|Home | About | Journals | Submit | Contact Us | Français|
We identify the leucine-rich repeat transmembrane protein LRRTM2 as a key regulator of excitatory synapse development and function. LRRTM2 localizes to excitatory synapses in transfected hippocampal neurons, and shRNA-mediated knockdown of LRRTM2 leads to a decrease in excitatory synapses without affecting inhibitory synapses. LRRTM2 interacts with PSD-95 and regulates surface expression of AMPA receptors, and lentivirus-mediated knockdown of LRRTM2 in vivo decreases the strength of evoked excitatory synaptic currents. Structure-function studies indicate that LRRTM2 induces presynaptic differentiation via the extracellular LRR domain. We identify Neurexin1 as a receptor for LRRTM2 based on affinity chromatography. LRRTM2 binds to both Neurexin 1α and Neurexin 1β, and shRNA-mediated knockdown of Neurexin1 abrogates LRRTM2-induced presynaptic differentiation. These observations indicate that an LRRTM2-Neurexin1 interaction plays a critical role in regulating excitatory synapse development.
The function of the brain critically depends on appropriate synaptic connectivity. Specific patterns of connections, together with the properties of synaptic transmission at particular synapses, determine how the brain processes information. Although significant progress has been made in elucidating the pre- and postsynaptic organization of synapses (Sheng and Hoogenraad, 2007; Sudhof, 2004), our understanding of the molecular mechanisms that regulate the development and function of CNS synapses is limited.
Synaptic cell adhesion molecules are key players in organizing synapse formation between appropriate synaptic partners. Hetero- or homophilic adhesive interactions across a nascent synapse allow target cell recognition and subsequent bidirectional differentiation of pre- and postsynaptic elements. Previous work has identified several synaptic cell adhesion molecules that can mediate trans-synaptic interaction to induce pre- and postsynaptic differentiation. These synaptogenic adhesion molecules include the neuroligins (Ichtchenko et al., 1995; Scheiffele et al., 2000; Song et al., 1999), synCAM (Biederer et al., 2002), NGL-2 and NGL-3 (Kim et al., 2006; Woo et al., 2009) and EphB2 (Kayser et al., 2006).
To identify novel transmembrane proteins that might regulate the development and function of synapses, we carried out an in silico screen for genes that showed subregion-specific expression in the adult hippocampus as reported in the Allen Brain Atlas (Lein et al., 2007) and encoded proteins with extracellular protein-protein interaction domains. Candidate genes were subsequently tested for putative synaptogenic properties in heterologous cell-neuron coculture assays (Biederer and Scheiffele, 2007; Scheiffele et al., 2000). This approach led to the identification of Leucine-Rich Repeat Transmembrane (LRRTM) proteins as proteins that could induce presynaptic differentiation. During the course of the study Linhoff and colleagues reported that LRRTM proteins can induce synapse formation in hippocampal neurons (Linhoff et al., 2009), but several major questions remained unanswered. Importantly, whether endogenous LRRTMs contribute to synaptic function in vivo was not known, and the receptors that mediate the effects of LRRTM had not been identified. Here we provide evidence that endogenous LRRTM2 regulates excitatory synapse development and function in vitro and in vivo, and identify Neurexins as functional receptors for LRRTM2.
The LRRTM gene family consists of four genes, LRRTM1-4. Analysis of the LRRTM gene expression pattern by in situ hybridization showed that these genes are differentially expressed in the developing cortex and hippocampus (Figure S1). To determine the subcellular localization of LRRTM proteins, we transfected 293T cells with tagged constructs, and examined their distribution by immunofluorescence. Myc-tagged LRRTM2-4 and LRRTM4L all localized to the cell membrane in transfected 293T cells (Figure S1). LRRTM1 however remained largely intracellular, despite the presence of a predicted transmembrane domain (Laurén et al., 2003), suggesting that LRRTM1 may require additional proteins for proper membrane targeting.
In a heterologous synapse induction assay, in which LRRTM proteins expressed in 293T cells were co-cultured with hippocampal neurons, we found that LRRTM2 was more effective than the other LRRTM genes in inducing presynaptic differentiation (Figure S2). We therefore decided to investigate whether endogenous LRRTM2 contributed to synapse formation. To determine if LRRTM2 was targeted to excitatory synapses, hippocampal neurons were cotransfected with myc-tagged LRRTM2 and GFP, and immunostained for synaptic markers. Myc-LRRTM2 clustered in the heads of dendritic spines, where it colocalized with the excitatory postsynaptic marker PSD-95, but not with the inhibitory postsynaptic marker gephyrin (Figure 1A). The quality of immunofluorescence with currently available LRRTM2 antibodies is not suitable to draw definitive conclusion about the localization of the endogenous protein, but the localization of the tagged proteins suggests that LRRTM2 primarily localizes to the postsynaptic density of excitatory synapses.
To examine the consequences of LRRTM2 loss of function, we generated an shRNA construct against LRRTM2 to knock down LRRTM2 expression in hippocampal neurons. The LRRTM2 shRNA reduced expression of myc-tagged mouse LRRTM2 in 293T cells by ~90% (Figure 1B). Myc-tagged human LRRTM2, which differs from mouse LRRTM2 by only one nucleotide in the shRNA target sequence, was not affected by the LRRTM2 shRNA, demonstrating target specificity (Figure 1B). To test the efficiency of sh-LRRTM2 in knocking down endogenous LRRTM2 protein levels in hippocampal neurons, the H1 promoter and shRNA were cloned into a lentiviral vector. Dissociated hippocampal neurons were infected at DIV 6 and cell lysate was analyzed by Western blot on DIV 12. The LRRTM2 antibody (Figure S3) detects a weak band around 90 kDa in lysates of DIV 12 hippocampal neurons infected with the control virus, which was strongly reduced in neurons infected with the virus encoding the LRRTM2 shRNA (Figure 1C). Expression of sh-LRRTM2 did not affect protein levels of PSD-95 and βIII-tubulin in the same samples (Figure 1C).
To determine the effects of LRRTM2 knockdown on synapse development, we electroporated hippocampal neurons with LRRTM2 shRNA and control constructs and immunostained the cultures for the excitatory synapse markers PSD-95 and VGlut1 at DIV 14. Expression of sh-LRRTM2 in hippocampal neurons resulted in a 40% reduction in excitatory synapse density (Figure 1D,E), suggesting that endogenous LRRTM2 significantly contributes to excitatory synapse development. The decrease in excitatory synapse density following LRRTM2 knockdown was rescued by coexpression of human myc-LRRTM2 (Figure 1D,E), supporting target specificity of sh-LRRTM2. Expression of sh-LRRTM2 in hippocampal neurons had no effect on inhibitory synapse density (Figure 1F), indicating that expression of sh-LRRTM2 does not cause a general decrease in synapse density. These results indicate that endogenous LRRTM2 contributes to the development of excitatory, but not inhibitory, synapses in cultured hippocampal neurons.
The LRRTM2 loss of function studies suggest that LRRTM2 acts as a postsynaptic organizer of excitatory synapses in hippocampal neurons. A salient feature of hippocampal excitatory synapses is the presence of postsynaptic AMPA and NMDA receptors. To determine whether LRRTM2 recruits AMPA receptors to the postsynaptic density, we analyzed GluR1 surface expression following knockdown of LRRTM2. Neurons were electroporated with shRNA and control constructs, surface labeled for GluR1 at DIV 15 and subsequently permeabilized and immunostained for GFP and VGlut1. Knockdown of LRRTM2 resulted in a 33% decrease in the density of synaptic GluR1 surface puncta (Figure 2A, B). The decrease in density of synaptic GluR1 surface puncta following knockdown of LRRTM2 was rescued by coexpression of human LRRTM2 (Figure 2A, B). The reduction in synaptic GluR1 surface expression following LRRTM2 knockdown was caused primarily by a decrease in the density of GluR1 surface puncta (Figure 2C), although the density of presynaptic VGlut1 puncta was also somewhat reduced (Figure 2D). This suggests that LRRTM2 acts as a postsynaptic organizer of the excitatory synapse and recruits AMPA receptors to the postsynaptic density.
To further investigate the interaction between LRRTM2 and glutamate receptors, we co-expressed myc-LRRTM2 together with GFP-tagged glutamate receptor subunits in 293T cells, and examined their association by immunoprecipitation. Remarkably, GluR1-GFP, GluR2-GFP and NR1-GFP all co-precipitated with full-length LRRTM2 (Figure 2E-G and data not shown). Analysis of the LRRTM2 domain involved in glutamate receptor binding showed that LRRTM2 interacts with GluR1-GFP and GluR2-GFP via the LRR domain (Figure 2E-G). EphB2-YFP did not co-precipitate with LRRTM2, indicating that co-precipitation with AMPA and NMDA receptor subunits does not reflect non-specific interactions with transmembrane proteins. These data indicate that LRRTM2 can bind to glutamate receptor subunits via its extracellular LRR domain. Determination of whether endogenous LRRTM2 associates with glutamate receptors in vivo will require the development of suitable antibodies.
It is also possible that LRRTM2 recruits postsynaptic components via an interaction with synaptic scaffolding proteins. LRRTM proteins contain a highly conserved four amino acid sequence (ECEV) at their C terminus that is reminiscent of a PDZ-binding domain, suggesting that LRRTM2 might interact with PSD-95, an abundant PDZ-domain containing scaffolding protein (Sheng and Hoogenraad, 2007). Furthermore, several proteomic analyses of purified postsynaptic density fractions or immunopurified PSD-95 complexes have identified LRRTM1, LRRTM2 and LRRTM4 as constituents of the postsynaptic density (Dosemeci et al., 2007; Jordan et al., 2004; Yoshimura et al., 2004). To test whether LRRTM2 can interact with PSD-95, myc-tagged LRRTM2 was coexpressed with PSD-95-mCherry in 293T cells. In control cells expressing GFP, PSD-95-mCherry was distributed throughout the cytosol (Figure 2H). Upon coexpression of myc-LRRTM2, PSD-95-mCherry translocated to the cell membrane, where it colocalized with LRRTM2, indicating that LRRTM2 can recruit PSD-95 to the cell membrane (Figure 2H). To identify the domain of LRRTM2 that mediates binding to PSD-95, we generated LRRTM2 deletion constructs lacking the entire cytoplasmic domain (LRRTM2 ΔC) or the C-terminal 4 amino acids (LRRTM2 ΔECEV). Both LRRTM2 mutants were no longer able to bind PSD-95 in 293T cells (Figure 2I), indicating that the C-terminal ECEV motif in LRRTM2 mediates binding to PSD-95.
To examine the role of LRRTM2 in synaptic function in vivo, we generated and stereotaxically injected lentiviruses expressing sh-LRRTM2 and GFP or a control vector into the dentate gyrus of P5-6 rat pups. Hippocampal slices were cut between P13 and P16 and simultaneous recordings were made from uninfected and nearby infected dentate granule cells. A stimulating electrode placed in the outer half of the molecular layer was used to stimulate perforant path (PP) inputs onto granule cells (Figure 3A, B). No differences in cellular measures of integrity such as input resistance were found between treatment conditions (data not shown). Simultaneous recordings showed that knockdown of LRRTM2 in granule cells revealed a 58% reduction in the strength of AMPAR-mediated EPSCs compared to neighboring uninfected cells (Figure 3C). In addition to AMPAR-mediated EPSCs, the strength of NMDAR-mediated PP inputs onto sh-LRRTM2 expressing granule cells was also significantly reduced by 54% (Figure 3D), indicating that endogenous LRRTM2 contributes significantly to excitatory synaptic transmission in vivo. LRRTM2 knockdown caused a small reduction in the ratio of AMPAR-mediated synaptic current to that carried by NMDARs (Figure 3E). LRRTM2 knockdown did not affect paired pulse ratios (PPRs), a measure of presynaptic release probability, of the same inputs (Figure 3F). Infection with a control virus did not affect synaptic function by any of these measures (Figure S4). Together, these results demonstrate that loss of LRRTM2 function in vivo leads to a reduction in the glutamatergic transmission.
The reduction in evoked EPSCs in granule cells expressing sh-LRRTM2 could indicate a reduced number of synapses onto granule cells or a reduced glutamate receptor density at these synapses. To gain more insight in the consequences of loss of LRRTM2 in vivo we recorded spontaneous AMPAR-mediated miniature excitatory postsynaptic currents (mEPSCs) from lentivirus infected granule cells. Whole-cell recordings were made from infected cells, identified by GFP fluorescence, and AMPA mEPSCs were isolated pharmacologically. Analysis of AMPAR-mediated mEPSCs showed that loss of LRRTM2 in granule cells caused a small, but significant decrease in mEPSC amplitude compared to control infected neurons (Figure S5A, B). The mEPSC frequency was not significantly affected in sh-LRRTM2 infected cells (Figure S5C). The decrease in mEPSC amplitude in sh-LRRTM2 infected granule cells indicates a reduction in strength of individual synapses and suggests that loss of LRRTM2 in hippocampal granule neurons reduces the density of postsynaptic AMPA receptors. The reduction in mEPSC amplitude, however, does not seem to be sufficient to account for the ~50% decrease in evoked synaptic currents (Figure 3), suggesting that LRRTM2 regulates both the efficacy of individual synapses and the number of synaptic inputs.
To identify the domains of LRRTM2 that mediate its synaptogenic activity, we carried out gain of function experiments with full length LRRTM2, and mutants that lack various domains. Overexpression of LRRTM2 in hippocampal cultures significantly increased the density of VGlut1/PSD-95 positive puncta along dendrites of transfected neurons compared to control neurons expressing GFP alone, indicating that LRRTM2 expression was sufficient to induce excitatory synapses (Figure S6A, B). In contrast, overexpression of LRRTM2 had no effect on the density of inhibitory synapses (Figure S6C, D).
To determine the relative contribution of the extra- and intracellular domains of LRRTM2 to excitatory synapse formation, hippocampal neurons were transfected with LRRTM2 deletion constructs lacking the intra- (LRRTM2 ΔC) or extracellular LRR (LRRTM2 ΔLRR) domains and immunostained for VGlut1 and PSD-95. Surface expression for all deletion mutant constructs was verified in 293T cells (data not shown). Overexpression of LRRTM2 ΔC resulted in a similar increase in excitatory synapse density as overexpression of full-length LRRTM2, but overexpression of the ΔLRR mutant abolished LRRTM2’s synaptogenic effect (Figure (Figure4A4A and S6E, F). This indicates that the LRR domain is required for the LRRTM2-mediated increase in excitatory synapse density. Neutravidin beads coated with LRRTM2-ecto-Fc induced clustering of VGlut1 puncta in contacting neurites, indicating that the ectodomain is sufficient to induce presynaptic differentiation (Figure S6G). To determine whether the binding partner for LRRTM2 was localized to excitatory synapses, we incubated hippocampal cultures with clustered LRRTM2-ecto-Fc or control Fc, and examined the localization of LRRTM2 binding sites by co-staining with anti-VGlut1 antibodies. LRRTM2-ecto-Fc and VGlut1 puncta were co-localized, suggesting that the LRRTM2 receptor is a synaptically-localized protein (Figure S6H).
The receptors for LRRTM proteins are not known. To identify the receptor for LRRTM2, we coupled LRRTM2-ecto-Fc or control Fc protein to proteinA beads, and column purified synaptosomal membrane proteins from P18 rat brains that bound Fc-LRRTM2. LRRTM2-associated proteins were then analyzed by MudPit tandem mass spectrometry (Washburn et al., 2001). We repeatedly found ~300 proteins to co-purify with the ecto domain of LRRTM2 while fewer than 100 proteins co-purified with Fc alone. Judging by the number of peptides and spectra hits, Neurexin family members were the most abundant proteins identified as potentially binding Fc-LRRTM2 (Figure 4B). We performed our screen with low (150mM) and high (500mM) salt washes, and under both conditions we found no Neurexin peptides in the Fc negative control indicating a high level specificity in our assay (Table S1). Since many of the peptides are common to different Neurexin proteins, these results alone do not identify a unique Neurexin protein as the major binding partner, but strongly suggest that a Neurexin isoform is an LRRTM2 receptor.
To determine whether LRRTM2 binds to specific Neurexin family proteins, we performed a series of experiments to examine the binding of Fc-LRRTM2 to the surface of 293T cells expressing different Neurexin isoforms (Figure 4C-F). 293T cells were transfected with the indicated isoforms, and 24 hr later the cells were exposed to clustered LRRTM2-ecto-Fc or control Fc proteins. Relative expression of Neurexin constructs is shown in Figure 4E. Following a 1 hr incubation, cultures were fixed and Fc binding assessed by immunofluorescence. As shown in Figure 4D, F there was no significant LRRTM2 binding to control transfected cells (GFP, N-Cadherin, NGL2). There was statistically significant binding to Neurexin 1α, but not Neurexin 2α or 3α. The highest level of binding was seen with Neurexin 1β (Figure 4D, F).
Neurexin 1β contains only one extracellular LNS domain that has previously been implicated in its interaction with Neuroligin. To determine whether the LNS domain of Neurexin 1β was specifically required for binding to LRRTM2, we tested two additional constructs in which the LNS domain of Neurexin 1β was replaced by the corresponding domain from Neurexin 2 or Neurexin 3. The Neurexin 2α1β construct showed significant, but highly reduced binding compared to Neurexin1β, whereas the Neurexin 3α1β construct showed no specific binding (Figure 4D, F). Consistent with a ligand-receptor relationship, Neurexins and LRRTM2 appear not to interact in cis, as they did not co-precipitate when co-expressed in 293T cells (data not shown). These results indicate that Neurexin 1α and Neurexin 1β are LRRTM2 receptors, and that the LNS domain of Neurexin 1β is critical for its ability to bind LRRTM2.
To estimate the binding affinity of LRRTM2 to Neurexin 1α and 1β, we examined relative binding of LRRTM2 at various concentrations to 293T cells expressing Neurexin constructs. Scatchard plot analysis indicates LRRTM2 binding affinities as ~16nM for Neurexin 1α and ~7nM for Neurexin 1β (Figure 4G, H). It should be noted, however, that these represent affinities of LRRTM2 for Neurexin 1α and 1β expressed on the cell surface; an accurate inter-molecular binding affinity would require analysis with recombinant purified proteins. To determine whether LRRTM2 directly binds to Neurexin 1α and 1β, we carried out a direct binding assay using recombinant LRRTM2 and Neurexin proteins. As shown in Figure 4I, J, both Neurexin 1α and 1β bound to LRRTM2-ecto-Fc immobilized on protein A beads and could be eluted from the column (lanes E2 and E3), indicating a direct interaction between these proteins.
Finally, to address whether Neurexin1 function was needed for LRRTM2-induced synaptic differentiation, we used the synapse induction assay in which 293T cells expressing LRRTM2 can induce presynaptic differentiation in co-cultured hippocampal neurons (Figure S2). Hippocampal neurons were electroporated with either GFP or GFP together with an shRNA directed against Neurexin 1. At DIV 5, 293T cells expressing GFP alone or GFP+myc-LRRTM2 were overlayed on the neuronal culture. These cocultures were fixed and immunostained for GFP to visualize transfected axon/293T overlap and synapsin to indentify presynaptic terminals, and the number of synaptic puncta per 10 μm of axon/293T overlap was quantified. As shown in Figure 5K,L[Fauxsmith2], LRRTM2 induced presynaptic differentiation in control hippocampal axons. However, this effect was abolished in hippocampal neurons expressing shNrxn1, indicating that Neurexin1 function is necessary for LRRTM2-induced presynaptic differentiation.
The observations reported here identify LRRTM2 as a key regulator of excitatory synapse formation. LRRTM2 localizes to the postsynaptic density of excitatory synapses and induces functional presynaptic differentiation in contacting axons in coculture assays. A role for LRRTM2 in regulating postsynaptic differentiation is supported by the fact that LRRTM2 binds PSD-95 and regulates AMPA receptor surface expression, and knockdown of LRRTM2 in vivo decreases the strength of glutamatergic synaptic transmission without affecting presynaptic properties. Overexpression of LRRTM2 in cultured hippocampal neurons increases the density of excitatory synapses, without affecting inhibitory synapse density and knockdown of LRRTM2 in cultured neurons selectively decreases excitatory synapse density. Finally, we find that Neurexin1 acts as a functional receptor for LRRTM2. Neurexin 1α and 1β both show isoform-specific binding to LRRTM2, and knockdown of Neurexin1 blocks the ability of LRRTM2 to induce presynaptic differentiation. Together, these observations indicate that LRRTM2-Neurexin1 interactions play a critical role in the formation of excitatory synapses (Figure 4M).
Our observations suggest that LRRTM2 regulates postsynaptic function by interacting with key postsynaptic components. LRRTM2 interacts with PSD-95, and knockdown of LRRTM2 in dissociated neurons leads to a reduction in the density of PSD-95 puncta. Our data suggest that LRRTM2 can recruit PSD-95 to the postsynaptic density by binding to PSD-95 via the cytoplasmic ECEV motif in LRRTM2. The interaction of LRRTM2 with PSD-95 is significant as PSD-95 plays an important role in regulating synaptic strength and plasticity by interacting with glutamate receptors and other synaptic signaling proteins (Kim and Sheng, 2004; Sheng and Hoogenraad, 2007; Irie et al., 1997; Kim et al., 2006; Ko et al., 2006; Wang et al., 2006; Han and Kim, 2008).
In further support of a postsynaptic role for LRRTM2, knockdown of LRRTM2 expression in cultured neurons strongly reduced synaptic GluR1 surface expression. This suggests that LRRTM2 recruits AMPARs to the postsynaptic density. One possibility is that LRRTM2 recruits AMPARs indirectly, through PSD-95 and TARPs (Chen et al., 2000; Schnell et al., 2002; Béïque et al., 2006; Elias et al., 2006). Alternatively, LRRTM2 could directly recruit AMPARs and NMDARs. We find that LRRTM2 can interact with AMPA and NMDA receptor subunits in heterologous cells via its extracellular domain, which suggests that one mechanism by which LRRTM2 promotes excitatory, but not inhibitory, synapse development could be through direct coupling with glutamate receptors in the postsynaptic membrane. This is an interesting possibility, but the interactions need to be confirmed for endogenous proteins.
In addition to regulating postsynaptic function, LRRTM2 is able to induce presynaptic differentiation. Our search for the presynaptic LRRTM2 receptor led to the identification of Neurexin 1α and Neurexin 1β as LRRTM2 receptors. This is striking, as these Neurexins have previously been identified as the major receptors for Neuroligin1 (Dean et al., 2003; Ichtchenko et al., 1995), and Neuroligin1 has been widely studied as a regulator of excitatory synapse formation (reviewed in Craig and Kang, 2007; Südhof, 2008). Thus it appears that neurexins function as receptors for two major classes of synaptogenic proteins. Finally, recent findings suggest that mutations in LRRTMs, neuroligins, and neurexins are associated with autism and other neurological and psychiatric disorders (Majercak et al., 2006; Francks et al., 2007; Edwards et al., 2009; Cohen and Greenberg, 2008; Südhof, 2008; Etherton et al., 2009). The identification of an LRRTM-Neurexin interaction as a key regulator of excitatory synapses should provide important insight into the molecular understanding of developmental cognitive disorders.
We thank members of the Ghosh lab for comments and helpful discussions, and Richard Huganir (Glutamate receptors), Ann-Marie Craig (Neurexins), Matthew Dalva (EphB2-YFP) and Eunjoon Kim (NGL-2) for plasmids. This work was supported by a Rubicon Fellowship from the Netherlands Organisation for Scientific Research (NWO) (JdW), NIH grant NS052772 (AG), and an Autism Speaks grant #2617 (DC).
Experimental Procedures Please see Supplemental Information for details of experimental procedures.
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.