The role of CaMKII as an F-actin-bundling protein
crucial for maintenance of dendritic spine structure
Ken-Ichi Okamoto*, Radhakrishnan Narayanan*, Sang H. Lee†, Kazuyoshi Murata‡, and Yasunori Hayashi*§
*RIKEN-MIT Neuroscience Research Center, The Picower Institute for Learning and Memory, Department of Brain and Cognitive Sciences, Massachusetts
Institute of Technology, Cambridge, MA 02139;†Department of Pharmacology, BSB 608, Medical College of Wisconsin, 8701 Watertown Plank Road,
Milwaukee, WI 53226; and‡Whitehead Institute, Department of Biology and Division of Biological Engineering, Massachusetts Institute
of Technology, Cambridge, MA 02142
Communicated by Susumu Tonegawa, Massachusetts Institute of Technology, Cambridge, MA, February 27, 2007 (received for review October 25, 2006)
Ca2?-calmodulin-dependent protein kinase II (CaMKII) is a serine/
threonine protein kinase critically involved in synaptic plasticity in
the brain. It is highly concentrated in the postsynaptic density
fraction, exceeding the amount of any other signal transduction
molecules. Because kinase signaling can be amplified by catalytic
reaction, why CaMKII exists in such a large quantity has been a
mystery. Here, we provide biochemical evidence that CaMKII is
capable of bundling F-actin through a stoichiometric interaction.
Consistent with this evidence, in hippocampal neurons, RNAi-
mediated down-regulation of CaMKII leads to a reduction in the
volume of dendritic spine head that is mediated by F-actin dynam-
ics. An overexpression of CaMKII slowed down the actin turnover
in the spine head. This activity was associated with ? subunit of
CaMKII in a manner requiring its actin-binding and association
domains but not the kinase domain. This finding indicates that
CaMKII serves as a central signaling molecule in both functional
and structural changes during synaptic plasticity.
cytoskeleton ? plasticity ? synapse
This phenomenon, long-term potentiation (LTP), has been con-
sidered as a cellular counterpart of memory, and its molecular
mechanisms have been intensively studied. A strong postsynaptic
depolarization caused by high-frequency input induces a Ca2?
influx through postsynaptic NMDA receptors, which then triggers
a series of biochemical processes, including an activation of a
serine/threonine protein kinase, Ca2?/calmodulin-dependent pro-
tein kinase II (CaMKII). The activation of CaMKII is followed by
a series of autophosphorylation events, which enables active
CaMKII to stay activated until all subunits are dephosphorylated
(2). This activated CaMKII phosphorylates AMPA receptor and
various other postsynaptic proteins, which triggers a mechanism
synaptic transmission (2–9).
In addition to the functional change typically measured by
electrophysiological recordings, recent works revealed another
aspect of synaptic plasticity: the structural changes associated with
the functional modifications (10–13). During LTP of hippocampal
CA1 pyramidal cells, a dendritic spine, the tiny protrusion where
most of excitatory synapses are harbored in this class of neurons,
expands within ?60 s after the LTP-inducing stimulation (12, 13).
Antagonists for both NMDA receptor and CaMKII blocked the
effect of tetanic stimulation; therefore, the activation of NMDA
receptor and CaMKII are both important for structural plasticity
(12, 13). This observation encouraged us to explore the role that
CaMKII activity plays in the regulation of postsynaptic structure.
CaMKII is found in a large quantity in postsynaptic density
fraction much more than any other signal transduction molecules,
almost comparable with cytoskeletal components (14, 15). This
knowledge led us and others to speculate that CaMKII may serve
not only as a signal transduction molecule but also as a structural
element at synapses (2, 14). In relation to this speculation, it has
n hippocampal CA1 pyramidal cells, a transient burst of synaptic
input potentiates the efficiency of subsequent transmission (1).
sequence present in the ? subunit (16–20), but this interaction has
been considered to anchor CaMKII to the postsynaptic sites.
However, considering that CaMKII is an oligomeric protein com-
posed of 12–14 subunits with rotational symmetry (2), it is possible
that one oligomer can simultaneously bind to multiple F-actin,
thereby acting as an F-actin-bundling protein at postsynaptic sites.
Here, we show evidence that CaMKII is in fact not only a
signaling protein, but also a structural protein that bundles F-actin
through a specific and stoichiometric interaction mediated by the ?
subunit, and that such capacity is required for the maintenance of
the postsynaptic architecture of the dendritic spine. This work
ascribes a specific, previously undescribed role for CaMKII as a
cytoskeletal component in the postsynaptic structure.
F-Actin Is Bundled by CaMKII. To demonstrate the F-actin-bundling
activity of CaMKII, we first used the conventional F-actin cosedi-
mentation assay (21). Purified non-muscle actin was polymerized
and coincubated with purified CaMKII (heteromeric complex of ?
and ?). The resultant complex was centrifuged at 10,000 ? g, at
which linear, unbundled F-actin stays in supernatant but bundled
F-actin precipitates (21). Upon separation of both fractions on an
was absent, but in the pellet when CaMKII was present (Fig. 1a).
CaMKII without F-actin did not form a pellet, indicating that
CaMKII itself did not form complexes that precipitate, and that an
interaction between F-actin and CaMKII was indeed needed to
form the pellet. This activity was associated with ? subunit of
CaMKII, but not with the ? subunit, a finding consistent with the
presence of an F-actin binding sequence in CaMKII? but not in ?.
The concentration response curve of CaMKII? at a fixed concen-
tration of actin indicates that the maximum effect was attained at
a molar ratio of actin to CaMKII? of higher than 1:1 (or 1:?0.08,
as CaMKII? holoenzyme; Fig. 1b). This stoichiometry of actin and
CaMKII is reasonable in view of the amount of these two proteins
present in the postsynaptic density fraction (15).
We next attempted to visualize the F-actin bundling by CaMKII.
We negatively stained F-actin filaments formed in the presence or
absence of CaMKII and observed them using electron microscopy
In contrast, in the presence of CaMKII, F-actin formed clear
bundles. Again, this activity was associated with CaMKII?; homo-
Author contributions: K.-I.O. and R.N. contributed equally to this work; K.-I.O., R.N., S.H.L.,
and Y.H. designed research; K.-I.O., R.N., and S.H.L. performed research; S.H.L. and K.M.
contributed new reagents/analytic tools; K.-I.O. and R.N. analyzed data; and K.-I.O., R.N.,
and Y.H. wrote the paper.
The authors declare no conflict of interest.
Abbreviations: LTP, long-term potentiation; CaMKII, Ca2?-calmodulin-dependent protein
kinase II; FRAP, fluorescent recovery after photobleaching; shRNA, short hairpin RNA; CBB,
Coomassie Brilliant Blue.
§To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
This article contains supporting information online at www.pnas.org/cgi/content/full/
© 2007 by The National Academy of Sciences of the USA
April 10, 2007 ?
vol. 104 ?
no. 15 www.pnas.org?cgi?doi?10.1073?pnas.0701656104
meric CaMKII? failed to show such activity. This assay, in con-
junction with the cosedimentation assay, establishes that CaMKII
is an F-actin-bundling protein, consistent with a recent report (20).
CaMKII Does Not Affect Kinetics of Polymerization and Depolymer-
ization of Actin. Given the bundling activity of CaMKII?, we
wondered whether binding of CaMKII affects the kinetics of actin
polymerization or depolymerization. To test this hypothesis, we
used standard fluorescence measurement of actin polymerization/
depolymerization using pyrene conjugated actin (22). The poly-
merization was initiated by the addition of actin polymerization
buffer to G-actin in the presence and absence of CaMKII. A
sigmoidal curve showing the lag phase, the log phase and the
plateau phase of polymerization was observed. The addition of
CaMKII, both unphosphorylated (Fig. 2a) and phosphorylated
(data not shown), did not largely change the actin polymerization
2b). We induced depolymerization of preformed pyrene-labeled
F-actin by diluting 10-fold in the presence or absence of CaMKII?.
The addition of CaMKII did not change the rate of depolymeriza-
tion significantly. We used the actin polymerizing drug phalloidin
as a control for these experiments (data not shown). These results
indicate that despite clear bundling of F-actin, the kinetics of
polymerization and depolymerization are largely unaltered by
CaMKII? but Not -? Is Necessary for Maintaining Synaptic Structure.
F-actin composes the primary cytoskeleton in the dendritic spine,
23–25). Because CaMKII bundles F-actin in vitro, it is likely that
CaMKII also plays a structural role in dendritic spines by bundling
F-actin. To test whether endogenous CaMKII in dendritic spines
has such a role, we first tested the effects of specific reduction of
CaMKII? and -? proteins by using RNAi and then observing the
resulting dendritic spine structure. We constructed plasmid-based
short hairpin RNA (shRNA) expression vectors, which specifically
down-regulate CaMKII? or -? respectively (Fig. 3a). They were
biolistically transfected in CA1 pyramidal cells in hippocampal
organotypic slice culture along with GFP to visualize the structure
have any effect on the number and structure of dendritic spines
compared with vector alone (Fig. 3 b–e). In contrast, the shRNA
against CaMKII? significantly reduced the number of mature
dendritic spines and increased the number of long filopodial
structures often seen in immature neurons (Fig. 3 b–d). Impor-
tantly, the total number of protrusions was unchanged with these
manipulations, indicating that the effect of CaMKII? shRNA was
not to generate new filopodial structures (Fig. 3e). This effect was
completely rescued by cotransfecting an expression vector of
CaMKII? cDNA with silent mutations at the shRNA target
sequence (indicated as ?Rin Fig. 3) but to a lesser extent by using
the expression vector carrying the original CaMKII? cDNA, con-
firming the specificity of the effect of shRNA. Thus, we show that
CaMKII, specifically the ? subunit, is necessary for the mainte-
nance of mature spine structure.
F-actin. F-actin formed in vitro was allowed to react with purified CaMKII?/?, -?,
Under this condition, the linear, unbundled F-actin stays in supernatant but
SDS/PAGE and stained with CBB. (b) Dose-dependent bundling of F-actin by
CaMKII?. F-actin (3 ?M as monomer) was incubated with increasing concentra-
tions of CaMKII? (0.1–4.25 ?M as monomer) and actin-bundling assays were
performed. The amount of bundled F-actin was plotted against the concentra-
purified CaMKII, all at 3 ?M as monomer.
CaMKII bundles F-actin. (a) Cosedimentation of purified CaMKII and
unchanged by CaMKII. Effect of CaMKII on the kinetics of polymerization (a)
and depolymerization (b) was examined by using pyrene-labeled actin. (a)
polymerization buffer in the presence (gray line) or absence (black line) of
CaMKII? (2 ?M as monomer). (b) Actin filament disassembly was induced in
the absence (black line) or presence of 2 ?M CaMKII? (gray line) by dilution of
preformed 5% pyrene-labeled actin filaments to a final concentration of 0.2
?M. Error bars indicate SEM of three experiments, shown every 1 min.
Kinetics of polymerization and depolymerization of actin is largely
Okamoto et al.PNAS ?
April 10, 2007 ?
vol. 104 ?
no. 15 ?
Maintenance of Dendritic Spine Structure Is Independent of Kinase
Activity or Domain Itself. We next tested whether kinase activity is
necessary for maintaining dendritic spine structure by CaMKII?.
This hypothesis was tested by rescuing the effect of shRNA using
CaMKII?RK43R, a mutant lacking kinase activity combined with
actin-bundling activity comparable with wild type CaMKII? in a
biochemical experiment (see below). If the kinase function is not
necessary for spine structure maintenance, this mutant should be
sufficient to rescue the spine structure phenotype observed with
CaMKII? shRNA. As predicted, we found that CaMKII?RK43R
fully rescued the phenotypes observed with shRNA against
CaMKII? (Fig. 3 b–e).
independent of the kinase activity. For example, a synaptic scaf-
folding protein CASK/mLIN-2 has a domain similar to the kinase
domain of CaMKII, which serves as an interaction interface with
Mint-1/mLIN-10 but has no kinase activity (26). Therefore, it is
possible the K43R mutant may still have such function. To rule out
this possibility, we tested two deletion mutants CaMKII?285–542
and ?344–542 lacking kinase domain but still having the F-actin
binding and association domains. These deletion mutants also fully
rescued the effect of CaMKII? shRNA (Fig. 3 b–e). This result
indicates that the F-actin binding domain oligomerized together
through the associating domain is sufficient for the rescue. These
results are most compatible with the idea that the major role of
CaMKII in unstimulated synapse is to bundle F-actin and such
activity is important for the maintenance of mature dendritic spine
CaMKII? but Not -? Affects the Turnover of Actin in Dendritic Spine.
Because we have shown that CaMKII? is an F-actin-bundling
protein in vitro, we next attempted to observe evidence of the
bundling of F-actin by CaMKII in the postsynaptic structure.
However, likely because of high density of proteins including actin
itself and possible instability during processing, it has been difficult
to reveal discrete structure of F-actin in dendritic spine both under
not necessarily give insight into the chemical and functional prop-
erties of actin. It has been demonstrated that the actin molecules in
dendritic spines undergo constant turnover at a rate regulated by
neuronal activity (12, 25). Therefore, we used fluorescent recovery
after photobleaching (FRAP) assay to monitor the turnover of
actin molecules in dendritic spines. We overexpressed untagged
CaMKII? along with GFP-actin. We assumed that such an over-
expression would result in the formation of an oligomer dominant
in the ? subunit, having a greater number of F-actin binding sites
and therefore more efficient stabilization of F-actin at dendritic
spines compared with endogenous CaMKII, which is composed of
both ? and ? subunits. GFP-actin in a single dendritic spine was
photobleached and thereafter the recovery of its fluorescence was
monitored (Fig. 4). The GFP signal recovered to its original level
of overexpressed CaMKII?, the turnover was significantly slower,
whereas CaMKII? did not have such an effect (Fig. 4). We could
not compare actin turnover in shRNA expressing neurons because
the structure of spines were altered dramatically by reduction of
CaMKII? (Fig. 3).
The effect of CaMKII? on actin turnover did not require kinase
Spine head size
( /10 µm)
- - -
- - -
- - -
duces mature dendritic spines and
increases filopodial protrusions. (a)
The effectiveness and specificity of
RNAi. GFP-tagged CaMKII? and -?
were transfected in the Cos7 cells
with and without shRNA expression
vectors and then blotted with anti-
GFP antibody. (b) Two-photon mi-
croscopic images of dendritic seg-
ment in CA1 pyramidal neurons
transfected with control shRNA vec-
tor or expression vectors for shRNA
against CaMKII? or -?. To test the
specificity of the effect, the expres-
sion vectors for CaMKII? cDNA with
(indicated by ?R) or without silent
mutations at the shRNA target se-
quence or deletion mutants were
used to rescue the observed pheno-
type. CaMKII?RK43R and two dele-
deletion mutants. The neuronal
structure was visualized by coex-
pressing with GFP. (c–e) Plots of the
average of the spine length (c), the
size of spine head (d), and the den-
sity of spines including filopodial
protrusions (e). The size of spine
head was measured by ratio of flu-
orescent intensity at the spine head
(sh) and the dendritic shaft (ds).
Number of observations was as fol-
lows (number of cells/number of
spines); RNAi vector only, 18/172;
CaMKII? shRNA, 14/261; CaMKII? shRNA, 15/163; CaMKII? shRNA plus CaMKII? cDNA, 24/174; CaMKII? shRNA plus CaMKII?RcDNA, 17/210; CaMKII? shRNA plus
CaMKII?RK43R cDNA, 20/253, CaMKII? shRNA plus CaMKII? 285–542, 20/297; CaMKII? shRNA plus CaMKII? 344–542, 20/197. Statistical significance compared
with vector only is indicated. There was a slight increase in the head volume in cells expressing CaMKII? shRNA plus CaMKII? 285–542 (d) and in the density in
cells expressing CaMKII? shRNA plus CaMKII? 344–542 (e). The exact mechanism is currently unknown.
Reduction of CaMKII? re-
www.pnas.org?cgi?doi?10.1073?pnas.0701656104 Okamoto et al.
same effect on actin turnover as wild-type CaMKII? (Fig. 4 d and
e). Also, with deletion mutants lacking the kinase domain but
containing actin binding and association domains (CaMKII?285–
542 and ?344–542), we observed the same effect as full length
CaMKII? (Fig. 4 d and e). Rather it requires oligomerization of
actin-binding domain of CaMKII?, because a deletion mutant of
on actin turnover (Fig. 4 d and e).
Presently, it is not clear how CaMKII? slows down FRAP of
GFP-actin. Whereas CaMKII did not have a direct impact on
polymerization and depolymerization kinetics of actin (Fig. 2), in
dendrites, CaMKII bundling of F-actin may cause inhibition of
diffusion or access to actin depolymerizing/severing proteins that
affect F-actin dynamics. Given this, however, the most important
finding here is that CaMKII affects the actin turnover in a way
independent of kinase activity or domain and rather requires an
intact actin binding domain associated with each other. This is not
expected if CaMKII is purely a signal transduction protein or if the
Rather, it is most reasonably explained by a structural function of
CaMKII to bundle F-actin.
Activation of CaMKII Leads to a Loss of Bundling Activity. We finally
wanted to see whether the kinase activity is involved in the
regulation of actin bundling. To test this, we preformed bundled
F-actin with CaMKII? and subsequently activated the enzyme by
addition of Ca2?/calmodulin. To reveal a specific effect of auto-
Ca2?/calmodulin complex from CaMKII and the coincubation was
after CaMKII was activated by Ca2?/calmodulin (Fig. 5a). The
reversal of bundling after activation by Ca2?/calmodulin increased
with longer exposure of CaMKII to Ca2?/calmodulin. This was not
observed with a K43R mutant, confirming the importance of
autophosphorylation (Fig. 5b). However, the F-actin-bundling ac-
tivity of K43R was also blocked by Ca2?/calmodulin if EGTA was
not added and Ca2?/calmodulin remained attached to the kinase
throughout the bundling and sedimentation steps (Fig. 5b), con-
firming previous results that calmodulin binding impairs F-actin
binding (18, 19). Taken together, both Ca2?/calmodulin binding
the actin binding region of CaMKII contains multiple serines/
threonines, many of which are actually autophosphorylated rapidly
upon activation (M. K. Hayashi, R. Narayanan, and Y. Hayashi,
0 120 240 360
0 120 240 360
300 60 90 (sec) -15-30 -45
Time constant (sec)
* * * *
Time constant (sec)
F-actin binding domain
slows down the turnover of actin in the den-
dritic spine. (a) Images of dendritic segments
expressing GFP-actin with (Right) or without
(Left) untagged CaMKII?. (b) An example of
FRAP assay. Images shown are time-lapse of sin-
gle dendritic spine head expressing GFP-actin
before and after photobleaching. (c) Averaged
FRAP of GFP-actin alone (Left) and GFP-actin
coexpressed with untagged CaMKII? (Right).
The first time point after photobleaching was
taken as time 0. (d) CaMKII? deletion mutants
used in FRAP assay. (e) Effect of CaMKII?,
CaMKII?, ?K43R, and deletion mutants of
The time course of FRAP in individual spine was
obtained by fitting with a single exponential
curve and plotted. (Upper) Plots of the same
data. (Lower) Cumulative plot of the time con-
as follows: GFP-actin only, 13; CaMKII?, 11;
CaMKII?, 12; CaMKII? K43R, 9; ?1–401, 11;
?285–401, 10; ?285–542, 11; ?344–542, 10.*,
Statistical significance compared with GFP-actin
only (P ? 0.05).
Interaction between actin and CaMKII
Ca2?/calmodulin and resultant autophosphorylation. (a) CaMKII? was preincu-
then followed by F-actin bundling and sedimentation steps. ?*, Gel-shifted
with CBB likely because of heavy autophosphorylation and resulting incorpora-
tion of negative charges. (b) A similar experiment to a using K43R (kinase-null)
mutant of CaMKII?. EGTA –, A sample in which EGTA was not added and
Ca2?/calmodulin binding independent of autophosphorylation.
Inhibition of F-actin-bundling activity of CaMKII? after activation by
Okamoto et al. PNAS ?
April 10, 2007 ?
vol. 104 ?
no. 15 ?
F-actin. This requires the presence of the ? subunit in the CaMKII
oligomer but not kinase activity itself. This feature of CaMKII is
necessary for maintaining the dendritic spine structure. Not only
that, when kinase is activated, CaMKII looses the ability to bundle
F-actin. Both binding with Ca2?/calmodulin and autophosphory-
lation of CaMKII contribute to this event. It has been speculated
that CaMKII might have a structural role because of its abundance
in the postsynaptic density (2, 14); our report demonstrates a
specific structural role for CaMKII as an F-actin-bundling protein
necessary for maintenance of postsynaptic structure.
Previous reports indicate that the overexpression of CaMKII?
but not ? in neurons in dissociated culture increases filopodia
motility, dendritic arborization, and spine density (27, 28). These
effects were abolished by blocking kinase activity with a point
mutation or a kinase inhibitor. Whereas it is possible that the
precise location of kinase activity is important for these effects
(27), it is also possible that the bundling of F-actin and its
CaMKII accumulates on the dendritic spine upon neuronal
stimulation (19, 20, 29–31). Because CaMKII? does not bundle
F-actin but still shows activity dependent accumulation, the accu-
mulation is independent of F-actin binding (19). Instead, self-
association of CaMKII (20, 30) or binding with the NR2B subunit
of NMDA receptors (31), densin-180 (32, 33), and/or ?-actinin
In view of these results, we suggest the following model for the
role of CaMKII in structural plasticity (Fig. 6). At resting synapses,
F-actin is bundled by CaMKII, thereby maintaining a stable struc-
ture. When CaMKII is activated by neuronal activity and the
resultant Ca2?influx, it detaches from F-actin and allows its
reorganization by other signal transduction machineries, such as
small G proteins. Meanwhile, more CaMKII are recruited to the
synapse by self-association or interaction of NR2B to play a role as
level for the amount of CaMKII at dendritic spines, which may in
turn affect the amount of bundled F-actin in the spines, and
consequently the structure of the dendritic spines. Upon returning
to the unphosphorylated state, CaMKII bundles the newly reorga-
nized F-actin and maintains the remodeled spine structure. In this
dendritic spine structure constant at resting Ca2?levels, allowing
modification when the Ca2?level increases, and subsequently
preserving the plastic change. Duration of activated CaMKII gives
a time-window during which actin can be remodeled. Consistent
with this view, the structural plasticity of dendritic spines requires
CaMKII activation (13). In this way, CaMKII plays a dual role in
excitatory synaptic plasticity: a signaling role during neuronal
activity, and an actin-bundling role during the basal state in
dendritic spines, thereby affecting both functional and structural
plasticity of dendritic spines
Expression Vectors. BaculoviralexpressionvectorsforratCaMKII?
is from Clontech (Mountain View, CA). Untagged rat CaMKII?
and ? were expressed by using Clontech’s pEGFP-C1 vector by
replacing the coding region of EGFP with respective cDNAs.
Deletion mutants were constructed by PCR-mediated method and
expressed by using pCMV-myc expression vector (Clontech). For
CaMKII? and ? RNAi constructs, the following sequences were
chosen as target sequence: for CaMKII?; 5?-CCACTACCT-
TATCTTCGAT-3?; for CaMKII?; 5?-GAGTATGCAGCTAA-
GATCA-3?. The shRNAs were expressed by using a plasmid-based
expression vector, pSuper, and their effectiveness and specificity
were confirmed by cotransfecting them with GFP-tagged CaMKII
subunits in Cos7 cells (12) and immunoblotting with anti-GFP
antibody (Fig. 3a). To rescue the phenotype observed with RNAi
experiments, CaMKII? cDNA was made resistant to RNAi by
incorporating five mismatched silent mutations at the shRNA
target region (indicated as CaMKII?R). The resistance of
CaMKII?RcDNA to RNAi was confirmed by Western blotting.
Actin-Bundling Assay. Non-muscle actin (Cytoskeleton) was poly-
merized in F-actin buffer containing 10 mM Tris?HCl, 0.2 mM
DTT, 0.2 mM CaCl2, 2 mM MgCl2, 50 mM KCl, and 0.5 mM ATP,
pH 7.5. The purified CaMKII [for additional information, see
supporting information (SI) Methods], precentrifuged to remove
After incubation at room temperature (homomeric CaMKII? or
was centrifuged at 10,000 ? g for 10 min to sediment bundled
F-actin but not linear, unbundled F-actin (21). The pellet and
supernatant were subjected to SDS/PAGE and the gel was stained
with Coomassie Brilliant Blue (CBB). To estimate binding affinity,
structure. (Center) When CaMKII is activated by neuronal activity and the resultant Ca2?influx, it detaches from F-actin and allows its reorganization by other
plays a role as a signal transduction molecule. Such mechanism may set a new level for the amount of CaMKII at dendritic spines. (Right) Upon returning to the
unphosphorylated state, CaMKII bundles the newly reorganized F-actin and maintains the remodeled spine structure. The new level of CaMKII may affect the
amount of bundled F-actin in the spines, and consequently the structure of the dendritic spines.
www.pnas.org?cgi?doi?10.1073?pnas.0701656104 Okamoto et al.