Abstract
In humans, mutations or deletions of atypical FAT cadherin genes are linked to autism spectrum disorder and cerebellar ataxia. However, their large genomic size and the enormous size of their encoded proteins have hampered functional studies, leaving the roles of FAT cadherins poorly understood. To address this gap, we investigated FAT2—an atypical cadherin selectively expressed in cerebellar granule cells—in murine cerebellar function. We demonstrate that FAT2 directly binds Cbln1, a secreted molecule essential for synapse formation and plasticity at Purkinje cell synapses. Furthermore, Fat2 deletion mice of both sexes selectively weakened the synaptic strength of parallel fiber synapses in the cerebellum and impaired motor behaviors. These findings reveal that FAT2 is indispensable for motor behaviors, likely through regulating Cbln1-dependent synaptic integrity.
Significance Statement
Abnormal motor behavior is a hallmark of many neurological and psychiatric disorders and a common symptom across numerous diseases. Growing evidence highlights the critical role of the motor system in elucidating the pathophysiology and treatment of mental disorders. Digging behavior—a movement characterized by forefeet scratching and/or hindfeet substrate kicking—is poorly understood at the genetic level. Here, we identify FAT Atypical Cadherin 2 (FAT2) as a binding partner of Cbln1, a synaptic organizer for cerebellar parallel fiber synapses. We demonstrate that deletion of cerebellar granule cell FAT2 impairs synaptic integrity and motor behaviors. These findings establish FAT2 as essential for synaptic integrity and the execution of fine motor and digging behaviors.
Introduction
The atypical FAT cadherins family in vertebrates consists of FAT1-4 (Sadeqzadeh et al., 2014; Avilés and Goodrich, 2017), and mutations or deletions of FAT cadherins have been linked to multiple neurological diseases, including autism and spinocerebellar ataxia (Blair et al., 2006; Bendavid et al., 2007; Abou Jamra et al., 2008; Sadeqzadeh et al., 2014; Butler et al., 2015; Chen et al., 2015; Nibbeling et al., 2017; Ameri et al., 2023; Tuncay et al., 2023). With large extracellular domains and intracellular domains that harbor several protein interaction sites, FAT cadherins are well positioned to mediate multiple cellular functions. The intracellular domains of FAT cadherins are intensively investigated in invertebrates and vertebrates. Intracellular domains of FAT2 regulate axon terminal organization via binding to scaffold proteins in Drosophila olfactory receptor neurons (Vien et al., 2024). Intracellular domains of FAT3 control neuronal morphology (Deans et al., 2011), synapse formation, and high temporal frequency light response in mouse retina (Avilés et al., 2025), likely via cytoskeletal regulators and synaptic proteins (Avilés et al., 2022, 2025). In contrast, the ligand(s) and functions of the extracellular domains of FAT cadherins are largely unknown. FAT2 transcellularly interacts with an unknown protein to maintain the WAVE (WASP family verprolin homolog regulatory) complex in follicular epithelial cells of Drosophila melanogaster (Williams et al., 2022). FAT4 likely transcellularly interacts with Dachsous1 cadherins to regulate the apical plasma membrane organization in the embryonic cerebral cortex (Ishiuchi et al., 2009). Thus, the functions of FAT cadherins, especially the extracellular domain, in vertebrates are still largely unknown.
Abnormal motor behavior is a key feature of several neurological and psychiatric disorders and is a prevalent symptom within many other diseases. There is an increasing awareness of the essential role of the motor system in understanding the pathophysiology and treatment of mental disorders (Northoff et al., 2021). A key element of the motor behavior repertoire of rodents is digging behavior. Digging behavior is a movement consisting of scratching with the forefeet and/or kicking out substrate with the hindfeet. Previous research has shown that lesions of different brain regions lead to impaired digging behavior (Thompson et al., 1989, 1990). Comparative studies provided the first insight into relevant genetic factors. For example, the field group of old-field mice dig larger burrows with longer entrances than deer mice, and the tunnel length was found to be associated with distinct genetic regions (Weber et al., 2013). Moreover, aberrant digging behavior is observed in genetic mouse models for neurological and psychiatric conditions (Deacon et al., 2001; Deacon, 2006; Kouser et al., 2013; Jirkof, 2014; Sungur et al., 2014; de Brouwer et al., 2019). However, those genetic models display a variety of additional behavioral phenotypes, complicating the interpretation of the role of the specific set of genes involved in regulating this type of fine motor behavior. Therefore, the gene(s) and circuit(s) that are selectively essential for this motor behavior are still largely unclear.
Through combining a newly generated FAT2 conditional knock-out mouse line, electrophysiology, and biochemistry methods, we demonstrate here that Cbln1 is the endogenous ligand of FAT2. In addition, we reveal that the deletion of cerebellar granule cell FAT2 impairs synaptic strength and digging behaviors. We conclude that FAT2 is required for synaptic integrity and motor behaviors, likely through a Cbln1-dependent mechanism in the cerebellum.
Materials and Methods
Animals
Mice were housed in the animal facility of the Peking University Shenzhen Graduate School. The mouse line on a C57BL/6J background carrying flanked-loxP (flox) FAT2 alleles (Fat2fl/fl) was generated by Biocytogen Pharmaceuticals. Female Fat2fl/fl mice were crossed with male Atoh-1 Cre mice (Matei et al., 2005) to obtain Atoh-1Cre; Fat2fl/wt mice. Male Atoh-1Cre; Fat2fl/wt mice were further crossed with female Fat2fl/fl to obtain Atoh-1Cre; Fat2fl/fl mice (cKO) for experiments. C57BL/6J mice were purchased from Guangdong Medical Laboratory Animal Center (China) and the Ai14 mice were obtained from Jackson lab (JAX#007908). The littermate control mice and cKO mice (Fat2fl/fl) of both sexes were used for electrophysiological experiments (between Day 21 and Day 28) and of male mice were used for behavior tests (male mice at the age of 2–5 months). Transgenic mice were genotyped by PCR of tail DNA, using the following primers spanning exons 5–11 of Fat2: Forward: 5′ GTTTATACAATGCCACTAG CCAACCA 3′; Reverse: 5′ TGTATCGGAGGTCAGAAGCTGAATAG 3′. Genotyping for the Cre allele was conducted by PCR with the following primers: Forward: 5′ GAAACTCATCAAATATGCGTGTTAGTG 3′; Reverse: 5′ AGTGCCCCTCAATCTCTTCAAATTCTG 3′.
All mice were maintained under a 12 h light/dark cycle at a temperature of 22–25°C with unrestricted access to tap water and standard chow. All animal procedures followed the guidelines approved by the Peking University Shenzhen Graduate School Animal Care and Use Committee and Shenzhen Bay Laboratory Animal Care and Use Committee.
Construction of expression vectors
We designed PCR amplification primers based on the CDS region sequence of Fat2 (Gene ID: 245827) in the NCBI database. Total RNA was isolated from the mouse cerebellum and reverse transcribed to obtain cDNA, which served as the template for PCR amplification using the designed Fat2 primers. The amplified Fat2 DNA fragment was then inserted into the linear pcDNA-3.1 plasmid backbone using the In-Fusion recombinase, yielding the full-length Fat2 expression plasmid (pcDNA3.1-Fat2). The extracellular domain of Fat2 consists of 33 cadherin repeats. Specific primers were designed according to the DNA sequences corresponding to these cadherin repeats. Using pcDNA3.1-Fat2 plasmid as a template, various Fat2 DNA truncations were PCR amplified. The resulting amplicons were then subjected to homologous recombination using the In-Fusion recombinase to obtain the corresponding truncated plasmids, including the DNA fragment containing only the first two cadherin repeats (Fat2-EC1-2); the DNA fragment lacking the first two cadherin repeats (Fat2-ΔEC1-2); the DNA fragment lacking the first five cadherin repeats (Fat2-ΔEC1-5); the DNA fragment lacking the first eleven cadherin repeats (Fat2-ΔEC1-11); and the DNA fragment devoid of all extracellular cadherin repeats (Fat2-ΔEC33).
Primary cell culture
Primary granule cell cultures were prepared as previously described (Yang et al., 2019). Briefly, cerebella were isolated from postnatal days (P) 5 to 7 of C57BL/6J pups. Cerebellar tissues were dissected out and treated with 0.25% trypsin at 37°C for 20–25 min and then centrifuged (×1,000 rpm) for 1 min. The supernatant was removed and the precipitate was added with 10% fetal bovine serum (FBS). Stood for 5 min and the middle layer of white granule cells was pipetted onto a 70 μm cell strainer and the filtrate was collected. The mixture was then centrifuged (×1,000 rpm) for 5 min, and the cell pellet was gently resuspended in low-glucose cultured medium [DMEM supplemented with 10% horse serum, 1% Penicillin-Streptomycin Solution (PS), Invitrogen, #10567014] and plated onto poly-d-lysine coated wells. After 24 h, the cell culture solution was replaced with a high-glucose cultured medium (DMEM supplemented with 10% horse serum, 1% PS, 4.5 g/L glucose). The length of axons was analyzed every 12 h until 96 h after culturing.
Cell aggregation and cell surface binding assays
Cell aggregation experiments were carried out essentially the same way as in our previous study (Qin et al., 2025). Cultured HEK293T cells were cotransfected with plasmids (tdT; Nrxn1β-SS4++eGFP, Nrxn1β-SS4-+eGFP; Nlgn4+tdT; Fat2+eGFP; Fat2+tdT) by Lipo2000 (Invitrogen) and waited for 48 h for expression. Cells were washed with prewarmed PBS and gently tapped on the bottom of the dishes to detach those cells. Cells were transferred into 2 ml tubes and centrifuged (×1,000 rpm) for 5 min. Those resuspended cells were gently pipetted up and down after removing the supernatant. Cells from two groups were mixed in equal volumes and incubated under gentle rotation at room temperature for 1 h using a tube rotator. Subsequently, 40 μl aliquots of the mixture were deposited onto electrostatically charged glass slides, followed by immediate imaging.
Expression vectors for GluD2, Fat2, and all truncations together with or without pEGFP-C1 were transfected into HEK293T cells. Transfected cells expressing GluD2, Fat2, or Fat2 truncations were incubated with recombinant Flag-Cbln1 (20 μg/ml each) and/or the supernate with soluble Nrxn1α in HBSS containing 2 mM CaCl2 for 1 h at room temperature. The plasmid containing ectodomain of Nrxn1α-SS4+ (sNrxn1α) was overexpressed in HEK293T cells and the supernate was concentrated and collected. After washing, cells were fixed with 4% paraformaldehyde (PFA) and immunostained with rabbit anti-HA and mouse anti-Flag, followed by incubation with species-specific Alexa Fluor 488- or 546-conjugated secondary antibodies.
Immunocytochemistry
Coverslips with HEK293T cells or cultured primary cerebellar granule cells were gently washed three times with PBS before fixation at room temperature with 4% PFA plus 0.3% Triton X-100 in PBS for 10 min and then blocked with 5% BSA in PBS for 1 h at room temperature. A primary incubation with primary antibodies (anti-HA, 1:1,000, rabbit monoclonal antibody #3727, Cell Signaling; anti-Flag, 1:1,000, mouse monoclonal antibody #F1804, Sigma-Aldrich; anti-NeuN, 1:1,000, rabbit polyclonal antibody, #ABN78, Millipore; anti-tubulin, 1:1,000, mouse monoclonal antibody #HC101, TransGen Biotech; anti-vGluT1, 1:1,000, guinea pig polyclonal antibody #135304, Synaptic Systems) overnight at 4°C followed by three washes in PBS and incubated with fluorescence-conjugated secondary antibodies (1:1,000, Alexa 488, 546, 633, Invitrogen) for 1 h at room temperature. Excess secondary antibody was washed off with another round of PBS followed by mounting on a coverslip with Fluoromount-G (SouthernBiotech, 0100-20) for the subsequent image.
RNAscope fluorescent in situ hybridization and immunohistochemistry
Mice were anesthetized and perfused with 1× PBS for 10 min followed by 4% PFA for 10 min. Brains were removed and stored in 4% PFA overnight at 4°C. The brain tissues were cryoprotected with 10, 20, and 30% sucrose gradient and then embedded in OCT (catalog #4586, Scigen) and rapidly frozen on dry ice. Transverse sections at 20 μm were cut at −20°C using a cryostat (CM1050, Leica) and mounted directly on Superfrost Plus slides (catalog #J1800AMNZ, Epredia). Fluorescent in situ hybridization (FISH) and immunohistochemistry were performed using the RNAscope multiplex platform (Multiplex Reagent Kit, catalog #323100; Co-detection ancillary Kit, catalog #323180, Advanced Cell Diagnostics) following the manufacturer's instructions. RNA probes for FAT2 (catalog #507561, Advanced Cell Diagnostics) and primary antibody against NeuN (1:300, rabbit polyclonal antibody, #ABN78, Millipore) and calbindin (1:300, mouse monoclonal antibody, C9848, Sigma-Aldrich) were used. Secondary antibodies were Alexa Fluor conjugates (1:500; Thermo Fisher Scientific). Samples were mounted with Fluoromount-G (SouthernBiotech, 0100-20). Images were acquired using an OLYMPUS FV3000 confocal microscope with a 60× oil-immersion objective or OLYMPUS SLIDEVIEW VS200 with a 20× objective.
qRT-PCR
The cerebellar tissue was isolated from cKO and littermate control mice at the age of P60. A TRIzol-up kit was used to extract and purify total mRNA according to the manufacturer's instructions, and complementary DNA synthesis was performed by reverse transcription of each sample using a TransScript One-Step gDNA Removal and cDNA Synthesis SuperMix (AT311, TransGen). Real-time PCR was performed using the QuantStudio 5 Real-Time PCR System (Applied Biosystems) with SYBR Green detection (AQ131, TransGen) in a two-step reaction. Gene expression fold changes were calculated with the comparative Ct method and normalized to Gapdh. The forward (F) and reverse (R) qPCR primers (5′–3′) used were as follows: Gapdh F: GGCATTGCTCTCAATGACAA; R: CCCTGTTGCTGTAGCCGTAT. Fat2 F: GACGACCTGGCCTCTTCTGGT; R: TGGGCATCTCTTGGTCTCGCA.
Immunohistochemistry
Immunohistochemistry experiments were performed similarly to our previous studies (Zhang et al., 2015; Qin et al., 2025). Adult mice from cKO and littermate control mice were perfused with PBS after anesthetization with isoflurane and then perfused with ice-cold 4% PFA in PBS. The cerebellum was dissected out and further fixed in ice-cold 4% PFA overnight and cryoprotected in ice-cold 30% sucrose for 48 h. Then, 30 µm sagittal sections were collected using a cryostat (Leica CM1050). Sections were intensively washed with PBS, then blocked with blocking buffer (PBS with 0.3% Triton X-100 plus 5% BSA) at room temperature for 30 min, and then incubated with primary antibodies overnight at 4°C (parvalbumin, 1:1,000, mouse monoclonal antibody, P3088, Sigma-Aldrich; S100, 1:1,000, guinea pig polyclonal antibody, 287004, Synaptic Systems; vGluT1, 1:1,000, guinea pig polyclonal antibody, #135304, Synaptic Systems; vGluT2, 1:1,000, rabbit polyclonal antibody, #135402, Synaptic Systems; GluA2, 1:500, mouse monoclonal antibody, MAB397, Sigma-Aldrich; GluA4, 1:1,000, rabbit polyclonal antibody, AB1508, Millipore; vGAT, 1:1,000, rabbit polyclonal antibody, #131002, Synaptic Systems; vGAT, 1:1,000, guinea pig polyclonal antibody, #131004; calbindin, 1:1,000, mouse monoclonal antibody, #C9848, Sigma-Aldrich). Sections were then washed three times in PBS followed by the incubation of secondary antibodies (1:1,000, Alexa 488, 546, 633, Invitrogen) for 2 h at room temperature and washed three times again with PBS. Sections were then mounted on a coverslip with Fluoromount-G (SouthernBiotech, 0100-20), and images were collected from cerebellar lobules IV/V by using an Olympus confocal microscope (FV3000). All acquisition and analyzed parameters were kept constant among different conditions within experiments.
For imaging of thick slices after recording, biocytin-injected neurons via patch pipette and then were stained with streptavidin Alexa 555 (1:1,000, #S21381, Invitrogen) and images of labeled individual PCs were acquired using an Olympus confocal microscope (FV3000) with a 60× oil-immersion objective. Sequential optical sections of 1,024 × 1,024 pixels were taken at 1.0 μm intervals along the z-axis, and three-dimensional (3D) reconstruction of whole cells was made using Imaris 10.0 (Bitplane) to calculate the number of intersections and the dendritic filament length.
Nissl staining
Nissl staining was performed as described previously (Zhang et al., 2015). Briefly, following perfusion and fixation, cerebellar sagittal sections (30 µm) were collected with a cryostat (Leica CM1050). Sections were washed with PBS and put on a coverslip. Then those sections were immersed with 4% PFA for 15 min and rinsed with double distilled water followed by immersion in Nissl Staining Solution (Beyotime, C0117) at 55°C for 20 min. The sections were rinsed in double distilled water and differentiated in 95% ethanol to get the best results. After that, sections were cleared in xylene. Sections were finally mounted with Fluoromount-G (SouthernBiotech, 0100-01).
Electrophysiology
Electrophysiological recordings were performed as described previously (Zhang et al., 2015; Qin et al., 2025). Control littermates and cKO mice (aged P21–P28) were anesthetized with isoflurane, and brains were rapidly transferred into ice-cold oxygenated low-Ca2+ artificial CSF (aCSF) containing the following(in mM): 125 NaCl, 2.5 KCl, 3 MgCl2, 0.1 CaCl2, 25 glucose, 1.25 NaH2PO4, 0.4 ascorbic acid, 3 myo-inositol, 2 Na-pyruvate, and 25 NaHCO3; pH was adjusted to 7.4. Sagittal slices of 250 µm cerebellum were collected using a vibratome (VT 1200S; Leica) and incubated in a 34°C thermostatic water bath for 35 min and then transferred to room temperature (22–25°C) for >1 h before recordings. The recording chamber was perfused with oxygenated aCSF containing the following (in mM): 1.3 MgSO4, 2.5 CaCl2 instead of 3 MgCl2, 0.1 CaCl2. Purkinje cells (PCs) in cerebellar lobules IV/V were voltage-clamped at −70 mV with borosilicate glass pipettes (3–4 MΩ for PCs). The internal pipette solutions contained (in mM) the following components for the following recordings: (1) EPSC recordings—140 Cs-methanesulfonate, 10 CsCl, 5 Na2-phosphocreatine, 2 Mg-ATP, 0.3 Na2-GTP, 0.5 EGTA, 2 QX314 (pH 7.3, adjusted with CsOH); (2) IPSC recordings—140 CsCl, 10 TEACl, 5 phosphocreatine, 4 Mg-ATP, 0.3 Na2-GTP, 0.5 EGTA, 2 QX314 (pH 7.3, adjusted with CsOH). For biocytin-injected test, internal solution was made with 2 mg/ml biocytin (Sigma, #B4261). Focal square pulse stimuli (duration 50 µs, amplitude 0–30 V or 1–100 µA) were applied with a bipolar stimulation electrode (FHC). CF-PC EPSCs were recorded with a cocktail of NBQX (0.5 µM, to reduce the EPSC size and minimize the clamping error), picrotoxin (50 µM), and APV (10 µM) in the aCSF and were identified by their characteristic all-or-none response and paired-pulse depression. PF-PC EPSCs were recorded with a cocktail of picrotoxin (50 µM) and APV (10 µM) in the aCSF and were identified by their characteristic paired-pulse facilitation. Basket cell-PC IPSCs were recorded with a cocktail of NBQX (20 µM) and APV (10 µM) in the aCSF. The basket cell-PC IPSCs recordings were achieved through two complementary strategies: (1) positioning the recording electrode near Purkinje cells to minimize activation of stellate cells located in the outer half of the molecular layer and (2) leveraging the all-or-none inhibitory postsynaptic currents (IPSCs) characteristic of basket cell–Purkinje cell synapses for unambiguous identification. Miniature EPSCs (mEPSCs) were recorded with TTX (1 µM) and picrotoxin (50 µM) in the aCSF. Miniature IPSCs (mIPSCs) were recorded with TTX (1 µM) and NBQX (20 µM) in the aCSF. Only those recordings of PCs with a stable series resistance (<15 MΩ) were used for further analysis. The series resistance (measured by applying 10 ms, −10 mV voltage pulses) was not compensated and was monitored before and after experiments. All electrophysiology data was acquired with a MultiClamp 700B amplifier and Digidata 1550A (Molecular Devices) and analyzed using Clampfit software (v10.7, Molecular Devices). In all electrophysiology experiments, the experimenter was blind to the conditions, both during acquisition and quantification.
Behavioral tests
Behavioral tests were conducted during the day (light-on period) using age-matched male littermates (2–5 months). The ages of those mice within the same cohort were within 1 month difference. All mice for each behavioral experiment were handled by the operator for 2–3 min per day for 3 consecutive days for habituation. Behaviors were tested in the following order: gait analysis, open field, grip strength test, sucrose preference test, grooming, marble-burying test, single-pellet reaching task, three-chamber sociability paradigm, novel object recognition test, rotarod, balance beam assay, elevated plus maze, and Y-maze. All experiments were performed between 8:00 A.M. and 8:00 P.M. All behaviors were measured under conditions of white light (∼40 lux). Mice were moved from the holding facility to the testing room at least 1 h before testing began. Mice rest for at least 1 d between different tests if with the same cohort. All behavior assays were conducted and analyzed by researchers blinded to genotypes.
Open field test
Open field test was performed in a white plastic box (45 cm length × 45 cm width × 45 cm height), illuminated at both ∼40 lux or using infrared light. The central zone was a 15 cm × 15 cm square. Mice were put into the center zone and their movements were recorded for 30 min with a video camera. Boxes were cleaned with 75% ethanol between trials. Distance traveled and time spent in each area were analyzed by Viewer 3.0.
Grooming test
An adult mouse was placed individually into a standard mouse cage (43 cm length × 26 cm width × 18 cm height, illuminated at ∼40 lux). A thin (1 cm) layer of bedding reduced neophobia and prevented digging, a potentially competing behavior. After a 10 min habituation period in the test cage, each mouse was scored using a camera for 10 min for cumulative time spent grooming all body regions. The total time spent grooming was analyzed manually by an experienced observer watching the recorded video.
Gait analysis
For the gait analysis, the hindfeet and forefeet of the mice were coated with green and purple nontoxic paints, respectively. The mice were then trained to walk along a 50-cm-long, 10-cm-wide open-top runway (with 10-cm-high walls) into an enclosed box. All mice were given three runs per day for 3 consecutive days. A fresh sheet of white paper was placed on the floor of the runway for each run. The footprint patterns were assessed quantitatively by four measurements: front base, hind base, forefoot stride, and hindfoot stride.
Grip strength test
Grip strength was performed using a high-precision digital force gauge (PUYAN). Mice were held by the tail, allowed to grasp a triangular bar with their forelimbs or all four limbs, and pulled away from the bar horizontally until they lost their grasp. Force was recorded for three trials; the values reported are the averages of these trials. The same experimenter, who was blinded to genotype, conducted all measurements. Results are reported as gram force (gf).
Single-pellet reaching task
Single-pellet reaching task were performed as described previously (Esposito et al., 2014). For food deprivation, baseline body weight was first determined. Then, food was restricted to 0.1 g per 1 g body weight daily for 2 d, adjusting portions to maintain ∼90% baseline body weight throughout training. During the shaping stage (Days 3–7), mice were acclimated to the training chamber; first in pairs (Day 1: 20 seeds/mouse, 20 min) and then individually (Day 2). On the subsequent days, forefoot dominance was determined and mice had to achieve ≥20 reaching attempts in 20 min with ≥70% preference for one forefoot to proceed. If mice used their tongue, the tray was slightly repositioned to encourage forefoot use. Mice failing to meet the criteria within a week were excluded. For the training stage (Days 8+), the double-slit chamber side was used, with seeds placed on the preferred foot's side, and the following reaching outcomes were recorded daily: (1) success, retrieve and eat seed; (2) drop, grasp but lose seed; and (3) fail, miss or knock seed off. Sessions were conducted for 30 attempts or 20 min/day, and then mice were returned to their home cages with their daily food ration. The reaching process was recorded by a high-speed video (1080p/120 fps) in the training stages and then employed DeepLabCut to identify each mouse's preferred forefeet. The trajectory was reconstructed and analyzed with a custom program.
Balance beam test
A 100 cm and 1 cm in diameter long stainless-steel beam was suspended 100 cm in the air. A dark box was positioned at one end of the beam (end) and a bright light was positioned above the other end (starting point). A starting line was marked ∼10 cm from the starting point so that the total distance between the starting line and the goal box was 80 cm. Mice were trained in 20 mm square rods for three consecutive days and three trials each day before testing. On the fourth day, mice were sequentially subjected to rod-based motor coordination tests with varying dimensions (20 mm square, 10 mm square, 20 mm round, and 10 mm round rods). Mice were given three trials with 60 s between each trial, and their best trial was used for analysis. The beam was cleaned with 75% ethanol between trials.
Three-chamber test
The three-chamber test followed our previous study (Qin et al., 2025) in a three-chambered setup (60 cm length × 40 cm width × 25 cm height with 20 cm width side and middle chambers). Littermate controls and cKO mice were first allowed to explore both the left and right chambers for 10 min. Then, an age-matched stranger was placed and caged in one side chamber, and the test mouse was allowed to explore for 10 min freely. Immediately after, another age-matched stranger was placed and caged in the other side chamber, and the test mouse was allowed to freely explore for another 10 min. The box was wiped with 75% ethanol and air-dried between each mouse. Behaviors were under video monitoring. Time in close interaction was defined as instances where the mouse's nose contacted the side chamber or when the mouse exhibited orientation toward the side chamber within a distance of 2 cm.
Rotarod test
Rotarod testing consisted of three trials per day over 4 consecutive days (trials 1–12), similar to our previous study (Qin et al., 2025). The accelerating rotarod equipment was programmed as starting from 4–40 rpm within 5 min (trial 1–6) or 8–80 rpm (trial 7–12) within 5 min, and the latency to fall was recorded for each trial. Each trial ended when a mouse fell off, made one complete backward revolution while hanging on, or reached 5 min.
Novel object recognition test
Each mouse was placed in an empty squared open field (45 cm length × 45 cm width × 45 cm height) to freely explore it for 10 min. The mouse was returned to its home cage for at least a 40 min pause during which two identical objects were put on the three equinoxes of the diagonal line of the open field for the mice to explore for 10 min. The mouse was returned to its home cage for at least a 40 min pause during which a novel object replaced one previously explored object for the mice to explore for another 10 min. All objects were cleaned with a 75% ethanol solution before their introduction in the open field. The time spent exploring each object was recorded.
Y-maze test
The Y-maze setup contained three white arms placed at 120° angles. Mice were placed in the center of three arms and allowed to freely explore the maze for 5 min under video monitoring. Only consecutive visiting all three different arms were counted as a correct trial; the correct alternation percentage was calculated as [number of correct trials / total trials] × 100%.
Sucrose preference test
Mice were water-deprived overnight and subsequently housed individually in standard cages. Each mouse was provided with two drinking bottles: one containing tap water and the other one containing 1% sucrose solution. After a 12 h exposure period, all bottles were removed, and individual consumption of each solution was recorded. Sucrose preference was calculated as the ratio of sucrose solution consumption to total fluid intake (the sum of sucrose and water consumption).
Elevated plus maze test
The elevated plus maze device consisted of two open arms (30 cm length × 5 cm width) and two closed arms (30 cm length × 5 cm width × 10 cm high walls) crossed at 90° angles, with a square platform (5 cm length × 5 cm width) at the middle intersection. The device was 50 cm from the ground. Mice were put onto the central platform and allowed to explore freely for 5 min under video monitoring. The elevated plus maze was cleaned with 75% ethanol between mice.
Marble-burying test
The marble-burying test was used to measure anxiety-like and digging behaviors. The marble-burying test was performed in a cage (43 cm length × 26 cm width × 18 cm height) with 5-cm-thick corncob bedding. Twenty glass marbles were evenly placed on the surface of the bedding in 4 × 5 rows. Mice were put into the cage and allowed to move and dig freely for 30 min. Between mice, bedding was changed and marbles were cleaned with 75% ethanol. At the end of the session, a picture of the marbles was taken and the number of marbles buried (more than two-thirds of the volume buried) was counted.
Digging behavior was defined as a movement consisting of scratching with the forefeet and/or kicking out with the hindfeet. The duration and location of every digging bout were recorded. All digging bouts observed were either classified as “stereotypic” or “nonstereotypic.” A digging bout was defined as stereotypic if the digging movements were repeated for longer than 8 s. This limit was set by referring to the distribution of all bout lengths and separating locally fixed, long-lasting (i.e., stereotypic) digging bouts from the variable, short (i.e., nonstereotypic) digging bouts (Wiedenmayer, 1997).
Statistical analysis
Investigators were blinded to animal genotypes. Between-group comparisons were performed using Welch's t tests. For multiple comparisons, data were analyzed using two-way ANOVAs. Cumulative distributions were analyzed using Kolmogorov–Smirnov tests. For behavior tests, two-way ANOVAs were applied, followed by post hoc comparisons using the Holm–Sidak test, if appropriate. Significance levels were set as *p < 0.05; **p < 0.01; ***p < 0.001. Data shown are means ± SEM.
Results
FAT2 is a Cbln1 binding partner
Cbln1 bridges presynaptic Neurexin and postsynaptic GluD2 to regulate parallel fiber synaptic structure and function (Matsuda et al., 2010; Uemura et al., 2010). FAT2 is copurified with Cbln1 in cultured cerebellar granule cells (Uemura et al., 2010) and is enriched on Purkinje cell surfaces via in vivo proximity labeling (Shuster et al., 2022), suggesting that FAT2 interact with components of the Neurexin/Cbln1/GluD2 complex. Since FAT2 does not directly bind GluD2 in vitro (Uemura et al., 2010), we tested interactions between FAT2 and Neurexin or Cbln1. Using the cell aggregation assay, we detected no direct FAT2-FAT2 or FAT2-Neurexin-1β (SS4+/SS4−) binding (Fig. 1B,C). However, the cell surface binding assay revealed that FAT2 directly interacted with Cbln1 (Fig. 1D–F). To localize the Cbln1-binding region on FAT2, we generated truncation mutants and found that Cbln1 exclusively bound FAT2 constructs containing the first two cadherin repeats (Fig. 1G,H). These findings confirm FAT2 as a Cbln1 binding partner.
FAT2 is a Cbln1 binding partner. A, Experimental strategy of the cell aggregation assay with cultured HEK293 cells. B, C, Representative images and analysis graphs show that FAT2 does not bind to GluD2, Neurexin-1, or FAT2 itself. Scale bar in B, 50 μm. D, Experimental strategy of the cell surface binding assay with cultured HEK293 cells. E, F, Representative images and analysis showing FAT2 binds to Cbln1. Scale bar in E, 10 μm. G, H, Representative images and analysis showing the first two cadherin repeats bind to Cbln1. Scale bar in G, 10 μm. I, K, Representative images and analyses showing the effects of soluble Nrxns on the interaction between FAT2 and Cbln1. The insert in I with Western blotting confirms the production of soluble Nrxn1α protein in the supernatant. Scale bars, 10 μm. All data are shown as means ± SEM, analyzed by Welch's t test, *p < 0.05; **p < 0.01; ***p < 0.001.
Since Cbln1 interacts with both presynaptic Neurexin and postsynaptic GluD2, FAT2 may either mediate Neurexin-Cbln1 binding or compete with Neurexin for Cbln1. To distinguish between these two alternatives, we tested Cbln1-FAT2 binding in the presence or absence of soluble Nrxn1α (SS4+). Soluble neurexin-1 (SS4+) abolished interactions between Cbln1 and full-length FAT2 or its membrane-bound EC1/2 fragment (Fig. 1I–K), whereas it did not affect Cbln1-GluD2 binding (Fig. 1I). These results indicate that Neurexin outcompetes FAT2 for Cbln1 interaction, suggesting weaker affinity between FAT2 and Cbln1 compared with Neurexin-Cbln1 binding. Notably, FAT2 is enriched in presynaptic nonsynaptic vesicles of granule cells (Nakayama et al., 2002), whereas Cbln1 is activity-dependently stored and released from lysosomes (Ibata et al., 2019). This spatial segregation implies that Cbln1 may initially bind FAT2 extrasynaptically before engaging Neurexin at parallel fiber synapses, although in vivo confirmation is needed. Altogether, our results confirm FAT2 as a binding partner of Cbln1 and imply a functional role in cerebellar processes.
FAT2 is dispensable for the gross structure of the cerebellum
To investigate FAT2's role, we first examined its expression profile using RNAscope, a highly sensitive RNA in situ hybridization method. FAT2 mRNA was specifically enriched in the cerebellar granule cell layer, with no detectable expression in Purkinje cells (Fig. 2A,B), aligning with prior studies demonstrating Fat2 gene specificity for cerebellar granule cells (Nakayama et al., 2002). This restricted expression pattern supports the notion of FAT2’s functional specialization in cerebellar granule cells. To assess the physiological roles of Fat2 in the cerebellum, we first generated a mouse line carrying flanked-flox Fat2 alleles (Fat2fl/fl) and then crossed it with a mouse line carrying Cre recombinase under the granule cell-specific promotor Atoh-1 to selectively delete Fat2 in cerebellar granule cells (Atoh-1 Cre; Fat2fl/fl, cKO; Fig. 2C,D). The Fat2fl/fl mice served as controls in our following experiments. The mRNA quantification confirmed the deletion of Fat2 in the cerebellum in Fat2-cKO mice (Fig. 2E).
Fat2 is dispensable for the gross structure of the cerebellum. A, B, Representative RNAscope images of the sagittal section of the whole brain (A) and cerebellar cortex (B) showing Fat2 mRNA specifically located in the cerebellar granule cells but not in the Purkinje cell (white arrow). Scale bar in A, 1 mm. Scale bars in B, 20 μm. C, D, Strategy (C) and genotyping results (D) of Fat2-cKO mice. E, RT-qPCR analysis of Fat2 gene in the cerebellum between P24 and P28 in control and Atoh1Cre; cKO mice. F, Representative Nissl was stained with the cerebellar cortex of cKO mice and their littermate controls. G, Quantification of the number of Purkinje cells, the thickness of the granule cell layer (GCL), and the thickness of the molecular layer (ML) in F. H, Representative images (72 h) and plot of the axon length of parallel fibers in cultured granule cells from cKO mice and littermate controls. P5–P7 of cKO mice and their littermate controls were used for granule cell culture. Scale bar, 5 μm. I, Representative images and plots of the intersection numbers and total length of biocytin-labeled Purkinje cells in cKO mice and littermate controls. Scale bar, 50 μm. All data are shown as means ± SEM, analyzed by Welch's t test (E, G, I) or two-way ANOVA with or without the Holm–Sidak post hoc test (H), *p < 0.05; **p < 0.01; ***p < 0.001.
FAT2’s selective expression in cerebellar granule cells prompted the investigation into its roles in cerebellar function (Thompson et al., 1990). Gross cerebellar morphology in adult Fat2-deficient mice appeared normal, with no changes in Purkinje cell density, granule cell layer thickness, or molecular layer dimensions (Fig. 2F,G), suggesting limited structural roles. Given granule cells' axonogenesis and synaptic connections to Purkinje cells, we assessed axonal growth in cultured P5–7 granule cells. No significant differences in axon growth rates were observed between Fat2-deficient and control neurons (Fig. 2H), indicating the dispensability of FAT2 in axon extension. Furthermore, Purkinje cell morphology remained grossly intact following Fat2 deletion (Fig. 2I), again, supporting limited structural reorganization after Fat2 deletion.
FAT2 maintains the synaptic integrity of parallel fiber synapses
We next assessed FAT2's role in parallel fiber-Purkinje cell (PF-PC) synapses. Cryosections from P21–28 littermate control and Fat2-cKO mice were immunolabeled with antibodies against vGluT1, vGluT2, and vGAT to mark axonal terminals of parallel fibers, climbing fibers, and inhibitory synapses, respectively. In cKO mice, vGluT1 intensity was reduced (Fig. 3A,B), whereas vGluT2 density and puncta size increased (Fig. 3C–E), and vGAT density and puncta size remained unchanged (Fig. 3F–H). The altered vGluT1 and vGluT2 density after Fat2 deletion does not necessarily represent the change of synapse number, likely as vesicle pool alteration. Notably, prior attempts to localize FAT2 via a knock-in 3× HA/miniSOG tag system failed due to disrupted protein expression (Extended Data Fig. 1). Given FAT2's selective expression in cerebellar granule cells, we tested whether it might act as a postsynaptic regulator by examining the expression level of two glutamate receptors GluA2 and GluA4 levels in the granule cell layer (Yamazaki et al., 2010). However, receptor densities were comparable between Fat2-cKO and control mice (Fig. 3I,J). Collectively, those data indicate that FAT2 is critical for maintaining presynaptic terminal integrity.
Fat2 deletion reduces vGluT1 in the cerebellar cortex. A, B, Representative images and analysis showing vGluT1 labeling in the cerebellar cortex of cKO mice and littermate controls. Scale bar in A, 20 μm. C–E, Representative images and analysis showing vGluT2 labeling in the cerebellar cortex of cKO mice and littermate controls. Scale bar in C, 20 μm. F–H, Representative images and analysis showing vGAT labeling in the cerebellar cortex of cKO mice and littermate controls. Scale bar in F, 20 μm. I, J, Representative images and analyses showing GluA2 and GluA4 labeling in the granule cell layer of cerebellar cortex from cKO mice and their littermate controls. Scale bar in I, 50 μm. All data are shown as means ± SEM, statistical analysis was performed by Welch's t test, *p < 0.05; **p < 0.01; ***p < 0.001.
The altered presynaptic properties in Fat2-cKO mice prompted the investigation of synaptic transmission. Using whole-cell patch-clamp recordings in acute cerebellar slices from P21–28 littermates, we assessed excitatory and inhibitory synaptic transmission onto Purkinje cells. Evoked excitatory postsynaptic currents (EPSCs) at parallel fiber-Purkinje cell (PF-PC) synapses exhibited reduced amplitude and slope in Fat2-cKO mice (Fig. 4A–C), with no change in presynaptic release probability, as indicated by paired-pulse ratios (Fig. 4D), suggesting functional impairment of parallel fiber synapses. The ∼30% reduction in PF-PC synaptic transmission correlated with the ∼25% decrease in vGluT1 density, suggesting FAT2 regulates activity-dependent vesicle recruitment at parallel fiber synapses. Climbing fiber-Purkinje cell EPSCs, however, showed no deficits in amplitude or release probability (Fig. 4E,F). Consistent with these findings, miniature EPSC (mEPSC) frequency trended lower in Fat2-cKO mice, whereas amplitude and kinetics remained unchanged (Fig. 4G–J). The mismatch between increased vGluT2 immunoreactivity and stable CF-PC synaptic currents likely reflects methodological sensitivity: electrophysiology captures dynamic vesicle release, while immunohistochemistry measures static transporter density.
Fat2 deletion reduces parallel fiber synaptic transmission. A–C, Representative traces and summary graphs for the input–output curve of parallel fibers evoked EPSCs from Purkinje cells of cKO mice and littermate controls. Calibration in A, 50 pA, 10 ms. D, Summary graph of the paired-pulse ratio of parallel fibers evoked EPSCs recorded from Purkinje cells. The interval of pulses is 50 ms. E, F, Representative traces and summary graphs of climbing fiber-EPSCs recorded from Purkinje cells. The climbing fibers were activated by paired stimulation with an interval of 50 ms. Calibration in E, 100 pA, 20 ms. G–J, Sample traces, frequency, amplitude, and kinetics of mEPSC recorded from Purkinje cells. Calibration in G, 10 pA, 1 s. K–N, Sample traces, frequency, amplitude, and kinetics of mIPSC recorded from Purkinje cells. Calibration in K, 40 pA, 1 s. O, P, Representative traces and summary graphs of basket cell-PC IPSCs recorded from Purkinje cells. The basket cell synapses were activated by paired stimulation with an interval of 50 ms. Calibration in O, 200 pA, 20 ms. All data are shown as means ± SEM; statistical analysis was performed by Welch's t test (bar diagrams) or Kolmogorov–Smirnov test (cumulative distributions), *p < 0.05; **p < 0.01.
We next evaluated inhibitory synaptic transmission in Purkinje cells. Strikingly, Fat2-cKO mice exhibited enhanced frequency and amplitude of mIPSCs (Fig. 4K–N). We positioned the recording electrode near Purkinje cells to activate Basket cells but minimize stellate cells to obtain all-or-none Basket cell-PC IPSCs for synapse identification—we observed similarly augmented basket cell-Purkinje cell synapses-mediated IPSC amplitudes in cKOs, with no change in presynaptic release probability (Fig. 4O,P). Notably, Cbln1 deficiency also increases inhibitory synapse number/function and climbing fiber synaptic input onto Purkinje cells (Hirai et al., 2005; Ito-Ishida et al., 2014). Given that Fat2 mRNA is selectively expressed in cerebellar granule cells (Nakayama et al., 2002a) and FAT2 interacts with Cbln1 (Fig. 1), these findings suggest the enhanced inhibitory signals likely reflect an indirect consequence of Cbln1-mediated mechanisms.
FAT2 is required for fine motor coordination
Given FAT2’s selective expression in cerebellar granule cells (Fig. 2) and its essential roles in synaptic integrity (Figs. 3 and 4), we next investigated the behavioral consequences of Fat2 deficiency. Our behavioral assessments focused on motor function, including grip strength, gait analysis, balance beam traversal, accelerating rotarod, and single-pellet reaching tests (Fig. 5A–J). Fat2-cKO mice exhibited prolonged traversal times on a 10 mm cylindrical beam (Fig. 5C) and increased foot slips on a 20 mm rectangular beam (Fig. 5D) while maintaining intact grip strength (Fig. 5A) and baseline motor coordination (Fig. 5B). These findings suggest that Fat2 deletion selectively impairs motor coordination under challenging conditions rather than reflecting musculoskeletal deficits.
Fat2 is required for fine motor coordination. A, Forelimbs and all limbs grip strength analysis for control and FAT2 cKO mice. B, Quantitative gait analysis. C, Quantitative analysis of the balanced beam test using different beams in cKO mice and littermate controls. D, E, Quantitative behavioral analysis of the accelerating rotarod test in cKO mice and littermate controls. F, Quantitative analysis of the marble-burying test in cKO mice and littermate controls. G, The success rate for single-pellet reaching task of cKO mice and littermate controls in training stages. H, Analysis of success, miss, and drop rate from cKO mice and littermate controls on Day 7 of training stages. I, J, The average trajectory of >6 successful reaches, from lifting foot off the ground to pellet contact (trajectory distance, average speed, maximum speed, and forefeet lift height). All data are shown as means ± SEM, analyzed by Welch's t test (A–C, F, H, J) or two-way ANOVA with or without the Holm–Sidak post hoc test (D, E), *p < 0.05; **p < 0.01; ***p < 0.001.
Specifically, we evaluated motor performance under progressively challenging conditions. On a standard accelerating rotarod (4–40 rpm), Fat2-cKO mice performed comparably to controls (Fig. 5D), suggesting normal basal coordination and motor learning. However, when challenged with faster acceleration (8–80 rpm), cKOs showed blunted improvement across trials, with diminished latency gains compared with controls (Fig. 5D). Terminal speed analysis across 12 trials confirmed cKO deficits at higher rotational velocities (Fig. 5E). Impaired motor coordination was further evidenced by reduced marble-burying behavior (Fig. 5F). Interestingly, motor learning and execution during single-pellet retrieval were intact: cKOs achieved comparable success rates with controls by Day 7 (Fig. 5G,H), with no differences in kinematic parameters (travel distance, velocity, or forefoot elevation; Fig. 5I,J), suggesting that FAT2 is specifically required for context-dependent motor tasks.
We conducted additional behavioral assessments to evaluate cognitive function in Fat2-cKO mice. Tests including open field (Fig. 6A,B), elevated plus maze (Fig. 6C), Y-maze (Fig. 6D), self-grooming (Fig. 6E), sucrose preference (Fig. 6F), three-chamber sociability (Fig. 6G), and novel object recognition (Fig. 6H) revealed no significant alterations in motility, basal coordination, exploratory activity, spatial memory, anxiety-like behavior, stress-induced anhedonia, sociability, or short-term object recognition memory. Notably, the deficits in marble-burying and rotarod performance observed earlier were not driven by increased levels of anxiety or reduced motivation, as Fat2-cKO mice exhibited normal performance in the elevated plus maze and sucrose preference tests (Fig. 6C,E,F). These data collectively suggest that FAT2 deficiency disrupts context-dependent motor tasks without affecting broader behavioral or emotional processes.
Fat2 is dispensable for social behaviors and cognitive functions. A–H, Behavioral assessment included the open field test (under normal light conditions; A), the open field test (under infrared conditions; B), the elevated plus maze test (C), the Y-maze (D), the self-grooming test (E), the sucrose preference test (F), the three-chamber sociability paradigm (G), and the novel object recognition test (H). All data are shown as means ± SEM and analyzed by Welch's t test, *p < 0.05; **p < 0.01; ***p < 0.001.
FAT2 is essential for digging behaviors
After observing deficits in fine motor coordination and marble-burying behavior, we investigated whether these impairments might manifest in a more physiologically relevant context. Digging behavior, a natural and critical survival activity for mice, was systematically analyzed during the 30 min marble-burying test. Initially, mice explored the arena before initiating digging, which we categorized into two main types: type-1 (burrow-down and forefeet digging) and type-2 (head-assisted, two-sided, or single-side digging; Fig. 7A,B). The deletion of Fat2 did not affect the proportion of time spent engaging in those two digging types (Fig. 7C). Moreover, the motivation to dig appeared to be intact given that the latency to dig was unchanged in cKO mice (Fig. 7D). An in-depth analysis of digging behavior, however, revealed that Fat2 deletion reduced the frequency of digging bouts, particularly type-1 digging (Fig. 7E–G).
Fat2 is required for digging behaviors. A–C, Digging behavior subtypes (A), representative patterns (B), and fraction of digging types (C) of digging behavior displayed by adult cKO mice and their littermate controls. D–G, Quantitative analysis of type-1 and type-2 digging behaviors from A. H–I, Quantitative analysis of stereotypic and nonstereotypical digging behaviors (E). J, Fraction of stereotypic and nonstereotypical digging behaviors. K, The total duration of digging behavior. L, Working model of FAT2 in cerebellar parallel fiber-Purkinje cell synapses. All data are shown as means ± SEM, analyzed by Welch's t test (D–I, L) or two-way ANOVA with or without the Holm–Sidak post hoc test (K) are shown in the bar graphs or plots; *p < 0.05; **p < 0.01; ***p < 0.001.
Digging behavior was further classified as stereotypic (>8 s per bout) or nonstereotypic (<8 s per bout) based on duration (Wiedenmayer, 1997). Both categories were impaired in Fat2-cKO mice, with type-1 stereotypic digging disproportionately affected (Fig. 7H,I). Total digging time was also reduced in cKO mice (Fig. 7J,K). These findings were specific to fine motor coordination, as gait analysis under resting conditions revealed normal basal motor coordination (Fig. 5B), and unaltered grooming time suggested that reduced digging was not linked to repetitive behavior changes (Fig. 6E). These data collectively suggest that FAT2 deficiency impair digging behaviors.
Discussion
In this study, we identified FAT2 as a binding partner of Cbln1 via its first two cadherin repeats. We then revealed that Fat2 was required for the maintenance of the synaptic integrity of the Purkinje cells. Finally, we found that the deletion of FAT2 in cerebellar granule cells impaired fine motor coordination and digging behaviors. Therefore, our results suggest that Fat2 is required for synaptic integrity and motor behaviors, likely through a Cbln1-dependent mechanism in the cerebellum.
FAT cadherins possess large extracellular domains and intracellular regions containing multiple protein interaction sites that mediate diverse cellular functions. While the roles of FAT cadherin intracellular domains are relatively well characterized—such as regulating axon terminal organization in Drosophila olfactory neurons (Vien et al., 2024) and controlling neuronal morphology, synapse formation, and light responses in mouse retinas (Deans et al., 2011; Avilés et al., 2022, 2025)—the ligands and extracellular functions of these domains remain poorly understood. FAT2 exhibits homophilic binding in vitro (Nakayama et al., 2002; Nibbeling et al., 2017), but Fat2 mRNA is absent in Purkinje cells (Fig. 2A,B), making homophilic interactions between FAT2 in parallel fiber terminals and Purkinje cells unlikely. Notably, FAT2 has been shown to transcellularly interact with an unknown protein to stabilize the WAVE complex in Drosophila follicular epithelia (Williams et al., 2022). Here, we demonstrate that FAT2 directly binds Cbln1 via its extracellular cadherin domain (Fig. 1), with truncation experiments pinpointing the first two cadherin repeats as critical for this interaction (Fig. 1G,H). Further structural studies remain essential to elucidate the precise mechanism underlying this specificity.
Synaptic adhesion molecules, such as neuroligins and Cblns, are essential for synapse formation and function (Hirai et al., 2005; Matsuda et al., 2010; Zhang et al., 2015, 2017, 2018; Zhang and Südhof, 2016; Südhof, 2017; Yuzaki, 2018; Qin et al., 2020, 2025; Han et al., 2022). Cbln1 is highly expressed in cerebellar granule cells (Miura et al., 2006) and binds presynaptic neurexin and postsynaptic GluD2 (Matsuda et al., 2010; Uemura et al., 2010). Similar to Cbln1-KO mice (Hirai et al., 2005), Fat2 cKO mice exhibit reduced vGluT1 staining (Fig. 3), a parameter usually used to quantify the parallel fiber-Purkinje cell (PF-PC) synapse numbers. FAT2 directly binds Cbln1 in vitro (Fig. 1), and Cbln1 deficiency also increases inhibitory synaptic strength and climbing fiber synapse density (Hirai et al., 2005; Ito-Ishida et al., 2014). Given that Fat2 and Cbln1 mRNAs are both enriched in cerebellar granule cells, and their proteins likely localize to presynaptic terminals (Nakayama et al., 2002; Ibata et al., 2019), we propose that FAT2 and Cbln1 coexist in these terminals and might bind with each other before neurexin outcompete FAT2 for binding to Cbln1 within the synaptic cleft at parallel fiber synapses (Fig. 7L). We propose that elevated vGluT2 might drive postsynaptic compensatory mechanisms (e.g., receptor scaling), balancing functional output. While Cbln1 is also expressed outside the cerebellum and regulates other synapses (Kusnoor et al., 2010; Otsuka et al., 2016; Krishnan et al., 2017), it is unknown whether FAT cadherins have any function outside of the cerebellum (Yuzaki, 2018).
In this study, we found that Fat2 deletion in cerebellar granule cells impaired motor behaviors, particularly digging behaviors. Digging behavior, like other motor actions, involves initiation, execution, and termination stages, potentially governed by distinct brain regions (Lammers et al., 1987; Thompson et al., 1990; Deacon et al., 2003; Deacon and Rawlins, 2005; Puighermanal et al., 2020). Our results indicate that the cerebellum is critical for spontaneous digging behavior (Fig. 7), as evidenced by reduced digging in cKO mice. However, the latency to initiate digging did not differ between cKO and control mice (Fig. 7D), suggesting the cerebellum is less involved in initiating digging, which may depend on other brain regions. Spontaneous digging can be categorized into stereotyped (short-duration) and nonstereotyped (longer-duration) subtypes (Fig. 7). While learned digging relies on broader learning systems (Thompson et al., 1989), the neural circuits underlying the fine motor coordination deficits in cKO mice remain unclear. The altered synaptic integrity from parallel fiber, climbing fiber, and inhibitory synapses observed in cKO mice likely contributes to deficits in fine motor coordination and natural digging behavior, similar to what was seen in mGluR4-KO mice (Pekhletski et al., 1996). In addition, further studies are needed to determine whether impaired digging behavior directly stems from these motor coordination deficits or involves other mechanisms.
Motor abnormalities are key diagnostic criteria for many neuropsychiatric disorders, with important implications for their etiology, nosology, pathophysiology, and management (Peralta and Cuesta, 2017). Dissecting the molecular mechanisms of fine motor behaviors and other motor abnormalities can significantly impact the future development of methods to treat motor and neuropsychiatric disorders.
Footnotes
We thank Dr. Haitao Wu for the gift of Atoh1Cre mouse line, Liangyu Liao for the gift of Cbln1 protein, Dr. Junhai Han for the gift of soluble Neurexin-1α plasmid, Dr. Shuwen Zhang for the gift of GluA2 antibody, Drs. Fujun Luo and Zhihui Liu for comments on an earlier version of the manuscript, and Bioimaging Core of Shenzhen Bay Laboratory for assistance with confocal microscopy. This work was supported by the following funding: National Natural Science Foundation of China (82022018, 32070958, 82161138025, to B.Z.; 82304385, to L.G.); Shenzhen-Hong Kong Institute of Brain Science-Shenzhen Fundamental Research Institutions (2024SHIBS0004, to B.Z.); Guangdong Pearl River Funding (to B.Z.); Major Program of Shenzhen Bay Laboratory (S241101002, to B.Z.); China Postdoctoral Science Foundation (2021M700223, to J.M.); Fonds Wetenschappelijk Onderzoek–Vlaanderen (FWO; Research Foundation – Flanders; G0E6722N, to M.W.).
↵*X.W., Y.P., and J.M. contributed equally to this work.
The authors declare no competing financial interests.
This paper contains supplemental material available at: https://doi.org/10.1523/JNEUROSCI.2345-24.2025
- Correspondence should be addressed to Bo Zhang at zbo{at}pku.edu.cn or zbo{at}szbl.ac.cn.