Synaptogenesis in the CNS: An Odyssey from Wiring Together to Firing Together

Abstract

To acquire a better comprehension of nervous system function, it is imperative to understand how synapses are assembled during development and subsequently altered throughout life. Despite recent advances in the fields of neurodevelopment and synaptic plasticity, relatively little is known about the mechanisms that guide synapse formation in the central nervous system (CNS). Although many structural components of the synaptic machinery are pre-assembled prior to the arrival of growth cones at the site of their potential targets, innumerable changes, central to the proper wiring of the brain, must subsequently take place through contact-mediated cell-cell communications. Identification of such signalling molecules and a characterization of various events underlying synaptogenesis are pivotal to our understanding of how a brain cell completes its odyssey from ‘wiring together to firing together’. Here we attempt to provide a comprehensive overview that pertains directly to the cellular and molecular mechanisms of selection, formation and refinement of synapses during the development of the CNS in both vertebrates and invertebrates.

Synaptic communications between neurons, which occur through specialized structures termed synapses, are essential for all nervous system functions. The term ‘synapse’ denotes a specialized morphological structure, which resides between two synaptic partners. Physiological communications at a synapse take place between neurons through the release of chemical messenger from one cell (the presynaptic neuron), which diffuses across the synaptic cleft and binds to its appropriate receptor on the target cell membrane (the postsynaptic cell). Neither the anatomical nor physiological attributes of a synapse are hardwired entities. Rather, recent studies have clearly demonstrated that during development and throughout life, synapses are continuously reconfigured both structurally and functionally –a process that is generally referred to as synaptic plasticity, which forms the basis for learning and memory in a normal brain. Neuronal inability to exhibit such plastic changes is a root cause for various neurodegenerative and psychological disorders.

In this review, we provide an overview of recent progress that elucidates cellular and molecular mechanisms underlying target cell selection, synaptogenesis and synapse refinement during CNS development. A detailed account of other aspects of nervous system development, such as cell migration, proliferation, differentiation, survival and axonal pathfinding is beyond the scope of this review and the reader is directed towards some excellent articles on these topics (Francis & Landis, 1999; Hatten, 1999; Lee & Jessell, 1999; Mueller, 1999; Sanes & Lichtman, 1999; Huang & Reichardt, 2001; Lemke, 2001; Rice & Curran, 2001; Sofroniew et al. 2001).

General principles of neurodevelopment and synapse formation

Roger Sperry (1963) proposed that each of the billions of neurons should have its own uniquely identifiable chemical characteristics to be recognized by a specific synaptic partner. Given the enormous number of neurons in the nervous system, and the paucity of potential cell-cell recognition molecules identified to date, it is rather difficult to envisage that as many different recognition molecules would exist as there are neurons. Alternatively spliced proteins from the gene families including cadherins (Shapiro & Colman, 1999; Wu & Maniatis, 1999), protocadherins (Wu & Maniatis, 1999; Kim et al. 2000), neurexins and neuroligins (Missler et al. 1998a,b; Missler & Sudhof, 1998; Cantallops & Cline, 2000) and integrins (Einheber et al. 1996) may facilitate the recognition of synaptic partners, though their numbers do not match the myriad numbers of synaptic connections that exist in the nervous system (Missler & Sudhof, 1998). Instead, it appears that target recognition cues such as cadherins (Redies & Takeichi, 1996) and integrins (Anton et al. 1999; Graus-Porta et al. 2001) direct migrating neurons to specific regions or layers of tissue, rather than specific cells. Thus, it may seem safe to infer that due to the enormity of the vertebrate brain, cell-cell specificity and target selection may occur at the level of groups of (nearly) identical neurons, rather than single cells.

In the human brain, there are an estimated 10¹⁰ neurons, which exhibit surmounting complexity vis-à-vis patterns of synaptic connectivity, whereas invertebrates possess far fewer neurons, with C. elegans being the extreme example, having only 302 neurons. Notwithstanding this disparity in the number of neurons, developmental neurobiologists have reached a consensus that mechanisms leading up to the final wiring of the brain are equally complex in both vertebrates and invertebrates.

Following proliferation, migration and differentiation, a developing neuron reaches its final destination in the nervous system. To establish contacts with its synaptic partners, a neuron must extend axonal and dendritic processes towards targets located significant distances away. Extending axons and dendrites are guided toward their targets by a specialized structure, termed the growth cone, which is located at the tip of a developing neurite (Tessier-Lavigne & Goodman, 1996; Mueller, 1999). A growth cone does not proceed randomly en route to its targets –rather it is guided by a variety of other cells and molecules present in its extracellular milieu (van Vactor & Flanagan, 1999; Yu & Kolodkin, 1999; Brose & Tessier-Lavigne, 2000; Giger & Kolodkin, 2001; Lemke, 2001; Liu & Strittmatter, 2001; McAllister, 2002). In general, an axon's journey towards its synaptic partner is regulated by a series of intermediate targets (Mueller, 1999), and gradients of diffusible chemotrophic factors, which can either be attractive (Kolodkin, 1996; Lemke, 2001) or repulsive in nature (Yu & Kolodkin, 1999; Lemke, 2001). Upon approach to the target innervation site, a presynaptic growth cone slows its advance, makes a physical contact and transforms itself into a rudimentary presynaptic ending (Levitan & Kaczmarek, 2002). Neurotransmitters or other secreted molecules, such as agrin, then diffuse across the potential presynaptic membrane and bind to their appropriate receptor on the target cell. Various molecules diffused from the presynaptic growth cones, in conjunction with interactions between cell adhesion molecules, then induce clustering of postsynaptic receptors and other components of the synaptic machinery (Sanes & Lichtman, 1999, 2001).

Recent studies by Nimchinsky et al. (2002) show that a postsynaptic neuron in the CNS is not merely a passive recipient of instructions from the presynaptic growth cone. Rather its dendritic filopodia often actively tango with a potential partner –a dancing ritual that enables it to engage the presynaptic axon for synapse formation between the cells (Nimchinsky et al. 2002). Thus, from studies published to date, it is apparent that both pre- and postsynaptic neurons play active roles in identifying and attracting their potential synaptic partners through a series of cell-cell interactions and either diffusible or membrane bound molecules.

The vertebrate neuromuscular junction (NMJ) and various CNS preparations have helped significantly in advancing our knowledge regarding mechanisms of target cell selection, specific synapse formation and synaptic refinement (Sanes & Lichtman, 1999, 2001). However, due to the sheer numbers of neurons in the mammalian CNS and the rate at which the synapses are formed (an estimated 10 000 per 15 min), it has been difficult to define fully the precise anatomical, physiological, molecular and genetic mechanisms of synaptic connectivity. Various invertebrate models, such as C. elegans and Drosophila, with a battery of powerful genetic and molecular tools, have proven useful for our understanding of many key elements of the synaptogenic programme. These model organisms are, however, limited in their ability to reveal physiological changes that may occur throughout synaptogenesis. Specifically, because of their smaller sizes, the C. elegans and Drosophila neurons (for instance, a C. elegans neuron is only 1μm in diameter, which is equivalent to the tip of an intracellular electrode) may be considered less favourable by an electrophysiologist whose primary interest resides in physiological changes occurring during early synapse formation. Mollusks, such as Aplysia, Lymnaea, and Helisoma (Haydon, 1988; Glanzman et al. 1989; Haydon & Zoran, 1989; Syed et al. 1990; Martin et al. 1997; Schacher & Wu, 2002; Munno et al. 2003), and annelids such as leeches (Fernandez de Miguel & Drapeau, 1995) offer better solutions to cellular accessibility. Moreover, identified, adult neurons from these invertebrates can also be isolated in cell culture where they recapitulate their patterns of development (regeneration and synapse formation) with remarkable accuracy (Bulloch & Syed, 1992). Together, with genetic and molecular data obtained from vertebrate models and their invertebrate counterparts, some fundamental principles of the synaptogenic programme are beginning to merge and we are now ever closer to solving the mystery of synapse formation.

Is the synaptogenic programme pre-determined?

Although the final programme orchestrating the patterns of connectivity in the CNS may not be pre-determined, various components of the synaptic machinery appear to be ‘ready-made’ prior to contact between potential synaptic partners. These pre-assembled and ‘ready-to-use’ components of both pre- and postsynaptic machinery are dispatched towards the axonal and dendritic processes where they can function as rudimentary synaptic specializations. For instance, utilizing real-time fluorescence imaging of GFP-tagged presynaptic vesicle-associated membrane protein (VAMP; also known as synaptobrevin), Ahmari et al. (2000) found that ready-made ‘packets’ of presynaptic machinery were often discernable prior to contact between synaptic partners and these were subsequently dispatched to future synaptic sites once the physical contacts between the potential targets were established.

Figure 1. Wiring together to firing together.

A, as pre- and postsynaptic growth cones approach each other, transmitter-receptor interactions, via the release of presynaptic vesicles, attract appropriate target growth cones by binding to and stimulating postsynaptic receptors. Various components of the pre- and postsynaptic specializations, including presynaptic packets containing synaptic machinery and channels and postsynapatic proteins such as PSD-95 are mobile prior to contact. B, as the extending growth cones contact, the growth cones smoothe, and asymmetric interactions between membrane-bound molecules such as neurexins/neuroligins, cadherins, and integrins mark the synaptic site and stabilize pre- and postsynaptic scaffolding proteins such as CASK and PSD-95. C, subsequent maturation of the synaptic contact and interactions between trophic ligands and receptors leads to clustering of calcium channels and synaptic vesicles at the presynaptic terminal and transmitter receptors at the postsynaptic bouton.

The presynaptic growth cones of developing neurons in both vertebrates (Gao & van den Pol, 2000) and invertebrates (Spencer et al. 1998, 2000; Yao et al. 2000) are endowed with transmitter secretory capabilities. Likewise, extensive evidence indicates that prior to innervation, the postsynaptic neurons possess functional transmitter receptors that are responsive to exogenously applied neurotransmitters (Haydon et al. 1985; Lankford et al. 1988; McCobb et al. 1988; Zheng et al. 1994, 1996; Spencer et al. 2000). If both pre-and postsynaptic components of the synaptic machinery are ‘ready, set and go’ (Haydon & Drapeau, 1995) for synapse formation prior to contact between potential synaptic partners, then one would predict that synaptic transmission should be feasible immediately after contact between the neurons. Consistent with this notion, a number of studies have shown that synaptic transmission is indeed possible within seconds to minutes (Haydon & Drapeau, 1995) of contact between the growth cones and their potential synaptic targets (see Lauder, 1993; Spencer et al. 1998). Whether these physiological indicators of synaptic communication are real synapses or a ‘love song’ (nascent synapse) for attracting future synaptic partners remains polemical. This idea is, however, in line with recent work from our laboratory, which showed that dopamine released spontaneously from developing growth cones of Lymnaea neurons can coax the growth cones from its potential target cells at some distance, while repelling the non-target growth cones from hundreds of microns away (Spencer et al. 2000). While the transmitter-receptor interactions were shown to play an important role in defining the initial patterns of target cell selectivity, they did not appear to regulate synaptogenesis between the central neurons (Spencer et al. 2000). It therefore seems plausible that although pre-assembled components of the synaptic machinery may serve as a beacon for attracting synaptic partners or to assist in target cell selection, a lasting synaptic partnership would demand a rigorous scrutiny of the key elements that ensure long-term synaptic compatibility.

In addition to their involvement in assisting growth cones in target cell selection, the transmitter-receptor interactions between neurons have also been shown to regulate the timing of synapse formation between the synaptic partners (Lovell et al. 2002). Specifically, utilizing soma-soma synapses between Lymnaea neurons we have identified a novel form of synaptic competition that regulates the temporal patterns of connectivity between cells that establish mutually inhibitory synaptic connections both in vivo and in vitro. These studies have demonstrated that during early synapse formation, an identified neuron, termed visceral dorsal 4 (VD4) out-competes its counterpart (right pedal dorsal 1 –RPeD1) for synapse formation by suppressing its transmitter-releasing machinery (Lovell et al. 2002). These effects are time dependent, involve transmitter-receptor interactions and are mediated through the metabolites of the arachidonic acid pathway. Once the synapse between VD4 and RPeD1 is fully matured, VD4 then allows RPeD1 to regain its transmitter release properties, thus enabling it to establish a reciprocal inhibitory synapse. VD4-induced effects on RPeD1′s secretory machinery are mimicked by its transmitter. This study provides a unique example of how synapse formation between reciprocally connected neurons may be regulated. Furthermore, these data underscore the importance of transmitter-receptor interactions during early synapse formation in defining the hierarchical patterning of synaptic connectivity whereby the synaptogenic programme of one cell takes precedence over its counterparts.

Although a number of studies allude towards a possible involvement of transmitter-receptor interactions in defining at least some aspects of target cell recognition and synapse formation (Lauder 1993; Spencer et al. 1998), other studies have shown that synapse formation in the CNS can proceed normally in the absence of transmitter-release machinery. For instance, in munc18-1 knockout mice that lack spontaneous and induced transmitter release capabilities, both neuronal projections and connectivity appear normal (Verhage et al. 2000). This suggests that alternative cell-cell or ligand-receptor interactions, such as those mediated by Netrin/DCC and Slit/Robo (Lemke, 2001), may guide the assembly of synaptic circuits. Although transmitter-receptor interactions do not appear necessary for an initial assembly of neuronal circuits in munc18-1-deficient animals, the neurons in this knockout mouse undergo accelerated apoptosis and neurodegeneration. This suggests, at the least, that transmitter-receptor interactions between newly formed synapses are important for neuronal viability, survival and perhaps also for final patterns of connectivity.

Similarly, in Drosophila mutants, in which pan-neuronal expression of tetanus toxin light chain (which cleaves the synaptic vesicle protein synaptobrevin and blocks evoked transmitter release) was induced in specific neurons, normal morphological profiles of synaptic connectivity were observed in the nervous system (Baines et al. 2001). However, several electrical properties, including the expression of voltage-gated K⁺ and Na⁺ channels, conductance, excitability and action potential parameters were perturbed in these mutant flies. Taken together, these studies demonstrate that the nervous system may exhibit morphologically normal patterns of connectivity in the absence of initial transmitter-receptor interactions. However, the neuronal survival, viability, intrinsic membrane, synaptic and connectivity patterns may be compromised when transmitter-release machinery is altered during early development of the nervous system.

Cell-cell interactions induce specific structural and functional changes in their respective targets

The protrusion of postsynaptic dendritic spine is perhaps the first physical sign of early synapse formation during development (Okabe et al. 2001a). Similarly, during long-term plasticity (LTP), which also involves the formation of new synapses, albeit in the mature systems, dendritic spines appear specifically at sites exhibiting activity dependent LTP (Engert & Bonhoeffer, 1999). Lipophilic dyes, such as DiI, and variants of the green fluorescent protein either expressed alone or tagged to postsynaptic density protein 95 (PSD-95) and its analogous proteins, have helped to visualize the appearance of dendritic spines concomitant with clusters of PSD-95 in rat hippocampal cultures (Okabe et al. 2001a,b). PSD-95 appears at the synaptic contact site within 1-3 h after spine formation and is dynamically regulated, in terms of its entry and exit from dendritic spines. Retrospective analysis of these cultures revealed that PSD-95 clusters were associated with transmitter receptors such as the AMPA receptor subunits GluR1. The analogous postsynaptic density protein for metabotropic glutamate synapses PSD-ZIP45 (Okabe et al. 2001b) was also shown to have similar dynamics of appearance in dendritic spines following contact with axonal targets. The use of fluorescently labelled PSD proteins is proving to be a powerful technique for identifying proteins that accumulate immediately after contact between pre- and postsynaptic partners. On the postsynaptic side of the synapse, PSD-95 is essential for localizing proteins required for neurotransmission. PSD-95 may also play a more active role by directly recruiting neurotransmitter receptors and ion channels to the intercellular junctions formed by neurexin-neuroligin interactions (Irie et al. 1997; Nguyen & Südhof, 1997; Butz et al. 1998). Intracellularly, neuroligins bind to a PDZ domain (an acronym for the proteins PSD-95/SAP90, Drosophila Discs-large, and the epithelial tight junction protein ZO-1) of PSD-95 to form a complex between these two proteins (Irie et al. 1997; Song et al. 1999; for review see Sheng & Sala, 2001). This interaction effectively marks the site of the postsynaptic terminal, thus providing a beacon to attract all of the other postsynaptic proteins.

PSD-95 also localizes to newly expressed receptors and is essential for synapse formation. Molecules such as stargazin enable receptors that do not directly bind to PSD-95 to be delivered to synapses. Following delivery of these receptors to the synapse, additional interactions with NSF (N-ethylmaleimide-sensitive fusion protein), and other PDZ proteins such as GRIP/ABP (glutamate receptor-interacting protein/AMPA receptor binding protein) and PICK-1 (protein interacting with C-kinase 1) stabilize the glutamate receptor subunits at the synapse and enable their retention at synaptic sites (Braithwaite et al. 2002; Chetkovich et al. 2002a,b; Schnell et al. 2002). Further retrospective analysis of newly formed PSD-95 clusters (and analogous protein clusters) using immunocytochemistry will yield a precise timeline for the assembly of the postsynaptic specialization.

The protrusion of dendritic spines and accumulation of PSD and receptor proteins mark ‘immature’ or nascent synapses and require further cell-cell interactions to make future synaptic transmission compatible with the functional needs of a neuronal network. In the central nervous system, transynaptic interactions between the cell surface proteins neurexins and neuroligins are thought to induce differentiation at both sides of the synapse (Rao et al. 2000). While neuroligin-neurexin models of synapse formation still require critical testing in vivo, recent in vitro experiments suggest that asymmetric interactions between these membrane-associated proteins mark synaptic sites and enable the localization of essential synaptic machinery at the site of contact between target neurons.

Neurexins are presynaptic transmembrane proteins whose extracellular domains bind postsynaptic neuroligins to induce transynaptic signalling cascades. Intracellularly, neuroligin-induced signalling via neurexins is transduced through ‘supramolecular complexes’ (Sheng & Sala, 2001) with other presynaptic proteins. The cytoplasmic tail of neurexins binds to the PDZ domains of membrane-associated guanylate kinase proteins (MAGUKs). The primary MAGUK protein with which neurexins bind is CASK (calcium/calmodulin-dependent serine protein kinase), a homologue of the LIN-2 protein produced by the C. elegans lin-2 gene (Butz et al. 1998; Sheng & Sala, 2001).

CASK is involved in multiple downstream signalling cascades and it forms complexes with proteins that are assembled during synapse formation, and is essential for synaptic transmission. Butz et al. (1998) identified CASK as part of an evolutionarily conserved, tripartite protein complex including Mint1 and Velis, which are homologues of the C. elegans LIN-7 protein. Mint1 (Munc18-interacting protein 1), binds to the vesicle trafficking protein Munc18-1 (Okamoto & Südhof, 1997). Munc18 (the product of a rat homologue of the C. elegans unc-18 gene), in turn strongly binds syntaxin, which is part of the presynaptic SNARE protein complex required for synaptic vesicle exocytosis (Bennett et al. 1992, 1993; Sollner et al. 1993a,b; Richmond et al. 2001; for review see Wu & Bellen, 1997; Jahn, 2000). In addition, CASK binds protein 4.1, which connects to the actin cytoskeleton, (Cohen et al. 1998; Biederer & Südhof, 2001), calmodulin (Hata et al. 1996) and calcium channels (Maximov et al. 1999). Binding of CASK to calcium channels, directly (Maximov et al. 1999) or indirectly via protein 4.1, provides a mechanism by which calcium can enter the synaptic terminal in close vicinity of the synaptic vesicle and release machinery. Interestingly, CASK-dependent activity, either as a transcription regulator or during assembly of the presynaptic specialization, is required immediately following contact between synaptic partners (Spafford et al. 2003). Once the synapse is formed, competitive inhibition of CASK activity –via synthetic peptides targeting the PxxP sequence necessary for CASK binding with interacting proteins –does not acutely affect synaptic transmission. Thus, CASK effectively acts as a beacon for synaptic assembly, by endowing the terminal with vesicular release capabilities essential for chemical neurotransmission.

An important step in understanding localization of synaptic proteins to contact sites has been taken in the invertebrate nervous system at which cell-cell interactions specifically between target and non-target cells can be easily investigated in vitro. Using the Lymnaea soma-soma model of synapse formation, in which synapses form between the cell bodies of contacting synaptic partners (Smit et al. 2001; Munno et al. 2003), we have found that ‘hot spots’ of calcium channels (Feng et al. 2002) and transmitter-release machinery (Munno et al. 2003) appear at the presynaptic contact site of synaptic partners that have formed synapses. These Ca²⁺ hotspots are target cell-and contact site-specific (Feng et al. 2002). Interestingly, the total calcium current between single cells and synapsed pairs did not change. However, using two-photon confocal microscopy, we demonstrated that there was a redistribution of calcium signal such that influx was selectively enhanced at the site of synaptic contact (Feng et al. 2002) as compared with non-contacted sites. These studies, coupled with cloning of specific Ca²⁺ channels in Lymnaea, led us to believe that non-L-type calcium channels (Ca_v2) specifically localize at the synapse after hours of cell-cell contact. These channels, as opposed to L-type (Ca_v1) calcium channels, are specifically required for functional synapse formation and synaptic transmission, as perturbation of L-type Ca²⁺ channels during early synapse formation not only rendered the synapse incapable of function but also blocked synaptogenesis between the synaptic partners (Spafford et al. 2003).

It is important to realize that additional as yet unidentified proteins at the pre-postsynaptic interface probably perform similar functions as CASK, PSD-95, and their respective adaptor proteins. Although experimental evidence implicates the role of neurexin-neuroligin binding in transynaptic signalling at cell-cell contacts and establishing beacons of PDZ proteins for recruiting synaptic proteins, ion channel, and receptors, these in vitro studies now need to be confirmed in vivo.

Extrinsic trophic factors and synapse formation

In addition to cell-cell signalling, various extrinsic factors also serve during synapse formation and the best studied among these are the neurotrophic factors. Neurotrophic factors such as those that comprise the neurotrophin family were originally identified as target-derived signals that regulate survival and differentiation of neurons (Levi-Montalcini et al. 1996). Neurotrophins, such as nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF), and others such as glial cell line-derived neurotrophic factor (GDNF) support synaptic and neuronal survival and maturation (Tomac et al. 1995a,b; Fagan et al. 1996; Choi-Lundberg et al. 1997; Martinez et al. 1998; Brennan et al. 1999; Kaplan & Miller, 2000; Erickson et al. 2001; Huang & Reichardt, 2001). The neurotrophins are expressed widely in the CNS and may act as both anterograde (Kohara et al. 2001; Heerssen & Segal, 2002; Spalding et al. 2002; Wahle et al. 2003) and retrograde neurotrophic signalling molecules (Watson et al. 1999; Miller & Kaplan, 2001; Ginty & Segal, 2002). Recent evidence suggests that neurotrophins can be released by presynaptic neurons to stimulate survival of postsynaptic neurons (for review see, Heerssen & Segal, 2002) and influence synapse formation and synaptic activity.

In addition to their conventional roles in cell survival, differentiation, proliferation and neurite outgrowth a variety of trophic factors have now been shown to alter neuronal excitability and these effects range from ion channel modulation to synaptic plasticity (Poo, 2001). More recently, we have demonstrated the requirement of extrinsic trophic factors for excitatory (Hamakawa et al. 1999; Woodin et al. 1999) but not inhibitory synapse formation (Feng et al. 1997). Specifically, we have shown that excitatory synapses between Lymnaea neurons, in a soma-soma configuration, fail to develop in the absence of brain conditioned medium (CM)-derived trophic factors. The trophic factor-induced excitatory synapse formation requires gene transcription and de novo protein synthesis and these effects are mediated through receptor tyrosine kinases. Interestingly, in the absence of trophic factors, neurons that form excitatory synapses in vivo, establish inappropriate inhibitory synapses, which we have referred to as ‘default’ synapses (Woodin et al. 1999). These inappropriate inhibitory synapses, which do not normally exist in vivo, can be corrected by the addition of trophic factors to the culture dish (Woodin et al. 1999, 2002). Utilizing synapses between the presynaptic somata and the postsynaptic neurites severed from their cell bodies, we have demonstrated that the extrinsic trophic factors may either assist in stabilizing and/or localizing excitatory (cholinergic) postsynaptic receptors at the synaptic sites (Meems et al. 2003), which in turn provides the basis for excitatory synapse formation. These studies uncover a novel role for trophic factors, which spans beyond development and may form the basis for synaptic plasticity underlying learning and memory.

Genes involved in synaptogenesis

In contrast with the cellular and molecular makeup of synaptic architecture, little is known about the genetic machinery that orchestrates synaptogenesis in the nervous system. A recent flurry of activity in this area of developmental neuroscience has resulted in the identification and characterization of various synapse-specific genes and their products at the NMJ. Specifically, mutation of a gene termed late bloomer, which is involved in synapse formation in Drosophila, delayed, but did not prevent synapse formation (Kopczynski et al. 1996). In addition, in the highwire (hiw) Drosophila mutant, elaborate synapses are formed, both in numbers of synaptic boutons and lengths of synaptic branches (Wan et al. 2000). Although these exuberant synapses have reduced quantal contents, the ultrastructure, axonal pathfinding and synapse formation appeared normal. In the C. elegans homologue of hiw, rpm-1, loss of function mutations produce abnormalities of presynaptic terminals at the GABAergic NMJ (Zhen et al. 2000). Specifically, rpm-1 is believed to regulate either the spatial arrangement of synapses or restrict the formation of presynaptic structures to localized areas. The involvement of rpm-1 in the regulation of synaptic growth was also demonstrated independently by Schaefer et al. (2000). Finally, a gene termed futsch has been shown to regulate synaptic organization at the Drosophila NMJ (Roos et al. 2000). Futsch mutations were shown to reduce synaptic microtubular organization, and reduce bouton numbers, while increasing their size.

Although perturbations of all of the above genes resulted in aberrant morphological organization of synapses, they did not appear to block synapse formation completely. Moreover, in contrast with the identification of synapse-specific genes at the NMJ, very little is known about genes that regulate synapse formation in the CNS. A notable exception is the work of van Kesteren et al. (2001), who identified a human homologue of the MEN1 tumour suppressor gene in Lymnaea, which codes for the trascription factor menin. Menin was shown subsequently to be a critical mediator of synapse formation between soma-soma paired Lymnaea neurons (van Kesteren et al. 2001). This study showed that the MEN1 gene was upregulated during synapse formation between identified neurons in cell culture and antisense knockdown of MEN1 mRNA blocked the formation of functionally mature synaptic connections. Moreover, using immunocytochemistry and cell-specific antisense knockdown of MEN1 mRNA, we demonstrated that postsynaptic, but not presynaptic, expression was required for synapse formation. This study now opens the possibility that various synapse-specific genes can indeed be identified, characterized and manipulated in the CNS.

While invertebrate systems such as Lymnaea have been well suited for studying synaptic physiology during synapse formation and have yielded important insights into the genetic determinants of synapse formation, other model systems promise more detailed analysis of the genetic and molecular determinants of synapse formation. The completion of the Drosophila, C. elegans, and mouse genomes will enable the use of large-scale DNA microarray techniques to study the dynamic changes in gene expression during developmental periods of neurogenesis, differentiation and synapse formation. In the mouse hippocampus, more than 1900 genes were shown to be dynamically regulated during critical periods of embryonic and postnatal development when neurons are born and begin to establish functional synapses with target neurons (Mody et al. 2001). Among the many genes that were found to be differentially regulated during this period of nervous system development were those coding for synaptic vesicle proteins (VAMP2 and synaptophysin), neurotrophic factors (BDNF), transmitter receptors (glutamate receptor subunits GluR1, GluR2, NR1, and muscarinic acetylcholine receptor subunit M1), signal transductions enzymes and proteins (calcineurin B, ras, RAB-3A), and transcription and translation regulators (DNA binding protein SMBP2, transcription factor Sox-M). Not surprisingly, there was also a gradual increase in genes necessary for glucose metabolism as the brain's energy requirements shift from ketone in the neonate to glucose in adult animals.

The use of microarrays to screen for genes that are dynamically regulated during nervous system development and synapse formation complements the growing understanding of genetic mechanisms of synapse formation from mouse, fly and worm knockouts that are fatal due to the lack of synaptic transmission or display malfunctions in synapse formation and synaptic function (Brenner, 1974; Aravamudan et al. 1999; Verhage et al. 2000; Godenschwege et al. 2002; Ho et al. 2003; Shin et al. 2003). There is no substitute, however, for precisely manipulating the expression of particular genes in a single cell to deduce the impact of such perturbations on synapse formation. As genetic information and tools are becoming more prevalent in model systems such as Lymnaea, it has been possible to use antisense and RNA interference techniques to knock down particular genes thought to be involved in synapse formation (van Kesteren et al. 2001; Korneev et al. 2002; Spafford et al. 2003). This approach may yield further information about the site of action (pre- vs.postsynaptic neuron) and hierarchy of genes that control synapse formation. Continued use of genetic screens at the level of single-cell analysis will provide a comprehensive picture of how various genes interplay temporally to regulate synapse formation.

Summary and future perspective

From studies summarized in this review, it appears that the fundamental principles governing synapse formation in both vertebrates and invertebrates are probably conserved. Even if the key players of the synaptogenic programme turn out to be system specific, studies on simple animal models will pave the way for our understanding of how basic elements of synapse formation –which in most mammalian preparations remain elusive –are configured to generate synapse specificity. Some of the confounding elements that deter our ability to identify both synapse-specific molecules and the underlying mechanisms include: complexity, redundancy and identification of nascent synapses. Nothing can be done to resolve the issue of complexity and we will have to continue to exercise reductionism or to adopt novel in vitro cell culture approaches to get around this problem. Regarding redundancy, it is often difficult to evaluate the outcome of experiments where specific genes or their products are perturbed to determine their involvement in synapse formation. This is because the neurons involved compensate immediately for the loss of function, sparing the physiology of a synapse from significant disruption. The most reasonable conclusion from this observation is that the events of synapse formation are not controlled by one molecule or one gene, rather dynamic interactions between various genes and their encoded proteins must occur throughout development to generate synapse specificity. The genes and molecules involved in the proper wiring of the nervous system are symbiotic to the extent that the loss of function of one component is not only sensed by its counterparts, but that the remaining genes and molecules march forward to compensate for the functional deficit. Thus, the compensatory traits of the brain evoke admiration from developmental neurobiologists, and at the same time present daunting challenges for data interpretation for those who hone in on any give molecule or gene of interest.

In terms of identification of nascent synapses, there is currently a lack of consensus between anatomical and physiological definitions of these predecessors to mature synapses. Ready-made elements of synaptic architecture in the absence of physiological activity are often identified by anatomists as ‘nascent’ synapses. In contrast, the initial events of chemical and electrical communication between neurons are hallmarks of a physiologist's version of ‘nascent synapse’, even if the cells lack the necessary structural components. Efforts should therefore be made to first confirm both the morphological and electrophysiological attributes of a synapse before designating it as a ‘synapse’. This would most certainly require a multidisciplinary approach that should take advantage of both morphological and electrophysiological techniques (Ahmari & Smith, 2002).

Understanding the mechanisms of synapse formation is not only important for our basic knowledge of the development of the nervous system but also for the developing techniques to facilitate the regeneration of the nervous system in the face of neurodegenerative diseases and injuries. This will be particularly important for cell replacement therapies, most notably using stem cell-derived neurons. While neurogenesis from stem cells can produce neurons that are capable of extending processes and contacting potential synaptic partners, what good would this ‘wiring’ bring if the newly born cells failed to ‘fire’ together?