Adhesion Molecules in Synapse Assembly and Function

Abstract

As the nervous system develops, cell adhesion molecules (CAMs) interact across the synaptic junction to govern synapse assembly and function. The CAM interactions are highly dependent on the molecular composition of the extracellular domains, of which there is a large variety, and on the intracellular domains, which provide a structural and signaling platform for the pre- and the postsynaptic terminals. The diversity of CAMs generates a myriad of potential extracellular interactions, yet CAM binding is highly specific and gives rise to unique adhesive codes that guide synapse assembly and tune the efficacy of synaptic communication. The aim of this chapter is to provide an overview of some of the major synaptic CAMs, highlighting their structural features that contribute to the specificity and promiscuity of CAM interactions, and to discuss how these mechanisms intersect to control synapse formation, organization, and plasticity.

1 Introduction

Synapses mediate spatially and temporally controlled cell-to-cell communication in the nervous system. The proper functioning of synapses requires a wide array of subcellular signaling components across both sides of the synaptic cleft. This transcellular organization at the synapse is highly dependent on cell adhesion molecules (CAMs) . CAMs mechanically link the presynaptic and postsynaptic membranes to provide a trans-synaptic bridge. They also bind other CAMs or membrane-bound proteins along the same membrane surface in cis and contribute to synapse organization. Furthermore, they anchor to intracellular scaffolds to promote synapse stability and initiate intracellular signaling events to regulate synapse structure and function.

There exists an incredible diversity of CAMs involved in distinct aspects of synapse function (Fig. 17.1), although in general, CAMs fall into two major classes: those that bind to the members of the same family (homophilic binding) or those that interact with different family members (heterophilic binding). Within each family, CAMs are further subdivided based on the structure and binding properties of their extracellular adhesion domains. The numerous classes of CAMs and the potential for heterophilic interactions between CAMs of different classes create an overwhelming abundance of CAM binding interactions. Such complexity of CAM binding interactions is underscored by observations that different classes of CAMs can cooperate or compete in synapse formation (Stan et al. 2010; Aiga et al. 2010). Nevertheless, CAM binding itself is highly specific and depends not only on the presence of specific ligands, but on their relative abundance and binding affinities (Table 17.1). Moreover, the binding affinities between CAMs and their ligands can be tuned by a variety of activity-dependent modifications (Ullrich et al. 1995; Iijima et al. 2011; Aoto et al. 2013) (Table 17.2 and Box 17.2).

Fig. 17.1

Distribution profile of adhesion molecules across the dendritic tree. (a) Adhesion molecules that are localized to excitatory synapses (along the green axons) include the integrins, cadherins, NCAMs, neurexins, neuroligins (NLGN1 and NLGN3 isoforms), and LRRTMs. (b) Integrins, neurexins, and neuroligins (NLGN2 and NLGN3 isoforms) are also localized to inhibitory synapses (along the red axon), while a PTPδ/Slitrk constitute one of the only inhibitory synapse-specific adhesive complexes described to date. (c) Adhesion molecules such as NCAM and the integrins also distribute to extrasynaptic regions, as do NMDARs containing GluN2B subunits. Extrasynaptic NCAM may constitute a reserve pool that is shuttled into synaptic regions (arrows) to replenish and maintain the abundance of NCAM during ongoing turnover

Table 17.1 Binding affinities of common CAMs and their ligands

Table 17.2 Common modifications that influence CAM activity

In this chapter we will overview the key structural determinants that govern specific CAM interactions and outline how regulating such CAM interactions can influence the position, type, strength, and plasticity of particular types of synapses. While new synaptic CAMs and their ligands are continually being discovered (Biederer et al. 2002; Woo et al. 2009; Linhoff et al. 2009; Takahashi et al. 2012; de Wit et al. 2013) (Box 17.1), here we will highlight the following CAMs for which we have accumulated considerable knowledge: among the homophilic synaptic CAMs, the cadherins and Ig superfamily (IgSF) molecules, and among the heterophilic CAMs, the neuroligins (NLGNs), leucine-rich repeat transmembrane proteins (LRRTMs), netrin-G ligands (NGLs), and integrins. We end with a brief summary on the future outlook of how further knowledge on synapse adhesion proteins might contribute to the next advances in neuroscience research.

Box 17.1: Common Methods Used to Assess Adhesion Molecule Function in Neural Circuits

Assays

1.1 Cell Aggregation Assay

Some of the earliest investigations of adhesion molecules have made use of the cell aggregation assay (Rutishauser et al. 1982; Chuong et al. 1982; Miyatani et al. 1989). This assay involves assessing the degree to which suspended cells adhere to each other under different conditions. Prior transfection of cells with different adhesion molecules, their mutated isoforms, or treatment with antibodies or other agents can reveal the specifics of cell–cell adhesion.

1.2 Artificial Synapse Formation Assay

Artificial synapse formation assay s have proved to be a powerful tool in identifying the role of adhesion complexes in synapse assembly and function (Scheiffele et al. 2000; Graf et al. 2004; Linhoff et al. 2009). Two kinds of artificial synapse formation are used: one involves transfecting nonneuronal cells with DNA constructs encoding different adhesion molecule isoforms, while the other uses beads coated with particular ligands. Neurons are grown in the presence of transfected cells or coated beads, and synapse formation is assessed by monitoring the clustering of pre- or postsynaptic components to the sites of contact between neuron and the bead or the transfected cell. Other assays, such as styryl dye uptake into recycling vesicles, electron microscopy, and electrophysiology provide functional read-outs of synapses (Scheiffele et al. 2000; Graf et al. 2004; Sara et al. 2005).

1.3 Pharmacological or Antibody-Mediated Disruption

In some cases, adhesion molecule binding can be disrupted by introducing function-blocking antibodies or by naturally occurring toxins that target specific CAM interactions. These can be administered acutely, and the resulting effects can be monitored in real time. However, the results of these experiments should be interpreted carefully, as exogenous peptides mimicking ligand-binding domains could either inhibit or stimulate adhesion molecule signaling depending on the abundance of the endogenously bound ligand and/or the state of cell signaling. Ideally, these experiments are paired with gene manipulation approaches involving transgenic mice or acute overexpression or knockdown of CAM or its substrate expression, based on methods such as cell transfection, viral vectors, or electroporation.

Box 17.2: Common Activity-Dependent Modifications That Influence Adhesion Molecule Binding

1.1 Glycosylation

The addition or removal of oligosaccharides typically at N-linked (to an amide group of an asparagine residue) or O-linked (to a hydroxyl group of serine or threonine residues) glycosylation sites may influence protein folding to present or mask potential ligand-binding sites. Moreover, oligosaccharides may contain additional sugars such as polysialic acid (PSA) that contribute a strong negative charge and leads to membrane repulsion. Many adhesion molecules are regulated by glycosylation, with one prime example being the Ig superfamily molecules NCAM (Crossin et al. 1984) and SynCAM (Galuska et al. 2010; Fogel et al. 2010).

1.2 Alternative Splicing

Many adhesion molecules possess several alternative splice sites that include key residues that directly influence the three-dimensional structure of the protein and/or modify its binding properties. Consequently, proteins with distinct binding properties can be generated from single transcripts depending on the cellular context. For example, the neurexin/neuroligin complex is tightly regulated by alternative splicing (Ullrich et al. 1995; Boucard et al. 2005; Iijima et al. 2011).

1.3 Cleavage or Ectodomain Shedding

Several proteases are upregulated and secreted in an activity-dependent manner, and they target the extracellular domains of adhesion molecules. Cleavage normally leads to the generation of a soluble extracellular fragment and an intracellular fragment that may interact with additional components that would not be possible otherwise. N-cadherin and neuroligins are among those adhesion molecules whose specific susceptibility to proteinase-dependent cleavage has been reported (Reiss et al. 2005; Peixoto et al. 2012; Suzuki et al. 2012).

1.4 Phosphorylation

The addition of phosphate groups to proteins by protein kinases is one of the most common and potent signaling mechanisms in synapse cell biology. Adhesion molecules can be a substrate for phosphorylation (Giannone et al. 2013; Bemben et al. 2013) or can regulate the phosphorylation of other synaptic proteins (Bernard-Trifilo et al. 2005; Bodrikov et al. 2008). Given the proximity of synaptic adhesion molecules to neurotransmitter receptors and other synaptic machinery, these effects can be propagated quickly to tightly regulate synaptic strength and plasticity.

2 Homophilic Adhesion Molecules

2.1 The Cadherin Family

The cadherins are a superfamily of homophilic cell adhesion molecules that are classified into several subfamilies based on the organization of their protein domains and genomic structure. These include the classical (type I) cadherins, the atypical (type II) cadherins, the desmosomal cadherins, the flamingo cadherins, the protocadherins, and several independent cadherin isoforms (Nollet et al. 2000; Niessen et al. 2011). We will focus on the classical cadherins, which include the N-, E-, P-, and R-cadherins. Notably, the N- and E-cadherins are the most highly expressed in the nervous system.

2.1.1 Structure and Binding

All classical cadherins are single-pass type I transmembrane glycoproteins that possess a conserved extracellular domain consisting of five repeating modules (named EC1-5 from the most distal to the membrane proximal) (Fig. 17.2a) (Hatta et al. 1988; Niessen et al. 2011). In the cytoplasmic region, classical cadherins possess three major functional domains: a juxtamembrane domain that contributes to cis dimerization (Yap et al. 1998), a membrane proximal domain that participates in p120-catenin and δ-catenin binding (Arikkath and Reichardt 2008), and the highly conserved β-catenin -binding domain on the distal C-terminus (Niessen et al. 2011). β-Catenin binds to α-catenin , which in turn binds to filamentous actin (F-actin), thereby linking cadherin with the actin cytoskeleton. Interestingly, at an intermediate location between the p120-/δ-catenin and β-catenin-binding domains, an additional binding site for the glutamate receptor-interacting protein-1 (GRIP1) has recently been identified (Heisler et al. 2014) and could play a role in postsynaptic regulation (see below).

Fig. 17.2

The cadherins. (a) Cadherin domain organization. EC cadherin extracellular domain, PM plasma membrane, TMD transmembrane domain, JMD juxtamembrane domain, HP hydrophobic pocket, BD binding domain. (b) Illustration of the proposed “strand dimer”-binding mechanism for classical cadherins based on the crystal structure of C-cadherin described in Boggon et al. (2002)

In the extracellular domain, each EC region folds into a seven-stranded β-barrel containing three Ca²⁺-binding sequences. Specific homophilic binding is thought to occur when a conserved tryptophan side chain (Trp2 or W2) in the EC1 domain of one cadherin molecule inserts into a hydrophobic pocket in the EC1 domain of a partner molecule (Fig. 17.2b) (Nose et al. 1990; Miyatani et al. 1989; Shapiro et al. 1995; Boggon et al. 2002). This “strand dimer ” formation is essential for binding both in cis along the plane of the membrane (Shapiro et al. 1995) and in trans across the two membranes (Pertz et al. 1999; Boggon et al. 2002). While many structural studies have identified interactions between EC1 domains, additional binding mechanisms have been proposed that involve contributions from several other EC domains (Sivasankar et al. 2001; Chappuis-Flament et al. 2001; Zhu et al. 2003; Shan et al. 2004; Prakasam et al. 2006; Guo et al. 2009). Importantly, Ca²⁺ binds near the interface region between each EC domain (Nagar et al. 1996); Ca²⁺ binding facilitates trans cadherin interaction by stabilizing and extending the curved rod structure into the extracellular space and also by enhancing cis dimerization (Nagar et al. 1996; Brieher et al. 1996; Yap et al. 1998; Pertz et al. 1999).

Although homophilic cadherin interaction is highly specific (Miyatani et al. 1989), heterodimers composed of different cadherin isoforms have been observed (Niessen and Gumbiner 2002; Katsamba et al. 2009). N- and E-cadherin heterodimers bind with affinities that are intermediate between those observed for homodimers of either type, thus generating a hierarchy of cadherin-binding potency (N-N>N-E>E-E) (Katsamba et al. 2009).

2.1.2 Cadherins at Synapses

During neuronal development, N-cadherin and β-catenin are initially diffusely expressed along growing neurites and growth cones, and they subsequently become localized to synapses (Uchida et al. 1996; Fannon and Colman 1996; Benson and Tanaka 1998; Shan et al. 2000; Togashi et al. 2002). Ultrastructural analysis has revealed that N-cadherin and α- and β-catenin are localized to regions of both pre- and postsynaptic membranes adjacent to active zones and postsynaptic densities (Uchida et al. 1996; Fannon and Colman 1996), although a subsequent study has shown that cadherins are distributed evenly throughout the synaptic cleft in immature synapses and later become concentrated into hotspots within or adjacent to the synaptic cleft (Elste and Benson 2006). N-cadherin is initially present at both excitatory and inhibitory synapses in culture, but is progressively lost from inhibitory contacts as synapses mature (Benson and Tanaka 1998). Collectively these observations suggest that cadherin function in synapse organization is developmentally regulated and that N-cadherin plays an ongoing role only at excitatory synapses.

Interestingly, despite the early expression of N-cadherin in neurons prior to synapse formation, in artificial synapse formation assays N-cadherin does not induce synaptogenesis (Sara et al. 2005; Linhoff et al. 2009). Therefore, N-cadherin may contribute to the maturation of synapses rather than to their formation per se. Indeed, disrupting N-cadherin or β-catenin in cultured hippocampal neurons slows synapse formation and maturation (Togashi et al. 2002; Okuda et al. 2007); dendritic spines are longer with smaller spine heads, and exhibit reduced GluA2 surface expression and quantal amplitudes (Togashi et al. 2002; Okamura et al. 2004; Okuda et al. 2007; Vitureira et al. 2011). Moreover, in neuronal cultures prepared from αN-catenin knockout mice, dendritic spines remain immature and are less stable than wild-type spines (Togashi et al. 2002; Abe et al. 2004). Genetic deletion of p120-catenin also leads to reduced spine density, altered spine morphology, and reduced dendritic arbor complexity (Elia et al. 2006). Together, these observations demonstrate that synapse maturation, but not formation, is altered in the absence of N-cadherin and several of its intracellular partners. Whereas perturbing N-cadherin and its binding partners potently affect dendritic spine morphology, its effects on synapse function are more complex. For example, β-catenin null synapses show deficits in the postsynaptic receptor abundance, and this cellular phenotype could be rescued by exogenously expressing β-catenin mutants unable to bind to α-catenin but not by those harboring mutations in the PDZ-binding region (Okuda et al. 2007). Therefore, whereas β- and α-catenin binding is crucial for the link to actin and the spine morphology, it is dispensable for basal synaptic function.

The complexity of N-cadherin-dependent synapse regulation is further highlighted by a study in cultured hippocampal neurons that has addressed the extent of symmetric requirement across the synaptic cleft of this homophilic adhesion complex. Disrupting postsynaptic N-cadherin but not β-catenin retrogradely reduces basal release probability (p_r) of contacting presynaptic terminals (Vitureira et al. 2011). Interestingly, however, disrupting presynaptic N-cadherin does not replicate the effects on p_r, arguing against the role of N-cadherin homophilic binding for controlling p_r. Instead, postsynaptic N-cadherin may interact with the GluA2 AMPA receptor subunit to control presynaptic p_r trans-synaptically (Saglietti et al. 2007; Vitureira et al. 2011). Presynaptic N-cadherin/β-catenin complex however does play a role in presynaptic organization, specifically in recruiting the reserve and resting pool synaptic vesicles to sites of cadherin clustering (Bamji et al. 2003, 2006).

Synaptic activity strongly regulates the distribution pattern, internalization rate, and molecular composition of cadherin/catenin complexes (Tanaka et al. 2000; Bozdagi et al. 2000; Murase et al. 2002; Abe et al. 2004; Okamura et al. 2004; Tai et al. 2007; Brigidi et al. 2014). Following elevated network activity, N-cadherin redistributes along the surface of the membrane, forms cis-dimers, and becomes protease resistant (Tanaka et al. 2000). Activity also changes N-cadherin surface abundance in an NMDA receptor-dependent manner by slowing N-cadherin endocytosis via enhanced intracellular binding to β-catenin (Bozdagi et al. 2000; Tai et al. 2007). These activity-dependent changes point to the role of cadherins as transducers of synaptic signaling.

The relationship between activity and cadherins is reciprocal in that cadherins also impact activity-dependent synaptic processes. Infusing function-blocking antibodies against N- and E-cadherin have been reported to attenuate LTP in hippocampal slices (Tang et al. 1998; Bozdagi et al. 2000), and introducing mutations into the W2 domain of N-cadherin alters depolarization-induced spine dynamics in culture (Okamura et al. 2004). Some of these effects may be due to disrupting a direct interaction between the extracellular domains of N-cadherin and the GluA2 subunit of the AMPA receptor (Saglietti et al. 2007) or compromising an intracellular interaction between N-cadherin and GRIP1 (Heisler et al. 2014) or other intracellular signaling molecules, including δ-catenin , which has been shown to be essential for spine head expansion and surface AMPA receptor insertion (Brigidi et al. 2014). N-cadherin has also been implicated in homeostatic modulation of postsynaptic strength in response to chronic activity alterations (Vitureira et al. 2011). However, contrary to its role in basal presynaptic strength regulation, postsynaptic N-cadherin does not retrogradely influence the homeostatic plasticity of p_r, although it is required for homeostatic modulation of postsynaptic AMPA receptors. Postsynaptic β-catenin, on the other hand, is required for homeostatically adjusting both pre- and postsynaptic strengths, the former occurring in a retrograde fashion independently of N-cadherin (Okuda et al. 2007; Vitureira et al. 2011).

Altogether, the cadherins and their accessory proteins play important roles in controlling synapse structural maturation, organization, and function. The highly complex nature of the regulation is reflected not only by the numerous interacting proteins, particularly in the intracellular domain of cadherin, but also by the developmental changes that are observed even at the same synapse type.

2.2 Ig Superfamily

The Ig superfamily (IgSF) CAMs participate in a wide range of cell-to-cell and cell-to-ECM interactions. All IgSF members are characterized by the number of extracellular Ig-like domains that participate in homophilic (Zhou et al. 1993, 1998; Rao et al. 1994) and heterophilic adhesion (Felding-Habermann et al. 1997; Blaess et al. 1998; Fogel et al. 2007). Most IgSF members possess a variable number of fibronectin type III (FNIII) domains, followed either by a GPI membrane anchor or a transmembrane domain (Walsh and Doherty 1997; Maness and Schachner 2007). Unlike cadherins, the intracellular domain of IgSF CAMs is highly variable, ranging from sequences that confer PDZ-like binding (Polo-Parada et al. 2005; Biederer 2006) and cytoskeletal interactions (Davis and Bennett 1994; Dahlin-Huppe et al. 1997; Leshchyns’ka et al. 2003) to phosphatase domains (Johnson and Van Vactor 2003).

Individual Ig domains are approximately 100 amino acids in length that fold into a tertiary structure composed of two β-sheets stabilized by disulfide bonds (Wang 2013). Each β-sheet is itself composed of a variable number of β-strands that generate three distinct subgroups: a variable set (V-set), a conserved set (C-set; further subdivided into C1-set and C2-set), and an intermediate set (I-set; further subdivided into I1-set and I2-set) (Smith and Xue 1997; Wang 2013). Different IgSF CAMs are constructed of different combinations of these Ig domains and this variability contributes to their specific binding affinities (Zhou et al. 1993; Biederer 2006).

In this chapter, we will focus on the neural cell adhesion molecule (NCAM) and synaptic cell adhesion molecule (synCAM) subfamilies that have received considerable attention for their roles at central synapses.

2.2.1 NCAM

NCAM is a Ca²⁺-independent CAM widely expressed throughout the developing and the adult nervous system. Although considered primarily a homophilic CAM, NCAM can also participate in a multitude of heterophilic interactions with other members of the Ig superfamily such as L1 (Horstkorte et al. 1993) and TAG-1/axonin (Milev et al. 1996) or with several growth factors and growth factor receptors such as the fibroblast growth factor receptor (FGFR) (Kiselyov et al. 2005), the glial cell-line-derived neurotrophic factor (GDNF) (Paratcha et al. 2003), and the brain-derived neurotrophic factor (BDNF) (Vutskits et al. 2001). Moreover, NCAM robustly interacts with several ECM -related molecules including the chondroitin sulfate proteoglycans, neurocan (Friedlander et al. 1994) and phosphacan (Milev et al. 1994), collagens (Probstmeier et al. 1992), heparin (Cole et al. 1986), and agrin (Storms and Rutishauser 1998).

2.2.1.1 NCAM Structure and Binding

Alternative splicing of a single NCAM gene generates a subfamily that is composed of at least three main isoforms expressed in neural tissue and named according to their molecular weights – NCAM120, NCAM140, and NCAM180 (Fig. 17.3a). All isoforms possess identical extracellular domains that mediate Ca²⁺-independent adhesion, but differ in the presence or absence of the transmembrane domain and the intracellular sequence. NCAM120 is linked to the membrane via a glycophosphatidylinositol (GPI) anchor, while NCAM140 and NCAM180 both contain transmembrane domains followed by an intracellular domain that includes a 40 kDa insert containing a PDZ-like (KANESKA) motif on the distal C-terminus (Polo-Parada et al. 2005). An additional alternatively spliced sequence in the intracellular domain of NCAM180 contains dynein and cytoskeletal-binding domains (Polo-Parada et al. 2005; Perlson et al. 2013).

Fig. 17.3

Immunoglobulin superfamily members NCAM and SynCAM. (a) Domain organization of NCAM120, NCAM140, and NCAM180 isoforms and SynCAM. Ig immunoglobulin domain; FN fibronectin type III domain; GPI glycophosphatidylinositol; 180-SD NCAM180-specific domain; PDZ BD postsynaptic density 95, disc large, zona occludens 1 binding domain. Key amino acid sequences that participate in synaptic functions are shown in brackets. (b) Illustration of proposed homophilic binding mechanisms of NCAM in cis and trans based on crystal structures described in Soroka et al. (2003)

The extracellular domain of NCAM is composed of five I-type Ig domains linked by disulfide bonds followed by two FNIII domains (Jensen et al. 1999; Kiselyov et al. 2005). A long, negatively charged carbohydrate chain known as α2,8-linked polysialic acid (PSA) can bind to the fifth Ig domain to inhibit NCAM-mediated adhesion (Rutishauser et al. 1985). The addition of 10-amino acid sequence, termed VASE (variable alternatively spliced exon), to the fourth Ig domain of NCAM transforms it to resemble a V-type Ig domain (Small et al. 1988; Small and Akeson 1990) and may influence its binding affinities (Chen et al. 1994).

How NCAM Ig domains bind to each other has been disputed, and several different binding conformations have been proposed. These include a model where all five Ig domains participate in binding in an antiparallel manner (Rao et al. 1994; Ranheim et al. 1996), a model where only Ig domain 3 self-associates (Rao et al. 1993), and a model where Ig domains 1 and 2 bind in an antiparallel manner (Jensen et al. 1999; Kasper et al. 2000). More recent work has suggested that cis-binding occurs between Ig domains 1 and 2, while trans-binding occurs via three possible antiparallel conformations: between Ig domains 2 and 3, 1 and 3, or 2 and 2 (Fig. 17.2b) (Soroka et al. 2003). NCAM “zippers” would then be assembled as clusters of cis-dimers interacting in trans across the membrane.

2.2.1.2 NCAM at the Synapse

NCAM isoforms are shuttled to regions of neurite contact where they accumulate during the early stages of synaptogenesis (Sytnyk et al. 2002) and persist into maturity (Schuster et al. 2001). During initial stages of synapse formation, NCAM may facilitate the transition from neurite outgrowth (Leshchyns’ka et al. 2003; Chernyshova et al. 2011) to contact stabilization by regulating the delivery of trans-Golgi-derived cargo to the plasma membrane at specific sites in response to target-derived cues (Hata et al. 2007; Chernyshova et al. 2011). As such, several structural and functional deficits in both pre- and postsynaptic compartments have been observed at synapses from NCAM knockout animals (Rafuse et al. 2000; Polo-Parada et al. 2001; Dityatev et al. 2004). However, NCAM is not required for synapse formation. Like N-cadherin, NCAM is unable to induce synapse formation in artificial synapse formation assays (Sara et al. 2005), and instead, NCAM likely functions to promote synapse maturation and activity-dependent adjustments in synaptic strengths (Dityatev et al. 2000; Polo-Parada et al. 2001; Cambon et al. 2004; Puchkov et al. 2011; Chipman et al. 2014a).

NCAM140 and NCAM180 are most abundantly expressed at synapses. Presynaptically, they influence p_r, possibly by regulating a developmental switch in synaptic vesicle recycling and/or myosin light chain kinase activity involved in controlling synaptic vesicle pools (Polo-Parada et al. 2004; Hata et al. 2007; Shetty et al. 2013; Chipman et al. 2014a, b). NCAM180 is highly enriched in postsynaptic membranes where it is thought to contribute to postsynaptic organization and plasticity (Schuster et al. 2001; Sytnyk et al. 2006; Leshchyns’ka et al. 2011). Specifically, NCAM associates with α1β1-spectrin that links NCAM to the NMDA receptor subunits, NR1 and NR2B, and CaMKIIα (Sytnyk et al. 2006). These interactions are critical for the appropriate clustering (Sytnyk et al. 2006), stability (Puchkov et al. 2011), and functionality (Kochlamazashvili et al. 2012) of these postsynaptic components. Moreover, perturbing NCAM function disrupts NMDAR-dependent LTP and LTD in the hippocampus (Lüthi et al. 1994; Muller et al. 1996; Eckhardt et al. 2000; Bukalo et al. 2004; Kochlamazashvili et al. 2012). However, some of these effects may represent the indirect consequences of losing NCAM as a carrier for PSA (Muller et al. 1996; Eckhardt et al. 2000).

NCAM adhesion and signaling are regulated by activity-dependent mechanisms. For instance, in hippocampal slices, PSA expression is modulated by activity in a spatially and temporally defined manner (Muller et al. 1996). Recent studies have demonstrated that ectodomain shedding of NCAM140 and NCAM180, a process that reduces their homophilic binding, can be induced by matrix metalloproteases such as ADAM (a disintegrin and metalloprotease)-8, ADAM-10, and ADAM-17 (Hübschmann et al. 2005; Hinkle et al. 2006; Kalus et al. 2006; Brennaman et al. 2013). Furthermore, once released, soluble NCAM ectodomain fragments have the potential to accumulate in the extracellular matrix and influence subsequent NCAM substrate adhesion and the synaptic activities that depend on it. Accordingly, transgenic mouse lines that overexpress NCAM ectodomain fragments in neurons show deficits in LTP and impaired inhibitory synapse formation in the cortex (Pillai-Nair et al. 2005; Brennaman et al. 2011). One study has reported increased amounts of extracellular NCAM following LTP induction in dentate gyrus (Fazeli et al. 1994), suggesting that activity may regulate NCAM ectodomain shedding. Notably, accumulation of NCAM ectodomain has been identified in patients with schizophrenia , and NCAM dysfunction is associated with cognitive disorders (Sandi 2004). These observations underscore the importance of NCAM-dependent mechanisms of synaptic circuit regulation.

2.2.2 SynCAM

The SynCAM subfamily is composed of four genes that code for four isoforms (SynCAM1–SynCAM4) that share significant sequence homology (Biederer 2006). In the brain, SynCAM1 and SynCAM4 are expressed embryonically, while SynCAM2 and SynCAM3 are detectable only postnatally (Fogel et al. 2007).

2.2.2.1 Structure and Binding

All SynCAM family members are Ca²⁺-independent CAMs characterized by three extracellular Ig domains, a transmembrane domain, and a carboxyl tail that contains a PDZ-like type II-binding domain (Fig. 17.3a) (Biederer et al. 2002; Biederer 2006). According to the classification of Ig-like domain subclasses, the sequence information of SynCAM indicates that the first Ig domain of all SynCAM isoforms belongs to the so-called V-set, while the second belongs to the C1-set, and the third to the I-set (Biederer 2006). All Ig domains possess N-glycosylation sites (Biederer et al. 2002; Biederer 2006), but only the first Ig domains of SynCAM1 and 2 can be heavily glycosylated (Fogel et al. 2007, 2010) or polysialylated, both of which influence binding affinities (Galuska et al. 2010; Rollenhagen et al. 2012).

All SynCAM isoforms, with the exception of SynCAM4, participate in homophilic binding but preferentially assemble into heterodimers composed of either SynCAM1 and SynCAM2, or SynCAM3 and SynCAM4 (Fogel et al. 2007). The first Ig domain of SynCAM1 and SynCAM2 mediate their trans-synaptic adhesion (Fogel et al. 2010), while the second and third Ig domains of SynCAM1 engage in cis assembly in a manner that strengthens trans interactions (Fogel et al. 2011). N-glycosylation of each isoform modifies their binding affinities to inhibit SynCAM2 binding while promoting SynCAM1 binding (Fogel et al. 2010). Polysialylation also modifies SynCAM’s adhesive properties (Galuska et al. 2010). These modifications occur in a developmental and region-specific manner and can potently modulate synapse formation.

2.2.2.2 SynCAM at the Synapse

All SynCAM isoforms have been detected at nascent and established excitatory and inhibitory synapses, although their distribution profiles differ somewhat between characterizations in vitro and in vivo (Biederer et al. 2002; Sara et al. 2005; Fogel et al. 2007; Stagi et al. 2010; Robbins et al. 2010). In vivo, SynCAM1 has been found to distribute primarily to excitatory synapses in the hippocampus (Robbins et al. 2010). SynCAM1 and SynCAM2 are potent synapse-organizing CAMs (Sara et al. 2005; Fogel et al. 2007; Stagi et al. 2010). When expressed in nonneuronal cells, they induce the clustering of functional presynaptic components, including synaptic vesicle proteins and SNARE machinery in contacting axons, and when overexpressed in neurons, they enhance excitatory but not inhibitory synapse formation (Biederer et al. 2002; Sara et al. 2005; Robbins et al. 2010). Sustained SynCAM expression into adulthood is required for the maintenance of excitatory synapses. Functionally, SynCAM constrains LTD but has little effect on LTP (Robbins et al. 2010).

Many of the synaptic effects of SynCAM disruption involve its intracellular domains: the juxtamembrane motif that is predicted to bind to members of the protein 4.1 family, and the PDZ type II-binding motif (Biederer et al. 2002; Sara et al. 2005). The protein 4.1 family consists of a cytoskeleton-interacting protein that stabilizes the spectrin /actin cytoskeleton (Hoover and Bryant 2000), while the PDZ type II-binding domain contains an EYFI sequence that can bind to a range of synapse-specific proteins implicated in synapse formation and plasticity (Sheng and Sala 2001). Altogether, among the homophilic synapse adhesion proteins, SynCAM displays the most potent synaptogenic activity. Future studies may further expand its repertoire in regulating aspects of synapse function.

3 Heterophilic Adhesion Molecules

3.1 Neuroligin/Neurexin Complex

The neuroligin (NLGN)/neurexin (NRXN) complex is one of the most widely studied synaptic adhesion complexes, in part because of its association with human cognitive disorders such as autism spectrum disorders , Tourette syndrome , and schizophrenia (Südhof 2008). As Ca²⁺-dependent heterophilic CAMs, postsynaptic NLGNs bind to presynaptic NRXNs to promote the formation and maturation of stable synaptic contacts (Nguyen and Südhof 1997; Scheiffele et al. 2000; Missler et al. 2003; Graf et al. 2004; Varoqueaux et al. 2006; Chubykin et al. 2007; Berninghausen et al. 2007).

3.1.1 NRXN Structure

NRXNs are encoded by three genes (NRXN1-3), each of which generates a long isoform (α-NRXNs ) and a short isoform (β-NRXNs ) that are driven by distinct promoter regions (Ushkaryov et al. 1992). α-NRXNs possess six laminin, NRXN, sex-hormone-binding globulin (LNS) domains with three EGF (epidermal growth factor )-like domains that are interdigitated between the first and second, third and fourth, and fifth and sixth LNS domains to form three repeating LNS(A)-EGF-LNS(B) cassettes (Fig. 17.4a) (Ushkaryov et al. 1992; Missler et al. 1998). β-NRXNs, on the other hand, possess only one LNS(B) domain, which is identical to the sixth LNS domain of α-NRXN. In both α- and β-NRXN isoforms, the membrane proximal LNS(B) domain is followed by a linker region containing O-linked glycosylation sites, which, when occupied, may help extend the extracellular domains into the synaptic cleft. This region is followed by a transmembrane domain and a PDZ type II-binding domain, which can interact with N-type Ca²⁺ channels and the presynaptic actin network through CASK and protein 4.1 (Ushkaryov et al. 1992; Hata et al. 1996; Missler et al. 1998, 2003; Biederer and Südhof 2001; Mukherjee et al. 2008).

Fig. 17.4

Neurexins and neuroligins. (a) α- and β-neurexin domain organization. LNS, laminin, NRXN, sex-hormone-binding globulin domains; EGF epidermal growth factor domain; SS1–5 splice sites 1–5. (b) Neuroligin domain organization. CLD cholinesterase-like domain, SSA splice site A, SSB splice site B – only present in NLGN1. (c) Illustration of proposed binding mechanism between α- and β-NRXN and NLGN based on crystal structures derived by Araç et al. (2007) and presented in Sudhof (2008)

3.1.2 NLGN Structure

Neuroligins are single-pass transmembrane proteins and belong to the class of α/β hydrolase superfamily of proteins, which share significant sequence homology with acetylcholinesterases (Araç et al. 2007; Fabrichny et al. 2007). Four genes (NLGN1-4) code for NLGNs in mammals, and humans possess a fifth (NLGN5). Each isoform contains a highly conserved cholinesterase-like domain (CLD) that participates in heterophilic binding with NRXNs in trans and homophilic dimerization in cis (Fig. 17.4b) (Dean et al. 2003; Araç et al. 2007; Fabrichny et al. 2007; Chen et al. 2007). The CLD domain, which contains several N-glycosylation sites as well as a splice site (Ichtchenko et al. 1995; Hoffman et al. 2004; Araç et al. 2007), adopts an ellipsoid shape composed of a twisted β-sheet surrounded by multiple helices (Araç et al. 2007; Fabrichny et al. 2007). A linker sequence bridges the CLD and the transmembrane region and contains several O-glycosylation sites (Ichtchenko et al. 1995; Hoffman et al. 2004) that may help extend the globular CLD into the synaptic cleft in a manner similar to NRXN (Südhof 2008). The intracellular C-terminal end of NLGNs contains several domains that interact with postsynaptic signaling proteins and scaffolding molecules. For instance, NLGN1 contains a type I PDZ domain-binding motif and interacts with proteins such as PSD-95 (Irie et al. 1997).

NLGN constitutively forms dimers via key residues in the alpha helix region at the base of the CLD (Dean et al. 2003; Comoletti et al. 2007; Araç et al. 2007; Fabrichny et al. 2007; Chen et al. 2007; Shipman and Nicoll 2012a). Homo- and heterodimers, composed of NLGN1 and NLGN3 and NLGN2 and NLGN3, have been identified in the brain (Comoletti et al. 2006; Budreck and Scheiffele 2007; Shipman et al. 2011).

3.1.3 Binding

Both NLGNs and NRXNs are alternatively spliced to generate a wide array of potential binding interactions (Ullrich et al. 1995; Ichtchenko et al. 1995, 1996; Boucard et al. 2005) that may function to specify cell-type and circuit-specific connections with unique synaptic strength and plasticity properties (Südhof 2008). All NLGN isoforms possess a splice site within the CLD (called SSA) (Ichtchenko et al. 1995, 1996; Boucard et al. 2005). A second splice site (called SSB) is active only in NLGN1 and controls NLGN’s binding affinity for α- and β-NRXNs (Boucard et al. 2005; Koehnke et al. 2010). α-NRXNs contain 5 splice sites in their extracellular domains, termed SS1–5; SS1 is after the first EGF-like domain; SS2, SS3, and SS4 are in the LNS(B) domains of cassettes 1, 2, and 3, respectively; and SS5 resides in the O-glycosylated proximal domain (Fig. 17.4a) (Missler et al. 1998). Due to their shorter extracellular domains β-NRXNs possess only SS4 and SS5. Altogether, alternative splicing of NRXN can generate thousands of individual isoforms (Ullrich et al. 1995), each of which may have different binding affinities for their postsynaptic partners and thus may vary in their potency of synaptic influence.

The heterophilic NLGN/NRXN adhesion is mediated by binding between the CLD of NLGNs and the sixth LNS domains of α- and β-NRXNs (Fig. 17.4b) (Boucard et al. 2005; Südhof 2008). However, the β-unique N-terminal domain must be present for β-NRXN binding (Nguyen and Südhof 1997) while the third EGF domain is required for α-NRXN binding (Boucard et al. 2005). Different NLGN isoforms exhibit different affinities for β-NRXNs (Comoletti et al. 2006), indicating that dimer composition could potently influence the binding affinity. Crystal structures of NLGN/β-NRXN1 complexes demonstrate that two β-NRXN1 monomers bind to each NLGN molecule on either side of the dimer interface (Araç et al. 2007; Fabrichny et al. 2007). Within the binding interface, Ca²⁺ ions occupy two distinct sites, and are coordinated by ligands from either molecule (Araç et al. 2007). The β-NRXN binding site on NLGN1 is located adjacent to the SSB site in the CLD (Boucard et al. 2005; Araç et al. 2007). Likewise, the NLGN binding site of β-NRXN is located next to the SS4 site in the LNS domain of β-NRXN. The proximity of these splice sites to the binding regions confers significant regulation of binding affinity when different splice variants are present at the synapse (Boucard et al. 2005; Comoletti et al. 2006; Siddiqui et al. 2010).

3.1.4 NLGN/NRXN at the Synapse

Both NLGNs and NRXNs are highly enriched throughout the developing and adult brain (Ullrich et al. 1995; Varoqueaux et al. 2006). Whereas NLGN1 distributes primarily to the PSD of excitatory synapses, NLGN2 distributes primarily to inhibitory synapses, and NLGN3 can heterodimerize with NLGN1 or NLGN2 and distributes to both types of synapses (Song et al. 1999; Graf et al. 2004; Chih et al. 2006; Chubykin et al. 2007; Budreck and Scheiffele 2007). Nevertheless, these associations are highly variable (see below) (Chih et al. 2006; Levinson et al. 2010).

The synaptogenic activity of NLGN (Scheiffele et al. 2000) and NRXN (Graf et al. 2004) is supported by artificial synapse formation assays. However, synapse formation proceeds rather normally in the complete absence of NLGNs, whereas synapses fail to develop mature properties of synaptic transmission (Varoqueaux et al. 2006). Interestingly, a recent study has shown that synapse density is significantly reduced only when NLGN1 is knocked-down in a subset of neurons rather than across all neurons (Kwon et al. 2012). Moreover, the same study suggests that relative NLGN abundance across neurons correlates with excitatory synapse density, although synapse-to-synapse variability has not been assessed. Therefore, competitive mechanisms may underlie NLGN/NRXN-dependent synapse formation.

Further experiments have focused on understanding the isoform and splice site variations in both NLGN and NRXN and how they contribute to specific heterophilic interactions. The SSB site of NLGN strongly regulates NRXN binding: the SSB-containing NLGN [NLGN(+SSB)] binds only to β-NRXNs that lack the SS4 site, whereas the SSB-lacking NLGNs [NLGN(−SSB)] bind to both α- and β-NRXNs regardless of the presence of SS4 (Ichtchenko et al. 1995; Boucard et al. 2005; Comoletti et al. 2006). Specific heterophilic interactions created by the isoform and splice site variations result in differences in synaptogenic activity. Neurons overexpressing full-length NLGN bear more numerous and larger synapses than those expressing NLGN(−SSA/−SSB) (Boucard et al. 2005). The subtle regulation of synapse assembly by the presence or absence of SSB in NLGN is further highlighted in the artificial synapse formation assay: postsynaptic cells expressing NLGN(+SSB) slow the accumulation of presynaptic components as compared to cells expressing NLGN(−SSB) (Lee et al. 2010). Impairing the N-linked glycosylation within the SSB that promotes binding to α-NRXN (Boucard et al. 2005) hastens synaptogenesis; this suggests a key role for N-glycosylation of SSB in slowing the rate of synaptogenesis by favoring NLGN binding to β-NRXNs (Lee et al. 2010).

Alternative splicing of NRXN isoforms also impacts synapse formation. In a well-studied case, β-NRXN(−SS4) isoforms potently cluster excitatory and inhibitory postsynaptic components in artificial synapse assembly assays. The presence of SS4 significantly reduces clustering of PSD-95 and NLGN1, NLGN3, and NLGN4 without an effect on gephyrin or NLGN2 (Graf et al. 2006).

Different isoforms and alternative splicing of NLGNs not only regulate the rate and extent of synaptogenesis, but they themselves are differentially localized to excitatory and inhibitory synapses. Such differential distribution of NLGNs has been proposed to contribute to the fine balance between excitation and inhibition that exists in neural circuits (Graf et al. 2004; Chih et al. 2005, 2006; Levinson and El-Husseini 2005; Chubykin et al. 2007). For example, NLGN1(+SSA/+SSB) is preferentially localized to excitatory synapses, while NLGN1(−SSA/−SSB) is comparably distributed to both excitatory and inhibitory synapses; NLGN1(+SSA/−SSB) is mostly present at inhibitory synapses, whereas NLGN1(−SSA/+SSB) is targeted to excitatory synapses (Chih et al. 2006). Likewise, synapse targeting of NLGN2 is dependent on the SSA region in that NLGN2(+SSA) is preferentially localized to inhibitory synapses, while NLGN2(−SSA) is equally distributed between excitatory and inhibitory synapses (Chih et al. 2006). As one might expect, the ability of these isoforms to promote the assembly of a particular synapse type parallels their distribution patterns (Chih et al. 2006; Chubykin et al. 2007). The stoichiometry of PSD-95/gephyrin association with NLGNs is also important. Disrupting PSD-95 levels substantially alters the balance between excitation and inhibition in an NLGN-dependent manner (Prange et al. 2004). Moreover, knockdown of gephyrin shifts the association of NLGN2 from inhibitory to excitatory synapses, while a knockdown of PSD-95 shifts NLGN2 and NLGN3 from excitatory to inhibitory synapses (Levinson et al. 2010). However, several of NLGN’s major synaptic effects have been shown to occur independently of its PDZ and gephyrin-binding domains (Shipman et al. 2011), suggesting that multiple intracellular domains participate in NLGN function at the synapse.

Interestingly, to add another layer of complexity, the synaptogenesis-promoting effect of NLGN isoforms is activity-dependent. NMDAR-mediated Ca²⁺-influx and CaMKII activation are required for NLGN1-induced enhancement of excitatory synapse formation, while network activity is required for NLGN2-induced inhibitory synapse formation (Chubykin et al. 2007). Furthermore, a recent study has shown that activity shifts the interaction of NLGN1 with PSD-95 in favor of gephyrin by phosphorylation of NLGN1 at a tyrosine residue in its C-terminus (Giannone et al. 2013), which is homologous to a sequence in NLGN2 responsible for gephyrin interaction (Poulopoulos et al. 2009). Additionally, some recent studies have demonstrated activity-dependent cleavage of the extracellular domain of NLGNs by matrix metalloprotease 9 (MMP-9) and ADAM-10 (Peixoto et al. 2012; Suzuki et al. 2012). NLGN cleavage occurs within its juxtamembrane region and generates a secreted form of NLGN and a C-terminal stalk (Peixoto et al. 2012; Suzuki et al. 2012). Induction of NLGN cleavage destabilizes presynaptic β-NRXNs and reduces presynaptic p_r (Peixoto et al. 2012), whereas inhibiting NLGN cleavage increases dendritic spine density (Suzuki et al. 2012).

Altogether, NLGN/NRXN heterophilic adhesion controls many aspects of synapse assembly and function. Although not exclusively, NLGN’s influence at excitatory synapses are executed via modifying postsynaptic AMPA (Heine et al. 2008; Etherton et al. 2011) and NMDA (Chubykin et al. 2007; Shipman and Nicoll 2012b; Kwon et al. 2012) receptor function at excitatory synapses or via gephyrin at inhibitory synapses (Poulopoulos et al. 2009; Levinson et al. 2010). NLGNs also influence presynaptic maturation in a retrograde manner (Futai et al. 2007; Wittenmayer et al. 2009; Lee et al. 2010; Peixoto et al. 2012).

3.2 Leucine-Rich Repeat-Containing Molecules (LRRTM/Slitrk/NGL/SALM)

Several known synaptic CAMs possess extracellular leucine-rich repeat (LRR) domains and contribute to synapse formation, maturation, and plasticity. Similar to NLGNs, LRR-containing CAMs have recently received considerable attention because their mutations have been associated with autism spectrum disorders and schizophrenia (Francks et al. 2007; Sousa et al. 2010; Takashima et al. 2011; Ko 2012). The LRR-containing CAM families include the LRRTMs that bind to presynaptic NRXNs (Lauren et al. 2003; Linhoff et al. 2009; de Wit et al. 2009; Ko et al. 2009) and other ligands (de Wit et al. 2013; Siddiqui et al. 2013), the Slit and Trk-like family that bind to presynaptic protein tyrosine phosphatases (PTPs) (Takahashi et al. 2011, 2012), netrin-G ligands (NGLs) that bind to GPI-linked netrin-Gs and PTPs (Nakashiba et al. 2000; Lin et al. 2003b; Kim et al. 2006; Woo et al. 2009; DeNardo et al. 2012), and synapse adhesion-like molecules (SALMs) that are homophilic CAMs (Wang 2006; Ko et al. 2006; Seabold et al. 2008). Many members of these LRR-containing CAMs have been discovered using artificial synapse formation assays, suggesting that LRR-containing molecules are potent synaptic organizing CAMs. Moreover, these adhesion complexes can cooperate or compete with each other or other synaptic CAMs to profoundly influence the specifics of synapse formation and function (Siddiqui et al. 2010; Soler-Llavina et al. 2011).

3.2.1 The Structure of the LRR Domain

LRR domains are composed of repeating sequences of 20–29 amino acids with a conserved 11 amino acid sequence of LxxLxLxxN/CXL, where x can be any amino acid and L positions can be occupied also by valine, isoleucine, or phenylalanine (Kobe and Kajava 2001). The earlier crystal structures of the LRR domain have revealed that individual LRRs fold into a non-globular, horseshoe-shape (Fig. 17.5a) (Kobe and Deisenhofer 1993, 1995; Kobe and Kajava 2001); the parallel arrangement of α-helices and β-sheets forms an outer surface of α-helices that influence the curvature and an inner surface lined by β-sheets that provide a number of potential ligand-binding sites (Kobe and Kajava 2001; Ko 2012). The structure and binding specificity of the LRR domain can vary significantly in the number of repeats as well as in the composition of α-helix, β-sheets, and intermediate regions (Kobe and Kajava 2001).

Fig. 17.5

Leucine-rich repeat-containing cell adhesion molecules, LRRTM, and netrin-G ligand. (a) Illustration of domain organization of LRRTM. The organization of LRR domains are based on crystal structures described by Kobe and Deisenhofer (1995). LRR domain leucine-rich repeat domain, N-terminal SS N-terminal signal sequence. (b) Illustration of the domain organization of netrin-G ligand

3.2.1.1 LRRTM/Neurexin Complex

LRRTMs are type I transmembrane proteins, and in human and mice they are encoded by four genes (LRRTM1-4), three of which (LRRTM1-3) are located within the gene coding for α-catenin (Lauren et al. 2003). All isoforms are highly expressed in the brain shortly after birth in a region-specific and non-overlapping manner (Lauren et al. 2003). During development, they become tightly localized to synapses (Linhoff et al. 2009; de Wit et al. 2009).

3.2.2 Binding

All LRRTM isoforms possess similar domain structures composed of an N-terminal signal sequence, ten LRRs flanked by cysteine-rich sequences, a transmembrane domain, and an intracellular tail that contains a highly conserved PDZ-like motif that binds to PSD-95 (Fig. 17.5a) (Lauren et al. 2003; de Wit et al. 2009). All isoforms have N-glycosylation sites in the third LRR domain, although additional sites are differentially distributed across isoforms (Lauren et al. 2003).

LRRTM1 and LRRTM2 can bind to all α- and β-NRXN isoforms lacking the SS4 insert in a Ca²⁺-dependent manner (de Wit et al. 2009; Ko et al. 2009; Siddiqui et al. 2010). Specific extracellular domain interactions between NRXN and LRRTM have not been directly solved, but mutagenesis studies have shown that D260 and T262 residues within the concave β-sheet of the LRR domain of LRRTM2 are responsible for β-NRXN binding (Siddiqui et al. 2010). Moreover, the binding domain on NRXNs for LRRTMs is likely shared with that for NLGNs, leading to potential competing interactions between these three protein family members. Interestingly, however, LRRTMs and NLGNs cooperate in synapse development (Siddiqui et al. 2010), suggesting a complex interaction between NRXN/LRRTM and NRXN/NLGN complexes during synapse formation and maturation. The extracellular domain of LRRTM4 has recently been shown to bind in trans to the GPI-linked heparin sulfate proteoglycan, glypican, in an NRXN-dependent manner (de Wit et al. 2013; Siddiqui et al. 2013), thus expanding the molecular repertoire of LRRTM interactions.

3.2.3 LRRTMs at the Synapse

All LRRTM isoforms promote excitatory synapse formation in the artificial synapse formation assay: LRRTM1 and LRRTM2 are the most potent organizers, while LRRTM3 and LRRTM4 are less potent (Linhoff et al. 2009). Acute knockdown of LRRTM1 and LRRTM2 leads to a specific deficit in AMPA, but not NMDAR-mediated basal excitatory synaptic responses (de Wit et al. 2009; Soler-Llavina et al. 2011) along with deficits in LTP in young and mature hippocampal CA1 regions (Soler-Llavina et al. 2013). Knockdown of LRRTM4 in a subset of cortical neurons via in utero electroporation leads to a loss of synapses (de Wit et al. 2013). Constitutive loss of LRRTM1 (Linhoff et al. 2009) and LRRTM4 (Siddiqui et al. 2013) in mice show distinct phenotypes, which likely reflect their differential expression and binding partners. Whereas LRRTM1 knockout mice exhibit enhanced vesicle accumulation at synapses in CA1 stratum radiatum and stratum oriens, LRRTM4 knockout mice have normal synaptic vesicle accumulation in CA1 but reduced synaptic vesicle accumulation throughout the dentate gyrus. The latter phenotype is accompanied by deficits in DG granule cells of dendritic spine density, PSD-95 accumulation, and activity-dependent AMPA receptor surface expression (Soler-Llavina et al. 2013; Siddiqui et al. 2013). While these findings demonstrate distinct functions of LRRTM isoforms according to their binding affinity and distribution patterns, they also suggest that the shared actions of LRRTMs converge on the regulation of AMPAR abundance and surface expression.

3.2.3.1 Netrin-G Ligands

Netrin-G ligand-1 (NGL1) was originally identified as a growth-promoting ligand of netrin-G in thalamocortical axons (Lin et al. 2003b). Subsequent studies have identified two additional NGL isoforms, NGL2 and NGL3, in a yeast two-hybrid screen for PSD-95-interacting molecules (Kim et al. 2006). All isoforms are Ca²⁺-independent heterophilic CAMs and are highly enriched in and distributed throughout the brain from the early postnatal period where they localize to central regions of the postsynaptic density (Kim et al. 2006).

3.2.4 NGL Structure and Binding

NGLs are composed of an N-terminal signal peptide, followed by nine LRR domains flanked by N and C-terminal LRR domains, an Ig domain, a type 1 transmembrane domain, and a C-terminal domain containing an ETQI PDZ-binding motif (Fig. 17.5b) (Lin et al. 2003b; Kim et al. 2006). NLG3 isoforms possess an additional N-terminal motif which is absent from NGL1 and NGL2, along with an extended linker sequence between the transmembrane and Ig domains.

NGL1 binds to netrin-G1 (also called laminet-1) but not netrin-G2 (also called laminet-2), whereas NGL2 show preferential binding to netrin-G2 over netrin-G1 (Lin et al. 2003b; Kim et al. 2006; Nishimura-Akiyoshi et al. 2007). NGL1 binding to netrin-G1 requires the first four LRR domains and the N-terminal domains, but does not require the Ig domain (Lin et al. 2003b). NLG3 binds to leucocyte common antigen-related (LAR) molecule s (Woo et al. 2009), PTPσ, and PTPδ (Kwon et al. 2010), but does not bind to netrin-G1 or netrin-G2 (Kim et al. 2006). NGL3 binding is enabled by interactions between an NGL3 unique glutamine (Q96) in the first LRR domain and the first two FNIII domains of LAR, PTPσ, and PTPδ (Kwon et al. 2010).

3.2.5 NGLs at the Synapse

All NLG isoforms induce clustering of presynaptic proteins when expressed in nonneuronal cells in artificial synapse formation assays (Woo et al. 2009). Specifically, the NLG2/netrin-G2 complex contributes exclusively to excitatory, but not inhibitory synapse formation by inducing the clustering of functional presynaptic machinery and postsynaptic components such as NMDARs and scaffolding proteins, PSD-95 and Shank (Kim et al. 2006). Likewise, NLG3/LAR complex also contributes specifically to excitatory synapse formation. In addition to NMDARs and scaffolding proteins, NLG3 promotes the clustering of AMPARs, a property which may contribute to its greater apparent synaptogenic efficiency as compared to NLG1 and NLG2 (Woo et al. 2009). Notably, the NGL3/PTPσ complex induces the clustering of postsynaptic components (i.e., PSD-95 ), but the NGL3/PTPδ complex does not (Kwon et al. 2010). This suggests that upon interacting with PTPδ, NGL3 may alter its intracellular binding properties. Moreover, given that PTPδ and PTPσ also engage different postsynaptic ligands to induce inhibitory and excitatory synapse formation (Takahashi et al. 2011, 2012), combinatorial expression of these molecules during synapse formation or plasticity may tightly regulate the organization of postsynaptic density components to impact synaptic strength.

A feature of NGLs distinct from other synaptic CAMs discussed thus far is its distribution to specific sub-compartments of neuronal dendrites where they have been suggested to contribute to the functional compartmentalization of neural circuit signaling (Nishimura-Akiyoshi et al. 2007; DeNardo et al. 2012; Soto et al. 2013). For example, in the hippocampal CA1 region, NGL1 distributes to dendritic segments within the stratum lacunosum moleculare (SLM) where it encounters netrin-G1 positive axons originating from the entorhinal cortex. NGL2, on the other hand, distributes to dendritic segments in the stratum radiatum (SR) and stratum oriens (SO) of CA1 where it encounters netrin-G2-positive axons originating from the CA3 region (Nishimura-Akiyoshi et al. 2007). Both the extracellular LRR domain and the PDZ-binding domain are essential for this subcellular compartmentalization. Moreover, this differential distribution of NGLs has functional consequences: knocking out NGL2 expression selectively decreases the number of excitatory synapses within the SR but not the SLM, leading to altered neurotransmission and impaired dendritic integration of SR and SLM inputs (DeNardo et al. 2012).

3.3 Integrins

Integrins are allosteric, Ca²⁺-dependent, heterodimeric cell adhesion molecules that participate primarily in cell-to-ECM but also in cell-to-cell adhesion (Hynes 2002). Major ligands include fibronectins, vitronectins, collagens, laminins, and IgSF members like L1 (Ruppert et al. 1995; Montgomery et al. 1996; Felding-Habermann et al. 1997; Yip et al. 1998; Hynes 2002). In vertebrates, integrin heterodimers are composed of 1 of 18 α and 1 of 8 β subunits, forming up to 24 distinct integrin αβ pairs (Luo et al. 2007). Two major integrin subgroupings are based on a common β1 subunit (binds αV, α1–11 subunits) and a common αV subunit (binds β1, 3, 5, 6, 8 subunits). Integrins indirectly link to the actin cytoskeleton and are capable of initiating diverse signaling events involving “outside-in” signaling (from outside of the cell to control the intracellular environment) and “inside-out” signaling (from the inside of the cell to adjust the extracellular ligand-binding state) (Luo et al. 2007). The ability to mediate such dynamic signaling is reflected in the modular and flexible structure of the α and β subunits that allows the molecule to adopt a wide range of conformational states depending on the cellular context. The complex signaling capacity makes the integrins an effective and highly versatile CAM at the synapse.

3.3.1 Integrin Structure and Binding

Both α- and β-integrins are type I transmembrane glycoproteins with a globular extracellular head domain whose α and β subunit interface in the heterodimer supports ligand binding. Also present in the extracellular domain is a “leg” or stalk region that is essential for integrin activation, which is followed by a single-pass transmembrane domain and a short cytoplasmic domain (Fig. 17.6a) (Nermut et al. 1988; Xiong et al. 2001; Luo et al. 2007).

Fig. 17.6

The integrins. (a) Domain organization of α- and β-integrins. PSI plexin/semaphorin/integrin domain. (b) Illustration of conformational changes in the structure and interaction of α- and β-integrins upon activation by extracellular or intracellular ligands based on crystal structures described in Kim et al. (2003). The extended open conformation presents a ligand-binding site near the interface region of α- and β-integrin

The extracellular globular head domain of α-integrin s consists of a seven-segment β-propeller structure containing multiple Ca²⁺-binding sites (Xiong et al. 2001, 2002; Luo et al. 2007). The “leg” region contains three β-sandwich domains that are subdivided into a “thigh” domain and two “calf” domains (calf-1 and calf-2). An additional Ca²⁺-binding loop separates the thigh region from calf-1 and calf-2 and is termed the “genu” because it is a key pivot point during integrin activation (Fig. 17.6b). The head domain of β-integrin s is composed of a βI domain that participates in α subunit binding and is flanked by a hybrid domain that forms the upper part of the β “leg” region. The hybrid domain itself is positioned within a plexin/semaphorin/integrin (PSI) domain that contributes to the stability of the head region. The remainder of the leg region is composed of four EGF-like domains followed by a β-tail domain. When inactive, the leg of β subunit is thought to bend at a region between EGF domains 1 and 2.

The conformation of an inactive integrin αβ heterodimer is heavily bent with the head region drooping toward the membrane (Xiong et al. 2001). In this conformation, the extracellular domain forms an interface with the lower legs (Takagi et al. 2002), and the molecule is not effectively positioned for ligand binding. However, upon ligand binding (Takagi et al. 2002; Xiao et al. 2004) or cytoplasmic activation (Calderwood et al. 1999; Tsuboi 2002; Vinogradova et al. 2002), the αβ heterodimer adopts an active conformation where the head region is extended away from the membrane (Xiao et al. 2004; Luo et al. 2007). The extended conformation can exist in one of two orientations: an open headpiece conformation that presents the ligand-binding region and a closed headpiece conformation where the ligand-binding region is more obscured (Fig. 17.6b) (Luo et al. 2007). The difference between the two is the relative positions of the subunit legs, transmembrane domains, and intracellular domains that become separated in the open conformation (Kim et al. 2003) and promote the swing-out of the hybrid domain of the β subunit.

3.3.2 Integrins at the Synapse

Integrin subtypes are present throughout the brain in a region-specific manner (Pinkstaff et al. 1999). Different αβ pairs localize to specific domains within the hippocampus (McGeachie et al. 2011), where they are enriched at synapses (Shi and Ethell 2006) and participate in distinct functions in synapse maturation and plasticity (Chavis and Westbrook 2001; Kramár et al. 2003; Bernard-Trifilo et al. 2005; Chan et al. 2006; Cingolani et al. 2008; McGeachie et al. 2012). Two of the major integrin subtypes in the nervous system that participate in synaptic function are composed of either β1 or β3 subunits and recognize ligands containing the Arg-Gly-Asp (RGD) sequence present in ECM-related molecules such as fibronectin (Hynes 2002).

Many of the studies that have addressed the contributions of integrins to synaptic regulation have used the infusion of peptides containing an RGD sequence to interfere with the ECM interaction of integrins. Notably, these studies have revealed acute effects on AMPA (Cingolani et al. 2008) and NMDA (Lin et al. 2003a; Bernard-Trifilo et al. 2005) receptor currents, although the latter may have involved direct effect of peptides on the glycine-binding site of NMDAR (Cingolani et al. 2008). In cultured hippocampal neurons, RGD peptide application rapidly reduces the surface abundance of AMPARs, accompanied by a concomitant decrease in quantal amplitudes and evoked synaptic transmission (Cingolani et al. 2008). The specific interaction between β3-containing integrins with the GluA2 subunit (Pozo et al. 2012) is likely involved in controlling the abundance and composition of synaptic AMPARs. Whereas the β3-integrin modulation of synaptic AMPARs occurs independently of actin, RGD peptide application over a longer duration elicits changes in synaptic actin organization and dendrite morphology, where F-actin within spine heads is reduced and dendritic spines become elongated (Shi and Ethell 2006). Such effects on spine morphology could be mediated by β1-integrin (Shi and Ethell 2006), which plays a role in LTP stability (see below).

Chronic application of RGD peptides during synapse maturation in cultured hippocampal neurons arrests synaptic development in a state characterized by high p_r and the prevalence of NR2B-containing NMDARs. These changes can be reproduced if RGD peptides are replaced with function-blocking antibodies directed toward β3 but not β1-containing integrins (Chavis and Westbrook 2001). Likewise, glycinergic synapses of spinal cord neurons also show a differential effect of integrin subtypes on synaptic strength. While application of RGD peptides increases overall glycine receptor surface levels, this effect is the sum of very different influences of β1- and β3-containing integrins: function-blocking antibodies against β1-integrins decrease, while those against β3-integrins increase surface glycine receptor abundance (Charrier et al. 2010). These findings reveal a complex interplay between integrin subtypes in regulating basal synaptic activity.

Several integrins play a role in long-term synaptic plasticity (McGeachie et al. 2011). In particular, infusing RGD peptides or anti-integrin antibodies before or shortly after LTP induction disrupts the stability of LTP in the hippocampus (Staubli et al. 1990; Bahr et al. 1997; Stäubli et al. 1998). Studies using conditional β1-integrin knockout mice support the role for β1-integrins in the stabilization but not the induction of LTP (Chan et al. 2006; Huang et al. 2006; McGeachie et al. 2012). Interestingly, β3-containing integrins do not seem to influence LTP (McGeachie et al. 2012).

In contrast to LTP, homeostatic synaptic plasticity seems particularly dependent on β3- rather than β1-integrins (Cingolani et al. 2008). Eliciting compensatory, homeostatic changes in synaptic AMPAR levels by chronically altering network activity also induces a parallel change in surface β3-integrin levels without a change in β1-integrin. The requirement for β3-integrin in homeostatic plasticity is supported by observations that β3-integrin-deficient neurons do not show a compensatory increase in synaptic AMPARs upon chronic activity blockade. Together, these findings highlight the differential roles for integrin subtypes in distinct forms of synaptic plasticity: whereas the β1-integrins regulate Hebbian synaptic plasticity, β3-integrins are involved in homeostatic synaptic plasticity. Much remains to be explored on the molecular mechanisms by which β1- and β3-integrins exert differential control over the two seemingly opposing synaptic processes that nonetheless target the AMPARs, as well as uncovering the roles for other integrin subtypes in synaptic regulation.

4 Outlook

We have discussed how synapse formation and synaptic strength can be dynamically controlled by the action of a variety of synaptic CAMs. Apart from their traditional role in conferring the stability of cell–cell junctions – i.e., at synapses – we have shown how different CAMs organize the pre- and the postsynaptic assemblies to affect the efficacy of neurotransmitter release and the availability of functional neurotransmitter receptors and participate in synaptic signaling events. The degree to which particular CAMs are able to control these functions depends on the binding interactions in which they are engaged, extracellularly and intracellularly. The binding activity of CAMs, in turn, is finely tuned by the presence or absence of additional ligands, their glycosylation state, activity of proteases that act on CAMs, and intracellular signaling events, just to name a few. Despite the enormous complexity of CAMs at the synapse, the rules that govern specific CAM interaction modes are slowly being revealed. The difficult task ahead will be to sort out how different synaptic CAMs cooperate with each other in time and space at individual synapses and across synapses. In addition, some very basic questions still remain unknown, such as the minimal number of CAM complexes needed at each synapse to effect a particular adhesive or regulatory function, or their spatiotemporal dynamics in relation to key pre- and postsynaptic components such as the readily releasable vesicles and the number of postsynaptic receptors. In this respect, recent developments in super-resolution imaging are highly promising in being able to tackle detailed arrangements of CAMs in synapses undergoing plastic changes. The strong implications for synaptic CAMs in neurological diseases should fuel further research efforts on this fascinating topic.