(Received for publication, February 12, 1997, and in revised form, March 27, 1997)
From the Department of Molecular Physiology and Biophysics, Vanderbilt University Medical Center, Nashville, Tennessee 37232-0615
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
-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.
CaMKII1 isoforms comprise a
family of broad specificity, calcium-activated kinases (1, 2). The and
isoforms are abundantly expressed in the brain, with
making
up as much as 2% of total protein in certain brain regions (3). CaMKII
is particularly enriched in PSDs (4, 5), cytoskeletal specializations
apposed to the postsynaptic membrane of excitatory synapses that are
thought to be scaffolds for neurotransmitter receptors, ion channels, and their postsynaptic modulators and effectors (reviewed in Refs. 6
and 7). Earlier reports suggested that CaMKII
constitutes as much as
50% of total PSD protein (8-10), but PSDs prepared from rapidly
homogenized brains are only 2-3-fold enriched in CaMKII
compared
with whole forebrain extracts (3, 11). CaMKII
knockout mice show
impaired hippocampal long term potentiation, a cellular model for
learning and memory (12). Conversely, introduction of CaMKII
into
neurons augments postsynaptic responses and occludes further
electrically induced long term potentiation (13, 14).
CaMKII undergoes calcium/calmodulin-dependent
autophosphorylation on Thr286 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, Thr286 autophosphorylation promotes binding of
CaMKII
to a 190-kDa PSD protein by gel overlay (17). The present
results extend these findings, demonstrating that Thr286
autophosphorylation controls subcellular targeting of CaMKII in neurons
with important functional consequences.
CaMKII was expressed in Sf9 cells and purified (17).
[35S]CaMKII
(
1200 cpm/pmol) was purified from cells
metabolically labeled with 35 µCi/ml [35S]methionine
for 58 h prior to harvesting. Kinase was autophosphorylated in the
presence (Thr286) or the absence
(Thr305/Thr306) of calcium/calmodulin (0.2-0.6
and 0.6-1.2 mol 32P/mol kinase subunit, respectively)
(17). Appropriate autophosphorylation of [35S]CaMKII
with unlabeled ATP was confirmed by assaying
calcium-dependent and -independent kinase activities (18)
using specific CaMKII substrate autocamtide-2 (10 µM)
(19).
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 µg/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 µM) 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 µM autocamtide-2).
Calcium-independent kinase activity was <2% and 10% of
calcium/calmodulin-dependent kinase activity when PSDs were
isolated in the absence or the presence of microcystin-LR,
respectively. Therefore, CaMKII (Thr286) was almost
completely dephosphorylated during normal PSD isolation (minus
microcystin-LR).
PSDs and 35S- or
32P-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 µg/ml leupeptin) containing 1 mg/ml bovine serum albumin, 0.1 M NaCl, 1 µM
microcystin-LR, and EDTA and EGTA at 2.5-fold molar excess over
Mg2+ and Ca2+, respectively, carried over from
the autophosphorylation. Binding was terminated by centrifugation of 90 µl through 0.5 ml of HEPES-buffered 0.5 M sucrose cushion
into 20 µl of 10% glutaraldehyde in 0.6 M sucrose in a
horizontal rotor (5000 × g, 45-60 s). The sucrose cushions were aspirated (excess ligand), and the tube bottoms (PSD
pellets) were cut off and counted.
Soluble
[T286-32P]CaMKII (0.1-0.2 µM subunit)
was incubated with 25-50 µg/ml isolated PSDs, whereas resuspended
PSD·[T286-32P]CaMKII
complexes (2-5 µg kinase/mg
PSD) were incubated at a final concentration of 70 µg/ml PSD protein.
Incubations were conducted for 30 min at 30 °C in Buffer A
containing 1 mM EGTA, 1 mg/ml bovine serum albumin, and 0.1 M NaCl, plus specific inhibitors or activators indicated
below. Phosphatase activity was quantitated as trichloroacetic
acid-soluble (20%, w/v) [32P]phosphate by scintillation
counting. Blanks, with no PSDs in soluble substrate assays or 2.5 µM microcystin-LR in "PSD complex" substrate assays,
were subtracted from all assays to control for [32P]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 µM 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 µM 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 µM 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.
After incubation without or with 1 µM [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 µM
microcystin-LR, and either 0.5 mM CaCl2, 6 µM calmodulin, or 1 mM EGTA. Phosphorylations
were started by adding 0.5 µM [
-32P]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).
Preparation
and electrophysiological recordings from hippocampal slice were
described previously (23). Four to six slices (400 µm, 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 µM for 60 min) or tetraethylammonium (TEA) (25 mM for 10 min, 15-min
wash) slices were flash frozen in liquid N2. 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 µM microcystin-LR, 1 mM benzamidine, 20 µg/ml leupeptin, and the PSD-enriched particulate fraction was isolated by centrifugation (30 min at 14,000 × g).
Crude PSD fractions were immunoblotted in triplicate for the indicated
proteins using mixed alkaline phosphatase- and 125I-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.
Thr286 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 [35S]CaMKII
to isolated PSDs (Fig.
1A). Calcium/calmodulin-dependent autophosphorylation at Thr306, generating the
calcium-independent form of the kinase ([T286-P]CaMKII), enhanced
in vitro binding about 5-fold.
Calcium/calmodulin-independent autophosphorylation at
Thr305/Thr306, 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 due to exchange of [35S]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 (
25%
dissociated in 3 h), binding of [T286-P]CaMKII
appeared
essentially irreversible under these conditions (<3% 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% ± 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
[T286-32P]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 IC50
1.1
µM) (Fig. 1D). Because this IC50
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 IC50 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
[T286-32P]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
[T286-32P]-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 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 Thr286 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 calcium-independent activity
toward autocamtide-2 peptide substrate by an amount corresponding to
binding of [T286-32P]CaMKII
quantified in parallel
reactions (not shown). More importantly, binding of [T286-P]CaMKII
(
10 µg/mg PSD protein) led to a
2-fold increase of
calcium-dependent phosphorylation and a
13-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.
Brain injuries involving increased intracellular calcium such as
ischemia/hypoxia (11, 35), hypoglycemia (36), and excitotoxic insults
(37) cause CaMKII translocation to the cytoskeleton, including PSDs.
Ischemia also transiently increases calcium-independent CaMKII activity
(38). We therefore tested the hypothesis that Thr286
autophosphorylation is sufficient for translocation of CaMKII to occur
in neurons. Calyculin A, a cell-permeant protein phosphatase inhibitor,
was shown previously to increase Thr286 phosphorylation in
hippocampal slices (39) without affecting general excitability or
viability (40). Here, calyculin A increased calcium-independent CaMKII
activity in hippocampal slice extracts from 25.6 ± 1.2% to
43.4 ± 3.7% (n = 6, p < 0.01),
accompanied by a >2-fold increase in CaMKII associated with
PSD-enriched fractions (Fig. 4, A and
C). We next induced a form of long term potentiation with
the K+-channel blocker TEA (41). There was a 83.6 ± 21.7% (n = 4, p < 0.05) enhancement
of transmission at CA3-CA1 synapses (Fig. 4B) 15 min after
TEA removal, accompanied by increased calcium-independent CaMKII
activity (20.3 ± 1.9% to 26.7 ± 4.2%, n = 5, p = 0.07) and a 70-80% increase of CaMKII protein
in the PSD-enriched fraction (Fig. 4C). Thus, CaMKII
translocation occurs in intact neurons in response to treatments that
induce CaMKII autophosphorylation and potentiate synaptic
transmission.
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).