It had been widely accepted that synapses are stable structures once they are formed in the period of early postnatal development. To see if newly formed synapses are quickly stabilized, we measured the lifetime of individual synapses by using GFP-tagged scaffolding proteins localized in the postsynaptic density (PSD). Time-lapse imaging of GFP-labeled PSDs revealed the elimination of a predominant proportion of newly formed synapses even in the early period of monotonous synapse increase. Within one day, 10-20% of total synapses were eliminated, and a roughly equal number of synapses were newly generated to counterbalance the elimination. Detailed analysis of the balance between synapse elimination and formation indicated a slight bias toward synapse formation, which underlies a gradual increase in the total number of synapses. The high rate of synapse elimination is thought to underlie the mechanism of flexible adjustment of neuronal connection during postnatal development.
When new synapses are formed, both presynaptic release machinery and postsynaptic receptor should be recruited. Molecular assembly and structural change of single synapses were visualized by using synaptophysin-CFP as a presynaptic marker and PSD-95-YFP as a postsynaptic marker. This imaging analysis revealed the rapid formation of synaptic molecular specialization in the time scale of 30 min to several hours. Global analysis of synapse formation indicated a gradual increase of synapse density, which led to the contention that the formation and maturation of single synapses are also slow and progressive. However, the visualization of individual synapses clearly illustrated the rapid establishment of single synapses.

To obtain an integrated view on both the synaptic structure and dynamics of synaptic molecules, we should build a quantitative model of molecular assembly in both the PSD and the presynaptic active zone. Essential information for creating such a model is the absolute numbers of molecules localized within single synapses. To reveal the number of specific PSD scaffolding proteins per synapse, we developed a new quantitative technique of fluorescence microscopy. We first quantitated fluorescence intensities of single GFP molecules and estimated the number of GFP-tagged synaptic molecules within individual synaptic sites. Next, we performed quantitative immunocytochemistry and estimated the absolute numbers of endogenous PSD scaffolding proteins localized within single synaptic sites. This approach enabled us to obtain the absolute synaptic contents of multiple PSD scaffolding proteins, which are in the range of 100 to 500 molecules per synapse. These numbers indicate a relatively dense distribution of PSD scaffolding proteins in the PSD structure and a high probability of interaction with glutamate receptors.
By using a technique of fluorescence recovery after photobleaching (FRAP), turnover rates of PSD scaffolding molecules were estimated at the level of single synapses. Time constants of turnover were in a range of several minutes to several tens of minutes, indicating a faster assembly-disassembly rate of scaffolding molecules compared with the lifetime of the PSDs themselves. These data suggest that PSD scaffolding molecules are densely packed and have multiple interactions, but in the process of frequent exchange with the soluble cytoplasmic pool, which enables continual changes in the size and structure of the PSDs.

The technology of single synapse visualization can be applied to the analysis of the interaction between synapses and surrounding non-synaptic components. Major structural elements present in the vicinity of synapses are processes of glial cells. Previous studies have shown that diffusible factors released from astroglia facilitate the formation and maturation of synapses. It has not yet been clarified if there is any local regulation on the development and maturation of single synapses by the direct contact of astroglial processes to synapses. In hippocampal slices, we observed contact events between neurons and astrocytes, which were labeled by different fluorescent probes. Selective stabilization and maturation of spines after astrocytic contacts were demonstrated. Single astrocytes contact with thousands of synapses within their domains. Therefore astrocytes have the potential to either regulate individual synapses locally or influence thousands of synapses within their domains at once.

Excitatory synapses between pyramidal neurons in the hippocampus and the neocortex have been the default model of synapse formation in the CNS. However, the principles of synapse formation identified in one synapse subtype may not be applied to other synapse subtypes. A clear example was found in excitatory synapses on dendrites of cortical interneurons. The cortical interneurons are inhibitory neurons and use GABA as a primary neurotransmitter. We found retrograde transport of synaptic structure along dendritic protrusions in immature interneurons. Translocating synapses reach the main dendritic shafts and stop their movement. Synapse translocation was driven by dynein motor proteins, which move along the rails of microtubules. Synapse mobility based on microtubule-dependent motor molecules is a novel finding and may contribute to the clarification of cytoskeletal contribution in synapse formation.
The second example of unique strategies taken by specific neurons in synapse formation was found in the cerebellum. Detailed analyses of postnatal synapse development between cerebellar Purkinje cells and granule cell axons (parallel fibers) led to the discovery of small axonal protrusions at the contact sites with Purkinje cell dendrites. These axonal protrusions of granule cells wrap dendritic spines of Purkinje cells and facilitate synapse maturation. The formation of axonal protrusion was mediated by a signaling pathway triggered by a soluble factor cbln1. This finding revealed sequential interplay between dendrites and axons plays an essential role in synapse maturation.
Even in the case of excitatory synapses in pyramidal neurons, which have been studied extensively, new principles and new regulatory molecules could be discovered. Microtubule-associated protein DCLK1 shows the robust activity of facilitating microtubule assembly and stabilization. Interestingly, DCLK1 is preferentially localized at the tips of growing dendrites. This unique localization of DCLK1 helps the facilitation of microtubule assembly at the tips of dendrites. On the other hand, DCLK1 shows suppressive roles in spine growth, assembly of PSD molecules, and postsynaptic receptor functions. Namely, DCLK1 controls dendritic growth positively and synapse maturation negatively at the tips of growing dendrites. The dual roles of DCLK1 may be critical in local control of dendritic development.
After the discovery of a new mechanism of synapse maturation based on presynaptically released cbln1, we initiated the screening of other molecular candidates that are released from the axons and regulate synapse formation. We identified BMP4 as a negative regulator of synapse formation in the hippocampal excitatory synapses. BMP4 acts on BMP receptors on the surface of the axonal membrane and activates the signaling cascade that facilitates the deconstruction of the presynaptic structure. BMP4 keeps the synapse density within the physiological range by removing excess and unnecessary synapses in the hippocampal neural circuits.

The technology of synapse imaging is now expanded to the field of in vivo synapse analyses. This expansion was enabled by the introduction of two-photon excitation laser scanning microscopy in the studies of neuronal morphology and function in intact brain tissues. With adequate surgical techniques and cranial window preparations, the same cortical regions can be imaged repetitively, and time-lapse images of dendrites and synapses in vivo can be recorded with time scales of minutes to months. By applying this imaging technology to the analyses of model mice of mental disorders, neural circuit-level dysfunctions in vivo may be identified. Autism spectrum disorders (ASDs) are early-onset mental disorders characterized by deficits in social behaviors, restricted interest, and repetitive behaviors. Recent genetic studies revealed a strong link to genes encoding synaptic cell adhesion molecules and scaffolding proteins, suggesting functional changes of synapses in ASD patients. We analyzed three mouse models of ASDs with distinct genetic backgrounds and identified common phenotypes in synapse dynamics. Both synapse formation and elimination are enhanced in cortical pyramidal neurons during the early postnatal period. If future studies confirm the enhancement of synapse turnover as a core feature of neural circuits in mouse models of ASDs, this finding will significantly contribute to our understanding of the pathophysiology in ASDs.

Synapses and spines are the structures with their sizes in the order of microns or sub-micron. Therefore the internal structure of synapses and spines is thought to be impossible to be reliably quantified by light microscopy. However, recent advancements in super-resolution microscopy provided new tools that can quantitatively analyze the internal structure of spine synapses. We applied the technique of structured illumination microscopy (SIM) in the analyses of nano-scale spine structure. SIM imaging revealed the presence of a concave surface on the head of mature spines in contact with the presynaptic membrane. Thus, the expansion of the concave surface in the spine head contributes to the stabilization of synaptic junctions after induction of spine structural plasticity.
The application of fluorescence correlation techniques can report molecular diffusion within the spine cytoplasm. We found the intra-spine suppression of diffusion for the molecules with their sizes larger than 100 kDa. The diffusional barrier is actin-filament dependent and can be disassembled by induction of spine structural plasticity only for 5 minutes after plasticity-inducing stimulation. The highly temporally restricted window of molecular mobilization is a new concept, which may contribute to define a distinct stage in synaptic plasticity.