Mice With A Cleavage-Resistant N-Cadherin Exhibit Synapse Anomaly in the Hippocampus and Outperformance in Spatial Memory Tasks

N-cadherin is a homophilic cell adhesion molecule that stabilizes excitatory synapses, by connecting pre-and post-synaptic termini. Upon NMDA receptor (NMDAR) activation by glutamate, membrane-proximal domains of N-cadherin are cleaved serially by a-disintegrin-and-metalloprotease 10 (ADAM10) and then presenilin 1(PS1, catalytic subunit of the γ-secretase complex). To assess the physiological signicance of the initial N-cadherin cleavage, we engineer the mouse genome to create a knock-in allele with tandem missense mutations in the mouse N-cadherin/Cadherin-2 gene (Cdh2 R714G, I715D , or GD) that confers resistance to proteolysis by ADAM10 (GD mice). GD mice showed a better performance in the radial maze test, with signicantly less revisiting errors after intervals of 30 and 300 sec than WT and a tendency for enhanced freezing in fear conditioning. Interestingly, GD mice reveal higher complexity in the tufts of thorny excrescence in the CA3 region of the hippocampus. Fine morphometry with serial section transmission electron microscopy (ssTEM) and three-dimensional (3D) reconstruction reveals signicantly higher synaptic density, signicantly smaller PSD area, and normal dendritic spine volume in GD mice. This knock-in mouse has provided in vivo evidence that ADAM10-mediated cleavage is a critical step in N-cadherin shedding and degradation and involved in the structure and function of glutamatergic synapses, which affect the memory function.


Introduction
N-cadherin is a key cell adhesion molecule required not only for dendrite arborization and axon guidance during development, but also for post-developmental generation, maintenance, and remodeling of synapses [1,2,3,4,5,6]. N-cadherin abounds in active zones of immature synapses, and at the periphery of mature synapses 7 , indicating that N-cadherin changes its location as well as its functional role between synaptogenesis and mature synapses. Transsynaptic contact mediated by the N-cadherin/catenin complex is a determinant of neuro-transmission and spine morphology, and reciprocally, synaptic activity modulates the expression, conformation, targeting, degradation and proteolytic cleavage of N-cadherin [8,9,10,11,12,13].
Inhibition of the ADAM10-meditated initial cleavage increases the steady-state level of the full-length Ncadherin. Notably, stabilization of N-cadherin by inhibition of ADAM10 activity increases the spine head volume and the recruitment of GluR1into the postsynaptic compartment, thereby regulating function of AMPA receptors at the hippocampal excitarory synapses. (28) However, most of these ndings are based on pharmacological experiments with cultured neurons or brain slices, and in vivo relevance had never been tested. (28) Thus, we set up to test the impact of ADAM10-mediated N-cadherin cleavage on animal behavior. To analyze this, on the basis of our previous studies [20], we created a line of knock-in mice that express N-cadherin with a mutation resistant to ADAM10-mediated cleavage (GD mice).
Here, we demonstrate that the GD mice show higher delity in a working memory-dependent task than WT mice in the radial maze test. In the fear conditioning test, GD mice show a tendency for enhanced freezing. This behavioral change is accompanied by higher complexity in the tufts of thorny excrescence and signi cantly higher synaptic density, signi cantly smaller PSD area, and normal dendritic spine volume in the CA3 region of the hippocampus. Our mouse model helps address the open questions as to the physiological and pathological impacts of the cleavage-defective N-cadherin on synapse morphology and transmission, learning and memory, and other brain functions.

Results
Generation of a knock-in allele encoding the ADAM10resistant mutant of N-cadherin Our previous in vitro analysis showed that a tandem substitution mutation near the transmembrane domain of mouse N-cadherin (R714G/I715D, hereafter GD; Fig. 1A) confers resistance to ADAM10 [20]. To assess the physiological signi cance of ADAM10-mediated cleavage of N-cadherin in the brain, we designed a targeting vector to replace exon 13 of the mouse Cdh2 gene with the GD mutant exon, by homologous recombination (Fig. 1A). After gene targeting of an ES cell line (Bruce4) derived from C57BL/6J strain, we screened Neomycin-resistant clones for a GD allele (Cdh2 R714G, I715D ) by southern blotting, PCR, and sequencing of the genomic DNA (data not shown). Chimeric mice generated through blastocyst injection of one of Cdh2R714G, I 715D/+ ES cell clones into C57BL/6J embryos were bred with WT mice to yield heterozygous (Cdh2R714G, I715D/+) mice with the genetic background of C57BL/6J. The oxed Neomycin-resistant gene cassette was removed by breeding with CAG-Cre driver C57BL/6J mice. After con rming mendelian inheritance of the GD allele, we consistently compared homozygous GD mutant (Cdh2 R714G, I715D/R714G, I715D ) and WT (Cdh2 +/+ ) male littermates born from heterozygous parents.
A mutant N-cadherin that is uncleavable by ADAM10 is relatively stable on the cell surface First, to assess the resistance of GD mutant cadherin to ADAM10-mediated cleavage, we expressed wildtype (WT) or GD mutant N-cadherins, each tagged with HA epitope, in CHO cells. Their distribution patterns to the endoplasmic reticulum (ER), Golgi apparatus, and plasma membrane were largely comparable (Fig. 1B). On the other hand, surface biotinylation pulse-labeling followed by serial immunoblot assay for the HA epitope revealed that the cell surface fraction of GD N-cadherin was signi cantly higher than that of WT, indicative of a longer half-life (~ 12 h vs. ~4 h; Fig. 2A, B). These in vitro data indicate that GD mutation prolongs the retention of the full-length N-cadherin on the plasma membrane, indicating that ADAM-10 mediated cleavage of N-cadherin affects the level of cell-surface Ncadherin.
GD mutation does not show obvious difference in the content of major synapse proteins in synaptosome Previous reports demonstrated that systemic and brain-selective null mutant mice of N-cadherin respectively exhibit severe systemic and neural anomalies [2,4,22], indicating a requirement of N-cadherin for development. In contrast, GD mice grow and breed normally, suggesting that the perturbation of the Ncadherin cleavage at the site does not cause obvious defects in embryonic and post-natal development.
Since cleaved ectodomain fragments of N-cadherin were barely detectable from the brain tissue samples, we examined the cytoplasmic fragment CTF1 in primary cultured cerebrocortical neurons and synaptosomal fractions from WT and GD mice. An antibody directed against the carboxyl terminus of Ncadherin detected native full-length N-cadherin and CTF1 in WT neurons, while CTF1 was not detected in GD neurons (Fig. 3B). CTF2, a g-secretase-mediated cleavage product of CTF1, was barely visible in this setting due to proteosomal degradation [20]. These data indicate the ADAM10-mediated cleavageresistance of N-cadherin R714G, I715D in vivo, in line with the enhanced stability of exogenously expressed Ncadherin R714G, I715D in the surface bio-tinylation assay ( Fig. 2A, B) and previous in vitro studies [20,21].
Next, we asked whether the content of excitatory synapse-related proteins is altered in the total brain lysate and the synaptosomal fraction of the GD mice. Immunoblot showed no obvious difference between steady-state levels of full length N-cadherins (GD and WT), and as to most of major post/pre/peri-synaptic proteins and activity-regulated proteins in those mice; i.e., ADAM-10, PS1, GluA1, GluA2, AKT, phospho-AKT, syntaxin 6, synaptophysin, and GLT-1 in synaptosome. (Fig. 3A).
Behavioral pro le of GD mice is largely normal, except for correct responses in a spatial working memory task Next, we asked whether the insertion of GD mutation alter mouse behavior. As a means of unbiased functional screening for possible alterations in the GD mouse brain, we systematically conducted a learning memory test at 4 months of age. Male GD and WT littermates (n = 12, 9) were comparable in most of the physical and behavioral indices measured, but signi cantly differed in hippocampusdependent tests and body weight, as follows (Supplemental Fig. S1-S4).
GD mice exhibited an unexpected phenotype in a spatial working memory task using 8-arm radial maze ( Fig. 4A-G): The time spent for the acquisition of the reward from the opening of the gates was signi cantly longer in GD mice (Fig. 4C, G). Intriguingly, however, the revisiting errors after intervals of 30 and 300 secs were signi cantly less in GD mice than in WT (Fig. 4E, F), indicating that GD mice execute a working memory-dependent task with higher delity than WT mice. A similar trend, albeit statistically insigni cant, was observed consistently in the probe test of the Barnes maze test, where GD mice tended to reduce the latency to nd the escape hole and the number of errors made (Suppl. Figure 3).
In the fear conditioning test, GD mice showed a tendency for enhanced freezing (an index of associative memory) in the late phases of fear memory acquisition on Day 1 (Fig. 5A, B, left) and of re-exposure to the shock box (without tone cue or foot shock) on Day 2 (Fig. 5A, B, middle). Interestingly, in the cued testing in altered context on Day 2, mutants showed enhanced freezing even before CS (Fig. 5A, B, right), indicating GD mice carried over learned behavior (freezing) the to a different context. Thus, this aberrant behavior could be interpreted as 'generalization'. Both wild-type and mutant mice reacted to unconditioned stimuli, although the time course of reaction was slightly different (Fig. 5C).

Hippocampal CA3 pyramidal neurons in GD mice develop anomalous synapses
To explore histological anomalies responsible for the behavioral alterations in GD mice, we conducted morphological analysis. The total weight and macroscopic architecture of the brain are normal in homozygous GD mice (Suppl. Figure 4). Given the abundance of Cdh2 mRNA in the hippocampus, the highest in the CA3 pyramidal neurons, followed by the dentate gyrus granule cells (http://mouse.brainmap.org/experiment/show/79632275), we rst examined the large specialized synapses between them.
Golgi stain of the stratum lucidum of the CA3 region revealed signi cantly higher complexity in the tufts of thorny excrescence (Fig. 6A), which receive mossy ber inputs from the granule cells. Statistical analysis revealed that thorny excrescence area ( Then, we focused on the synapses formed between associational/commissural bers and pyramidal neurons in the stratum radiatum of the CA3 region to conduct serial section transmission electron microscopy (ssTEM) for 3D morphometry (Fig. 7). The density of asymmetrical (mostly glutamatergic) synapses measured using the dissector method was signi cantly higher in GD mice than WT mice The associational/commissural ber-pyramidal neuron synapses in CA3 exhibit marginal transmission anomalies To explore possible functional alterations in the synapses formed between associational/commissural (A/C) bers and pyramidal neurons in the stratum radiatum of the CA3 region, we conducted electrophysiological recording of eld excitatory postsynaptic potentials (EPSPs), using acute hippocampal slices from GD and WT mice. Indices of basal synaptic transmission (i.e., ber volley amplitude, eld EPSP slope, and paired-pulse facilitation) were largely comparable between GD and WT slices ( Fig. 8A-C). However, the dependence of the synaptic response on the stimulus intensity was altered, with a trend of reduced EPSP slopes at higher stimulus intensities in GD mice (Fig. 8B). While high-frequency stimulation induced stable long-term potentiation (LTP) in both genotypes, the magnitude of LTP tended to be larger in GD slices (Fig. 8D, E).

Discussion
In this study, we generated the rst knock-in mouse model in which N-cadherin is systemically replaced into an ADAM10-uncleavable form. In contrast to the previous reports of N-cadherin knock-out mice, showing severe systemic and neural anomalies [2,4,22], GD mutant mice grow and breed normally, but exhibit a set of unique phenotypes that include 1) enhanced delity (or reduced error rate) in a spatial working memory task, and 2) synapse anomaly (more distribution density, less PSD area) in the CA3 regions.
N-cadherin-mediated cell/synapse adhesion is augmented by the cytoplasmic anchorage to the actin cytoskeleton via p120/β/α-catenins and terminated by the proteolytic cleavages, which release extracellular and cytoplasmic fragments [23,24]. The relative stability of GD mutant N-cadherin on the cell surface ( Fig. 2A, B) and the absence of the cytoplasmic fragment CTF1 (Fig. 3B) indicate augmentation of N-cadherin-based cell-cell adhesion in GD mice. Previous report by other group demonstrated that expression of a stable β-catenin mutant augments N-cadherin-mediated synaptic adhesion in vivo, without signi cantly affecting synaptic transmission and LTP [25,26]. Interestingly, in the mouse model, N-cadherin stabilization slows synapse remodeling and causes behavioral in exibility in reversal learning and memory extinction, and drug addiction [25,26]. The phenotype of GD mice is concordant as to the normal synaptic transmission and LTP. Interestingly, behavioral change similar to 'in exibility' in associative learning and memory was observed in GDmice, i.e, excessive freezing in the late phases of fear memory acquisition and of re-exposure to the context (Fig. 5A, B). Interestingly, in the cued testing in altered context on Day 2, mutants showed enhanced freezing even in the pre-CS period (Fig. 5A), indicating the occurrence of 'fear generalization'. This phenotype might be due to failure to eliminate previous fear memory when exposed to an altered context. Fear generalization has been attributed to hippocampal CA3 and dentate gyrus [27,28], thus, behavioral change observed in GD mice, i.e. better working memory and fear generalization, might be closely related to synapse anomaly in CA3 region of GD mice, observed in the present study.
Conditional knockout of ADAM10 in post-natal mouse brain perturbs N-cadherin cleavage, LTP, dendritic spine morphology (i.e., reduced density, stubby shape), and causes seizure [29]. Pharmacological inhibition of ADAM10 results in spine enlargement in vitro and in vivo [30]. These cases contrast sharply with quasi-normal phenotype of GD mice with normal-sized dendritic spines (Fig. 6A). Given that ADAM10 cleaves not only N-cadherin, but also many other synaptic substrates that modulate spine shape and adhesion to the presynaptic counterparts [31], the phenotypic difference is attributed to the pleiotropy of ADAM10. On the other hand, spine enlargement caused by GD mutant expression (in addition to the endogenous N-cadherin) [30] may be interpreted as a phenotype caused by simple Ncadherin excess [13], because spine volume is demonstrated to be normal in GD mice, which express normal level of full-length N-cadherin (Fig. 3A).
Dendritic spine volume and PSD area, two principal morphometric indices that re ect the excitatory synapse activity, tightly correlate [32]. Intriguingly, however, CA3 regions of GD mice densely contain disproportionate synapses with small PSD area on normal-sized spines (Fig. 7B-D). The mechanism underlying this anomaly is unclear, but there are a few possibilities that are not mutually exclusive. 1) Circuit immaturity: Synapses prematurely stabilized by excess N-cadherin may resist activity-dependent competitive elimination and pruning [33,34]. 2) Adaptive compensation (homeostatic scaling): Too many excitatory inputs into a given neuron may suppress maturation of individual synapses [35]. 3) Active zone con nement: N-cadherin excess may expand "non-active" synaptic contact around active zone [36]. Given that activity-dependent N-cadherin upregulation contributes to post-epileptic rewiring in CA3 region [37], similar disproportion and/or high-density synapse anomalies may constitute the pathology.

Conclusions
Overall, this study has provided in vivo evidence that ADAM10-mediated cleavage is a critical step in Ncadherin shedding and degradation at glutamatergic synapses, which alters hippocampal synapse morphology and animal behaviors related to spatial working memory and contextual fear memory. The unique mouse model established in this study will serve as a tool to explore the roles of N-cadherin processing in physiological synapse remodeling, and the pathological dysregulations which accompany epilepsy, Alzheimer's disease, anxiety disorders and other neurological disorders.

Animal experiments
Animal experiments were approved by institutional review committees and conducted in accordance with the regulations for the care and use of animals at Kyoto University and Nippon Medical School. We compared male littermates raised in the same cages, unless otherwise noted.
Generation and establishment of a line of Cdh2 R714G, I715D knock-in mice We designed a targeting vector that spans a 9.6 kb region of the mouse the Cdh2 gene to replace six bases in exon 13 from 6656 AGGATC to GGGGAT. After electroporation of the linearized vector into C57BL/6 mouse-derived Bruce 4 embryonic stem cells and selection of neomycin-resistant clones, we veri ed homologous recombination by Southern blotting and PCR (5' side probes; Fw-GATGCTGCTAACAGATGACTACAGA, Rv-AAAGGTACTGACAATAGGGCTCATA and 3' side probes; Fw-TCTCAAAGACTCCTATTGCTGTTCT, Rv-GTGTCTATAAGCTCCCATCAATGTC) of the genomic DNA. After blastocyst injection of the recombinant clones, we obtained chimera mice, which were crossbred with transgenic mice that ubiquitously express Cre-recombinase (CAG-Cre, C57BL/6, RBRC01828). Through backcrossing with C57BL/6J mice for more than two generations, we veri ed the removal of the neomycin-resistant gene cassette and the CAG-Cre allele by Southern blotting and PCR, and Mendelian transmission of the Cdh2 R714G, I715D allele. We bred heterozyzous Cdh2 R714G, I715D/+ mice to generate homozygous (Cdh2 R714G, I715D/R714G, I715D ; GD) and control (Cdh2 +/+ ; WT) littermates for experiments. The line has been deposited with the Center for Animal Resources and Development, Kumamoto University (ID 2027).
Genotyping DNA from tail snips was puri ed using the automatic DNA isolation system PI-50(KURABO, Japan).
Mouse Cdh2 gene were ampli ed using the following primers: Forward = CCA CTT CTA AGC ATG CAG GT; Reverse = AAT GAC TCC TAT TTG AGC ACA

Behavioral analyses
We conducted behavioral tests with male littermates during 3-5 months of age, using an established protocol [38]. The behavioral tests were conducted in the following order: general health and neurological screening (including body weight and temperature measurements, grip strength test, and righting, whisker touch, and ear twitch re exes), wire hang test, Barnes maze test, eight-arm radial maze test, and fear conditioning test.

Eight-arm radial maze test
The protocol was as previously described [39]. We used an apparatus with a central platform connected to eight arms (40-cm long with 25-cm high transparent walls, food pellet wells and sensors at distal ends) with automated shutters, which was placed 75 cm above the oor in a dim room with several spatial cues. The animals were starved for over a week to induce 15-20% weight loss and started on pre-training on the eighth day. We allowed a mouse to explore and eat food pellets for 30 min. Then, we set a pellet in a well and let a mouse explore and eat it, which was repeated eight times, once for each arm. In the spatial working memory task, we set a pellet in each well, and observed until a mouse ate the eight pellets. After each visit to an arm, the shutters were closed for 5 s with mice at the center. We videomonitored and analyzed arm choices, latency to acquire eight pellets, distance traveled, the number of times empty arms were chosen in the rst eight choices, and the number of revisiting errors. Image RM software was used for the control of shutters, data acquisition and analysis (see below for 'IMAGE ANALYSIS').

Contextual and cued fear conditioning test
To assess fear memory [40], mice were placed in a conditioning chamber (26×34×29 cm) in a sound attenuated room and allowed to explore freely for 2 min. The animals were presented with an auditory cue (55 dB white noise) which served as a conditioned stimulus (CS) for 30 sec. During the last 2 sec of the CS, mice were given a mild foot shock (2 sec, 0.35 mA) as an unconditioned stimulus (US). Two more CS-US pairings were presented with 120 sec interval. 24 hrs later, context test was performed. Cued test in an altered context was performed using a triangular box (35×35×40 cm) made of white opaque plexiglas, located in a different room. Following initial 3-min of pre-CS period, the CS was presented for 3 min. Data acquisition, control of stimuli (white noise and foot shock), and data analysis were performed automatically using Image FZ software.

Image analysis
The application programs for behavioral data acquisition and analysis (Image BM, RM, FZ) were created on the platform of ImageJ (http://rsb.info.nih.gov/ij/) by TM.

Plasmids
Ncad HA and NcadGD HA constructs, expressing the full-length human N-cadherin tagged with HA in Cterminus, as described elsewhere [20].

Preparation of synaptosome
Synaptosome was prepared according to the previous report of P.R. DODD et al [41]. In brief, mouse brain tissue was homogenized in 10 volumes of homogenization buffer (5 mM HEPES buffer, pH 7.4, containing 0.32 M sucrose, 50 mM sodium uoride, with phosphatase and protease inhibitor cocktail) in a Potter-Elvehjem homogenizer. The homogenate was centrifuged at 1000 g at 4°C for 5 min, twice. Supernatant was layered directly onto 1ml 1.2 M sucrose and centrifuged at 50,000 rpm at 4℃ for 10 min. The Pellet of intermediate layer was collected and diluted with ice-cold 0.32 M sucrose to a nal volume of 1 ml. The diluted suspension was then layered onto 0.8 ml of 0.8 M sucrose and centrifuged once again at 50,000 rpm at 4°C for 15 min. The pellet was suspended in RIPA buffer (with protease and phosphatase inhibitors. The pellet was then layered over discontinuous sucrose gradient (0.85-1.0-1.2 M) and centrifuged at 85,000g for 2h at 4°C. Synaptosomal fraction was obtained at the interface of 1 and 1.2 M sucrose.

Golgi Staining
Golgi staining was performed using a Rapid GolgiStain kit according to the manufacturer's protocol (FD NeuroTechnologies, Columbia,MD). Mice were sacri ced at the age of 4 months. The brains were immersed in impregnation solution for 2 weeks in the dark, and then stored at 4℃ for a week. The brain samples were dipped into isopentane pre-cooled with dry ice, mounted with TFM (TBS,Durham,Nc,USA), sliced (80μm) with a cryostat ,and then mounted on gelatin-coated glass slides. Subsequently, the sections were stained with staining solution. Measurements of the spine area were made by Image J and Metamorph (Molecular Devices).

Electrophysiology
Mice were decapitated under deep halothane anesthesia at the age of 14 to 15 weeks, and both hippocampi were isolated. Transverse hippocampal slices (380 μm) were cut using a tissue slicer and maintained in a humidi ed interface holding chamber. Electro-physiological recordings were performed as described [44]. Recordings were made in a submersion-type chamber maintained at 27.0 -27.5 ºC and superfused at 2 ml/min with saline composed of (in mM): NaCl, 125; KCl, 2.5; NaH 2 PO 4 , 1.0; NaHCO 3 , 26.2; glucose, 11; CaCl 2 , 2.5; MgCl 2 , 1.3. For recording EPSPs arising from the A/C ber-pyramidal cell synapses, a glass recording pipette lled with 2 M NaCl and bipolar stimulating electrodes were placed in the stratum radiatum of the CA3 region. The initial slope of EPSPs was measured on analysis. Single electrical stimulation was delivered at a frequency of 0.05 Hz, unless otherwise speci ed. To induce LTP, high-frequency burst stimulation (5 pulse at 100 Hz, repeated 10 times at 5 Hz) was delivered 3 times at an interval of 20 s. All recordings were made using a Multiclamp 700B ampli er (Molecular Devices, Sunnyvale, CA, USA).

Statistical analysis
Quantitative data were expressed as mean ± SEM. For statistical analyses, either two tailed t-test or ANOVA (one-way, two-way repeated measures, and repeated measures with two factors) was applied using StatView (SAS institute). F and p values represent the effects of genotype, unless otherwise noted.
conducted the EM analysis under the supervision of MK. MK and AK wrote the manuscript. MU and RT gave a critical advice throughout the experiments. All authors read and approved the manuscript.

Availabilities of data and materials
Data sharing not applicable to this article as no datasets were generated.

Figure 1
Introducing ADAM10-uncleavable GD mutations into N-cadherin/Cdh2 in the mouse genome, and testing the mutant distribution in cultured cells (A) Genomic DNA map of the mouse Cdh2 gene locus, mutant design, and targeting strategy. The gene targeting vector was designed to introduce tandem missense mutations (6656AGG ATC for 714RI → GGG GAT for GD) in the exon 13 of the mouse Cdh2 gene. The targeting vector was introduced into C57BL/6-derived ES cells by electroporation. Neomycin-resistant clones were screened for homologous recombination with genomic Southern blotting using two probes, and one such clone was injected into ICR blastocysts. The resulting chimera mice were intercrossed with C57BL/6N mice. The oxed neomycin-resistant cassette was removed from the allele by intercrossing the F1 mice with a line of CAG-Cre driver mice of C57BL/6 strain. The mutant and wild type alleles were discriminated routinely by PCR genotyping, using a pair of primers (arrowheads). See Materials and Methods for details. (B) Representative immuno uorescence images of FLAG-tagged, WT and GD mutant N-cadherin expressed in CHO cells. Their distribution patterns on ER, Golgi apparatus, and plasma membrane (green), as well as colocalization with endogenous β-catenin (red), were comparable. Nuclear DNA is stained with DAPI (cyan). Protein pro les of GD mouse brain is normal but for the absence of CTF1 (A) Representative immunoblot of endogenous proteins related to glutamatergic synapse and/or N-cadherin in the total brain lysate and a synaptosomal fraction from GD and WT mice. No recognizable difference was found in syntaxin 6, PSD95, synaptophysin, full-length N-cadherin, phospho-GluA1, GluA1, GluA2, GLT-1, phospho-AKT, AKT, αtubulin). (B) Immunoblot detection of endogenous N-cadherin with an antibody against a C-terminal region in the total lysates of primary cultured cerebrocortical neurons from WT and GD mice. While the full-length form (FL, 130kDa) was detected both in WT and GD samples, the cytoplasmic fragment CTF1 (45kDa) was detected only in those from WT, but not in those from GD mice (left). Similar results were obtained from the synaptosomal fraction (right). Non-speci c bands appeared around 50kDa (*) and above.