Neuroligin-2 as a central organizer of inhibitory synapses in health and disease

See allHide authors and affiliations

Science Signaling  22 Dec 2020:
Vol. 13, Issue 663, eabd8379
DOI: 10.1126/scisignal.abd8379


Postsynaptic organizational protein complexes play central roles both in orchestrating synapse formation and in defining the functional properties of synaptic transmission that together shape the flow of information through neuronal networks. A key component of these organizational protein complexes is the family of synaptic adhesion proteins called neuroligins. Neuroligins form transsynaptic bridges with presynaptic neurexins to regulate various aspects of excitatory and inhibitory synaptic transmission. Neuroligin-2 (NLGN2) is the only member that acts exclusively at GABAergic inhibitory synapses. Altered expression and mutations in NLGN2 and several of its interacting partners are linked to cognitive and psychiatric disorders, including schizophrenia, autism, and anxiety. Research on NLGN2 has fundamentally shaped our understanding of the molecular architecture of inhibitory synapses. Here, we discuss the current knowledge on the molecular and cellular functions of mammalian NLGN2 and its role in the neuronal circuitry that regulates behavior in rodents and humans.


For information to travel throughout networks in the brain, it is transmitted from neuron to neuron through chemical synapses, highly specialized subcellular compartments that not only relay but also fundamentally shape the information being transmitted. Accordingly, the molecular machinery that constitutes these subcellular structures plays a key role in determining information flow throughout the network, and alterations in these molecules can contribute substantially to the emergence of psychiatric disorders. In addition to the machinery that releases neurotransmitters and the neurotransmitter receptors that receive them, all synapses contain a large number of organizational proteins that form intracellular scaffolds or transsynaptic bridges that connect the pre- and postsynaptic terminals. Prominent representatives of the latter group are the neuroligin family of postsynaptic adhesion proteins, which bind to presynaptic neurexins to regulate synapse formation and function and which have been the focus of extensive investigations due to their role in disorders such as autism and schizophrenia (15).

The neuroligin protein family consists of five members in humans (NLGN1, 2, 3, 4X, and 4Y) and four in mice (NLGN1, 2, 3, and 4), each of which regulates synaptic function at a distinct albeit partially overlapping subset of synapses (13). A key aspect of their functional specificity arises from their differential effects at the two primary types of synapses in forebrain networks, namely, excitatory synapses (which use the neurotransmitter glutamate) and inhibitory synapses [which use γ-aminobutyric acid (GABA) or glycine]. In particular, NLGN1 was found to be a key component specifically of excitatory synapses (6, 7), whereas NLGN2, first identified in 1996 (8), is unique in being the only family member found to localize exclusively to inhibitory synapses (7, 9). NLGN3 shows a more complex pattern of selectivity, with effects specific to either excitatory or inhibitory synapses in a context-dependent manner (5, 1015). Arguably, the most enigmatic neuroligin in this respect, however, is NLGN4, a poorly evolutionarily conserved family member (16) that was originally linked to inhibitory synapse function in mice (1719) but was subsequently found to regulate excitatory synapses in humans (5, 20, 21).

Because of their importance as synapse organizers, research on all neuroligin family members has fundamentally shaped our knowledge of synapse structure and function, but perhaps none more so than NLGN2 at inhibitory synapses. Inhibitory synapses contain a molecular machinery that is entirely distinct from that at excitatory synapses, and the enormous complexity and diversity of this machinery are only now becoming appreciated (22). As one of the prototypical organizer proteins at inhibitory synapses, NLGN2 has played a pivotal role in driving much of our understanding of the molecular architecture of these synapses in health and disease. In this Review, we summarize the current state of knowledge on the molecular and cellular functions of NLGN2 at inhibitory synapses and beyond, as well as its role in the networks and circuits that underlie psychiatrically relevant behaviors in mice and humans.


Like all neuroligin family members, NLGN2 is a type I transmembrane protein with a large extracellular domain that mediates transsynaptic binding to presynaptic neurexins as well as neuroligin dimerization, a single-pass α-helical transmembrane domain, and an intracellular domain, which acts as a hub for neurotransmitter receptor recruitment and downstream signaling pathways (Fig. 1A) (8).

Fig. 1 Structure and function of NLGN2 at inhibitory synapses.

(A) NLGN2 dimers form transsynaptic interactions with presynaptic neurexins (NRXN), as well as postsynaptic interactions with Cb and gephyrin, which, in turn, recruit GABAARs to the synapse. Other interaction partners include the scaffolding protein S-SCAM, the peptidyl-prolyl isomerase PIN1, and the endosome-cargo adaptor SNX27. (B) Schematic representation of the splice sites that determine recruitment of neuroligins to glutamatergic versus GABAergic synapses. The figure is adapted from (1). B(+) and B(−), splice site B in the neuroligins with and without the splice insert (depicted as a light orange triangle). NLGN2 only exists in the B(−) form. SS4(+) and SS4(−), splice site 4 in the neurexins with and without the splice insert (depicted as a half circle). Neurexins containing the SS4(+) splice insert bind with high affinity to neuroligin B(−) and are more likely to recruit NLGN2 to GABAergic synapses. (C) Proposed synapse specificity of NLGN2 function at different inhibitory interneuron subtypes in the cortex and hippocampus. VIP, vasoactive intestinal peptide; CCK, cholecystokinin. NLGN2 is thought to primarily function at perisomatic synapses formed by PV-positive basket cells. (D) Schematic representation of the mode of interaction of the five proposed interaction partners of NLGN2. NLGN2 is represented in dark red, and α-neurexin is represented in light blue (depicted schematically for reasons of space). All interaction partners are labeled. Credit: Adapted by A. Kitterman/Science Signaling from D. Krueger-Burg/Max Planck Institute of Experimental Medicine and University Medical Center Göttingen.

Extracellular interactions

The N-terminal extracellular domain of the neuroligins bears substantial homology to the cholinesterase family of esterases, albeit without an intact and active catalytic site. Instead, a crystal structure of the extracellular domain of NLGN2 reveals that it contains two primary interaction sites located on opposite sides of the molecule: a neurexin binding site and a neuroligin dimerization interface (23).

The neurexin binding site regulates interaction with the only known presynaptic interaction partner of NLGN2, the neurexin family of presynaptic adhesion molecules. This interaction is regulated by a highly complex code of alternative splicing, which governs the differential localization of the neuroligins to glutamatergic versus GABAergic synapses (1, 3, 24). Neurexins are expressed as a long α-isoform and a short β-isoform that each contain several alternative splice sites, including the splice site SS4, which is localized at the binding site with neuroligins (Fig. 1B). α-Neurexins and SS4(+)-neurexins show a stronger association with NLGN2 and a greater localization to GABAergic synapses than β-neurexins and SS4(−)-neurexins (6, 25, 26). Similarly, neuroligin family members contain two splice sites, A and B, which have differential effects on localization to GABAergic synapses. Inclusion of splice site A and exclusion of splice site B (which is never present in NLGN2) promote localization to GABAergic synapses (6, 27). Importantly, it appears that these splice form-specific extracellular interactions with neurexins are the primary determinants of the specific localization of NLGN2 to inhibitory synapses and that interactions of the intracellular domain (described in detail below) are necessary for the subsequent recruitment of GABAergic postsynaptic components but not for the initial NLGN2 localization (2730).

In addition to neurexin binding, the extracellular domain of all neuroligin family members is also responsible for their dimerization at the dimerization interface, which is a prerequisite for their function. This dimerization effectively clusters the associated neurexin monomers and thereby regulates the assembly and function of the appropriate presynaptic machinery. Interestingly, the dimerization interface is highly conserved among all neuroligin family members, in principle permitting the formation of hetero- and homodimers (23). In practice, however, it has been shown that NLGN2 forms primarily homodimers, although traces of NLGN1/2 (but not NLGN2/3) heterodimers were also detected, the functional relevance of which is currently unclear (31).

Intracellular interactions

In addition to the large extracellular domain, NLGN2 contains a shorter C-terminal intracellular domain, which mediates interactions with the postsynaptic complex at GABAergic synapses. This complex contains GABAA receptors (GABAARs) that mediate fast inhibitory synaptic transmission through their function as ligand-gated chloride channels, as well as organizer proteins such as the scaffolding protein gephyrin and others that anchor GABAARs at synapses and regulate GABAergic transmission (22). NLGN2 was shown to play a central role in organizing this complex by recruiting gephyrin and GABAARs to synaptic clusters in a process that depends on the guanosine diphosphate/guanosine triphosphate exchange factor collybistin (Cb; encoded by the gene ARHGEF9) (32). In the model arising from these studies, NLGN2 accumulates across from GABAergic presynaptic terminals because of its interactions with neurexins, where it then forms a tripartite complex with Cb and gephyrin (Fig. 1A). In this complex, NLGN2 binds to gephyrin through a motif containing an essential tyrosine (Tyr770 in rat NLGN2) and to Cb through a proline-rich motif, thereby stabilizing an open, active conformation of Cb and leading to the local assembly of a gephyrin scaffold and the recruitment of GABAARs (3234).

In addition to the gephyrin and Cb binding motifs, the intracellular domain of NLGN2 also contains a PDZ binding motif that can bind to PDZ domain–containing proteins, such as the scaffolding protein synaptic scaffolding molecule (S-SCAM), also known as membrane-associated guanylate kinase inverted 2 (MAGI-2) (35). Moreover, NLGN2 can be phosphorylated at a conserved serine (Ser714 in the mouse) in a proline-directed manner, leading to recruitment of the peptidyl-prolyl isomerase PIN1, which causes a conformational change in NLGN2 and disrupts its ability to bind to gephyrin (36). It has been proposed that this signaling pathway represents a mechanism for NLGN2 function at inhibitory synapses that is parallel to and independent from the NLGN2/Cb/gephyrin tripartite complex and that disruption of both pathways is necessary to completely abolish the effect of NLGN2 on inhibitory synapse function (27). NLGN2 also contains several conserved sites that were shown to be phosphorylated in other Neuroligin family members, including Tyr770, but their relevance in NLGN2 remains to be assessed (2, 36).

Last, the intracellular domain of NLGN2 is also involved in its trafficking and particularly in the endocytosis of NLGN2 in response to ligand binding (37). This process is reportedly regulated by direct interaction of NLGN2 with the endosome-cargo adaptor sorting nexin 27 (SNX27), which prevents lysosomal degradation and promotes recycling of NLGN2 to the inhibitory synapse, thereby strengthening inhibitory synaptic transmission and NLGN2-containing synapses (38, 39).


Early evidence that NLGN2 may play a central role in the development and function specifically of inhibitory synapses was derived from experiments in neuronal cultures and transgenic mice, which demonstrated that overexpression of NLGN2 selectively enhances the function of inhibitory but not excitatory synapses (4042), and from studies in heterologous human embryonic kidney cells showing that NLGN2 expression results in clustering of functional GABAAR (42, 43). Since then, a range of studies have used constitutive or conditional Nlgn2 knockout (KO) mice to investigate the role of endogenous NLGN2 in inhibitory synapse function across multiple brain regions (summarized in Table 1). Whereas the detailed consequences of Nlgn2 deletion appear to differ subtly between brain regions, a number of common themes emerge from these studies, as discussed below.

Table 1 Consequences of Nlgn2 deletion or overexpression at inhibitory synapses.

The table displays changes in GABAergic and glycinergic markers and synaptic transmission observed in the various brain regions in response to loss or gain of NLGN2 in mice, as reported in the studies cited. N/A, not assessed. ↑, increase; ↓, decrease; ↔, no change; OE, overexpression; cNlgn2, conditional Nlgn2 deletion; freq, frequency; amp, amplitude; presyn, presynaptic; postsyn, postsynaptic; S.P., stratum pyramidale; S.R., stratum radiatum; GCL, granule cell layer; ML, molecular layer; CB1, cannabinoid receptor 1; GAD65, glutamate decarboxylase 65; VGLUT3, vesicular glutamate transporter 3; VIAAT, vesicular inhibitory amino acid transporter; Gly, glycine; GlyR, glycine receptor; PV-pos, PV-positive; SOM-pos, SOM-positive; PPI, paired-pulse inhibition; LTP, long-term potentiation; mIPSCs, miniature inhibitory postsynaptic currents; miRNA, microRNA.

View this table:

Effects on synaptic protein composition

One of the most consistent findings across virtually all brain regions is that protein components of the GABAergic postsynaptic complex (particularly the GABAARs and the scaffolding protein gephyrin), but not the GABAergic presynapse [particularly the vesicular inhibitory amino acid transporter (VIAAT)], are reduced in Nlgn2 KO mice, indicating an alteration in the composition of the postsynapse but not in the total number of inhibitory synapses. Interestingly, there is some variability in the extent to which the different postsynaptic protein components are affected in different regions (detailed in Table 1), particularly with respect to the different subunits of the pentameric GABAAR, which each play distinct roles in receptor function (44). It is currently unclear whether this discrepancy represents a genuine regional difference in NLGN2 function or a consequence of technical distinctions between studies. In hippocampal area CA1, deletion of Nlgn2 additionally leads to a marked increase in intracellular gephyrin clusters, likely reflecting a disruption of the normal recruitment of gephyrin to postsynaptic sites (32).

Effects on inhibitory synaptic transmission

Consistent with this reduction in postsynaptic components, GABAergic and glycinergic inhibitory synaptic transmission, as assessed in the form of miniature or spontaneous inhibitory postsynaptic currents (mIPSCs or sIPSCs), is markedly reduced in Nlgn2 KO mice in the vast majority of regions investigated (see Table 1). Notable exceptions appear to be the central amygdala (CeA) and the thalamic reticular nucleus (nRT), the only regions reported to date to have near-normal inhibitory synaptic transmission in the absence of NLGN2 (4547). In most regions, Nlgn2 deletion results in a decrease in both mIPSC amplitude, which may reflect a reduction in the number or function of postsynaptic GABAARs, and mIPSC frequency, which is most often considered to reflect a change in presynaptic GABA release. Given that NLGN2 is a postsynaptic protein and its deletion affects primarily components of the postsynaptic machinery but not the presynaptic terminal, the most parsimonious mechanistic explanation for this observation is that mIPSCs with very small amplitudes can no longer be detected reliably, resulting in an apparent decrease in frequency. Nevertheless, it is also conceivable that transsynaptic signaling mediated by neurexin may directly affect presynaptic GABA release and, hence, mIPSC frequency. The relative contribution of these two potential mechanisms, which need not be mutually exclusive, remains to be assessed. Again, there appears to be substantial variability in the relative importance of changes in mIPSC amplitude and frequency (see Table 1), reflecting either technical aspects of the individual studies or unexplored regional or cell type–specific differences in the underlying molecular mechanisms.

It is also important to note that NLGN2 not only is required during development to initially establish appropriate inhibitory synapse function but also appears to play an ongoing role during adulthood in synapse maintenance. This is evidenced by the observation that virally mediated deletion of Nlgn2 in the medial prefrontal cortex (mPFC) (48) or lateral septum (LS) (49) after completion of synaptogenesis results in reduced inhibitory synaptic transmission that is similar to that observed in constitutive Nlgn2 KO mice.

Synapse specificity of NLGN2 effects

A particularly intriguing finding in several forebrain regions is the observation that constitutive deletion of Nlgn2 appears to affect only a subset of inhibitory synapses onto a given postsynaptic neuron, despite the presence of NLGN2 at virtually all inhibitory synapses. Specifically, these include perisomatic synapses formed by fast-spiking interneurons, likely parvalbumin (PV)–positive basket cells (Fig. 1C). In the hippocampus and basolateral amygdala (BLA) of Nlgn2 KO mice, gephyrin and GABAAR puncta are reduced specifically in the perisomatic but not the dendritic region of neurons (32, 45, 50). Moreover, in the somatosensory cortex, Nlgn2 KO mice show a decreased amplitude of IPSCs originating from fast-spiking and PV-positive interneurons [which are thought to target primarily the perisomatic region (51)] but not from somatostatin (SOM)–positive interneurons (which target the dendritic region) (52). It should be noted, however, that microRNA-mediated knockdown (KD) of Nlgn2 expression in organotypic hippocampal slices has been reported to reduce light-evoked IPSCs from both PV- and SOM-positive interneurons (53). It remains to be determined whether this discrepancy results from regional differences in synapse specificity, from differences between constitutive Nlgn2 KO and acute Nlgn2 KD, or from other as yet unknown factors. Either way, the mechanisms governing synapse specificity of NLGN2 function are largely unknown, and it is likely that they are far more complex than is currently appreciated. Of particular interest in this context is the notion that NLGN2 function may be modulated by interaction partners in a synapse subtype–specific manner (discussed further in the next section), highlighting the importance of understanding the rules that govern the diversity of organizer complexes at inhibitory synapses (22).


Although NLGN2 is undoubtedly one of the most central organizer proteins at inhibitory synapses, it is by far not the only one, and growing evidence indicates that it is the synapse-specific combination of multiple organizer proteins and their mutual interactions that confer distinct properties to individual inhibitory synapse subtypes (22). Accordingly, deciphering the code by which NLGN2 interacts with its various partners will be critical in fully understanding its function at inhibitory synapses (Fig. 1D).

Dystroglycan complex

One of the first postsynaptic interaction partners of NLGN2 to be identified was β-dystroglycan (β-DG), which was reported to bind both to neurexins (54) and to NLGN2 by means of the synaptic scaffolding protein S-SCAM (35). On the basis of the preferential localization of β-DG to perisomatic synapses, it was originally proposed that the interaction between β-DG and NLGN2 may explain why Nlgn2 deletion affects primarily perisomatic synapses despite being present at all inhibitory synapses (29). However, subsequent studies provided no evidence for a functional interaction of β-DG and NLGN2 at perisomatic synapses in the hippocampus (55), and thus, further research will be needed to clarify the significance of this complex.

Immunoglobulin superfamily member 9b

Another transmembrane cell adhesion protein that, like β-DG, was originally proposed to interact with NLGN2 through interaction with S-SCAM is the immunoglobulin superfamily member 9b (IgSF9b) (56). More recently, it was shown that although NLGN2 and IgSF9b do share some overlap in localization, their effects at inhibitory synapses are very distinct: Unlike NLGN2, whose deletion impairs inhibitory synaptic transmission throughout the brain, the only study investigating the function of IgSF9b in vivo to date found that deletion of Igsf9b, in fact, increased inhibitory synaptic transmission in the CeA (but not in the BLA), likely by increasing the number of presynaptic GABAergic vesicles (46). These findings indicate that although it is still unclear whether NLGN2 and IgSF9b act in a complex structurally in vivo, their functions appear to be largely distinct and independent, at least in the amygdala.

Membrane-associated mucin domain–containing glycosylphosphatidylinositol anchor 1/2

A different model of interaction has been proposed for the membrane-associated mucin domain–containing glycosylphosphatidylinositol anchor (MDGA) proteins, which are located at the synaptic cleft and are attached to the postsynaptic membrane via a glycosylphosphatidylinositol anchor (57). There are two family members, MDGA1 and MDGA2, both of which have been proposed to inhibit the function of neuroligins by binding to their extracellular domain and, thus, sterically preventing the transsynaptic interaction between neuroligins and neurexins (5861). To date, however, it is not yet clear which of the two MDGA family members plays a greater role in specifically regulating the function of NLGN2 at inhibitory synapses. Proteomic analysis of synaptic clefts in cultured cortical neurons using proximity labeling indicated that MDGA1 and MDGA2 were primarily present in the excitatory and inhibitory synaptic cleft, respectively (62), implying that MDGA2 would be the most likely interaction partner for NLGN2. In contrast, analysis of Mdga1 KO and Mdga2 heterozygous mice (Mdga2 KO being lethal) revealed specific increases in inhibitory and excitatory synaptic transmission, respectively, in hippocampal sections (63, 64), implicating MDGA1 as the primary regulator of NLGN2 at GABAergic synapses. Whether these apparently contradictory findings result from differences in brain region, gene dosage, or technical aspects remains to be determined. Direct analysis of the functional consequences of the NLGN2-MDGA interaction using either double deletion or binding-deficient mutants will be necessary to clarify this question.


A clearer mode of interaction was recently demonstrated for Slit and tropomyosin receptor kinase-like family, member 3 (SLITRK3), which bind to presynaptic protein tyrosine phosphatases to regulate the formation and maturation of synapses. SLITRK3 was the first and, to date, only SLITRK family member found to be specifically involved in the development of inhibitory synapses (65). Consistent with this role, SLITRK3 was shown to interact directly with NLGN2 via its extracellular domain, and this interaction was required for inhibitory synapse development in mature but not in immature hippocampal neurons (66). Accordingly, disruption of the NLGN2-SLITRK3 interaction in vivo results in a reduction in inhibitory synaptic transmission, a disturbance of the GABAergic hippocampal network, and increased seizure susceptibility (66).


GABAAR regulatory lipoma high-mobility group protein C (HMGIC) fusion partner–like 4 (GARLH; also known as LHFPL4) is a four membrane-spanning protein that was recently identified as an auxiliary subunit for GABAARs (67). GARLH was found to form a complex with NLGN2 and γ2 (but not δ) subunit–containing GABAARs, and this complex was proposed to be necessary for the localization of γ2 subunit–containing GABAARs to synapses but not for their surface localization or function (6769). Consistent with an essential role for this complex in determining inhibitory synapse function in specific neuronal subtypes, Garlh KO mice showed a reduction in GABAAR clustering and synaptic transmission in pyramidal neurons but not interneurons in the hippocampus (69), as well as in the granular layer but not the molecular layer of the cerebellum (70).


In addition to the well-established role of NLGN2 at inhibitory synapses, a small number of studies have provided first indications that NLGN2 may also contribute to other biological pathways within and outside of the central nervous system. These findings have yet to receive independent validation, and in most cases, the physiological role and biochemical mechanisms through which NLGN2 may act in these contexts are currently unknown. Nevertheless, when interpreting the consequences of Nlgn2 deletion on neuronal networks and behavioral outcomes in mice and humans, it is worth bearing in mind that functions of NLGN2 beyond the inhibitory synapse may exist.

NLGN2 at other synapse subtypes

Several studies have indicated that synapse subtypes other than inhibitory synapses may also contain NLGN2. Using immuno–electron microscopy, one study reported NLGN2 to be present at cholinergic synapses throughout the brain (71). Moreover, NLGN2 was also found to be present at GABAergic postsynaptic sites opposing dopaminergic presynaptic terminals in the striatum (72). Although dopamine and GABA can be co-released from the same presynaptic terminals (73), no GABA was observed in dopaminergic terminals opposing NLGN2-containing postsynapses, leading the authors to conclude that NLGN2 may provide a competitive advantage to heterologous dopaminergic synapses over traditional GABAergic synapses (72). Last, it has been reported that NLGN2 functions in a complex with the serotonin transporter SERT in brainstem dorsal raphe neurons and that Nlgn2 KO mice have alterations in serotonergic function that may contribute to their behavioral phenotype (74).

NLGN2 in astrocytes

Astrocytes reportedly express several Neuroligin family members, including NLGN2, which were proposed to interact with presynaptic neurexins to regulate presynaptic function (75). Deletion of Nlgn2 specifically from astrocytes using conditional Nlgn2 KO mice resulted in a reduction in the number of excitatory (vesicular glutamate transporter 1–positive) but not inhibitory (VIAAT-positive) synapses, as well as a reduction in excitatory synaptic transmission and an increase in inhibitory synaptic transmission (75). How this occurs mechanistically and how these findings can be reconciled with the inhibitory synapse–specific effects of the constitutive Nlgn2 KO remain to be determined.

NLGN2 outside of the central nervous system

Last, NLGN2 has also been proposed to play a role in insulin secretion from pancreatic islet β cells. Stimulation of β cells by transcellular NLGN2 interactions was reported to enhance insulin secretion (76), and targeting β cells with nanoparticles functionalized with a NLGN2-derived peptide has been proposed as a novel approach for treating diabetes (77). Moreover, NLGN2 is reportedly expressed in vascular endothelial cells, where it may regulate angiogenesis through modulation of the release of vascular regulators (78).


In light of its central role in regulating inhibitory synaptic transmission and hence the functional connectivity and excitation/inhibition (E/I) balance throughout the neural network, NLGN2 is perfectly positioned to exert a major influence on the behavioral output of the brain. Accordingly, numerous studies (summarized in Table 2) have used genetic manipulations of Nlgn2 in rodents to understand its contribution to individual behaviors and their underlying circuit mechanisms. The picture that emerges from these studies indicates that NLGN2 may play a particularly prominent role in behaviors with an aversive component, such as anxiety and fear learning, with more subtle consequences for other behaviors.

Table 2 Consequences of Nlgn2 deletion or overexpression on behavior in mice.

The table displays various behavioral and sleep changes observed in response to loss or gain of NLGN2 in mice, as reported in the studies cited. ↑, increase; ↓, decrease; ↔, no change; OE, overexpression; “c” (in cNlgn2 and cNLGN2), conditional manipulation of expression; OF, open field test; EPM, elevated plus maze; LDB, light-dark box; TST, tail suspension test; FST, forced swim test; ASR, acoustic startle response; FC, fear conditioning; 5CSRTT, five-choice serial reaction time task; REM, rapid eye movement; NREM, non-rapid eye movement; EMG, electromyography; HPC, hippocampus; NR, nucleus reuniens; USV, ultrasonic vocalizations; D1, dopamine receptor 1; D2, dopamine receptor 2.

View this table:

Nlgn2 in anxiety-like defensive behaviors

One of the most robust and extensively studied behavioral consequences of NLGN2 manipulations is alterations in anxiety-like defensive behaviors (Fig. 2). Using approach avoidance paradigms, such as the open field test (OF), the elevated plus maze (EPM), and the light-dark box (LDB), it was shown that constitutive global deletion of Nlgn2 in mice results in a pronounced increase in avoidance behaviors (Fig. 2A) that may, in their detail, depend on the strain background (4547, 7981). Interestingly, moderate overexpression of NLGN2 resulted in an anxiety-like phenotype that very much resembled that of the constitutive KO, indicating that shifting the E/I balance at NLGN2-containing synapses in either direction may exacerbate these behaviors (41). Analysis of the neural circuits underlying these alterations in defensive behaviors using adeno-associated viral (AAV) vector-mediated local deletion of Nlgn2, immediate-early gene assays, retrograde tracing, and in vivo electrophysiology revealed important contributions of several brain regions that are known to play a role in anxiety processing (Fig. 2B). In particular, constitutive deletion of NLGN2 resulted in an exaggerated activation of the BLA and its projections to the CeA under aversive conditions, consistent with the notion that the reduced synaptic inhibition results in neuronal overactivation in key regions mediating affective behaviors (45, 46). Accordingly, local Cre-mediated deletion of NLGN2 from the BLA recapitulated some (but not all) aspects of the anxiety phenotype of the constitutive Nlgn2 KO mouse (80). Similarly, loss of NLGN2 from the ventral hippocampus (vHPC), another key region in the anxiety circuitry, resulted in an increase in freezing behaviors under specific anxiogenic conditions (80). In contrast, local deletion of Nlgn2 from the mPFC (48) or the LS (49) led to a pronounced reduction in anxiety-like behaviors, likely reflecting a loss of inhibition onto anxiolytic projections from the mPFC to subcortical structures (48) or from the LS to the lateral hypothalamus (49). Intriguingly, multisite local field potential (LFP) recordings from constitutive Nlgn2 KO mice during exploration of an OF arena indicated that alterations in the LFP coherence in the vHPC-mPFC-BLA circuit more strongly predict defensive behaviors than alterations in LFP power within a given brain region, indicating that a key role of NLGN2-mediated synaptic inhibition within the anxiety network is to regulate long-range connectivity between brain regions (80).

Fig. 2 Role of NLGN2 in anxiety-like behaviors in mice.

(A) Overview of the OF and EPM tasks used to assess anxiety-like behaviors. Representative track patterns of WT and Nlgn2 KO mice show that Nlgn2 KO mice spend less time in the anxiogenic center of the OF and the anxiogenic open arms of the EPM. This figure is adapted from (46). (B) Consequences of loss or gain of Nlgn2 expression in anxiety-associated circuitry. Global deletion of Nlgn2 results in increased anxiety-like behaviors, as well as an exaggerated activation of the basal amygdala (BA) and centromedial amygdala (CeM) and an altered coherence in the vHPC-mPFC-BLA circuitry upon exposure to an OF. Whether the exaggerated amygdala activation or the altered coherence is causal to the anxiety phenotype remains to be determined. Local deletion of Nlgn2 results in increased (BLA and vHPC) or decreased (mPFC and LS) anxiety-like behaviors. Global overexpression results in increased anxiety-like behaviors. Credit: Adapted by A. Kitterman/Science Signaling from H. Ali/Max Planck Institute of Experimental Medicine and University of Göttingen and D. Krueger-Burg/Max Planck Institute of Experimental Medicine and University Medical Center Göttingen.

In addition to the consequences for anxiety behaviors, constitutive deletion of Nlgn2 in mice was also found to have an antidepressant-like effect in the forced swim test and the tail suspension test (74). How this relates to the increased anxiety phenotype remains to be determined.

NLGN2 in sensory functions

Unlike its consequences for anxiety behaviors, constitutive deletion of Nlgn2 appears to have only very minor effects on sensory functions. Auditory function as assessed by an acoustic startle response test was unaltered (81), and in an auditory discrimination paradigm, auditory acuity as assessed by tone generalization did not differ between wild-type (WT) and Nlgn2 KO mice (82). Similarly, olfactory function was not affected by Nlgn2 deletion (81). To our knowledge, visual acuity has not yet been assessed, although electroretinogram and multielectrode array recordings of retinal function in Nlgn2 KO mice revealed modest alterations in light processing (83). Pain sensitivity was mildly reduced in a hot plate sensitivity test and a shock threshold analysis but not in a tail flick test of pain sensitivity (79, 81). Similarly, no alterations in any sensory functions were observed in transgenic mice with a modest overexpression of NLGN2 (41).

NLGN2 in motor behavior

Similar to sensory functions, motor behaviors are affected only very mildly by either loss of function or gain of function of NLGN2. A detailed analysis of basic motor functions, stereotypies, and reflexes revealed no major phenotypic changes in Nlgn2 KO mice (81), and in the accelerated rotarod test, Nlgn2 KO mice showed subtle impairments in motor coordination and/or motor learning, which may depend on strain background (79, 81). Analysis of locomotor activity in a homecage setting indicated that Nlgn2 KO mice spent less time immobile than their WT littermates (45); this, however, may be a consequence of disruptions in sleep architecture (84) rather than a change in locomotor function. In transgenic mice with a moderate overexpression of NLGN2, basic motor functions and locomotor activity were also largely unaltered, although minor abnormalities were detected (41).

NLGN2 in seizure phenotypes

Despite the consistent reductions in synaptic inhibition, global deletion of Nlgn2 has never been reported to result in generalized tonic-clonic motor seizures. However, abnormal spike and wave discharges as well as behavioral arrests characteristic of absence seizures were observed in Nlgn2 KO mice, and these were ameliorated by the anti-absence seizure drug ethosuximide (47). Intriguingly, restoring GABAergic transmission in a projection from the nRT to the ventrobasal thalamus also partially normalized these phenotypes, indicating that they may result from loss of feedforward inhibition in the nRT-thalamic circuit (47).

NLGN2 in social behavior

Perhaps surprisingly, given the association of Neuroligin family members with psychiatric disorders (detailed below), constitutive deletion of Nlgn2 appears to have very little impact on social behaviors. In particular, Nlgn2 KO mice performed comparably to WT littermates on multiple tests of sociability (defined as a preference for exploration of a novel mouse over a novel object) and showed inconsistent changes in social learning and memory (defined as an acute or long-term preference for exploration of a novel mouse over a familiar one) (47, 79, 81). The only socially relevant phenotypic changes observed in these mice were a reduction in ultrasonic vocalizations in Nlgn2 KO pups in response to maternal separation (81) and a reduction in social dominance in the tube test (74). Interestingly, conditional deletion or KD of Nlgn2 from adult mPFC (48) or from dopamine D1 receptor–expressing neurons in the nucleus accumbens (85) resulted in a reduction in sociability that was not observed after constitutive deletion, raising the question whether compensatory mechanisms during development may obscure some social abnormalities in Nlgn2 KO mice. Moreover, KD of Nlgn2 expression specifically in D1 versus D2 receptor–expressing neurons in the nucleus accumbens resulted in opposite effects on social dominance in a tube test, highlighting the complexity of assessing the molecular basis of behavioral phenotypes in constitutive Nlgn2 KO mice.

In contrast to Nlgn2 deletion, gain of function or local inhibition of NLGN2 yielded a range of alterations in social function that are complex to interpret (41, 8688). It is possible that the heterogeneity of NLGN2 levels and spatial expression patterns that is inherent with viral overexpression or peptide inhibitor approaches may lead to distinct behavioral phenotypes depending on the precise nature of the manipulation.

NLGN2 in cognitive function

Surprisingly, little is known to date about the role of NLGN2 in cognition, but the few studies that exist indicate that NLGN2 may particularly modulate associative learning paradigms involving an aversive component. For example, conditional deletion of Nlgn2 from the mPFC of adult mice led to a deficit in cued and contextual fear conditioning, which was due to disrupted fear memory acquisition or consolidation but not retrieval (48). This behavioral phenotype was accompanied by an attenuation of training-induced immediate early gene expression, indicating that impaired inhibition may disrupt mPFC function by uncoupling it from experience-dependent inputs (48). Interestingly, the same manipulation had no effect on spatial memory as assessed in a Y-maze paradigm. In another study, local KD of Nlgn2 expression in the nucleus reuniens of the thalamus, which receives inputs from the mPFC, improved context discrimination in a fear conditioning paradigm by modulating inhibition in a circuit that is implicated in fear generalization (89). In an inhibitory avoidance task, inhibition of NLGN2 function in the prelimbic region of the mPFC in rats resulted in a persistent disruption of memory retention and an enhancement of immediate early gene expression after a retention trial (90). Last, in an auditory discrimination task, constitutive Nlgn2 KO mice showed decreased avoidance of a punished tone despite normal auditory acuity, indicating a potential impairment in encoding the valence of the auditory stimulus (82). To our knowledge, only one study has identified a role for NLGN2 in cognitive functions other than aversive conditioning, showing that viral overexpression of NLGN2 rescued a stress-induced attention deficit in the five-choice serial reaction time task (91). Further studies will be essential in determining the full extent to which NLGN2 contributes to the regulation of the neural circuits underlying cognition.


As a key regulator of inhibition in the brain, NLGN2 is also a candidate of particular interest in the etiology of neuropsychiatric disorders, several of which have been associated with disturbed E/I ratios due to synaptic dysfunction within specific neuronal circuits. Consistent with this notion, several NLGN2 variants have been identified in individuals with schizophrenia, autism, and anxiety disorder, and mouse models of these variants mimic psychiatrically relevant phenotypes.

NLGN2 variants in human patients

The first evidence for an involvement of NLGN2 in psychiatric pathophysiology arose when six rare missense mutations in the exons and promotor region of the NLGN2 gene were found in patients with schizophrenia (92). Among these, an R215H point mutation was predicted to have damaging effects, albeit with incomplete clinical penetrance. In cell-based assays, the R215H mutation was found to encode a loss-of-function variant due to a failure in protein maturation and glycosylation, resulting in deficient translocation to the cell membrane and subsequent impairments in GABAergic synaptogenesis (92, 93). Consistent with this notion, a subsequent genomic study screening subjects with schizophrenia detected an overall increased level of rare mutations in mostly noncoding areas of the NLGN2 gene, although none of these mutations was identified as directly increasing the risk for this condition (94). More recently, an apparently pathogenic heterozygous de novo nonsense mutation (Y147*) in NLGN2 was identified in a case report on a 15-year-old male patient showing a unique phenotype including autism spectrum disorder, obsessive-compulsive behaviors, anxiety, macrocephaly, and obesity (95). Moreover, a decreased amount of NLGN2 expression was found in the nucleus accumbens of patients with major depressive disorder (85), as well as in iPSC-derived cortical interneurons from patients with schizophrenia (96). Last, mutations in several of the interaction partners of NLGN2 at inhibitory synapses, most notably neurexins, gephyrin, Cb, S-SCAM, IgSF9b, and the MDGAs, have also been implicated in psychiatric disorders, with partial overlap of patient phenotypes (Fig. 3). Together, these findings support the notion that alterations in the molecular complex containing NLGN2 may substantially contribute to the genetic and/or environmental risk of mental disorders. Nevertheless, further analyses of both the known NLGN2 variants and potential additional mutations will be essential for understanding the full extent and clinical impact of this contribution.

Fig. 3 NLGN2 and interacting partners in psychiatric disorders.

Variants in NLGN2 and several of its interacting partners have been linked to psychiatric disorders, including neurexin (NRXN) (102), NLGN3 and NLGN4 (103), IgSF9b (104106), dystroglycan complex (107), MDGA1/2 (108), gephyrin (109, 110), Cb (also known as ARHGEF9) (111), and S-SCAM (112). The figure is adapted from (113). ADHD, attention deficit hyperactivity disorder. Credit: Adapted by A. Kitterman/Science Signaling from D. Krueger-Burg/Max Planck Institute of Experimental Medicine and University Medical Center Göttingen.

Mouse models of disease-associated NLGN2 variants

Animal models of NLGN2 mutations discovered in human patients provide useful tools to further investigate the precise role of aberrant NLGN2 in neuropsychiatric symptoms. Accordingly, on the basis of the R215H point mutation identified in a patient with schizophrenia (92), two homozygous R215H knock-in (KI) mouse lines were generated (93, 97). As expected on the basis of evidence from cell-based assays (92), these mouse lines expressed little or no NLGN2 protein, effectively rendering them constitutive KOs, although minor residual expression of immature nonglycosylated protein could be detected in the line reported by Jiang et al. (93, 98) but not the other reported by Chen et al. (97). Despite this residual expression, the Jiang et al. line showed the expected impairment in inhibitory synapse function but not excitatory synapse function, with decreased levels of the postsynaptic GABA-related proteins gephyrin and GABAARγ2 as well as reduced mIPSC frequency and amplitude in both the dentate gyrus of the hippocampus (93) and the mPFC (98). Notably and in discrepancy with the constitutive Nlgn2 KO mice, levels of the presynaptic markers PV and VIAAT were also found to be reduced in both brain regions (93, 98), indicating that the residual NLGN2-R215H mutant protein may confer a gain of function with respect to presynaptic development rather than mimicking loss of function, as has previously been reported for an autism-related mutation in NLGN3 (11). Moreover, electrophysiological approaches revealed abnormal GABAergic network activity in the mPFC, with a specific reduction in the power of fast gamma oscillations at 60 to 100 Hz (98). This phenomenon may be attributable to an impaired function of PV-positive synapses, which are most prominently affected by Nlgn2 deletion (see above) and are known to be essential for the generation of gamma oscillations (51). Intriguingly, a disruption in the function of PV-positive interneurons in the cortex has long been postulated to be a key pathophysiological mechanism in schizophrenia (99), strengthening the notion that NLGN2 variants may contribute to this disorder through their effects at PV-positive synapses in the mPFC and beyond.

In line with the molecular and cellular consequences of the R215H mutation in NLGN2, both KI mouse lines show behavioral phenotypes that partially recapitulate the constitutive KO mice, including, most notably, the prominent increase in anxiety behaviors in the OF and EPM (93, 97, 98). They also display a far wider range of cognitive impairments than have previously been reported for the constitutive KO mice, including impaired spatial learning and memory in the Morris water maze and the novel location recognition test but not the novel object recognition test (97), as well as impairments in spontaneous alternation, contextual fear conditioning (93), and cued fear conditioning (98). Last, alterations in prepulse inhibition and exaggerated stress responses after restraint stress were also reported (93, 97), supporting the notion that these models recapitulate several core phenotypes of schizophrenia. Together, these data indicate that the R215H KI mouse line may serve as novel in vivo model for this condition and may additionally help to understand the precise role of NLGN2 in pathophysiological mechanisms.


Although NLGN2 has emerged as a central player at inhibitory synapses in the almost 25 years since it was first identified, we are still far from truly understanding the mechanisms by which it regulates synapse assembly and function and, in consequence, the flow of information through neuronal networks in health and disease. At the cell biological level, it will be crucial to resolve current controversies regarding the synapse specificity of NLGN2 function at PV-positive versus other inhibitory synapse subtypes, as well as identifying its interaction partners at these different synapses and the mechanisms by which they together recruit synapse-specific GABAAR subtypes. Similarly, it will be essential to determine how these mechanisms contribute to the regulation of local circuit function and long-range connectivity by GABAergic neurons, which, in turn, play key roles in governing the behavioral output of the brain. Given an increasing realization that alterations in the function of GABAergic synapses and neurons are major contributors to a wide range of psychiatric and neurodevelopmental disorders, understanding the detailed function of NLGN2 and its partners at these synapses may open interesting new avenues for the development of therapeutic strategies to these disorders.


Acknowledgments: We are grateful to N. Brose for continuous advice and support of their research in the Department of Molecular Neurobiology and for valuable feedback on this review article. Funding: D.K.-B. was the recipient of a NARSAD Young Investigator Grant (Brain and Behavior Research Foundation). Competing interests: The authors declare that they have no competing interests.

Stay Connected to Science Signaling

Navigate This Article