Cbln1 Is a Ligand for an Orphan Glutamate Receptor δ2, a Bidirectional Synapse Organizer

See allHide authors and affiliations

Science  16 Apr 2010:
Vol. 328, Issue 5976, pp. 363-368
DOI: 10.1126/science.1185152


Cbln1, secreted from cerebellar granule cells, and the orphan glutamate receptor δ2 (GluD2), expressed by Purkinje cells, are essential for synapse integrity between these neurons in adult mice. Nevertheless, no endogenous binding partners for these molecules have been identified. We found that Cbln1 binds directly to the N-terminal domain of GluD2. GluD2 expression by postsynaptic cells, combined with exogenously applied Cbln1, was necessary and sufficient to induce new synapses in vitro and in the adult cerebellum in vivo. Further, beads coated with recombinant Cbln1 directly induced presynaptic differentiation and indirectly caused clustering of postsynaptic molecules via GluD2. These results indicate that the Cbln1-GluD2 complex is a unique synapse organizer that acts bidirectionally on both pre- and postsynaptic components.

Glutamate and its receptors mediate fast excitatory neurotransmission in the mammalian brain. Although the glutamate receptor δ2 (GluD2) belongs to the ionotropic glutamate receptor (iGluR) family, it is referred to as an orphan receptor because it has no known endogenous ligands. Nevertheless, GluD2 is essential for the normal development of cerebellar circuits; the numbers of parallel fiber (PF)–Purkinje cell synapses are specifically and dramatically reduced in the GluD2-null cerebellum (1). In addition, GluD2 rapidly induces synapse formation (2) and is essential for synapse maintenance in the adult cerebellum (3). Cbln1, a member of the C1q tumor necrosis factor superfamily (4), is expressed and secreted from cerebellar granule cells. The behavioral, physiological, and anatomical phenotypes of cbln1-null mice precisely mimic those of GluD2-null mice (5), which suggests that Cbln1 and GluD2 may be engaged in a common signaling pathway that is required for the formation and maintenance of PF synapses (6, 7). Nevertheless, a binding partner for Cbln1 has remained elusive. Therefore, we examined the hypothesis that GluD2 serves as a receptor for Cbln1 and that this receptor complex may regulate cerebellar synapse formation and maintenance.

First, we tested if hemagglutinin (HA)–tagged recombinant Cbln1 could bind to GluD2 using cell-based assays (8). Immunohistochemical (Fig. 1A) and immunoblot (Fig. 1B) analyses revealed that HA-Cbln1 could bind to human embryonic kidney (HEK) 293 cells expressing GluD2. In contrast, HA-Cbln1 did not bind to cells expressing the GluK2 kainate receptor, a different member of the iGluR family. To examine the GluD2 region responsible for this binding, we interchanged the extracellular domains of GluD2 and GluK2 (Fig. 1A). HA-Cbln1 could bind to HEK293 cells expressing GluD2ext-GluK2, in which the extracellular domain of GluD2 replaced that of GluK2, whereas no binding was observed for HEK293 cells expressing GluK2ext-GluD2, in which the extracellular domain of GluK2 replaced that of GluD2 (Fig. 1, A and B). To further narrow down the binding site for Cbln1 within the extracellular domain, which comprised an N-terminal domain (NTD) and a ligand-binding domain (LBD), we introduced various mutations in these domains. Most mutants did not reach the cell surface (fig. S1), except for GluD2∆NTD, in which the NTD was deleted (2). Nevertheless, no binding was observed for HA-Cbln1 in cells expressing GluD2∆NTD (Fig. 1, A and B).

Fig. 1

Cbln1 binds to the NTD of GluD2. (A and B) Cbln1 binding assay. Binding to HA-Cbln1 (red) on HEK293 cells (GFP, green) expressing GluD2, GluK2, or chimeric constructs was visualized by immunostaining (A) and quantified using immunoblot analysis (B). The mean amount of HA-Cbln1 bound to cells expressing GluD2 was defined as 100%. Three independent experiments. **P = 3.56 × 10−5. Scale bar, 50 μm. (C) Direct interaction between HA-Cbln1 and the NTD of GluD2. The NTD of iGluRs or the extracellular domain of CD4 was fused to Fc and coupled to protein G beads. Bound HA-Cbln1 was quantified using an immunoblot analysis. (D) Dose-dependent interaction between HA-Cbln1 and the NTD of GluD2 immobilized on beads. Each point represents the mean amount of bound HA-Cbln1 (filled circles) or HA-CS-Cbln1 (open circles) for three independent experiments, quantified using an immunoblot analysis. The line indicates a logistic equation. (E) Confocal images of Purkinje cells immunostained for calbindin (calb; green) and HA (red). Purkinje cell dendrites marked by the white boxes are enlarged below. Scale bars, 50 μm and 20 μm. (Right) HA-Cbln1 immunoreactivity along the dendrites is quantified. At least 27 cells for each group were analyzed in three independent experiments. **P = 2.37 × 10−16. (F) Wild-type (wt) and GluD2-null cerebella immunostained for Cbln1 (red) and calbindin (green). (Bottom) Enlarged images of the molecular layer. Arrows indicate Cbln1 immunoreactivity on the dendritic spines. Scale bars, 20 μm and 1 μm. (G) Postembedding immunogold EM images of endogenous Cbln1. Cbln1 and vesicular glutamate transporter 1 (a marker for parallel fiber) were immunolabeled with 10-nm (arrows) and 15-nm gold particles, respectively. Arrowheads indicate the edges of the PSD on Purkinje cell spines (sp). Scale bar, 100 nm.

To confirm whether the NTD of GluD2 (GluD2NTD) was a binding site for Cbln1, we purified GluD2NTD by fusing it to an immunoglobulin Fc fragment. The NTDs of GluK2, the AMPA receptors GluA1 and GluA2, and the extracellular domain of the CD4 lymphocyte surface protein were used as controls. Immunoblot analysis showed that HA-Cbln1 bound specifically to GluD2NTD-Fc (Fig. 1C). The amount of bound HA-Cbln1 increased with high doses of HA-Cbln1 (apparent binding dissociation constant of ~167 nM) (Fig. 1D).

We assumed that if GluD2 is a receptor for Cbln1, the binding of exogenously applied Cbln1, as well as endogenous Cbln1, should be reduced for GluD2-null Purkinje cells. First, we incubated cerebellar cultures [14 to 16 days in vitro (DIV)] with HA-Cbln1 and performed immunohistochemical analysis with antibodies against HA and calbindin (a marker for Purkinje cells). A mutant Cbln1 (HA-CS-Cbln1), in which hexamer formation was disrupted by replacing two cysteines with serines (9), was used as a negative control (10). Strong punctate immunoreactivity for HA-Cbln1, but not for HA-CS-Cbln1, was observed along the dendrites of wild-type Purkinje cells. In contrast, HA-Cbln1 did not bind to GluD2-null Purkinje cells (Fig. 1E). Next, we used antigen-exposing methods, i.e., pepsin pretreatment for light microscopy and postembedding immunogold for electron microscopy, to visualize endogenous Cbln1 in the cerebellum (11) using specific antibody (fig. S2). Immunohistochemical analysis of cerebellar slices detected punctate Cbln1 immunoreactivity in the molecular layer of the wild-type, but not the GluD2-null, cerebellum (Fig. 1F). Similarly, postembedding immunogold electron microscopic (EM) analysis found that Cbln1 immunoreactivity, which was highly abundant at PF–Purkinje cell synaptic clefts in the wild-type cerebellum [7.8 ± 0.7 particles/μm of postsynaptic density (PSD), n = 100 spines; versus 0.01 ± 0.01 particles/μm of nonsynaptic membranes], was essentially absent in the GluD2-null cerebellum (0.2 ± 0.1 particles/μm of PSD, n = 70 spines) (Fig. 1G). In contrast, immunoblot analysis indicated that the Cbln1 protein levels for wild-type and GluD2-null cerebella were similar (fig. S3) (7), which suggested that endogenous Cbln1 may have been washed away from the GluD2-null cerebellum during the immunostaining procedures.

To identify a role for Cbln1-GluD2 interactions in synapse formation, we performed a neuron–HEK293 cell coculture assay. GluD2 expressed in HEK293 cells has been shown to induce synapse formation with cocultured wild-type granule cells in vitro (Fig. 2A) (2, 12, 13). However, HEK293 cells expressing GluD2 did not accumulate in the synaptic terminals (immunopositive for synaptophysin) of cbln1-null granule cells (Fig. 2A). Similarly, HEK293 cells expressing GluD2∆NTD did not accumulate in the synaptic terminals of wild-type granule cells (Fig. 2A). In contrast, when HEK293 cells expressing GluD2, but not those expressing GluK2, were incubated with HA-Cbln1, they were able to form synapses even with cbln1-null granule cells (Fig. 2A). Although applying HA-Cbln1 induced PF synapse formation in dissociated cbln1-null Purkinje cells in culture, it failed to do so in GluD2-null or cbln1- and GluD2-double null (cbln1/GluD2-null) Purkinje cells (fig. S4). Finally, wild-type Purkinje cells incubated with GluD2NTD-Fc showed a significant reduction in synaptophysin immunoreactivity on their dendrites compared with those treated with HA-Fc (fig. S5), which suggested that GluD2NTD-Fc inhibited synaptogenesis, probably by disrupting endogenous Cbln1-GluD2 interactions.

Fig. 2

Cbln1 and the NTD of GluD2 cooperatively induce synaptogenesis in vitro and in vivo. (A) HEK293 cells expressing GluD2, GluD2∆NTD, or GluK2 and GFP were cocultured with wild-type (wt) or cbln1-null cerebellar granule cells with or without HA-Cbln1 or HA-CS-Cbln1. HEK293 cells were immunostained for GFP (green) and synaptophysin (Syn; red) and the mean intensities of synaptophysin immunoreactivity in the GFP-positive area are quantified. Scale bar, 50 μm. At least 24 fields were analyzed in three independent experiments. **P = 9.22 × 10−22. (B) GluD2NTD and Cbln1 induced functional synapses of granule cells. GluD2NTD-Fc or HA-Fc was conjugated with protein G beads (green) and added to cbln1-null cerebellar granule cells with or without HA-Cbln1. Functional presynaptic terminals were labeled with FM4-64 (red), and the mean intensities in the bead areas are quantified. Arrows indicate FM4-64 fluorescence around the beads. Scale bar, 20 μm. At least 150 beads were analyzed for each group in two independent experiments. **P = 9.52 × 10−49. (C) EM analysis on the effect of HA-Cbln1. Two days after we injected HA-Cbln1, the percentage of normal synapses (asterisks) was counted in cbln1-null, GluD2-null, and cbln1/GluD2-null mice. Free spines and mismatched synapses are indicated by f and m, respectively. Scale bars, 500 nm. **P = 5.14 × 10−4. (D) Functional restoration of PF synapses. PF-evoked EPSC traces (middle traces) were recorded in cbln1/GluD2-null Purkinje cells transduced with Sin-GFP or Sin-GluD2-GFP after the subdural injection of HA-Cbln1 in acutely prepared cerebellar slices. GluD2 (red) is expressed in GFP (green)–positive Purkinje cell spines (arrows). Scale bars, 20 μm and 2 μm. The averaged input-output relationship of PF-EPSCs for each condition (n = 30 each) is summarized in the lower graph. **P = 0.0032 (+HA-Cbln1+GluD2 versus +cont.), **P = 0.001 (+HA-Cbln1+GluD2 versus +GluD2) at 200 μA.

To determine whether synapses that formed by the Cbln1-GluD2 interactions were functional, we assessed the uptake of the fluorescent styryl dye FM4-64 into presynaptic vesicles (14). We incubated cbln1-null granule cells with protein G beads coupled to GluD2NTD-Fc or HA-Fc along with HA-Cbln1. Significant FM4-64 fluorescence was detected on the beads coupled to GluD2NTD-Fc only when HA-Cbln1 was coapplied (Fig. 2B).

To determine whether the Cbln1-GluD2 complex was also capable of inducing functional synapses in vivo, we injected HA-Cbln1 into the subarachnoid supracerebellar space above the cerebellar vermis in adult cbln1-null (10), GluD2-null, or cbln1/GluD2-null mice. Cbln1/GluD2-null mice displayed ataxia (fig. S6) and a reduced number of PF-Purkinje cell synapses (Fig. 2C) (5). Although a single injection of HA-Cbln1 (1 μg/g body weight) restored the ultrastructures of PF–Purkinje cell synapses in adult cbln1-null mice (10), it was ineffective in the GluD2-null or cbln1/GluD2-null cerebellum (Fig. 2C). We also assessed the recovery of functional PF–Purkinje cell synapses by recording PF-evoked excitatory postsynaptic currents (PF-EPSCs) using the whole-cell recordings in cerebellar slices. We injected Sindbis virus carrying a gene encoding GluD2 with green fluorescent protein (GFP) (Sin-GluD2-GFP) into the subarachnoidal space of cbln1/GluD2-null mice (aged 26 to 33 days) with or without HA-Cbln1 (Fig. 2D). Although the expression of GluD2 restored the reduced PF-EPSCs in adult GluD2-null mice (2), it was ineffective for cbln1/GluD2-null Purkinje cells (Fig. 2D). When HA-Cbln1 was injected together with Sin-GluD2-GFP, cbln1/GluD2-null Purkinje cells displayed PF-EPSCs comparable to those in wild-type Purkinje cells, and these were significantly larger than those transduced with Sin-GluD2-GFP or Sin-GFP [n = 30 each; analysis of variance (ANOVA) followed by Bonferroni test, P < 0.001] (Fig. 2D).

To further define the roles of each component of the Cbln1-GluD2 complex for organizing synapses, we used a bead-induced synaptic differentiation assay. First, to examine if Cbln1 alone could induce presynaptic terminals, we incubated cbln1-null granule cells with beads coated with HA-CS-Cbln1 or HA-Cbln1 (Fig. 3A). Immunocytochemical analysis using an antibody against synapsin I (a marker for presynaptic terminals) revealed that presynaptic terminals accumulated only on beads coated with HA-Cbln1, even in the absence of GluD2 (Fig. 3A). In addition, significant FM4-64 fluorescence was detected on the beads coated with HA-Cbln1 (Fig. 3B). Further, synapsin I–immunopositive terminals accumulated around HA-Cbln1–coated beads at extrasynaptic sites that lacked endogenous AMPA receptor clusters (Fig. 3C).

Fig. 3

Cbln1 is a direct presynaptic organizer. (A to C) HA-Cbln1 or HA-CS-Cbln1 was conjugated with beads and added to cbln1-null granule cells at 8 DIV as illustrated in the schematic diagram. (A) HA-Cbln1 immobilized on beads was sufficient for the accumulation of presynaptic terminals. (Left) Confocal images of granule cells (13 DIV) immunostained for synapsin I (red) and beads (green). Regions marked by the white boxes are enlarged below. Scale bars, 50 μm and 20 μm. The averaged intensity of synapsin I in the bead area is quantified on the right. At least 30 fields were analyzed in three independent experiments. **P = 1.05 × 10−11. (B) HA-Cbln1 immobilized on beads was sufficient for the accumulation of functional presynaptic terminals. At 13 DIV, functional presynaptic terminals were labeled with FM4-64 (red). Scale bar, 20 μm. (C) Synaptic terminals were directly induced by HA-Cbln1–coated beads. Synapsin I–immunopositive terminals (red) were induced around HA-Cbln1–coated beads (arrowheads), which were located at extrasynaptic sites lacking endogenous AMPA receptors (pan GluA, green). Scale bar, 20 μm.

To determine whether the Cbln1-GluD2 complex could also convey signals to postsynaptic sites, we incubated cbln1-null Purkinje cells with beads coated with HA-Cbln1. Immunocytochemical analysis showed that GluD2 clustered in Purkinje cells around beads coated with HA-Cbln1, but not those coated with HA-CS-Cbln1 (not shown) or beads alone (Fig. 4A). GluD2 clustering was also induced around HA-Cbln1–coated beads at extrasynaptic sites that lacked endogenous synapsin I–immunopositive presynaptic terminals (fig. S7). Further, HA-Cbln1–coated beads induced clustering of GluD2 and GluD2ext-GluK2, but not GluD2∆NTD, GluK2, or GluK2ext-GluD2, in HEK293 cells (fig. S8).

Fig. 4

Cbln1 directly promotes the clustering of GluD2 and other postsynaptic molecules. (A to C) Control or HA-Cbln1–conjugated beads were added to cbln1-null Purkinje cells at 7 DIV, as illustrated in the schematic diagram. (A) HA-Cbln1 immobilized on beads caused the clustering of endogenous GluD2 (red) in Purkinje cells (calbindin; green) at 10 DIV. Dendrites marked by the white boxes in the top panels are enlarged below. Scale bars, 50 μm and 20 μm. (B) HA-Cbln1–coated beads (blue) caused the clustering of GluD2 (green) and indicated postsynaptic proteins (red) in cbln1-null Purkinje cells. Arrowheads indicate accumulated GluD2 around the beads. Scale bar, 20 μm. (C) Accumulation of postsynaptic molecules induced by HA-Cbln1–coated beads requires GluD2. HA-Cbln1–conjugated beads (blue) were added to cbln1-null or cbln1/GluD2-null Purkinje cells and immunostained for calbindin (green) and each postsynaptic marker (red). Scale bar, 20 μm. (D) The density of endogenous GluD2 was reduced in cbln1-null synapses in vivo. GluD2 immunopositive particles (blue) were visualized using the SDS-FRL method in the molecular layer of the wild-type or cbln1-null cerebellum. The density of GluD2 in intact synapses accompanied by the presynaptic protoplasmic face is plotted as an occurrence histogram. At least three replicates were analyzed for two mice. **P = 2.55 × 10−12. (E) Proposed mechanism for Cbln1-GluD2 signaling as a bidirectional synaptic organizer.

The C terminus of GluD2 interacts directly with several intracellular molecules, such as shank-2 (15) and PSD-93/95 (16); many of these serve as scaffolds for other postsynaptic molecules, including homer-3, transmembrane AMPA receptor regulatory protein (TARP), and AMPA glutamate receptors (GluAs). Thus, we examined if the clustering of GluD2 induced by the Cbln1-coated beads might accumulate with other postsynaptic molecules in Purkinje cells. With GluD2, shank-2, homer-3, and GluA2 clustered in Purkinje cells around the beads coated with HA-Cbln1 (Fig. 4B). In contrast, HA-Cbln1–coated beads did not accumulate gephyrin [an anchoring protein for the γ-aminobutyric acid (GABA) receptor] (Fig. 4B) or excitatory amino acid transporter 4 (EAAT4, a neuronal glutamate transporter) (fig. S9) in Purkinje cells. HA-Cbln1–coated beads did not induce clustering of shank-2 or GluA2 in cbln1/GluD2-null Purkinje cells (Fig. 4C). Shank-2 and PSD-95 accumulated around HA-Cbln1–coated beads only when the responsible C-terminal domains of GluD2 were intact (fig. S10).

To further identify a role for Cbln1 as a postsynaptic organizer in vivo, we examined if the distribution of GluD2 was affected in cbln1-null Purkinje cells using the SDS-digested freeze-fracture replica labeling (SDS-FRL) method, which has a nearly one-to-one detection sensitivity for each iGluR on the surface of the postsynaptic membrane specialization (17). To exclude a possible effect of the presence of noninnervated spines in the cbln1-null cerebellum (Fig. 2C), we counted the number of immunoparticles detected by GluD2-specific antibody (fig. S11) in intact synapses, which were accompanied by the presynaptic protoplasmic face. The number of GluD2 immunoparticles located on postsynaptic membranes was significantly reduced in cbln1-null Purkinje cells (Fig. 4D, P < 0.001), which indicated that Cbln1 serves as a postsynaptic organizer in vivo and contributes to the clustering of postsynaptic GluD2.

We have demonstrated that Cbln1 is a ligand for the orphan receptor GluD2. Among known synapse-organizing molecules, such as neuroligin-neurexin (18), SynCAM-SynCAM (19), EphrinB-EphB (20), fibroblast growth factor (FGF) 22–FGF receptor 2b (21), Narp-GluAs (22), and netrin-G ligand-3 and leukocyte common antigen-related (NGL-3–LAR) (23), Cbln1-GluD2 signaling is unique in that without each component, synapse formation was severely abrogated in the cerebellum in vivo as well as in heterologous cells in vitro. Its bidirectional mode of action is also unique; at synaptic junctions, presynaptically derived Cbln1 accumulates and directly induces presynaptic differentiation, possibly by interacting with unidentified proteins on the presynaptic membrane (Fig. 4E). Because beads coated with HA-Cbln1 induced accumulation of functional presynaptic terminals (Fig. 3), GluD2 may simply serve as a scaffold to accumulate and stabilize Cbln1 at synaptic junctions. Conversely, Cbln1 probably serves as a postsynaptic organizer by clustering GluD2, which may regulate synaptic plasticity via its interacting intracellular proteins (24).

Cbln1 is also expressed in various brain regions where GluD2 is not expressed, such as the olfactory bulb, the entorhinal cortex, and certain thalamic nuclei (25), which indicates that Cbln1 may bind to other receptors in these regions. An alternative candidate receptor is GluD1, which is highly expressed in these brain regions, especially during development (26). Indeed, HA-Cbln1 could bind to HEK293 cells that expressed GluD1 (fig. S12A) or beads coated with GluD1NTD-Fc (fig. S12B). Furthermore, other Cbln family proteins (Cbln2 and Cbln4) are expressed in various brain regions (25). Therefore, further studies are warranted to elucidate the synaptic roles of Cbln and GluD family proteins in normal and pathological conditions in the CNS.

Supporting Online Material

Materials and Methods

Figs. S1 to S12


References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. We thank J. Motohashi for excellent technical support. This work was supported by a grant-in-aid from Japanese Government Ministry of Education, Culture, Sports, Science and Technology (MEXT) (M.Y.), the Takeda Science Foundation (M.Y.), the Naito Memorial Grant for Female Researchers (K.M.), and the Core Research for Evolutional Science and Technology from the Japanese Science and Technology Agency (M.Y.).
View Abstract

Stay Connected to Science

Navigate This Article