GRID1/GluD1 homozygous variants linked to intellectual disability and spastic paraplegia impair mGlu1/5 receptor signaling and excitatory synapses

The ionotropic glutamate delta receptor GluD1, encoded by the GRID1 gene, is involved in synapse formation, function, and plasticity. GluD1 does not bind glutamate, but instead cerebellin and D-serine, which allow the formation of trans-synaptic bridges, and trigger transmembrane signaling. Despite wide expression in the nervous system, pathogenic GRID1 variants have not been characterized in humans so far. We report homozygous missense GRID1 variants in five individuals from two unrelated consanguineous families presenting with intellectual disability and spastic paraplegia, without (p.Thr752Met) or with (p.Arg161His) diagnosis of glaucoma, a threefold phenotypic association whose genetic bases had not been elucidated previously. Molecular modeling and electrophysiological recordings indicated that Arg161His and Thr752Met mutations alter the hinge between GluD1 cerebellin and D-serine binding domains and the function of this latter domain, respectively. Expression, trafficking, physical interaction with metabotropic glutamate receptor mGlu1, and cerebellin binding of GluD1 mutants were not conspicuously altered. Conversely, upon expression in neurons of dissociated or organotypic slice cultures, we found that both GluD1 mutants hampered metabotropic glutamate receptor mGlu1/5 signaling via Ca2+ and the ERK pathway and impaired dendrite morphology and excitatory synapse density. These results show that the clinical phenotypes are distinct entities segregating in the families as an autosomal recessive trait, and caused by pathophysiological effects of GluD1 mutants involving metabotropic glutamate receptor signaling and neuronal connectivity. Our findings unravel the importance of GluD1 receptor signaling in sensory, cognitive and motor functions of the human nervous system.


Genome wide-linkage analysis and whole exome sequencing
Genomic DNA samples were extracted from peripheral blood following standard protocols.Genotyping of Family A (three affected children, one healthy child and both consanguineous parents) was performed on Genechip® human 250K NspI array (Affymetrix) according to manufacturer's instructions.Briefly 250 ng of genomic DNA were restricted with NspI.NspI adaptators were then ligated to restricted fragments followed by PCR using universal primer PCR002.PCR fragments were purified and 90 µg were used for fragmentation and end-labelling with biotin using Terminal Transferase.Labelled targets were then hybridized overnight to Genechip® human 250K NspI array (Affymetrix) at 49°C.Chips were washed on the fluidic station FS450 following specific protocols (Affymetrix) and scanned using the GCS3000 7G.The image was then analyzed with GCOS software to obtain raw data (CEL files).Genotypes were called by the Affymetrix GType software using Dynamic Model (DM) and Bayesian Robust Linear Model with Mahalanobis (BRLMM) mapping algorithms.Homozygosity regions were obtained using MERLIN software assuming a recessive model with complete penetrance (disease allele frequency of 0.0001).
WES study was performed using Agilent SureSelect Human All Exon kit (V2; Agilent technologies).Genomic DNA was captured with biotinylated oligonucleotides probes library (Agilent technologies), followed by paired-end 75 bases massive parallel sequencing on Illumina HiSEQ 2000.Image analysis and base calling were performed using the Illumina Real-Time Analysis Pipeline version 1.14 with default parameters.Sequencing data was analyzed according to the Illumina pipeline (CASAVA1.7)and aligned with the Human reference genome (hg19) using the ELANDv2 algorithm.Genetic variation annotation was performed with the IntegraGen in-house pipeline (IntegraGen).Filtering was performed using Eris software (IntegraGen) with an autosomal recessive hypothesis.Variants with minor allele frequency (MAF) >1% in either the 1000 Genomes Project, the EXAC, or the gnomAD databases were excluded.Genetic segregation of the candidate variant with the disease in Family A was confirmed by Sanger sequencing of GRID1 exon 3.
For Family B, DNA sample of the proband was shipped to Otogenetics, USA (CLIA lab).~50 Mb of genomic DNA were captured on HiSeq 2500.Fragments were read 100-125 bp, paired end.The sample was uploaded onto DNAnexus software and 71.5 million reads were aligned to the reference human genome (Hg19) (Mean on target coverage, X118).Variants which were low covered, off target (>6bp from splice site), synonymous, heterozygous, predicted as benign, MAF>0.5% on ExAC and MAF>4% in the Hadassah in-house dbSNP were removed.Thirty-one homozygous variants survived this filtering.

Molecular modeling of GluD1 mutants structure
The protein was generated using Rat GluD1 receptor in complex with 7-chloro-kynurenate and calcium ions (PDB codes: 6KSS and 6KSP) as structure templates (Burada et al., 2020).The system with proteins and ligands was prepared in the CHARMM-GUI web server (Jo et al., 2008) in order to generate a membrane around the protein and solvate with water and ions.A heterogeneous membrane made of POPC was chosen, and a TIP3 water model with NaCl (0.15 M) counter ions was chosen for the solvation.The system was typed with a CHARMM36m force field, and NAMD protocol was used.The system was equilibrated through six constrained simulations for a total of 500 ps by gradually diminishing the force constraints at each steps.The following constraints were applied (each value represents an equilibration step): protein backbone (5/2.5/1/0.5/0.1 kcal/mol), protein side chains (5/2.5/1.25/0.5/0.25/0.05kcal/mol), lipid heads (5/5/2/1/0.2/0kcal/mol), and dihedral bonds (500/200/100/100/50/0 kcal/mol).Then, a production dynamic of 10 ns was carried out in NPT conditions at 303.15 K without any constraints.
Mutant models were generated using Built Mutant protocol from Discovery Studio 2019.A set of 100 structures was created and ranked for their Dope score.The best model was then minimized using Adopted Basis Newton-Raphson algorithm (a Newton-Raphson algorithm applied to a subspace of the coordinate vector spanned by the displacement coordinates of the last positions) until a RMS gradient of 0.001 was obtained.
Molecular docking experiments of D-Serine, glycine and kynurenic acid at the active site were performed as described (Ducassou et al., 2015, Dhers et al., 2017), using default parameters from CDocker (Wu et al., 2003) with Discovery Studio 2020 and a sphere radius of 10 Å in rigid mode.Flex Dock (Discovery Studio) was used for ligand-protein flexible docking.Dockings were performed on minimized structure of receptor before dynamics experiments.
Mouse GluD1ATD-LBD-Fc was constructed by fusing the GluD1 amino-terminal domain and ligandbinding domain (ATD-LBD; D21-T814; with a GT linker between K547 and P664 to remove the M1-M2-M3 transmembrane helices) with the fragment crystallizable (Fc) region of human IgG1.Mutations R 161 H or T 752 M were introduced onto this scaffold.WT, R 161 H or T 752 M GluD1ATD-LBD-Fc proteins were transiently expressed in HEK293T cells, and conditioned expression media containing the secreted proteins were collected 48-60h after transfection.
The direct interaction between Cbln1FL and GluD1ATD-LBD-Fc (WT, R 161 H and T 752 M) was measured using a BLI Octet® R8 (Sartorius).Octet® AHC biosensors (Sartorius) were coated by dipping them into conditioned expression media containing GluD1ATD-LBD-Fc (WT, R 161 H and T 752 M) at 25 °C to a final optical thickness of 1.1 nm.Assays were performed at 25 °C in a volume of 200 μL in 20 mM HEPES pH 7.4, 150 mM NaCl, 3 mM CaCl2, 0.1% BSA and 0.005% TWEEN 20.A duplicate set of sensors without protein was used as a background binding control.Association of Cbln1FL (5000, 2500, 1250, 625, 312.5, 156.25, and78.125 nM) to coated and uncoated reference sensors was measured over 300 s and dissociation over 1600 s after switching to Cbln1FL-free buffer.Data analysis on the Octet® R8 instrument was performed using a double reference subtraction (sample and sensor references) in the Octet® Analysis studio software (version 13), which accounts for nonspecific binding, background, and signal drift and minimizes well-based and sensor variability.KD values were fitted locally and separately using a 2:1 heterogeneous binding model to account for avidity effects stemming from the multivalent nature of Cbln1FL (dimer of trimers) and GluD1ATD-LBD-Fc (dimer).Finally, processed raw data and fitted curves were exported from Octet® Analysis studio software and plotted in GraphPad Prism (Version 10.0.2).The Ka1/Kdis1 and Ka2/Kdis2 association/dissociation values for the individual concentrations were plotted onto iso-affinity graphs for easier visualization.

Electrophysiology on Xenopus laevis oocytes
To obtain oocytes, parts of the ovaries were surgically removed from Xenopus laevis (Xenopus 1, Dexter, MI, USA) anaesthetised with 3-aminobenzoic acid ethylester (1.5 g/L, Sigma, Taufkirchen, Germany).To remove the follicular cell layer, the ovary clippings were digested with collagenase type I (4 mg/mL, Worthington, Lakewood, NJ, USA) in Ca 2+ -free Barth's solution (88 mM NaCl, 1.1 mM KCl, 2.4 mM NaHCO3, 0.8 mM MgSO4, 15 mM HEPES-NaOH, pH 7.6) for 1.5-2 h at 20 °C and then washed with Barth's solution (88 mM NaCl, 1.1 mM KCl, 2.4 mM NaHCO3, 0.3 mM Ca(NO3)2, 0.4 mM CaCl2, 0.8 mM MgSO4, 15 mM HEPES-NaOH, pH 7.6) to stop the digestion.Complementary RNA (cRNA) for injection into oocytes was synthesized from 1 μg linearized plasmid DNA with the mMESSAGE mMACHINE T7 in vitro transcription kit (Ambion, Austin, TX, USA).Defolliculated oocytes of stages V and VI were manually selected, maintained at 16 °C in Barth's solution supplemented with gentamicin (100 µg/mL), streptomycin (40 µg/mL), and penicillin (63 µg/mL), and injected with 10 fmol (11 ng) of cRNA using a nanoliter injector (WPI, Sarasota, FL, USA).Four to five days after cRNA injection, current responses were recorded under voltage clamp at -70 mV with a Turbo Tec-10CX amplifier (npi electronic, Tamm, Germany) controlled by Patchmaster software (HEKA, Lambrecht, Germany).Currents were filtered with a 20 Hz low-pass filter and then digitised with a sampling rate of 50 Hz.Recording electrodes with resistances of 0.5-1.5 MΩ were pulled from borosilicate glass (Science Products GmbH, Hofheim, Germany) with an L/M-3P-A vertical pipette puller (List-Medical) and filled with 3 M KCl.Recordings were performed in a 50 µL chamber under constant superfusion at a flow rate of 3-5 mL/min.To assess the background current without any Na + flux, the oocyte was superfused with a Na + -free solution containing the impermeable cation N-methyl-D-glucamine (NMDG; 115 mM NMDG-Cl, 2.5 mM KCl, 1.8 mM BaCl2, 10 mM HEPES, pH 7.2) until a stable baseline was reached.The recording was started, and after 10 s, Na + -containing Ba 2+ Ringer's solution (115 mM NaCl, 2.5 mM KCl, 1.8 mM BaCl2, 10 mM HEPES-NaOH, pH 7.2) was applied for 30 s to determine the spontaneous Na + current through permanently open channels.Then, the substance to be tested for its effect on the spontaneous current (3 mM D-Ser, 3 mM Gly or 100 µM pentamidine in Ba 2+ Ringer's solution) was applied for 20 s, then washed out for 20 s with Ba 2+ Ringer's solution and finally for 40 s with Na + -free NMDG solution to check for reversibility.

HEK293T cell culture and transfection
HEK293T cells (ATCC Number: CRL-3216, authenticated using Short Tandem Repeat analysis by the ATCC cell authentication service, mycoplasma-free) were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum, 100 U/ml penicillin and 100 µg/ml streptomycin (Life Technologies).For immunostaining or cerebellin binding experiments, cells were seeded at 8.10 5 cells per well on glass coverslips coated with poly D-lysine (Sigma Aldrich P7280) and cultured in 12-well plates.For membrane protein isolation and immunoprecipitation experiments, cells were seeded in 10 cm dishes coated with poly D-lysine at a density of 2.10 6 cells/dish.Transient plasmid transfection was performed the next day using the calcium phosphate precipitation method (6 µg plasmid per 12 well-plate or per dish) or using Lipofectamine 2000 (Invitrogen, 2,5 µg plasmid and 6µL reagent per 6 well-plate).Plasmids encoding mGlu1-YFP and GluD1 were mixed at a ratio 1:1 for cotransfection.Culture medium was renewed 6 h after transfection, and cells cultured overnight.

Immunostaining on HEK cells
Transfected HEK cells were fixed with 4% paraformaldehyde in 0.1 M sodium phosphate buffer (PB) during 20 min, and then washed with Dulbecco's phosphate-buffered saline (D-PBS).All the procedure was performed at room temperature.After fixation, cells were incubated in PBS containing fish skin gelatin (2g/l) and Triton X100 0,25% (PBS-GT) for 1 hour.Triton X100 was omitted from incubation medium (PBS-G) when cells were not permeabilized.Next, cells were incubated for 2 to 4 hours with primary antibodies (see Suppl.Table for antibodies) diluted in PBS-GT/PBS-G, washed 3 times 15 minutes with PBS, and incubated with secondary antibodies (Suppl.Table ) and DAPI nuclear stain (300 nM, Invitrogen) diluted in PBS-GT for 2 hours.After PBS washes, samples were mounted on glass slides using Fluoromount-G (Biovalley 0100-01), and images were acquired using an epifluorescence microscope (DMR, Leica), or a confocal microscope (SP5, Leica).

Isolation of membrane proteins from HEK cells and western blotting
Total membrane proteins were extracted from HEK cells expressing HA-GluD1 WT , HA-GluD1 R161H or HA-GluD1 T752M using the MEM-Per TM Plus Membrane Protein extraction kit (Thermoscientific) according to manufacturer's protocol.Proteins lysates were separated on 4-20% Mini-PROTEAN® TGX Stain-Free Precast electrophoresis Gels (Bio-Rad) and transferred using Trans Blot Turbo system (Bio-Rad) on nitrocellulose membranes (Bio-Rad).Membranes were then incubated in blocking buffer with 5% milk in a mixture of Tris-buffered saline and Tween 0,002% (TBST, Fisher) for 1 hour at room temperature.Next, membranes were incubated with rat anti-HA antibody (Suppl.Table ) overnight at 4°C in 5% milk diluted in TBST.After three washes of 10 min, membranes were incubated in 5% milk-TBST with secondary anti-rabbit and anti-beta-actin antibodies conjugated with horseradish peroxydase (HRP, Suppl.Table ) for 45 min.HRP was revealed through chemiluminescence using Clarity TM Western ECL substrate (Bio-rad), visualized on a ChemiDoc TM Touch imaging system (Bio-rad), and quantified using the ImageJ software (U.S. National Institutes of Health, Bethesda, MD, USA; http://rsbweb.nih.gov/ij/).

Immunoprecipitation from HEK cells
HEK cells co-expressing HA-mGlu1a-Venus and GluD1, GluD1 R161H , or GluD1 T752M were washed twice with ice cold PBS and lysed in 500 µl lysis buffer containing 50mM Tris-HCl pH 7.5, 150mM NaCl, 1% Nonidet P40, 0.5% sodium deoxycholate, and protease inhibitor (Complete Ultra Tablets, Roche) according to manufacturer's instructions.The whole immunoprecipitation procedure was carried out at 4 °C.Lysates were centrifuged 13000g for 15 min and protein concentration was determined in the supernatant by the Bradford's method using BSA as standard.Supernatants were then pre-cleared with Protein A Plus Agarose beads (Pierce).Specific immunoprecipitation were performed overnight by incubating 250 µg proteins of the precleared lysates with specific antibodies or control rabbit anti-mouse antibodies (Suppl.Table ).Protein complexes bound to rabbit anti-GluD1 antibodies were precipitated with Protein A Plus agarose beads for 4 h.Protein complexes bound to mouse anti-HA antibodies were precipitated with beads coupled to rabbit anti-mouse antibodies.Precipitates were washed twice with lysis buffer, twice with 50 mM Tris Hcl pH7.5, 500 mM Nacl, 0.1% Nonidet P40, 0.05% Sodium deoxycholate and once with 50 mM Tris Hcl 0.1% Nonidet P40, 0.05% Sodium deoxycholate.Proteins were eluted from the beads with 30 µl LDS sample buffer (Invitrogen), separated on 4-15 % polyacrylamide gels (Biorad), and transferred onto nitrocellulose membranes.Western blots were carried out using standard protocols and antibodies listed in Suppl.Table .Detection was performed with the Odyssey detection system (LI-COR Bioscience) using secondary anti-IgG antibodies coupled to infrared dyes (Suppl.Table ).Band intensity was determined using ImageJ.

Primary cortical or hippocampal cell cultures
All components for cell cultures were from Thermo Fisher Scientific unless otherwise stated.Primary cortical or hippocampal cell cultures were prepared essentially as described (Ung et al., 2018) from E17-E18 Grid1 -/-or Grid1 +/+ mice embryos, respectively.Cortices and hippocampi were dissected in ice cold PBS containing 100 U/ml penicillin and 100 µg/ml streptomycin, and kept in Hibernate E medium supplemented with 2% B27, while genotyping using the Phire Animal Tissue Direct PCR Kit (Thermo Fisher, for primers see Hepp et al., 2015).Tissues were pooled, dissociated with papain (Worthington).Tissues were then triturated in DMEM-F12 containing 10 % heat inactivated fetal calf serum and cells transferred to a new tube and centrifuged 250 g for 4 minutes.Cells were resuspended in Neurobasal medium supplemented with 2% B27, 0.5 mM glutamax (complete Neurobasal medium), and counted.Hippocampal cells were seeded at a density of 6x10 4 cells/500 µl medium per well on glass coverslips coated with poly D-lysine and laminin (Sigma Aldrich) in 24 well plates.For Ca 2+ imaging, cortical cells were cultured as described for hippocampal cells, whereas for western blot analyses, cortical cells were plated at 10 6 cells per dish on 35 mm culture dishes coated with poly D-lysine and laminin.Cells were grown at 35°C, under 5% CO2 atmosphere.Half of the medium was changed every 3 to 4 days.
For measurements of lentiviral transduction efficiency, cultures were next fixed and processed for DAPI staining and immunolabelling (see Suppl.Table ) as described above for HEK cells.
For test of mGlu1/5 signaling via the ERK pathway, cultures were next rinsed once with warm HBSS-TTX-APV medium containing 2 mM Ca2+, 1 mM Mg2+, 300 nM TTX (Latoxan), and 50 µM of the NMDAR antagonist APV (Hello Bio).Cells were then incubated for 1 hour at 35° in HBSS-TTX-APV.The medium was next removed, and cells were incubated for 5 minutes at 35°C in HBSS-TTX-APV medium, in the presence or absence of the mGlu1/5 agonist RS-3,5-dihydroxyphenylglycine 2 (DHPG, 100 µM, Hello Bio).Cells were then lysed in ice cold 50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1% Nonidet P40, 0.5% sodium deoxycholate, supplemented with protease (Complete Ultra Tablets, Roche) and phosphatase inhibitors (PhoStop, Roche) according to manufacturer's instructions.Protein concentration was determined using the Bradford's method with BSA as standard.Proteins (10 µg/lane) were separated by SDS-PAGE, transferred onto nitrocellulose sheets, and western blots were carried out using standard protocols.Primary antibodies and secondary anti-IgG antibodies coupled to infrared dyes are listed in Suppl.Table .Detection was performed with the LI-COR Odyssey detection system.Band intensity was determined using the ImageJ software.
For live imaging of mGlu1/5-dependent Ca 2+ signaling, 0.5µl of the recombinant Sin-Twitch-2B pseudo-virion solution (10 4 i.p.) was added per well onto cortical neurons cultured on glass coverslips.The next day, glass coverslips were transferred to a recording chamber and neurons were perfused with a solution containing (in mM): 110 NaCl, 5.4 KCl, 1.8 CaCl2, 0.8 MgCl2, 10 HEPES, 10 D-Glucose at a rate of 2 ml/min at room temperature.The mGlu1/5 agonist S-3,5-dihydroxyphenylglycine 2 (S-DHPG, 50 µM, Hello Bio) was applied through the bath perfusion.Two photon image stacks were obtained every 15 seconds with a custom-built 2-photon laser scanning microscope as described (Bonnot et al., 2014) using a 20x (0.5 NA) water immersion objective.Image acquisition was performed using the Matlab routine ScanImage (Pologruto et al., 2003).The ratiometric Ca 2+ sensor Twitch-2B is based on the fluorescent CFP/YFP variants of GFP as donor/acceptor pair (Thestrup et al. 2014).Two-photon excitation was performed at 850 nm for CFP with a tunable femtosecond Ti:sapphire pulsed laser (MaiTai HP; Spectra Physics, Ellicot City, MD, USA).CFP and YFP fluorescence were collected on two independent photomultipliers.Images were analyzed using custom written macros in ImageJ and the emission ratio YFP/CFP calculated for regions of interest (ROI) drawn around the somata of Twitch-2B-expressing neurons.Fluorescence intensity variations in a given ROI was expressed as the ratio ΔR/R0, were ΔR= R-R0.R is the YFP/CFP ratio in a ROI at a given time point, and R0 corresponds to the YFP/CFP ratio in the same ROI during control baseline prior to drug application.

Plasmid transfection of hippocampal cell cultures and analyses of neurites and excitatory synapses
Cultures were transfected at DIV4 for morphometric analyses of dendrites, and at DIV11-DIV13 for analyses of spine morphology and synapse counting.Cells were transfected with plasmid pCMX-GFP alone or in combination with pDEST26-GluD1 WT , pDEST26-GluD1 R161H , or pDEST26-GluD1 T752M (1:1 ratio) using Lipofectamine 2000 (Invitrogen).Plasmids pmaxGFP, pCMV-HA-GluD1 WT and pCMV-HA-GluD1 R161H were used instead of above plasmids for experiments shown in Figures S3 and S8.Lipofectamine (1 µl/well) and plasmids (500 ng/well) were diluted in Neurobasal medium (100 µl/well).Prior to transfection, 300 µl medium was collected from each well and diluted by half with fresh complete Neurobasal medium.This conditioned medium was kept in the incubator for the duration of the transfection.Next, 200 µl complement-free Neurobasal medium and 100 µl of the lipofectamine-DNA solution was added in each well.After 2 h incubation at 35°C, cells were washed twice with complete Neurobasal medium before adding 500 µl conditioned medium.Cells were then incubated for 48 h before fixation.Cells were fixed with 4 % paraformaldehyde for 20-30 min, permeabilized with PBS-GT unless otherwise stated, and processed for immunolabelling and DAPI staining as described above for HEK cells.Immunolabelling of co-transfected cultures using chicken anti GFP, and rabbit anti-GluD1 or rat anti-HA primary antibodies (Suppl.Table ), demonstrated that more than 96 % of GFP-expressing neurons also over-expressed either GluD1 WT , GluD1 R161H , GluD1 T752M , HA-GluD1 WT , HA-GluD1 R161H , or HA-GluD1 T752M .
For morphometric analyses of dendrites, images of isolated GFP-expressing neurons were acquired with an epifluorescence microscope (DMR, Leica).The Sholl analysis was performed upon conversion to binary images using the SNT module of ImageJ/Fiji software (https://imagej.nih.gov/ij/,Schindelin et al., 2012;Schneider et al., 2012;Ferreira et al., 2014).An ROI delimiting the soma was used to define the cell center from which concentric circles of 20 pixels (5 µm) apart were drawn on a radius of 700 pixels.For each cell, the number of neurites crossing along the radius and the total crossings were determined.The total length of neurites per cell was manually determined using the segmented line tool of ImageJ/Fiji.Lines were converted to ROIs to determine the length of all the segment per cells in order to sum them up.Neurites extending beyond the field were not included.
For analyses of dendritic spine morphology, images of isolated GFP-expressing spiny dendrites were acquired by confocal microscopy (TCS SP8-STED, Leica) using a 63X objective with a zoom of 2 and were z-sectioned at 0.3 µm increments.Morphological analysis of the GFP-labelled spines was performed manually according to Zagrebelsky et al. (2005), based on measurements of spine length and of the ratio between neck and head diameters of the spine.We distinguished immature spines comprising both long thin (length: 1<x<3 µm, head/neck diameter<2) and filopodia-shaped (length >3 µm) spines, versus mature spines comprising both mushroom-shaped (length: 1<x<3 µm, head/neck diameter>2) and stubby (length<1 µm) spines.
For synapse counting, cells were labelled using primary antibodies: chicken anti GFP, mouse anti Bassoon, rabbit anti Homer1, and secondary antibodies: goat anti-chicken Alexa 488, goat anti-mouse-RRX, goat anti-rabbit-Alexa 647 (Suppl.Table ).Images were acquired with a DMR Leica epifluorescence microscope.The density of glutamatergic synapses was measured by counting manually Homer1/Bassoon co-labelled spots present on merged images of GFP positive dendrites processed with ImageJ/Fiji.Only spiny neurons exhibiting a pyramidal cell-like morphology with pyramidal-shaped soma and prominent apical dendrite were analysed.
For visualization of recombinant HA-GluD1 WT and dendritic spine counting in CA1 neurons expressing tdTomato, organotypic slices were fixed with 4% paraformaldehyde-4% sucrose in PBS for >4 h before the permeabilization of membranes with 0.25% Triton in PBS.Slices were subsequently incubated with a rat anti-HA antibody followed by a donkey anti-rat Alexa 488 antibody (Suppl.Table ).Images were acquired on a Leica DM6 CFS TCS SP8 microscope using a 63 × /1.4 NA oil objective and a pinhole opened to 1 time the Airy disk.Images with pixel size of 70 nm were acquired at a scanning frequency of 400 Hz.The vertical step size was set at 0.3 µm.The number of spines per unit dendrite length of tdTomato-positive cells was calculated manually using Image J.

Statistical analyses
All experiments were repeated at least three times, and GraphPad prism6 software (Instat) was used for statistical analyses and graphical representations.When d'Agostino-Pearson normality tests were successfully passed, we conducted parametric test using One-way ANOVA.Then, Tukey's post hoc method was used to determine statistical significance in multiple comparisons and to reveal the contribution of the genotype in the variability between each test.For samples that did not pass the normality test, we used Kruskal-Wallis method followed by Dunn's post hoc test.Results are given as mean ± standard error of the mean.Differences were considered significant if p<0.05.

Clinical description of Family B
The proband (designated Patient 4), a 6 years and 3 months old girl, presented for evaluation due to intellectual disability, unique facial features, brachycephaly, recurrent episodes of hypereosinophilia and additional findings.
She is the seventh of seven children born to consanguineous (first and second degree cousin) parents of Arab-Muslim descent.She was reported to have had intrauterine growth retardation (IUGR), and was born at term (38 weeks of gestation), birth weight of 1800 grams.The parents reported of failure to thrive, hypotonia and significantly delayed acquisition of developmental milestones: she turned over at 8 months, sat up at 12 months, and began walking at the age of 2.5 years.She began talking at 5 years of age, and at 6 years was reported to have a vocabulary of ~20 words.Socially, she did not interact with her peers.
Upon physical examination (at 6 years of age), she was alert and vital, in no respiratory distress, and was hyperactive throughout the exam.She showed brachycephaly, flat occiput, low set ears and a unique structure of the upper eyelids resembling bilateral ptosis.Persistent flexion of the neck raised suspicion for an anomaly of the cervical spine.She further had sparse hair, also notable for a reddish tinge in the hair ends.Skin was translucent with no hypo-or hyperpigmented skin lesions.She had pectus excavatum, and her limbs were notable for brachydactyly in both hands with short fifth digits, syndactyly of II-III toes bilaterally, with metatarsal shortening of the fifth toes.Neurological examination showed no nystagmus, symmetrical facial movements, normal sensation, decreased muscle tone and brisk deep tendon reflexes (DTR) in the lower limbs.Of note, she had an abnormal and unstable walk, although not purely atactic.Her height was 106 cm (3 rd centile, Z score -1.83) and head circumference was 45 cm (<1 st centile, -4.7 SD).
Previous clinical investigation included a brain CT in early childhood that was reportedly notable for microcephaly and craniosynostosis, and a brain MRI performed at 3 years of age, and considered to demonstrate mild diffuse cortical atrophy.Due to chronic diarrhea and persistent hypereosinophilia and elevated serum IgE levels initially attributed to a Strongyloides infection, the patient required several hospitalizations and extensive investigations which were noncontributory.These included abdominal ultrasound, and laboratory tests for immunological and infectious etiologies.Of note, the diarrhea responded to steroid treatment, but repeatedly resumed upon its cessation.Echocardiogram was performed and considered to be normal.She did not have a history of recurrent infections.
Family history is notable for an additional affected elder sister (designated Patient 5), 24 years old, who is reportedly similarly but more severely affected.She had developmental delay, and had begun walking at 4 years of age.She shows intellectual disability, facial features resembling those of the proband, and instability when walking, with kyphosis of the cervical spine.She reportedly has spastic paraplegia, and requires a wheelchair when leaving the house.
The other five siblings (males and females) are reportedly healthy, apart from eczema/atopic dermatitis and arthralgia in one male sibling.

The R 161 H and T 752 M mutations do not hamper cerebellin binding to the extracellular aminoterminal domain of GluD1
The kinetics and affinity of cerebellin binding to GluD1 were investigated using Bio-Layer Interferometry (BLI) measurements on recombinant full-length Cerebellin-1 (Cbln1FL) and a fusion of WT or mutant GluD1 amino-terminal and ligand-binding domains with the fragment crystallizable (Fc) region of human IgG1 (WT, R 161 H or T 752 M GluD1ATD-LBD-Fc, see Suppl.Methods).Measurements were performed at seven different Cbln1FL concentrations ranging from 75 to 5000 nM.Consistent with the multimeric nature of cerebellin (Elegheert et al., 2016), association and dissociation kinetics (Ka and Kdis, respectively) were biphasic, and were fitted with a model (see Suppl.Methods) yielding two dissociation constants (KD1 and KD2) for each Cbln1FL concentration (Figure S6).The values of these dissociation constants did not reveal differences in binding affinity between WT, R 161 H and T 752 M GluD1ATD-LBD-Fc (Figure S6) that may hamper cerebellin binding to GluD1 mutants.
We thus examined the impact of the GluD1 R 161 H mutation on dendritic spine density and morphology in mature hippocampal primary neuronal cultures from Grid1 +/+ mice using co-transfection of plasmids encoding HA-GluD1 WT or HA-GluD1 R161H together with a GFP-expressing plasmid.We found a significant increase in the density of dendritic spines in neurons overexpressing GluD1 WT as compared to GFP-only, control neurons (spine number per 10 µm dendritic segment; control: 5.3 ± 0.2, n=6 neurons; GluD1 WT : 7.0 ± 0.3, n=9 neurons), consistent with the reported spine-promoting function of GluD1 (Gupta et al., 2015).Conversely, neurons overexpressing GluD1 R161H exhibited a spine density (5.2 ± 0.2 per 10 µm segment, n=8 neurons) similar to that of control neurons (Figure S8).We also observed, in the same dendritic sections, that the proportion of immature spines (see Suppl.Methods and Zagrebelsky et al., 2005) was significantly enhanced in GluD1 R161H -transfected neurons as compared to control neurons (control: 20.1 ± 2.3, GluD1 WT : 23.4 ± 1.6, GluD1 R161H : 28.3 ± 1.4 %; GluD1 R161H >control, p<0.05; Figure S8).These results indicate that the R 161 H mutation impairs GluD1 stimulatory effects on dendritic spine formation and maturation.