Translocation of autophosphorylated calcium/calmodulin-dependent protein kinase II to the postsynaptic density.

Calcium/calmodulin-dependent protein kinase II (CaMKII) undergoes calcium-dependent autophosphorylation, generating a calcium-independent form that may serve as a molecular substrate for memory. Here we show that calcium-independent CaMKII specifically binds to isolated postsynaptic densities (PSDs), leading to enhanced phosphorylation of many PSD proteins including the alpha-amino-3-hydroxy-5-methyl-4-isoxazole-propionic acid (AMPA)-type glutamate receptor. Furthermore, binding to PSDs changes CaMKII from a substrate for protein phosphatase 2A to a protein phosphatase 1 substrate. Translocation of CaMKII to PSDs occurs in hippocampal slices following treatments that induce CaMKII autophosphorylation and a form of long term potentiation. Thus, synaptic activation leads to accumulation of autophosphorylated, activated CaMKII in the PSD. This increases substrate phosphorylation and affects regulation of the kinase by protein phosphatases, which may contribute to enhancement of synaptic strength.

CaMKII␣ undergoes calcium/calmodulin-dependent autophosphorylation on Thr 286 in its regulatory domain, rendering the kinase partially calcium-independent (1,2). This reaction has been proposed as a "molecular switch," translating transient calcium elevation into prolonged kinase activity (15,16), which becomes subject to regulation by protein phosphatases. In addition, Thr 286 autophosphorylation promotes binding of CaMKII␣ to a 190-kDa PSD protein by gel overlay (17). The present results extend these findings, demonstrating that Thr 286 autophosphorylation controls subcellular targeting of CaMKII in neurons with important functional consequences.
Isolation of PSDs-PSDs were prepared from adult rat forebrains flash frozen within 45 s of euthanasia by detergent lysis of synaptosomes (20) except that 1 mM dithiothreitol, 1 mM phenylmethylsulfonyl fluoride, 1 mM benzamidine, and 10 g/ml leupeptin were included in all buffers. Synaptosomes were lysed in 1% (v/v) Triton X-100 and 150 mM KCl, and a second subsequent sucrose gradient was omitted because it yielded no further purification. PSDs displayed typical "donut" morphology by video-enhanced differential-interference contrast microscopy (21). PSDs prepared in the absence or the presence of the protein phosphatase inhibitor microcystin-LR (1 M) contained similar amounts of CaMKII protein (5-10% by quantitative immunoblotting (17)) and calcium/calmodulin-dependent CaMKII activity (100 -150 nmol/min/mg with 10 M autocamtide-2). Calcium-independent kinase activity was Ͻ2% and Ϸ10% of calcium/calmodulin-dependent kinase activity when PSDs were isolated in the absence or the presence of microcystin-LR, respectively. Therefore, CaMKII (Thr 286 ) was almost completely dephosphorylated during normal PSD isolation (minus microcystin-LR).
In Vitro Binding of CaMKII␣-PSDs and 35 S-or 32 P-labeled CaMKII␣ were incubated at the indicated concentrations, temperatures, and times in Buffer A (20 mM HEPES, pH 7.5, 10 mM dithiothreitol, 0.1% (v/v) Triton X-100, 1 mM benzamidine, 20 g/ml leupeptin) containing 1 mg/ml bovine serum albumin, 0.1 M NaCl, 1 M microcystin-LR, and EDTA and EGTA at 2.5-fold molar excess over Mg 2ϩ and Ca 2ϩ , respectively, carried over from the autophosphorylation. Binding was terminated by centrifugation of 90 l through 0.5 ml of HEPESbuffered 0.5 M sucrose cushion into 20 l of 10% glutaraldehyde in 0.6 M sucrose in a horizontal rotor (5000 ϫ g, 45-60 s). The sucrose cushions were aspirated (excess ligand), and the tube bottoms (PSD pellets) were cut off and counted.
Protein was quantitated as trichloroacetic acid-soluble (20%, w/v) [ 32 P]phosphate by scintillation counting. Blanks, with no PSDs in soluble substrate assays or 2.5 M microcystin-LR in "PSD complex" substrate assays, were subtracted from all assays to control for [ 32 P]phosphate present at the beginning of the incubation. Microcystin-LR completely blocked dephosphorylation in both assays. PP1 activity was calculated by subtraction of dephosphorylation measured in the presence of 0.2 M Inhibitor-2 (Upstate Biotechnology, Inc.) from total "constitutive" dephosphorylation. PP2A activity was defined as activity inhibited by 2.5 nM okadaic acid (LC Laboratories); activity measured in the presence of 2.5 M okadaic acid was assigned to unknown okadaic acid-resistant phosphatase(s); PP2C activity was defined as that stimulated by 10 mM magnesium acetate in the presence of 2.5 M okadaic acid. Each protein phosphatase activity was expressed as a percentage of constitutive activity. Activities do not add up to 100% because inhibitor concentrations were chosen for optimum selectivity, not maximal efficacy.
Phosphorylation of PSD Proteins and Immunoprecipitation-After incubation without or with 1 M [T286-P]CaMKII␣, PSDs (0.3 mg/ml) were washed by pelleting twice and suspended in Buffer A containing 10 mM magnesium acetate, 0.5 mg/ml bovine serum albumin, 1 M microcystin-LR, and either 0.5 mM CaCl 2 , 6 M calmodulin, or 1 mM EGTA. Phosphorylations were started by adding 0.5 M [␥-32 P]ATP (100 cpm/nmol) and stopped with 25 mM EDTA after 1 min on ice. GluR1 was immunoprecipitated (22) using antibody from Chemicon. No phosphorylated band co-migrating with GluR1 was detected in control precipitations with normal serum (not shown).
Translocation of CaMKII in Hippocampal Slices-Preparation and electrophysiological recordings from hippocampal slice were described previously (23). Four to six slices (400 m, 3-6 week old rats) were divided into control/experimental hemislices and maintained in oxygenated, artificial cerebrospinal fluid for 60 min at 32°C. After treatment with or without calyculin A (1 M for 60 min) or tetraethylammonium (TEA) (25 mM for 10 min, 15-min wash) slices were flash frozen in liquid N 2 . Slices were homogenized in 50 mM Tris, pH 7.5, 1% (v/v) Triton X-100, 0.1 M KCl, 20 mM sodium pyrophosphate, 50 mM NaF, 50 mM sodium ␤-glycerophosphate, 1 M microcystin-LR, 1 mM benzamidine, 20 g/ml leupeptin, and the PSD-enriched particulate fraction was isolated by centrifugation (30 min at 14,000 ϫ g). Crude PSD fractions were immunoblotted in triplicate for the indicated proteins using mixed alkaline phosphatase-and 125 I-labeled secondary antibodies. Gamma counts for CaMKII␣/␤ and GluR1 in each lane were normalized to neurofilament heavy subunit (NF-H) immunoreactivity. Normalized values for experimental slices were expressed as a percentage of control slice values. Quantifications were performed in the linear range of the assay, which was determined separately for each protein by probing serial extract dilutions.

RESULTS AND DISCUSSION
Thr 286 autophosphorylated CaMKII␣ has been shown previously to bind a 190-kDa PSD protein on gel overlays (17). One criticism of gel overlays is that denatured proteins are partially renatured prior to binding, potentially exposing binding domain(s) that are cryptic in native protein or PSDs. To examine possible interactions of CaMKII under native conditions, we investigated binding of [ 35 S]CaMKII␣ to isolated PSDs (Fig.  1A). Calcium/calmodulin-dependent autophosphorylation at Thr 306 , generating the calcium-independent form of the kinase ([T286-P]CaMKII), enhanced in vitro binding about 5-fold. Calcium/calmodulin-independent autophosphorylation at Thr 305 / Thr 306 , inactivating the enzyme ([T306-P]CaMKII), reduced binding somewhat below nonphosphorylated kinase. Binding of [T286-P]CaMKII by PSDs was independent of calcium/calmodulin or magnesium (not shown). The apparent binding was not Translocation of CaMKII to PSDs 13468 due to exchange of [ 35 S]CaMKII for endogenous enzyme but represented true accumulation of CaMKII␣ in the PSD (Fig.  1A, inset). [T286-P]CaMKII␣ binding to PSDs was specific and may be functionally relevant, because the low levels of binding to mitochondria and membranes were unaffected by autophosphorylation (Fig. 1A).
[T286-P]CaMKII␣ binding to PSDs was rapid, reaching saturation after 15 min at 4°C (Fig. 1B) or 5 min at 25°C (not shown). Whereas nonphosphorylated kinase bound reversibly (Ϸ25% dissociated in 3 h), binding of [T286-P]CaMKII␣ appeared essentially irreversible under these conditions (Ͻ3% dissociated in 3 h) (Fig. 1C). Interestingly, although binding to PSDs was enhanced by autophosphorylation, continued phosphorylation was not required to maintain the interaction because dissociation was not accelerated when dephosphorylation of bound kinase by endogenous protein phosphatases was allowed (calcium-independent CaMKII activity 17 versus 10% Ϯ microcystin-LR at 60 min) (Fig. 1C, open circle). This may explain why during PSD isolation CaMKII remains PSD-associated even though it is mostly dephosphorylated (see "Experimental Procedures"). In fact, [T286-P]CaMKII␣ dissociated from PSDs as slowly as endogenous CaMKII␣ measured by immunoblotting (not shown), suggesting that in vitro and in vivo association of CaMKII with the PSD are mechanistically similar. The reason for this very slow reversibility is not known, but mechanisms such as proteolysis (24,25) may be required to dissociate CaMKII from the PSD in vivo.
To assess the affinity of binding to PSDs, initial rates of [T286-32 P]CaMKII␣ binding were measured in the absence or the presence of nonradioactive kinase. Whereas non-P CaMKII␣ and [T306-P]CaMKII␣ were poor competitors, [T286-P]CaMKII␣ was an effective competitor (apparent IC 50 Ϸ1.1 M) (Fig. 1D). Because this IC 50 is approximately 10 times lower than the average concentration of CaMKII in forebrain (3,17), sufficient CaMKII exists for binding to be regulated in vivo. Significantly, this IC 50 is very similar to the affinity of [T286-P]CaMKII␣ for p190 estimated by gel overlay (17); p190 is therefore a candidate for targeting the calcium-independent form of CaMKII to PSDs.
It seemed important to determine which protein phosphatases (PPs) influence the lifetime of calcium-independent CaMKII in the PSD. CaMKII can be dephosphorylated by purified PP1 (26), PP2A (27), and PP2C (28), but not by PP2B. Each of these major phosphatases is present in PSDs, although only PP1 is enriched in this fraction (29). Their activities can be distinguished based on requirements for divalent cations and sensitivity to specific inhibitors (30). With soluble exogenous [T286-32 P]CaMKII␣ as the substrate, the PP2A:PP1 activity ratio was 3:1 (Fig. 2), in agreement with previous data (29). However, under these conditions, 5-10% of the substrate binds to PSDs. When assays were repeated using [T286-32 P]-CaMKII␣ previously bound to the PSD as a substrate, PP1 appeared to be mostly responsible for dephosphorylation (PP2A:PP1 activity ratio, 1:6; Fig. 2). 2 This is consistent with 2 Total activities in soluble and PSD-bound CaMKII phosphatase assays were 7.8 and 0.6 pmol/min/mg, respectively. It is difficult to interpret this difference because both substrate and enzyme are immobilized on the PSD in the latter assay. The apparent decrease in phosphatase activity after binding may reflect a preferential loss of the PP2A component or may simply be due to restricted mobility of PSDbound CaMKII.

FIG. 2. Dephosphorylation of [T286-P]CaMKII␣.
Protein phosphatases endogenous to isolated PSDs were allowed to dephosphorylate soluble and [T286-32 P]CaMKII␣ previously bound to PSDs. Specific inhibitors and activators were used to define contributions of PP1, PP2A, PP2C, and an okadaic acid-resistant activity (see "Experimental Procedures"). Binding of CaMKII␣ to PSDs had no effect on the dephosphorylation of glycogen phosphorylase a, a selective PP1 substrate, or casein, a substrate for PP2A and PP2C (not shown), demonstrating that CaMKII␣ binding to PSDs does not activate PP1 or inactivate PP2A. Similar data were obtained in four or five independent experiments. previous reports that dephosphorylation of CaMKII endogenous to PSDs is catalyzed by PP1 (29,31,32); however, CaMKII endogenous to isolated PSDs (i.e. in vivo translocated CaMKII) may represent a modified form of the kinase (46), possibly due to post-mortem ischemic conditions (11,47). The present data demonstrate that interaction of the calcium-independent form of CaMKII with PSDs directly regulates its inactivation, in that binding to the PSD in vitro converts CaMKII from a PP2A substrate to a substrate for PSD-bound PP1. The mechanism for this change in protein phosphatases responsible for dephosphorylating Thr 286 is unknown. CaMKII may undergo a conformational change after binding to PSDs that favors dephosphorylation by PP1. Alternatively, dephosphorylation by PP1 may be enhanced by physical proximity in the PSD.
AKAP79 is a protein that is thought to anchor inactive forms of protein kinase A, protein kinase C, and PP2B to PSDs (33,34). In contrast, CaMKII bound to PSDs remains active, because binding of [T286-P]CaMKII␣ to PSDs increased calciumindependent activity toward autocamtide-2 peptide substrate by an amount corresponding to binding of [T286-32 P]CaMKII␣ quantified in parallel reactions (not shown). More importantly, binding of [T286-P]CaMKII␣ (Ϸ10 g/mg PSD protein) led to a Ϸ2-fold increase of calcium-dependent phosphorylation and a Ϸ13-fold increase of calcium-independent phosphorylation of many proteins in PSDs (Fig. 3A).
In the presence of calcium/calmodulin, CaMKII phosphorylates GluR1 subunits of AMPA-type glutamate receptors in PSDs, increasing channel permeability and postsynaptic responses (22). GluR1 immunoprecipitation experiments were carried out with PSDs phosphorylated in the absence of calcium, with or without bound [T286-P]CaMKII␣. Phosphorylated GluR1 was immunoprecipitated only from PSDs to which [T286-P]CaMKII␣ had been bound (Fig. 3B), identifying GluR1 as a substrate for the PSD-bound, autophosphorylated, calcium-independent form of CaMKII.
In conclusion, the present data suggest a molecular mechanism for association of CaMKII with the PSD that is likely to be important in physiological and pathological states. Stimulation of CaMKII autophosphorylation by calcium influx into dendritic spines may increase the amount of active CaMKII in the PSD resulting in enhanced phosphorylation of GluR1 and other key substrates, in turn leading to an enhanced postsynaptic response. In addition, translocation may sequester CaMKII away from the cytosolic phosphatase activity of PP2A, making it available for dephosphorylation by PSD-bound PP1 only. Interestingly, PP1 itself is highly regulated by phosphorylation and association with targeting and inhibitory subunits (42) and is also involved in synaptic plasticity (43). Finally, because CaMKII has been proposed to play a structural role in the PSD (5), translocation of substantial amounts of cytosolic CaMKII to the PSD in response to calcium signals may result in long term changes in synapse morphology (44,45).