If we accept the hypothesis that information storage within the brain occurs at synaptic contacts, the neuronal network of synapses requires two contradictory but essential properties: stability and plasticity. Stability of synaptic contacts is certainly a general tendency that increases with age and experience and that is needed to leave a trace or “memory” of previous activity. In contrast, plasticity, that represents the capacity of the system to allow modifications of its structure and function and that is required to “learn” seems to occur principally at two occasions: during
development and sustain activity. While not exclusive but certainly profoundly interconnected this two circumstances are treated separately within the following chapters
Development
Formation of synapses starts during embryogenesis but continues during the early postnatal life when dendritic spines begin to appear. Although formation of spines and synapses are relatively independent of activity levels during development, further refinement of connections through spine pruning requires sensory activity.
Synaptogenesis
During development, synaptogenesis begins with the extension of axons and dendrites that
accompanies neuronal differentiation. However it seems that the specificity of the synaptogenesis is regulated spatially and temporally to retain connection between inappropriate cells.
Synaptogenesis is a multistep process. First molecules like fibroblast growth factor (FGF) or cholesterol are secreted by target neurons and their surrounding glia to promote axonal and
dendritic maturation (Scheiffele, 2003) (Ullian et al., 2004). Then the initial contact between axones and dendrites is stabilized by cell adhesion molecules like Cadherins (Takai et al., 2003). Finally, cell adhesion molecules like SynCAM (Biederer et al., 2002), Neuroligin-Neurexin (Scheiffele et al., 2000) and EphB/EphrinB induce the formation of the pre and postsynaptic zones (Dalva et al., 2000).
Presynaptic assembly starts shortly after the clustering of cell adhesion molecules like SynCam at the plasma membrane and depends on the delivery of presynaptic elements by pleomorphic vesicles to nascent axonal boutons (Ziv and Garner, 2004).
Postsynaptic assembly seems to occur by gradual accumulation of molecules rather than by delivery of an integer number of vesicles as for presynaptic zone. One of the earliest event is the recruitment of PSD-95 to the postsynaptic side that occurs within minutes after axo-dendritic contact in cultures (Bresler et al., 2001). Then follows the independent insertion of NMDARs through their interaction with different PDZ domains. Local protein synthesis may also contribute to synapse assembly particularly for proteins like CaMKII and Shank which mRNA is abundant in dendrites. Finally, considering the heterogenous postsynaptic composition and the possibility of its regulation by
Introduction
activity-dependent mechanisms, we may expect that each component of the PSD is at least partially independently inserted (Garner et al., 2006).
Synaptic maturation implies morphological and functional changes at both synaptic sides that grow in a coordinated fashion. Morphologically, long and thin spines bearing a small head shorten and widen as they stabilize. Functionally, the probability of transmitter release decreases (Bolshakov and Siegelbaum, 1995) but the reserve of pool vesicles increases at the presynaptic side. There are also postsynaptic changes in glutamate receptor composition: replacement of NR2B subunits by NR2A that decreases the duration of NMDARs currents (Bellone and Nicoll, 2007) and addition of AMPARs to initial silent synapses (Isaac et al., 1999).
Although activity modulates synaptogenesis from postnatal periods to adult stages, its function is probably less crucial during development (Hua and Smith, 2004). Indeed, synapses with normal morphology can be formed in mice lacking neurotransmitter release (Verhage et al., 2000),
suggesting that activity is presumably more involved in synapse stabilization and elimination than in synapse formation during this period.
Spinogenesis
In most regions of the developing brain, formation of dendritic spines coincides with the main period of synaptogenesis.
Density
An initial overproduction of dendritic spines during fetal and early postnatal life is followed by a later decrease of their density in different cortical areas (Zecevic and Rakic, 1991) (Bourgeois and Rakic, 1993). Spinogenesis starts in the middle of the first postnatal week, short after the beginning of the synaptogenesis in rodent CA1 (Fiala et al., 1998). During the next four weeks spine density increase to finally decrease and stabilize in adulthood (Nunzi et al., 1987). Similarly, to
developmental synaptogenesis, the early postnatal burst of spinogenesis doesn’t seem depend on extrinsic activity. However, sensory inputs and NMDA activity are crucial for the following refinement of contacts that operates trough a reduction in spine number (Zuo et al., 2005b).
Morphology
Considering asymmetric synapses in the hippocampal CA1 region, the approximate proportion of filopodia and shaft synapses that are 20% and 60% at P1 are divided by two at P12. Inversely, spine synapses increase from 20% at P1 to 60% at P12. Taken together, shaft and filopodia synapses predominate at early stages but are progressively replaced by spine synapses later in development (Fiala et al., 1998). At adulthood, spines predominate and comprise mainly thin and mushroom types. Nearly all spines have synapses in mature brain and naked spines are reported only occasionally (Arellano et al., 2007).
Origin
There is no consensus for the mechanism of spinogenesis, but three models summary the different views. In the first one, filopodia catch axons terminals and further transform into spines. In the
Introduction! Plasticity conditions
second one, axons terminals induces spine formation after contacting the dendritic shaft. In the third one, spine emergence is independent from synaptic contact with an axon terminal.
From filopodia to spines:
In this model, filopodia probe space around dendrites to find an appropriate presynaptic partner.
The release of glutamate by axons could be a mechanism to trigger filopodia extension and to guide them to presynaptic terminals (Portera-Cailliau et al., 2003). During early stages of synaptogenesis, dendrites are covered by many highly motile filopodia (Dunaevsky et al., 1999) (Lendvai et al., 2000) that can express PSD-95 and contact axon terminals (Marrs et al., 2001). With development and as synaptic density increases, more stable spines progressively replace these transient filopodia.
As suggested by the chronology of events and the collocation of structures, filopodia could play the role of spine precursors. However this function is controversial as many cells have filopodia during development but do never develop spines (Wong et al., 1992). Indeed, without continuous images showing the transition from filopodia to spines in a developing network the proper demonstration is difficult.
From shaft synapse to spine:
Spines are proposed to arise from shaft synapses because majority of excitatory contacts on pyramidal neurons occur on shaft rather than on filopodia early in development (Fiala et al., 1998) and because emergence of mature spines from shaft synapses is reported (Marrs et al., 2001).
Moreover, the number of spine synapses increases as shaft synapses decrease with network maturation (Harris et al., 1992) (Miller and Peters, 1981). In this model, axonal filopodia are involved in prospection for postsynaptic partners and establish synaptic contacts with dendritic shafts (Fiala et al., 1998). We would expect axons to have convoluted trajectories at least during development if this model was generally applied. In contrast, axonal trajectories are straight and contacts are mainly en passant (Yuste and Bonhoeffer, 2004).
Directly spines:
Temporally, synaptogenesis precedes spinogenesis in hippocampus and neocortex, and
mechanistically, the loss of synaptic proteins correlates with the disappearance of mature spines (Okabe et al., 1999). However, presynaptic terminals are not absolutely required for spinogenesis during development (Sotelo, 1990). Spinogenesis can also precede synaptogenesis in late
developing slices after activity-induced plasticity (Nagerl et al., 2007) and even in anesthetized adult mice (Knott et al., 2006). At this age, new spines often start by contacting multisynapse boutons (MSBs) before showing ultrastructural characteristic of a synapse after hours to days depending on the exact developmental stage and experimental model.
Giving the high prevalence of filopodia and shaft synapses early in development, some authors also imagined a model in which filopodia and shaft synapses share the role of spine precursor. Filopodia would first establish a synaptic contact before retracting and pulling the synapse to the dendritic shaft from where a spine would rise (Fiala et al., 1998) (Marrs et al., 2001).
In summary, live imaging experiments that document how spines emerge during early development are missing from the literature. Different models certainly coexist in different types of neurons or even in the same cell at different developmental stages, reflecting the diversity of molecular
Introduction! Plasticity conditions
mechanisms. For example, when synaptic contact is required before spine formation35 cell adhesion molecules could trigger the genesis of dendritic spines, but when spine formation is independent of previous contact a different process should occur.
Activity
Even after development, activity can still influence brain connectivity. Within the network
plasticity and stability of synapses are finely balanced to allow the “learning” of new constraints but also the “memory” of changes. Traditionally, these changes are considered to be mainly functional, modifying transmission at individual synapses. More recently, it seems that structural modification alter not only spine morphology but also spine growth and elimination.