Cbln family proteins promote synapse formation by regulating distinct neurexin signaling pathways in various brain regions

Authors

  • Keiko Matsuda,

    1. Department of Physiology, School of Medicine, Keio University, 35 Shinanomachi, Shinjuku-ku, Tokyo 160-8582, Japan
    2. Core Research for Evolutional Science and Technology (CREST), Japan Science and Technology Corporation, Saitama, Japan
    Search for more papers by this author
  • Michisuke Yuzaki

    1. Department of Physiology, School of Medicine, Keio University, 35 Shinanomachi, Shinjuku-ku, Tokyo 160-8582, Japan
    2. Core Research for Evolutional Science and Technology (CREST), Japan Science and Technology Corporation, Saitama, Japan
    Search for more papers by this author

M. Yuzaki, Department of Physiology, as above.
E-mail: myuzaki@a5.keio.jp

Abstract

Cbln1 (a.k.a. precerebellin) is a unique bidirectional synaptic organizer that plays an essential role in the formation and maintenance of excitatory synapses between granule cells and Purkinje cells in the mouse cerebellum. Cbln1 secreted from cerebellar granule cells directly induces presynaptic differentiation and indirectly serves as a postsynaptic organizer by binding to its receptor, the δ2 glutamate receptor. However, it remains unclear how Cbln1 binds to the presynaptic sites and interacts with other synaptic organizers. Furthermore, although Cbln1 and its family members Cbln2 and Cbln4 are expressed in brain regions other than the cerebellum, it is unknown whether they regulate synapse formation in these brain regions. In this study, we showed that Cbln1 and Cbln2, but not Cbln4, specifically bound to its presynaptic receptor –α and β isoforms of neurexin carrying the splice site 4 insert [NRXs(S4+)] – and induced synaptogenesis in cerebellar, hippocampal and cortical neurons in vitro. Cbln1 competed with synaptogenesis mediated by neuroligin 1, which lacks the splice sites A and B, but not leucine-rich repeat transmembrane protein 2, possibly by sharing the presynaptic receptor NRXs(S4+). However, unlike neurexins/neuroligins or neurexins/leucine-rich repeat transmembrane proteins, the interaction between NRX1β(S4+) and Cbln1 was insensitive to extracellular Ca2+ concentrations. These findings revealed the unique and general roles of Cbln family proteins in mediating the formation and maintenance of synapses not only in the cerebellum but also in various other brain regions.

Introduction

Presynaptic neurexins (NRXs) and postsynaptic neuroligins (NLs) are the best-known trans-synaptic cell adhesion molecules (Craig & Kang, 2007) that are associated with various psychiatric and neurodevelopmental disorders (Sudhof, 2008). In mammals, three NRX genes, each producing long NRXαs and short NRXβs in multiple splice forms, are present (Ullrich et al., 1995). NLs, encoded by four genes in rodents, also undergo alternative splicing (Ichtchenko et al., 1996). Such diversity suggests that different combinations of NRX/NL isoforms may mediate distinct recognition events between neurons (Craig & Kang, 2007). Indeed, NRXβs carrying the splice site 4 insert [NRXβ(S4+)] were reported to preferentially bind to NLs that lacked splice site B, such as NL1(−), and promote inhibitory synapse formation (Chih et al., 2006; Graf et al., 2006). In contrast, NRXαs and NRXβs lacking the splice site 4 insert [NRXα(S4−) and NRXβ(S4−), respectively] also bind to NLs carrying the splice site B insert and also promote excitatory synapse formation. Recently, leucine-rich repeat transmembrane proteins (LRRTMs) were shown to bind to presynaptic NRXα(S4−) and NRXβ(S4−) receptors, leading to excitatory-specific synapse formation (Ko et al., 2009; de Wit et al., 2009; Siddiqui et al., 2010). Nevertheless, the density of excitatory or inhibitory synapses is not severely reduced in NL- or LRRTM1-null mice (Varoqueaux et al., 2006; Linhoff et al., 2009). Therefore, the exact roles of the interactions of NRXs/NLs and NRXs/LRRTMs in synapse formation remain unclear.

Cbln1 is one of the most recently identified bidirectional synaptic organizers. Cbln1 is secreted from cerebellar granule cells and highly accumulated in the synaptic cleft of parallel fiber (PF)–Purkinje cell synapses (Hirai et al., 2005; Miura et al., 2009). It directly induces presynaptic differentiation and indirectly serves as a postsynaptic organizer by binding to its receptor, the δ2 glutamate receptor (GluD2), which is specifically expressed in cerebellar Purkinje cells (Matsuda et al., 2010); the number of excitatory synapses between PFs (axons of granule cells) and Purkinje cells is severely reduced in cbln1- or GluD2-null cerebellum (Yuzaki, 2009). However, Cbln1 and other Cbln family proteins are expressed in various brain regions (Miura et al., 2006) where GluD2 is not expressed. Therefore, it remains unclear whether and how Cbln family proteins are involved in synaptic functions in these brain regions. The more fundamental question is how Cbln1 binds to presynaptic sites. The mechanism by which the Cbln1/GluD2 pathway interacts with other synaptic organizers, such as NRXs/NLs and NRXs/LRRTMs, remains unclear.

In this study, we showed that Cbln1 and Cbln2 but not Cbln4 bound to presynaptic NRX1α(S4+) and NRXβs(S4+) and induced synaptogenesis in cultured cerebellar, hippocampal and cortical neurons. Cbln1 competed with synaptogenesis mediated by NL1(−) but not by LRRTMs, possibly by sharing the presynaptic receptor NRX(S4+). However, unlike NRXs/NLs or NRXs/LRRTMs, the interaction between NRX1β and Cbln1 was insensitive to extracellular Ca2+ concentrations. These findings revealed the unique and general roles of Cbln family proteins in mediating the formation and maintenance of synapses, not only in the cerebellum but also in various other brain regions.

Materials and methods

cDNA constructs

cDNA encoding hemagglutinin (HA) was added to the 5′ end of mouse Cbln1, Cbln2 and Cbln4 cDNA (Iijima et al., 2007; Matsuda et al., 2009). Furthermore, two cysteine residues at amino acid positions 34 and 38 were replaced with serine residues to create CS-Cbln1, as previously described (Matsuda et al., 2009). Rat cDNA encoding GluD2 was a gift from Dr J. Boulter (University of California at Los Angeles, Los Angeles, CA, USA). Mouse cDNAs encoding NL1(−) and NRX2β were gifts from Dr P. Sheiffele (University of Basel, Basel, Switzerland). cDNA encoding Flag was added to the 3′ end of mouse NRXs or LRRTM2 cDNA. For green fluorescent protein (GFP)-tagged NL1(−), cDNA encoding enhanced GFP was inserted between amino acids 776 and 777. For immunoglobulin Fc fragment-fusion constructs, the N-terminal domain (NTD) of GluD2 (amino acids 1–430), the extracellular domain of NRX1β(S4+) (amino acids 1–393), LRRTM2 (amino acids 1–421) or NL1(−) (amino acids 1–696) and CD4 (a gift from Dr Y. Oike, School of Medicine, Keio University, Tokyo, Japan) were added immediately before the Fc fragment of human IgG1. The cDNA constructs were cloned in pCAGGS vector (provided by Dr J. Miyazaki, Osaka University, Osaka, Japan).

Preparation of recombinant proteins

The HA-tagged Cblns or Fc fusion proteins were expressed in human embryonic kidney (HEK)293 tSA cells (a gift from Dr R. Horn, Thomas Jefferson University Medical School, Philadelphia, PA, USA) as previously described (Matsuda et al., 2009). The concentration of each recombinant protein was quantified by immunoblot analyses with purified 6 × histidine-tagged HA-Cbln1 or purified TrkB-Fc (R&D Systems, Inc., Minneapolis, MN, USA) as the standard (Ito-Ishida et al., 2008). HA-Cbln1, 2 or 4, or Fc fusion proteins were incubated with biotinylated anti-HA (BIOT-101L mouse; Covance Research Products, Berkeley, CA, USA) or biotinylated anti-Fc (609-1602 goat; Rockland Immunochemicals, Gilbertsville, PA, USA) and then immobilized to avidin beads (Dynabeads M-280 Streptavidin; Invitrogen).

Cell cultures

Mixed cerebellar cultures were prepared from embryonic day 17 to day-of-birth ICR or cbln1-null mice as previously described (Matsuda et al., 2009). Cells were plated at a density of 2 × 105 cells on plastic coverslips (13.5 mm in diameter) and maintained in Dulbecco’s modified Eagle medium/F12 containing 100 μm putrescine, 30 nm sodium selenite, 0.5 ng/mL tri-iodothyronine, 0.25 mg/mL bovine serum albumin, 3.9 mm glutamate and N3 supplement (100 μg/mL apotransferrin, 10 μg/mL insulin and 20 nm progesterone) in 5% CO2 at 37 °C. Dissociated cultures of hippocampal or cortical neurons were prepared from embryonic day 17–18 mice as previously described Forrest et al., 1994) and maintained in Neurobasal medium supplemented with NS21 (Chen et al., 2008) and l-glutamine (Invitrogen). Cultured neurons were transfected at 7–8 days in vitro (DIV) using Lipofectamine 2000 (Invitrogen). HA-Cbln or NRX1β beads were added to the culture medium at 8–11 DIV and incubated for 3–4 days. Heterologous synapse formation assays were performed using HEK293 cells as previously described (Kakegawa et al., 2009). Briefly, HEK293 cells transfected with each cDNA construct were added to primary cultured neurons at 7 DIV and cocultured for 4 days in the presence of 5-fluoro-2-deoxyuridine (10 μm) with or without recombinant Cbln1. The mice were killed by decapitation under tribromoethanol anesthesia (125–250 mg/kg body weight). All animal care and treatment procedures were approved by the Animal Resource Committee of the School of Medicine, Keio University.

Cbln1-binding assay using intact cells

The HEK293 cells expressing Flag-tagged NRX were incubated with HA-Cbln1 (2 μg/mL) or Fc fusion proteins (2 μg/mL) [i.e. NL1(−)-Fc or LRRTM2 fused to the Fc fragment (LRRTM2-Fc)] for 4 h and reacted with anti-HA antibody or Alexa 546-conjugated anti-human IgG (1 : 2000; Invitrogen) without permeabilization to selectively stain proteins on the cell surface. For immunoblot analyses, cells treated with HA-Cbln1 for 4 h were washed four times with ice-cold phosphate-buffered saline (PBS) and solubilized in a buffer (50 mm Tris–HCl, pH 8.0, 50 mm NaCl, 20 mm EDTA, 1% Nonidet P-40) containing 0.1% sodium dodecyl sulfate. The cell lysate was subjected to immunoblot analyses using anti-HA antibody. HEK293 cells expressing GFP-NL1(−) were incubated with NRX1β(S4+ or S4−)-Fc (5 μg/mL) for 4 h in the presence or absence of HA-Cbln1 (40 μg/mL). Alternatively, HEK293 cells expressing NRX1β(S4+ or S4−) were incubated with NL1(−)-Fc (2 μg/mL) for 4 h in the presence or absence of HA-Cbln1 or CS-Cbln1. Cells were then incubated with Alexa 546-conjugated anti-human IgG without permeabilization.

Immunohistochemistry

Cells in dissociated cultures were fixed using PBS containing 4% paraformaldehyde for 20 min on ice, followed by 100% methanol at −20 °C for 10 min for immunostaining synaptic markers (synapsin I or synaptophysin). After permeabilization with 0.4% Triton X-100 in PBS containing 2% bovine serum albumin and 2% normal goat serum for 1 h at room temperature (22 °C), cells were treated with primary antibodies (see below) and subsequently treated with secondary antibodies that were conjugated with Alexa 405, 488 or 546 (1 : 2000; Invitrogen). Fluorescence images were captured using a CCD camera attached to a fluorescence or confocal microscope. To quantify the accumulation of each synaptic marker on transfected HEK293 cells or on HA-Cbln1 beads, or to quantify the intensity of NRX1β(S4+ or S4−), NL1(−), LRRTM2-Fc or HA-Cbln1 bound on transfected HEK293 cells, images were randomly captured in eight or more fields (each field corresponds to 450 × 600 μm containing at least five transfected HEK293 cells) using fixed gains and exposures for each fluorescent channel. The images were analyzed using IP-lab software (version 3.61). GFP- or Flag-immunopositive cell areas or bead areas were selected using macro auto-segmentation. The intensity of immunoreactivity within the segmented area was averaged and background immunoreactivity within the nonsegmented area was subtracted.

Functional labeling of presynaptic terminals with FM4-64

Cultured hippocampal neurons were incubated with HA-Cbln1-coated beads from 13 to 17 DIV. Functional presynaptic terminals were labeled with FM4-64 (Invitrogen) as previously described (Iijima et al., 2009). Briefly, cells were incubated with Tyrode’s solution (20 mm HEPES, pH 7.2, 30 mm glucose, 129 mm NaCl, 5 mm KCl) for 15 min at room temperature and treated for 5 min with Tyrode’s solution to which 5 μm FM4-64, 80 mm KCl and 4 mm CaCl2 were added. Immediately after FM4-64 loading, cells were fixed using PBS containing 4% paraformaldehyde and beads were stained with Alexa 488-conjugated anti-mouse IgG (Invitrogen).

Binding assay using immobilized beads in vitro

The NRX1β(S4+ or S4−)-Fc or CD4-Fc was immobilized on magnetic protein G beads (Dynabeads Protein G; Invitrogen) and incubated overnight in the presence of HA-Cbln1 (2 μg/mL) in cerebellar culture medium containing 1.4% bovine serum albumin. Bound HA-Cbln1 was recovered by magnetic separation and washed four times with ice-cold PBS. The final pellet was analyzed by immunoblotting using anti-HA antibody. HA-Cbln1 was incubated with anti-HA antibody and conjugated to magnetic avidin beads. HEK293 cells expressing Flag-tagged NRX1β(S4+) were solubilized in PBS containing 1% Triton X-100, and its supernatant was incubated with immobilized Cbln beads. Bound NRX1β(S4+) was recovered by magnetic separation and washed four times with 1% Triton X-100 in PBS. The final pellet was analyzed by immunoblotting using anti-Flag antibody.

Primary antibodies

The following dilutions of antibodies were used: anti-GFP (AB16901 chicken, 1 : 2000; Millipore, Temecula, CA, USA), anti-synaptophysin (S5768 mouse, 1 : 500; Sigma), anti-HA (MMS-101P mouse, 1 : 1000; Covance Research Products), anti-Flag (F3165 mouse, 1 : 1000 and F7425 rabbit, 1 : 1000; Sigma), anti-actin (A4700 mouse, 1 : 1000; Sigma), anti-Fc (I9135 rabbit, 1 : 1000; Sigma), anti-synapsin I (AB1543 rabbit, 1 : 1000; Millipore), anti-pan α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate (AMPA) receptors (guinea pig, 1 : 500) (Fukaya et al., 2006), anti-GluD2 (rabbit, 1 : 2000 and guinea pig, 1 : 250) (Takeuchi et al., 2005), anti-calbindin (C8666 mouse, 1 : 1000; Sigma), anti-shank2 (rabbit; 1 : 500) (Matsuda et al., 2010) and anti-Cbln1 (rabbit; 1 : 300) (Iijima et al., 2007). Antibody against NRX (chicken; 1 : 500) (Dean et al., 2003) was kindly provided by Dr P. Sheiffele.

Data analyses and statistics

Data are presented as the mean ± SEM and statistical significance was defined as < 0.05 as determined using anova or the Kruskal–Wallis test followed by the Bartlett test for multiple comparisons or paired Student’s t-test.

Results

Cbln1 inhibits synapse formation induced by NL1, but not by LRRTM2

To clarify how Cbln1 interacts with other synaptic organizers, such as NRXs/NLs and NRXs/LRRTMs, we performed artificial synapse-forming assays using HEK293 cells and cbln1-null granule cells. We previously reported that HEK293 cells expressing GluD2 accumulated synaptophysin-positive presynaptic terminals of cbln1-null granule cells when recombinant HA-Cbln1 protein was added to the culture medium (Matsuda et al., 2010). Similarly, HEK293 cells expressing NL1(−), which lacked splice sites A and B, or HEK293 cells expressing LRRTM2 accumulated synaptophysin-positive presynaptic terminals of cbln1-null granule cells (Fig. 1A). Thus, presynaptic terminals of granule cells probably express NRX isoforms that could bind to both NL1(−) and LRRTM2. Interestingly, when HA-Cbln1 was applied to HEK293 cells expressing NL1(−), synaptogenesis was significantly inhibited (Fig. 1A). In contrast, HA-Cbln1 did not affect synaptogenesis observed in HEK293 cells expressing LRRTM2 (Fig. 1A). HA-Cbln1 did not directly bind to HEK293 cells expressing NL1(−) or LRRTM2 (data not shown). LRRTM2 is reported to bind to NRXβ(S4−), which lacks a splice site 4 insert, whereas NL1(−) binds to both NRXβ(S4−) and NRXβ(S4+) (Boucard et al., 2005; Ko et al., 2009). Indeed, presynaptic terminals of cbln1-null granule cells accumulated on HEK293 cells expressing LRRTM2 were preferentially inhibited by NRX1β(S4−)-Fc. In contrast, synaptogenesis induced by NL1(−)cells was preferentially inhibited by NRX1β(S4+)-Fc (Supporting Information Fig. S1). Therefore, we hypothesized that Cbln1 may interact with NRXβ(S4+) expressed at presynaptic sites in granule cells and, thus, specifically interfere with NL1(−)-induced synaptogenesis.

Figure 1.

 Cbln1 interferes with synaptogenesis induced by NL1(−) but not by LRRTM2 by competing with NRX1β(S4+). (A) Effects of HA-Cbln1 on artificial synapse formation induced by NL1(−) or LRRTM2. HEK293 cells expressing GFP-NL1(−) or LRRTM2 plus GFP were cocultured with cbln1-null cerebellar granule cells. Representative confocal images of cells immunostained for synaptophysin (Syn) (red or white) and GFP (green) are shown. Application of HA-Cbln1 (40 μg/mL) blocked NL1 but not LRRTM2 synaptogenic activity. Scale bar, 25 μm. Mean intensities of synaptophysin immunoreactivity in the GFP-positive area in the presence of HA-Cbln1 are normalized to those in the absence of HA-Cbln1 and are shown in the graph at the right. Error bars represent SEMs. At least = 26 fields were analyzed in three independent experiments. **= 7.51 × 10−15. (B) Effects of HA-Cbln1 on binding between NRXs and NL1(−). The ectodomain of NRX1β(S4+) or NRX1β(S4−) was fused to the human immunoglobulin G1 Fc domain (5 μg/mL each) and incubated with HEK293 cells expressing GFP-NL1(−). Representative confocal images of cells immunostained for Fc (red or white) and GFP (green) are shown. Scale bar, 100 μm. Application of HA-Cbln1 (40 μg/mL) blocked NRX1β(S4+) but not NRX1β(S4−)-Fc binding to NL1(−). Mean intensities of bound NRX1β-Fc in the GFP-positive area are summarized and shown in the graph at the right. Error bars represent SEMs. At least = 26 fields were analyzed in three independent experiments. **= 3.52 × 10−18.

To examine this hypothesis, we next expressed GFP-tagged NL1(−) in HEK293 cells and examined whether HA-Cbln1 affected the binding between NL1(−) and NRX1β(S4+). The extracellular domains of NRX1β isoforms were attached to the Fc fragment of IgG. We confirmed that both NRX1β(S4+)-Fc and NRX1β(S4−)-Fc bound to HEK293 cells expressing NL1(−) (Fig. 1B). Application of HA-Cbln1 to the culture medium specifically and significantly reduced the interaction between NL1(−) and NRX1β(S4+)-Fc (Fig. 1B). These results indicate that Cbln1 interacts with NRX(S4+) and competes with the NL1(−)-NRX(S4+) pathway.

Cbln1 specifically binds to NRXα and NRXβ containing splice site 4 insert

To confirm that Cbln1 bound to NRX(S4+), we performed cell-based binding assays, which were previously used for the characterization of interaction between GluD2 and Cbln1 (Matsuda et al., 2010). GluD2 served as a positive control, and GluD2 lacking the NTD (GluD2ΔNTD), to which Cbln1 did not bind, served as a negative control for the binding assays. At 2 days after transfection, cells were incubated with recombinant HA-Cbln1 for 4 h. Immunoblot analyses (Fig. 2A) showed that HA-Cbln1 bound to HEK293 cells expressing NRX1β(S4+) or GluD2, but not to cells expressing GluD2ΔNTD. Immunocytochemical analyses also showed that HA-Cbln1 bound to HEK293 cells expressing NRX1β(S4+), whereas HA-CS-Cbln1, a trimeric complex that did not possess synaptogenic activities (Matsuda et al., 2010), did not bind (Fig. 2B).

Figure 2.

 Cbln1 specifically binds to NRXs carrying the splice site 4 insert. (A) Immunoblot analysis of binding of HA-Cbln1 on HEK293 cells expressing NRX1β isoforms with (S4+) or without (S4−) the splice site 4 insert, GluD2 or GluD2ΔNTD. Bound HA-Cbln1 was analyzed by an immunoblot assay using anti-HA (upper panel) and anti-actin (lower panel) antibodies. (B) Cell-based binding assay of HA-Cbln1 or CS-Cbln1. HA-Cbln1 or HA-CS-Cbln1 (red or white) was incubated with HEK293 cells expressing Flag-tagged NRX1β(S4+) (green). Scale bar, 25 μm. The mean intensity of HA immunoreactivity in the Flag-immunopositive area in HEK293 cells expressing Flag-tagged NRX1β(S4+) incubated with HA-Cbln1 was defined as 100%. HEK293 cells expressing Flag-tagged GluD2ΔNTD was used as negative control for binding. Error bars represent SEMs. At least = 30 fields were analyzed in three independent experiments. **= 1.92 × 10−13. (C) Cell-based binding assay of HA-Cbln1. HA-Cbln1 was incubated with HEK293 cells expressing indicated Flag-tagged NRX isoforms with (S4+) or without (S4−) the splice site 4 insert. Confocal images of cells immunostained for HA (red or white) and Flag (green) are shown. Scale bar, 25 μm. The mean intensity of HA immunoreactivity in the Flag-immunopositive area in HEK293 cells expressing Flag-tagged NRX1β(S4+) incubated with HA-Cbln1 was defined as 100% in the lower graph. HEK293 cells expressing Flag-tagged GluD2ΔNTD was used as negative control for binding. Error bars represent SEMs. At least = 22 fields were analyzed in two independent experiments. **P (NRX1β) = 9.3 × 10−17, **P (NRX1α) = 3.4 × 10−17, **P (NRX2β) = 1.41 × 10−17, **P (NRX3β) = 3.61 × 10−17. (D) Cell-based binding assay of LRRTM2 or NL1(−). Ectodomain of LRRTM2 or NL1(−) fused to the Fc fragment was incubated with HEK293 cells expressing NRX1β(S4+)- or (S4−)-Flag. Confocal images of cells immunostained for Fc (red or white) and Flag (green) are shown. Scale bar, 25 μm. The mean intensity of Fc immunoreactivity in the Flag-immunopositive area in HEK293 cells expressing Flag-tagged NRX1β(S4−) incubated with LRRTM2-Fc or NL1(−)-Fc was defined as 100% in the right graph. Error bars represent SEMs. At least = 25 fields were analyzed in two independent experiments. **P (LRRTM2) = 1.34 × 10−13, **P (NL1) = 1.3 × 10−6. (E) Inhibitory effects of HA-Cbln1 on binding between NL1(−) and NRX1β(S4+). NL1(−)-Fc was incubated with HEK293 cells expressing NRX1β(S4+)- or (S4−)-Flag in the presence of HA-Cbln1 or HA-CS-Cbln1 (60 μg/mL each). The mean intensity of Fc immunoreactivity in the Flag-immunopositive area in HEK293 cells expressing NRX1β(S4+)- or (S4−)-Flag in the absence of HA-Cbln1 was defined as 100% in the right graph. Error bars represent SEMs. At least = 24 fields were analyzed in two independent experiments. **= 3.99 × 10−12.

Although LRRTMs interact with both NRXα(S4−) and NRXβ(S4−) (Ko et al., 2009; de Wit et al., 2009; Siddiqui et al., 2010), certain NL isoforms bind preferentially to β-isoforms of NRXs. Thus, we examined which isoforms of NRXs interacted with Cbln1 in the cell-based binding assays. Immunocytochemical analyses of the surface HA-Cbln1 showed that Cbln1 bound to HEK293 cells expressing three subtypes of NRXβ [NRX1β(S4+), NRX2β(S4+) and NRX3β(S4+)] and one NRX1α [NRX1α(S4+)], all of which contained the splice site 4 insert (Fig. 2C), but not subtypes without the splice site 4 insert. The specific interaction with NRXs containing the splice site 4 insert was also observed by the immunoblot analysis (Fig. 2A). In contrast, the extracellular domain of LRRTM2 fused to the Fc fragment (LRRTM2-Fc) bound to HEK293 cells expressing NRX1β(S4−), but not to cells expressing NRX1β(S4+) (Fig. 2D). The extracellular domain of NL1(−) fused to the Fc fragment, i.e., NL1(−) Fc, bound to HEK293 cells expressing NRX1β(S4−) or NRX1β(S4+) (Fig. 2D). The addition of HA-Cbln1, but not HA-CS-Cbln1, significantly inhibited the binding between NL1(−)-Fc and NRX1β(S4+), whereas it did not affect the binding between NL1(−)-Fc and NRX1β(S4−) (Fig. 2E). Together, these results indicate that, unlike LRRTM2 and NL1(−), hexametric Cbln1 binds to α- and β-isoforms of NRXs in a manner dependent on the splice site 4 insert, which probably determines the interaction with Cbln1.

Binding of neurexins to Cbln1 is resistant to low extracellular Ca 2+

The binding of NLs and LRRTMs to NRXs has been reported to require extracellular Ca2+ (Ko et al., 2009; Siddiqui et al., 2010), which binds to the interface between these molecules (Koehnke et al., 2008). To examine whether the binding of Cbln1 to NRX(S4+) was also sensitive to extracellular Ca2+, we performed a cell-based binding assay in a medium containing low (56 nm, according to Ca-ethylene glycol tetra-acetic acid calculator) (Schoenmakers et al., 1992) Ca2+ concentrations. The binding of NL1(−)-Fc to HEK293 cells expressing NRX1β(S4+) under normal Ca2+ concentrations completely disappeared under low Ca2+ concentrations (Fig. 3A and B). Similarly, the binding of LRRTM2-Fc to NRX1β(S4−) was completely inhibited under low Ca2+ concentrations (Fig. 3C and D). In contrast, binding of Cbln1 to NRX1β(S4+) was observed even under low extracellular Ca2+ concentrations (Fig. 3E and F), suggesting that the mode of interaction between NRX1β(S4+) and Cbln1 was distinct from that between NRX1β(S4+) and NL1(−).

Figure 3.

 Cbln1 binds to NRX1β(S4+) under low Ca2+ conditions. (A–F) Cell-based binding assay between NRXs and NL1(−), LRRTM2 or Cbln1 in normal and low extracellular Ca2+ concentrations. The Fc-fused ectodomain of NL1(−) (A), LRRTM2 (C) or HA-Cbln1 (E) was incubated with HEK293 cells expressing Flag-tagged NRX1β(S4−) or NRX1β(S4+) in a medium containing normal (1.8 mm) or low [56 nm by addition of 5 mm ethylene glycol tetra-acetic acid (EGTA)] Ca2+ concentrations. Confocal images of cells immunostained for ligands (red or white) and Flag (green) are shown. Scale bar, 25 μm. The mean intensity of HA or Fc immunoreactivity in the Flag positive area in HEK293 cells expressing Flag-tagged NRX1β(S4+) or NRX1β(S4−) in normal Ca2+ concentrations was defined as 100% in the right graphs (B, D and F). HEK293 cells expressing GFP or Flag-tagged GluD2ΔNTD were used as negative control for binding. Error bars represent SEMs. At least = 24 fields were analyzed in two independent experiments. **P (HA-Cbln1) = 0.003, **P (NL1-Fc) = 6.84 × 10−11, **P (LRRTM2-Fc) = 4.76 × 10−11. (G and H) Cell-based binding assay of mutant NRXs and NL1(−) or Cbln1. The Fc-fused NL1(−) ectodomain (G) or HA-Cbln1 (H) was incubated with HEK293 cells expressing NRX1β(S4+)-Flag, which has mutations in Ca2+ binding sites (N238A or D137A). Confocal images of cells immunostained for ligands (red or white) and Flag (green) are shown. Scale bar, 25 μm. Quantified data are summarized in B and F, respectively. (I) In vitro binding assay of HA-Cbln1 and NRX at low Ca2+ concentrations. HA-Cbln1 was incubated with a column immobilized with CD4-Fc or NRX1β(S4+)-Fc in a buffer containing normal (2 mm) or low (in the presence of 5 mm EGTA) Ca2+ concentrations. Immunoblot analyses of column-bound HA-Cbln1 using anti-HA or anti-Fc antibody are shown. (J) Cbln1 binds to GluD2 under low Ca2+ conditions. HA-Cbln1 (red or white) was incubated with HEK293 cells expressing GluD2 (green) in a medium containing normal (1.8 mm) or low (56 nm by addition of 5 mm EGTA) Ca2+ concentrations. Scale bar, 25 μm. WT, wild-type.

To further confirm the distinct binding mode of Cbln1 to NRX1β(S4+), we mutated Ca2+ binding sites of NRX1β(S4+) (Fabrichny et al., 2007; Reissner et al., 2008). Even under normal extracellular Ca2+ concentrations, NL1(−)-Fc did not bind to HEK293 cells expressing NRX1β(S4+)N238A in which an alanine residue replaced an asparagine residue at position 238 or cells expressing NRX1β(S4+)D137A in which an alanine residue replaced an aspartate residue at position 137 (Fig. 3B and G). In contrast, HA-Cbln1 bound to HEK293 cells expressing NRX1β(S4+)D137A or NRX1β(S4+)N238A in a manner similar to cells expressing wild-type NRX1β(S4+) (Fig. 3F and H). To examine whether Ca2+ concentrations did not affect the direct binding between Cbln1 and NRX1β(S4+), we performed an in vitro binding assay using HA-Cbln1 and NRX1β(S4+)-Fc or CD4-Fc. Immunoblot analyses demonstrated that binding of HA-Cbln1 to the NRX1β(S4+)-Fc column was similarly observed in both normal and low Ca2+ concentrations (Fig. 3I). These results indicate that Cbln1 bound to NRXs in a manner distinct from NLs or LRRTMs.

As Cbln1 binds to GluD2 at the postsynaptic site, we next examined whether the binding between Cbln1 and GluD2 was affected by extracellular Ca2+ concentrations. Immunocytochemical analyses of the surface HA-Cbln1 revealed that HA-Cbln1 bound to HEK293 expressing GluD2 under low extracellular Ca2+ concentrations (Fig. 3J). Together, these results indicate that, unlike NRX/NL- or NRX/LRRTM-based cell adhesion, trans-synaptic cell adhesion mediated by NRX1β(S4+)/Cbln1/GluD2 is resistant to low extracellular Ca2+ concentrations.

Neurexin/Cbln1/GluD2 serves as a presynaptic organizer

Cbln1, which accumulates at the synaptic junction by binding to GluD2, serves as a presynaptic organizer (Matsuda et al., 2010). As NRX is known to recruit synaptic vesicles (Dean et al., 2003), it probably mediates the presynaptic organizing function of Cbln1. To examine this hypothesis, we first examined whether Cbln1 and GluD2 formed a tripartite complex with NRXs. Immunocytochemical analyses showed that NRX1β(S4+)-Fc but not NRX1β(S4−)-Fc specifically bound to HEK293 cells expressing GluD2 only when HA-Cbln1 was added to the culture medium (Fig. 4A). Similarly, when NRX1β(S4+) and GFP were coexpressed in cbln1-null cerebellar granule cells, NRX1β(S4+) accumulated in GFP-positive axons around the beads coated with HA-Cbln1 but not around uncoated beads (Fig. 4B). We expressed NRX1β(S4+)-Flag, in which the region necessary for binding to presynaptic organizing proteins such as calcium/calmodulin-dependent serine protein kinase (CASK) (Hata et al., 1996; Dean et al., 2003) was disrupted by attaching the Flag tag at the extreme C-terminus of NRX1β(S4+) (Fairless et al., 2008) in wild-type hippocampal neurons. Importantly, NRX1β(S4+)-Flag also accumulated in axons contacting the beads coated with HA-Cbln1 without recruiting the presynaptic marker synapsin I (Supporting information Fig. S2A), indicating that accumulation of NRX1β(S4+) was directly caused by HA-Cbln1 and not by other presynaptic molecules that bound to the C-terminus of NRX1β(S4+). In addition, not only overexpressed NRX1β(S4+), but also endogenous NRXs in cbln1-null granule cells preferentially accumulated in axons contacting the beads coated with HA-Cbln1 (Supporting Information Fig. S2B). Furthermore, NRX1β(S4+)-Flag expressed in cbln1-null granule cells accumulated in axons that crossed Purkinje cells only when HA-Cbln1 was added to the culture medium (Supporting Information Fig. S2C), indicating that Cbln1, which was bound to GluD2 on Purkinje cell dendrites, induced clustering of NRX1β(S4+) at presynaptic terminals. Although beads coated with Cbln1 accumulated synapsin I-positive synaptic vesicles in cbln1-null granule cell axons (Matsuda et al., 2010), addition of NRX1β(S4+)-Fc and not NRX1β(S4−)-Fc to the culture medium significantly inhibited Cbln1 presynaptic organizing function (Fig. 4C). Lastly, immunocytochemical staining of endogenous NRXs using pan-NRX antibody revealed that NRX immunoreactivity (green) associated with GluD2 puncta (red) on Purkinje cell dendrites (blue) was lower in cbln1-null cerebellar neurons (Fig. 4D, middle panels) than in wild-type neurons (Fig. 4D, upper panels). Addition of HA-Cbln1 to the culture medium restored accumulation of endogenous NRXs associated with GluD2 puncta on cbln1-null Purkinje cell dendrites (Fig. 4D, lower panels). Together, these results indicate that Cbln1/GluD2 serves as a presynaptic organizer by directly accumulating its presynaptic receptor NRXs(S4+).

Figure 4.

 The Cbln1/GluD2 complex serves as a presynaptic organizer by recruiting presynaptic NRX(S4+). (A) Cbln1 and GluD2 form a tripartite complex with NRX1β(S4+). The Fc fusion protein of NRX1β(S4+) or NRX1β(S4−) ectodomain (50 μg/mL each) was applied to HEK293 cells expressing GluD2 or GluD2ΔNTD in the presence of HA-Cbln1 or HA-CS-Cbln1 (40 μg/mL). NRX1β(S4+)-Fc (red or white) bound to cells expressing GluD2 (blue) only in the presence of HA-Cbln1 (green or white). Scale bar, 50 μm. (B) Accumulation of axonal NRX1β(S4+) on HA-Cbln1-coated beads in granule cells. Cbln1-null cerebellar granule cells expressing NRX1β(S4+)-Flag plus GFP were cocultured with HA-Cbln1-conjugated beads or uncoated beads from 9 to 11 DIV. HA-Cbln1-conjugated beads but not control beads (blue) caused clustering of axonal NRX1β(S4+) (red or white) in transfected granule cells. Red arrowheads indicate beads that induced clustering of axonal NRX1β(S4+) and white arrowheads indicate beads that were close to axons but did not cause the clustering. Scale bar, 20 μm. Mean intensities of NRX1β(S4+)-Flag immunoreactivity in the distal part of GFP-positive axons that were in contact with beads are summarized in the right graph. Error bars represent SEMs. At least = 15 neurons were analyzed in two independent experiments. **= 1.57 × 10−4. (C) Inhibition of Cbln1-induced presynaptic differentiation by NRX1β(S4+)-Fc. HA-Cbln1-conjugated beads were cocultured with cbln1-null cerebellar granule cells in the presence of the Fc fusion protein of NRX1β(S4+) or NRX1β(S4−) ectodomain (50 μg/mL each). Representative images of cells immunostained for synapsin I (red or white) and beads (green) are shown. Scale bar, 20 μm. Mean intensity of synapsin I on the beads is summarized and shown in the graph. Error bars are SEMs. At least = 24 fields were analyzed in three independent experiments. **= 2.52 × 10−11. (D) Accumulation of endogenous NRX on Purkinje cells requires Cbln1. Cerebellar mixed cell cultures prepared from wild-type (wt) or cbln1-null mice were cultured with or without HA-Cbln1 (40 μg/mL for 4 days). Cells were fixed at 12–14 DIV and immunostained for pan-NRX (green or white), GluD2 (red or white) and calbindin (blue). Boxed regions are enlarged in the right panels. Scale bar, 50 μm.

Neurexin/Cbln1/GluD2 serves as a postsynaptic organizer

Cbln1 also serves as a postsynaptic organizer that induces clustering of GluD2 and its associated proteins at the postsynaptic site. To examine whether NRX functions as a postsynaptic organizer by forming a tripartite complex with Cbln1 and GluD2, we cultured HEK293 cells expressing GluD2 with beads coated with NRX1β. GluD2 clustering was induced around beads coated with NRX1β(S4+) only when HA-Cbln1 was added to the culture medium (Fig. 5A). However, beads coated with NRX1β(S4−) did not cause clustering of GluD2 even in the presence of HA-Cbln1 (Fig. 5A), suggesting that NRX1β(S4+) caused GluD2 clustering in HEK293 cells by forming a complex with Cbln1.

Figure 5.

 NRX1β(S4+) functions as a postsynaptic organizer by forming a tripartite complex with Cbln1 and GluD2. (A) NRX1β(S4+) induced clustering of GluD2 in HEK293 cells in a Cbln1-dependent manner. The Fc fusion protein of NRX1β(S4+) or NRX1β(S4−) ectodomain was conjugated with avidin beads by biotinylated anti-Fc antibody. Beads were incubated with HEK293 cells expressing GluD2 (red or white) with GFP (green or white) for 2 days in the presence or absence of HA-Cbln1 (40 μg/mL). Red arrowheads indicate beads that induced clustering of GluD2 and white arrowheads indicate those that did not. Scale bar, 20 μm. (B) NRX1β(S4+) induced clustering of endogenous GluD2 and shank2 in Purkinje cells in a Cbln1-dependent manner. NRX1β(S4+)- or NRX1β(S4−)-conjugated beads were cocultured with cbln1-null Purkinje cells in the presence or absence of HA-Cbln1 (40 μg/mL) from 10 to 13 DIV. NRX1β(S4+)-conjugated beads induced clustering of endogenous GluD2 (green) and its associated postsynaptic protein shank2 (red) in cbln1-null Purkinje cells only in the presence of HA-Cbln1. Red arrowheads indicate GluD2 accumulated around the beads. Scale bar, 20 μm. (C) Clustering of GluD2 in Purkinje cells was directly induced by NRX1β(S4+)-conjugated beads. In the presence of HA-Cbln1, clustering of GluD2 (green) was induced in Purkinje cells around NRX1β(S4+)-conjugated beads (red; arrowheads), which was located at extrasynaptic sites lacking endogenous presynaptic terminals (detected by synapsin I; blue). Scale bar, 20 μm.

The C-terminus of GluD2 interacts directly with several intracellular molecules in neurons; many of these serve as scaffolds for other postsynaptic molecules. Thus, to examine whether NRX also functions as a postsynaptic organizer in neurons, we cultured cbln1-null Purkinje cells with beads coated with NRX1β(S4+) from 10 to 13 DIV. Immunocytochemical analyses showed that GluD2 clustering was induced around beads only in the presence of HA-Cbln1 (Fig. 5B). Similarly, shank2, a scaffold protein that binds to the C-terminus of GluD2, clustered around beads coated with NRX1β(S4+) (Fig. 5B). In contrast, beads coated with NRX1β(S4−) did not cause clustering of GluD2 or shank2 even in the presence of HA-Cbln1 (Fig. 5B). Coimmunostaining of presynaptic synapsin I and postsynaptic GluD2 showed that GluD2 puncta induced by beads coated with NRX1β(S4+) in the presence of HA-Cbln1 were not associated with synapsin I-positive presynaptic terminals (Fig. 5C), indicating that NRX1β(S4+)-beads directly induced GluD2 clustering at the contact sites. These results indicated that the tripartite complex consisting of NRX, Cbln1 and GluD2 serves as a bidirectional synaptic organizer.

Synapse-organizing activities of other Cbln family members

Of the Cbln family members, Cbln1, Cbln2 and Cbln4 mRNAs are expressed in various brain regions outside the cerebellum, including the olfactory bulb, entorhinal cortex and certain thalamic nuclei (Miura et al., 2006). As NRXs(S4+) are also highly expressed in these regions (Ichtchenko et al., 1995), Cbln family members may also be involved in synapse formation by forming complexes with NRXs. To explore this possibility, we first performed an in vitro binding assay using a column that immobilized HA-tagged Cbln family proteins and cell lysates from HEK293 cells expressing Flag-tagged NRX1β(S4+) as ligands. Immunoblot analyses showed that NRX1β(S4+)-Flag bound to the columns conjugated with HA-Cbln2 as well as HA-Cbln1, whereas it did not bind to columns conjugated with Cbln4 or CS-Cbln1 (Fig. 6A). Similarly, HA-Cbln1 and HA-Cbln2 but not HA-Cbln4 or HA-CS-Cbln1 bound to columns that immobilized NRX1β(S4+)-Fc, whereas none of the Cbln family members bound to NRX1β(S4−)-Fc columns (Fig. 6B). Furthermore, beads coated with HA-Cbln1 or HA-Cbln2, but not those coated with HA-Cbln4 or HA-CS-Cbln1, caused clustering of NRX1β(S4+) expressed in HEK293 cells (Supporting Information Fig. S3). These results indicate that, like Cbln1, Cbln2 also binds and accumulates NRXs carrying the splice site 4 insert.

Figure 6.

 Direct interaction with NRX1β(S4+) and synaptogenic activities in cerebellar granule cells of Cbln family proteins. (A) Binding of NRX1β(S4+)-Flag to HA-Cblns conjugated on columns. Lysates of HEK293 cells expressing NRX1β(S4+)-Flag were loaded on columns conjugated with HA-Cbln family proteins (Cbln1, Cbln2, Cbln4 and CS-Cbln1). Bound NRX1β(S4+)-Flag was visualized by immunoblot analyses using anti-Flag antibody (upper panel). The amount of HA-Cbln protein conjugated to columns was also analyzed by immunoblotting using anti-HA antibody (lower panel). (B) Binding of Cbln family proteins to columns conjugated with Fc fusion protein of NRX1β(S4+) or NRX1β(S4−) ectodomain. Immunoblots of bound and applied HA-Cbln proteins (Cbln1, Cbln2, Cbln4 or CS-Cbln1) using anti-HA antibody are shown in the top and middle panel, respectively. The amount of NRX protein conjugated to columns was also analyzed by immunoblotting using anti-Fc antibody (lower panel). (C) Cbln1 and Cbln2 but not Cbln4 induced presynaptic differentiation of cerebellar granule cells. Beads coated with HA-Cbln1, Cbln2, Cbln4 or CS-Cbln1 were cocultured with cbln1-null cerebellar granule cells from 11 DIV for 3 days. Confocal images of granule cells immunostained for synapsin I (red or white) and beads (green) are shown. Scale bar, 20 μm. Mean intensity of synapsin I on the beads is summarized and shown in the graph at the right. Error bars are SEMs. At least = 22 fields were analyzed in three independent experiments. **P (Cbln1 vs. CS-Cbln1) = 2.43 × 10−8, **P (Cbln2 vs. CS-Cbln1) = 2.63 × 10−9.

To examine whether Cbln family proteins had direct synaptogenic activities in cerebellar granule cells, we performed artificial synapse-forming assays using beads coated with HA-Cbln1, HA-CS-Cbln1, HA-Cbln2 or HA-Cbln4. Beads were incubated with cbln1-null cerebellar granule cells for 3 days and presynaptic terminals were immunostained with synapsin I. Like HA-Cbln1, HA-Cbln2 significantly induced clustering of synapsin I-positive presynaptic terminals on the beads (Fig. 6C). Although the amount of Cbln proteins on the beads was adjusted, the intensity of synapsin I immunoreactivity on Cbln2-coated beads was weaker than that on Cbln1-coated beads (= 0.015; Fig. 6C). Thus, Cbln2 may have weaker affinity to NRX1β(S4+) (Fig. 6A and B) and weaker synaptogenic activity (Fig. 6C) than Cbln1. Consistent with the finding that HA-Cbln4 did not bind to NRXs (Fig. 6A and B), HA-Cbln4 as well as HA-CS-Cbln1 did not induce accumulation of presynaptic terminals of cbln1-null granule cells (Fig. 6C).

Cbln1 induces presynaptic differentiation in hippocampal neurons

The NRX(S4+) is widely expressed in the central nervous system, including the hippocampus and cerebral cortex (Ichtchenko et al., 1995). Although GluD2 is specifically expressed in Purkinje cells, its family member δ1 glutamate receptor (GluD1), which also binds to Cbln1 (Matsuda et al., 2010), is highly expressed in various brain regions, such as the striatum, especially during development (Lomeli et al., 1993). Indeed, Cbln1 is expressed in the thalamic parafascicular nucleus that sends axons to the striatal neurons (Kusnoor et al., 2010). Thus, the NRX/Cbln1/GluD1 complex might be involved in synaptic functions in these brain regions. As a first step to explore this possibility, we performed artificial synapse-forming assays using HEK293 cells and wild-type hippocampal neurons as a model system, taking advantage of the fact that hippocampal neurons do not express endogenous Cbln1 (Miura et al., 2006). Immunocytochemical analyses showed that HEK293 cells expressing GluD2 but not those expressing GluD2ΔNTD accumulated synaptophysin-positive presynaptic terminals of hippocampal neurons only when recombinant HA-Cbln1 protein was added to the culture medium (Fig. 7A). In addition, beads coated with HA-Cbln1 directly accumulated synapsin I-positive presynaptic terminals of hippocampal neurons (Fig. 7B). FM4-64 fluorescence, which is taken up by functional presynaptic terminals, was also detected on beads coated with HA-Cbln1 (Supporting Information Fig. S4A). Furthermore, synapsin I-immunopositive terminals accumulated around HA-Cbln1-coated beads at extrasynaptic sites that lacked endogenous AMPA receptor clusters (Supporting Information Fig. S4B). These results indicate that exogenous Cbln1 is capable of directly inducing the accumulation of functional synaptic vesicles in non cerebellar neurons.

Figure 7.

 Cbln1 and Cbln2 serve as a presynaptic organizer in hippocampal neurons. (A) Cbln1 and GluD2 induced accumulation of presynaptic sites of hippocampal neurons on HEK293 cells. HEK293 cells expressing GFP and GluD2 or GluD2ΔNTD were cocultured with wild-type (wt) hippocampal neurons in the presence or absence of HA-Cbln1 (2 μg/mL). Confocal images of HEK293 cells immunostained for GFP (green) and synaptophysin (Syn) (red or white) are shown. Scale bar, 50 μm. Mean intensities of synaptophysin immunoreactivity in the GFP-positive area are summarized and shown in the graph at the right. Error bars are SEMs. At least = 21 fields were analyzed in two independent experiments. **= 2.07 × 10−11. (B) Beads coated with HA-Cbln1 directly accumulated presynaptic sites of hippocampal neurons. Confocal images of hippocampal neurons immunostained for synapsin I (red or white) and beads (green). Regions marked by white boxes are enlarged in the panels on the right. Scale bar, 20 μm. Mean intensity of synapsin I immunoreactivity on the bead area is summarized and shown in the graph at the right. Error bars are SEMs. At least = 20 fields were analyzed in two independent experiments. **= 8.67 × 10−11. (C) Cbln1, Cbln2 and Cbln4/Cbln1 but not Cbln4 induced presynaptic differentiation of hippocampal neurons. Cbln1-, Cbln2-, Cbln4-, Cbln4/Cbln1- or CS-Cbln1-conjugated beads were cocultured with wild-type hippocampal neurons. Confocal images of neurons immunostained for synapsin I (red or white) and beads (green) are shown. Scale bar, 20 μm.

To further evaluate the synaptogenic activity of Cbln family proteins that are expressed outside the cerebellum, we incubated the beads coated with HA-Cbln1, 2 and 4 with hippocampal and cortical neurons. Immunocytochemical analyses of synapsin I showed that HA-Cbln2 but not HA-Cbln4 or HA-CS-Cbln1 accumulated presynaptic terminals of hippocampal (Fig. 7C) and cortical (Supporting Information Fig. S5) neurons on the beads. As Cbln4 and Cbln1 are coexpressed in certain brain regions, such as the entorhinal cortex and thalamus, Cbln4 may still work as a heteromeric complex with Cbln1 (Miura et al., 2006; Iijima et al., 2007). To test this possibility, HA-Cbln4 and nontagged Cbln1 were coexpressed in HEK293 cells and HA-Cbln4 homomers and HA-Cbln4/Cbln1 heteromers were recovered by biotinylated anti-HA antibody and immobilized on avidin beads. Immunocytochemical analyses showed that, unlike beads coated with HA-Cbln4, beads containing HA-Cbln4/Cbln1 heteromers accumulated presynaptic terminals of hippocampal neurons (Fig. 7C). Together, these results indicate that, of the Cbln family proteins, Cbln1, Cbln2 and Cbln4/Cbln1 heteromers function as presynaptic organizers by associating with NRXs with the splice site 4 insert in various brain regions at least in vitro.

Discussion

Cbln1 is one of the most recently identified bidirectional synaptic organizers in the cerebellum; Cbln1 secreted from cerebellar granule cells indirectly serves as a postsynaptic organizer by binding to its postsynaptic receptor GluD2 expressed in Purkinje cells and directly induces presynaptic differentiation (Matsuda et al., 2010). However, it remained unclear how Cbln1 binds to the presynaptic sites and interacts with other synaptic organizers. In this study, we found that Cbln1 competed with synaptogenesis mediated by NL-NRX and identified NRX1α(S4+) and NRXβs(S4+) as presynaptic receptors for Cbln1. While this manuscript was in preparation, Uemura et al. (2010) also reported the interaction of Cbln1 with NRXs in the cerebellum. We further showed that not only Cbln1, but also its family member Cbln2 but not Cbln4 specifically bound to NRX1β(S4+) even under low Ca2+-concentrations, which was distinct from the interaction between NRXs and NLs or NRXs and LRRTM2. We also characterized in detail the nature of the tripartite complex NRXs/Cbln1/GluD2 as a bidirectional organizer. Finally, we showed that Cbln members induced synaptogenesis in hippocampal and cortical neurons as well as in the cerebellar neurons (Fig. 7 and Supporting Information Fig. S5).

Unique synaptic signaling by the tripartite complex neurexin/Cbln1/GluD2

The cell-based (Fig. 2) and in vitro (Fig. 6) binding assays showed that NRX1α and NRX1–3β carrying the splice site 4 insert specifically bound to Cbln1. Cbln1 coated on beads directly accumulated NRX1β(S4+) on granule cell axons (Fig. 4B and Supporting Information Fig. S2A) and Cbln1-induced presynaptic differentiation was specifically inhibited by soluble NRX1β(S4+)-Fc (Fig. 4C), indicating that NRXs(S4+) serves as a presynaptic receptor for Cbln1. In addition, NRX1β(S4+) coated on beads clustered GluD2 and its interacting intracellular protein shank2 in postsynaptic Purkinje cells in a Cbln1-dependent manner (Fig. 5B). These results indicate that the tripartite complex consisting of NRX(S4+), Cbln1 and GluD2 could serve as a bidirectional synaptic organizer.

The NRX/Cbln1/GluD2 complex has several unique features as a synapse organizer (Fig. 8). First, unlike NRXs/NLs (Nguyen & Sudhof, 1997) or NRXs/LRRTMs (Ko et al., 2009; Siddiqui et al., 2010), this complex was resistant to low extracellular Ca2+ concentrations. The crystal structure of NRX1β indicates that Ca2+ binding is essential for binding to NLs (Koehnke et al., 2008). Similarly, other NRX ligands, such as LRRTMs and α-dystroglycan (Sugita et al., 2001), also bind to NRX in a Ca2+-dependent manner. In contrast, neurexophilins bind to the second laminin, NRX, sex-hormone-binding protein (LNS) domain in NRXα in a Ca2+-independent manner (Missler et al., 1998). Unlike neurexophilins but like NLs and LRRTMs, Cbln1 binds to both NRXα and NRXβ, suggesting that Cbln1 binds to the sixth LNS domain in which the splice site 4 insert is located (Craig & Kang, 2007). Structural studies on NRX1β(S4+) have shown that the splice site 4 insert is unstructured and remains partially disordered in the complex with NLs despite its high level of sequence conservation, suggesting that it has a distinct functional role in binding to partner molecules other than NLs (Koehnke et al., 2008). Together, these findings indicate that Cbln1 binds to the region involving the splice site 4 insert of NRXs in a manner distinct from NLs or LRRTMs. Although it remains unclear whether Cbln1 and NLs compete for presynaptic NRXs in vivo, Cbln1 inhibited the interaction between NL1(−) and NRX(S4+) in vitro (Fig. 1) probably by steric hindrance because Cbln1 and NL1(−) are unlikely to share the same binding site of NRX(S4+).

Figure 8.

 Schematic drawing summarizing a unique synaptic signaling by the tripartite complex NRX/Cbln1/GluD2. Cbln1 serves as a bidirectional synaptic organizer by binding to presynaptic NRXs(S4+) and postsynaptic GluD2 in hippocampal, cortical and cerebellar neurons. Cbln2 but not Cbln4 is likely to have a similar synaptogenic function. Although LRRTMs and NLs bind to NRXs in a Ca2+-dependent manner, the NRX/Cbln1/GluD2 complex is maintained under low Ca2+ concentration. In contrast to LRRTMs, which bind to NRXs(S4−) and mediate excitatory synaptogenesis, Cbln1/GluD2 specifically binds to NRXs(S4+) and regulates excitatory and inhibitory synaptogenesis, depending on the type of neurons expressing Cbln proteins. NMDAR, N-methyl-D-aspartate receptor; PSD95, postsynaptic density 95; VGluT, vesicular glutamate transporter; VGAT, vesicular GABA transporter.

Although various cell adhesion molecules (such as cadherins, protocadherins, NRXs/NLs and NRXs/LRRTMs) require extracellular Ca2+, synaptic adhesion itself is independent of Ca2+ (Sudhof, 2001). cbln1- and GluD2-null mice are ataxic, showing a markedly impaired performance on the rotorod test. Although their cerebellum appears grossly normal, detailed electrophysiological and electron microscopic analyses of these mice revealed that the number of PF–Purkinje cell synapses is markedly reduced, and most dendritic spines have lost synaptic contact with the PFs. In addition, in the remaining PF–Purkinje cell synapses, the postsynaptic densities are disproportionally longer than the presynaptic active zones. These unique morphological phenotypes and Ca2+-resistant binding of the NRX/Cbln1/GluD2 complex is consistent with the function of the complex as synaptic glue, connecting pre- and postsynaptic elements.

The second unique feature of the NRX/Cbln1/GluD2 complex is that the secreted Cbln1 works by being sandwiched between presynaptic NRX and postsynaptic GluD2. In central nervous system synapses, synaptic organizers are classified into two categories: cell adhesion molecules that directly link pre- and postsynaptic elements and soluble factors. Most soluble synaptic organizers in the central nervous system, such as neuronal pentraxins (Xu et al., 2003), fibroblast growth factors (Terauchi et al., 2010) and Wnt-7a (Hall et al., 2000), work on either the pre- or postsynaptic site, depending on the location of their receptors (Johnson-Venkatesh & Umemori, 2010). Thus, the sandwich-type signaling by the NRX/Cbln1/GluD2 complex is unique in that secreted Cbln1 serves as a bidirectional synaptic organizer. For Cbln1 to bind to pre- and postsynaptic receptors simultaneously, Cbln1 needs to have at least two binding sites. This could have been achieved by the presence of multiple binding sites within single Cbln1 monomers or by the presentation of single binding sites in different directions by forming a multimeric Cbln1 complex (Iijima et al., 2007). Recently, glial-derived neurotrophic factor was also proposed to serve as a synaptic adhesion molecule being sandwiched by its receptor glial-derived neurotrophic factor family receptor (GFR)α1 located at pre- and postsynaptic neurons (Ledda et al., 2007). In addition, leucine-rich glioma inactivated 1 was recently shown to be secreted from neurons and to organize presynaptic potassium channels and postsynaptic AMPA receptors by binding to its pre- and postsynaptic receptors, a disintegrin and metalloproteinase (ADAM) 22 and ADAM23, respectively (Fukata et al., 2010). These recent findings indicate that the sandwich type constitutes the third category of synaptic organizers.

Advantages of sandwich-type synaptic organizers may include an additional level of regulation of synapse formation and its functions. For example, the expression of cbln1 mRNA is completely shut down in granule cells when neuronal activity is increased for several hours (Iijima et al., 2009). Similarly, a sustained increase in neuronal activity causes the internalization of GluD2 from the postsynaptic site of cultured Purkinje cells (Hirai, 2001). As Cbln1 and NLs compete for NRXs, such activity-dependent regulation of Cbln1 and GluD2 might lead to switching between NRX/NL and NRX/Cbln1/GluD2 modes of synaptogenesis. Furthermore, the splicing of site 4 of NRXs was also shown to be regulated during development and in response to neurotrophic factors in chicken (Patzke & Ernsberger, 2000) and by ischemia in rats (Sun et al., 2000). Thus, each component of the NRX/Cbln1/GluD2 complex may be differentially regulated at the transcriptional and post-translational levels and such fine tuning of the NRX/Cbln1/GluD2 complex may play a role in the structural changes observed at PF synapses following increased neuronal activity in the adult cerebellum (Black et al., 1990).

Cbln1 and Cbln2 serve as synaptic organizers in various brain regions

Cbln1 mRNA is highly expressed in the cerebellum, but it is also enriched in a subset of neurons in various brain regions, including the mitral layer of the olfactory bulb, retrosplenial granular cortex, entorhinal cortex and thalamic parafascicular nucleus (Miura et al., 2006). Nevertheless, it is unclear whether Cbln1 is involved in synaptogenesis in these brain regions. We showed that Cbln1-coated beads were capable of inducing hemisynaptic differentiation of hippocampal and cortical neurons in vitro. Interestingly, in cbln1-null mice the spine density of medial spiny neurons in the striatum, which receive inputs from the Cbln1-positive thalamic parafascicular nucleus, was markedly increased, suggesting that Cbln1 determines the dendritic structure of striatal neurons with effects distinct from those seen in the cerebellum (Kusnoor et al., 2010). Although GluD2 is not expressed, its family member GluD1, which also binds to HA-Cbln1 (Matsuda et al., 2010), is highly expressed in these brain regions, especially during development (Lomeli et al., 1993). Therefore, a possible explanation for this difference is that GluD1 may mediate postsynaptic effects distinct from those regulated by GluD2. Indeed, Cbln1-coated beads did not accumulate AMPA receptors in hippocampal neurons (Supporting Information Fig. S4B) although endogenous GluD1 is expressed in these neurons (data not shown), suggesting that, unlike GluD2, GluD1 may not associate with scaffolding proteins such as shank2. Further studies are required to determine the signaling pathways regulated by Cbln1 outside the cerebellum.

The Cbln family consists of four members, Cbln1–Cbln4. Although Cbln3 is specifically expressed in cerebellar granule cells, other members are expressed in various brain regions (Miura et al., 2006). We showed that Cbln1 and Cbln2 but not Cbln4 were capable of binding to NRX1β(S4+) and inducing hemisynaptic differentiation of cerebellar, hippocampal and cortical neurons in vitro. Such differential effects were rather unexpected, as the amino acid sequences of the coding regions of Cbln1, Cbln2 and Cbln4 are very similar to each other (87–91%) (Yuzaki, 2008). As Cbln4 is always coexpressed with Cbln1 or Cbln2 in most brain regions (Miura et al., 2006), such as the entorhinal cortex and thalamic parafascicular nucleus, Cbln4 may serve as a synaptic organizer by forming a heteromer complex (Fig. 7C), and possibly by modulating the synaptogenic activities of Cbln1 and Cbln2. Future studies using gene knockout mice are required to elucidate the synaptic roles of Cbln and GluD family proteins in the central nervous system.

Acknowledgements

We thank Dr J.I. Morgan for cbln1-null mice and J. Motohashi and S. Narumi for their technical support. This work was supported by MEXT and/or JSPS KAKENHI to K.M. and M.Y., the CREST from the JST Agency (M.Y.), the Takeda Science Foundation (K.M. and M.Y.), and the Naito Memorial Grant for Female Researchers (K.M.)

Abbreviations
AMPA

α-amino-3-hydroxyl-5-methyl-4-isoxazole-propionate

DIV

days in vitro

GFP

green fluorescent protein

GluD1

δ1 glutamate receptor

GluD2

δ2 glutamate receptor

HA

hemagglutinin

HEK

human embryonic kidney

LRRTM

leucine-rich repeat transmembrane protein

NL

neuroligin

NMDAR

N-methyl-D-aspartate receptor

NRX

neurexin

NTD

N-terminal domain

PBS

phosphate-buffered saline

PF

parallel fiber

PSD95

postsynaptic density 95

VGluT

vesicular glutamate transporter

VGAT

vesicular GABA transporter

Ancillary