Nrxn expression profiles of medial ganglionic eminence (MGE)-derived interneurons (such as PV and SOM) and caudal ganglionic eminence (CGE)-derived interneurons (such as CCK and VIP) (Kepecs and Fishell, 2014) are embryonically determined and remain stable into adulthood (Lukacsovich et al., 2019). While all three Nrxn genes are typically expressed in a single neuron, Nrxn2 expression is typically the lowest and is sometimes too low to confidently perform analyses on isoform expression and/or alternative exon usage (Aoto et al., 2013; Fuccillo et al., 2015; Lukacsovich et al., 2019; Boxer et al., 2021). Nrxn1 and Nrxn3 transcripts containing the alternative SS4 exon are dominantly expressed in SOM, PV, and CCK interneurons (Fuccillo et al., 2015; Nguyen T. -M. et al., 2016; Winterer et al., 2019). SS4 is located in LNS6, which is a critical binding domain for many Nrxn ligands, and the inclusion (SS4+) or exclusion (SS4-) of the SS4 insert can regulate binding to neuroligins and cerebellins, and dystroglycan (Table 2). Curiously, despite its unique expression and significance to ligand binding, studies have failed to identify the biological significance for the alternative exon usage at SS4 in mediating overall inhibitory synaptic function (discussed later) and further experiments will be required to determine its functional contribution (Nguyen T. -M. et al., 2016; Trotter et al., 2022).

The molecular mechanisms that regulate the selective cell-type specific usage of the alternative SS4 exon are being examined. Convincing evidence has implicated three members of the signal transduction and activation of RNA (STAR) domain family of RNA binding proteins—SAM68, SLM1, and SLM2—in regulating splicing at Nrxn SS4. SAM68, SLM1, and SLM2 possess a single central RNA-binding KH domain and their expression in neurons promotes the skipping of the SS4 exon to produce SS4- Nrxns (Chawla et al., 2009; Iijima et al., 2011, 2014; Ehrmann et al., 2013; Feracci et al., 2016; Nguyen T. -M. et al., 2016; Traunmüller et al., 2016). Of the three molecules, SLM2 appears to critically regulate the alternative splicing at SS4 of all Nrxns. SLM2 is highly expressed in hippocampal principal neurons, which results in the production of Nrxns lacking the SS4 insert. By contrast, a significant fraction of hippocampal PV neurons do not express SLM2 and express SAM68 and SLM1 at levels lower than in principal neurons, which results in the inclusion of the SS4 insert (Nguyen T. -M. et al., 2016). Curiously, while SLM2 appears to be a key determinant for splicing in PV interneurons, the majority of CA1 CCK and SOM interneurons express SLM1 or SLM2, respectively, yet exhibit SS4 exon usage similar to PV neurons (Iijima et al., 2014; Fuccillo et al., 2015; Nguyen T. -M. et al., 2016; Winterer et al., 2019). Although SAM68, SLM1, and SLM2 each have the capability to regulate SS4 usage, it will be important to further assess how each RNA-binding protein is regulated and utilized by different classes of GABAergic interneurons.

The functional relevance of the other Nrxn splice sites is not as well characterized as SS4, even at excitatory synapses. However, the other splice sites could play significant roles in inhibitory synaptic function, as their expression profiles are highly region-and cell-type specific. Notably, the alternative SS2 insert appears to be excluded (SS2-) from Nrxns 1 and 3 of both MGE and CGE-derived interneurons, but is included in Nrxn1 in pyramidal neurons (Lukacsovich et al., 2019). Accordingly, SS2- is required for binding to the inhibitory synapse Nrxn ligands dystroglycan and neurexophilins 1 and 3 (Table 2). Another notable expression pattern involves SS3, which is included in PV neurons, but excluded in CCKs in both hippocampus and cortex (Fuccillo et al., 2015; Lukacsovich et al., 2019). Interestingly, based on the possible combinations of Nrxn alternative exon usage, Nrxn3 has the largest potential for transcriptional diversity, but interestingly does not harness this potential. Instead, alternative exon usage of Nrxn3 is more restricted than predicted and displays the least possible diversity of the three neurexins at the mRNA levels (Schreiner et al., 2014).

In addition to alternative splicing, post-translational modifications further add to the diversity of Nrxns. While all Nrxns are N-glycosylated, the sequences between LNS6 and the transmembrane region are subject to O-linked glycosylation (Ushkaryov et al., 1994). More recently, it was revealed that Nrxns1–3 share a conserved serine residue located immediately upstream of the conserved Cys-loop sequence that is modified by the addition of heparin sulfate (Zhang et al., 2018). Controlled by CA10 and FAM19A1-4, the heparin sulfate modification of Nrxns facilitates interactions with Nlgns and LRRTMs and has been proposed to expand the binding diversity by recruiting heparin-sulfate binding proteins to Nrxn complexes (Zhang et al., 2018; Khalaj et al., 2020; Noborn and Sterky, 2021). Additionally, heparin sulfate modifications may be regulated in an activity-dependent manner by FAM19A1-4 (Khalaj et al., 2020).

Unique to Nrxn3 is the presence of an in-frame stop codon encoded in the SS5 exon (Nrxn3 SS5+), that results in the production of Nrxn3 without transmembrane and intracellular domains (Ushkaryov and Südhof, 1993; Tabuchi and Südhof, 2002). This truncated splice isoform of Nrxn3 is post-translationally modified by the addition of a glycosylphosphatidylinositol (GPI) membrane anchor (Hauser et al., 2022). The GPI membrane anchor attaches the extracellular sequences of Nrxn3 to the plasma membrane but does not permit intracellular signaling. Furthermore, the Nrxn3 SS5 exon encodes for mRNA sequences that are subject to strong translational repression resulting in cell-type specific protein expression. Nrxn3 SS5+ mRNA is detected in pyramidal neurons and GABAergic neurons in hippocampus but its protein is only detected in inhibitory neurons (Hauser et al., 2022).

Measuring the expression and localization of Nrxn protein is no easy task due to the lack of antibodies that detect individual Nrxns (Südhof, 2017). Currently, the most reliable approach to characterize Nrxn protein requires the generation of endogenously tagged Nrxns (Anderson et al., 2015; Ribeiro et al., 2019; Trotter et al., 2019; Klatt et al., 2021; Hauser et al., 2022). However, analysis of Nrxn expression at the protein level and its targeting to specific synapses will be an important undertaking for future studies. One important but unanswered question is whether Nrxn molecules are equally distributed among synapses of a given neuron, or whether specific isoforms are selectively trafficked or retained at certain synapses depending on factors such as the identity or activity of the post-synaptic neuron (Boxer et al., 2021). Moreover, examination of the subsynaptic localization of individual Nrxn molecules, enabled by super resolution microscopy, will expand our understanding of how Nrxns interact with and control synapse organization (Trotter et al., 2019).

Specific to Nrxn3 is a SS5 alternative exon that was recently demonstrated to regulate SOM and CCK transmission in hippocampus. Fascinatingly, the inclusion of the SS5 exon in Nrxn3 mRNA is observed in both glutamatergic and GABAergic neurons in hippocampus, but Nrxn3 SS5+ protein is completely absent from glutamatergic cells, suggesting translational repression in glutamatergic neurons (Hauser et al., 2022). In dentate gyrus, Nrxn3 SS5+ protein (visualized via an epitope tag inserted into endogenous Nrxn3 SS5) is present in PV, CCK, and SOM-expressing interneurons, however, it exhibits selective presynaptic enrichment only in synapses made by SOM and CCK neurons (Hauser et al., 2022). Deletion of the SS5 exon resulted in the selective impairment of SOM-and CCK-mediated transmission without altering PV-mediated transmission or excitatory transmission. This study suggested that Nrxn3 SS5 controls CCK synapse formation or stabilization as the authors observed a slight but significant reduction in the number of synapses formed by Nrxn3 SS5 KO CCK neurons in dentate gyrus, but the synaptic property controlled by Nrxn3 SS5 at SOM synapses was not identified. However, cellular binding assays and elegant proteomics on novel Nrxn3 SS5 HA-tagged mice revealed neurexophilin 1 (Nxph1) and Fam19a2, which bind to Nrxns in cis, as binding partners that may mediate these synaptic functions (Hauser et al., 2022).

Nrxn SS4 is a critical binding site for many inhibitory Nrxn ligands such as neuroligins, dystroglycan, and cerebellins, and governs properties at excitatory synapses in hippocampus. Specifically, Nrxn1 SS4+ controls NMDAR currents whereas Nrxn3 SS4- is required for AMPAR-mediated synaptic transmission (Aoto et al., 2013; Dai et al., 2019, 2021). Inhibitory neurons throughout forebrain predominantly express Nrxn1 and Nrxn3 mRNAs that include the SS4 exon (Fuccillo et al., 2015; Nguyen T. -M. et al., 2016; Lukacsovich et al., 2019; Boxer et al., 2021; Que et al., 2021). Together, these observations have led to the prediction that SS4 must be critical for inhibitory synaptic function. Surprisingly, however, the selective genetic manipulation of SS4 in Nrxn1 and Nrxn3 in PV interneurons had only modest effects on behavior and had no effect on PV synapse numbers (Nguyen T. -M. et al., 2016). Further, a recent study indicated that SS4 of Nrxn3 may not be necessary for inhibitory function in olfactory bulb or mPFC, despite Nrxn3 being predominantly expressed as the SS4+ variant in inhibitory neurons in these regions as well. Interestingly, inhibition in olfactory bulb and mPFC may depend on alternative exon usage at both SS2 and SS4, which in concert control inhibitory synaptic function via interactions with dystroglycan (Trotter et al., 2022).

The studies reviewed above demonstrate that Nrxns play pivotal roles in regulating multiple aspects of inhibitory synaptic transmission that is dependent on the brain region, GABAergic cell-type and synapse studied. At PV- and SOM-expressing GABAergic synapses made onto layer 5 pyramidal neurons in mPFC, the ablation of all Nrxns resulted in a reduction in synapse density and presynaptic release probability, respectively (Chen et al., 2017). At CCK synapses in hippocampal slice cultures, Nrxn triple KO impaired release probability (Uchigashima et al., 2020a). When studying the contributions of individual Nrxns to synaptic inhibition, the field has focused primarily on interrogating the functional properties of GABAergic synapses controlled by Nrxn3. Genetic manipulation of Nrxn3 in Layer 2/3 SOM neurons in ACC resulted in reduced presynaptic release probability, which recapitulated the triple Nrxn KO phenotype (Chen et al., 2017; Keum et al., 2018). In the dentate gyrus of hippocampus, manipulation of the Nrxn3 SS5 insert revealed a role for Nrxn3 SS5 in controlling the strength of CCK- and SOM-expressing synapses onto pyramidal neurons (Hauser et al., 2022). In the subiculum of hippocampus, we found that Nrxn3 plays a critical sex- and synapse-specific role in regulating inhibition exclusively at synapses made by PV interneurons onto regular spiking neurons. In males, Nrxn3 is required to promote synapse maintenance and post-synaptic strength. In females, Nrxn3 suppresses presynaptic release probability (Boxer et al., 2021). Together these findings raise two important questions: (1) How can Nrxn3 display such striking functional heterogeneity? and (2) Do Nrxns other than Nrxn3 also play critical roles in synaptic inhibition?

The function of Nrxns can be influenced by alternative exon usage (Aoto et al., 2013; Dai et al., 2019, 2021), synaptic localization (Hauser et al., 2022), and ligand availability (Uchigashima et al., 2020a). Nrxn alternative splicing varies among classes of hippocampal interneurons, which may result in the formation of different trans-synaptic complexes to differentially sculpt synaptic transmission onto principal neurons (Fuccillo et al., 2015; Nguyen T. -M. et al., 2016; Uchigashima et al., 2020a). As discussed above, studies have failed to find a critical role for the SS4 insert in Nrxn-dependent synaptic inhibition, however, SS2 and SS3 exhibit differential exon usage in distinct classes of GABAergic neurons and SS2 is a binding site for some inhibitory ligands (discussed below) (Fuccillo et al., 2015; Südhof, 2017; Lukacsovich et al., 2019). Although the precise mechanisms that influence the trafficking of Nrxns to presynaptic terminals are poorly understood, Hauser et al. (2022) revealed that the inclusion of the SS5 insert in Nrxn3 selectively localizes Nrxn3 to the axon terminals of CCK and SOM, but not PV interneurons. As expected, deletion of the SS5 insert selectively impacted GABAergic synaptic transmission at synapses made by CCK and SOM interneurons but not PV interneurons. These findings emphasize the need to not only understand Nrxn mRNA expression profiles but also determine the subcellular properties of Nrxns. To circumvent the lack of reliable Nrxn antibodies, groups have developed elegant epitope-tagged Nrxn mice to profile full-length Nrxn1 (Ribeiro et al., 2019; Trotter et al., 2019), all β-Nrxns (Anderson et al., 2015; Klatt et al., 2021) and specifically Nrxn3 with an insert at SS5 (Hauser et al., 2022). However, given that Nrxn3 plays an important role in synaptic inhibition, the development of epitope-tagged mice that permit the assessment of full-length Nrxn3 is necessary. Moreover, as discussed below, Nrxn ligands may not be uniformly expressed in all pyramidal neurons and may not be uniformly distributed to all inhibitory synapses; instead, some ligands are differentially expressed in a brain region-specific manner and can be preferentially enriched post-synaptically at subsets of inhibitory synapses (Lukacsovich et al., 2019; Uchigashima et al., 2020a). Another potential explanation for why Nrxn3 exhibits pleiotropic function at different inhibitory synapses is that signaling mediated by the cytoplasmic sequences of Nrxn3 may be differentially utilized in distinct classes of GABAergic neurons.

Finally, along with Nrxn3, Nrxn1 is highly expressed in most GABAergic neurons in cortex and hippocampus (Nguyen T. -M. et al., 2016; Lukacsovich et al., 2019; Uchigashima et al., 2020a; Boxer et al., 2021; Que et al., 2021) yet the impact of Nrxn1 on synaptic inhibition is poorly characterized. Overexpression experiments hint that Nrxn1 and Nrxn3 may be differentially used at inhibitory synapses: at CCK GABAergic synapses onto CA1 pyramidal neurons in organotypic slices, overexpression of Nrxn1, but not Nrxn3, was capable of partially rescuing the Nrxn triple KO phenotype (Uchigashima et al., 2020a). A systematic assessment of individual neurexin function at GABAergic synapses may provide new insight into how individual Nrxns contribute to altered inhibition observed in neuropsychiatric, neurodevelopmental and substance use disorders.

There are four Nlgn genes, Nlgns 1–4, but Nlgns 1–3 are most often studied in rodents, as Nlgn4 function is not conserved between rodents and humans (Nlgn4 was originally linked to inhibitory synapses in mice but regulates excitatory synapses in humans). Nlgns were quickly appreciated for their powerful synaptogenic properties- when expressed in non-neuronal cells or coated onto beads, these molecules rapidly induced the recruitment of presynaptic specializations (Scheiffele et al., 2000; Graf et al., 2004; Nam and Chen, 2005; Südhof, 2017). However, knockout studies revealed no differences in synapse formation, but significant differences in synapse function (Varoqueaux et al., 2006; Chubykin et al., 2007; Liang et al., 2015; Zhang et al., 2015; Chanda et al., 2017). Nlgn1 is exclusive to excitatory synapses and will not be discussed in detail here. Relevant to inhibitory synapses are Nlgn2 and Nlgn3. Nlgn2 is localized to GABAergic (Graf et al., 2004; Varoqueaux et al., 2004), dopaminergic, and cholinergic synapses (Takács et al., 2013; Uchigashima et al., 2016). Its localization to dopaminergic and cholinergic synapses may potentially be because they co-release GABA (Südhof, 2017). Nrxns are the only known presynaptic receptor for Nlgn2 and engage in trans-synaptic interactions to control synaptic function. Trans-synaptic interactions between Nrxns and Nlgn2 are regulated by alternative splicing at splice-site 4 of Nrxns and by alternative splicing of splice-site A of Nlgn2 (Ichtchenko et al., 1996; Chih et al., 2006; Koehnke et al., 2010). Splice-site A is conserved in all Nlgns. Unlike Nlgn1, Nlgn2 is competent to interact with both α- and β-Nrxn isoforms. Nlgn1 uniquely possesses a second splice-site (splice-site B) that dictates binding to α-Nrxns, which is absent in Nlgn2 (Chih et al., 2005, 2006; Patrizi et al., 2008; Nguyen Q. -A. et al., 2016).

Nlgn2 is recruited to inhibitory synapses independent of GABA_AR activity and then recruits gephyrin, the central scaffolding molecule at inhibitory synapses, and activates collybistin to recruit and organize inhibitory post-synaptic proteins, including GABA_ARs (Graf et al., 2004; Patrizi et al., 2008; Poulopoulos et al., 2009; Ali et al., 2020). Knockout studies indicate that Nlgn2 is required both during development to assemble the post-synaptic components of inhibitory synapses, and throughout adulthood for continuous synapse maintenance (Liang et al., 2015; Troyano-Rodriguez et al., 2019). Across brain regions, imaging studies consistently reveal that Nlgn2 KO impairs post-synaptic but not presynaptic morphology, indicating that Nlgn2 does not control presynaptic organization (Ali et al., 2020). Interestingly, this is in contrast to Nlgns 1 and 3, which participate in the regulation of both post- and pre-synaptic properties (Uchigashima et al., 2021).

Curiously, while Nlgn2 is ubiquitous at inhibitory synapses, Nlgn2 KO primarily affects perisomatic synapses. In hippocampus and basolateral amygdala, Nlgn2 KO selectively impacts GABA_ARs and gephyrin at perisomatic synapses (Poulopoulos et al., 2009; Jedlicka et al., 2011; Babaev et al., 2016); in cortex, Nlgn2 KO mice exhibit reduced inhibition at PV, but not SOM synapses (Gibson et al., 2009). However, note that in addition to impairment at PV synapses, Horn et al. did find that Nlgn2 knock-down in organotypic hippocampal slices impaired SOM-interneuron transmission (Horn and Nicoll, 2018). Nlgn2 KO from Purkinje cells in cerebellum reduced total inhibitory synaptic strength but the relative impact of the KO on perisomatic-targeting basket cells and distal-targeting stellate cells was not tested (Zhang et al., 2015). Surprisingly, despite the fact that CCKs are the other primary perisomatic targeting interneuron in the forebrain, the requirement for Nlgn2 in CCK inhibitory transmission has not been directly assessed. Moreover, Nlgn2 localizes with dystroglycan at somatic synapses, further indicating Nlgn2 may play a yet tested role at CCK synapses (see below for further discussion in section “Dystroglycan”).

MAM-domain containing GPI anchor proteins, MDGAs, are post-synaptic membrane proteins that bind to Nlgn2 with high affinity and compete with Nrxns for Nlgn binding. Overexpression and knock-down studies demonstrate that MGDAs limit the abundance of inhibitory, but not excitatory synapses, in a Nlgn2-dependent manner, and thus MGDAs inhibit Nlgn2’s ability to function at inhibitory synapses (Lee et al., 2013). MGDAs have an important role in neuronal migration and neurogenesis early in development, and MGDA2 KO is lethal.

Of the four Nlgns, Nlgn3 is perhaps the most extensively studied, despite its genetic KO having modest impacts on synaptic function (Südhof, 2017). Nlgn3 is found at both excitatory and inhibitory synapses. Its localization to inhibitory synapses appears to rely on extracellular interactions in cis with Nlgn2 (Nguyen Q. -A. et al., 2016). Functional studies examining Nlgn3 have utilized multiple approaches including ASD-associated Nlgn3 R704C or R451C KI mice, Nlgn3 KO models or Nlgn3 KD/Nlgn3 overexpression via biolistic transfection or viral transduction (see Uchigashima et al., 2021 for in-depth review of these models). These manipulations, extensively studied in hippocampus and somatosensory cortex, have identified inhibitory synaptic properties controlled by Nlgn3 that are distinct from Nlgn2 (Földy et al., 2013; Speed et al., 2015; Horn and Nicoll, 2018).

The Nlgn3 R451C mutation impairs surface trafficking when expressed in non-neuronal cells (Comoletti, 2004; Chubykin et al., 2005) and reduces the ability of Nlgn3 to traffic to the synapse in neurons (Chih et al., 2004; Tabuchi et al., 2007) and produces brain-region specific changes in inhibition (Tabuchi et al., 2007; Etherton et al., 2011). In somatosensory cortex of Nlgn3 R451C mice, inhibitory synaptic transmission is increased without discernable differences in excitatory synaptic transmission (Tabuchi et al., 2007). A subsequent study revealed that the increased inhibition in somatosensory cortex stemmed from decreased endocannabinoid (eCB)-mediated suppression of IPSCs and that PV and SOM-mediated IPSCs were unaltered (Speed et al., 2015), indicating that the increased inhibition could be driven by increased release from eCB sensitive CCK interneurons. By contrast, in CA1 of hippocampus, excitatory but not inhibitory synaptic transmission was increased in the R451C mutant mice (Etherton et al., 2011). Intriguingly, although total inhibition in CA1 was unaltered, further studies revealed synapse-specific R451C-dependent phenotypes. Synaptically connected paired recordings revealed that CCK-mediated inhibition was increased due to impaired tonic, but not phasic eCB signaling (Földy et al., 2013). Additionally, the R451C mutation decreased PV-mediated inhibition onto CA1 pyramidal neurons (Földy et al., 2013). The contrasting effects of the R451C mutation at synapses made by CCK and PV neurons on pyramidal neurons may begin to explain why changes in total inhibition was not observed in CA1. Further adding to the complexity of Nlgn3 R451C function, inhibitory synaptic transmission is reduced in basolateral amygdala (Hosie et al., 2018).

While eCB dependent phenotypes observed at CCK-positive synapses onto CA1 pyramidal neurons could be hypothesized to arise from impaired communication of Nlgn3 R451C with Nrxns, which have been shown to trans-synaptically mediate eCB signaling at excitatory synapses (Anderson et al., 2015). These effects may instead be mediated via impaired Nlgn3 binding to protein tyrosine phosphatase delta (PTPδ). PTPδ is a member of the type-II receptor protein tyrosine phosphatases, which, like Nrxns, are important presynaptic hubs for synaptic organization (Takahashi and Craig, 2013; Yoshida et al., 2021). Yoshida et al. recently found that PTPδ competes with Nrxn1 binding to Nlgn3. The selective disruption of PTPδ-Nlgn3 binding recapitulated many of the biochemical, physiological, and behavioral endophenotypes specifically observed in R451C KI mice. By contrast, disruption of Nlgn3-Nrxn binding did not resemble R451C KI mice, suggesting that the R451C KI phenotype is largely caused by impaired PTPδ binding. In fact, impairing Nlgn-Nrxn binding actually improved some of the social behavioral deficits that are observed in the R451C mice. These unexpected findings underscore how critical the diversity and precision of CAM trans-synaptic signaling is to the shaping neural circuits and behavior (Yoshida et al., 2021).

Interestingly, most of the phenotypes observed in cortex and hippocampus of Nlgn3 R451C KI mice are not phenocopied in Nlgn3 KO mice (Tabuchi et al., 2007; Etherton et al., 2009). However, the increased inhibition specifically at CCK synapses in CA1 of R451C mice is also present in Nlgn3 KO mice (Földy et al., 2013). By contrast, sparse knock-down (KD) of Nlgn3 in CA1 pyramidal neurons of hippocampal slice cultures caused a reduction of CCK IPSC amplitudes (Uchigashima et al., 2020a). The disparate Nlgn3 phenotypes observed in R451C KI mice vs. constitutive KO mice vs. sparse KD in slices could reflect differences in approach and/or requirement of neurexins. While the R451C mutation selectively disrupts PTPδ binding in cortex, Nlgn3 KO prevents both PTPδ and Nrxn binding. Differences in constitutive Nlgn3 KO vs. sparse Nlgn3 KD in-vitro are also distinct manipulations that could explain the different phenotypes. Nlgn3 KO could produce yet-unidentified developmental or compensatory effects that drive the observed functional differences. Similarly, shRNA knockdown or overexpression of Nrxns and their ligands may produce unintended consequences to perisomatic inhibition. Importantly, regardless of the differential approaches, these studies show that Nlgn3 signaling is critical at CCK synapses.

Finally, overexpression of certain Nlgn3 isoforms in CA1 pyramidal neurons can either enhance or suppress IPSCs. Splice-site A of Nlgn3 consists of two exons: A1 and A2, whose inclusion or exclusion yields four Nlgn3 splice isoforms which differ in their extracellular sequences (Nlgn3Δ, +A1, +A2, and +A1A2). Overexpression of Nlgn3 lacking splice-site A (Nlgn3Δ) or only including splice-site A2 enhances IPSCs, whereas inclusion of just A1 or both A1A2 decreases IPSCs (Uchigashima et al., 2020b). Furthermore, Nlgn3Δ and A1 mediate CCK synapses and Nlgn3 A2 mediates SOM synapses, whereas Nlgn3A1A2 may function at excitatory synapses (Horn and Nicoll, 2018; Uchigashima et al., 2020a,b).

Beyond cortex and hippocampus, Nlgn3 also mediates cerebellar and striatal inhibitory synapses in a cell-type and input-specific manner. In cerebellum, Nlgn3 is localized at a subset of inhibitory synapses, including at molecular layer interneuron-Purkinje cell synapses and at synapses in the inner granular layer (Baudouin et al., 2012; Lai et al., 2021). In nucleus accumbens, Nlgn3 expression is highly enriched in D1R compared to D2R MSNs, and accordingly, Nlgn3 KO impairs mIPSCs in D1R, but not D2R MSNs (Rothwell et al., 2014).

Only α and β neurexins with an insert at splice site 4 (SS4+) bind to secreted cerebellins (Cblns)1, 2, or 4 and form a tripartite complex with post-synaptic ionotropic glutamate delta receptors, GluD1 or GluD2 (Yuzaki, 2017). Cerebellins are hexameric glycoproteins of the C1q and tumor necrosis factor superfamily (Kishore et al., 2004) that, depending on the synapse in question, are secreted pre- or post-synaptically. Tripartite Nrxn-Cbln-GluD complex formation regulates trans-synaptic organization, synapse maintenance and synaptic plasticity independent of ion flux through their pore (Fossati et al., 2019; Dai et al., 2021). The functional relevance of these tripartite signaling complexes have historically been studied at excitatory synapses in cerebellum, and more recently at excitatory synapses in subiculum, however, their roles at inhibitory synapses have only recently garnered interest.

The expression patterns of Cbln1-4 are developmentally regulated and vary dramatically by brain region and cell type. Cbln1, 2, and 4 are expressed abundantly throughout the brain while Cbln3 expression is largely restricted to neurons in cerebellum (Pang et al., 2000; Miura et al., 2006; Seigneur and Südhof, 2017). Triple knock-out of Cbln1, 2, and 4 causes reductions in excitatory synapse density in old (6 month) but not young (1–2 month) adult mice at select synapses, indicating Cblns are not required for initial synapse formation, but are utilized dynamically through development to maintain established synapses (Seigneur and Südhof, 2018). In inhibitory neurons, Cblns are differentially expressed in a cell-type-specific manner. For example, in hippocampus, PV inhibitory neurons highly express Cbln4 but not Cbln2 (Nguyen Q. -A. et al., 2016; Seigneur and Südhof, 2017; Boxer et al., 2021). Curiously, in cortex, PV neurons do not express Cblns (Lukacsovich et al., 2019). Note that this does not exclude their involvement at PV inhibitory synapses in these regions, as Cblns could still be localized to synapses by secretion from the post-synaptic glutamatergic neuron. The interrogation of Cbln signaling at inhibitory synapses is at its infancy and continued study to characterize the possible synaptic functions mediated by Cblns and their binding partners is necessary.

Studies of GluD1 and GluD2 have focused on their functional roles at excitatory synapses in cerebellum and hippocampus (Konno et al., 2014; Hepp et al., 2015; Yuzaki, 2017; Dai et al., 2021). In addition to their roles at excitatory synapses, Nrxn signaling via Cbln/GluD complexes are prime binding candidates to regulate inhibitory synapse function because Nrxns predominantly express the alternative SS4 exon, which is required to bind to Cblns to generate a synaptic Nrxn-Cbln-GluD tripartite complex. Indeed, several studies have shown that GluD1 and Cblns are critical for inhibitory synapse formation and specification. GluD1 expressed in HEK cells primarily induced inhibitory presynaptic differentiation when co-cultured with cortical neurons when Cbln1 or 2 were added to the media (Yasumura et al., 2012). Similarly, inhibitory pre-synapses were also formed if HEK cells were co-cultured with entorhinal cortex neurons, which intrinsically express Cbln1 (Ryu et al., 2012). In another study, GluD1 was observed at 50% of inhibitory synapses on pyramidal neurons in layer 2/3 of somatosensory cortex, and GluD1 knockdown decreased the number of inhibitory synapses formed on the dendrites of pyramidal neurons. Consistent with the morphological phenotype, GluD1 knockdown resulted in a significant reduction in mIPSC frequency but not amplitude. The regulation of synapse density by GluD1 in somatosensory cortex is dependent on binding to Cbln4, which is expressed in presynaptic somatostatin interneurons (Fossati et al., 2019). Whether neurexins are required for GluD1/Cbln4 signaling at SOM synapses in somatosensory cortex remains untested. Finally, in nucleus accumbens core of ventral striatum, GluD1 loss led to reduced mIPSC amplitude and frequency, density of inhibitory terminals, and inhibitory release probability, suggesting pre- and post-synaptic effects on synaptic transmission (Gawande et al., 2021). Loss of GluD1 in dorsal striatum, however, produced excitatory but not inhibitory deficits, indicating region-specific localization and function of GluD1.

Dystroglycan is a Nrxn ligand found exclusively at inhibitory synapses (Lévi et al., 2002). The dystroglycan (DG) gene, Dag1, encodes a single polypeptide that is cleaved into α-DG and β-DG (Ibraghimov-Beskrovnaya et al., 1992). α-DG binds to the LNS2 region of α-Nrxns and to the LNS6 region of α and β-Nrxns (Sugita et al., 2001). Importantly, α-DG binding is dependent on alternative splicing: α-DG only binds to Nrxns that lack inserts at splice-site 2 (located in LNS2) and/or splice-site 4 (located in LNS6). Thus, α-DG competes with neurexophilin for α-Nrxn binding at the LNS2 site, and competes with Nlgns for α- and β-Nrxn binding at the LNS6 site (Missler et al., 1998; Reissner et al., 2014). Both α- and β-DG associate with the post-synaptic dystrophin-glycoprotein complex (DGC), which also contains dystrophin, dystrobrevin, and sarcoglycans. The DGC is found exclusively at perisomatic synapses in neocortex (Lidov et al., 1990) but can be found at both perisomatic and dendritic inhibitory synapses in purkinje cells of cerebellum (Sassoè-Pognetto et al., 2011; Briatore et al., 2020).

Several studies have emphasized a critical role for dystroglycan in the assembly and maintenance of CCK synapses in hippocampus and cortex: α-DG is not found opposite PV basket cell terminals, and deletion of Dag1 from pyramidal neurons selectively impairs CCK synapse assembly and maintenance as well as CCK neuron survival, without affecting PV or SOM interneurons in these regions (Früh et al., 2016; Miller and Wright, 2020, 2021). The role that Nrxn signaling plays in dystroglycan-mediated CCK maintenance is unclear and should be tested because in addition to Nrxns, dystroglycan also binds laminin and other LNS-domain containing proteins. Interestingly, Nlgn2 binds to the dystrophin-glycoprotein complex via S-SCAM, yet a functional role for Nlgn2 at CCK synapses has not been explored (Sumita et al., 2007; Patrizi et al., 2008). Furthermore, Nlgn3 and 4 may also interact with the DGC via syntrophin (Yamakawa et al., 2007). Multiple studies suggest that dystroglycan helps to recruit Nlgn2 to DGC+ inhibitory synapses, and that dystroglycan and Nlgn2 may then recruit GABAARs (Brünig et al., 2002; Panzanelli et al., 2017; Briatore et al., 2020). Together these studies suggest that the dystrophin-glycoprotein complex and the α-DG adhesion molecule are critical organizers of the CCK inhibitory postsynapse, recruiting Nlgns 2, 3, 4, and GABA_ARs.

α-DG is required for homeostatic synaptic plasticity by increasing GABAergic currents following prolonged elevation of neuronal activity in hippocampus (Pribiag et al., 2014). Furthermore, chronic stress downregulates α-DG expression in ventral hippocampus (Xie et al., 2022). In-vivo administration of agrin, a dystroglycan ligand, rescues the stress-induced behavioral impairments. A role for neurexins in these intriguing phenotypes was not tested, but may be worth investigating in future studies, as a role for Nrxns in mediating plasticity at inhibitory synapses remains untested.

Neurexophilins are a family of small, secreted glycoproteins that are encoded by four genes in mouse (Nxphs1–4). Nxph genes are highly conserved in vertebrates but are absent in invertebrates (Wilson et al., 2019). Intriguingly, the Nxphs do not share sequence homology with any known Nrxn ligand. Nxph mRNAs exhibit differential expression patterns in brain: Nrxph1 is enriched in hippocampal interneurons, Nrxph3 is enriched in excitatory cortical neurons and Nxph4 in inhibitory hindbrain neurons (Petrenko et al., 1996; Földy et al., 2016; Tasic et al., 2016; Saunders et al., 2018; Zeisel et al., 2018). Nxph1 is arguably the most well-studied of the Nxphs and elegant genetic experiments revealed that Nxph1 signaling through α-Nrxns is required to recruit and stabilize metabotropic GABA_BRs and GABA_ARs at inhibitory synapses in thalamus (Born et al., 2014). Nxph1 was recently identified as a prominent binding partner of Nrxn3 SS5+ expressed in inhibitory neurons in hippocampus (Hauser et al., 2022). Recently, Nxph4, expressed in presynaptic Golgi cells in the cerebellum, was revealed to control GABAergic transmission at Golgi cell—granule cell synapses. Nxph4 KO mice displayed reduced inhibitory synapses as revealed by imaging and electrophysiology experiments (Meng et al., 2019). Further investigation is required to determine if Nxphs control synapse-specific properties in other brain regions in a manner similar to the Nlgns and Cblns, and to continue to elucidate the unique role of α-Nrxns at inhibitory synapses.

Calsyntenins are post-synaptic transmembrane proteins encoded by three evolutionarily conserved genes (Cstn1-3). The extracellular sequences of Cstns contain two cadherin domains and an LNS domain (Vogt et al., 2001; Araki et al., 2003). Although there is conflicting evidence, Cstn3 was identified as an α-Nrxn specific ligand in an unbiased screen designed to identify synaptogenic proteins (Pettem et al., 2013). When expressed in non-neuronal cells and co-cultured with primary hippocampal neurons, Cstn3, but not Cstn1 or Cstn2, robustly recruited excitatory and inhibitory presynaptic specializations (Pettem et al., 2013; Um et al., 2014). However, the manipulation of Cstn3 in neurons has produced differing phenotypes. Genetic deletion of Cstn3 reduced the number of excitatory and inhibitory synapses and reduced the frequency of mEPSCs and mIPSCs in acute hippocampal slices (Pettem et al., 2013). In contrast to these findings, no changes in inhibitory synapse density were observed after shRNA-mediated knockdown of Cstn3 in cultured hippocampal neurons or in hippocampal slices from an independently generated Cstn3 KO mouse (Um et al., 2014; Kim et al., 2020). In support of the notion that Cstn3 functions at inhibitory synapses, Südhof et al. recently studied the impact of CRISPR knockdown of Cstn3 in cerebellar Purkinje neurons and found that the loss of Cstn3 reduced the density of inhibitory synapses formed by basket and stellate cells and significantly reduced mIPSC frequency and evoked IPSC amplitudes (Liu et al., 2022).

There are conflicting reports regarding the binding of Cstn3 with Nrxns. A series of elegant cell binding and biochemical and electron tomography assays identified direct interactions between Cstn3 and α-Nrxns, but not β-Nrxns, which occurred independent of SS4 (Pettem et al., 2013; Lu et al., 2014). This interaction likely requires the LNS domain of Cstn3 with extracellular sequences unique to α-Nrxns (Lu et al., 2014). However, a separate study used mass spectrometry and biochemical and cellular assays and found that Cstn3 binds to both α-Nrxns and β-Nrxns in a SS4-dependent manner (Kim et al., 2020). To add further complexity, another study failed to identify direct interactions between Cstn3 and Nrxns (Um et al., 2014). While the role for Cstn3 at inhibitory synapses is promising, it will be critical to ascertain whether the KO phenotypes reported by Pettem et al. (2013) and Liu et al. (2022) are due to the disruption of Nrxn-Cstn3 interactions or through the disruption of Cstn3 interactions with other binding partners. Additionally, it will be important to address the confusion about the Nrxn isoforms and alternative SS4 usage required for Cstn3 binding.

In addition to interacting with post-synaptic ligands to indirectly regulate the post-synaptic strength of inhibitory synapses, Nrxns directly interact with post-synaptic GABA_ARs to impair the maturation of inhibitory synapses (Zhang et al., 2010). This direct interaction was demonstrated by multiple biochemical assays including affinity chromatography, reciprocal precipitation with immobilized protein and by surface-plasmon resonance. In cultured neurons, overexpression of neurexins mediated by lentiviral transduction or transient transfection reduced the strength of synaptic inhibition. Although the properties underlying the impact of Nrxn overexpression on inhibitory synaptic transmission are largely unknown, the Nrxn-GABA_AR is interaction is cell autonomous and independent of Nlgn2 binding. The mapping of the precise interaction interface and whether this interaction occurs in cis or trans remains to be tested.

Immunoglobulin superfamily member 21 (IgSF21) was recently identified as a Nrxn2α-specific interacting protein (Tanabe et al., 2017). IgSF21 contains two immunoglobulin domains (Ig1 and Ig2) and is anchored to the post-synaptic membrane via a GPI modification. Cellular binding assays revealed that the interaction between IgSF21 and Nrxn2α required Ig1 of IgSF21 and LNS1 of Nrxn2α (Tanabe et al., 2017). IgSF21 gene is expressed at both the embryonic and postnatal stages and its protein is highly expressed in cortex, hippocampus, thalamus, and pons. Constitutive deletion of IgSF21 reduced the levels of inhibitory synaptic proteins, reduced inhibitory synapse densities and functionally impaired inhibitory synaptic strength. Whether the phenotypes observed in the IgSF21 KO are a consequence of developmental or postnatal loss of trans-synaptic signaling with Nrxn2α or other interacting proteins is unknown.