Dynamic Control of CaMKII Translocation and Localization in Hippocampal Neurons by NMDA Receptor Stimulation

See allHide authors and affiliations

Science  02 Apr 1999:
Vol. 284, Issue 5411, pp. 162-167
DOI: 10.1126/science.284.5411.162


Calcium-calmodulin–dependent protein kinase II (CaMKII) is thought to increase synaptic strength by phosphorylating postsynaptic density (PSD) ion channels and signaling proteins. It is shown that N-methyl-D-aspartate (NMDA) receptor stimulation reversibly translocates green fluorescent protein–tagged CaMKII from an F-actin–bound to a PSD-bound state. The translocation time was controlled by the ratio of expressed β-CaMKII to α-CaMKII isoforms. Although F-actin dissociation into the cytosol required autophosphorylation of or calcium-calmodulin binding to β-CaMKII, PSD translocation required binding of calcium-calmodulin to either the α- or β-CaMKII subunits. Autophosphorylation of CaMKII indirectly prolongs its PSD localization by increasing the calmodulin-binding affinity.

CaMKII is involved in synaptic plasticity (1) and in the maintenance of the dendritic architecture (2). These different postsynaptic and dendritic functions of CaMKII make this enzyme a good model to investigate if and how signaling proteins are recruited to their sites of action. We studied the translocation of CaMKII isoforms by monitoring green fluorescent protein (GFP)–tagged CaMKII isoforms in cultured hippocampal CA1-CA3 neurons (3). GFP–α-CaMKII was uniformly distributed in the cell body and main processes of living neurons [Fig. 1A, (I)]. Glutamate stimulation induced a rapid enrichment of fluorescence in punctate structures along the dendritic processes and somatic cortex [Fig. 1A, (II)] (n = 32) (4). Alternatively, 10-s electric field stimuli of 50 Hz also triggered GFP-CaMKII translocation to similar punctate sites (5). The CaMKII translocation sites were likely in postsynaptic terminals, since they correlated with cross-points of transfected dendrites and untransfected neurites. Panel III of Fig. 1A shows an overlay of a fluorescence image of GFP–α-CaMKII (green) and a differential interference contrast (DIC) image before glutamate stimulation. The sites of GFP–α-CaMKII translocation can be visualized in a difference image (red, panel IV) by subtracting the fluorescence image before glutamate stimulation from the image after stimulation.

Figure 1

Glutamate-induced translocation of α-CaMKII to PSD. (A) (panel I) GFP–α-CaMKII–transfected CA3-CA1 neurons (12 days in culture) showed a uniform cytosolic fluorescence distribution. In panel II, the same neuron is shown after stimulation with 100 μM glutamate plus 10 μM glycine. A marked translocation of GFP–α-CaMKIIα to local sites can be observed after 20 s. Panel III shows an overlay of a DIC image and a fluorescence image of a neuron expressing GFP–α-CaMKII before glutamate stimulation (green). Panel IV shows the translocation to punctate sites using a subtraction analysis (red; the fluorescence image after glutamate addition was subtracted from the fluorescence image before addition). Arrows point to local sites where transfected dendrites make physical contacts with untransfected neurites. (B) Colocalization of GFP–α-CaMKII (left, green) with antibodies to the PSD marker PSD-95 (middle, red). Colocalization is apparent as a yellow color in the merged image (right). (C) (panel I) GFP–α-CaMKII transfected CA1-CA3 neuron before glutamate addition. In panel II, the same neuron is shown 20 s later after addition of 50 μM glutamate plus 5 μM glycine. In panel III, the same neuron is shown 2 min later, after the extracellular buffer was exchanged to a low Ca2+ solution without glutamate. In panel IV, the same neuron is shown 20 s later after the solution with Ca2+ and glutamate was added. Scale bars, 5 μm.

What are these punctate dendritic target sites? Because previous biochemical studies suggested that CaMKII is associated with postsynaptic densities (PSDs) and enriches in PSD fractions during ischemic conditions or tetraethylammonium stimulation protocols (6), we compared the localization of GFP–α-CaMKII to that of the PSD marker PSD-95 (Fig. 1B) (7). The punctate distribution of GFP–α-CaMKII after glutamate stimulation (green) showed a marked colocalization with antibodies to PSD-95 (red) (n = 7), suggesting that CaMKII translocates to PSD. A similar colocalization was observed between GFP–α-CaMKII and FM 4-64, a marker for functional presynaptic terminals juxtaposed to PSD (4). Together, these results suggest that GFP–α-CaMKII translocates to PSDs after glutamate or electrical stimulation.

The localization of GFP–α-CaMKII to PSD is a Ca2+-dependent process, because the translocation could readily be reversed by chelating extracellular Ca2+concentrations after glutamate addition. Panels I and II of Fig. 1C show the glutamate-triggered PSD translocation of GFP-CaMKII, while panel III shows the reversal by lowering extracellular Ca2+. Panel IV shows that the translocation to PSD could be repeated by subsequent addition of glutamate and extracellular Ca2+. This glutamate-induced translocation was repeatable at least four times (n = 3) (5). These findings not only show that the translocation of CaMKII is a Ca2+-mediated process but also suggest that PSD translocation is a dynamic state of CaMKII that reversibly enhances the phosphorylation of PSD-localized substrates.

Although the α isoform of CaMKII has been extensively studied, much less is known about β-CaMKII. We recently reported that β-CaMKII can bind to F-actin in vitro and colocalizes with F-actin in vivo when expressed in different cell types. Coexpression of β- and α-CaMKII leads to the formation of large hetero-oligomers which bind to F-actin (8). This F-actin localization of GFP–β-CaMKII can be seen in Fig. 2A (left). Glutamate stimulation triggered a rapid dissociation of GFP–β-CaMKII from F-actin and led to a nearly uniform cytosolic distribution of the kinase (Fig. 2A, middle). In a subsequent step, GFP–β-CaMKII translocated from the cytosol to PSD (Fig. 2A, right) (n = 22). The same F-actin dissociation and PSD translocation of CaMKII hetero-oligomers was also observed when GFP–α-CaMKII was expressed in the same neurons together with β-CaMKII (5).

Figure 2

Differential translocation kinetics of β- and α-CaMKII homo-oligomers and α/β-CaMKII hetero-oligomers. (A) Two-step translocation of GFP–β-CaMKII from F-actin to cytosol to PSD. GFP–β-CaMKII–transfected CA3-CA1 neurons showed a localized and filamentous staining pattern reminiscent of the distribution of F-actin (left) (3). Stimulation with 100 μM glutamate and 10 μM glycine induced a rapid loss in the F-actin staining pattern (middle). Persistent glutamate stimuli induced a second translocation step to the same punctate pattern observed for CaMKIIα (arrows, right). (B) Fluorescence recovery after photobleaching (FRAP) experiments in neurons transfected with GFP–α-CaMKII. The distribution before the photobleaching protocol (left). An individual dendritic site was photobleached 1 min after glutamate addition as indicated by the box in the middle image. Almost no recovery was seen 5 min after photobleaching (right). (C) Control measurement, showing rapid recovery of cytosolic GFP–α-CaMKII fluorescence in unstimulated neurons. (D) Time course of GFP–α-CaMKII translocation. Glutamate was added att = 0. (E) Time course of PSD translocation of α-CaMKII (square) and β-CaMKII homo-oligomers (circle), and α/β hetero-oligomers (triangle). Half-maximal translocation was 20, 80, and 280 s for α, α/β, and β oligomers, respectively. Traces represent average value of different experiments (n = 12 for α, n = 8 for β, n = 6 for α/β). (F) Time course of F-actin dissociation of GFP–β-CaMKII after glutamate addition (n = 6). Scale bars: 10 μm in (A) through (C), and 5 μm in (D).

We used photobleaching recovery to compare the binding of CaMKII to F-actin and PSD (9). Whereas F-actin–localized GFP–β-CaMKII recovered in less than 5 s from a local photobleaching laser pulse (8), the fluorescence recovery of PSD-bound GFP–α-CaMKII was longer than 5 min, suggesting that PSD-localized CaMKII is nearly immobile for the duration of glutamate stimulation (Fig. 2B). The same PSD immobilization was observed for GFP–β-CaMKII (5). As a control shows (Fig. 2C), cytosolic GFP–α-CaMKII recovered much more rapidly from a local photobleaching pulse. This suggests that the binding of CaMKII to PSD has a much higher affinity than its binding to F-actin.

Because the time required to recruit CaMKII to PSD is important to convert the duration of the synaptic input to the activation of CaMKII targets, we measured the translocation kinetics for CaMKII homo- and hetero-oligomers. Figure 2D shows a series of images used to analyze the translocation time course of GFP–α-CaMKII. The average relative increase in fluorescence intensity at each PSD was plotted as a function of time (10). The half-maximal time required for translocation was ∼20 s for GFP–α-CaMKII, increased to ∼80 s for GFP–α/β-CaMKII hetero-oligomers (1:1 ratio), and was longest (∼280 s) for GFP–β-CaMKII (Fig. 2E). In contrast, the dissociation of GFP–β-CaMKII from F-actin was half-completed within 5 s of glutamate addition (Fig. 2F). Thus, dissociation from F-actin is a rapid process, whereas translocation to PSD has a delay time that can be regulated in neurons by the ratio of β- to α-CaMKII subunits in each oligomer.

These measurements define three distinct localization states and two translocation steps of α/β-CaMKII hetero-oligomers. The first rapid translocation step requires a dissociation of β-CaMKII from F-actin, and the second delayed translocation step requires high-affinity binding of α- or β-CaMKII subunits to PSD-localized binding partners. What is the role of these two separate localization states? F-actin and PSD localization states might correspond to the two separate functions of CaMKII in the stabilization of dendritic arbors and synaptic plasticity, respectively. Although potential substrates of F-actin–localized CaMKII are unknown, the function of PSD-localized CaMKII is likely to phosphorylateN-methyl-D-aspartate (NMDA) receptors (11), AMPA (α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid) receptors (12), SynGAP (13), as well as other PSD-localized substrates involved in synaptic plasticity. Because an earlier study showed that autophosphorylated CaMKII can bind to the NR2B subunit of the NMDA receptor (14), this substrate of CaMKII is also a potential candidate for a PSD-binding partner.

In addition to being a substrate and potential binding partner, the NMDA receptor is a Ca2+ channel that may itself mediate CaMKII translocation by increasing Ca2+ concentration in the postsynaptic terminal. Indeed, the glutamate-induced translocation of GFP–α-CaMKII to PSD was blocked byD-2-amino-5-phosphonovalerate (APV), a selective NMDA receptor blocker (Fig. 3A) (n = 9), suggesting that NMDA receptors are necessary for CaMKII translocation. Furthermore, when NMDA, AMPA, and metabotropic glutamate receptors were selectively stimulated with NMDA, AMPA, and ACPD [trans-(1S, 3R)-1-amino-1, 3-cyclopentanedicarboxylic], respectively, only NMDA induced the PSD translocation of GFP-CaMKIIα (Fig. 3B) (n = 11). This suggests that Ca2+ influx through NMDA receptors is not only necessary but also sufficient for PSD translocation.

Figure 3

Regulation of PSD translocation of CaMKII by NMDA receptors, calmodulin-binding, and autophosphorylation. (A) PSD translocation does not occur in the presence of the NMDA-receptor blocker APV. Fluorescence images of a GFP–α-CaMKII–transfected neuron are shown before (left) and after (right) glutamate addition in the presence of 300 μM APV. (B) Addition of the NMDA-receptor agonist NMDA triggers translocation. Fluorescence images of a GFP–α-CaMKII–transfected neuron are shown before (left) and after (right) addition of 30 μM NMDA. (C) Schematic representation of mutant α-CaMKII constructs used for the characterization of the molecular requirements for PSD translocation. (D) Glutamate-induced translocation of a catalytically deficient GFP–α-CaMKII to PSD (K42R mutant). Fluorescence images of a neuron transfected with a K42R mutant GFP–α-CaMKII are shown before (left) and ∼1 min after (right) addition of 100 μM glutamate and 10 μM glycine. (E) Glutamate-induced translocation of an autophosphorylation-deficient GFP–α-CaMKII to PSD [T286A mutant, same protocol as in (D)]. (F) A GFP–α-CaMKII mutant deficient in calmodulin binding (A302R) fails to translocate to PSDs [same protocol as in (D)]. (G) Potential role of Thr286 phosphorylation in delaying the CaMKII dissociation from PSD. Time course of dissociation of GFP–α-CaMKII wild type (WT) (n = 8) and a T286A mutant (n = 8) from PSD. The glutamate solution was exchanged to a low Ca2+ solution without glutamate at t = 0. Scale bars, 5 μm.

Which molecular events determine the localization and translocation of CaMKII? Calmodulin binding and autophosphorylation are the two key events in CaMKII activation. Earlier studies showed that NMDA receptor stimuli activate α-CaMKII and trigger autophosphorylation of α- and β-CaMKII (15). Autophosphorylation leads to a partially active kinase (16) with a several hundred-fold increase in Ca2+-calmodulin–binding affinity (17).

CaMKII mutants were used to define the molecular steps of CaMKII translocation (18) (Fig. 3C). Because α-CaMKII does not bind F-actin, we used GFP–α-CaMKII instead of hetero-oligomers for the PSD translocation studies. Surprisingly, a kinase-inactive mutant [Lys42 → Arg42 (K42R)] (19) still translocated to the PSD, suggesting that kinase activity is not needed for PSD translocation (Fig. 3D) (n = 7). Consistent with this result, the autophosphorylation-deficient mutant [Thr286 → Ala286 (T286A)] (20) translocated to PSD, showing that autophosphorylation is also not a requirement for PSD translocation (Fig. 3E) (n = 12). Furthermore, the Thr286 → Asp286 mutant, which simulates the autophosphorylated CaMKII, did not prelocalize to PSDs (5) (n = 8). However, when calmodulin binding was disrupted in the Ala302 → Arg302(A302R) mutant (21), no PSD translocation was observed (Fig. 3F) (n = 6). This suggests that calmodulin binding, but not autophosphorylation, is necessary and sufficient for PSD translocation.

Even though autophosphorylation does not directly mediate PSD translocation, we determined whether autophosphorylation has an indirect role in PSD translocation by increasing the calmodulin-binding affinity (17). The prediction is that such an increase in binding affinity should prolong the association of CaMKII with PSD after a transient Ca2+ increase. Indeed, when the neuronal Ca2+ concentration was lowered by removing external glutamate and Ca2+, the autophosphorylation-deficient T286A mutant dissociated much more rapidly from PSD compared to the wild type (Fig. 3G). This finding adds a new perspective to the earlier suggestion that autophosphorylation is necessary for binding of CaMKII to PSD (6) or NMDA receptors (14). In the new model, autophosphorylation of CaMKII may function as a short-term biochemical “memory” by prolonging the localization of active CaMKII near its site of action.

We then explored the molecular requirements for β-CaMKII dissociation from F-actin, using model cells that contain F-actin but lack the PSD translocation event. Actin localization of GFP–β-CaMKII can be readily visualized in glial cells, which are rich in actin stress fibers (Fig. 4A, left), and in rat basophilic leukemia (RBL) cells, which are rich in cortical actin (Fig. 4B, left). The F-actin localization of β-CaMKII was lost after an increase in Ca2+ concentration (Fig. 4, A and B; right) (n = 25).

Figure 4

Dissociation of β-CaMKII from F-actin was triggered by either binding of Ca2+/calmodulin or Thr287 autophosphorylation. (A andB) F-actin dissociation of GFP–β-CaMKII in model cells. (A) Hippocampal glial cell transfected with GFP–β-CaMKII showed a distinct stress fiber staining pattern (left). Loss of stress fiber staining pattern 20 s after stimulation with 1 μM ionomycin (right). (B) RBL cells transfected with GFP-CaMKIIβ showed a cortical staining pattern reminiscent of the abundant cortical F-actin in these cells (left). Loss of cortical staining after ionomycin stimulation (right). (C) Schematic representation of mutant β-CaMKII constructs used for the characterization of the molecular requirements for F-actin dissociation. (D) Calcium ionophore induced F-actin dissociation of a catalytic activity–deficient GFP–β-CaMKII (K43R mutant). (Left) Distribution of the mutant before the Ca2+ increase. (Right) Distribution 2 min after addition of 1 μM ionomycin. (E) A GFP–β-CaMKII mutant deficient in calmodulin binding (A303R) fails to dissociate from F-actin after ionomycin addition. The F-actin prelocalization of this mutant construct is prominently visible in the left and right images. (F) A mutant GFP–β-CaMKII with a simulated autophosphorylation at Thr287 (T287D) fails to prelocalize to F-actin. The addition of Ca2+ionophore had no apparent effect on the localization of this mutant in the model cells. Scale bars, 5 μm.

We tested the β-CaMKII mutants (Fig. 4C) in the RBL cell model. The kinase-deficient GFP–β-CaMKII mutant (K43R) showed prelocalization and dissociation similar to wild-type GFP–β-CaMKII, suggesting that kinase activity is not required for F-actin dissociation (Fig. 4D, n = 14). A GFP–β-CaMKII deficient in calmodulin binding (A303R) bound effectively to F-actin and failed to dissociate after an increase in Ca2+concentration (Fig. 4E, n = 16). Consistent with this result, W-7, a calmodulin antagonist, also blocked the dissociation of GFP–β-CaMKII from actin cytoskeleton (5). However, no F-actin localization was observed when Thr287 was replaced by aspartic acid (T287D), even in the absence of an increase in Ca2+ concentration (Fig. 4F, left) (n = 18). This suggests that either Ca2+-calmodulin–binding or autophosphorylation at Thr287 are both sufficient to induce the dissociation of β-CaMKII from F-actin. These measurements define a separate cytosolic autophosphorylated state of CaMKII which cannot translocate to PSD unless calmodulin is bound.

We demonstrate the existence of distinct molecular states of α- and β-CaMKII, corresponding to an F-actin–bound, a cytosolic, and a PSD-bound form of the hetero-oligomeric kinase. Whereas calmodulin binding or autophosphorylation of β-CaMKII were both sufficient for the dissociation of CaMKII from F-actin, PSD translocation of CaMKII required the binding of calmodulin to α- or β-CaMKII. Because autophosphorylation induces a several hundred-fold increase in calmodulin-binding affinity (17), a role for autophosphorylation of α- or β-CaMKII is to delay calmodulin dissociation and thereby prolong the stimulus-induced CaMKII localization to PSD. Thus, two molecular events—calmodulin binding and autophosphorylation—are not only used to control CaMKII enzyme activity but also to dynamically regulate the translocation and localization of CaMKII with its cytoskeletal and synaptic target sites.

  • * To whom correspondence should be addressed. E-mail: tobias{at}


View Abstract

Navigate This Article