The Cellular and Molecular Landscape of Neuroligins

Conservation among subtypes

Neuroligins are a class of adhesion molecules found at postsynaptic sites in neurons (Nguyen and Sudhof, 1997). The four major members of the class are neuroligins (NLGNs) 1, 2, 3, and 4, which each share gross structural similarity and substantial conservation in amino acid sequence (Fig 1A). Each neuroligin has a single transmembrane region separating an acetylcholinesterase-like domain in the extracellular space from a short cytoplasmic tail (Choi et al., 2011, Nguyen and Sudhof, 1997, Sudhof, 2008). The extracellular domains of NLGNs1-3 contain two notable alternative splice sites that can affect the specificity of trans-synaptic adhesive interactions (Chih et al., 2006, Koehnke et al., 2010) with repercussions for cellular function (Lee et al., 2010, Shipman and Nicoll, 2012b). However, the extracellular domain is otherwise more conserved than the intracellular domain (Fig 1B), which will be the primary focus of the molecular aspects in this review. There is an additional fifth gene in humans – NLGN4Y (occasionally referred to as NLGN5) – found on the Y chromosome that is greater than 97% identical in protein sequence to NLGN4X, on the X chromosome (Fig 1C). Due to the high level of molecular similarity, this review will discuss NLGN4X and NLGN4Y together as NLGN4.

Neuroligins are situated within a larger superfamily of post-synaptic adhesion molecules, including, but not limited to: cadherins (Arikkath and Reichardt, 2008), SynCAMs (Biederer et al., 2002), LRRTMs (de Wit et al., 2009, Ko et al., 2009a, Linhoff et al., 2009), GluD2 (Uemura et al., 2010), CL1 (Boucard et al., 2012), NGLs (Woo et al., 2009), and teneurins (Mosca et al., 2012). On the presynaptic side, the primary adhesion partner of neuroligins is the neurexin family, which is itself composed of many subtypes and a vast number of splice variants (Treutlein et al., 2014, Ullrich et al., 1995). This diversity of synaptic adhesion molecules suggests the possibility that a trans-synaptic molecular code could participate in the formation or ongoing function of the diverse synapses found in the brain (Fig 1D).

Subcellular Localization

All neuroligins have a conserved domain structure and are postsynaptic molecules. However, the different subtypes of neuroligin each display unique subcellular localizations. Classically, NLGN1 is localized to excitatory (glutamatergic) synapses; NLGN2 is localized to inhibitory (GABAergic) synapses; and NLGN3 is localized to both excitatory and inhibitory synapses (Fig 2A). This pattern was initially discovered via immunostaining and biochemical analysis (Budreck and Scheiffele, 2007, Graf et al., 2004, Song et al., 1999, Varoqueaux et al., 2004), and has been subsequently supported by functional experiments (Chubykin et al., 2007, Shipman et al., 2011). The odd member of the family, NLGN4, unlike NLGNs1-3, is poorly conserved from humans to mice (~58%), despite strong conservation in other species such as chicken (>93%) and opossum (>95%) among others, and the mouse NLGN4-like gene is not even fully conserved among strains (Bolliger et al., 2008). Nonetheless, a NLGN4-like protein in mice has been shown to localize both to GABAergic synapses in the CNS as well as glycinergic synapses in the retina (Hoon et al., 2011).

Outside the nervous system, NLGN2 has been found on the surface of pancreatic β cells, where it stimulates the secretion of insulin through a trans-cellular, clustering-dependent, but neurexin-independent mechanism (Suckow et al., 2012). Perhaps further study of this related, but clearly unique, system could yield insight into the most conserved core process by which neuroligin induces trans-cellular sites of release.

Dimerization

The canonical interactions between neuroligin molecules follow a pattern that aligns with their subcellular localization. The primary species of neuroligin in the cell is a dimer, if not a higher-order oligomer. The neuroligin dimer was initially inferred from similarity to the protein sequence of acetylcholinesterase (Comoletti et al., 2003), and the presence of a dimer has been supported both biochemically as well as functionally: mutations in the extracellular domain based on the acetylcholinesterase structure abolish both the biochemical dimer and the overexpression phenotype of the molecule (Dean et al., 2003). When the crystal structure of the neuroligin extracellular domain was later determined, it confirmed dimerization highly reminiscent of cholinesterases (Arac et al., 2007, Comoletti et al., 2007, Fabrichny et al., 2007, Koehnke et al., 2008). With structure in hand, function was revisited: a series of mutations that abolish dimerization were found to induce a corresponding reduction in function (Ko et al., 2009b). There is some evidence that dimerization is a necessary step in the trafficking of neuroligin (Poulopoulos et al., 2012), although this is not undisputed (Shipman and Nicoll, 2012a). Finally, using variants of neuroligin in which dimerization of neuroligin can be controlled via a small molecule, it was shown that dimerization is, indeed, necessary for neuroligin’s synaptogenic effect (Shipman and Nicoll, 2012a).

Most of the structural and functional work has focused on neuroligin homodimerization. However, biochemical studies have given us insight into the rules that govern inter-type, or heterodimerization. Using co-immunoprecipitation, NLGNs 1–3 were each shown to homodimerize, but; whereas, NLGN3 was able to heterodimerize with both NLGN1 and NLGN2, there was no apparent dimerization between NLGN1 and NLGN2 (Budreck and Scheiffele, 2007). The heterodimerization between NLGN1 and NLGN3 has been multiply confirmed (Poulopoulos et al., 2012, Shipman et al., 2011), whereas a separate investigation of endogenous neuroligin found traces of a NLGN1/2 dimer and the absence of a NLGN2/3 dimer (Poulopoulos et al., 2012). Moreover, homo- or hetero-dimerization of NLGN4 is completely untested (Fig 2B). Rigorous determination of the most prevalent complexes as well as whether there may be cell-type specific complexes is an area that deserves further experimentation.

Finally, it is not clear whether individual dimers of neuroligin or in fact higher-order oligomers are the primary species present at a synaptic site (Fig 2C–D). There is evidence that a multimer of neuroligin must cluster at least four neurexin molecules to achieve synaptogenic effects (Dean et al., 2003). Moreover, the NLGN1/neurexin1β extracellular domains can form a higher-order lattice sheet when crystalized, whereas the NLGN1/neurexin1α domains cannot (Tanaka et al., 2012). Although yet to be tested functionally, perhaps this is another mechanism by which adhesion molecules may contribute to synaptic diversity.

Binding Partners

After neuroligin isoforms were found to induce the genesis of distinct synaptic subtypes, researchers looked for the critical domains and protein interactions that are required. The first identified postsynaptic binding partner with the cytoplasmic tail (c-tail) of neuroligins was postsynaptic density protein 95 (PSD-95), a canonical constituent of glutamatergic synapses (Irie et al., 1997). The PDZ ligand is conserved among all neuroligin isoforms (Fig 3) and interacts with the third PDZ domain within PSD-95. In addition to PSD-95, neuroligins bind to other members of the membrane-associate guanylate kinase (MAGUK) family (Meyer et al., 2004), although the functional significance of these interactions remains to be determined. Interestingly, NLGN1 successfully traffics to the synapse (Dresbach et al., 2004) and NLGN3 is able to robustly potentiate synapses independent of their PDZ ligands (Shipman et al., 2011). Due to complete conservation throughout the family, it seems unlikely the PDZ ligand is the dominant factor that dictates the differential localization of neuroligin subtypes.

Similar to PSD-95, Synaptic Scaffolding Molecule (S-SCAM) was shown to bind to NLGN1 and NLGN2 via a yeast-two hybrid screen (Hirao et al., 1998, Meyer et al., 2004). NLGN1 and NLGN2 bind to two different regions within S-SCAM: a PDZ- and a WW-domain (Iida et al., 2004, Sumita et al., 2007). Researchers only tested S-SCAM binding to NLGN1 and NLGN2, but both the PDZ ligand and WW-binding domain are conserved among all neuroligin subtypes (Fig 3). Therefore, it is conceivable that all neuroligins bind to S-SCAM, similar to PSD-95 and, indeed, S-SCAM is present at both excitatory and inhibitory synapses. To rigorously test the importance of this interaction, both the PDZ- and WW-interacting sites would need to be simultaneously mutated in NLGN1 or NLGN2.

At GABAergic synapses, NLGN2 was shown to co-localize with and recruit gephyrin (an inhibitory scaffolding molecule) to nascent postsynaptic specializations (Graf et al., 2004, Varoqueaux et al., 2004) and an unbiased screen further identified gephyrin as a direct NLGN2 interactor. However, like PSD-95 and S-SCAM, the gephyrin-binding domain in NLGN2 is conserved and binds equally well to NLGN1, NLGN2, and NLGN3 (Fig 3) (Poulopoulos et al., 2009). Therefore NLGN2’s specificity to inhibitory synapses cannot solely be attributed to the gephyrin-binding sequence. Recent studies suggest phosphorylation of NLGN1 and NLGN2 may, at least in part, dictate the affinity of gephyrin to a particular neuroligin isoform (discussed further below; (Giannone et al., 2013)), providing a possible explanation for synapse specificity.

Research into collybistin may provide another clue to the localization and function of NLGN2. Collybistin is a key inhibitory synapse molecule that binds gephyrin and recruits it to the plasma membrane, where gephyrin can self-assemble and cluster critical inhibitory receptors, a prevailing model for inhibitory postsynaptic differentiation (Harvey et al., 2004, Kins et al., 2000). Remarkably, only inhibitory neuroligins (NLGN2 and mouse NLGN4) bind to collybistin’s SH3 domain, promote an open/active confirmation, and traffic gephyrin and collybistin to plasma membrane phosphoinositides (Hoon et al., 2011, Poulopoulos et al., 2009, Soykan et al., 2014). The collybistin-binding domain is unknown but has been predicted to involve the poly-proline region near the terminal end of the NLGN2 c-tail (Fig 3). This region is not conserved in NLGN1, NLGN3, (Papadopoulos and Soykan, 2011) or NLGN4 (mouse), suggestive of an independent binding site in NLGN4. A direct test of this two-step, two-domain, three-protein interaction is an area worthy of future research.

All of the early identified postsynaptic interactions with neuroligins (discussed above) involved their c-tails; however, recent studies show that their extracellular domains are also involved in cis interactions in the postsynaptic cell. For example, the ectodomains of NLGN1 and GluN1 interact, and NLGN1 binding acts as a shuttle to deliver NMDA receptors to the synapse (Budreck et al., 2013). This binding is noteworthy because a phenotype of the NLGN1^−/− is reduced NMDA currents (Blundell et al., 2010), suggesting a model where NLGN1 can directly interact with a glutamate receptor and affect synaptic strength, in contrast to the canonical model in which scaffolding molecules (such as PSD-95, S-SCAM, and gephyrin) link neuroligins to the receptors at the synapse. A different ectodomain association between NLGN2 and MDGA1 (MAM domain-containing GPI anchor protein) breaks the intercellular interaction between NLGN2 and neurexin, inhibiting NLGN2-mediated presynaptic differentiation (Lee et al., 2013, Pettem et al., 2013). The binding sites within the extracellular domains of NLGN1 and NLGN2 that are required for these interactions remain an enigma, but these studies provide evidence that neuroligin ectodomains’ responsibility is not only for dimerization and to trans-synaptically interact with neurexins.

It is likely that the library of neuroligin interacting proteins will continue to expand. Two independent screens identified many proteins that interact with NLGN2 and NLGN3 c-tails, and these interactions have yet to be characterized (Poulopoulos et al., 2009, Shen et al., 2014). Moreover, a critical region was identified in the c-tails of NLGN1 and NLGN3 that is required for these molecules to potentiate excitatory synaptic transmission (Shipman et al., 2011) without a known direct binding partner (Fig 3). Additionally, a motif in the c-tail of NLGN1 required for dendritic sorting has been uncovered (Fig 3), but the molecular basis of this trafficking pathway is unknown and likely requires unidentified interacting proteins (Rosales et al., 2005). A subset of the known molecular constituents of an excitatory and an inhibitory synapse are modeled in Figure 4, with a focus on the proteins that interact with neuroligins, providing context for the molecular discussion of neuroligin.

Post-translational Modifications

The Function of Neuroligin in Development and Plasticity

Can we extrapolate from the two decades of experimental work on neuroligin to define an endogenous function for the molecule at the synapse? Based on overexpression and knock-down both in vitro and in vivo, the global picture is that increasing neuroligin expression results in more synapses (Boucard et al., 2005, Chih et al., 2005, Dahlhaus et al., 2010, Hines et al., 2008, Shipman et al., 2011), whereas reducing neuroligin expression results in fewer synapses (Chih et al., 2005, Shipman and Nicoll, 2012b, Shipman et al., 2011). If we attempt to map this phenomenology onto our accumulated knowledge of the brain, we can broadly segregate a non-pathological change in the number of synapses into two biological categories: those that occur as a process of development; and those that occur as a process of plasticity. The crucial difference is one of initial conditions: development is a semi-determined progressive process from inchoate material and plasticity is a change from a steady-state. Molecularly, the two may have both shared and divergent properties.

In development, neuroligin appears to have a prominent role (Fig 6A). Synapses can still form following the knockout of NLGNs1-3 (Varoqueaux et al., 2006), therefore the molecules are not strictly necessary for synapse formation. Yet, at least brainstem synapses are deficient in synaptic transmission following the triple knockout, suggesting a potential role in the assembly or maturation of the postsynaptic site beyond formation. Moreover, the manipulation of neuroligin levels during synaptogenesis can grossly alter the number of synapses formed onto a given neuron in a bi-directional manner (Chih et al., 2005), and the simple presence of neuroligin can induce the formation of a functional synapse onto a non-neuronal cell (Fu et al., 2003, Scheiffele et al., 2000). This sensitivity to neuroligin during development, coupled with its localization and molecular interactions, would suggest an endogenous function at the site of the developing synapse. The lack of an absolute requirement may owe to the presence of other compensatory synaptic adhesion molecules (Tallafuss et al., 2010). An emerging model has neuroligin coming into physical proximity with neurexin and catalyzing the clustering of synaptic molecules on both sides of the nascent cleft, thus forming a physical synaptic site (Dean et al., 2003, Shipman and Nicoll, 2012a). It is important to note that this is a model of synaptogenesis that does not require any synaptic transmission, which is consistent with the evidence that synaptogenesis proceeds in the absence of pre-synaptic release of transmitter (Varoqueaux et al., 2002) or post-synaptic reception of glutamate (Lu et al., 2013).

Now let us consider the role of neuroligin during the ongoing formation of synapses that occurs after development (i.e. plasticity). Mechanistically, there is no reason to believe that the final molecular events leading to the formation of a synaptic site would differ after establishment of the baseline system – these final events are likely to involve neuroligin in an identical manner as those that occur during development. However, a more interesting question is whether neuroligin plays any part in the directed, activity-dependent reshaping of synaptic circuits that underlie learning and memory. To do so, neuroligin must possess the ability to change in response to ongoing activity in a cell in such a way that enacts directed synaptic remodeling. There exists evidence that neuroligin is, in fact, changed by ongoing activity in the cell and sensory experience in an animal – multiple sites in the cytoplasmic tail of NLGN1 are phosphorylated, in at least one case by a calcium-dependent kinase (Bemben et al., 2014, Giannone et al., 2013). Moreover, it has been shown that the c-tail phosphorylation of NLGN1 by CaMKII results in an increase in the surface trafficking of neuroligin (Bemben et al., 2014). Mutations that interfere with surface trafficking of neuroligin abolish its synaptogenic properties (Comoletti et al., 2004, Ko et al., 2009b), thus it is specifically the surface expression of neuroligin that must influence synapse density. The CaMKII-phosphorylation-dependent trafficking mechanism provides a direct link from calcium to synaptogenesis (Fig 6B) and may explain not only the finding that pharmacological blockade of neuronal activity can diminish the enhancement of synapses by NLGN1 expression (Bemben et al., 2014, Chubykin et al., 2007), but even the finding that this pharmacological reduction can be overcome by stronger overexpression (Shipman et al., 2011).

So, calcium can influence the density of synapses onto a cell through neuroligin, but what sort of plasticity is this? Is neuroligin a player in long-term plasticity (LTP) directly? Although knock-out and knock-down of NLGN1 has been shown to reduce the expression of LTP (Blundell et al., 2010, Jedlicka et al., 2013, Shipman and Nicoll, 2012b), several lines of evidence would argue against a direct role. First, LTP is defined as the specific recruitment of new AMPARs to synapses (be it through un-silencing or increases in synaptic strength) (Kerchner and Nicoll, 2008), whereas manipulations of NLGN result in changes to the number of whole synapses (including both AMPARs and NMDARs) (Shipman and Nicoll, 2012b). Furthermore, the acute loss of NLGN1 in area CA1 after development has no apparent effect on LTP (Shipman and Nicoll, 2012b), arguing instead that the molecular composition of a NLGN1-formed synapse may pre-set the conditions for LTP rather than NLGN1 directly participating in the plasticity. A different emerging role for neuroligin in synaptic plasticity is discussed in Box 2.

Concluding Remarks

The discovery of neuroligins twenty years ago unlocked new avenues of research into the molecular dynamics of the synaptic site. The research that has followed has greatly enhanced our understanding of functional synaptogenesis in the developing and mature brain. The involvement of neuroligins in cognitive disorders has enabled researchers to develop genetic models to study the etiology of these diseases. Yet, despite outstanding advances in the field, many key questions remain unsolved. For example, how do neuroligins localize to their respective synapses? How does a single molecule recruit the machinery to form a functional synapse? What are the functional similarities and differences among neuroligins during development and plasticity? The answers to these questions will provide exciting insights in synapse biology for decades to come.