The requirement of a relatively high level of Ca2+ for LTP induction is usually explained by the low affinity of CaM to CaMKII, whose activation is required for LTP expression. In this study, we explored the hypothesis that CaM localization within the dendritic spine regulates LTP induction. First, we demonstrated that CaM exhibits a non-random distribution in the spine where a significantly greater fraction of CaM is localized close to the plasma membrane than would be predicted from a random distribution. We also found that the lateral distribution of CaM resembles that of CaMKII. Interestingly, increasing Ng expression in CA1 hippocampal neurons concentrates CaM closer to the extrasynaptic plasma membrane and results in lowering the threshold of LTP induction. These results reveal a novel mechanism by which CaM can control plasticity within dendritic spines.
CaM is a regulatory protein that modulates the activity of many signaling molecules in the cell. Intriguingly, some of these targets have apparent opposing roles, e.g. Ca2+
/CaM-dependent kinase and Ca2+
/CaM-dependent phosphatase. Therefore, it has been postulated that CaM can regulate its targets through its compartmentalization in the cell 
. For example, hormonal stimulation translocates CaM to the nucleus and enhances its activation of nuclear targets 
. This model of redistribution between compartments, however, cannot explain CaM-mediated regulation of opposing targets within a small compartment such as the dendritic spine. Here, we propose that CaM ultrastructural localization within dendritic spines can be regulated by Ng and that this localization influences synaptic plasticity.
The binding of Ca2+
enhances the affinity of CaM to most of its target proteins. Ng, however, is an exception because its affinity for CaM decreases with increased Ca2+
. This unique property of Ng has led to two opposing views on its function. One view postulates that Ng sequesters CaM and constrains Ca2+
/CaM-regulated signaling 
. The other view suggests that Ng concentrates and targets CaM within the spine to facilitate LTP induction 
. In the present study, we provide direct evidence that increasing Ng expression in CA1 hippocampal neurons causes CaM to translocate closer to the plasma membrane. We have previously shown that a significant fraction of Ng (31.3% of immunolabeling of endogenous Ng labeling) is present at the extrasynaptic plasma membrane next to the PSD, suggesting that Ng itself is targeted to the plasma membrane within dendritic spines 
. Since the only known function of Ng is to bind CaM, overexpression of Ng within the spine causes more CaM to shift to where Ng is localized and as a result, CaM is translocated closer to the plasma membrane. In this study, we also show that Ng overexpression, which results in CaM targeting within the spine, is sufficient to lower LTP induction threshold. These results also provide an explanation of our previous study where overexpression of CaM, unlike Ng, was not able to enhance synaptic strength. Together, these results highlight the physiological relevance of CaM distribution within the spine.
There are quite a few CaM-regulated enzymes that are of interest for synaptic plasticity, e.g. adenylyl cyclases, calcineurin and protein kinases. CaMKII, however, has been of special interest because of its unique features. For example, CaMKII can act as a molecular switch, i.e. once activated by Ca2+
-CaM, its activity can persist through its autophosphorylation even after the return of Ca2+
signal to baseline 
. CaMKII is also required 
and sufficient 
for LTP induction. Excellent reviews for the roles and mechanisms of CaMKII in LTP are available 
. Interestingly, there is a close correlation between Ng level and CaMKII activity 
. Nonetheless, while Ng knockout mice show decreased CaMKII activity compared to wild type 
, overexpression of Ng shows increased CaMKII activity in the synaptosomal fraction 
. Our findings that CaM and CaMKII have a similar lateral, but not vertical, distribution within dendritic spines highlight the importance of their ultrastructural localization.
Analysis of CaM distribution within the spine reveals that there is a relatively high fraction of CaM at the PSD. This observation supports previous biochemical findings that identified CaM as a major component of the PSD 
. In contrast to CaM, only 5.7% of Ng immunolabeling was found at the PSD 
, suggesting that a large fraction of CaM at the PSD is not associated with Ng, and thus is unlikely to be involved in Ng-CaM-CaMKII signaling pathway. Such finding is not surprising given the plethora of targets that bind to CaM at the PSD 
. Thus, the fact that a fraction of the intraspine CaM is specifically translocated closer to the membrane, when Ng is overexpressed, highlights the relevance of this particular CaM pool in LTP induction.
Our data support a model in which Ng targets CaM closer to the plasma membrane within dendritic spines, where CaMKII is also more concentrated, and supports LTP through the preferential activation of CaMKII. CaM targeting within the spine is apparently essential for efficient CaMKII and LTP induction. For example, Ng knockout mice have impaired CaMKII activation and LTP induction 
. Even acute knockdown of Ng resulted in failure of LTP induction 
. Moreover, a computational study has supported a need for high CaM concentration within the spine to be able to activate CaMKII by short Ca2+
. Accordingly, it is likely that a lack of Ng may lead to an inadequate level of CaM being targeted close to the plasma membrane, which subsequently would fail to fully activate CaMKII.
This model predicts that CaM targeting by Ng may regulate learning and memory. This notion is supported by Ng knockout studies, which showed that Ng knockout mice exhibit spatial memory deficits 
. Interestingly, studies with mice that are heterozygotes for Ng showed that there is a correlation between Ng levels and memory performance, where lower Ng levels correlated with lower memory performance 
. Importantly, there are a number of neurological disorders, which are characterized by memory deficits, that are accompanied with decreased Ng levels. For example, Ng immunoreactivity was dramatically reduced in areas of prefrontal cortex in postmortem schizophrenia brain tissues 
. As a result, a number of studies have investigated the potential involvement of Ng in schizophrenia 
. A significant role of Ng in schizophrenia is strongly supported by a genome-wide scan of thousands of schizophrenia and control cases that identified Ng as one of four major variants associated with the disease 
. Similarly, chromosomal microarray mapping suggests a role of Ng in cognitive deficits in Jacobsen syndrome 
. Alzheimer’s disease is also accompanied by decreased Ng levels 
and the Ng mRNA fails to be delivered to neuronal dendrites in Alzheimer’s disease, suggesting that Ng protein level is decreased locally at synapses in this disease 
. An interesting question that arises from the current study is whether CaM targeting is disrupted in the before-mentioned diseases as well as in the Ng knockout mice. Our model predicts that CaM will not be enriched close to the plasma membrane in dendrtitic spines of affected neurons. Another interesting question that also warrants further testing is whether increasing Ng (and hence CaM targeting) will be sufficient to enhance learning and memory and/or capable of enhancing memory performance in animal models of the before-mentioned conditions.
An additional important question remains to be answered is how Ng itself is targeted to the plasma membrane. While, through its IQ motif, Ng can bind to CaM and target it closer to the plasma membrane, it is not well understood how Ng concentrates at the plasma membrane. In vitro
study showed that Ng can bind to phosphatidic acid (PA) 
. Thus, it is possible that through this interaction, Ng is targeted to the plasma membrane. Another potential mechanism by which Ng may bind and accumulate at the plasma membrane is through possible palmitoylation of two cysteine residues at its N-terminus. Interestingly, growth-associated protein 43 (GAP-43), another neural-specific CaM-binding protein expressed mainly presynaptically, is tethered to presynaptic membrane through two palmitoylated cysteine residues 
. Finally, it is possible that Ng binds to an unidentified protein that targets it specifically at the spine plasma membrane. Future studies are needed to explore these possibilities.
In conclusion, our results provide insight into the significant physiological role of CaM targeting within the spine and introduces a novel mechanism by which LTP threshold is regulated.