Competitive stabilization and elimination of synaptic connections play a key role in the refinement of neuronal networks during development, learning and memory. Cellular mechanisms mediating heterosynaptic competition, however, remain poorly understood.

In the cerebral cortex, excitatory synapses are made on small protrusions of dendrites, dendritic spines, in the pyramidal neurons. Selection of dendritic spines should be achieved by structural plasticity of spines, their enlargement and shrinkage, which are associated with long-term potentiation (LTP) and depression (LTD), respectively. It has been proposed that the modest increases in cytosolic Ca2+ concentrations ([Ca2+]i) selectively activate protein phosphatase PP2B, calcineurin, while large increases in [Ca2+]i activate Ca2+/calmodulin kinase II (CaMKII) to induce LTD and LTP, respectively. It has also been proposed that spine shrinkage is induced by a signaling cascade, involving calcineurin and actin depolymerizing factor, ADF/cofilin. ADF/cofilin is activated upon dephosphorylation by slingshot, and severs and depolymerizes actin fibers. On the other hand, spine enlargement is induced by a signaling cascade, involving activation of CaMKII and Rac1, which causes phosphorylation of PAK, LIMK, and cofilin. Cofilin is deactivated by LIMK-induced phosphorylation. The spatial organizations of the two pathways and their interactions, however, have not been investigated.

Two-photon glutamate uncaging has enabled stimulation of individual spines, and demonstrated the enlargement of dendritic spines underlies LTP at the level of single spines. Although electrical stimulation of presynaptic fibers can induce spine shrinkage and LTD, spine shrinkage has not been induced with stimulation of identified spines, which have hampered the understanding of synaptic competition where spine shrinkage and elimination would play a major role.

In this thesis, I report that spine shrinkage could be readily induced when repetitive glutamate uncaging was applied in the presence of GABAergic inhibition. This could be achieved either by the spike-timing protocol where glutamate uncaging was paired with a spike in the postsynaptic neuron, provided that each spike was associated with uncaging of caged-GABA compound (1 Hz, 80 times), or with the continuous presence of a selective GABAA receptor agonist, muscimol. Importantly, I have found that, although spine enlargement was confined to the stimulated spine, spine shrinkage tended to spread into neighboring spines over 10 μm, even when only one spine was stimulated (Fig. 1).

I next examined whether this spreading shrinkage induced heterosynaptic competition between enlargement and shrinkage of spines along a dendrite. For this purpose, two neighboring spines were challenged with repetitive pairing with a spike and glutamate uncaging, one with LTP and the other with LTD timing. I found that spreading spine shrinkage was induced using the LTD protocol except for the spine which was specifically stimulated with the LTP protocol, where spine enlargement was induced. Thus, spine enlargement could outcompete the spreading shrinkage, while shrinkage spread beyond the enlarged spines. Moreover, I found that the bidirectional structural plasticity was mediated by the competition between phosphorylation and dephosphorylation of cofilin (Fig. 2).

Since cofilin regulates both directions of the plasticity depending on its phosphorylation, I hypothesized cofilin is a pivotal molecule for synaptic competition. I therefore investigated the diffusive properties of cofilin using photoactivatable GFP (PAGFP) fused with rat cofilin1 or S3A mutant of PAGFP-cofilin, which mimics dephosphorylated active cofilin. In the resting spine head, the most PAGFP-cofilin or PAGFP-cofilin (S3A) molecules gradually diffused out from the photoactivated spines and readily spread along dendrites over 10 μm within a few minute (Fig. 3). Thus, active cofilin readily diffuses along a dendrite. Moreover, to test whether cofilin could induce spine shrinkage, I applied cofilin protein via whole-cell pipette into the soma of pyramidal neurons. Cofilin perfusion induced shrinkage and reduction of PSD95 of many spines in the dendrite within 30 min (Fig. 4). These experiments indicate that cofilin can diffuse effectively along dendrites, and induces spreading spine shrinkage.

Finally, I examined the diffusive properties of wild type (WT), S3E or 3SA mutants of cofilin during spine enlargement. The S3E mutant of cofilin is known to mimic phosphorylated cofilin. The expression of cofilin and its mutants did not significantly affect the spine enlargement induced by repetitive glutamate uncaging in a Mg2+-free solution. I found that WT and S3E mutant of PAGFP-cofilin were accumulated in the stimulated spines over 30 min during the spine enlargement, similarly as the stable enlargement pool of F-actin. In contrast, S3A mutant of PAGFP-cofilin rapidly diffused out from the enlarged spine, as it did in the resting spines (Fig. 5). These data indicate that LTP protocol induced selective generation and accumulation of phosphorylated cofilin, while dephosphorylated cofilin was excluded from the stable F-actin. Thus, the diffusion of cofilin was dependent on the type of stimulation in the way accounting for the synaptic competition at the level of single spine.

In summary, I have established the conditions to reliably induce shrinkage of identified spines, and found that it tended to spread due to the diffusion of cofilin, and gave rise to local competition of spine synapses. Moreover, the spine shrinkage was greatly promoted by the activation of GABAA receptors (Fig. 6). My study thus revealed the highly interactive nature of synapses along dendrites.

Figure 1. Spine shrinkage induced by two-color uncaging of glutamate and GABA.

Spine shrinkage was reliably induced when GABA uncaging was applied at the onset of spike at the dendritic shaft close to the stimulated spine. Spine shrinkage tended to spread into neighboring spines over 10 μm and was dependent on Ca2+ entry through NMDA receptors.

Figure 2. Competition of enlargement and shrinkage of spines with the spike-timing protocol.

Spine enlargement was confined to the stimulated spine even when spine shrinkage spread to the nearby spines. Competitive structural plasticity reflected the competitive phophorylation of cofilin, because the phosphorylated cofilin peptide (p-cofilin peptid) abolished spine shrinkage, and it hastened the enlargement, while the dephosphorylated cofilin peptide blocked spine enlargement. Cofilin peptide and p-cofilin peptide inhibits phosphorylation and dephosphorylation of cofilin, respectively.

Figure 3. Spread of cofilin along dendrites imaged with photoactivation of PAGFP-cofilin.

In the resting spine, PAGFP-cofilin (S3A) molecules gradually diffused out from the photoactivated spines and readily spread along dendrites over 10 pm within a few minute.

Figure 4. Spine shrinkage induced by whole-cell perfusion of cofilin protein.

Cofilin protein was applied via whole-cell pipette into the soma of pyramidal neurons. Cofilin perfusion induced shrinkage and reduction of PSD95 of many spines in the dendrite, while heat inactivated (HI) cofilin did not.

Figure 5. Accumulation of PAGFP-cofilin in the enlarged spines.

WT and S3E mutant of PAGFP-cofilin were accumulated in the stimulated spines over 3D min during the spine enlargement. In contrast, S3A mutant of PAGFP-cofilin rapidly diffused out from the enlarged spine, as it did in the resting spines. S3E and S3A mutant of cofilin is known to mimic phosphorylated and dephosphorylated cofilin, respectively.

Figure 6. Current hypothesis of the local competition of spine enlargement and shrinkage along a dendrite.

本研究は記憶や学習などの高次脳機能において重要な役割を担っていると考えられる競合的シナプス可塑性の機構を明らかにするため、ラット海馬CA1錐体細胞においてケイジドグルタミン酸とケイジドGABAの2色光刺激法を用いて、スパインの収縮及び局所競合の解析を試みたものであり、下記の結果を得ている。

1.ラット海馬スライス培養系において、ケイジドグルタミン酸とケイジドGABAの2色光刺激法を用いることで、単一スパインレベルでのスパイン収縮誘発法を確立した。また、ケイジドGABA刺激は、GABAA受容体の作動薬であるmuscimolによって代替出来ることが示された。この時、スパイン増大は刺激スパインに限局するのに対し、スパイン収縮は周囲のスパインに拡散することが示された。また、スパイン収縮はグルタミン酸感受性の減少(長期抑圧)とともに起きていることが示された。

2.長期増強刺激と長期抑圧刺激を隣り合ったスパインに対して行ったところ、長期抑圧刺激を行ったスパイン及びその周囲のスパインは収縮するのに対し、長期増強刺激を行ったスパインは増大することが示された。つまり、スパイン増大はスパイン収縮と競合することが示された。この時、p-cofilin peptideを用いてcofilinの脱リン酸化を抑制すると、スパイン収縮が阻害された。反対に、cofilin peptideを用いてcofilinのリン酸化を抑制すると、スパイン増大が阻害された。このことより、スパインの増大・収縮は、cofilinのリン酸化・脱リン酸化によって制御されていることが示された。

3.Photoactivatable GFPを結合したcofilin1タンパク質を発現させて、cofilinタンパク質の細胞内動態の解析を行った。その結果、cofilinは数分以内に光活性化を行ったスパイン内から周囲のスパインに拡散していくことが示された。また、パッチクランプ電極からヒトcofilin1タンパク質を細胞体に注入することで、スパインの収縮及びPSD95の減少が起こることが示された。これらの結果より、cofilinは拡散性のスパイン収縮因子であることが示された。

4.Photoactivatable GFP-cofilin発現細胞に対して、photoactivatable GFPの光活性化と同時にケイジドグルタミン酸による長期増強刺激を行い、スパイン増大時におけるcofilin動態の観察を試みた。その際、野生型のcofilinの他に、脱リン酸化型を模倣した変異体であるcofilin (S3A)とリン酸化型を模倣した変異体であるcofilin (S3E)のそれぞれを用いて、リン酸化状態におけるcofilin動態の差異を解析した。その結果、スパイン増大時にはリン酸化cofilinが生成され、増大したスパインに集積することが示された。

以上、本論文はラット海馬CA1錐体細胞において、単一スパインレベルでのスパイン収縮誘発法の確立に成功し、スパイン収縮及び増大機構の解析から、樹状突起局所での競合的シナプス可塑性機構の存在を明らかにした。本研究はこれまで未知であった、脳内における競合的なシナプスの除去及び維持の機構の解明に重要な貢献をなすと考えられ、学位の授与に値するものと考えられる。