|Home | About | Journals | Submit | Contact Us | Français|
Recently, leucine-rich repeat transmembrane proteins (LRRTMs) were found to be synaptic cell-adhesion molecules that, when expressed in non-neuronal cells, induce presynaptic differentiation in contacting axons. We now demonstrate that LRRTM2 induces only excitatory synapses, and that it also acts in transfected neurons similar to neuroligin-1. Using affinity chromatography, we identified α- and β-neurexins as LRRTM2 ligands, again rendering LRRTM2 similar to neuroligin-1. However, whereas neuroligins bind neurexins containing or lacking an insert in splice site #4, LRRTM2 only binds neurexins lacking an insert in splice site #4. Binding of neurexins to LRRTM2 can produce cell-adhesion junctions, consistent with a trans-interaction regulated by neurexin alternative splicing, and recombinant neurexin-1β blocks LRRTM2's ability to promote presynaptic differentiation. Thus, our data suggest that two unrelated postsynaptic cell-adhesion molecules, LRRTMs and neuroligins, unexpectedly bind to neurexins as the same presynaptic receptor, but that their binding is subject to distinct regulatory mechanisms.
Synapse assembly, maturation, specification, and maintenance are likely driven by a multitude of trans-synaptic cell-adhesion molecules. Multiple synaptic cell-adhesion molecules may contribute to these processes, including but not limited to neurexins and neuroligins (Ushkaryov et al., 1992; Ichtchenko et al., 1995), ephrins and Eph receptors (Torres et al., 1998), SynCAMs (Biederer et al., 2002), and netrin G-ligands (Kim et al., 2006). A key technical advance in studying synaptic cell-adhesion molecules was the discovery that expression of such proteins in non-neuronal cells can potently enhance formation of synapses onto these cells (i.e., induce presynaptic differentiation of axons), when these cells are co-cultured with neurons (Scheiffele et al., 2000; Biederer et al., 2002; Graf et al., 2004; Kim et al., 2006). In this assay, referred to as the artificial synapse-formation assay, SynCAMs, neuroligins/neurexins, and NGLs are active (see references cited above). Most recently, a family of neuronal leucine-rich repeat proteins called LRRTMs was also identified as postsynaptic proteins that are active in this assay (Linhoff et al., 2009; Brose, 2009).
LRRTMs comprise a family of four homologous leucine-rich repeat proteins that are selectively expressed in neurons with a differential distribution in brain (Lauren et al., 2003). LRRTM1 is a maternally suppressed gene that is associated paternally with handedness and schizophrenia (Francks et al., 2007; Ludwig et al., 2009). All LRRTMs induce presynaptic differentiation in artificial synapse-formation assays, and LRRTM2 is localized to excitatory synapses (Linhoff et al., 2009). Moreover, deletion of LRRTM1 in mice causes an increase in the immunoreactivity for the vesicular glutamate transporter VGLUT1 (Linhoff et al., 2009), a morphological change similar to that observed in neuroligin-3 R451C knockin mice (Tabuchi et al., 2007). Together, these data indicate that LRRTMs may be postsynaptic cell-adhesion molecules similar to neuroligins. However, these data raise important new questions, for example whether LRRTMs also alter synapse numbers in neurons, and more significantly, which presynaptic molecules they might interact with.
Here, we examined the role of LRRTMs in neurons, focusing on LRRTM2 because or its well-documented localization to synapses (Linhoff et al., 2009). We demonstrate that LRRTM2 selectively induces excitatory synapse formation in the artificial synapse-formation assay, and increases excitatory synapse density in transfected neurons. Moreover, we identify neurexins as the presynaptic receptors for LRRTM2, and demonstrate that neurexin-binding to LRRTM2 is tightly regulated by alternative splicing of neurexins at splice site #4 (SS#4). Our data expand the trans-synaptic interaction network mediating synaptic cell adhesion, and suggest that neurexins generally nucleate trans-synaptic signaling.
We transfected COS cells with plasmids encoding only mVenus (control), or mVenus-fusion proteins of LRRTM2 or neuroligin-1, and co-cultured the transfected COS cells with cultured hippocampal neurons. After two days of co-culture, samples were fixed, immunolabeled for mVenus and synaptic markers, and analyzed by quantitative fluorescence microscopy (Figures 1A–1B).
Immunostaining for synapsin confirmed that LRRTM2, similar to neuroligin-1, induced formation of stable presynaptic terminals onto transfected COS cells (Linhoff et al., 2009). Analysis with antibodies to the vesicular glutamate- and GABA-transporters (VGLUT1 and VGAT, respectively) demonstrated that LRRTM2 only induced formation of excitatory, VGLUT1-containing synapses on transfected COS cells, whereas neuroligin-1 induced formation of both excitatory and inhibitory-synapses (Figures 1A–1B and S1A). Neither LRRTM2 nor neuroligin-1 produced formation of synaptic specializations containing the postsynaptic marker PSD-95.
We next examined whether the effect of LRRTM2 can also affect synapse density in neurons. We transfected cultured hippocampal neurons at 10 days in vitro (DIV10) with plasmids encoding only mVenus (control), or mVenus-fusion proteins of LRRTM2 or neuroligin-1, and fixed and immunostained the neurons at DIV14. Using image analysis, we quantified the density and signal intensity of immunoreactive synaptic puncta identified with antibodies for various pre- and postsynaptic markers and for mVenus (Figures 1C–1E). LRRTM2 potently increased the density of excitatory, but not inhibitory synapses on transfected neurons, as measured with both pre- and postsynaptic marker proteins. In this assay, LRRTM2 thus again acted similar to neuroligin-1, but was more effective (Figures 1C–1E). Moreover, LRRTM2 strongly increased the signal of presynaptic markers per punctum on the transfected neurons, also similar to neuroligin-1. However, different from neuroligin-1, LRRTM2 had no significant effect on the postsynaptic signal (Figures 1C–1E).
To search for presynaptic ligands of postsynaptic LRRTM2, we produced recombinant LRRTM2 composed of the entire extracellular sequence of LRRTM2 fused to the Fc-domain of human IG (Ig-LRRTM2), analogous to previously generated neurexin fusion proteins (Ushkaryov et al., 1994). As a control, we used IgC that is composed of the first 18 residues of mature neurexin-1α fused to the Fc-domain of human IG. Ig-fusion proteins were produced in transfected HEK293 cells, and purified on protein A-Sepharose (Figure 2A). We then performed affinity chromatography experiments with solubilized rat brains on the immobilized Ig-fusion proteins (Figure 2B).
Silver staining of SDS polyacrylamide gels loaded with the input, wash and eluate fractions from affinity chromatography experiments suggested that multiple bands are purified on immobilized LRRTM2. We performed mass spectrometry analyses of all proteins in the silver-stained gels larger than Ig-LRRTM2. Of 140 identified peptides, 31 peptides were derived from neurexins, with all three α-neurexins represented, and 18 peptides were from synaptotagmin-1, which binds to neurexins (Figure 2B; Hata et al., 1993). Besides neurexins, no other cell-surface proteins were identified in the LRRTM2 affinity-purified fractions, suggesting that neurexins are the most abundant and/or the most tightly binding extracellular interaction partners of LRRTM2.
To confirm that LRRTM2 indeed pulls down neurexins present in detergent-solubilized membrane fractions, we used immunoblotting to analyze which proteins were captured by immobilized LRRTM2 (Figure 2C). We observed a high degree of enrichment of α- and β-neurexins in the LRRTM2-bound fraction, but not of any other cell-surface protein tested, confirming the mass spectroscopy results.
To test whether neurexins can directly bind to LRRTM2 on the surface of a cell, we bound various recombinant Ig-fusion proteins of neurexins to HEK293T cells that express full-length LRRTM2-mVenus. We fixed the cells without detergent, and measured surface-bound Ig-fusion proteins by indirect immunofluorescence (Figure 3A). As a negative control, we used IgC, and as a positive control, we employed cells transfected with two different neuroligin-1 splice variants. We found that both neurexin-1α and -1β avidly bound to LRRTM2 in this assay, but only when the neurexins lacked an insert in SS#4. In contrast (but as reported previously, see Boucard et al., 2005), neurexin-1α containing or lacking an insert in SS#4 did not bind to neuroligin-1 containing an insert in splice sites A and B, but did bind to neuroligin-1 lacking an insert in splice sites A and B. Neurexin-1β, similar to neurexin-1α, also bound to LRRTM2 dependent on SS#4 of neurexin-1β, but was independent of either neurexin-1β or neuroligin-1 alternative splicing (Figure 3A). Thus, binding of neurexin-1α and -1β to LRRTM2 and to neuroligin-1 is differentially controlled by alternative splicing at SS#4 of neurexins, suggesting that neurexin-binding to LRRTM2 and neuroligins operates via distinct but related mechanisms. This conclusion is reinforced by the finding that binding of both neurexin ligands is reversed by EGTA, i.e., is Ca2+-dependent (Figure 3A).
To investigate whether binding of LRRTM2 to neurexins is a trans-interaction, i.e. capable of promoting intercellular cell-adhesion (as expected for an interaction between a postsynaptic cell-adhesion molecule with presynaptic neurexins), we examined the ability of surface-expressed LRRTM2 to mediate cell-adhesion with cells expressing neurexins. We transfected HEK293T cells with (i) mVenus or tdTomato alone; (ii) mVenus-fusion proteins of LRRTM2; (iii) mVenus-fusion protein of neuroligin-1, (iv) mCherry-fusion proteins of various neurexins; and (v) tdTomato co-transfected with untagged LRRTM2. One day after transfection, cells were dissociated, and mVenus-expressing and tdTomato- or mCherry-expressing cells were mixed. Cells were imaged immediately after mixing and after a 60 min incubation at room temperature with mild agitation, and free cells were counted at each time point to quantify cell adhesion (Figures 3B–3C).
Since at least in some instances, leucine-rich repeat proteins mediate homophilic cell adhesion (e.g., see the role of connectin in Drosophila synapse formation; Nose et al., 1992), we first examined whether LRRTM2 mediates homophilic cell adhesion by mixing red (td-Tomato) and green (mVenus) LRRTM2-expressing cells. However, we observed no cell-adhesion (Figure S3A). We next examined whether LRRTM2-binding to neurexins can mediate cell adhesion. Indeed, cells expressing LRRTM2 formed large aggregates with cells expressing neurexin-1α or -1β, provided that the neurexins lacked an insert in SS#4 (Nrx1αSS4− and Nrx1βSS4−, but not Nrx1αSS4+ or Nrx1βSS4+; Figures 3B–3C). All three β-neurexins bound with the same splice-site dependence (Figures S3B–S3C). This splice-site dependence was different from that observed with neuroligin-1, where cells expressing neuroligin-1 lacking an insert in the neuroligin-splice sites A and B (NL1ΔAB) formed aggregates with all neurexins tested, consistent with the binding data shown in Figure 3A.
However, similar to neuroligin-1, LRRTM2-mediated cell-adhesion to neurexin-containing cells was strictly dependent on Ca2+-binding to neurexins. Binding was abolished by EGTA, and was blocked by a single amino-acid substitution in neurexin-2β that abolishes Ca2+-binding to the neurexin LNS domain, and that also abolishes neuroligin-1 mediated cell adhesion (Figures 3B–3C, S3B, and S4). No cell adhesion in negative controls was observed (Figure 3B and Figure S4). Taken together, these data demonstrate that LRRTM2 forms an intercellular junction via a trans-interaction with α- and β-neurexins in a manner that resembles that of neuroligin-1, but differs from neuroligin-1 in its distinct regulation by alternative splicing of the neurexins.
To corroborate the surface-binding and cell-adhesion assays with an independent approach, and to additionally test whether neurexins also bind to other LRRTM isoforms, we performed pull-down experiments using the immobilized Ig-fusion protein of neurexin-1β lacking an insert in SS#4 (IgNrx1αSS4−; Figure 4A). We found that immobilized Ig-neurexin-1β effectively captured all LRRTM isoforms (LRRTM1–4) and neuroligin-1 expressed in HEK293 cells. However, Ig-neurexin-1β did not bind to NGL-3, and IgC used as a negative control was unable to bind to either LRRTMs or neuroligin-1.
We next reversed the assay, and employed the Ig-fusion protein of LRRTM2 described above (see Figure 2) to pull down neurexins expressed as mCherry fusion proteins in HEK293 cells. Strikingly, LRRTM2 pulled down only neurexins lacking an insert in SS#4, and exhibited no activity towards neuroligin-1, confirming the binding specificity observed in the cell-surface binding and cell-adhesion assays (Figure 4B).
Neurexin binding by both neuroligin-1 and LRRTM2 depends on Ca2+-binding to neurexins, and both binding reactions are regulated, although differentially, by SS#4 of neurexins. These observations indicate that despite their lack of homology, LRRTM2 and neuroligin-1 may bind to overlapping sites on the neurexin LNS domain. To test this hypothesis, we investigated whether LRRTM2 and neuroligin-1 simultaneously bind to neurexins. We produced recombinant Ig-LRRTM2, FLAG-tagged neuroligin-1 lacking an insert in splice sites A and B (FLAG-NL1ΔAB), and HA-tagged neurexin-1β lacking an insert in SS#4 (HA-Nr×1βSS4−). We then mixed these proteins in different combinations, and tested whether pulldowns of neuroligin-1 or LRRTM2 would capture not only neurexin-1β, but also LRRTM2 or neuroligin-1, respectively (Figures 4C and 4D).
Unequivocally, only neurexin-1β but not with LRRTM2 was brought down with neuroligin-1. Similarly, only neurexin-1β but not neuroligin-1 was brought down with LRRTM2 (Figures 4C and 4D). These data demonstrate that neuroligin-1 and LRRTM2 cannot simultaneously bind to neurexin-1β.
Finally, we estimated the binding affinity for the interaction of LRRTM2 and neurexin-1β. LRRTM2-mVenus expressing and control HEK293T cells were incubated with increasing amounts of Ig-neurexin-1β fusion protein lacking an insert in SS#4, and proteins bound to the cell surface were measured with an HRP-tagged secondary antibody. After subtraction of non-specific binding, we calculated a nanomolar Kd by Scatchard analysis (5.83 ± 1.47 nM; Figure 4E). This result indicates that LRRTM2 binds to neurexins with high affinity, although it should be noted that the dimeric neurexin ligands used in these experiments produce an increased binding affinity.
Since neurexins directly interact with LRRTM2 (Figure 2–Figure 4), the question arises whether this interaction is essential for the synaptogenic activity of LRRTM2 (Figure 1). As a first test of this question, we examined whether Ig-neurexin-1β lacking or containing an insert in SS#4 (IgNrx1βSS4− and IgNrx1βSS4+) can block the function of LRRTM2 in the artificial synapse formation assay (Figures 5A and 5B). In these experiments, we used IgC as a negative, and neuroligin-1 as a positive control, and restricted the co-culture time to 12 h to avoid cellular uptake and degradation of the Ig-fusion proteins (Chubykin et al., 2005).
IgNrx1βSS4− but not IgNrx1βSS4+ specifically reduced the synaptogenic activity of LRRTM2 (Figures 5A and 5B), suggesting that neurexins are indeed presynaptic receptors for LRRTM2 in the artificial synapse formation assay, similar to neuroligin (Ko et al., 2009). Interestingly, the synaptogenic activity of neuroligin-1 was also specifically reduced by IgNrx1βSS4− but not by IgNrx1βSS4+, although neuroligin-1 (NL1 ΔAB) binds to both neurexin isoforms in vitro (Figure 3), possibly because of the differences in neuroligin-1 binding affinities of the two different neurexin splice variants (Boucard et al., 2005), or because of the dependence of the artificial synapse-formation assay on α-neurexin binding by neuroligin-1 (Ko et al., 2009).
Synaptic cell adhesion not only mediates the initial establishment of synapses, but also directs synapse specification, controls synapse maintenance, and regulates synapses during long-term synaptic plasticity. Moreover, the recent discovery of multiple synaptic cell-adhesion molecules as candidate genes for cognitive diseases such as autism, schizophrenia, and addiction has moved synaptic cell-adhesion into the center of attention (reviewed in Südhof, 2008). In particular, neurexin-1 has been repeatedly linked to autism and schizophrenia (Kim et al., 2008; Kirov et al., 2009; Rujescu et al., 2009), LRRTM1 was linked to schizophrenia (Francks et al., 2007; Ludwig et al., 2009), and neurexin-3 has been associated with reward pathways and drug addiction (Bierut et al., 2007). Thus, the recent finding that LRRTMs are candidate synaptic cell-adhesion molecules that are potent effectors in the artificial synapse-formation assay (Linhoff et al., 2009) were of great interest because they suggested that LRRTMs may define a novel trans-synaptic cell-adhesion pathway that may contribute to cognitive diseases. However, these results also raised important questions, namely whether the artificial synapse-formation activity of LRRTMs truly reflects a role in synapses formed between neurons, whether this role applies to all types or specific subtypes of synapses, and whether LRRTMs act as homophilic cell-adhesion molecules analogous to the leucine-rich repeat protein connectin (Nose et al., 1992), or function by binding to an as yet unidentified presynaptic ligand.
In the present study, we have provided initial answers to these questions by focusing on LRRTM2, the most abundant LRRTM isoform. We show that LRRTM2 increases the abundance of excitatory but not inhibitory synapses, not only in the artificial synapse-formation assay, but also in transfected neurons. Moreover, we demonstrate that LRRTM2 is not a homophilic cell-adhesion molecule, but instead binds to presynaptic neurexins in a tight interaction that is regulated by alternative splicing of neurexins at SS#4. Thus, our data unexpectedly show that neurexins act by binding to two different downstream ligands, neuroligins and LRRTMs, in a mutually exclusive manner, and that their binding to these ligands is differentially regulated. As illustrated in the model (Figure 5C), these findings place neurexins at the core of two separate trans-synaptic cell-adhesion pathways, and indicate that the involvement of LRRTM1 and neurexin-1 in schizophrenia may delineate a common mechanism.
As synaptic cell-adhesion molecules, LRRTMs could be involved in one or several steps during synapse formation, from their initial establishment to their maturation and remodeling. The assays used in the present study – the artificial synapse-formation and neuronal transfection assays – do not allow conclusion about the steps in which a protein functions. For example, in the artificial synapse-formation assay, even control non-neuronal cells form transient synapses with co-cultured neurons; thus, if a molecule stabilizes or validates synapses, it could in this assay stabilize synapses that are initially established by an independent mechanism. In vivo, neuroligins appear to be more important for synapse specification, function, and plasticity than for the initial establishment of synapses (Varoqueaux et al., 2006; Chubykin et al., 2007), and neurexins have only a demonstrated role in synapse specification and function but not synapse establishment (Missler et al., 2003; Kattenstroth et al., 2004; Zhang et al., 2005). However, these findings do not necessarily mean that LRRTMs do not function in the establishment of initial synapses, and even neurexins might do so, since no complete neurexin deletion has yet been analyzed. Moreover, we do not know whether LRRTMs also bind only to other presynaptic ligands, which is a distinct possibility since neuroligins also appear to interact with other presynaptic ligands besides neurexins (Ko et al., 2009). Again, these are issues that will have to be addressed in future studies.
Artificial synapse-formation assays were performed with COS-7 cells and co-cultured hippocampal neurons as described (Ko et al., 2009), and analyzed after 48 h.
Affinity Chromatography experiments were performed with rat brain proteins solubilized with 1% Triton X-100, using immobilized LRRTM2-Ig or IgC fusion proteins as a matrix (Sugita et al., 2001; Boucard et al., 2005). Proteins bound to LRRTM2 affinity matrix were identified by mass spectrometry (Stanford University Mass Spectrometry Facility), and identifications were confirmed by immunoblotting (Figure 2).
Cell Surface Binding Assays were performed with HEK293T cells as described (Boucard et al., 2005).
Cell Aggregation Assays were performed with HEK293T cells essentially as described (Nguyen and Südhof, 1997). HEK293T cells expressing red or green fluorescent proteins were mixed 24 h after transfection, and incubated at 4°C under gentle agitation. Cell and aggregate numbers were counted immediately after mixing (N0), and after 60 min incubation (N60). Aggregation is calculated as N0/N60.
Cultured hippocampal neurons were transfected with LRRTM2-mVenus, NL1ΔAB-mVenus, or mVenus at DIV10, and immunostained at DIV14 by the indicated antibodies as described (Ko et al., 2009). Images of randomly chosen transfected neurons were acquired using a confocal microscope (LSM510, Zeiss or TCS2, Leica) with constant image settings. Z-stacked images were converted to maximal projection, and analyzed using MetaMorph Software with area size and density of spines and presynaptic terminals per 50µm of dendrite.
All biochemical procedures were performed as described (Boucard et al., 2005; Chubykin et al., 2007). Recombinant proteins were produced in transfected HEK293T cells (Ushkaryov et al., 1994). Constructs were generated as described in detail in the Suppl. Materials. Antibodies and other reagents were purchased commercially, or described previously (Boucard et al., 2005; Chubykin et al., 2007; for a detailed description, see Suppl. Materials).
All data are expressed as means ± SEMs; all experiments were performed with at least three independent cultures, and evaluated statistically by Student’s t test (n = culture numbers). Detailed methods for all procedures are described in the Suppl. Materials, and all numerical data are listed in Suppl. Table 1.
We thank Eunjoon Kim for the gifts of pan-NGL antibody and of expression construct (NGL3-EGFP). This work was supported by grants from the NIMH and Simons Foundation (to T.C.S.), and by fellowships from the Human Frontier Science Program (to J.K.) and the NIH (to M.V.F.).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.