A complex of synaptic adhesion molecule CADM1, a molecule related to autism spectrum disorder, with MUPP1 in the cerebellum


Address correspondence and reprint requests to Takashi Momoi, Center for Medical Science, International University of Health and Welfare, 2600-1, Kitakanemaru, Ohtawara, Tochigi, Japan. E-mail: momoi@iuhw.ac.jp. (or) Mariko Y. Momoi, Department of Pediatrics, Jichi Medical University, 3311-1 Yakushiji, Shimotsukeshi, Tochigi, Japan. E-mail: mymomoi@jichi.ac.jp


Mutations in the synaptic adhesion protein CADM1 (RA175/SynCAM1) are associated with autism spectrum disorder (ASD), a neurodevelopmental disorder of uncertain molecular origin. Cadm1-knock out (KO) mice exhibit smaller cerebella with decreased number of synapse of Purkinje cells and some ASD-like symptoms, including impaired ultrasonic vocalization. In this study, we examined the alteration of the Cadm1 synaptic complex in the mouse cerebellum at post-natal stages. The C-terminal peptide of Cadm1 associated with Mupp1 at PSD-95/Dlg/ZO-1 (PDZ)(1-5), a scaffold protein containing 13 PDZ domains, which interacted with gamma-aminobutyric acid type B receptor (GABBR)2 at PDZ13, but not with PSD-95. The GABBR2 was detected in a set of proteins interacting with Cadm1 C-terminal. Cadm1 colocalized with Mupp1 and GABBR2 on the dendrites of Purkinje cells in the molecular layers of the developing cerebellum and on the dendrites of hippocampal neurons cultured in vitro. These observations suggest that the Cadm1 synaptic receptor complex, including Mupp1–GABBR2, is located on the dendrites of Purkinje cells. The amount of GABBR2 protein, but not mRNA, was increased in the cerebella of Cadm1 KO mice, suggesting that lack of Cadm1 does not affect transcription of GABBR2, but may stabilize the Mupp1–GABBR2 complex; the Mupp1–GABBR2 interaction may be stabilized by conformational change in Mupp1 or association with other adhesion molecules and by anchorage to the post-synaptic membrane. Up-regulation of GABBR2 in the cerebellum in the absence of CADM1 may be associated with ASD pathogenesis.

Abbreviations used

Autism spectrum disorder


Cell adhesion molecule 1


calmodulin-associated serine/threonine kinase


Gamma-aminobutyric acid type B receptor


knock in


knock out


Multiple PDZ domain protein 1






ultrasonic vocalization

RA175/SynCAM1/CADM1 (CADM1), a member of the immunoglobulin superfamily, localizes to both sides of the synaptic cleft and functions as a synaptic cell–cell adhesion molecule. The extracellular domain of Cadm1 mediates calcium-independent, homophilic trans-interactions (Biederer et al. 2002; Fujita et al. 2003). Patients bearing the CADM1 mutations H246N and Y251S are diagnosed with autism spectrum disorder (ASD) (Zhiling et al. 2008), a heritable neurodevelopmental disorder (Pickett and London 2005). Cadm1-knock out (KO) mice (Fujita et al. 2006, 2007) exhibit abnormal social and emotional behavior similar to the human symptoms of ASD (Takayanagi et al. 2010), Cadm1-KO pups possess smaller cerebella than wild-type pups and emit impaired ultrasonic vocalization (USV) like Foxp2(R552H)-knock in (KI) pups, who carry a mutation related to speech–language disorder (Fujita et al. 2008, 2012a).

Cerebellar abnormalities with Purkinje cell loss occur in human autopsy samples from ASD patients (Ritvo et al. 1986; Bauman and Kemper 1994; Fatemi et al. 2000). While Cadm1 mRNA also appears in the Purkinje cells and external granular cells of the developing cerebellum (Urase et al. 2001), Cadm1 protein is expressed in the developing brain, including the cerebrum and cerebellum (Fujita et al. 2005, 2012a). Cadm1 partly colocalizes with VGluT1 and VGluT2 in the developing cerebellum (Miyazaki et al. 2003; Fujita et al. 2012a). Thus, it is likely that Cadm1 is involved in the cerebellular synaptic function.

In this study, we focused on the Cadm1 molecular complexes in cerebellar synapses. Cadm1 harbors a band 4.1 region and a PSD-95/Dlg/ZO-1 (PDZ)-binding motif in its cytoplasmic tail (Fujita et al. 2003). At the pre-synapse, Cadm1 associates with calmodulin-associated serine/threonine kinase (CASK) via the PDZ domain (Biederer et al., 2002). In contrast with pre-synapse, little is known about Cadm1 molecular complex at the post-synapse. The localization and clustering of Cadm1, neurotransmitter receptors, and channels can be crucial for the proper function of post-synaptic neurotransmitter receptor complexes, with PDZ domains acting as adaptors. Scaffolding molecules such as PSD-95 structurally organize macromolecular complexes in the post-synaptic density (Sala et al. 2001).

The C-terminal region of neuroligin (NLGN), another synaptic molecule, interacts with PSD-95, a multiscaffold protein which forms a complex with excitatory N-methyl-d-aspartate receptors (NMDAR) and Shank; PSD-95 assembly causes an imbalance of excitatory and inhibitory receptors (Gerrow and El-Husseini 2007). Multiple PDZ domain protein 1 (Mupp1) is another multi-PDZ scaffold protein containing a PDZ13 domain, and interacts with SynGAP and GABBR2 at PDZ13 (Kraplvinsky et al. 2004; Balasubramanian et al. 2007); Mupp1–SynGAP form a complex within the excitatory NMDAR signaling complex (Kraplvinsky et al. 2004). Although GABBR2 and Cadm1 carry PDZ-binding motifs in their terminal regions and localize to Purkinje cells and their dendrites (Luján and Shigemoto 2006; Fujita et al. 2012a), little is known about Cadm1–neurotransmitter receptor complexes at the post-synaptic membrane.

In this study, we examined Cadm1-binding scaffold proteins by pull-down assay. Cadm1 specifically interacts with Mupp1, but not with PSD-95, and may form a ternary complex with Mupp1–GABBR2 in the cerebellum. We also demonstrate that GABBR2 is up-regulated in Cadm1-KO mice

Materials and methods

Ethics statement

We followed the Fundamental Guidelines for Proper Conduct of Animal Experiments and Related Activities in Academic Research Institutions under the jurisdiction of the Ministry of Education, Culture, Sports, Science and Technology, and all of the protocols for animal handling and treatment were reviewed and approved by the Animal Care and Use Committee of Jichi University (approval numbers, H22-179, 10-179) and International University of Health and Welfare (approval numbers, D1008; 10118). Wild-type, Cadm1-KO mice (Fujita et al. 2006) (male mice) were used for the experiments.

Plasmid construct

GST-tagged pEF-BOS mammalian expression vector (pEB-GST)-RA175C and -RA175ΔC (lacking C-terminus EYFI), and pcDNA-Cadm1myc (pcDNA4/TO/myc-His expression vector) were previously described (Fujita et al. 2007, 2010). cDNA fragments encoding PDZ domains of rat Mupp1and mouse neuroligin3 (Nlgn3) were obtained from pcDNA3-GFP–Mupp1 (kindly gifted from Dr. Ullmer) and derived from mouse brain (C57BL/6J), respectively. They were amplified by polymerase chain reaction (PCR) using the primers: the forward primer for Mupp1(PDZ 1-5) and Mupp1(PDZ 1-8); 5′- AAGATGTTGGAAACCATAGAC -3′, reverse primer for Mupp1(PDZ 1-5); 5′- ATACCCTTCTAGAGGTCACAGAGGCA-3′, reverse primer for Mupp1(PDZ 1-8); 5′-CTGCTGCATTTCTAGATCATACGGCC-3′, reverse primer for Mupp1, the forward primer for Nlgn3: 5- GGATCCATGCAGACCTTGCACCCC-3′, reverse primer for Nlgn3: 5- GGATCCCTAGGAGTGTGAGTGGGGCAG-3′. PCR products for Mupp1 were amplified by one cycle of 95°C for 2 min, 10 cycles of 95°C for 30 s and 62°C for 6 min, and one cycle at 72°C for 7 min, and cloned into pGEM-T easy vector (Promega, Madison, WI, USA) and then subcloned in-frame into EcoRI site of the pEGFP-C1 vector (BD Bioscience-Clontech Laboratories, Inc., San Jose, CA, USA). PCR products for Nlgn3 were amplified by one cycle of 95°C for 2 min, 10 cycles of 95°C for 30 s and 62°C for 3 min, and one cycle at 72°C for 7 min, and cloned into pGEM-T easy vector, and then subcloned in-frame into BamHI sites of the pGEX-4T-3 vector (Amersham Pharmacia Biotech., Piscataway, NJ, USA). PSD-95–GFPGW1 (kindly gifted from Dr. El-Husseini) was subcloned into HindIII and EcoRI site of pEGFPN1 vector (Clontech).

Cell culture

COS and C2C5 cells were cultured in α-minimum essential medium (MEM) with 10% fetal bovine serum (FBS), at 37°C in a humidified atmosphere of 5% CO2. Neurons were isolated from the brains of rat at embryonic day 18 as described (Goslin and Banker 1998). Neurons were cultured using Neurobasal medium with 2% B27 supplement (Invitrogen, Carlsbad, CA, USA) and l-glutamine (0.5 mM). Cultures were incubated at 37°C in a humidified atmosphere containing 5% CO2. After 6 days in vitro (DIV), neurons were transfected with pcDNA3-GFP–Mupp1, pcDNA3-GFP (vacant vector), or pcDNA3-GABBR2 and myc-tagged CADM1 using the calcium phosphate method and incubated for 2 DIV.

Immunoblot analysis

Brain of male mice and transfected COS cells were lysed with the buffer (20 mM Tris-HCl, 150 mM NaCl, 1% Triton X-100, 1 mM ethylenediaminetetraacetic acid, 1 mM dithiothreitol, the protease inhibitor cocktail, and 1 mM phenylmethyl sulfonyl fluoride). After centrifugation at 15 000 g for 30 min, the cell extracts (50 μg protein) were subjected to pull-down assay, immunoprecipitation assay, and sodium dodecyl sulfate–polyacrylamide gel (7.5–12%) electrophoresis and immunoblot analysis using mouse and/or rabbit anti-GST (Santa Cruz Biotech., Santa Cruz, CA, USA), mouse anti-GFP (Roche Diagnostics, Rotkreuz, Switzerland) and/or rabbit anti-GFP (Clontech), mouse anti-Tubulin (Sigma-Aldrich, St Louis, MO, USA), mouse anti-Mupp1 (BD Biosciences, San Jose, CA, USA), and rabbit anti-GABBR2 (Alomone, Alomone Labs Ltd., Jerusalem, Israel).

Pull-down assays and immunoprecipitation assay

For in vitro pull-down assays, GST-fused Cadm1-C containing or not containing EYFI (GST–Cadm1-C, GST–Cadm1-delta-C, or GST), and Nlgn3-C were purified by glutathione sepharose beads from the extract of E. coli containing each plasmid. GST-fusion proteins (5 μg) was incubated for 3 h at 4°C with lysates from brain or transfected COS cells. The lysates were applied to glutathione column after centrifugation at 15 000 g for 10 min. After the column was washed with the lysate buffer, the bound proteins were eluted with glutathione buffer (10 mM glutathione), and subjected to the silver (Ag) staining and or immunoblot analysis using mouse anti-Mupp1, rabbit anti-GST, rabbit anti-GFP, and rabbit anti-GABBR2.

Immunoprecipitation was performed as previously described (Fujita et al. 2006, 2007). pBgs (mammalian expression vector)-GST–Cadm1 or pBgs-GST–Cadm1-delta-C (2 μg), and pcDNA3-GFP–Mupp1 (2 μg) or vehicle (pcDNA3 vector) were cotransfected into COS cells by Lipofectamine 2000 (Invitrogen) according to the specimen. COS cells were lysed with lysis buffer. Cell lysates were clarified by centrifugation at 16 000 g for 30 min. Cadm1 and Mupp1 were immunoprecipitated by mouse anti-GST or mouse anti-GFP. Cadm1 and Mupp1 in the immunoprecipitates were detected by immunoblot analysis using rabbit anti-GFP, rabbit anti-GST, respectively.


Wild-type and Cadm1-KO mouse brains were fixed in 4% paraformaldehyde in phosphate-buffered saline at 4°C overnight. Frozen sections (10-μm thick) were cut on a cryostat and immunostained with chicken anti-SynCAM1 (Cadm1; MBL), mouse anti-Mupp1, rabbit anti-GABBR2, mouse anti-myc (Nacarai tesque, Kyoto, Japan), rabbit anti-GFP. For the immunostaining for the C2C5 cells and neuronal cells, the transfected cells were fixed in 4% paraformaldehyde, washed with phosphate-buffered saline, and then incubated with rabbit anti-GFP and mouse anti-myc overnight at 4°C. Alexa Fluor 488- and Alexa Fluor 568-conjugated secondary antibodies against rabbit, mouse, and chicken IgG were purchased from Molecular Probes. Nuclei were detected using Hoechst 33342 (Molecular Probes, Eugene, OR, USA). The reactivity was viewed using a confocal laser-scanning microscope CSU-10 (Yokogawa Electric Corp., Tokyo, Japan) and a Leica SP5 confocal microscope (Leica Microsystems, Wetzlar, Germany). At least three animals per genotype were examined, and experiments were repeated three times. Quantification of staining intensities was done using LAS AF software (Leica Microsystems). The mean pixel value in the area of interest and in the same size area of the background was calculated. The background level was subtracted from the value found in the area of interest (in the molecular layer). Reported intensities were normalized to control, and the Student's t-test was performed for statistical analysis.

Quantitative real-time PCR

Total RNA was prepared from testes of wild-type and Cadm1-KO male mice (P 10) using RNeasy mini kit (QIAGEN Inc., Valencia, CA, USA) according to the manufacturer's specifications. Complementary DNAs were synthesized from total RNA (1 μg) using reverse transcriptase (Invitrogen) as described previously (Fujita et al. 2007). Real-time PCR analysis was performed using Applied Biosystems 7500 fast real-time PCR system (Applied Biosystems, Foster City, CA, USA) using the TaqMan Gene Expression Assays (Applied Biosystems) based on published sequences for genes encoding the respective mouse Mpdz (MUPP1) and GABBR2, and VIC-labeled mouse Gapd (VIC-labeled MGD probe; Applied Biosystems) as endogenous control. For each sample, the 20 μL total volume consisted of 10 μL TaqMan Fast Universal PCR Master Mix (2x; Applied Biosystems), 1 μL TaqMan Gene Expression Assays, and 5 μL of each first-strand cDNA sample. The real-time PCR fragments were amplified as follows: one cycle at 95°C for 20 s, 60 cycles at 95°C for 3 s, and 60°C for 30 s. Results were analyzed using student's t-tests (< 0.05 was considered statistically significant).


We focused on the interaction of proteins via their PDZ-binding motifs. We examined the proteins interacting with the sequence EYFI in the cytoplasmic tail of Cadm1 (Cadm1-C). Pull-down experiments using brain extracts revealed that purified GST-fused Cadm1-C peptides (GST–Cadm1-C) interacted with proteins with apparent molecular weights of 220 kDa, 110 kDa, and 85 kDa (Fig. 1a); via immunoblotting, we identified the 110-kDa protein as CASK located at the pre-synaptic membrane (unpublished data), and the 220-kDa protein as Mupp1, a scaffold protein containing 13 PDZ domains. However, the anti-Mupp1 antibody did not detect the band in the set of proteins bound to GST–Cadm1-C lacking the PDZ-binding motif EYFI (Cadm1-delta-C) (Fig. 1b), demonstrating that, in the brain, Cadm1 specifically interacts with Mupp1 via its EYFI-PDZ domain. The 85-kDa band (with asterisk) may contain both non-specific proteins interacting with GST and unknown proteins specifically interacting with Cadm1-C. However, Cadm1 did not interact with PSD-95, with which Nlgn3 can interact, whereas Nlgn3 did not interact with Mupp1 (Fig. 1c and d).

Figure 1.

Pull-down and immunoblot analysis for the proteins associated with cell adhesion molecule 1 (Cadm1)-C. (a) Proteins in the brain associated with Cadm1-C. The arrow (220 kDa) and open triangle (110 kDa) indicate the major proteins bound to GST–Cadm1-C. Asterisk (85 kDa) indicates the mixture of the non-specific band and the bands of unknown proteins specifically interacting with Cadm1-C. The major 220-kDa protein was identified as multiple PDZ domain protein 1 (Mupp1) by immunoblot analysis using anti-Mupp1, and the 110-kDa protein was identified as calmodulin-associated serine/threonine kinase (CASK) by anti-CASK immunoblotting (unpublished observation). (b) Mupp1 was associated with GST–Cadm1-C, but not with GST–Cadm1-delta-C lacking the PDZ-binding motif (EYFI). Interaction between Mupp1 (c) or PSD-95 (d) with Cadm1-C and Nlgn3-C was examined by pull-down and immunoblot analysis.

The intracellular association of Mupp1 with the Cadm1 PDZ-binding domain was further confirmed in COS and C2C5 cells by immunoprecipitation and immunostaining (Fig. 2). GFP–Mupp1 associated with Cadm1-C, but not with GST–Cadm1-delta-C (Fig. 2a), and Cadm1–myc and GFP–Mupp1 colocalized in C2C5 cells (Fig. 2b). Cadm1–myc alone or GFP–Mupp1 alone diffusely localized to the cytoplasm, whereas GFP–Mupp1 and Cadm1–myc were assembled and tightly colocalized in the intracellular region.

Figure 2.

Intracellular interaction between cell adhesion molecule 1 (Cadm1)-C and multiple PDZ domain protein 1 (Mupp1). (a) Immunoprecipitation and immunoblot analysis. pBgs-GST–Cadm1-C or pBgs-GST–Cadm1-delta-C was cotransfected with pcDNA3-GFP–Mupp1 or vehicle into COS cells. Cadm1 and Mupp1 were immunoprecipitated by mouse anti-GST and mouse anti-GFP, respectively. Cadm1 and Mupp1 in the immunoprecipitates were detected by immunoblot analysis using rabbit anti-GFP or rabbit anti-GST, respectively. (b) Colocalization of Cadm1 and Mupp1 in C2C5 cells. Cadm1–myc and GFP–Mupp1 were cotransfected into C2C5 cells and their intracellular localization was visualized by immunostaining with anti-myc and anti-GFP. Bar indicates 20 μm.

Mupp1 interacts with the PDZ regions of various receptors, including GABBR2 (Becamel et al. 2001; Balasubramanian et al. 2007). To determine which PDZ domain of Mupp1 associates with Cadm1-C, we implemented pull-down assays between Cadm1-C and truncated Mupp1 peptides of different molecular weights. Cadm1 interacted with Mupp1(PDZ1-5) with greater affinity than with Mupp1(PDZ1-13) or with Mupp1(PDZ1-8) (Fig. 3a). Multiple bands of GFP–PDZ(1-5) with molecular weight higher than 110 kDa were detected, and may be because of phosphorylation or ubiquitination of GFP–PDZ(1-5). Furthermore, Cadm1 colocalized with Mupp1(PDZ1-5), Mupp1(PDZ1-13), and Mupp1(PDZ1-8) in C2C5 cells (Fig. 3b).

Figure 3.

The region of multiple PDZ domain protein 1 (Mupp1) that interacts with cell adhesion molecule 1 (Cadm1)-C. (a) Identification of the region associating with Cadm1-C by pull-down and immunoblot analysis. Three different GFP–Mupp1 regions [GFP–PDZ(1-13), –PDZ(1-8), and –PDZ(1-5)] were transfected into COS cells, and cell lysates were incubated with Cadm1-C-conjugated beads; the bound GFP–Mupp1 domain was analyzed by immunoblot analysis using anti-GFP. (b) Identification of the Mupp1 region interacting with Cadm1 by immunostaining. Cadm1–myc and GFP–PDZ(1-13), –PDZ(1-8), –PDZ(1-5) were transfected. Colocalization of Cadm1–myc and the three GFP–Mupp1 regions was visualized with anti-myc (red) and anti-GFP (green). Bar indicates 20 μm.

GABBR2 has a PDZ-binding domain and interacts with Mupp1 (Balasubramanian et al. 2007). As both GABBR2 and Cadm1 are expressed in the developing cerebellum (Luján and Shigemoto 2006; Fujita et al. 2012a), we examined whether Cadm1 can associate and colocalize with GABBR2 in the developing cerebellum (Fig. 4). We detected not only Mupp1 and CASK but also GABBR2 in the set of proteins pulled down by GST–Cadm1C (Fig. 4a). Furthermore, Cadm1 colocalized with GABBR2 in the developing cerebellum (Fig. 4b). At day 5 post-natal (P5, an early stage), Cadm1 and GABBR2 colocalized in the soma of Purkinje cells before dendrite formation. At P8, GABBR2 staining was positive in the soma, dendrites, and dendritic arbors of Purkinje cells, whereas Cadm1 staining was positive around and along the dendrites of Purkinje cells. At P11, GABBR2 mainly occurred in the dendritic stems and soma of the Purkinje cells and also in the dendritic arbors, whereas Cadm1 expression increased and extended to the apical-distal region in the molecular layer. Cadm1 colocalized with GABBR2 throughout the dendritic arbors of Purkinje cells, in particular the basal region of the molecular layer, but Cadm1 staining was almost negative in the soma and the dendritic stems of Purkinje cells. Cadm1 was also expressed in granular cells, but GABBR2 expression was not detected in these cells. At P14, both Cadm1 and GABBR2 expression decreased in the molecular layer.

Figure 4.

Cell adhesion molecule 1 (Cadm1) and gamma-aminobutyric acid type B receptor (GABBR)2 form a complex in the cerebellum. (a) Interaction among Cadm1, multiple PDZ domain protein 1 (Mupp1), and GABBR2 in the mouse brain revealed by pull-down assay. GABBR2 (108 kDa) was detected in the proteins associated with Cadm1-C by immunoblotting with anti-GABBR2. (b) Distribution of Cadm1 and GABBR2 in the developing cerebellum. There are several types of synapses in the molecular layer of the cerebellum; Purkinje cells receive two excitatory afferents, parallel fibers, and climbing fibers. Most of the Cadm1 and GABBR2 signal colocalized in the soma of Purkinje cells at P5. At P8-11, the dendrites extended to the distal-apical region. Cadm1 expression increased in the distal-apical region in the molecular layer at P11. Cadm1 mainly colocalized with GABBR2 over the dendritic arbor of Purkinje cells. HM, highmagnified picture. Arrowheads indicate both Cadm1- and GABBR2-positive spines assembled on the dendrites. Cadm1 and GABBR2 intensities decreased at P14. Bars, 30 μm.

In the cerebellum, Mupp1 can be detected in Purkinje cells and in the granular and molecular layers (Sitek et al. 2003). In the molecular layer of the cerebellum at P11, Cadm1 and Mupp1 partly colocalized to the dendrites of Purkinje cells, although with different intensities; Mupp1 preferentially localized at the proximal dendritic portion, whereas Cadm1 appeared in the apical-distal dendritic portion (Fig. 5a). Compared with wild-type mice, anti-GABBR2 immunoreactivity was more intense in the molecular layer of Cadm1-KO mice, WT : KO = 1 : 1.45, whereas anti-Mupp1 immunoreactivity was slightly decreased in Cadm1 KO mice, WT : KO = 1 : 0.82 (Fig. 5a and b), and the levels of GABBR2 and Mupp1 in the cerebellum at Cadm1-KO mice were approximately double and a little lower, respectively (Fig. 5c). Thus, lack of Cadm1 up-regulates the level of GABBR2 in the cerebellum, although quantitative real-time PCR (qRT-PCR) indicated that the mRNA levels of Mupp1 and GABBR2 were not altered following knockout of Cadm1 (Figure S1).

Figure 5.

Alteration of gamma-aminobutyric acid type B receptor (GABBR)2 immunoreactivity in the cerebellum of Cadm1-KO mice (P 11). (a) Anti-Cadm1 (green) and anti-Mupp1 or anti-GABBR2 (red) visualized in the cerebellum of wild-type (WT) mice. Cell adhesion molecule 1 (Cadm1) and multiple PDZ domain protein 1 (Mupp1) were preferentially distributed in the distal-apical dendritic portion and the proximal dendritic region of the WT mice, respectively, but the Mupp1 intensity gradient was lost in the Cadm1-KO mice. Compared with WT, GABBR2 intensity increased in the molecular layer of the Cadm1-KO mice. Nuclei were stained with Hoechst (blue). Bar, 30 μm. To eliminate variations because of regional differences, sections were taken from the same region in the central part of the cerebella of the WT and homozygous Cadm1-KO mice. (b) The intensities of Mupp1 and/or GABBR2 in the WT and Cadm1-KO mice. Values are mean ± standard error (SEM). Student's t-test (*< 0.05). Pups: n = 3. Images: n = 8. (c) Immunoblot analysis of the alteration of the GABBR2 level in the cerebella of WT and Cadm1-KO mice with anti-Mupp1, anti-GABBR2, and anti-Tubulin.

We evaluated the colocalization of Cadm1 and Mupp1 or GABBR2 on neurons cultured in vitro. In hippocampal neurons, Cadm1–myc colocalized with GFP–Mupp1 at spines assembled on the dendrites (Fig. 6a) and also colocalized with GABBR2 (Fig. 6b).

Figure 6.

Localization of cell adhesion molecule 1 (Cadm1)-C, multiple PDZ domain protein 1 (Mupp1), and gamma-aminobutyric acid type B receptor (GABBR)2 in hippocampal neurons. Cadm1–myc and GFP–Mupp1 or GFP alone (a), or Cadm1–myc and GABBR2 (b) were cotransfected into hippocampal neurons and their intracellular colocalization was visualized with anti-GFP or anti-GABBR2 and anti-myc. Bar indicates 20 μm.


The Cadm1–Mupp1–GABBR2 complex in the cerebellum

Mupp1 is a multi-PDZ scaffold that is highly expressed in the brain and displays distinct expression patterns including hippocampal and cerebellar localization (Sitek et al. 2003). Mupp1 is highly enriched in synaptosomes, specifically in post-synaptic densities, and interacts with SynGAP and GABBR2 at PDZ13, 5-HTR2C at PDZ10, and JAM-A at PDZ9 (Becamel et al. 2001; Hamazaki et al. 2002; Kraplvinsky et al. 2004; Balasubramanian et al. 2007).

Cadm1 occurs at both sides of the synaptic cleft and is assumed to act as a synaptic adhesion molecule that clusters neurotransmitter receptors via the PDZ-binding domain in its C-terminus. Cadm1 interacted with Mupp1 at PDZ(1-5), and Mupp1 and GABBR2 were detected in Cadm1-associated proteins in the brain, suggesting that Cadm1 forms a ternary complex with synaptic receptors, including GABBR2 at the post-synaptic membrane via interaction with Mupp1. Therefore, it is likely that the Cadm1–Mupp1–GABBR2 complex localizes to the inhibitory synapses between Purkinje cells and parallel fibers. However, VGluT1 is down-regulated in the Cadm1-KO mice (Fujita et al. 2012a), suggesting that Cadm1 also localizes to the excitatory synapse. Within the excitatory NMDAR signaling complex, disruption of the Mupp1–SynGAP complex increases the number of synapses containing functional AMPA receptors by regulating p38 MAP kinase activity (Kraplvinsky et al. 2004). Cadm1 may anchor the GABBR2 complex and the SynGAP–NMDAR complex via Mupp1 to the inhibitory and excitatory post-synaptic membranes, respectively. A possible Cadm1–Mupp1–GABBR2 and Cadm1–Mupp1–SynGAP–NMDAR complex is illustrated in Fig. 7.

Figure 7.

Possible model for cell adhesion molecule 1 (Cadm1)–multiple PDZ domain protein 1 (Mupp1)–gamma-aminobutyric acid type B receptor (GABBR)2 complex at the post-synaptic membrane.

PSD-95 is the multiscaffold protein, which forms a complex with NMDAR at excitatory synapse (Gerrow and El-Husseini 2007). However, Cadm1 and Nlgn3 did not interact with PSD-95 and Mupp1, respectively (Fig. 1c and d). Therefore, the two synaptic adhesion molecules may form different receptor complexes, such as Cadm1-Mupp1 and Nlgn3-PSD-95, at the post-synaptic membrane. Cadm1–Mupp1 and Nlgn3–PSD-95 regulation system at excitatory and inhibitory synapse will be major issue for the future study.

Possible molecular mechanism of the up-regulation of GABBR2 in the Purkinje cells of Cadm1-KO mice

Cadm1 deficiency increased GABBR2 protein level, but not its mRNA level (Figure S1), suggesting that GABBR2 is more stable in the cerebellum lacking Cadm1. Cadm1 deficiency may cause a conformational change in Mupp1 and its destabilization, but also association of Mupp1 with other adhesion molecules that anchors the Mupp1–GABBR2 complex to the post-synaptic membrane and stabilizes it, resulting in an imbalance of inhibitory and excitatory synapses that may be linked with ASD pathogenesis.

Cadm1 and Mupp1 preferentially localized to the apical-distal and proximal dendritic portions of the molecular layer, respectively, suggesting that other cell adhesion molecules interact with Mupp1 in the proximal region. Cntnap2 may associate with the Mupp1–GABBR2 receptor complex because the cytoplasmic tail of Cntnap2 harbors a PDZ-binding motif, KKEWLI, similar to that of Cadm1, KKEYFI, (Biederer et al. 2002; Fujita et al. 2003) and interacts with Mupp1 (Horresh et al. 2008). In the cerebellum, Cntnap2 occurs in the proximal stems of dendrites and in the soma of Purkinje cells (Fujita et al. 2012b), and the Cntnap2-positive area was observed to extend to the apical-distal dendritic portion in the molecular layer of Cadm1-KO mice (unpublished data). Thus, it may be possible that the Cntnap2–Mupp1 complex contributes to GABBR2 stabilization by forming a ternary complex in place of the Cadm1–Mupp1 complex and the increased GABBR2 levels in the cerebella of Cadm1-KO mice.

However, Cadm1-KO mice exhibit some ASD-like symptoms and impaired USV,, suggesting that Cntnap2 or other cell adhesion molecules cannot sufficiently compensate for Cadm1 function (Fujita et al. 2012b). This deficiency may be because of differences in the targets of pre-synaptic members; the Cntnap2–Mupp1 complex may have a synaptic function distinct from that of the Cadm1–Mupp1 complex because Cntnap2 forms a cis-complex with members of the contactin family, which have their own target molecules for cell adhesion.

Loss of function of Cadm1 and up-regulation of GABBR2 in the cerebellum of Cadm1-KO mice

Infant rodents utilize USVs for critical mother–offspring interactions (Branchi et al. 2001). In addition to ASD symptoms (Takayanagi et al. 2010), Cadm1-KO pups (P10) display impaired USVs similar to the USVs of Foxp2(R552H)-KI pups (P10) who carry a mutation related to speech–anguage disorder (Fujita et al. 2008, 2012a). In contrast with Cadm1-KO pups, Foxp2(R552H)-KI pups carry a reduced number of synapses expressing Cadm1 and GABBR2 on the dendrites of Purkinje cells, whereas more GABBR2 occurs in the cerebella of Cadm1-KO mice. Furthermore, the NLGN3(R451C) mutation is found in ASD patients, and GABAergic transmission is increased in Nlgn3(R451C)-KI mice (Tabuchi et al. 2007), suggesting that GABBR2 up-regulation may not be related to the impaired USVs, but rather to other ASD-like symptoms of Cadm1-KO mice. The relationship between the physiological alteration of the Cadm–Mupp1-receptor complex and the impaired social interactions of Cadm1-KO mice, a symptom of ASD, will be a major focus of future study.

In conclusion, Cadm1 associated with Mupp1 at PDZ(1-5) and colocalized with Mupp1 and GABBR2 in the molecular layer of the cerebellum. Lack of Cadm1 increased the up-regulation of GABBR2 in the cerebellum. Cadm1 may therefore form a ternary complex with Mupp1–GABBR2 in the cerebellum.


This work was supported by Grants-in-Aid for Scientific Research (KAKENHI) of the Ministry of Education, Culture, Sports, Science and Technology, Japan (21200011, 21700377); Grants-in-Aid for Health Labour Scientific Research of the Ministry of the Health, Labour and Welfare, Japan (10103243). We are extremely grateful to all the families. We thank Dr. S. Tsukita, Dr. C. Ullmer, and Dr. A.E. El-Hussini for providing cDNA. The authors declare no conflict of interest.