}
  1. These authors contributed equally to this work

  • Present address: Universität Hamburg, ZMNH, Falkenried 94, 20251 Hamburg, Germany


Excitatory synaptic transmission and plasticity are critically modulated by N‐methyl‐d‐aspartate receptors (NMDARs). Activation of NMDARs elevates intracellular Ca2+ affecting several downstream signaling pathways that involve Ca2+/calmodulin‐dependent protein kinase II (CaMKII). Importantly, NMDAR activation triggers CaMKII translocation to synaptic sites. NMDAR activation failed to induce Ca2+ responses in hippocampal neurons lacking the mandatory NMDAR subunit NR1, and no EGFP‐CaMKIIα translocation was observed. In cells solely expressing Ca2+‐impermeable NMDARs containing NR1N598R‐mutant subunits, prolonged NMDA application elevated internal Ca2+ to the same degree as in wild‐type controls, yet failed to translocate CaMKIIα. Brief local NMDA application evoked smaller Ca2+ transients in dendritic spines of mutant compared to wild‐type cells. CaMKIIα mutants that increase binding to synaptic sites, namely CaMKII‐T286D and CaMKII‐TT305/306VA, rescued the translocation in NR1N598R cells in a glutamate receptor‐subtype‐specific manner. We conclude that CaMKII translocation requires Ca2+ entry directly through NMDARs, rather than other Ca2+ sources activated by NMDARs. Together with the requirement for activated, possibly ligand‐bound, NMDARs as CaMKII binding partners, this suggests that synaptic CaMKII accumulation is an input‐specific signaling event.


Activity‐dependent changes in synaptic strength underlie some forms of learning and memory, with the N‐methyl‐d‐aspartate receptor (NMDAR) serving as a molecular coincidence detector for pre‐ and postsynaptic activity (Bliss and Collingridge, 1993; Malenka and Nicoll, 1999).

The NMDAR‐effected postsynaptic Ca2+ rise activates calmodulin‐dependent protein kinase II (CaMKII) thus enabling synaptic long‐term potentiation (LTP) (Bliss and Collingridge, 1993; Malenka and Nicoll, 1999): CaMKII is autophosphorylated, switching to a Ca2+/CaM‐independent active state, which can persist after the intracellular Ca2+ concentration ([Ca2+]i) has returned to the resting level. CaMKII thus acts as a molecular memory device translating NMDAR‐activity into changes in synaptic strength (Fukunaga et al, 1995; Lisman et al, 2002), a mechanism corroborated by gene targeting experiments (Elgersma et al, 2004).

NMDAR‐mediated [Ca2+]i elevation is thought to be highly localized at the channel mouth thus encoding information for Ca2+ sensors in close proximity of this high‐Ca2+ nanodomain (Augustine et al, 2003). A NMDAR‐mediated local Ca2+ rise triggers translocation of CaMKII to the postsynaptic density (PSD) where the kinase then directly binds to the NMDAR (Shen and Meyer, 1999; Bayer et al, 2001). This binding is facilitated once CaMKII is autophosphorylated (Strack and Colbran, 1998; Leonard et al, 1999; Strack et al, 2000; Bayer et al, 2001). The PSD‐localized, activated CaMKII is capable of phosphorylating molecules important for synaptic transmission. Indeed, mass spectrometric analysis showed a multitude of sites in PSD phosphorylated in vivo by CaMKII (Trinidad et al, 2006).

To investigate if the spatial nature of NMDAR‐mediated Ca2+ signals determines the specificity of downstream signaling, we used cultured hippocampal neurons from a mouse model expressing Ca2+‐impermeable NMDARs (Rudhard et al, 2003). In these NR1R/− mice the mandatory NR1 subunit carries the point mutation N598R in the pore domain, abolishing Ca2+permeability of all NMDAR channels (Burnashev et al, 1992).

We found that prolonged application of NMDA could still mediate a robust rise of dendritic [Ca2+]i in NR1R/− cells which express only Ca2+‐impermeable NMDARs. Although a similar protocol routinely induced translocation of CaMKII to the synapse in wild‐type neurons, no translocation occurred in NR1R/− cells. Mutations in the EGFP‐tagged CaMKII that prolong its synaptic localization (Shen et al, 2000) completely recovered NMDAR‐dependent accumulation of the kinase at synaptic sites in NR1R/− cells. This suggests that NMDAR‐generated local Ca2+ signals are primarily responsible for the wild‐type CaMKII translocation. Indeed, brief local application of NMDA induced significantly smaller Ca2+ transients in dendritic spines of NR1R/− compared with wild‐type cells, indicating an important role of Ca2+ entry through NMDARs in generating local Ca2+ signals.

We conclude that a highly localized rise in Ca2+ concentration near the NMDAR pore, rather than a general diffuse increase in [Ca2+]i via other Ca2+ sources indirectly activated by NMDARs, is required for CaMKII trapping at the PSD. This provides evidence for the concept of functional Ca2+ nanodomains at the PSD.


To study the influence of local NMDAR‐mediated Ca2+ signals on NMDAR‐dependent translocation of CaMKIIα, we investigated neurons of NR1R/− animals exclusively expressing mutant NMDARs. These receptors contain the NR1(N598R) subunit which leads to Ca2+‐impermeable NMDAR channels (Burnashev et al, 1992; Rudhard et al, 2003).

Hippocampal neurons from wild‐type (NR1+/+, NR1+/−) and NR1R/− animals grew in culture and developed comparable morphologies (Supplementary Figure S1A) with no evidence of a significant difference in surface expression levels or distribution of Ca2+‐impermeable versus wild‐type NMDA receptor complexes (Supplementary Figure S1B and C). Similarly, expression and distribution of other postsynaptic components appeared equivalent in the two genotypes (Supplementary Figure S2), in line with expression studies in the mutant NR1R/− mice (Rudhard et al, 2003).

Ca2+ measurements in wild‐type and NR1R/− cells

We evaluated NMDA‐induced Ca2+ responses using fluorescence ratiometric measurements. Intriguingly, although NR1R/− cells express only Ca2+‐impermeable NMDARs, selective NMDAR stimulation (50 μM NMDA/10 μM Gly) induced a large Ca2+ signal throughout the cell soma and dendrites (Figure 1A).

Figure 1.

Ca2+ imaging in wild‐type and NR1R/−‐mutant cells. (A) Pseudocolor‐coded ratio images of emissions evoked by excitation at 340 and 380 nm (ratio 340/380) taken with a CCD camera before (‘unstim’) and during (‘NMDA’) stimulation on an NR1R/− cell; scale bar: 10 μm. Application of NMDA led to an increase in Ca2+ signal in soma and dendrites. Dashed rectangles indicate the position of the photodiode field during measurements at the soma. (B) Sample traces obtained from somatic measurements on wild‐type (+/+; black), NR1R/− (R/−; medium gray), and NR1−/− (−/−; light gray) neurons, shown as 340/380 ratios. Bars indicate application of NMDA (‘N’, 50 μM, 60 s), APV (150 μM), and Mg2+ ions (2 mM). Measurements in Na+‐free conditions were made after equilibration with choline chloride‐containing Ringer (0 Na+; indicated by the discontinuity in the time axis). Neurons lacking NR1 subunits (−/−) did not respond to NMDA application at all. However, cells with Ca2+‐impermeable NMDARs (R/−) showed a large rise in Ca2+ signal which was not blocked by Mg2+ ions. Removal of external Na+ resulted in reduced Ca2+ responses of wild‐type cells and abolished responses of NR1R/− cells. (C) apparent somatic [Ca2+]i at baseline (‘basal’) levels and after maximal response to NMDA (‘NMDA’, 50 μM) in wild‐type and NR1R/− cells (ncells/nexperiments=29/5 and 21/4, respectively). Paired data from single cells (lines) are shown together with mean±s.e.m. (D) DIC image illustrating the typical position of the photodiode field onto a dendrite branch of a pyramidal neuron as used for data in (E+F); scale bar: 20 μm. (E) NMDA (50 μM, 30 s)‐evoked changes in dendritic [Ca2+]i in cells expressing the wild‐type (+/+, black filled symbols) or mutant NR1 subunit (R/−, red open symbols), in the absence (circles) or presence (squares) of 150 μM APV. Dendritic recordings were made as in (D); mean±s.e.m. (+/+, n=6; R/−, n=8). Horizontal line indicates agonist application. (F) Apparent dendritic [Ca2+]i at baseline (‘basal’) levels and after maximal response to NMDA (‘NMDA’) in wild‐type and NR1R/− cells (ncells/nexperiments=6/2 and 8/2, respectively). Paired data from single cells (lines) are shown together with mean±s.e.m.

To quantify the rise in [Ca2+]i, we measured 340/380 ratios from isolated somatic regions (Figure 1A) and dendritic segments (Figure 1D). NMDAR stimulation (60 s) evoked a large increase in the ratio 340/380 at the soma, both in wild‐type and NR1R/− cells (Figure 1B). This Ca2+ rise was insignificant in the presence of 150 μM 2‐amino‐5‐phosphonopentanoic acid (APV), or in neurons lacking the NR1 subunit (NR1−/−). NR1R‐containing receptors are also lacking the Mg2+ block (Rudhard et al, 2003). The absence of this Mg2+ block in R/− cells (Figure 1B) confirms therefore, the presence of mutant NMDARs. As these NMDARs cannot serve as source of Ca2+ ions, the rise in [Ca2+]i must be caused by Ca2+ release or influx secondary to depolarization.

Baseline levels of the somatic apparent free [Ca2+]i in NR1R/− cells were similar with those in wild‐type neurons (R/−, 84±20 nM, n=21; +/+, 64±8 nM, n=29; mean±s.e.m., ncells; Figure 1C), and in line with (Garaschuk et al, 1997; Pozzo‐Miller et al, 1999; Shuttleworth and Connor, 2001). NMDAR‐evoked responses were also comparable between the two genotypes (apparent [Ca2+]i after stimulation: +/+, 631±75 nM, n=29; R/−, 629±86 nM, n=21; for derivation of [Ca2+]i values, see Supplementary data and Supplementary Methods).

Likewise, the dendritic Ca2+ responses evoked by NMDA (30 s) in both wild‐type and NR1R/− cells were APV‐sensitive and comparable in size (apparent [Ca2+]i after stimulation: 116±24 nM, n=6, and 115±20 nM, n=8, respectively; Figure 1E and F).

In summary, these measurements showed a comparable agonist‐induced increase of [Ca2+]i in wild‐type and NR1R/− cells after prolonged (⩾30 s) NMDA application.

Sources of internal Ca2+ rise following NMDAR activation

NMDAR‐evoked Ca2+ responses were eliminated in the absence of extracellular Na+ in NR1R/− cells (Figure 1B), whereas in wild‐type cells, a residual rise in [Ca2+]i could still be observed (25±4% of the full response, mean±s.e.m., n=7; Supplementary Figure S3). This residual Ca2+ rise in wild‐type cells is in accordance with earlier observations (Schneggenburger et al, 1993) that about 10% of the ions passing through wild‐type NMDAR channels are Ca2+ ions and is consistent with a Ca2+ influx directly through the activated wild‐type, but not NR1(N598R)‐mutant, NMDAR channels.

Taken together, these findings confirm that the NMDA‐induced Ca2+ responses in NR1R/− cells result from a Ca2+ influx that is triggered by NMDAR‐driven depolarization, but carried through sources other than the activated NR1(N598R) NMDAR channels. The potential sources of Ca2+ entry include VGCCs, Na+/Ca2+exchanger, and intracellular stores (see Supplementary data and Supplementary Figure S3).

NMDAR‐dependent local Ca2+ entry in dendritic spines

Because a long‐term, space‐equilibrated elevation of Ca2+ could in principle originate anywhere in the cell, it may not necessarily reflect large Ca2+ rises inside dendritic spines: endogenous Ca2+ buffers and pumps restrict Ca2+ rises away from the source. Therefore, we assessed the profile of NMDAR‐dependent Ca2+ entry in the immediate spine proximity on a faster time scale (Figure 2).

Figure 2.

Rapid local application of NMDA evokes a local transient rise of [Ca2+]i in dendrites and dendritic spines. (A) A dendritic fragment of a wild‐type cell (wt) filled, in whole‐cell mode, with cell‐impermeable 20 μM Alexa Fluor 594 and 200 μM Fluo‐4; upper panel Alexa (red) channel; lower panel, Fluo‐4 (green) channel. The fine morphology of dendritic spines is clearly seen. Dotted lines indicate the line scan position. (B–D) Line‐scan recordings (at 500 Hz for 3 s, dotted lines in A show scan positioning) of Alexa and Fluo‐4 fluorescence (B, upper and lower panels, respectively) following pressure application of NMDA (100 ms × 15 psi pressure pulse; Alexa added to the pipette medium for pulse monitoring) through a patch pipette located in a ∼15 μm vicinity (shown by a circle in C). (B) Upper panel (red channel) depicts the spatio‐temporal profile of NMDA applied through the pipette together with Alexa and diffusing in the bath medium; the corresponding time course is shown in (D, red trace). (B) Lower panel (green channel) depicts Ca2+ transients evoked in spine head and in dendritic stem; the corresponding time course traces are shown in (D; green, spine head; blue, dendrite; red, Alexa). Scale bars; (C), 30 μm; (D), 1 s and 100 nM. (E–H) The same experiment as in (B–D) but in NR1 mutant (R/−) cells. The corresponding notations are the same as in (E–H). Scale bars; (E), 1 μm; (F) 500 ms; (G), 30 μm; (H) 1 s and 100 nM.

A brief (100–150 ms) pulse of NMDA evoked a rapid, prominent Ca2+ response in both the spine head and the dendrite (Figure 2A–D). In wild‐type cells, the average Ca2+ transient was larger in the spine than the dendrite (619±127 and 366±82 nM, respectively, n=7, p<0.04; Figure 2A–D) reflecting the predominant occurrence of NMDARs at the spine head (Racca et al, 2000). In the NR1R/− cells, local application of NMDA also triggered Ca2+ responses, which, however, displayed smaller amplitudes and a slower time course compared with the wild‐type (Figure 2E–H). Importantly, these data represent the contribution of local Ca2+ sources to intracellular Ca2+ rises, whereas results obtained with bath application protocols are likely to reflect long‐term equilibration of Ca2+ across large distances (Figure 1, and dendritic spines at high resolution in Supplementary Figure S5). In addition, NMDA‐induced Ca2+ transients in the NR1R/− cells were of similar size in spines and shafts (to 63.7±5.6 and 65.4±5.7 nM, respectively, n=5), with the shaft transient generally showing faster rises (Supplementary Figure 2H). Again, this is likely to reflect the lacking NMDAR Ca2+ channel sources at spine heads.

In summary, these Ca2+ imaging data indicate that (a) wild‐type NMDAR Ca2+ channels provide a disproportionately large source of rapid Ca2+ entry in dendritic spines compared with shafts (Figure 2), (b) there still exist rapid local sources of Ca2+ entry in the dendritic spine vicinity in the NR1R/− cells, and (c) long‐term activation of NMDAR in wild‐type and mutant cells results in similar elevation of global internal Ca2+.

Translocation of CaMKIIα

Our Ca2+ measurements indicate that wild‐type and NR1‐mutant cells differ in their sources of free Ca2+ in response to NMDAR activation. Thus, the question arises whether signaling pathway molecules downstream of the NMDAR can discriminate between Ca2+ entry directly through NMDAR channels and Ca2+ originating from other sources.

The translocation of CaMKIIα from the cytosol to the PSD is known to depend on NMDAR activation and rise of [Ca2+]i (Strack et al, 1997; Shen and Meyer, 1999). To test if this process depends on the direct influx of Ca2+ through NMDARs, we transfected NR1+/+, NR1R/− and NR1−/− knockout cells with EGFP‐CaMKIIα and analyzed the intracellular fluorescence distribution in Glu‐treated and untreated cells (Figure 3A).

Figure 3.

Translocation of CaMKIIα to synapses is dependent on Ca2+influx through the NMDA receptor channel pore. (A) Stimulation of neurons (50 μM Glu/10 μM Gly; 3 min) expressing the wild‐type (+/+), but not the Ca2+‐impermeable NMDARs (R/−), increased the synaptic translocation of CaMKIIα. Axial projection of confocal overviews (EGFP‐CaMKII fluorescence, green, overlayed with anti‐synapsin1, red; scale bar: 10 μm); details enclosed in the white rectangles are expanded underneath (EGFP‐CaMKIIα, CaMKIIα, and anti‐synapsin1 fluorescence, ‘syn’, shown individually and merged; scale bar: 5 μm). Note the juxtaposition of EGFP‐CaMKII puncta with anti‐synapsin1 structures in wild‐type, but not NR1R/− cells, after Glu treatment. (B) Agonist‐induced CaMKII translocation: Glu‐ or NMDA‐induced formation of CaMKII puncta occurred in wild‐type cells (+/+), but not in cells lacking NR1 subunits (−/−), or in NR1R/− neurons (R/−). n=4–15. Agonists used at 50 μM with 10 μM Gly. No formation of puncta was observed with kainate or AMPA (both 100 μM). (C) NMDAR activation is sufficient for formation of CaMKII puncta. NMDA, 50 μM; Cd2+, 100 μM. n=4–9; mean±s.e.m. Brackets indicate statistically significant changes by treatment: (*) P<0.05, (**) P<0.01, (***) P<0.001. For a detailed listing of n for each condition, see Supplementary data.

The treatment of wild‐type cells (50 μM Glu/10 μM Gly, 2–3 min) changed the uniform intracellular distribution of CaMKIIα to a pattern with clusters of high fluorescence intensity within the dendritic arbor in line with previous observations (Shen and Meyer, 1999). The treatment was comparable to that used for Ca2+ imaging experiments with prolonged agonist application (1 min). The translocation effect was specific for the tagged kinase; the same experimental conditions did not influence the distribution of plain green fluorescent protein (GFP) (Supplementary Figure S6).

CaMKII clustering was observed mainly at synaptic sites, but also occurred at extrasynaptic sites as reported for similar stimulation protocols (Hudmon et al, 2005). To confirm that the increase in the number of CaMKII puncta indeed reflected a translocation to the PSD, we determined the number of CaMKII puncta localized in the proximity of structures that were immunoreactive for the synaptic marker synapsin1 (Figure 4A, bar graph).

Figure 4.

CaMKII mutants rescue PSD accumulation of CaMKII in NR1R/− neurons. Fluorescent puncta for EGFP‐tagged wild‐type (A, ‘CaMKIIα‘), constitutively active form (B, ‘T286D’), and inhibitory autophosphorylation‐deficient form (C, ‘TT305/306VA’) of CaMKIIα in stimulated (Glu or NMDA; 50 μM with 10 μM Gly) and unstimulated neurons from specific NR1 genotypes (wild type: +/+; NR1−/−: −/−; NR1R/−: R/−). Left: Examples taken from the groups depicted in the graphs on the right (axial‐projected confocal stacks showing fluorescent EGFP‐CaMKII puncta, scale bar: 10 μm); insets: details of dendritic segments (scale bar: 2 μm). Right: Densities (counts/100 μm dendrite length) of synaptically localized EGFP‐CaMKII puncta. Data are shown as mean±s.e.m.; n=7–14, number of independent experiments: 2–4. For clarity, statistical significance is shown only for NR1R/− genotype; (**) P<0.01, (***) p<0.001. (D+E) Influence of Cd2+ (100 μM) (D) and of AMPA (100 μM) (E) on synaptic puncta levels of EGFP‐tagged CaMKII clones expressed in wild‐type and NR1 N589R‐mutant cells.

Quantification revealed a three‐fold increase in synaptically localized CaMKII puncta (Figure 4A) in Glu‐stimulated wild‐type cells. This increase was statistically significant compared with unstimulated controls or agonist‐treated NR1−/− and NR1R/− cells, where Glu‐induced CaMKIIα translocation was absent (Figure 3A). Moreover, no increase in puncta could be observed after selective stimulation of AMPARs (Figure 3B). When cells were cotransfected with plain mRFP, we analyzed the translocation in relation to the red fluorescence acting as cell filler and observed the same genotype dependency for the translocation (Supplementary Figure S8), confirming that the change in fluorescence distribution, and therefore CaMKII translocation, is independent of spine volume effects potentially induced by the agonist.

Thus, our observations suggest that Ca2+ influx directly through the NMDARs was required for Glu‐evoked CaMKII translocation, hence that CaMKII can discriminate among different Ca2+ sources.

Translocation of CaMKIIα to synapses is dependent on Ca2+ influx through the NMDA receptor channel pore

To test the contribution of secondary, depolarization‐induced, Ca2+ sources to the translocation, we minimized NMDA‐induced depolarization by reducing extracellular Na+ to nominally zero mM (0 Na+, Figure 3C). Under these conditions, NMDAR activation still led to the formation of GFP‐CaMKII puncta in dendrites of wild‐type cells, revealing successful kinase translocation (Figure 3B).

Our 0 Na+ data further suggest that a possible buildup of internal Ca2+ owing to inactivity of the Na+/Ca2+ exchanger has no influence on the accumulation of CaMKII.

To test for the influence of VGCCs activated secondary to NMDAR depolarization, we studied CaMKII translocation in the presence of Cd2+ (100 μM; Figure 3C). VGCCs blocked by Cd2+ did not eliminate the increase in puncta, suggesting that the mere influx through NMDAR channels in wild‐type cells is sufficient for translocation of CaMKII. Even depletion of intracellular stores by CPA (10 min, 15 μM) in addition to Cd2+ block still resulted in puncta formation (see Supplementary Data).

These results clearly show that direct Ca2+ influx through the NMDAR is necessary and sufficient for the agonist‐evoked translocation of CaMKIIα to the synapse.

Translocation of CaMKII mutants; retention at the PSD

Activated CaMKII is translocated to the PSD, where it can bind to NR2B subunits (Bayer et al, 2001). Autophosphorylation facilitates and strengthens the binding of CaMKII to PSD components (Leonard et al, 1999; Bayer and Schulman, 2001; Bayer et al, 2001). Indeed, the introduction of an autophosphorylation mimicking mutation into CaMKII (T286D) results in prolonged retention times of the kinase at the PSD (Shen et al, 2000).

The clustering of the constitutively active form T286D CaMKII in unstimulated cells was not different from that of wild‐type CaMKII (Figure 4B), an observation previously reported for wild‐type neurons (Shen and Meyer, 1999). In addition, Glu‐induced translocation of T286D CaMKII in NR1+/+ cells was comparable to that of wild‐type CaMKII. Interestingly, application of Glu or NMDA led to a striking recovery of the translocation effect in cells of the Ca2+‐impermeable NR1 genotype (Figure 4B). This result suggests that the constitutively active T286D CaMKII is able to overcome the absence of direct Ca2+ influx through NMDARs, thereby recovering the NMDAR‐dependent CaMKII translocation to the PSD in NR1R/− neurons to levels observed in wild‐type cells. Furthermore, this recovery could be blocked by addition of Cd2+ (Figure 4D), suggesting that NMDAR‐mediated depolarization‐dependent Ca2+ sources are necessary and sufficient for the NMDAR‐evoked accumulation of constitutively active CaMKII (T286D) at the PSD.

Additionally, we made use of the EGFP‐tagged CaMKII double mutant TT305/306VA, where two major inhibitory autophosphorylation sites are removed (Patton et al, 1990; Colbran, 1993). The translocated TT305/306VA CaMKII has been shown to have increased net synaptic localization (Shen et al, 2000). In line with these findings, we observed increased basal synaptic kinase puncta in wild‐type and NR1R/− cells in the absence of Mg2+ (Figure 4C), as reported in vivo by Elgersma et al (2002).

Glutamate stimulation increased the number of synaptic CaMKII‐TT305/306VA in wild‐type cells. Crucially, CaMKII‐TT305/306VA was also able to recover the NR1 wild‐type phenotype in NR1R/− cells after stimulation with Glu or NMDA.

However, as for wild‐type CaMKII, stimulation protocols did not have any effect on the distribution of mutant CaMKII in NR1−/− cells, suggesting that the presence of NMDARs is required for CaMKII translocation to the PSD.

Specific stimulation of AMPARs with 100 μM AMPA failed to induce translocation of CaMKII‐T286D and CaMKII‐TT305/306VA to synaptic sites (Figure 4E). We obtained equivalent results with the nondesensitizing agonist kainate (100 μM) (Supplementary Figure S7). These findings suggest a principal functional difference between NMDA‐ and AMPA‐mediated depolarization, and perhaps a requirement of ligand binding to the NMDAR leading to a conformational change, in addition to opening of the channel (Vissel et al, 2001).

In summary, two different mutated forms of CaMKII that affect the duration time of the kinase at the PSD, namely T286D and TT305/306VA, were found to accumulate at synaptic sites upon stimulation of Ca2+‐impermeable NMDA receptors. Thus, conditions that facilitate the binding of CaMKII to PSD components and reduce its dissociation from the synaptic compartment can overcome the lack of a Ca2+ signal beneath the NMDA receptor pore. This effect, however, requires an agonist‐induced activation of NMDARs.


CaMKII translocation requires direct Ca2+ influx through the NMDAR channel

The activation of NMDARs and the subsequent rise in Ca2+ are necessary and sufficient for CaMKII translocation (Strack and Colbran, 1998; Shen and Meyer, 1999 and this manuscript).

The lack of NMDA‐induced translocation of CaMKII in neurons derived from NR1−/− animals was not surprising, as these cells do not express functional NMDARs and consequently do not exhibit NMDA‐evoked Ca2+ rises (Figure 1B). Although Ca2+ rises were evoked by AMPAR‐dependent depolarization in these cells (Supplementary Figure S4), no CaMKII translocation was observed with Glu (Figure 3B), indicating the requirement for NMDAR activation.

Deletion of NMDARs in NR1−/− cells, however, removes not only NMDAR‐dependent Ca2+ signals but also the corresponding CaMKII‐binding sites at PSDs. To test whether CaMKII translocation requires Ca2+ entry through NMDARs, in addition to NMDAR binding, we used the NR1R/− mouse model.

In NR1R/− neurons, distribution of NR1 subunits and the integrity of PSDs appeared preserved (Supplementary Figures S1 and S2). However, prolonged activation of NMDARs in these cells failed to translocate CaMKII (Supplementary Figures S3 and S4A) even though it produced long‐term Ca2+ elevation, throughout the dendritic compartments, similar to that in the wild‐type. The most likely explanation is that in NR1R/− neurons, ligand‐bound Ca2+‐impermeable NMDARs depolarized the cell via Na+ influx and thus activated VGCCs and/or the reverse Na+‐Ca2+exchange. Prolonged activation of these Ca2+ sources led to a rise and equilibration of Ca2+ throughout the cell. However, this global Ca2+ rise was insufficient to trigger CAMKII translocation, which suggests that the latter may require high‐concentration Ca2+ domains inside the spines.

Indeed, the concept of local Ca2+ nanodomains has been well established (Augustine et al, 2003). In line with this, activated wild‐type NMDAR Ca2+ channels, which remain open for tens of millisecond, could generate an exceptionally sharp local Ca2+ ‘hot‐spot’ (Supplementary Figure S9). It would then be reasonable to assume that each individual Ca2+ source produces its own Ca2+ domain determined by the source opening kinetics.

We found that a brief local pulse of NMDA evoked clear Ca2+ responses inside the dendritic spines, both in wild‐type and mutant cells. In NR1R/− cells, however, Ca2+ entry into spines was reduced (Figure 2G and H). This indicated that, although local Ca2+ sources (along with their respective Ca2+ nanodomains) still existed, their numbers were greatly reduced because of the lack of Ca2+ permeability by NMDARs. The non‐NMDAR Ca2+ sources and their associated Ca2+ domains thus appeared unable to induce CaMKII translocation.

In view of these findings, we propose that Ca2+ nanodomains generated by Ca2+ entry directly through wild‐type NMDAR channels are critical for the process of CaMKII translocation.

A model for the lack of CaMKII translocation in cells with Ca2+‐impermeable NMDARs

What are the possible reasons for the lack of CaMKII translocation in cells with Ca2+‐impermeable NMDARs?

In NR1R/− cells, the number and distribution of NMDARs may differ from wild‐type cells. However, wild‐type and mutant NMDARs showed equivalent membrane expression, synaptic distribution (Supplementary Figure S1B/C), and similar amplitudes of whole‐cell currents (Rudhard et al, 2003). Moreover, we failed to detect differences in number and synaptic localization of two key components of NMDAR complexes, namely PSD‐95 and NR2B (Supplementary Figure S2). Therefore, the lack of NMDA‐induced CaMKII translocation in NR1R/− cells is unlikely to arise from changes in number, distribution, and structural integrity of the Ca2+‐impermeable NMDAR complexes.

A restricted or otherwise changed Ca2+ availability for activation of CaMKII in the mutant neurons in comparison to wild‐type cells might affect the translocation of CaMKII. Most likely, CaMKII is recruited from the dendritic area to the synaptic site upon NMDAR stimulation (Meyer and Shen, 2000). However, with the experimental paradigm used in the translocation assays (i.e., prolonged agonist application, time scale of minutes), we found no difference between genotypes in the Ca2+ signal in dendritic segments. On the other hand, at a much faster time resolution (hundreds of ms), line scanning measurements in spines revealed strikingly smaller and slower Ca2+ responses to brief NMDA puffs. This suggests that the direct NMDAR Ca2+ signal is indeed required for trapping CaMKII at the PSD.

The results with CaMKII mutants suggest that a dynamic model for CaMKII binding to and unbinding from the NMDAR can provide a way to compensate for the lack of direct Ca2+ signals in NR1R/− cells (Figure 5).

Figure 5.

Local NMDA receptor‐mediated Ca2+ signaling in CaMKII translocation. (A) In wild‐type neurons, stimulation of NMDA receptors (NMDAR) leads to an increase in [Ca2+]i directly (i), as Ca2+ ions enter via the receptor to form a locally restricted Ca2+ hot‐spot beneath the NMDAR channel mouth and indirectly (ii) as NMDAR activation triggers other Ca2+ sources. Consequently, CaMKII is translocated to the PSD (a) where it can directly interact with and bind to the NMDAR (b) in a process favored by phosphorylation at residue T286. The translocation is reversible and CaMKII dissociation from the PSD (c) is furthered by inhibitory autophosphorylation. (B) In cells with Ca2+‐impermeable NMDARs the path (i) is defective and CaMKII translocation (a) is possibly counterbalanced by pathways (b) and (c). The wild‐type phenotype can, therefore, be rescued either by the autophosphorylation‐mimicking CaMKII mutant T286D, which strengthens interaction (b), or by the inhibitory autophosphorylation‐deficient mutant TT305/306VA, which weakens pathway (c).

The translocation process itself (Figure 5A, pathway (a)) has been shown to depend only on the binding of Ca2+/CaM to CaMKII. Nevertheless, this process is reversible within seconds, hence the ability to observe CaMKII translocated to the synapse critically depends on the time scale in which the kinase is present at the PSD (Shen and Meyer, 1999). Autophosphorylation of CaMKII at T286 prolongs this time scale (Shen et al, 2000), thus enabling CaMKII to form, or strengthen, its interactions with multiple binding partners, for example, NR2B, NR1, NR2A, densin‐180 (Strack and Colbran, 1998; Gardoni et al, 1999; Bayer et al, 2001, 2006; Walikonis et al, 2001; Leonard et al, 2002; Robison et al, 2005).

Indeed, a CaMKII mutant that mimicks autophosphorylation (T286D) overcame the apparent lack of CaMKII translocation in cells with Ca2+‐impermeable NMDARs (Figure 4B). The recovery of wild‐type translocation behavior in NR1R/− cells with the constitutively active form of CaMKII suggests that in the absence of a high local [Ca2+] at the NMDAR, wild‐type CaMKII might not reach sufficient levels of activity necessary for efficient autophosphorylation to interact with its binding partners (Figure 5, pathway (b)) and thus may not acquire a Ca2+‐independent active state.

Phosphorylation at residues 305/306 favors the dispersion of CaMKII subunits from the PSD (Strack et al, 1997) (Figure 5, pathway (c)), but is inhibited after the enzyme is trapped at the NMDAR (Bayer et al, 2001). Introducing TT305/306VA mutations into CaMKII rescued the wild‐type translocation phenotype in NR1R/− cells (Figure 4C). Most likely, net inhibitory TT305/306 phosphorylation is favored in these cells, as presumably less autophosphorylation at T286 and hence, less binding to NMDARs occurs.

Thus, we propose that in NR1R/− cells the local Ca2+ signal at the NMDAR pore is at a subthreshold level and is only compensated by mutations in CaMKII that strengthen retention of the kinase in the PSD. Consequently, the local Ca2+ signal created by direct influx through the NMDA receptor is necessary to maintain sufficient Ca2+/CaM‐dependent CaMKII activity for autophosphorylation to outcompete the main protein phosphatase at the PSD, PP1. This would then result in the ability of T286‐phosphorylated CaMKII to bind several PSD components (Figure 5, pathway (b)), locking it at the PSD.

Alternatively, the rescue by the CaMKII mutants could be explained in terms of CaM sensitivity. The T286D mutation mimicks autophosphorylation that increases affinity to CaM (Meyer et al, 1992). In CaMKII‐TT305/306VA, the inhibitory autophosphorylation sites that prevent CaM binding (Colbran, 1993) are abolished. This net increase in CaM affinity in both mutants should allow translocation also in response to a less‐pronounced elevation of internal Ca2+, possibly recruiting the kinase from longer distances or compensating for an eventual restricted availability of CaM in NR1R/− cells, as the shuttle of CaM between membrane and cytosolic compartments is itself a process regulated by Ca2+ (Persechini and Stemmer, 2002).

Input specificity of NMDAR‐mediated CaMKII translocation

The requirement of a locally restricted Ca2+ signal for CaMKII translocation, as indicated by our data, strongly suggest input specificity of the translocation process. While CaMKII mutants seem to accumulate at synapses with slow and indirect Ca2+ signals (as in NR1R/− cells), the wild‐type CaMKII seems to translocate only to an NMDAR‐activated synapse. No accumulation at a nearby synapse would occur. The biochemical properties of wild‐type CaMKII make it, therefore, an ideal candidate for input‐specific activation of NMDAR‐dependent signaling events such as synaptic plasticity. Indeed, activated CaMKII is translocated to the PSD where it can bind to NR2B subunits (Bayer et al, 2001), an interaction modulated by the strength of the synaptic input (Bayer et al, 2006) and required for expression of LTP (Barria and Malinow, 2005).

Implications of NMDAR Ca2+impermeability and lack of CaMKII translocation for the NR1R/− genotype

At the PSD, Ca2+ signals arise by influx via VGCCs and NMDARs or by release from intracellular stores and have been studied at single spine level (Emptage et al, 1999; Kovalchuk et al, 2000; Sabatini et al, 2002; Sobczyk et al, 2005). The PSD provides a high‐density network of signaling molecules where a variety of Ca2+ sensors (e.g., CaM, caldendrin) are positioned in close proximity of Ca2+ sources. This creates the conditions for a locally restricted Ca2+ signal, called ‘nanodomain’, as Ca2+ ions need to diffuse only molecular distances from its source to the target molecule (Augustine et al, 2003).

In NR1R/− cells, the lack of a local Ca2+ concentration right at the NMDAR channel pore will mainly affect proteins in close vicinity, especially those PSD components capable of direct interaction with the NMDAR itself, such as CaMKII.

The failure of NMDAR in NR1R/− cells to accumulate CaMKII at the synapse may result in a reduced presence of CaMKII at the PSD with consequences on the PSD composition. However, the overall levels of major neurotransmitter receptor subunits in membranes are normal (Rudhard et al, 2003) and expression of crucial synaptic components is not affected by the mutation in the NR1 subunit (Supplementary Figures S1 and S2).

Changes of CaMKII at the PSD can influence learning and memory (Elgersma et al, 2004). NR1R/− mice die at birth, hence no LTP or learning and memory experiments can be performed. However, the altered recruitment and distribution of CaMKII probably contribute to the observed lack of synaptic refinement of emerging somato‐sensory maps in NR1R/− mice (Rudhard et al, 2003). Effects of the NMDAR mutation on synaptic plasticity are therefore likely.

NMDAR‐dependent activation of CaMKII, along with its localized accumulation, will particularly alter the phosphorylation state of its PSD‐associated substrates (Trinidad et al, 2006). Therefore, a failure of NMDARs to activate and/or translocate CaMKII can affect further downstream signaling in NR1R/− synapses. Apart from CaMKII, other targets for NMDAR‐dependent Ca2+ signals exist, which in turn are linked to specific downstream signaling cascades (Hardingham et al, 2002), which might be affected by a lack of NMDAR‐dependent Ca2+ rise in the spine. For example, NMDA‐induced GluR2 endocytosis was found to require Ca2+ influx through a specific NMDAR subtype (Tigaret et al, 2006), which can be best explained by the functional Ca2+ nanodomain model.

In summary, Ca2+ influx through activated NMDAR channels generates a local Ca2+ signal at the channel pore that is necessary and sufficient for CaMKII translocation to and its retention at the PSD. This is functional evidence of an exquisite signal processing machinery associated very tightly with NMDARs and thus supports the concept of Ca2+ nano/microdomains in spines or at the PSD.

Materials and methods


Unless stated otherwise, all chemicals were from Sigma (St Louis, MO), Tocris (Tocris Cookson Ltd, Avonmouth, UK) or Invitrogen (Carlsbad, CA).

Hippocampal neuron cultures

Hippocampal primary neuronal cultures were prepared in accordance to a protocol by Banker and Goslin (1991), and described in detail by Specht et al (2005). Cells from individual pups were cultured separately.

Expression constructs

EGFP‐CaMKIIα. The expression construct pBS.EGFP‐MT‐CK2A codes for an N‐terminal fusion of CaMKIIα with EGFP similar to one described (Shen and Meyer, 1999). For details, see Supplementary methods.

For analysis of overall cell morphology, neurons were infected with Sindbis viral EGFP‐expression construct for 5 h (Specht et al, 2005).


Hippocampal neurons were typically transfected at day 14 in vitro (DIV) (range 13–15), with GFP‐CaMKIIα expression constructs with lipofectamine 2000 used as recommended by the manufacturer (Invitrogen).


For CaMKIIα translocation experiments, cells were washed 14–15 h after transfection with HEPES‐Ringer (in mM: HEPES pH 7.35, 5; KCl, 2.5; NaCl, 126; CaCl2, 2; MgCl2, 2; glucose, 11.1; tetrodotoxin (TTX), 0.0005; strychnine, 0.001; bicuculline methiodide, 0.02), and subsequently treated for 2–3 min with 50 μM Glu/10 μM Gly, or alternatively with AMPA or kainate at 100 μM in presence of 150 μM APV. For application of 50 μM NMDA/10 μM Gly, the same buffer was used without Mg2+. Controls were treated in parallel with the respective buffers containing no agonists. For solutions lacking Na+, NaCl was substituted with choline chloride.

Typically, data were obtained from three independent experiments. One experiment was performed per litter, and only litters were used which included at least one wild‐type pup. We analyzed the distribution of EGFP‐tagged CaMKII on fixed tissue, in analogy to (Rao and Craig, 1997; Ju et al, 2004; Hudmon et al, 2005), see below. Whereas this approach does not allow kinetic analysis of the translocation process, it permits reduced experimental variability by parallel processing of multiple samples under exactly the same conditions. In addition, data from more than one cell per sample can be obtained, which is more suited to the limited availability of cultures from mutant mice (see breeding schemes in Supplementary methods).


Fixation and staining. Cells (14–16 DIV for CaMKII translocation studies) were fixed with 4% paraformaldehyde/ 4% sucrose in PBS for 15–20 min at 35°C, permeabilized with 0.2% Triton X‐100 in PBS for 10 min, washed and blocked at 35°C in 5% horse serum (HS) in PBS. They were counterstained with rabbit anti‐synapsin1 (1:500; ab8, Abcam, Cambridge, UK) or mouse anti‐synapsin1 (1:500; clone 46.1, Synaptic Systems, Göttingen, Germany). Secondary antibodies conjugated with Cy3 and Cy5 were used 1:300 (at 2.5 μg/ml, Jackson ImmunoResearch, West Grove, PA). Coverslips were washed with PBS and mounted in GelMount.

Imaging. Cells were randomly selected on basis of healthy morphology. Confocal images were taken with a LSM 510 Meta laser scanning inverted microscope (Carl Zeiss Ltd, Rugby, UK), using a × 63 magnification oil immersion Plan‐Apochromat objective (NA 1.40), sequential line‐scan mode, × 2 scan averaging and 0.2 μm distance between optical sections. Images were analyzed using ImageJ software ( Confocal stacks were Z‐projected for each channel individually. Puncta were recognized as spherical structures of 0.2–0.5 μm, ⩾ × 2 brighter than the surrounding dendritic background and counted blindly with respect to the experimental conditions. This measurement was performed on dendritic segments (50–100 μm length) that typically belonged to second and third order branches. The selected segments fulfilled the criteria of minimal crossings with neighboring processes, and a detectable axial fluorescence. Synaptic location of NR1 or CaMKII puncta was determined by their proximity to synapsin1‐containing structures, within a mask of 0.28 × 0.28 μm, centered on the CaMKII (or NR1) spot. If immunopositive structures for synapsin1 were found within the field of the mask, the original structure was deemed as postsynaptic.

Ca2+ measurements

Solutions. Recording solutions contained in mM: HEPES, 5 (pH 7.35); KCl, 2.5; CaCl2, 2; glucose, 11.1; NaCl, 126; TTX, 0.5; TROLOX, 1; glycine, 0.01. In Ca2+‐free solutions, Ca2+ was exchanged for 1 mM of EGTA. For solutions lacking Na+, HEPES was pHed with KOH, KCl was added to 2.5 mM of K+, and NaCl was substituted with choline chloride. Agonist solutions contained in mM: NMDA, 0.05; APV, 0.15; nickel, 0.1; nimodipine, 0.01; nifedipine, 0.01; cyclopiazonic acid, 0.015; KB‐R7943, 0.03. Experiments testing nifedipine and KB‐R7943 effects were performed in carbonate‐buffered recording solution as described (Rudhard et al, 2003).

Recording. Cells (DIV12–20) were loaded for 20 min with 2 μM fura‐2‐AM (Molecular Probes Inc., Eugene, OR) in MM at 37°C. Coverslips were transferred into a recording chamber (Model RC‐26GLP, Warner Instruments Corp.) on a Nikon Eclipse E600FN microscope equipped with a × 40 and a × 60 oil immersion objective. They were perfused for 20 min with oxygenated recording solution to allow for hydrolysis of the fura‐ester. Recording and agonist solutions were applied at a flow rate of 2.2 ml/min. Fura excitation (340 and 380 nm) was achieved via a monochromator (Polychrom II, TILL Photonics, Martinsried, Germany) controlled by CellWorks (npi, Tamm, Germany). Emission was monitored via a photodiode (TILL Photonics), data were recorded with CellWorks and analyzed in Igor Pro (Wavemetrics Inc., Lake Oswego, OR). Alternatively, pictures were acquired with a CCD camera (Retiga 1300, Q Imaging, Burnaby, BC, Canada), and for illustration purposes thresholded to background values, ratio images calculated and false colored in Image J V1.29 (Rasband WS, NIH, Bethesda, MD).

High‐resolution Ca2+ imaging: fast confocal microscopy

Experiments were carried out in a dedicated BioRad Radiance 2100 imaging system integrated with patch‐clamp electrophysiology (Scott and Rusakov, 2006). Cultures were perfused with the bath medium at room temperature. Cells were filled with the cell‐impermeable fluorescent tracer Alexa Fluor 594 (20 μM) and Ca2+‐sensitive, cell‐impermeable indicator Fluo‐4 (200 μM, Kd=350 nM), in whole‐cell mode at Vh=−70 mV. The dyes were excited, respectively, at 568 and 488 nm, and the emission signals were chromatically separated at 560 nm and recorded using two photomultiplier tubes. Line‐scans were quantified as fluorescence values integrated over the imaged width of the cell compartment, with the background fluorescence subtracted. Recorded image stacks were analyzed offline (Scott and Rusakov, 2006).

To activate local NMDARs, we applied the agonist through a pressurized patch pipette (tip diameter 2–3 μm) placed in the 10–20 μm vicinity of the spine and filled with 100 μM NMDA in bath solution (plus 10 μM Alexa Fluor 594 to monitor pressure puffs).

Because the kinetics of Ca2+ sensitive responses in the spine vicinity was slower than the characteristic time of inter‐spine Ca2+‐and‐indicator equilibration due to diffusion reaction, we assessed Ca2+ concentration from Fluo‐4 fluorescence using the common steady‐state formula (Maravall et al, 2000).

Embedded Image

where Kd=350 nM, and F and Fmax stand for, respectively, the recorded fluorescence value and the maximum fluorescence recorded in conditions of dye saturation (achieved through bath application of ionomycin) in each cell.

Statistical analysis

Statistical analysis of data was performed using a two‐tailed Mann‐Whitney U‐test (Altman, 1999).

Supplementary data

Supplementary data are available at The EMBO Journal Online (

Supplementary Information

Supplemental Data [emboj7601420-sup-0001.doc]

Supplemental Methods [emboj7601420-sup-0002.doc]

Supplementary Figure S1 [emboj7601420-sup-0003.pdf]

Supplementary Figure S2 [emboj7601420-sup-0004.pdf]

Supplementary Figure S3 and S4 [emboj7601420-sup-0005.pdf]

Supplementary Figure S5 [emboj7601420-sup-0006.pdf]

Supplementary Figure S6 [emboj7601420-sup-0007.pdf]

Supplementary Figure S7, S8 and S9 [emboj7601420-sup-0008.pdf]


The NR1 antibody was a gift from FA Stephenson. We thank R Tsien for mRFP. We acknowledge C Specht and LA Cingolani for critical reading of the manuscript. This work was supported by the Wellcome Trust Senior Fellowships (RS and DAR), Wellcome Trust and BBSRC grants (RS); MRC and EU Promemoria grants (DAR). CMT was supported in part by a Wellcome Trust Traveling Fellowship.