Human-mouse chimeric Fragile X syndrome model reveals FMR1 dependent neuronal phenotypes

Abnormal neuronal development in Fragile X syndrome (FXS) is poorly understood. Data on FXS patients remain scarce and FXS animal models have failed to yield successful therapies. In vitro models do not fully recapitulate the morphology and function of human neurons. Here, we co-injected neural precursor cells (NPCs) from FXS patient-derived and corrected isogenic control induced pluripotent stem cells into the brain of neonatal immune-deprived mice. The cells populated the brain and differentiated into neurons and astrocytes. Single-cell RNA sequencing of transplanted cells revealed upregulated excitatory synaptic transmission and neuronal differentiation pathways in FXS neurons. Immunofluorescence analyses showed accelerated maturation of FXS neurons, an increased proportion of Arc-positive FXS neurons and increased dendritic protrusion width of FXS striatal medium spiny neurons. Our data show faster maturation and suggest increased synaptic activity and synaptic strength of FXS transplanted neurons. This model provides new insights into the alterations in FXS neuronal development. Here, we established a novel in vivo model of FXS using co-transplantation of hNPCs differentiated from FXS patient-derived iPSC and isogenic control iPSC in the brain of immune-deprived mouse neonates. We assessed the cell-autonomous effects of FMR1 silencing on the development of human neurons in an in vivo context. Gene ontology analysis of single cell RNA sequencing data from transplanted FXS and isogenic control neurons showed upregulation of pathways linked to neurogenesis, neuronal differentiation and synaptic activity in FXS neurons. Transplanted FXS neurons matured faster than isogenic control, and showed a persistent increase in Arc expression. Additionally, FXS striatal medium spiny neurons (MSNs) had wider dendritic synaptic protrusions at 6-7 months post-injection (PI). Our data suggest that FXS neurons transplanted in the mouse brain mature faster and are synaptically hyperactive compared to isogenic control. activity in FXS neurons. Together, these results suggest increased synaptic activity, but not synaptogenesis, in FXS neurons compared to control. our and and FXS direct FXS in these two studies both neurons and astrocytes were mutant for FMR1. In line with this hypothesis, adult astrocyte-specific FMR1 KO mice displayed increased spine density and a higher proportion of thin spines in the motor cortex, indicating that FXS astrocytes are necessary and sufficient for the dendritic spine phenotype of neurons in FXS 26 Our data on Arc expression and dendritic protrusion diameter suggest that FXS neurons are hyperactive compared to control. This is in line with previous work indicating that the excitation/inhibition balance is altered in FXS patients, and suggests that cell-autonomous effects are involved in the hyperexcitability of the neuronal networks. These results are in line with previous studies from our laboratory and others 11,12 that revealed hyperexcitability in cultured FXS neurons induced by overexpression of Neurogenin-2 compared to isogenic control. Interestingly, neurons develop to form hyperexcitable networks in other intellectual autism spectrum as Kleefstra syndrome,


Introduction
Fragile X syndrome (FXS) is characterized by physical abnormalities, anxiety, intellectual disability, hyperactivity, autistic behaviors and seizures 1 . In FXS, the expansion of the CGG triplet repeats in the FMR1 gene leads to their hypermethylation and the hypermethylation of the FMR1 promoter. This causes the transcriptional silencing of the FMR1 gene after the tenth week of gestation, and the reduction or absence of FMR1 protein (FMRP) 2 . FMRP is a RNAbinding protein and is thought to be a translational repressor at the synapse. Previous studies have suggested that its absence causes defects in neuronal development 1 . Although an imbalance of excitatory and inhibitory neuronal activity has been associated with seizure activity and autistic behaviors in several Autism Spectrum Disorders (ASDs) as well as in corresponding animal models [3][4][5] , data on FXS patients is scarce. Morphologically, increased spine density and increased numbers of long and immature-appearing spines were found in the cortices of FXS patients, highlighting defects in neuronal connectivity in FXS 6 .
So far, findings from animal models of FXS have failed to translate into successful therapies, highlighting the need to develop human cell-based models of FXS. Several research studies have used neurons derived from human induced pluripotent stem cells (iPSC) or human embryonic stem cells (ESC) cultured in vitro [7][8][9][10] , with some studies reporting impaired maturation and synaptic hypoactivity of FXS neurons 7,8,10 , and other studies showing synaptic hyperactivity and faster acquisition of synapses in FXS neurons 11,12 . Canonical two-dimensional culture conditions fail to fully recapitulate the morphological and functional characteristics of neurons in the human brain and show important intrinsic variation. Additionally, brain organoids, the most commonly used 3-dimensional in vitro model, contain different cell types interacting with each other, such as glial cells, neural progenitor cells and neurons, making it hard to dissect the cell-autonomous consequences of disease-associated mutations on developing neurons, as neurons grow while interacting with cell types also affected by these mutations.
Moreover, a recent study showed that stress pathways were ectopically activated in brain organoids, which impaired cell type specification. These defects were alleviated by transplantation of the organoids in the mouse brain cortex 13 . This suggests that transplantation of human neural progenitor cells (hNPCs) into the mouse brain may better reflect cellular behavior in the human brain than growth in the culture dish. Previous studies have shown that once transplanted into the neonatal brain, hNPCs migrate away from the injection site, differentiate into neurons, undergo maturation, express brain region-specific markers and become electrically active [14][15][16] .
Here, we established a novel in vivo model of FXS using co-transplantation of hNPCs differentiated from FXS patient-derived iPSC and isogenic control iPSC in the brain of immunedeprived mouse neonates. We assessed the cell-autonomous effects of FMR1 silencing on the development of human neurons in an in vivo context. Gene ontology analysis of single cell RNA sequencing data from transplanted FXS and isogenic control neurons showed upregulation of pathways linked to neurogenesis, neuronal differentiation and synaptic activity in FXS neurons.
Transplanted FXS neurons matured faster than isogenic control, and showed a persistent increase in Arc expression. Additionally, FXS striatal medium spiny neurons (MSNs) had wider dendritic synaptic protrusions at 6-7 months post-injection (PI). Our data suggest that FXS neurons transplanted in the mouse brain mature faster and are synaptically hyperactive compared to isogenic control.

Differentiation and migration of hNPCs transplanted into the neonatal brain
To investigate the developmental phenotypes of FXS neurons in vivo, we used two different FXS/corrected isogenic control pairs. In the FXS_SW/ C1_2_SW pair, the isogenic control C1_2_SW was generated by CRISPR-mediated deletion of the CGG repeats in the FXS patientderived male FXS_SW iPSC line 17 , leading to the reactivation of the FMR1 promoter and FMRP reexpression (deletion pair). In the dCas9-Tet1/dCas9-dTet1 pair, the FXS patientderived male FXS2 iPSC line was targeted with a catalytically inactive Cas9 fused to Tet1, a methylcytocine dioxygenase, as previously described 11,18 . This led to the demethylation of the CGG repeats, the reactivation of the FMR1 promoter and FMRP reexpression (dCas9-Tet1 iPSC line). To generate a control where the CGG repeats were not demethylated, the FXS2 iPSC line was targeted with a catalytically inactive Cas9 fused to a catalytically inactive Tet1 (dCas9-dTet1 iPSC line) (Suppl. Figure 1). To eliminate the potential off-target effects associated with methylation editing in our phenotypical analysis, we compared the dCas9-dTet1 iPSC line, which is expected to display an FXS phenotype, and its isogenic control dCas9-Tet1 iPSC line (demethylation pair) in all the experiments. Both the demethylation and deletion of the CGG repeats rescued FXS neuronal activity phenotypes in vitro 11,12 . hNPCs were generated from these iPSCs and sorted for PSA-NCAM expression using magnetic-activated cell sorting (MACS) in order to isolate hNPCs directed towards a neuronal fate. FXS and control NPCs were labeled with different reporters (GFP or tdTomato) in order to distinguish FXS cells from control cells after transplantation into the mouse brain ( Figure 1A).
We co-injected hNPCs derived from FXS and isogenic control cell lines into the brain ventricles of immune-deficient mouse neonates ( Figure 1A) and analyzed the brains at 1, 3 and 6 months post-injection (PI). Transplanted hNPCs migrated through the mouse brain and populated several areas including the hippocampus, cortex, striatum, thalamus, and midbrain ( Figure 1B Figure 4). Additionally, the demethylation of the FMR1 promoter was maintained in the epigenetically edited isogenic pair, and RT-qPCR of total mRNA showed robust expression of the FMR1 gene in C1_2_SW (control, deletion pair) cells at 1 month PI (suppl Figure 4D,E).

Accelerated maturation of transplanted FXS neurons
DAPI is predominantly impermeant to live cells, which allows it to be used as a cell viability dye, DAPI-negative cells being considered viable. In order to investigate cell-autonomous gene expression phenotypes of transplanted FXS neurons, we collected GFP+DAPI-and tdT+DAPIisogenic control and FXS viable transplanted neurons from the deletion pair by whole-brain extraction and subsequent FACS sorting and performed single-cell RNA sequencing on the cells using the 10x Genomics platform. We analyzed two different mouse brains engrafted with FXS and isogenic control hNPCs, at 1 month PI.
FXS and isogenic control cells got assigned to similar UMAP clusters, allowing us to compare FXS and isogenic cells within the same clusters ( Figure 2A). Cell types corresponding to the different clusters were annotated according to the expression of known marker genes in UMAP clusters of grouped FXS and control cells (clusters C0 to C7, Figure 2B). In line with immunofluorescence data, the cells did not express markers specific for microglia, oligodendrocyte precursor cells or oligodendrocytes (data not shown). To identify the cell types corresponding to the different clusters, we used canonical marker genes and cell cycle markers (Suppl. Figure 5A,B). Cluster C7 was very small and was excluded from further analyses. Most clusters were identified as neurons, with clusters C0 and C3 being immature neurons and cluster C5 and C6 being more mature inhibitory and excitatory neurons respectively. Cluster C4 was composed of astrocytes, and clusters C1 and C2 were identified as neural progenitor cells (NPC) via expression of markers for NPC markers and proliferation. To obtain an unbiased characterization of cellular dynamic processes, we used Slingshot 19 to assign a pseudotime to each cell, representing where the cell is along developmental trajectories. The trajectory analysis showed that cells transitioned from NPCs (clusters C1-NPC 1 and C2-NPC 2) to immature neurons (cluster C0-Immature neurons) to immature excitatory neurons (cluster C3-Immature excitatory neurons) and then to mature neurons (cluster C5-Inhibitory neurons and cluster C6-excitatory neurons ) (Suppl. Figure 5C). We also performed gene list enrichment analysis using ToppGene 20 , comparing the differentially expressed (DE) genes in cluster C3 and cluster C0, cluster C3 and C5, cluster C3 and C6. Neuronal differentiation pathways were upregulated in cluster C3 compared to cluster C0, and in clusters C5 and C6 compared to cluster C3, confirming the order of progression of these clusters in terms of neuronal maturation (Suppl. Figure 5D).
Gene list enrichment analysis of the differentially expressed genes between FXS and isogenic neuronal clusters (cluster C0-immature neurons, C3-immature excitatory neurons, C5-inhibitory neurons and C6-excitatory neurons) indicated the upregulation of neurogenesis, neuronal differentiation and glutamatergic synaptic transmission pathways ( Figure 2C). During neuronal differentiation, neural progenitors exit the cell cycle and differentiate into neuroblasts. Gene list enrichment analysis of the differentially expressed genes in the neuronal clusters revealed downregulation of pathways involved in cell division ( Figure 2D). We analyzed the proportion of FXS and isogenic control cells in each cluster, and found that the distribution of the FXS cells was shifted towards more mature clusters ( Figure 2E). Together, these data suggest that FXS cells were more mature than isogenic control cells from the deletion pair at 1 month PI. In summary, these results showed upregulation of neurogenesis, neuron development and glutamatergic synaptic transmission pathways in FXS cells, suggesting increased or accelerated neuronal maturation.
In order to further investigate alterations in the maturation and synaptic activity of FXS neurons, we analyzed transplanted neurons on brain slices using immunofluorescence. Neurons underwent morphological changes suggesting maturation from 1 to 6 months PI as indicated by the dendrites of the neurons getting thicker and longer and the cell body becoming rounder (suppl Figure 6). Additionally, at 6 months PI, some neurons in the striatum and other brain areas displayed dendritic spines and brain region-specific marker expression (suppl Figure 5), suggesting they had undergone brain region-specific differentiation and maturation. We investigated whether FXS neurons showed accelerated maturation as compared to isogenic controls using immunofluorescence on brain slices from transplanted mice. Neuroblasts initially express doublecortin (DCX), an immature neuron marker and gradually lose DCX expression and start expressing NeuN, a mature neuron marker 21 . hNPCs transplanted into the neonatal mouse brain lost DCX and acquired NeuN expression between 1 and 6 months PI (

Increased synaptic activity of transplanted FXS neurons
Previous in vitro experiments published by our group and other groups showed that FXS neurons in both isogenic pairs were hyperexcitable in vitro 11,12 . Furthermore, our single-cell RNA sequencing analysis at 1 month PI suggested an increase in excitatory synaptic activity in FXS neurons. Arc is an immediate early gene expressed during synaptic plasticity and is commonly used as a marker of neuronal activity. In addition, Arc is a direct inhibitor of FMRP: its translation is repressed by FMRP, and FMRP absence is expected to increase Arc expression.
To determine whether Arc was upregulated in FXS neurons, we used immunofluorescence on transplanted mouse brain slices to assess Arc expression in control and FXS neurons. Neurons were defined by DCX positivity at 1 month PI, as most neurons express DCX at this timepoint and by NeuN staining at 3 months and 6 months PI, as the majority of neurons express this marker at these timepoints. At 1 month PI, a higher percentage of FXS neurons from the deletion pair was Arc-positive (Arc+) compared to isogenic control ( Figure 4A). No significant difference was observed for the demethylation pair, although the percentage of FXS Arc+ neurons tended to be higher than in the isogenic control ( Figure 4B To further investigate changes in arborization complexity, excitatory synaptogenesis and synaptic activity in FXS neurons, we assessed the dendritic protrusion density and morphology of medium spiny FXS and isogenic control neurons in the striatum at 6-7 months PI. We used DARPP32 as a marker for mature medium spiny neurons (MSNs) (Suppl. Figure 7A). Control and FXS MSNs exhibited complex morphology at 6-7 months PI (Suppl. Figure 7B Figure 5G,H). This suggests that, although excitatory synaptogenesis in FXS MSNs is unchanged, excitatory synaptic strength is increased. This is in line with increased excitatory synaptic activity in FXS neurons. Together, these results suggest increased synaptic activity, but not synaptogenesis, in FXS neurons compared to control.

Discussion
In this study, we designed a new transplantation model of FXS, that allows FXS neurons to develop in the in vivo context of the mouse brain. The transplanted cells were sparsely integrated in the mouse brain and surrounded by mouse cells, allowing for the assessment of their cell-autonomous developmental defects. The transplanted hNPC were sorted for high PSA-NCAM expression but not directed towards a specific neuronal subtype, and gave rise to a high number of neurons at 1 and 3 months PI. Neurons were fully mature at 6 months PI as assessed by NeuN expression, and some neurons differentiated in MSNs in the striatum, displaying complex morphology characteristic of MSNs, a high dendritic spine density, and expressing DARPP32, a marker specific to striatal MSNs.
In this study, we found accelerated maturation of FXS neurons compared to control. In agreement with our results is another study that showed accelerated maturation of ASD neurons as compared to control neurons 23 . Furthermore, genetic haploinsufficiency of SYNGAP1/Syngap1, which commonly occurs in intellectual disability, ASD and epilepsy, also leads to accelerated maturation of neocortical pyramidal cells in a mouse model 24 .  6 . This discrepancy may be explained by the fact that the dendritic spine phenotype these studies reported was non-cell autonomous and likely caused by FXS astrocytes. Indeed, our transplanted FXS and control neurons were exposed to the same neurodevelopmental niche and FXS neurons were not in direct contact with FXS astrocytes, whereas in these two studies both neurons and astrocytes were mutant for FMR1. In line with this hypothesis, adult astrocyte-specific FMR1 KO mice displayed increased spine density and a higher proportion of thin spines in the motor cortex, indicating that FXS astrocytes are necessary and sufficient for the dendritic spine phenotype of neurons in FXS 26 .
Our data on Arc expression and dendritic protrusion diameter suggest that FXS neurons are hyperactive compared to control. This is in line with previous work indicating that the excitation/inhibition balance is altered in FXS patients, and suggests that cell-autonomous effects are involved in the hyperexcitability of the neuronal networks. These results are in line with previous studies from our laboratory and others 11,12 that revealed hyperexcitability in However, our work contradicts other in vitro studies that showed impaired maturation of FXS neurons and hypoexcitability 7,8,10 . It is of note that these studies did not use isogenic pairs of iPSC. Intrinsic variation between cell lines may be a confounding factor in studies using nonisogenic pairs. Notably, in our study, the two isogenic pairs did not display the same time course of maturation. The deletion pair matured more rapidly than the demethylation pair, as assessed by expression of DCX and NeuN.
Two genes of interest, OTX1 and TBX1 were strongly downregulated and upregulated, respectively, in FXS transplanted neurons at 1 month PI, as assessed by scRNA seq (Tables S1 and S2). OTX1 is associated with autistic behavior and epilepsy, two symptoms of FXS.
Additionally, OTX1 knock-down was previously shown to increase the astrocyte/neuron ratio in the developing cortex, and to maintain neural progenitors in the transient amplifying phase 31 . It is also required for the refinement of axonal projections of cortical neurons to subcortical targets 32 . This is in line with phenotypes previously described in FXS models: increased neural progenitor proliferation and aberrant axogenesis [33][34][35][36] . TBX1 is also associated with ASD and intellectual disability. In particular, TBX1 mutations leading to a gain of function were linked to intellectual disability 37,38 . This suggests that OTX1 and TBX1 may be interesting targets for rescuing the cell-autonomous phenotypes of FXS neurons.
A major challenge in drug development is assessing the ability of the molecules to pass the blood-brain barrier. Our human-mouse chimeric model provides us with a readout to test potential therapeutic molecules acting on the cell-autonomous phenotypes of neurons while accounting for their potential modifications in the organism and their ability to cross the bloodbrain barrier, in contrary to 3D culture models. Additionally, neurons appear to reach full maturation at 6 months PI, and display complex arborization. However, the number of transplanted neurons decreased between 3 and 6 months PI, and transplanted neurons were sparse at 6 months PI, making the analysis at later timepoints tedious. Additionally, our approach does not allow us to control the fate of the transplanted cells, making the study of different neuronal subtypes difficult. To solve this issue, one could consider injecting hNPCs directed towards a striatal, cortical or hippocampal fate, or, as done in previous studies, cortical or striatal neurons [39][40][41] . A potential limitation of the neuronal injection approach is the low contribution of the cells to the mouse brain, as neurons are less resistant to stress than neural precursor cells and do not proliferate after injection.

Experimental Animals
The animals used for this study were NOD/SCID/gamma mice. They were kept in group housing under standard barrier, light and temperature-controlled conditions. Food and water were available ad libitum. Every effort was made to minimize the number of animals used and their suffering, and all experiments were performed in accordance with the Department of Comparative Medicine and Massachusetts Institute of Technology animal husbandry standards.

Induced pluripotent stem cell (iPSC) culture and Neural progenitor cell (NPC) differentiation
FXS patient-derived iPSC lines and isogenic control cell lines used in this study are listed in Table S3. iPSCs were cultured in feeder-free conditions on Geltrex (ThermoFisher Scientific, A1413302) or Matrigel coated flasks in StemFlexTM medium (ThermoFisher Scientific, A3349401). Cells were passaged using ReLeSRTM (STEMCELL Technology, 05873) and tested for mycoplasma contamination and karyotypic abnormalities.
NPCs were induced with dual SMAD inhibition using PSC Neural Induction medium (ThermoFisher Scientific, A1647801) according to the manufacturer's protocol. Briefly, when reaching ~80% confluency, iPSCs were dissociated with AccutaseTM Cell Dissociation Reagent

Lentiviral labeling of NPC
Lentiviruses carrying the expression cassette of GFP or tdTomato were produced by transfecting HEK293T cells with FUW constructs together with standard packaging vectors (pCMV-dR8.74 and pCMV-VSVG) followed by ultra-centrifugation-based concentration. NPCs were infected with these viruses. Once strong expression of GFP or tdT was visible, labeled NPC were purified by FACS sorting and amplified on Geltrex or Matrigel using STEMdiff™ Neural Progenitor Medium.

NPC transplantation
Cultured NPCs were dissociated using Accutase and resuspended in phosphate buffer saline without calcium and magnesium prior to injection, at a concentration of 105 cells/μL. Post-natal day 0 to post-natal day 3 mouse pups of either gender were manually injected with a total of 4x105 NPC dispersed over four injection sites in the lateral ventricles (two injection sites per brain hemisphere, one anterior and one posterior) using glass micropipettes.

Transplanted human cell extraction for single-cell RNA sequencing
Mice were euthanized using cervical dislocation, and the brains were extracted and dissociated using Miltenyi Adult mouse and rat Brain Dissociation Kit and the gentleMACS Octo Dissociator

Single-cell RNA sequencing and data analysis
The cells were sequenced immediately after extraction and FACS sorting. 5000 cells were targeted. Sequencing data was mapped to a reference meta-genome composed of human hg38 (GRCh38), mouse mm10 (GRCm38), GFP and tdTomato sequences using Cell Ranger (v3.0.2) according to the 10x Genomics pipeline. We used Ensembl version 97 gene annotation to assign UMIs to genes. Contamination of the samples by murine cells was very small, as 0.9%, 2.8%, 6.0% and 3.2% of total cells had more than 70% of UMIs mapped to mouse genes in control 1, control 2, FXS 1, and FXS 2 respectively. Only the cells with more than 95% of UMIs mapped to human genes were used for subsequent analysis.
Single-cell RNA-seq analysis Seurat v3 42 was used for quality control and analysis of the single cell RNA-seq experiment.
We followed the Seurat Guided Clustering Tutorial (https://satijalab.org/seurat/pbmc3k_tutorial.html). To remove cells of low quality or with too few reads, the following cut offs were applied: "nFeature_RNA" above 1500, "nCount_RNA" less than 40000, "percent.mt" less than 20. Because differences in cell cycle among individual cells can be an unwanted source of heterogeneity, these effects were mitigated by regressing out the difference between the G2/M and S phase scores, where the scores were based upon canonical markers of cell cycle. Cells were embedded in a k-nearest neighbors (kNN) graph, which was based on the Euclidean distance in the space of the 30 leading principal components. The quality of the cell partitions to clusters was optimized by applying the Louvain algorithm with a resolution of 0.2.

Differential gene expression in scRNA-seq
Since our experiment had a complex design with multiple paired replicates, we used Muscat (Multi-sample multi-group scRNA-seq analysis tools) 43 for differential expression analysis.
Using the set of high-quality cells from Seurat, Muscat aggregated cells based on metadata such as samples, clusters and conditions, and created pseudo-bulk RNA-seq profiles using the sum of raw counts. Statistics for differentially expression were calculated with limma 44 using a paired design, and differential state analysis was performed with edgeR 45 . The genes with at least two-fold changes and adjusted p-values less than 0.05 were considered to be differentially expressed.

Trajectory
We performed lineage reconstruction and pseudotime inference with Slingshot 19 . We used dimensionality reduction produced by PCA, and NPC as a starting point. Since we didn't know the order of developmental stage in the two NPC clusters (clusters C1-NPC 1 and C2-NPC 2), we ran each of these two clusters as a starting point. In both cases, two lineages were identified. The first lineage was from NPCs (cluster C1-NPC 1 or C2-NPC 2) to immature neurons (cluster C0-Immature neurons) to immature excitatory neurons (cluster C3-Immature excitatory neurons) and then to mature neurons (cluster C5-Inhibitory neurons and cluster C6excitatory neurons). The second lineage was from NPCs (cluster C1-NPC 1 or C2-NPC 2) to C4-astrocytes.

Gene list enrichment analysis
We performed gene list enrichment analysis on the DE protein-coding genes between control and FXS using ToppFun from the ToppGene suite 20,46 . The probability density function method was used to estimate p-values and a pathway significance threshold of FDR<0.05 was set. The Benjamini-Hochberg procedure was used to correct for multiple hypothesis testing."

DNA Methylation analysis
Pyro-seq of all bisulfite-converted genomic DNA samples was performed with the PyroMark Q48 Autoprep (QIAGEN) according to the manufacturer's instructions. The primers for pyro-seq of the FMR1 promoter are listed below.

Immunohistochemistry and immunocytochemistry
Cells in culture were fixed using 4% PFA. For the IHC analysis of cells transplanted in the mouse brain, mice at 1, 3 or 6 months post-transplantation were perfused with 4% FPA through transcardial route. 100 µm-thick brain sagittal slices were sliced with Cryostat tissue slice instrument (Leica) after cryopreservation in 30% sucrose. Immunostainings were performed using antibodies listed in Table S4. For analyses of Arc expression by IHC, a rabbit anti-FMRP antibody (Cell Signaling Technology 7104) was labeled with Alexa fluor 555 fluorophore using an antibody labeling kit (Thermo Fisher Scientific A20187). Images were acquired using a Zeiss LSM700 confocal microscope. Antibodies used in this study are listed in Table S2.

Neuronal imaging, tracing and morphometric analysis
Cells were identified as striatal medium spiny neurons using colocalization of DARPP32 immunofluorescence staining with the cell body.

Neuronal arborization analysis
Confocal microscopy z-stacks of control and FXS neurons were acquired with a 20x dry objective with zoom 0.6 using a Zeiss LSM 700 confocal microscope.