The patient-specific mouse model with Foxg1 frameshift mutation uncovers the pathophysiology of FOXG1 syndrome

Single allelic mutations in the gene encoding the forebrain-specific transcription factor FOXG1 lead to FOXG1 syndrome (FS). Patient-specific animal models are needed to understand the etiology of FS, as FS patients show a wide spectrum of symptoms correlated with location and mutation type in the FOXG1 gene. Here we report the first patient-specific FS mouse model, Q84Pfs heterozygous (Q84Pfs-Het) mice, mimicking one of the most predominant single nucleotide variants in FS. Intriguingly, we found that Q84Pfs-Het mice faithfully recapitulate human FS phenotypes at the cellular, brain structural, and behavioral levels. Importantly, Q84Pfs-Het mice exhibited myelination deficits like FS patients. Further, our transcriptome analysis of Q84Pfs-Het cortex revealed a new role for FOXG1 in synapse and oligodendrocyte development. The dysregulated genes in Q84Pfs-Het brains also predicted motor dysfunction and autism-like phenotypes. Correspondingly, Q84Pfs-Het mice showed movement deficits, repetitive behaviors, increased anxiety, and prolonged behavior arrest. Together, our study revealed the crucial postnatal role of FOXG1 in neuronal maturation and myelination and elucidated the essential pathophysiology mechanisms of FS.


Introduction
Foxg1 is the forkhead family transcription factor whose expression is induced as the telencephalon begins to form in embryos 1 . The elimination of the Foxg1 gene results in a striking loss of the forebrain volume 2 , highlighting the crucial role of Foxg1 in forebrain generation.
The cortex comprises six cortical layers occupied by distinct subtypes of cortical projection neurons produced from the neural progenitors within the dorsal telencephalon 3 . In addition to excitatory projection neurons, the cortex contains inhibitory GABAergic interneurons that are generated in the ventral telencephalon and migrate tangentially to the developing cortex 4 . Foxg1 plays a critical role in multiple steps of cortex development. Foxg1 is important for properly patterning the telencephalon at early stages [5][6][7] . In neural progenitor cells, Foxg1 is required for the proliferation and maintenance of the progenitor population by blocking cell cycle exits and premature neuronal differentiation 2,5,8,9 . In cortical neurons, Foxg1 plays a crucial role in differentiation, migration, subtype speci cation, and axon projection [10][11][12][13][14] . The speci c and complete elimination of Foxg1 in postmitotic cortical projection neurons results in inverted cortical layers 11 , indicating that Foxg1 is needed for proper cortical lamination. Notably, deleting only a copy of the Foxg1 gene in cortical projection neurons is su cient to attenuate the production of upper-layer neurons, changing the composition of cortical layers 11 . Deleting an allele of the Foxg1 gene leads to reduced cortical interneurons 15 , while the complete inactivation of Foxg1 in cortical interneurons interferes with their proper integration into the cortical layer 16 . Therefore, Foxg1 coordinates the orderly production, migration, and integration of glutamatergic excitatory neurons and GABAergic inhibitory interneurons in the developing cortex. Importantly, these studies highlight that both copies of the Foxg1 gene are required for building the functional cortex.
In mice, neuronal migration and layer formation are largely completed by birth, and glial cells begin to emerge 3 . Oligodendrocyte precursor cells (OPCs) in the cortex proliferate and differentiate into myelinating oligodendrocytes (OLs) after birth, and the myelination continues to adulthood 17 .
In humans, the pathogenic variants in the FOXG1 gene lead to the debilitating neurodevelopmental disorder collectively termed FOXG1 syndrome (FS) 18, 19,19−23 . Most FS is caused by de novo mutations in a single allele of the FOXG1 gene. FS patients exhibit structural brain abnormalities, such as microcephaly and corpus callosum agenesis, and a delay in oligodendrocyte differentiation and myelination. FS is also characterized by severe intellectual disability, hyperkinetic-dyskinetic movement disorder, irritability, and epilepsy. Many FS patients have autistic features, such as repetitive movements, poor social interaction skills, and a near absence of verbal speech and language development. Hence, FS belongs to the autism spectrum disorder (ASD). Currently, the molecular and cellular mechanisms leading to the pathology of FS remain elusive. Notably, a wide spectrum of symptoms of FS patients is associated with the type and location of the causative FOXG1 variants 19,24 . This highlights the need for patient-speci c FS mouse models, which carry FS-causing genetic variants, to understand FS pathophysiology and develop therapeutic interventions for FS patients. The existing Foxg1-null mouse lines 2,9,25,26 are inadequate as they have the complete deletion of the Foxg1 gene.
In this study, we report the rst patient-speci c FS mouse model that precisely mimics the predominant genetic variation of FS; Q84Pfs-Het mice that carry a heterozygous frameshift mutation in the Foxg1 gene. Remarkably, Q84Pfs-Het mice recapitulate not only FS-causing genotype but also diverse human FS symptoms from cellular to behavioral phenotypes, including uncoordinated movements and ASD-like behaviors. By studying the dysregulated genes in the cortex of Q84Pfs-Het mice, we identi ed the novel molecular pathways of FS in neurons and OL lineage. Together, our study with the newly established patient-speci c FS mouse model provides important insights into the gene regulatory pathways of FOXG1 and the pathophysiology of FS. Further, our patient-speci c FS mouse model creates the platform to develop FS therapeutic strategies.

Results
The coding region of FOXG1's N-terminus is a mutation hotspot Generation of mice carrying Q84Pfs allele To understand the pathophysiology of FS, we generated FS mouse model bearing a speci c FS-causing FOXG1 variant. The knock-in mice carry c.250dupC (p.Q84Pfs*31, herein referred to as Q84Pfs), equivalent to human c.256dupC (p.Q86Pfs*35, referred to as Q86Pfs), representing one of the most prevalent variants in FS (Fig. 1a, 2a,b). Foxg1 Q84Pfs/Q84Pfs homozygous (Q84Pfs-Homo) mice exhibited a marked loss of forebrain tissue and died soon after birth (Fig. 2c, Supplementary Fig. 1). They also showed craniofacial defects, such as a short frontonasal process, and underdeveloped eye morphology ( Supplementary Fig. 1). In contrast, Foxg1 Q84Pfs/+ heterozygous (Q84Pfs-Het) mice survived into adulthood. Structural brain de cits in Q84Pfs-Het mice Compared to wild-type (WT) littermate controls, adult Q84Pfs-Het mice showed smaller brains, shortened corpus callosum, and profound septum defects (Fig. 2d,e). The hippocampus size was reduced, and the malformation of the dentate gyrus was notable (Fig. 2d). Thus, Q84Pfs-Het mice recapitulate the brain structural de cits in human FS with N-terminal frameshift variants (Fig. 1d,e).

RNA-seq analysis identi ed the dysregulated gene sets in Q84Pfs-Het cortex
To identify the dysregulated genes contributing to the de cits of Q84Pfs-Het brains, we performed RNAseq in P1 cortices of Q84Pfs-Het and littermate WT control mice and compared their gene expression patterns. 222 genes were signi cantly changed in the Q84Pfs-Het cortex relative to control WT cortex, representing differentially expressed genes (DEGs) (Fig. 3a, Supplementary Dataset). Interestingly, genes expressed in the cortical upper-layer projection neurons were downregulated, whereas deep-layer projection neuronal genes were upregulated (Fig. 3b), suggesting that the microcephaly of Q84Pfs-Het cortex is at least partly attributed to the reduction of the upper-layer, not deep-layer. Many genes involved in GABAergic interneuron development were also downregulated in Q84Pfs-Het cortex (Fig. 3b), indicating defects in the development of cortical interneurons.
Approximately 30% of DEGs in Q84Pfs-Het cortex were associated with Foxg1-bound ChIP-seq peaks in the developing cortex 11 regardless of up-or down-regulation (Fig. 3c), suggesting that Foxg1 regulates its direct target genes both negatively and positively. Consistently, the HOMER motif analysis 27 revealed that the top two signi cantly enriched motifs in Foxg1-bound peaks annotated to either up-or downregulated genes are the FOX motif and E-box. E-box serves as the binding site for basic helix-loop-helix (bHLH) and Zbtb18 (Rp58) transcription factors, both of which were shown to collaborate with Foxg1 11,28 (Fig. 3d). The NF1 site was also signi cantly associated with both up-and down-regulated Foxg1target genes (Fig. 3d). These data suggest that a subset of dysregulated genes in Q84Pfs-Het mice are direct target genes of Foxg1 and that Foxg1 cooperates with ZBTB18, bHLH transcription factor, and NF1 in the developing cortex.

The dysregulated pathways in Q84Pfs-Het cortex
To gain insights into molecular and cellular pathways leading to defects of Q84Pfs-Het mice, we performed the gene set enrichment analysis (GSEA) of DEGs using the Database for Annotation, Visualization and Integrated Discovery (DAVID) 29 and Enrichr 30 .
In tissue analysis, DEGs were enriched in neuronal epitheliums, prefrontal cortex, cerebral cortex, cingulate gyrus, and superior frontal gyrus (Fig. 3e), consistent with the use of cortex for RNA-seq. In cell type analysis, the upregulated genes were signi cantly associated with cortical layer 6 glutamatergic neurons and, interestingly, oligodendrocytes, in which Foxg1 function remains unclear (Fig. 3f). The downregulated genes were most highly enriched for the "interneuron: embryonic prefrontal cortex" (Fig. 3f). Thus, our bioinformatic analyses indicate that the three major cell types in the neonatal cortex, excitatory projection neurons, inhibitory GABAergic interneurons, and the oligodendrocyte lineage, have a de ciency in Q84Pfs-Het mice.
Consistent with the known function of Foxg1 10-14 , DEGs of Q84Pfs-Het cortex were enriched for the genes controlling neuronal projection development, neuronal cell body, axon, and transcription regulations (Fig. 3g,h). Interestingly, the gene involved in the negative regulation of cell proliferation was downregulated in Q84Pfs-Het cortex (Fig. 3h), suggesting that progenitor proliferation may increase in Q84Pfs-Het cortex unlike the heterozygous mice carrying Foxg1-null allele (global Foxg1-Het) 2,5,8,9 . Intriguingly, many synaptic genes were signi cantly dysregulated in Q84Pfs-Het cortex. The upregulated genes were enriched for biological process (BP) terms of synaptic vesicle transport, synaptic vesicle exocytosis and endocytosis, and presynaptic membrane assembly, and cellular component (CC) terms of synapse and presynapse (Fig. 3g,h). Further, the downregulated genes were also strongly associated with the CC terms of synapse and postsynaptic density (Fig. 3g,h). Our results underline the new role of Foxg1 in synapse development. It is also noteworthy that the upregulated and downregulated genes were more strongly associated with presynapse and postsynaptic density, respectively.
Remarkably, both up-and down-regulated genes were enriched for translation and ribosome (Fig. 3g,h), suggesting defects in translation in Q84Pfs-Het brains. Other enriched terms, such as locomotory behavior and memory, are related to FS symptoms, such as intellectual disability and hyperkineticdyskinetic movements [19][20][21][22][23] .
Together, our results highlight the new role of Foxg1 in the oligodendrocyte development and synapse formation and function.

Developmental defects of neural progenitors and excitatory and inhibitory neurons in Q84Pfs-Het cortex
We next assessed if the dysregulated genes in Q84Pfs-Het brains led to cortical defects suggested by our GO analyses. Interestingly, although Q84Pfs-Het cortex was thinner than the WT cortex by E16 (Fig. 4a,b),
The production of Cux1 + upper-layer neurons was delayed in E16 Q84Pfs-Het cortex, as shown by a lack of Cux1 + neurons in the super cial area of the cortex (Fig. 5a). Cux1 + upper-layer neurons remained signi cantly reduced in P30 Q84Pfs-Het cortex (Fig. 5b,e), indicating that the delayed upper-layer neuron generation is not compensated at the later stages. The deep-layer neurons showed a trend of increase in number, but the thickness of the deep-layer did not signi cantly increase ( Fig. 5a,b,d). Combined with reduced cortex thickness ( Fig. 5. a-c), these results show that mainly the reduction of the upper-layer accounts for the thinner cortex. Dlx1 + GABAergic interneurons were substantially decreased in Q84Pfs-Het cortex ( Fig. 5f,g), consistent with downregulation of genes involved in cerebral cortex tangential migration (i.e., GABAergic interneuron migration de cits) (Fig. 3f,h).
Q84Pfs-Het mice displayed two prominent axon projection de cits in the cortex, matching the dysregulation of neuron projection development genes (Fig. 3g,h). First, a substantial fraction of callosal axons was stalled at the midline, forming the Probst bundle (Fig. 5h). Second, Q84Pfs-Het showed misprojected L1 + axon bundles crossing the cortical plate toward the super cial area. These misrouted L1 + axons expressed Ntng1 (netrin G1) (Fig. 5i), indicating the defects of thalamocortical axon guidance in Q84Pfs-Het cortex.

Myelination defects in Q84Pfs-Het cortex
Although delayed myelination is one of the most common features of FS brains (Fig. 1d), the role of Foxg1 in the oligodendrocyte lineage development remains unclear. To test if the FS mouse model recapitulates this prominent white matter phenotype in humans, we monitored oligodendrocyte differentiation and myelination in Q84Pfs-Het mice at P30, the active myelination phase. Interestingly, Olig2 + oligodendrocyte lineage cells, many of which include oligodendrocyte progenitor cells (OPCs), signi cantly increased in Q84Pfs-Het cortex (Fig. 6a,b), consistent with the association of upregulated genes with oligodendrocyte (Fig. 3f). Despite of the increased Olig2 + cells, Mbp + myelinated area was substantially reduced in Q84Pfs-Het brains (Fig. 6c,d). Moreover, Q84Pfs-Het brains exhibited a marked reduction of the arborization and complexity of myelination patterns (Fig. 6c,e). Our results indicate that the oligodendrocyte development process was disrupted, and the myelination steps, including the microstructural organization of myelination, was impaired in Q84Pfs-Het brains, providing the pathological mechanisms underlying white matter de ciency in FS.
Dysregulated genes in Q84Pfs-Het cortex were associated with motor dysfunction and autistic-like behaviors To ask if the dysregulated genes in the Q84Pfs-Het cortex are connected to speci c human conditions, we performed GSEA with the DisGeNET database, which contains collections of genes and variants associated with human diseases 31 (Fig. 7a). The upregulated genes were signi cantly associated with impaired social interactions, Alzheimer's disease, ASD, Parkinson's disease, and autonomic nervous system disorders (Fig. 7a). The downregulated genes were enriched for dystonia, myoclonic encephalopathy, myoclonus, and ataxia, as well as major depressive disorder. Together, Q84Pfs-Het transcriptome changes strongly linked to impairments in movement, social interactions, and autonomic nervous system function in Q84Pfs-Het mice.
Given that both FS and Huntington's disease (HD) exhibit many neurological symptoms, including abnormal involuntary movements 22,32 , we compared the dysregulated genes between Q84Pfs mice and HD using the database of HD molecular signatures (HDSigDB, hdinhd.org) (Fig. 7b). This analysis revealed the striking resemblance between Q84Pfs-Het mice and HD mouse models 33,34 . The highranking categories that resemble the up-and down-regulated genes in Q84Pfs-Het mice were the up-and down-regulated genes in the hippocampus of the HD mouse model Q175 33 , respectively (Fig. 7b). Notably, the two categories among the top ve gene sets for the upregulated genes were oligodendrocyte progenitor cell (OPC) genes (Fig. 7b), consistent with our nding that Q84Pfs-Het cortex showed the increased Olig2 + oligodendrocyte lineage cells (Fig. 6a,b).
Collectively, the transcriptome changes in Q84Pfs-Het cortex are closely connected to the clinical features of FS, such as movement disorders, autism-like behaviors, and social impairment.
Q84Pfs-Het mice showed movement de cits, repetitive behaviors, and prolonged behavior arrest To investigate whether Q84Pfs-Het mice exhibit behavior phenotypes corresponding to the above molecular and cellular changes, we performed behavioral assessment tests at adolescence stage P30 and adult stages P60 and P90.
There was no signi cant difference in body weights of Q84Pfs-Het and WT mice at these stages (Fig. 7c).
In the wire hang test that evaluates the motor function and muscle strength, Q84Pfs-Het mice showed signi cantly reduced hang time, which worsened as mice aged (Fig. 7d). In the open eld test that monitors locomotor activity and exploratory behaviors, Q84Pfs-Het mice moved signi cantly less than WT mice at P60 and P90 (Fig. 7e). At P30, Q84Pfs-Het mice showed a tendency for travel distance reduction over time compared to WT mice, but the total travel distance did not differ signi cantly (Fig. 7e). These results indicate de cits in muscle strength and movement in Q84Pfs-Het mice.
Q84Pfs-Het mice exhibited a strikingly extended behavior arrest, de ned by paused locomotion for longer than 3 min at one episode during free exploration of the arena (Fig. 7f, Supplementary Fig. 2 and videos). Throughout these episodes of remarkably pronounced stops, Q84Pfs-Het mice showed a high degree of postural control, typically positioning their bodies facing the center of the arena, and no apparent overt movements, such as visual survey, rearing, or digging. The proportion of Q84Pfs-Het mice that exhibited the extended behavior arrest was 11%, 22%, and 47% at P30, P60, and P90, respectively, whereas WT mice did not show behavior arrest without purposeful movements under the same condition (Fig. 7f). Notably, Q84Pfs-Het mice without prolonged behavior arrests still showed a decreased travel distance, suggesting that the extended arrest is not the only contributing factor to movement reduction.
Next, we assessed ASD-related behaviors. The center time in the open eld test was markedly decreased as Q84Pfs-Het mice aged from P30 to P90, relative to WT mice (Fig. 7g), indicating increased anxiety levels. Q84Pfs-Het mice exhibited a signi cant increase in repetitive grooming behavior at all three stages, with a tendency for bigger differences from WT mice as they age (Fig. 7h). These results suggest heightened repetitive behavior in Q84Pfs-Het mice. Intriguingly, in the marble-burying test, Q84Pfs-Het mice buried more marbles at P30 but fewer marbles at the adult stages than WT mice at P60 and P90 (Fig. 7i), suggesting age-dependent changes in the marble burying behavior.
Together, our data uncovered that Q84Pfs-Het mice have motor coordination defects, increased anxiety levels, remarkably prolonged behavior arrest, and repetitive behaviors, all observed in FS [19][20][21][22][23] and strongly linked to molecular and cellular changes of Q84Pfs-Het brains.

Discussion
A majority of FS patients possess mutations within the FOXG1 gene coding region, which are likely to produce faulty Foxg1 protein products that impact disease mechanisms and progression. Therefore, the currently available Foxg1-null mouse lines are inadequate to investigate FS pathophysiology because they delete the entire Foxg1 coding region 2,9,25,35,36 . Here, we established Q84Pfs-Het mice as the rst patient-speci c FS mouse model that accurately replicates human genetic conditions of FS. Remarkably, the heterozygous mice carrying only a single allele of Q84Pfs variant, without creating the homozygous conditions, recapitulated a wide range of human FS symptoms, such as brain structural de cits and ASDlike behaviors. Using this new FS mouse model, we uncovered molecular and cellular changes, such as oligodendrocyte lineage de ciency and dysregulation of synaptic genes, leading to the key hallmarks of FS, such as delayed myelination, movement de cits, and ASD-like behaviors.
A tight genetic and phenotypic resemblance between Q84Pfs-Het mice and human FS patients provides clear advantages in understanding the pathophysiology of neurodevelopmental disorders and developing therapeutics, compared to many ASD mouse models, in which the ASD genes are eliminated, unlike the corresponding human conditions in which ASD genes are often present as heterozygous conditions or are epigenetically silenced 37 . First, it allowed us to identify dysregulated genes in FS brains, which are likely to lead to cellular and structural de cits and further behavioral outcomes. Second, our studies uncovered new mouse phenotypes highly correlated with human FS symptoms, such as de ciency in oligodendrocyte development and myelination, increased repetitive behavior and anxiety levels, and extensive behavior arrest. Thus, Q84Pfs-Het mice will serve as an ideal mouse model to identify the pathogenic mechanism for these mouse phenotypes and human symptoms. Third, Q84Pfs-Het mouse model enables investigation of the role of the truncated form of Foxg1 resulting from frameshift mutations in FS. Foxg1 mRNA level did not decrease in Q84Pfs-Het cortex (Supplementary dataset), suggesting that Foxg1 mRNA from Q84Pfs-Het allele did not undergo non-sense mediated decay. These data also suggest that Q84Pfs allele in mice (Q86Pfs allele in humans) expresses the N-terminal fragment of the Foxg1 protein. Notably, the Foxg1 fragment produced from mouse Q84Pfs and human Q86Pfs alleles would have interesting amino acid signatures of histidine, proline, and glutamine repeats (Fig. 1a). In this regard, it is noteworthy that Q84Pfs-Het mice show phenotypes distinct from the existing global Foxg1-Het mice. Q84Pfs-Het mice show increased NPC proliferation while global Foxg1-Het mice exhibited reduced NPC proliferation 38 . It will be interesting to examine the molecular basis of this stark difference. Fourth, the Foxg1 heterozygous condition of Q84Pfs-Het mice enables us to evaluate the therapeutics that will not be assessable in homozygous mouse models, such as an approach to enhance Foxg1 expression from the WT Foxg1 allele, resulting in a boost of functional Foxg1 proteins.
Our parallel studies of Q84Pfs-Het mice at molecular and behavioral levels provided important insights into FS pathogenesis. Q84Pfs-Het cortex showed the gene expression changes that predict various movement de cits and ASD-like behaviors. Consistently, adult Q84Pfs-Het mice exhibited the corresponding behavioral characteristics. This close matching of gene and behavior signatures pointed to the genes whose misexpression leads to movement de cits and ASD-like behaviors in Q84Pfs-Het mice. Fmr1, Nalcn, Il1rapl1, and Slc16a2, which are aberrantly upregulated in Q84Pfs-Het brains, are likely to be involved in intellectual disability, social impairment, and autism-like behaviors, given that these genes are associated with neurodevelopmental disorders with the described signature behaviors [39][40][41][42] . A subset of FS symptoms, such as dystonia, myoclonus, and ataxia, may be attributed to the downregulation of PLEKHG2, DST, ARX, AAAS, KCTD17, PRDM8, and CRTC1 in FS brains as these genes are downregulated in Q84Pfs-Het brains and play a role in these movement and muscle coordination [43][44][45][46][47][48][49] . The remarkable similarity of dysregulated genes between Q84Pfs-Het and HD mouse models suggests that the shared gene expression changes are responsible for common symptoms between FS and HD, such as involuntary jerky and writhing movement. Our studies also revealed that the genes involved in synaptic vesicle exocytosis, endocytosis, and synapse organization and membrane assembly are dysregulated in Q84Pfs-Het cortex, strongly implying synapse de ciency as a mechanism driving FS symptoms, such as prolonged behavior arrests and cognition impairments. Of note, the prolonged behavior arrest similar to the behavior of Q84Pfs-Het mice (Fig. 7f) has been described for seizures, like focal behavior arrest seizures (aka, focal impaired awareness seizures, hypokinetic seizure, localized seizure with behavior arrest, or complex partial seizures) [50][51][52][53] . As this type of seizure has been observed in FS 54 , it will be interesting to perform electroencephalogram (EEG) recording paired with behavioral assessment in Q84Pfs-Het mice and examine if behavior arrest is correlated with speci c types of EEG activities. Given the symptomatic similarity between FS and other neurological disorders, the pathogenic mechanism studies in Q84Pfs-Het mice will provide signi cant insights into a broader range of neurodevelopmental disorders, including ASD.
Q84Pfs-Het brains faithfully captured the structural de cits of the cortex, a brain region mainly derived from the dorsal telencephalon, in FS patients. The future investigation can determine if the ventral telencephalon-derived brain de cits, such as the striatum, match between Q84Pfs-Het mice and FS patients and if the structural changes and gene dysregulation in the striatum correlate with the behavior of Q84Pfs-Het mice. As the striatum is strongly involved in motor control and compulsive and anxiety behaviors and is one of the main areas affected in HD and Parkinson's diseases 32 , the striatal circuits may play a role in a subset of behavioral phenotypes of Q84Pfs-Het mice.
The myelination de cits in Q84Pfs-Het mice suggest a key role of Foxg1 in oligodendrocyte development.
In Q84Pfs-Het brains, Olig2 + oligodendrocyte lineage cells increase, despite disrupted myelination (Fig. 6), suggesting Foxg1 functions at multiple points of oligodendrocyte development. Overexpression of Foxg1 inhibits the generation of OPC from neural progenitors in vitro 55 and Foxg1 deletion speci cally in adult neural stem cells decreases OPC proliferation and facilitates remyelination in Cuprizone-induced demyelination conditions 56 . Combined with these reports, our results raise the possibility that Foxg1 haploinsu ciency triggers the overproduction of Olig2 + OPCs from neural progenitors but inhibits differentiation and maturation of oligodendrocytes, thereby resulting in accumulated OPCs and reduced myelination. Future studies are needed to test this possibility. The myelination de ciency likely contributes to various FS symptoms, such as dystonia. Further investigation into dysmyelination of Q84Pfs-Het brains will uncover not only pathogenic mechanisms of delayed myelination, a hallmark of FS brains, but also key regulatory steps in OL development. Considering that proper myelination of nerves is essential for movement and diverse brain functions and that myelination primarily begins after birth and continues into young adulthood in humans 57 , dysmyelination could be an excellent therapeutic target for FS.
Foxg1 plays cell type-speci c and developmental timing-dependent roles in forebrain development, at least via the two mechanisms. First, the dynamic expression of Foxg1 plays a key role. The downregulation of Foxg1 during neurogenesis is crucial for the cell cycle exit and differentiation of neural progenitors 8, 9,58,59 . After neuronal differentiation, Foxg1 expression is induced in postmitotic neurons, in which Foxg1 promotes neuronal morphological transition and migration 10,11 . Second, Foxg1 partners with other transcription factors in a cell context-speci c manner in selecting target genes. Our motif analysis of Foxg1-bound genomic loci of DEGs identi ed E-box, the binding sites for Foxg1's partner transcription factors Zbtb18 and Neurod1 11,28 , along with the Fox motif. Our data integrating RNAseq and ChIPseq datasets (Fig. 3d) also suggest that NF1 collaborates with Foxg1 in Foxg1-mediated transcriptional regulation in the cortex. Notably, NF1 is a family of closely related transcription factors, NF1A, NF1B, NF1C and NF1X, and in particular NFIA and NFIB have been found to be involved in haploinsu ciency syndromes that share some features with the FS, such as corpus callosum agenesis and intellectual disability 60-62 .
Collectively, our study established Q84Pfs-Het mice as the FS mouse model that will expedite our understanding of the pathophysiology of FS and serve as an essential platform for therapeutic development. Importantly, our study uncovered new cellular and behavioral phenotypes that are linked to Foxg1-dependent gene expression changes in Q84Pfs-Het mice, providing key insights into Foxg1 function and FS etiology.

FOXG1 Syndrome Research Registry & Ciitizen Databank
The International FOXG1 Syndrome Research Registry is an online platform that collects information from caregivers of children and adults with FS, including results of genetic testing, clinical phenotypes, and developmental outcomes [https://foxg1.beneu t.com/login.php].
Ciitizen®, a wholly owned subsidiary of Invitae Corporation, is a patient-centric real world data platform that transforms medical records into structured datasets that describe clinical phenotypes and medical interventions [https://www.medrxiv.org/content/10.1101/2023.03.02.23286645v1]. In partnership with the FOXG1 Research Foundation, > 120 individuals with a FOXG1 variant have enrolled in Ciitizen. In brief, the platform supports medical records that are collected on behalf of participants through the Health Insurance Portability and Accountability Act (HIPAA). Following record collection, each medical document is systematically reviewed for relevant data variables, including presenting diagnoses and diagnostic imaging results. Genetic variants were extracted from clinical genetic test reports; loss-of-function variants in FS, including frameshift variants, are known to be pathogenic. To support harmonization of disparate data sources, extracted phenotypic data were mapped to SNOMED CT, an internationally recognized terminology (US Edition, version 2022_03_01) [www.snomed.org]. The resulting data elements were stored securely in HIPAA-compliant, controlled access, indexed database.
Caregivers and/or legal guardians of study participants provided broad consent to share de-identi ed data for research. The generation and subsequent analysis of participant data used in this study received determinations of exemption through a central institutional review board, Pearl IRB. Reasonable requests for data access can be made to cii-research@invitae.com.

Animals
Q84Pfs mutant mice were generated with the CRISPR/Cas9 gene editing system using guide RNA (CCCAGCAGCCGACGACGACA) and single stranded oligonucleotides (ssODN) containing the mutation c.250dupC to introduce the Q84Pfs allele and the restriction enzyme site (HindIII) for genotyping (T CCA CCG CCG CCC CAG CAG CAG CAG CAG CAG CCG CCC CCG GCC CCG CAG CCC CCG C CAG GCG CGC GGC GCC CCA GCA GCC GAC GAC GAC AAG CTT CCC CAG CCG CTC CTG CTC CCG CCC TCC ACC).
Potential founder mice were genotyped by Sanger sequencing and restriction enzyme digestion for PCR product surrounding the target site. The targeted founder mice from CRISPR/Cas9 microinjection were backcrossed to C57BL/6J mice for at least ve generations to eliminate potential off-target mutations before any experiments.

Behavioral tests
Mice were maintained on a 12 h light/dark cycle, and all behavioral tests were performed during the light phase. All animals were allowed to habituate in the behavior testing room (with attenuated light and sound) for at least 30 min before the tests. All arenas and objects used in behavioral analysis were cleaned with Alconox and then distilled water between test sessions.
Open eld test: The open eld test was utilized to examine locomotor activity and to monitor anxiety-like behaviors as well as repetitive behaviors such as jumping and self-grooming. Mice were placed into a Plexiglas chamber with transparent walls and white oor (40cm W X 40cm D X 30cm H) and then allowed to move freely for 60min. All activity was monitored by SuperFlex Open Field system (Omnitech Electronics) and the total moved distance and the time spent in the center (20 cm X 20 cm within the center of the oor) were collected and analyzed by Fusion software (AccuScan Instruments Incorporated). Grooming: Grooming was assessed manually using video recordings from the open eld test. Each mouse was monitored for 60 minutes. For a grooming event to count, the mouse had to groom for at least 2 seconds. If the mouse stopped grooming, but resumed in 5 seconds or less, it was counted as the same grooming event. If the mouse moved to a new location in the cage and continued grooming, even if it resumed in 5 seconds or less, it was counted as a new grooming event. Prolonged behavior arrest: During the open eld test, prolonged behavior arrest was marked whenever a mouse remained sitting in one location in the cage, largely unmoving, for at least 3 minutes. During this time, the mice did not appear to perform any purposeful movement, such as grooming, rearing, digging, or other exploratory behavior. Wire hanging test: Mice were placed inside a basket with wire grids on the bottom, and the basket was slowly inverted to allow the animal to hang from the grid. All sessions were video-recorded, and the hanging time was analyzed manually. The hanging time was de ned from the time the basket was inverted to the time that the mouse fell off the wire grid. Mice were tested on three consecutive trials with at least 5 min intervals between trials. All trials ended upon reaching 1000 seconds. Marble burying: Mice were gently placed in the periphery of testing cages (arena size: 30 cm × 20 cm × 13 cm, corncob bedding depth: 5 cm) containing 18 glass marbles (evenly spaced with three rows of six marbles on the top of bedding), and then the cages were covered with the transparent top (height 13 cm). Mice were carefully removed from the cages, and the number of marbles buried was recorded at the end of the 30 min exploration period. Marbles were counted as buried when they have covered at least 2/3 depth with bedding.

Brain preparation and immunohistochemical analysis
Dissected brains were xed in 4% paraformaldehyde in phosphate-buffered saline (PBS) at 4°C overnight, equilibrated in 30% sucrose, and embedded in Tissue Freezing Medium™ (Electron Microscopy Sciences) for frozen sectioning for 18 µm brain sections using cryostat (CM1950, Leica). Position matched sections were then stained following the standard immunohistochemical method. Sections were incubated with the primary antibody and the secondary antibodies conjugated to uorophores (Jackson ImmunoResearch), and then counterstained with DAPI to reveal nuclei. Images were collected through DM6 uorescence microscope (analyzed by Leica Software LasX) or Eclipse Ti2 confocal microscope

Nissl staining
Slides with cryosectioned frozen sections were incubated in ethanol/chloroform mixture, mixed 1:1 volumetrically, at RT overnight. Incubated slides went through rehydration steps in three steps, 100% ethanol, 95% ethanol and then distilled water for 10 minutes each. After rehydration, slides were stained in warm 0.2% cresyl violet solution for 12 min, followed by brief washing and then incubating in 95% ethanol, 100% ethanol and xylene for 20 minutes each for differentiation, dehydration, and clearing, respectively. Coverslip was mounted with Permount™ Mounting Medium (Fisher Scienti c). Images from 18µm thick stained specimen were acquired by Axio Scan.Z1 (Zeiss) under bright eld with 10X objective.

RNA-seq data analyses
FastQC was used to check the quality of the raw RNA-seq data. Clean reads were aligned to the mouse reference genome (mm10) using STAR (v2.6.1d) 63 with default parameters and the gene expression levels were quanti ed using RSEM (v1.3.3) 64 . The DEGs were identi ed using EBseq 65 based on the expected counts. The absolute value of log2 fold change of 0.3 and false discovery rate (FDR) of 0.05 were set as criteria to determine if the expression level of a gene was signi cantly changed. In this manner, 222 DEGs including 118 upregulated and 104 downregulated DEGs were identi ed in total. We subsequently carried out GSEA using DAVID and Enrichr on the upregulated DEGs and the downregulated DEGs, separately.

Statistical analyses
The data were analyzed using GraphPad 9.5 (Prism, San Diego, CA). Student T-test were run using genotype as between-subject variables. A repeated-measures T-test was used to analyze activity in Olig2 positive cell counting, cortex thickness measuring, Pax6 positive area measuring, layer size measuring and DG length/angle measuring. Statistical signi cance was assessed using the Mann-Whitney or ANOVA test, as indicated in the gure legends, and was represented on the graphs as: * * * * p < 0.0001; * * * p < 0.001; * * p < 0.01; * p < 0.05; ns, not signi cant. All bars and error bars were speci ed in the gure legends.    The dysregulated genes and pathways in Q84Pfs-Het cortex at P1. a, The differentially expressed genes (DEGs) in P1 cortices of Q84Pfs-Het mice as shown by the volcano plot. b, Cortical upper layer and interneuron genes were downregulated, whereas the deep layer genes were upregulated. c, The integration of DEGs of Q84Pfs-Het cortex and Foxg1 ChIP-seq data revealed the fraction of the DEGs that directly recruits Foxg1, as marked by orange. d, The motif analysis of Foxg1 ChIP-seq peaks associated with up-a, d), 50mm (higher magni cation images in a, d), or 100mm (f). Thickness in (b,c) was measured in three independent areas per section and three sections were analyzed in each mouse (n = 3 mice per group). The error bars (b,c,e) represent the standard deviation of the mean. *, p<0.05, ****, p<0.0001 in unpaired two-tailed test. Only the representative images are shown.

Figure 5
Q84Pfs-Het mice showed a reduction of cortical thickness, upper cortical layer, and cortical interneurons, and axon misprojections in the cortex. a-i, The immunostaining analyses of Q84Pfs-Het and WT brains at areas per the total cortex area (d; in two independent areas per mouse, n = 3 mice per group), and the relative Mbp intensity in the cortex (e; in three independent areas per mouse, n = 3 mice per group). Scale bars, 1mm (lower magni cation images in a,c), 100mm (higher magni cation images in a), and 200mm (higher magni cation images in c). b,d,e, The error bars represent the standard deviation of the mean. **, p<0.01, ***, p<0.001, ****, p<0.0001 in unpaired two-tailed test. The representative images are shown. Figure 7