Cadherins in neuronal morphogenesis and function




Classic cadherins represent a family of calcium-dependent homophilic cell–cell adhesion molecules. They confer strong adhesiveness to animal cells when they are anchored to the actin cytoskeleton via their cytoplasmic binding partners, catenins. The cadherin/catenin adhesion system plays key roles in the morphogenesis and function of the vertebrate and invertebrate nervous systems. In early vertebrate development, cadherins are involved in multiple events of brain morphogenesis including the formation and maintenance of the neuroepithelium, neurite extension and migration of neuronal cells. In the invertebrate nervous system, classic cadherin-mediated cell–cell interaction plays important roles in wiring among neurons. For synaptogenesis, the cadherin/catenin system not only stabilizes cell–cell contacts at excitatory synapses but also assembles synaptic molecules at synaptic sites. Furthermore, this system is involved in synaptic plasticity. Recent studies on the role of individual cadherin subtypes at synapses indicate that individual cadherin subtypes play their own unique role to regulate synaptic activities.


Cell–cell adhesion is essential for multicellular organisms. Nearly a quarter of a century ago, cadherins were identified as calcium-dependent homophilic cell–cell adhesion molecules (Yoshida & Takeichi 1982; Gallin et al. 1983; Peyrieras et al. 1983). Since then, progress in molecular biology resulted in an increase in the number of classic cadherins identified in vertebrates as well as in invertebrates. A substantial number of studies have revealed that cadherin-mediated cell–cell adhesion plays important roles in morphogenesis and tissue integrity of multicellular organisms (Halbleib & Nelson 2006). On the other hand, more than 100 molecules with divergent structures, including desmosomal cadherins, protocadherins, Flamingo/Celsrs, and FAT cadherins, have been placed in the cadherin superfamily. These cadherins engage in various biological events (Hirano et al. 2003; Halbleib & Nelson 2006; Takeichi 2007). Hereafter we refer to classic cadherin as cadherin. Despite the importance of cadherins in morphogenesis and tissue integrity, previously we had little knowledge as to their roles in the nervous system. However, recently accumulating evidence indicates the involvement of cadherins in multiple processes of morphogenesis and function in the nervous system. Here, we review recent advances in our understanding of the roles of cadherins in neural morphogenesis and function in the nervous system. In the first part of this review, we briefly summarize the fundamental properties of cadherins. Next, we present an overview of the involvement of the cadherin-mediated cell adhesion system in neural morphogenesis. Finally, we review recent studies on the roles of cadherins in the morphogenesis and functioning of synapses.

Organization of classic cadherins

Classic cadherins are found in a wide range of multicellular organisms and defined as single-pass transmembrane proteins that possess tandemly repeated extracellular cadherin domains (EC) in their extracellular region, and a highly conserved intracellular part containing the juxtamembrane domain and the catenin-binding domain (Fig. 1A). This juxtamembrane domain contains the GGGEED sequence, by which p120 catenin family member proteins (p120 catenin: Ctnnd1; NPRAP/δ-catenin: Ctnnd2, p0071: Pkp4; and ARVCF: Arvcf) interact with cadherins (Fig. 1B). Although the functions of these proteins remain to be fully understood, these proteins regulate the surface stability of cadherins (Chen et al. 2003; Davis et al. 2003). At the catenin-binding domain, cadherins interact with β-catenin (Ctnnb1). In turn, β-catenin binds to α-catenin and through this catenin complex cadherins are anchored to the actin cytoskeleton (Fig. 1B). The anchoring of cadherins to the actin cytoskelton via these catenins is essential for cadherin-mediated cell–cell adhesion (Hirano et al. 1992; Kintner 1992; Fujimori & Takeichi 1993; Roe et al. 1998).

Figure 1.

Organization of cadherin-mediated cell–cell adhesion and phylogenetic tree of vertebrate cadherins. (A) Comparison of domain organization of type I/II and type III cadherin. Both types of cadherins have tandemly repeated extracellular cadherin (EC) domains and a conserved cytoplasmic domain. Although type I/II cadherins have five EC domains, the number of EC differs among type III cadherins. Moreover, type III cadherins have the primitive classic cadherin domain (PCCD) between their EC and transmembrane domain. (B) Schematic drawing of cadherin/catenin adhesion system. (C) A phylogenetic tree for cadherins based on the extracellular domains of human cadherins.

Classic cadherins are subdivided into two groups by organization of their extracellular region: type I/II cadherins and type III cadherins (Tanabe et al. 2004). Type I/II cadherins, which are only found in vertebrate and ascidians, have five EC. In contrast, the number of EC varies among type III cadherins; for example, Drosophila E-cadherin (shg) and N-cadherin (CadN) have six and 15 EC, respectively (Oda et al. 1994; Iwai et al. 1997), and chicken Hz-cadherin has 13 of them (Tanabe et al. 2004). Furthermore, these type III molecules have a unique domain called the primitive classic cadherin domain (PCCD) situated between their EC and transmembrane domain (Fig. 1A; Oda & Tsukita 1999). Interestingly, type III cadherins exist in invertebrates and restricted species of vertebrates such as fishes and birds, but not in mammals. This distribution may indicate that type III cadherins are prototype cadherins that have been eliminated during mammalian evolution (Tanabe et al. 2004).

One of the hallmarks of classic cadherin-mediated cell adhesion is its binding specificity: the same or similar cadherins bind together. In particular, the most distal EC (EC1) of type I/II cadherins is known to be responsible for the binding specificity of this type of cadherins (Nose et al. 1990; Patel et al. 2006). The type I/II cadherin group is further subdivided into two groups according to their amino acid sequences: type I and type II cadherins (Suzuki et al. 1991; Tanihara et al. 1994; Nollet et al. 2000). E-, N-, P-, and R-cadherin (Cdh1, 2, 3 and 4, respectively) belong to the type I cadherins. These cadherins have a conserved HAV sequence in their most distal EC (EC1). Cadherins designated numerically as cadherin-6, -7, and -8 (Cdh6, 7, and 8, respectively) belong to the type II cadherins (Fig. 1C). Basically, each subtype of type I cadherins binds just homophillically to the same type. However, N-cadherin and R-cadherin can also interact heterophilically. But, the strength of the heterophilic interaction between them is weaker than that of their homophilic interactions (Inuzuka et al. 1991). In contrast, interactions among type II cadherins are more complicated. In addition to homophilic interactions, heterophilic interactions are frequently observed among this type of cadherins (Nakagawa & Takeichi 1995; Shimoyama et al. 2000).

Cadherins in neuronal morphogenesis in vertebrates

The vertebrate central nervous system is derived from a sheet of neuroepithelial cells. Among the classic cadherins, N-cadherin (Cdh2, hereafter called N-cad) first appears and accumulates in the adherens junctions of the apical portion of neuroepithelial cells in chick and mouse embryos (Hatta & Takeichi 1986; Hatta et al. 1987). N-cad maintains the integrity of these cells during development of the brain. Inhibition of N-cad by a blocking antibody or genetic ablation results in disruption of the apical junctions in the neuroepithelium (Matsunaga et al. 1988b; Ganzler-Odenthal & Redies 1998; Luo et al. 2001; Pujic & Malicki 2001; Kadowaki et al. 2007). N-cad in the neuroepithelium is also important for rearrangement of neuroepithelial cells. In the zebrafish embryo, neuroectodermal cells first converge at the midline; and then the cavitation of the neural tube follows. In a zebrafish N-cad mutant, this convergence is impaired (Lele et al. 2002).

Early in vitro studies revealed that N-cad promotes neurite extension when neurons are cultured on non-neuronal cells expressing N-cad or on a substrate coated with the extracellular domain of N-cad (Matsunaga et al. 1988a; Bixby & Zhang 1990). However, we have little knowledge about the roles of N-cad in vivo, because of disruption of the neuroepithelium and the death of N-cad mutant animals at an early developmental stage. We only know that neurite outgrowth of amacrine cells and axonal projection of retinal ganglion cells are affected in zebrafish N-cad weak-allele mutants (Masai et al. 2003). Results consistent with these were obtained from studies in which the cadherin/catenin adhesion system is perturbed ubiquitously by overexpression of a dominant-negative cadherin. This type of cadherin molecule lacks most of the extracellular domain and has been used as a tool to elucidate roles of the cadherin/catenin adhesion system. Overexpression of this molecule can disrupt cadherin-mediated cell–cell adhesion by competing with endogenous cadherins for interaction with the actin cytoskeleton through catenins. Blocking of the cadherin/catenin adhesion system in retinal cells with dominant-negative cadherins impairs elongation of the axons or dendrites of these cells (Riehl et al. 1996; Tanabe et al. 2006).

Another way to ubiquitously perturb cadherin/catenin function is genetic ablation of catenins. α-Catenin functions as a linker between the cadherin/β-catenin complex and the actin cytoskeleton, and was shown to be essential for strong cell–cell adhesion mediated by the cadherin/catenin system in epithelial cells (Hirano et al. 1992). There are three subtypes of α-catenin differentially expressed in the mouse: αE-catenin (Ctnna1) in epithelial tissues, αN-catenin (Ctnna2) in the nervous system, and αT-catenin (Ctnna3) in restricted tissues (e.g., heart, testis; Janssens et al. 2001). Genetic ablation of αN-catenin causes morphological defects in restricted brain regions (Park et al. 2002a; Togashi et al. 2002; Uemura & Takeichi 2006). In αN-catenin mutants, the anterior commissure, which is an axonal tract interconnecting the left and right halves of the olfactory region, does not cross the midline (Fig. 2A,B); and the arrangement of neurons in the hippocampus and inferior olivary nucleus is also slightly disorganized (Uemura & Takeichi 2006). Interestingly, the localization of a subset of Purkinje cells is affected in the αN-catenin mutant cerebellum (Cook et al. 1997; Park et al. 2002a; Togashi et al. 2002). Normally, Purkinje cells migrate out from the ventricular zone of the fourth ventricle to just below the surface of the cerebellar cortex and form a single cell layer called the Purkinje cell layer in the mature cerebellum (Fig. 2C). In the cerebellum of αN-catenin mutants, however, about 40% of the Purkinje cells aberrantly stop their migration in the inner part of the cerebellum (Fig. 2D; Beierbach et al. 2001). Analysis of chimeric mice made of αN-catenin mutant and wild-type embryos revealed that αN-catenin intrinsically functions in a subpopulation of Purkinje cells when they migrate to their appropriate location (Park et al. 2002b).

Figure 2.

Morphological impairments of αN-catenin mutant mice. (A, B) Schematic drawings of the defect in the anterior commissure of an αN-catenin mutant mouse. Horizontal sections of the forebrain are depicted. Red lines in both drawings represent the anterior commissure. In the wild-type brain, the anterior commissure runs in front of the third ventricle (V) and interconnects the left and right sides of the olfactory region (dotted line). By contrast, in the αN-mutant, the anterior commissure does not pass the midline. (C, D) Migration of Purkinje cells is affected in αN-mutant mice. Parasagittal sections of cerebellum from wild-type (C) and αN-catenin mutant (D) mice are stained for a Purkinje cell marker, calbindin-D28K. (C) In the wild-type cerebellum, Purkinje cells have settled at the distal part of the cerebellum. (D) In the αN-mutant cerebellum, many Purkinje cells are aberrantly located in the inner part of the cerebellum.

In spite of the ubiquitous expression of αN-catenin in the brain, why are only restricted parts of the brain affected by the loss of αN-catenin? Intriguingly, the loss of αN-catenin seems to affect the migration of genetically defined Purkinje cells. Purkinje cells are known to be subdivided by the expression of various kinds of genes including those for type II cadherins, transcription factor, enzymes, and so on (Oberdick et al. 1998). One such gene, zebrin II/aldolase C, is expressed by a subset of Purkinje cells in such a way that parasagittal stripes of zebrain II/aldolase C-expressing Purkinje cells and those of zebrin II/aldolase C-negative Purkinje cells alternate in the normal cerebellar cortex. Loss of αN-catenin affects mainly the zebrin II/aldolase C-negative Purkinje cells (Beierbach et al. 2001).

Cadherins in invertebrate nervous system

Drosophila has three type III classic cadherins (D E-cadherin: shg; DN-cadherin: CadN, and DN-cadherin2: CadN2). Among them, CadN is the major neuronal classic cadherin in the fly. CadN is widely expressed in the Drosophila nervous system; and in CadN-null mutants, the amount of the neuronal armadillo protein Drosophilaβ-catenin is severely reduced (Iwai et al. 1997). Loss of CadN causes misrouting of growth cones and defasciculation of axon bundles in the embryonic nervous system, and results in embryonic lethality (Iwai et al. 1997). A hypomorphic mutant of CadN that survives to the adult stage has disorganized brain structures and exhibits uncoordinated locomotion, showing that CadN is essential for interneuronal connections (Iwai et al. 2002). Furthermore, roles of CadN in interneuronal connections have been uncovered in detail by sophisticated Drosophila genetic analyses focusing on the visual and olfactory system.

The compound eye of Drosophila comprises approximately 750 ommatidia, each of which has eight photoreceptor neurons (R1–R8). Axons of R1–R8 in the same ommatidium form a common bundle and project to the lamina and medulla. The lamina is an assembly of a columnar unit called the cartridge. Each cartridge contains five lamina neurons and axons of photoreceptors. At the lamina, axons of R1–R6 defasciculate from the common bundle, extend laterally to six neighboring cartridges in a stereotype manner, and make synapses with lamina neurons there (Fig. 3A). In contrast, axons of R7 and R8 run through the lamina and innervate to the medulla at two separate layers (Ting & Lee 2007). Removal of CadN in R1–R6 perturbs defasciculation of their axons from the common bundles (Fig. 3B; Lee et al. 2001; Prakash et al. 2005). In the absence of CadN in lamina neurons, physical interaction between these neurons and axons of R1–R6 in the cartridge is disorganized. Furthermore, analysis using a specific marker for R4 revealed that the stereotype innervation of R4 axons to their target cartridges is perturbed when mutation of CadN is induced in lamina neurons in their target cartridge (Prakash et al. 2005). Loss of CadN also affects projection of R7 to the medulla (Lee et al., 2001; Prakash et al. 2005; Ting et al. 2005). Axons of R7 follow R8 axons pioneering the medulla and terminate in a deeper layer of the medulla past R8 axons (Ting & Lee 2007). CadN-mediated adhesive interactions between R7 axons and their target medulla neurons are essential for the layer-specific targeting of R7 in the medulla (Yonekura et al. 2007). In summary, in the formation of the fly visual system, CadN functions to stabilize cell–cell interaction between photoreceptor axons and their target neurons.

Figure 3.

Effects of CadN mutation in the Drosophila nervous system. (A, B) Defects of axonal targeting of CadN-deficient photoreceptors in the lamina. (A) Association of an array of lamina neurons and an axon bundle of R1–R8 in the same ommatidia is depicted with a cross section of the lamina. In the lamina, cartridges, each of which is composed of five lamina neurons and photoreceptor axons, are arranged in a regular fashion. Axons of R1–R6 defasciculate from a common bundle consisting of R1–R8 axons and extend laterally to neighboring cartridges in a stereotype manner. (B) CadN-deficient R1–R6 axons do not make stable contacts with their target lamina neurons. (C–E) Defects caused by mutation of CadN in the olfactory system. (C) Olfactory receptor neurons (ORN) with the same olfactory receptors converge their axons to a single glomerulus. Each olfactory projection neuron (PN) also targets its dendrites to a single glomerulus, where they make synapses with ORN axons. (D) Although coarse projection patterns are preserved in the absence of CadN in PN, CadN-deficient PN project their dendrites to multiple glomeruli. A similar phenotype is observed with a wild-type PN on a CadN-deficient PN background. (E) Axons of CadN-deficient ORN still project to their appropriate glomerular regions but are segregated from the dendrites of their target PN.

In the fly olfactory system, about 1500 olfactory receptor neurons (ORN) are present in the antenna and the maxillary palps and innervate the antennal lobe of the brain (Laissue et al. 1999). In the antennal lobe, there are about 50 discrete spherical structures called glomeruli, which contain axons of ORN and dendrites of second-order olfactory projection neurons PN and local interneurons (Stocker 1994; Stocker et al. 1990). Axons of ORN expressing the same olfactory receptors as well as dendrites of individual PN converge to one of ∼50 glomeruli (Fig. 3C; Gao et al. 2000; Vosshall et al. 2000; Jefferis et al. 2001). Developmentally, dendrites of many PN project into the antennal lobe and form prototypic glomerular structures prior to the arrival of ORN axons. In turn, axons of ORN project to one of the prototypic glomerular structures, penetrate into it, and make synapses there (Jefferis et al. 2004). Removal of CadN in PN does not affect the coarse projection pattern of PN dendrites, but results in innervation of multiple glomeruli by PN dendrites. Similar defects are observed in the dendrite projection of a wild-type PN that is induced in an otherwise CadN-deficient PN environment (Fig. 3D; Zhu & Luo 2004). Taken together, the data indicate that CadN-mediated dendrite–dendrite interaction is essential for PN to restrict their dendrites to a single glomerulus. On the other hand, CadN is essential for ORN to make connection with their target PN. CadN-deficient ORN project their axons to appropriate glomerular regions but do not intermingle with dendrites of their target PN (Fig. 3E; Hummel & Zipursky 2004).

Caenorhabditis elegans has only one classic cadherin gene, termed hmr-1. The hmr-1 gene has two alternative promoters and two alternatively spliced transcripts: hmr-1a and hmr-1b. Whereas hmr-1a has two EC in its extracellular domain and is expressed in epithelial cells (Costa et al. 1998), hmr-1b has 15 EC and is expressed in neuronal cells (Broadbent & Pettitt 2002). Because hmr-1b exhibits its highest similarity to CadN among the known cadherins, hmr-1b is thought to be an orthologue of Drosophila CadN. This view is supported by defects caused by loss of hmr-1b; for example, removal of hmr-1b impairs fasciculation of axons, as seen in Drosophila CadN mutants (Broadbent & Pettitt 2002).

Functions of the cadherin/catenin system at synapses

The synapse is a cell–cell junction specialized for signal transmission via neurotransmitters from the presynaptic neurons to the postsynaptic ones. At the ultrastructural level, mature synapses are compartmentalized into two zones, that is, the transmitter release zone where signal transmission via synaptic vesicles occurs and a symmetrical junctional zone called the puncta adherentia, which looks like the adherens junction formed between epithelial cells. Cadherin/catenin adhesion complexes are localized at synapses (Yamagata et al. 1995; Fannon & Colman 1996; Uchida et al. 1996; Elste & Benson 2006). In particular, these complexes are confined to the puncta adherentia in mature synapses (Uchida et al. 1996).

The roles of cadherin/catenin adhesion complexes in synapses have been extensively studied in the context of formation of excitatory synapses by dissociated hippocampal neurons (reviewed in Salinas & Price 2005; Takeichi & Abe 2005). Most excitatory synapses are formed on the tips of mushroom-shaped protrusions from dendrites called dendritic spines in the mature brain (Harris & Kater 1994). In the early stage of spine synapse formation, cell–cell contacts are formed between axons and fine motile extensions of dendrites called dendritic filopodia. As the hippocampal culture matures, these dendritic filopodia transform into the mushroom-shaped dendritic spines. During this transformation, the length of the protrusions is shortened, the head region becomes enlarged, and motility of the protrusion is reduced (Fig. 4A,B). Even in mature neurons, this process still occurs, being associated with neural plasticity. Cadherin/catenin complexes accumulate at contacts between axons and dendritic filopodia soon after they meet (Togashi et al. 2002). Inhibition of cadherin-mediated cell–cell adhesion by overexpressed dominant-negative cadherins or genetic ablation of αN-catenin or β-catenin during synaptogenesis prevents the morphological change from dendritic filopodia to dendritic spine and destabilizes cell–cell contacts between axons and dendrites (Fig. 4C; Togashi et al. 2002; Abe et al. 2004; Bozdagi et al. 2004; Okamura et al. 2004; Okuda et al. 2007). In contrast, when the amount of cadherin/catenin complex in neurons is increased by overexpression of αN-catenin during synaptogenesis, the synaptic contacts become more stabilized: spine heads are larger, spine length is shorter, spines are less motile, and the synaptic density is increased (Fig. 4D; Abe et al. 2004). Interestingly, inhibition of cadherin/catenin complex formation in mature neurons has a lesser effect on the morphology of spines and stability of cell–cell contacts at synapses (Togashi et al. 2002). These observations probably indicate that cell adhesion molecules other than cadherin also contribute to cell–cell adhesion at mature synapses.

Figure 4.

Effects of either loss- or gain-of-function of cadherin/catenin adhesion system on spine morphology. (A, B) Morphological changes in spine synapse during maturation. (A) Dendritic filopodia are observed along an immature dendrite. (B) Spine synapses have formed between a mature axon and dendrite. (C) Effects of loss-of-function of cadherin/catenin adhesion system on spine synapse morphology. Overexpressing a dominant-negative cadherin or genetic ablation of αN-catenin or β-catenin results in unstable cell–cell contacts between axons and dendrites. (D) Overexpression of αN-catenin overstabilizes cell–cell contacts in the spine synapse.

The recruitment of various synaptic molecules to their appropriate location in synapses is essential for building functional synapses. The cadherin/catenin adhesion complex plays an important role by acting as a scaffold for synaptic molecules. The cytoplasmic tails of cadherins interact with various synaptic molecules directly or indirectly through catenins (reviewed in Takeichi and Abe 2005). For example, β-catenin has a PDZ (PSD-95/Discs large/ZO-1)-binding motif at its C-terminus, where a PDZ-domain-containing presynaptic protein, lin-7, and a postsynaptic protein, S-SCAM (Magi2), are known to interact (Dobrosotskaya & James 2000; Kawajiri, Itoh & Fukata 2000; Perego et al. 2000). δ-Catenin /NPRAP (neural plakophilin-related arm protein), which binds to the juxtamembrane domain of cadherins, also interacts with PDZ-domain-containing proteins including ABP (AMPA receptor binding protein: Grip2), GRIP (glutamate receptor interacting protein: Grip1), and PSD-95 (postsynaptic density-95: Dlg4) at the PDZ-binding motif at the C-terminus. In turn these PDZ-domain-containing proteins interact with glutamate receptors (Silverman et al. 2007). Blocking of the cadherin/catenin system in neurons by a dominant-negative cadherin or genetic ablation of β-catenin hampers the appropriate accumulation of synaptic molecules at synapses in vitro and in vivo (Togashi et al. 2002; Bamji et al. 2003; Tanabe et al. 2006). Consequently, synaptic transmission is impaired under such a condition (Bamji et al. 2003; Bozdagi et al. 2004; Okuda et al. 2007).

Changes in synaptic activity cause remodeling of synapses. Cadherin/catenin complexes are also dynamically regulated by synaptic activity. Enhanced neural activity increases the accumulation of cadherins and catenins at synapses (Murase & Schuman 1999; Tanaka et al. 2000; Abe et al. 2004; Abe & Takeichi 2007). In contrast, blocking of neural activity by tetrodotoxin (TTX) decreases the amount of αN-catenin in synapses (Abe et al. 2004). Recent work suggests the possibility that N-methyl D-aspartic acid (NMDA) receptor-dependent neural activity simultaneously regulates cell–cell adhesion and gene expression through cleavage of β-catenin (Abe & Takeichi 2007). β-catenin is known to be a signal mediator of the canonical Wnt pathway. In the absence of Wnt signals, the N-terminal region of cytoplasmic β-catenin is phosphorylated by GSK3β (Gsk3b). This phosphorylation causes ubiquitin-dependent proteolysis of β-catenin. In the presence of Wnt signals, however, this phosphorylation of β-catenin is suppressed; and the amount of cytoplasmic β-catenin thus increases. Consequently, β-catenin is translocated to the nucleus, where it then activates gene expression by making a complex with Tcf/lef-1 transcription factors (Moon et al. 2004). The activation of NMDA receptors induces cleavage of the N-terminal region of β-catenin by calpain. Because the truncated β-catenin is resistant to ubiquitin-dependent proteolysis primed by GSK3β-dependent phosphorylation, it is translocated to the nucleus and induces there the gene expression of Fosl1, which is known to be an immediate early gene (Abe & Takeichi 2007). In addition, calpain can cleave β-catenin already complexed with cadherins; and the truncated β-catenin is then released from the cadherins. Thus, there is a possibility that neural activity mediated by NMDA receptors may modulate cell–cell adhesion as well as gene expression through truncation of β-catenin by calpain. On the other hand, activation of NMDA receptors is reported to enhance the retention of N-cad on the neuronal surface (Tai et al. 2007).

So far, we have mainly discussed the roles of the cadherin/catenin system that are irrelevant to cadherin subtypes. As described above, nearly 20 subtypes of cadherins exist in a vertebrate species; and most of these cadherins are expressed in the brain. How is each subtype of cadherins involved in synapse formation and/or function?

N-cad is abundantly expressed in the postnatal and mature mouse brains and has been implicated in the targeting of retinal neurons to specific lamina in the chick optic tectum and induction of long-term potentiation (LTP) based on the results of in vitro experiments (Inoue & Sanes 1997; Tang et al. 1998). Nevertheless, because of lethality of the mutation in N-cad at an early developmental stage, we have still only limited knowledge as to the exact roles of N-cad in synapse formation and function in vivo obtained from analyses of N-cad-deficient animals. Results of recent analyses indicate that N-cad is dispensable for the morphogenesis of synapses (Erdmann et al. 2003; Jungling et al. 2006; Kadowaki et al. 2007). The retina of zebrafish N-cad mutants has a severely disorganized laminar structure in which islands of synaptic areas are dispersed in the cellular region. However, at the ultrastructural level, these areas contain synapses with a normal appearance; and there are as many synapses as in the wild-type retina (Erdmann et al. 2003). Consistent with this observation, N-cad-deficient hippocampal neurons in which N-cad is conditionally removed and N-cad-deficient ES cell-derived neurons both formed morphologically normal synapses (Jungling et al. 2006; Kadowaki et al. 2007). Furthermore, β-catenin as well as pre- and postsynaptic molecules were normally localized at synapses on N-cad-deficient hippocampal neurons (Kadowaki et al. 2007). This observation indicates that other classic cadherins still exist at synapses formed on N-cad-deficient neurons. Intriguingly, despite the normal appearance of synapses formed between neurons derived from N-cad-deficient ES cells, synaptic vesicles were soon depleted during prolonged stimulation (Jungling et al. 2006). Taken together, these observations suggest that N-cad facilitates efficient secretion of synaptic vesicles. Furthermore, given that cadherin/catenin complexes still exist at synapses on neurons derived from N-cad-deficient ES cells, this function might not be displaceable by other cadherins.

Type II cadherins have been implicated in selective neural circuit formation as evidenced by the fact that their expression patterns are differentially expressed in a way correlating with specific neural circuits (Suzuki et al. 1997; Inoue et al. 1998). Indeed, perturbing the expression of cadherins of this type affected fasciculation of axonal tracts (Treubert-Zimmermann et al. 2002) and segregation of motor columns (Price et al. 2002). Besides, cadherin-11 (Cdh11) is known to positively regulate the number of synapses, as concluded from the results of knockdown with siRNA and overexpression of this cadherin in cultured hippocampal neurons (Paradis et al. 2007). However, we have little knowledge about the roles of these cadherins in selective neural circuit formation as obtained by analysis of mutant animals for these cadherins. We only know that LTP is enhanced in the hippocampus of cdh11−/− mice and that they show reduced responses in fear-related behavioral tests (Manabe et al. 2000).

Our recent analysis of cadherin-8 (Cdh8) mutant mice provides evidence showing that cdh8 is essential for physiological coupling between cold temperature-activated sensory neurons in dorsal root ganglia (DRG) and their target neurons in the dorsal part of the spinal cord, called the dorsal horn (DH; Suzuki et al. 2007). Temperature-sensitive sensory neurons in the DRG detect changes in the ambient temperature of the body region through peripheral branches and project their central branches to the DH of the spinal cord, where they make synapses with DH neurons. Cdh8 is expressed by a small number of small-sized sensory neurons and a subset of DH neurons. Also, cdh8 protein is localized in a restricted lamina of the DH (Fig. 5). The relationship between these cdh8-expressing neurons in cdh8+/- was examined by immunoelectron microscopy (immuno-EM) utilizing lacZ markers expressed from the targeted cdh8 locus. This analysis revealed that these two groups of lacZ/cdh8-expressing neurons were connected via synapses (Fig. 5A). The same immuno-EM analysis using spinal cord slices of cdh8−/− revealed that synapses between these lacZ/cdh8-expressing neurons were still made in the absence of cdh8. Recently, ion channels belonging to the transient receptor potential (TRP) cation channel family were found to function as thermosensors in temperature-sensitive sensory neurons (Tominaga & Caterina 2004). Expression of cdh8 in DRG resembled that of the cold and menthol receptor TRPM8: most of the cdh8-expressing sensory neurons expressed TRPM8 and vice versa. Using menthol as a tool to activate axonal terminals of TRPM8-expressing sensory neurons, we performed whole-cell patch clamp recordings from lacZ/cdh8-expressing DH neurons in spinal cord slices from cdh8+/− and −/− mice. Menthol increased the frequency of inputs in lacZ/cdh8-expressing DH neurons in cdh8+/− slices but not in such neurons of cdh8−/− slices. In support of these results, cdh8 mutant mice were less sensitive to a cold temperature. Taken together, these results indicate that cdh8 is dispensable for making physical connection between lacZ/cdh8-expressing sensory neurons and DH neurons, but is indispensable for the physiological function of the synapse between them. As in the case of synapse formation between N-cad-deficient hippocampal neurons, the loss of cdh8 did not affect the localization of β- and αN-catenin at the synapses where cdh8 was originally localized. Also, we observed that a single sensory neuron expressed multiple cadherins. So, other cadherins probably mediated cell–cell adhesion at the originally cdh8-localized synapses in cdh8−/− mice. Impairment of physiological coupling between TRPM8-expressing sensory neurons and their target lacZ/cdh8-expressing DH neurons in the presence of other cadherins at the synapses suggests that the synaptic activity-regulating function of cdh8 is unique to cdh8 (Fig. 5B).

Figure 5.

Involvement of cdh8 in activities of synapses between TRPM8-expressing sensory neurons and their target neurons in the DH of the spinal cord. (A) Schematic drawing of connection between cdh8-expressing sensory neurons in DRG and neurons in the DH of the spinal cord. (B) A possible explanation of the phenotype of cdh8−/− in TRPM8-expressing sensory neurons. Cdh8 as well as other cadherins (cdhX) mediate cell–cell adhesion at synaptic contacts between TRPM8-expressing sensory neurons and their target neurons in cdh8+/−. Cdh8 may recruit unknown synaptic molecules (X, Y) to synaptic contacts that regulate synaptic activities (left). In cdh8−/−, cell–cell contact between TRPM8-expressing sensory neurons and their target neurons is maintained by cadherins other than cdh8. However these cadherins do not recruit cdh8-interacting synaptic molecules. Consequently, synaptic transmission between these neurons is impaired (right).

Conclusion and perspectives

Here we have discussed roles of the cadherin/catenin adhesion system in neural development and functions. This system is involved in multiple steps of morphogenesis of neuronal cells including neuroepithelial formation, neurite extension and migration of neurons, and synaptic function. These studies raise further questions. The phenotypes of αN-catenin mutant mice clearly show that the cadherin/catenin system is involved in axonal projection and migration of neuronal cells (Park et al. 2002a; Togashi et al. 2002; Uemura & Takeichi 2006). Despite ubiquitous expression of αN-catenin in the nervous system, abnormal phenotypes are observed only in restricted brain regions of αN-catenin mutants. What is responsible for the difference in susceptibility of cells to the loss of αN-catenin? How does the cadherin/catenin system regulate axonal projection and neuronal migration? Previous studies indicate that various kinds of molecules play essential roles in neurite extension and neural migration (Hatten 2002; Wen & Zheng 2006). Interaction between the cadherin/catenin system and such molecules must be clarified. In this respect, studies using powerful Drosophila genetics will surely expand our knowledge. The cadherin/catenin system is essential for not only stabilizing cell–cell adhesion at synapses but also tethering synaptic molecules at synaptic contacts (Togashi et al. 2002; Bamji et al. 2003; Abe et al. 2004; Bozdagi et al. 2004; Okamura et al. 2004; Tanabe et al. 2006; Silverman et al. 2007). Moreover, neuronal activities modulate both structure and functions of synapses through the cadherin/catenin system (Tanaka et al. 2000; Murase et al. 2002; Abe et al. 2004). In spite of the importance of this system in synapse structure and function, involvement of individual cadherin subtypes is less understood. Recent studies revealed that the loss of N-cad or cdh8 affects synaptic functions (Jungling et al. 2006; Kadowaki et al. 2007; Suzuki et al. 2007). How does each type of cadherin regulate synaptic activity?

Elucidation of the molecular mechanisms underlying wiring specificity in the nervous system is one of the most important subjects in the field of neural development. Cadherins have been implicated in wiring specificity in the nervous system, because their expression patterns correlate with neural connectivity (Suzuki et al. 1997; Inoue et al. 1998). However, it is still unclear whether cadherins contribute to selective wiring in the vertebrate nervous system. Because a single neuron usually expresses multiple cadherins, functional redundancy among cadherins must be overcome to examine whether cadherins contribute to wiring specificity in the nervous system. On the other hand, even when the cadherin/catenin system is perturbed by overexpression of a dominant-negative cadherin during synapse formation, cell–cell contacts between axons and dendrites are formed (Togashi et al. 2002). Furthermore, synaptic contacts are less affected by overexpression of a dominant-negative cadherin in mature primary hippocampal neurons (Togashi et al. 2002). These observations indicate that cell adhesion molecules in addition to cadherins mediate cell–cell adhesion in synapses. Indeed, some cell–cell adhesion molecules belonging to the immunoglobulin superfamily have been shown to engage in selective synapse formation in the vertebrate nervous system (Yamagata et al. 2002; Ango et al. 2004). To fully understand the mechanisms underlying specific wiring in the nervous system, we must yet clarify the functional redundancy among cell adhesion molecules.

Conflict of Interest

No conflict of interest has been declared by S. C. Suzuki or M. Takeichi.