Abstract

   FOXG1 syndrome is a developmental encephalopathy caused by FOXG1
   (Forkhead box G1) mutations, resulting in high phenotypic variability.
   However, the upstream transcriptional regulation of Foxg1 expression
   remains unclear. This report demonstrates that both deficiency and
   overexpression of Men1 (protein: menin, a pathogenic gene of MEN1
   syndrome known as multiple endocrine neoplasia type 1) lead to
   autism‐like behaviors, such as social defects, increased repetitive
   behaviors, and cognitive impairments. Multifaceted transcriptome
   analyses revealed that Foxg1 signaling is predominantly altered in Men1
   deficiency mice, through its regulation of the Alpha Thalassemia/Mental
   Retardation Syndrome X‐Linked (Atrx) factor. Atrx recruits menin to
   bind to the transcriptional start region of Foxg1 and mediates the
   regulation of Foxg1 expression by H3K4me3 (Trimethylation of histone H3
   lysine 4) modification. The deficits observed in menin deficient mice
   are rescued by the over‐expression of Foxg1, leading to normalized
   spine growth and restoration of hippocampal synaptic plasticity. These
   findings suggest that menin may have a putative role in the maintenance
   of Foxg1 expression, highlighting menin signaling as a potential
   therapeutic target for Foxg1‐related encephalopathy.

   Keywords: ASD‐like behaviors, Atrx, FOXG1 syndrome, menin
     __________________________________________________________________

   FOXG1 syndrome is a developmental encephalopathy resulting from the
   dysregulation of FOXG1. Here, it is reported that both deficiency and
   overexpression of Men1 result in autism‐like behaviors by
   transcriptional regulation of Foxg1 which is mediated by Atrx‐mediated
   H3K4me3 modification. This work reveals a putative role of menin in
   maintaining Foxg1 expression which will benefit the Foxg1‐related
   encephalopathy therapeutic intervention development.

   graphic file with name ADVS-11-2307953-g005.jpg

1. Introduction

   FOXG1 syndrome is a major developmental encephalopathy that belongs to
   the autism spectrum disorder (ASD). It exhibits both classical and
   atypical ASD phenotypes, including severe microcephaly (postnatal or
   congenital) and neurodevelopmental delay.^[ [56]^1 ^] FOXG1 is a
   crucial member of the Forkhead gene family and plays a vital role in
   brain development. FOXG1 haploinsufficiency in humans is associated
   with significant differences in brain size and impaired intellectual
   development in early childhood, while homozygous mutations are
   typically fatal.^[ [57]^2 ^] Despite the crucial roles of FOXG1 in the
   nervous system, limited knowledge exists regarding the upstream
   regulation mechanisms governing the FOXG1 gene. Interestingly, patients
   with micro‐fragment deletion of 11q13.1, where the MEN1 gene is
   located, exhibit autism‐like behaviors.^[ [58]^3 ^] This clinical
   observation implies a strong association between menin and
   neurodevelopmental disorders. We previously found that neuronal
   specific deletion of Men1 leads to neuronal impairments in dendritic
   branching, spine density and synaptic plasticity, as well as ASD‐like
   behaviors in mice (Men1‐CcKO mouse). To investigate the molecular
   mechanism mediated by menin in the nervous system, we performed
   menin‐ChIP‐Seq, menin‐CUT&Tag and RNA‐Seq to screen and identify the
   downstream genes regulated by menin in the nervous system. We found
   that the transcription of Foxg1, a pathogenic gene associated with
   FOXG1 syndrome,^[ [59]^1 , [60]^4 ^] significantly decreases under Men1
   deletion conditions. Menin typically interacts with other transcription
   factors to coordinate gene transcription. Through a menin‐binding
   protein interactome proteomics assay, we successfully identified Atrx
   (alpha‐thalassemia mental retardation X‐linked) as a menin‐associated
   protein. The Atrx protein is a member of SWI/SNF‐like chromatin
   remodeling protein, and its mutations are associated with X‐linked
   syndromes that exhibit cognitive disabilities.^[ [61]^5 ^] Menin
   collaborates with Atrx to regulate the transcription of Foxg1. The
   restoration of Foxg1 expression in Men1‐CcKO mouse brain mitigated the
   neuronal impairments and ASD‐like behaviors. Our findings demonstrate
   the crucial role of menin in neuronal development and highlight its
   significance for FOXG1 syndrome, underscoring the need for further
   investigation to facilitate therapeutic development.

   Epigenetic regulation is tightly associated with neurodevelopmental
   disorders such as autism disorder spectrum (ASD).^[ [62]^6 ^] Histone
   lysine methyltransferases play an important role in histone methylation
   and are commonly referred to as “writers” for catalyzing histone
   methylation.^[ [63]^7 ^] Mixed lineage leukemia proteins‐1 (MLL1) and
   Mixed lineage leukemia proteins‐2 (MLL2) are crucial members of histone
   methyltransferase family. MLL1/2 typically interact with multiple
   endocrine neoplasia type 1 (menin, gene for Men1) to increase the
   expression levels of the downstream target genes such as HOX, MEIS1,
   and EZH2.^[ [64]^8 ^] Histone methylation, mediated by MLL1 and MLL2,
   plays pivotal roles in the development of nervous system. Mice with
   neuronal‐specific Mll1 or Mll2 deletions exhibit reduced synaptic
   plasticity and impaired working memory formation by regulating
   important neuronal development genes.^[ [65]^9 ^] However, the function
   of menin in the central nervous system are largely unknown.

   Menin is encoded by the multiple endocrine neoplasia type 1 (MEN1) in
   humans (Men1 in mice). It serves as a crucial transcriptional activator
   for MLL1 and MLL2, thereby exerting significant influence on histone
   methylation modification.^[ [66]^10 ^] MEN1 syndrome is a autosomal
   dominant disorder characterized by multiple tumorigenesis in endocrine
   organs, caused by loss‐of‐function mutations of MEN1.^[ [67]^11 ^]
   Genetic deletion of Men1 induces embryonic lethality, including
   impairments in the nervous system development, by regulating a large
   number of developmental genes.^[ [68]^12 ^] However, the underlying
   mechanism of menin in nerve development is not well understood. Through
   multifaceted transcriptome analyses specifically in neurons, we found
   that the transactivation of Foxg1 is dramatically decreased in the
   absence of Men1, and which is attributed to non‐menin‐Mll1/2 mediated
   H3K4me3 modification.

2. Results

2.1. Menin Binds to the Foxg1 Promoter Locus and Promotes its Transcription
and Expression

   To explore the molecular pathways mediated by menin, we conducted
   RNA‐Seq and ChIP‐Seq analyses of menin from hippocampal tissues or
   primary neurons with Men1 deletion. The Men1 ^(F/F) mice (Control) were
   bred with CaMK2α‐Cre or Nestin‐Cre mice to generate CaMK2α‐Cre;Men1
   ^(F/F) conditional knockout (Men1‐CcKO) mice or Nestin‐Cre;Men1 ^(F/F)
   conditional knockout (Men1‐NcKO) mice, respectively.^[ [69]^13 ^] RNA
   sequencing (RNA‐seq) was performed on hippocampus tissues (from
   Men1‐CcKO mice and their littermate controls) and primary neurons (from
   Men1‐NcKO mice and their littermate controls) (Figure [70]1A,B). A
   total of 2402 and 527 differentially expressed genes (DEGs)
   (|log2FC|>0.263, p < 0.05) were identified in primary neurons and
   hippocampus, respectively, when comparing the control and Men1‐knockout
   groups (Figure [71]1C,D; and Data [72]S1 and [73]S2, Supporting
   Information). The DEGs found in primary neurons and hippocampus were
   subjected to overlapping analysis, revealing that 66 DEGs changed in
   the same way, with 24 genes being down‐regulated and 42 genes being
   up‐regulated (Figure [74]S1A, Supporting Information) (Data [75]S3,
   Supporting Information). The subsequent analysis of these 66 DEGs using
   Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG)
   revealed that menin‐mediated genes are closely related to
   neurodevelopmental and neuropeptide signaling pathways. (Figure
   [76]S1B,C, Supporting Information) (boxed in red). Using String v10 and
   Cytoscape analysis, we found that the protein‐protein interaction (PPI)
   network of these core genes clusters in neurodevelopment, which
   includes Foxg1 (Figure [77]S1D, Supporting Information).

Figure 1.

   Figure 1
   [78]Open in a new tab

   Menin binds to the promoter region of Foxg1 and regulates the
   expression of Foxg1. A) Schematic of the experimental strategy used to
   identify the direct target of menin from multiple‐omics data. The
   RNA‐Seq analysis was performed on three independent biological
   replicates of mouse hippocampus and four independent biological
   replicates of primary neurons, while ChIP‐Seq or CUT&Tag experiments
   were limited to a single trial. B) Representative genotyping results
   are shown for Men1 WT (Men1^(+/+) ), heterozygous (Men1^(F/+) ),
   homozygous (Men1^(F/F) ), as well as two Cre transgenic mice with
   expected PCR product size. C) Volcano plot showing differential
   expressed genes between Men1‐KO neurons and Control neurons (p < 0.05
   and |Log[2]Fold change|>0.2630), There are 1348 up‐regulated genes and
   1054 down‐regulated genes in Men1‐KO neurons. D) Another Volcano plot
   demonstrates differential expressed genes between CcKO mouse
   hippocampus and Control mouse hippocampus (p < 0.05 and |Log[2]Fold
   change|>0.2630), There are 166 up‐regulated genes and 361
   down‐regulated genes in the hippocampus of Men1‐CcKO mice. E) Coverage
   profiles show the signal of menin and H3K4me3 in DIV10 primary neurons
   from ChIP‐Seq data, along with a heatmap displaying menin‐associated
   peaks centered at TSS, similar to the H3K4me3. And coverage profiles
   depict the signal of menin in Men1‐KO neurons compared to controls
   using CUT&Tag dataset. F) Venn diagram showing the overlap between
   Up‐regulated genes (vs. the control group) in Men1‐KO neurons and the
   hippocampus of Men1‐CcKO mice (n = 24 DEGs), Down‐regulated genes
   (vs. the control group) in Men1‐KO neurons and the hippocampus of
   Men1‐CcKO mice (n = 42 DEGs), respectively. Foxg1 was further
   identified by intersecting the DEGs from RNA‐seq with the genes bound
   by menin in ChIP‐Seq/CUT&Tag analyses. G) The expression changes of
   Foxg1 were examined in mouse hippocampus and primary neurons. FPKM
   value of Foxg1 from different samples were normalized to the control
   group (n = 3 hippocampal samples per group and n = 4 primary neuronal
   samples per group). H) The IGV diagram showing menin binding position
   on the Foxg1 gene from ChIP‐Seq and CUT&Tag data performed in primary
   neurons. A significant number of binding signals on the Foxg1 TSS by
   menin are observed. And six ChIP‐PCR primer pairs were designed within
   a ≈2500 bps fragment (near the TSS) to validate the data (Boxed with
   the blue rectangle). (All mice were males. Data is represented by mean
   ±SEM. ****p < 0.0001 and *p < 0.05 indicate significance between the
   two indicated groups. G : unpaired t‐test.).

   Menin, an essential component of histone methyltransferases complex, is
   capable of regulating H3K4me3 modification to transactivate gene
   expression.^[ [79]^14 ^] However, the menin‐H3K4me3 mediated genes
   transcription has been little studied in the nervous system. To
   investigate this, menin and H3K4me3 ChIP‐seq analyses were conducted in
   primary neurons (Figure [80]1A). The binding signal of menin and
   H3K4me3 was notably enriched proximal to transcription start sites
   (TSS). Furthermore, CUT&Tag is an improved version of ChIP‐seq that
   produces higher quality results with lower noise. Consequently, it was
   selected for further analysis of menin‐binding peaks and genes.^[
   [81]^15 ^] By conducing CUT&Tag analysis, we further found that the
   menin binding signal was reduced near the TSS in Men1‐KO neurons
   (Figure [82]1E).

   A total of 2449 menin‐associated peaks were detected in the ChIP‐seq
   analysis, with 1287 peaks are distributed in the vicinity of 1000
   genes. In the CUT&Tag analysis, a set of 3143 menin‐binding genes were
   identified from a pool of 3330 unique peaks (Data [83]S4 and [84]S5,
   Supporting Information). The analyses of GO and KEGG pathways for these
   ChIP binding genes revealed a high degree of clustering in neuronal
   development and maturation (Figure [85]S1E, Supporting Information). As
   previously reported,^[ [86]^16 ^] menin predominantly occupies
   positions in intronic and intergenic regions (menin peaks: n = 1287,
   43.5% in introns and 44.5% in intergenic regions) (Figure [87]S1F,
   Supporting Information). Neuronal menin DNA binding motifs are also
   enriched, particularly the one characterized as “CAGCTG” (Figure
   [88]S1G, Supporting Information). We performed integrated analysis by
   combining and overlapping ChIP‐seq, CUT&Tag and two RNA‐seq datasets.
   In the intersection analyses, only one gene, Foxg1, was observed
   significantly changed (Figure [89]1F). The decreasing of Foxg1 was
   confirmed in menin deleted hippocampus and primary neurons
   (Figure [90]1G). The DNA binding peaks identified from the reading
   density of the ChIP‐Seq and CUT&Tag indicate that menin binds to the
   Foxg1 gene, particularly in proximity to its transcription start site,
   as shown in CUT&Tag. Six primer pairs were designed based on the
   sequence near the transcriptional start site (TSS) for following
   validation (Figure [91]1H, boxed with the blue rectangle). Figure
   [92]2A–C shows that all primers can bind to the Foxg1 promoter region,
   as indicated by ChIP‐PCR experiments and these binding significantly
   decreased in Men1 deleted primary culture neurons. Gel electrophoresis
   further demonstrated that menin is recruited to the Foxg1 promoter in
   primary neurons and hippocampus tissues (Figure [93]S2A,B, Supporting
   Information). A Foxg1 promoter of approximately 2500 bps (Figure
   [94]S2C, Supporting Information) was synthesized and fused to a
   luciferase reporter. Subsequently, this plasmid was then transfected
   along with control vector into both WT and Men1 knockout primary
   neurons. The Men1 deletion significantly decreased the activity of the
   Foxg1 promoter‐driven luciferase in primary neurons (Figure [95]S2D,
   Supporting Information).

Figure 2.

   Figure 2
   [96]Open in a new tab

   Menin transcriptionally regulates the expression of Foxg1. A) Schematic
   of the experimental strategy used to be verify the menin‐Foxg1 axis in
   vivo and in vitro. B) Six independent ChIP assays were performed in
   primary neurons to confirm menin binding to the Foxg1 promoter (n = 4
   independent experiments). C) The abundance of menin‐binding Foxg1
   promoter was compared between Men1‐CcKO mouse hippocampus and control.
   (n = 4 independent experiments). D) Men1 and Foxg1 mRNA levels were
   detected in Men1‐CcKO of mouse brain, including cortex and hippocampus
   (The cortical sample size was 12 in the control group and 10 in the
   Men1‐CcKO group; The hippocampal sample size was 12 in the control
   group and 12 in the Men1‐CcKO group). E) Menin and Foxg1 protein levels
   were measured by western blotting in Men1‐CcKO of mouse brain, GAPDH as
   internal control (The cortical sample size was 8 from 4 mice in the
   Control group and 8 from 4 mice in the Men1‐CcKO group; The hippocampal
   sample size was 8 from 4 mice in the control group and 7 from 4 mice in
   the Men1‐CcKO group). F) The diagram represents the statistics of the
   corresponding gray values in (E). G) Representative cortical (above)
   and hippocampal (below) brain sections from Control and Men1‐CcKO mice
   stained with Foxg1 (Red) antibody. Section was counterstained with DAPI
   (blue). Scale bar, 100 µm. (Control group: n = 6 slices from 3 mice;
   Men1‐CcKO group: n = 6 slices from 3 mice. Data was collected from
   cortex and hippocampus specimens). H) The corresponding statistical
   results for (G) are presented. I) Diagram of CaMK2α‐Cre mice were
   injected with AAV‐EF1α‐DIO‐MEN1‐GFP (MEN1 AAV) into hippocampus to
   construct a MEN1 overexpression model. J) Representative
   immunofluorescence image showing MEN1 AAV and Control AAV. The red
   channel is menin, while green indicated GFP. Scale bar, 200 µm. High
   magnification insets are provided at the top of each image. (Control
   AAV: n = 3 slices from 3 mice and MEN1 AAV: n = 3 slices from 3 mice).
   K) AAV‐mediated human MEN1 overexpression in the hippocampus
   significantly enhanced the enrichment of menin binding to the Foxg1
   promoter (n = 4 independent experiments). L) Men1 and Foxg1 mRNA levels
   were detected in MEN1‐overexpressing hippocampus. The mRNA level of
   human‐derived MEN1, but not mice‐derived Men1, showed a significant
   increase (Control AAV group: n = 10 samples from 10 mice; MEN1 AAV
   group: n = 12 sample from 12 mice). M) Menin and Foxg1 protein levels
   were measured by western blotting in MEN1 overexpressing hippocampus,
   GAPDH as an internal control (Control group: n = 4 samples from 4 mice;
   Men1‐CcKO group: n = 4 samples from 4 mice). N) The diagram represents
   the statistics of the corresponding gray values in (M). O)
   Representative images of Foxg1 immunofluorescence were obtained from
   the MEN1 AAV group and Control AAV group. The red channel represents
   Foxg1 and green is GFP. Scale bar, 200 µm (Control AAV group: n = 6
   slices from 3 mice; MEN1 AAV group: n = 6 slices from 3 mice.). P) The
   corresponding statistical results for (O) are shown. (All mice were
   males. Data is represented by mean ± SEM. ****p < 0.0001, ***p < 0.001,
   **p < 0.01, and *p < 0.05 indicate significance between the two
   indicated groups. B, C, D, F, K, L, N : Two‐way ANOVA with Tukey post
   hoc test, H, P : unpaired t‐test.).

2.2. Menin Transcriptionally Regulates the Expression of Foxg1

   The above data suggests that menin may directly regulate the expression
   of Foxg1. By RT‐PCR, western blotting and immunostaining, we found a
   significantly decreased in both mRNA and protein levels of Foxg1 in the
   hippocampus and cortex of Men1‐CcKO mouse brains compared to the
   control (Figure [97]2D–H). To further investigate the regulation of
   menin on Foxg1, menin overexpression was conducted by injecting
   AAV‐EF1α‐DIO‐MEN1 (MEN1 AAV) into the hippocampus of CaMK2α‐Cre mice
   (Figure [98]2I). The expression of menin in the CA1 region was
   confirmed by immunostaining (Figure [99]2J). As anticipated, the
   binding of menin on the Foxg1 promoter region was dramatically
   increased in the menin‐overexpressed hippocampus compared to control,
   as measured by menin‐ChIP‐PCR (Figure [100]2K). Concomitant with these
   findings, the mRNA and protein levels of Foxg1 were also increased by
   menin overexpression measured by RT‐PCR (Figure [101]2L), western
   blotting (Figure [102]2M,N) and immunostaining (Figure [103]2O,P).

   To investigate whether the transcription regulation of menin on Foxg1
   is dependent on H3K4me3 modification, we performed an H3K4me3 ChIP
   assay in primary neurons and hippocampus tissues. Notably, two distinct
   primer pairs (PP1&PP2) targeting the Foxg1 promoter locus both indicate
   the robust H3K4me3 at the Foxg1 promoter in WT neurons and control
   hippocampus. Conversely, H3K4me3 occupancy at the Foxg1 promoter locus
   dramatically decreased in Men1‐knockout neurons and hippocampus (Figure
   [104]S2E,F, Supporting Information). Vice versa, extra menin expression
   increased a higher density of H3K4me3 on the Foxg1 promoter was
   observed (Figure [105]S2G, Supporting Information). Taken together,
   these results indicate that Foxg1, a causative gene of Foxg1 syndrome,
   is regulated through modulation of histone H3 lysine 4 trimethylation
   by menin.

2.3. Deletion or Overexpression of Menin in Excitatory Neurons both Induces
ASD‐like Behaviors in Mice

   Patients with FOXG1 syndrome present with mental retardation, lack of
   social ability, and intellectual disabilities.^[ [106]^17 ^] Both male
   and female Men1‐CcKO mice as well as control mice were subjected to
   behavior tests. Notably, these mice showed social deficits, increased
   stereotyped repetitive behaviors and cognitive deficits (Figure
   [107]3A–C; Figure [108]S3A, Supporting Information). There were no
   differences in locomotor activity were observed between the two groups
   based on rotarod assays (Figure [109]S3B, Supporting Information).
   Furthermore, Men1‐deficient mice also exhibited impairments in hind
   limb clasping, as well as difficulties in the nest building and
   marble‐burying behaviors (Figure [110]S3C–E, Supporting Information).

Figure 3.

   Figure 3
   [111]Open in a new tab

   Excitatory neuron‐specific deletion or overexpression of menin both
   induce mice exhibiting ASD‐like behaviors. A) Establishment of the
   Men1‐CcKO transgenic model. B) Grooming tests were used to assess
   compulsive, rigid, repetitive behaviors in the four groups of mice.
   There were no significant differences in grooming time between males
   and females (Control‐male group: n = 12; Control‐female group: n = 12;
   Men1‐CcKO‐male group: n = 10; Men1‐CcKO‐female group: n = 10). C)
   Three‐chamber social tests were employed to analyze the social behavior
   in mice, including adaptive exploration stage (Left), social novelty
   tests (Middle) and social memory tests (Right) (Control‐male group:
   n = 12; Control‐female group: n = 12; Men1‐CcKO‐male group: n = 10;
   Men1‐CcKO‐female group: n = 10). D) The lateral ventricles of neonatal
   CaMK2α‐Cre male mice were intracerebroventricularly injected with
   Control AAV or MEN1 AAV on postnatal day 0 (P0) to generate a
   neuron‐specific MEN1‐overexpressing model. E) Statistics of grooming
   time within 10 min under the free exploration state (Control AAV group:
   n = 15; MEN1 AAV group: n = 14, all mice were males). F) Three‐chamber
   social experiments were employed to analyze the social behavior in the
   two groups of mice, including adaptive exploration stage (left), social
   novelty test (middle), and social memory test (right) (Control AAV
   group: n = 15; MEN1 AAV group: n = 14, all mice were males). (Data is
   represented by mean ±SEM. *p < 0.05 and **p < 0.01 indicate
   significance between the two indicated groups. B : one‐way ANOVA with
   Tukey post hoc test. C, F : two‐way ANOVA with Tukey post hoc test. E :
   unpaired t‐test.).

   Another notable feature of FOXG1 syndrome is its bidirectional dose
   effect on the causative gene. Both deletion and overexpression can lead
   to the emergence of ASD‐like behaviors in individuals.^[ [112]^18 ^] To
   investigate this further, we overexpressed menin by injecting MEN1 AAV
   into the lateral ventricle of neonatal CaMK2α‐Cre mice, as depicted in
   Figure [113]2I and Figure [114]S3J (Supporting Information)
   (Figure [115]3D). Overexpression of menin in excitatory neurons
   significantly induces ASD‐like behaviors in mice, accompanied by
   cognitive impairments assessed through grooming, social interaction,
   water maze and fear condition tests (Figure [116]3E,F; Figure
   [117]S3F,G, Supporting Information), without affecting locomotor
   activity (Figure [118]S3H,I, Supporting Information). These data
   indicate that enhanced the expression of excitatory neuronal menin is
   also associated with impairments in ASD‐like behaviors.

   Foxg1 is also expressed in inhibitory neurons and plays important
   function prompted us to investigate the involvement of menin‐Foxg1
   signaling pathway in interneurons for regulating ASD‐like behaviors.^[
   [119]^19 ^] The conditional Men1 interneuron‐specific deleted mice
   (Men1‐DcKO) were generated by crossing Men1 ^F/F mice (Control) with
   Dlx5/6α‐Cre‐IRES‐GFP mice (Figure [120]S3K, Supporting Information).
   Mice with Men1 deficiency in inhibitory neurons (Men1‐DcKO) showed no
   significant changes in these phenotypes when compared to wild‐type
   animals (Figure [121]S3L–N, Supporting Information). These findings
   suggest that the deficiency of Men1 in excitatory neurons, rather than
   inhibitory neurons, is associated with ASD‐like behaviors.

2.4. Atrx Knockdown Causes Autistic Behaviors and Contributes to the
Transcription Regulation of Foxg1 by Menin‐Mediated Histone H3 Lysine 4
Trimethylation

   Since MLLs maintain gene expression through interaction with menin,^[
   [122]^10a ^] we examined whether menin‐MLL is involved in the
   regulation of Foxg1 transcription. Primary neurons were subjected to
   treatment with menin‐MLL inhibitors MI‐2 and MI‐3, which specifically
   target the interaction between menin and MLLs.^[ [123]^10b ^] However,
   we observed no significant changes in Foxg1 expression following
   treatment with either MI‐2 or MI‐3 (Figure [124]S4A–D, Supporting
   Information). As the positive control, the mRNA level of Stab2 was
   decreased (Figure [125]S4E, Supporting Information).^[ [126]^9a ^] This
   data suggests that other factors are required for menin to regulate the
   transcription of Foxg1.

   To investigate the potential interacting proteins of menin associated
   with the transcription regulation of Foxg1, we performed
   menin‐immunoprecipitation (IP) and liquid chromatography‐tandem mass
   spectrometry (LC‐MS) in hippocampus tissue from Men1‐CcKO mice and the
   controls (Figure [127]4A). Label‐free LC‐MS/MS analysis identified 247
   potential menin‐associated proteins (≥one unique peptide with an FDR <
   1%) (Figure [128]4B) (Data [129]S6, Supporting Information). These
   proteins were mostly enriched in neuronal development and mature GO
   terms, as well as transcriptional process‐related GO terms
   (Figure [130]4C). Notably, seven of these overlapping proteins are
   involved in neuronal development and serve as transcription factors
   (Figure [131]4D). Among the aforementioned seven candidates, ATRX is
   found to be associated with Alpha Thalassemia/Mental Retardation
   Syndrome X‐Linked, a rare congenital disorder with cognitive
   impairment. Another reason for selecting Atrx is that the
   ATRX‐DAXX‐menin complex plays a crucial role in neuroendocrine
   tumors.^[ [132]^11 , [133]^20 ^] The immunoprecipitation (IP) analysis
   in hippocampus tissue further confirmed that Atrx binds to menin in
   vivo (Figure [134]4E). It is noteworthy that Atrx gene encodes two main
   splicing isoforms: the full length Atrx (Atrx‐FL) and a truncated
   isoform (Atrxt).^[ [135]^21 ^] Men1 deletion resulted in decreased
   expression levels of Foxg1 but not that of Atrx (Figure [136]S4F,G,
   Supporting Information). On the other hand, the knockdown of Atrx by
   small interfering RNAs reduced the mRNA and protein levels of Foxg1
   (Figure [137]4F–H). This reduction may be due to the decreased
   abundance of Atrx binding to the Foxg1 promoter (Figure [138]4I).
   Additionally, the signal intensity of Atrx‐binding to the Foxg1
   promoter was significantly reduced in Men1‐knockout neurons
   (Figure [139]4J). Atrx silencing reduced the abundance of H3K4me3 marks
   on the Foxg1 promoter in neurons (Figure [140]4K). By luciferase assay,
   we also found that the transcriptional activity of Foxg1 dramatically
   decreased when Atrx was knockdown by shAtrx virus (Figure [141]4L,M).

Figure 4.

   Figure 4
   [142]Open in a new tab

   Atrx knockdown causes autistic behaviors and contributes to the
   transcription regulation of Foxg1 by menin. A) The experimental
   strategy used to identify the interacting proteins of menin in control
   and Men1‐CcKO mouse brain is schematically depicted. Hippocampus
   tissues were obtained from 1‐month‐old Men1‐CcKO mice and littermate
   control mice (Men1^(F/F) mice). The menin‐IP‐MS experiment was
   conducted in two independent trials. B) Interaction analysis among four
   protein sets revealed menin‐specific interacting proteins and a total
   of 247 menin‐specific interacting proteins were selected. C) The top 10
   enriched cellular compound and biological process contributing to menin
   function were determined by menin‐specific interacting proteins. Red
   columns represent neuronal development and mature‐related GO terms, and
   green columns represent transcriptional process‐related GO terms, while
   blue columns represent other terms. D) 7 potential factors were
   screened from menin‐interacting proteins involved in both neuronal
   development and transcriptional regulation. Atrx may be an important
   factor involved in the regulation of menin‐Foxg1. E) Interactions
   between Atrx and menin were determined in control and Men1‐CcKO mouse
   brains. Hippocampus tissues came from 1‐month‐old Men1‐CcKO mice and
   littermate control mice (Men1^(F/F) mice) (n = 3 independent
   experiments). F) Primary neurons were transfected with three siAtrx
   sequence for 72 h, followed by Western blotting of Atrx, menin and
   Foxg1 protein levels. α‐tubulin was used as a loading control (n = 3
   independent experiments). G) The corresponding statistical results for
   (F) are shown respectively. H) Primary neurons were transfected with
   three siAtrx sequence for 72 h, followed by real‐time PCR to detect the
   mRNA expression of Atrx and Foxg1. β‐actin was used as a loading
   control. (n = 3 samples from 3 mice per group). I) The enrichment of
   Atrx binds to Foxg1 promoter after interfering with Atrx expression in
   primary neurons. (n = 4 independent experiments and 4 samples per
   group). J) The enrichment of Atrx binds to Foxg1 promoter was compared
   between Men1‐KO and WT primary neurons (Primary neurons are obtained
   from E16.5 embryos Men1‐NcKO mice and littermate control mice). (n = 4
   independent experiments and 4 samples per group). K) The signal
   intensity of H3K4me3 on the Foxg1 promoter after interfering with Atrx
   expression in primary neurons. (n = 4 independent experiments and 7
   samples per group). L) To assessment of Atrx‐mediated transactivation
   of Foxg1, transfecting primary neurons with luciferase reporter
   plasmids after interfering Atrx expression level. The same fragment of
   siAtrx‐3 was inserted into AAV‐CaMK2a‐EGFP‐3xFlag‐miR30shRNA‐WPRE. M)
   The luciferase activity of Atrx‐mediated transactivation of Foxg1 was
   measured. (n = 6 samples per group). (All mice were males. Data is
   represented by mean ± SEM.*p < 0.05, **p < 0.01, ***p < 0.001, and
   ****p < 0.0001 indicate significance between the two indicated groups.
   G, H, M : one‐way ANOVA with Tukey post hoc test. I, J, K : two‐way
   ANOVA with Tukey post hoc test.).

   Subsequently, the knockdown of Atrx in excitatory neurons by
   intraventricular injecting shAtrx and control virus mediated by CaMK2α
   promoter into wild‐type neonatal mice (Figure [143]S4H, Supporting
   Information). After 2 months, these mice were subjected to
   comprehensive behavioral and histological analyses (Figure [144]5A).
   The Atrx protein level was significantly decreased by shAtrx AAV
   infection, accompanied by down‐regulation of Foxg1 (Figure [145]5B,C;
   Figure [146]S4I–L, Supporting Information). Behavior tests indicate
   that knockdown of excitatory neuronal Atrx significantly induces mice
   to exhibited repeating grooming behaviors (Figure [147]5D), impaired
   social memory (Figure [148]5E), and cognitive deficits (Figure
   [149]S4M, Supporting Information). In summary, Atrx, a novel potential
   target of ASD, governs Foxg1 transcriptional regulation by
   menin‐mediated histone H3 lysine 4 trimethylation.

Figure 5.

   Figure 5
   [150]Open in a new tab

   Partially Atrx Deletion causes autistic like behaviors. A) The lateral
   ventricles of neonatal C57BL6/J mice were intracerebroventricularly
   injected with shControl AAV or shAtrx AAV on postnatal day 1 (P1) to
   generate an excitatory neuron‐specific Atrx deficiency mouse model. B)
   The injection of shControl AAV or shAtrx AAV was performed in newborn
   C57 mouse bilateral ventricles. Representative confocal images of brain
   sections stained with Atrx (red) antibody. Section was counterstained
   with DAPI (blue). The green channel represents GFP. Scale bar, 20 µm.
   Insets marked by white boxes show enlarged cells. C) The corresponding
   statistical results of the fluorescence of Atrx in (B) were shown
   respectively. The left graph shows cortical statistics, while the right
   graph shows hippocampal statistics. (n = 30 GFP‐positive neurons.
   shControl AAV group: n = 6 slices from 3 mice; shAtrx AAV group: n = 6
   slices from 3 mice, all mice were males). D) Grooming tests were used
   to assess compulsive, rigid, repetitive behaviors in the four groups.
   (shControl AAV‐male group: n = 13; shControl AAV‐female group: n = 17;
   shAtrx AAV‐male group: n = 14; shAtrx AAV‐female group: n = 13). E)
   Three‐chamber social tests were employed to analyze social behavior in
   mice, including adaptive exploration stage (left), social novelty tests
   (middle) and social memory tests (right) (shControl AAV‐male group:
   n = 13; shControl AAV‐female group: n = 17; shAtrx AAV‐male group:
   n = 14; shAtrx AAV‐female group: n = 13).(Data is represented by
   mean ± SEM. ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001,
   and ****p < 0.0001 indicate significance between the two indicated
   groups. D : one‐way ANOVA with Tukey post hoc test. E : two‐way ANOVA
   with Tukey post hoc test. C : unpaired t‐test).

2.5. Restoring Foxg1 Expression in Men1‐CcKO Mice Relieves ASD‐Like Behaviors
and Synaptic Deficits

   Finally, we reinstated the expression of menin and Foxg1 by
   administering AAV‐EF1α‐DIO‐MEN1‐GFP (MEN1 AAV) and
   AAV‐EF1α‐DIO‐Foxg1‐GFP (Foxg1 AAV) virus, as well as AAV‐EF1α‐DIO‐GFP
   (Control AAV) into the bilateral CA1 region of the hippocampus in
   Men1‐CcKO mice and their littermates (Figure [151]6A). The levels of
   synaptic proteins, including GluR1/2, NR2A/2B, synaptophysin and PSD95,
   were decreased in the hippocampus of Men1‐CcKO mice compared to
   controls. The expression of these proteins was normalized by restoring
   menin or Foxg1 (Figure [152]6B). The hippocampal long‐term potentiation
   (LTP) defects in Men1‐CcKO hippocampus were also attenuated by Foxg1
   restoration (Figure [153]6C,D). Overexpression of Foxg1 also
   ameliorated the neuronal development defects induced by menin
   deficiency, as it increased the synaptic number, the neuronal dendritic
   complexity and total dendritic length in Men1‐KO primary neurons
   (Figure [154]6E–I).

Figure 6.

   Figure 6
   [155]Open in a new tab

   Foxg1 alleviates ASD‐like behaviors and synaptic dysfunction caused by
   Men1 deficiency. A) The Men1‐CcKO mice and littermates were injected
   with AAV‐EF1α‐DIO‐GFP (Control AAV), AAV‐EF1α‐DIO‐MEN1‐GFP (MEN1 AAV),
   and AAV‐EF1α‐DIO‐Foxg1‐T2A‐GFP (Foxg1 AAV) into hippocampus
   respectively. B) Menin, Foxg1, NR2A, NR2B, PSD95, GluR1, GluR2,
   Synaptophysin (SYP), Tuj1, NeuN, and internal control proteins
   (α‐tubulin and GAPDH) were detected in different mouse hippocampal
   lysates using SDS‐PAGE electrophoresis. The corresponding statistics
   are shown on the right side. (n = 4 samples from 4 mice per group). C)
   LTP recordings of brain slices from litter‐ and age‐matched control and
   Men1‐CcKO mice that were injected with the indicated AAV viruses in
   panel A. D) The field excitatory postsynaptic potential (fEPSP)
   potentiation was quantified during the last 10 min of recording
   (Control‐Control AAV: n = 9 slices from 4 mice; Men1‐CcKO‐Control AAV:
   n = 9 slices from 4 mice; Men1‐CcKO‐Foxg1 AAV: n = 10 slices from 5
   mice). E) Primary neurons from Control and Men1‐KO mice were cultured
   in vitro and infected with AAV‐EF1α‐DIO‐GFP (Control AAV) and
   AAV‐EF1α‐DIO‐Foxg1‐GFP (Foxg1 AAV) on DIV 3. Confocal projection images
   of GFP‐positive, Map2 (Red) and Synaptophysin (White) stained neurons
   on DIV 12 are shown. Scale bar: 10 µm. (n = 60 dendrites in each
   condition). F) The corresponding statistical results for (E) are shown.
   G) Primary neurons from Control and Men1‐deficient mice were cultured
   in vitro and transduced with AAV‐EF1α‐DIO‐GFP and
   AAV‐EF1α‐DIO‐Foxg1‐GFP on DIV 3. Confocal projection images of Map2
   (red) and Synaptophysin (baby blue) stained neurons on DIV 12 are
   shown. Scale bar, 50 µm. (n = 60 neurons in each condition). H) Summary
   of dendritic arborization. I) Summary of total length. J) Workflow of
   stereotactic AAV injection and rescue experiments. K) Grooming tests
   were used to assess the repetitive patterns of behavior in the three
   groups. L) Three chamber social tests were used to assess the social
   ability of the three groups, including adaptive exploration stage,
   social novelty and social memory (From left to right). (All mice were
   males. CaMK2α‐mCherry AAV: n = 15; CaMK2α‐Cre‐mCherry AAV& GFP AAV:
   n = 15; CaMK2α‐Cre‐mCherry AAV& Foxg1 AAV: n = 15. Data is represented
   mean ± SEM. ns: not significant, *p < 0.05, **p < 0.01, ***p < 0.001,
   and ****p < 0.0001 indicate significance between the two indicated
   groups. D, G, I, L : one‐way ANOVA with Turkey post hoc test. B, E :
   two‐way ANOVA with Turkey post hoc test. K : paired t‐test.).

   Using another strategy, we injected CaMK2α‐Cre AAV and
   AAV‐EF1α‐DIO‐Foxg1‐GFP into Men1 ^F/F mice to simultaneously
   overexpress Foxg1 and down‐regulate Men1 in excitatory neurons
   (Figure [156]6J). The behavioral tests were subsequently performed 45
   days after the injection. Immunostaining confirmed a decreased menin
   expression and an increased Foxg1 expression in the cortex and
   hippocampus of the injected animals (Figure [157]S5A,B, Supporting
   Information). Notably, overexpression of Foxg1 partially rescued
   incremental repetitive behaviors (Figure [158]6K), social disorders
   (Figure [159]6L) and cognitive impairments (Figure [160]S5C, Supporting
   Information) in neuronal menin knockdown mice. Therefore, restoring
   Foxg1 expression can alleviate ASD‐like behaviors and synaptic
   dysfunction caused by Men1 deletion.

3. Conclusion

   Menin is a nuclear scaffold protein that participates in
   transcriptional regulation through histone modification, such as lysine
   trimethylation at H3K4 and H3K27.^[ [161]^22 ^] MLL1/2 belong to the
   trithorax group and specifically methylate histone H3K4 to produce the
   H3K4me3 modification. Deletion of Mll1 or Mll2 specifically in neurons
   in mice resulted in reduced synaptic plasticity and impaired working
   memory by regulating the expression of important genes involved in
   neuronal development.^[ [162]^9 ^] Menin also can coordinate with
   EZH2‐PRC2 to catalyze H3K27me3 to suppress the gene expression.^[
   [163]^23 ^] Recently, menin has also been identified as a new H3K79me2
   reader and is associated with the binding sites of H3K79me2 that span
   across their gene bodies.^[ [164]^24 ^] These findings indicate that
   menin plays a complex role in histone modifications. Menin is a
   multifunctional epigenetic regulator that interacts with various
   proteins to perform a variety of functions. Here, we found that menin
   associates with Atrx to regulate the transcription of Foxg1 by H3K4me3.

   Interestingly, the deletion or overexpression of menin both induces the
   mice exhibited autism‐like behaviors, suggesting that the expression
   level of menin is crucial for the neuronal functions. The bidirectional
   effects of menin also reflect its downstream gene Foxg1. Since both
   heterozygous FOXG1 duplication or FOXG1 deletion can lead to FOXG1
   syndrome.^[ [165]^1 , [166]^25 ^] Meanwhile, the bidirectional effect
   of menin on autism‐like behaviors is similar to that of another
   well‐known pathogenic gene in Rett syndrome: MeCP2. Loss of function of
   MeCP2 induces the Rett syndrome phenotype, while gain of function
   promotes MeCP2 duplication syndrome.^[ [167]^26 ^] Menin and MeCP2 are
   both important epigenetic regulators that play bidirectional roles in
   regulating gene expression.^[ [168]^27 ^]

   Furthermore, single‐nucleus RNA sequencing of cortical tissues from
   patients with ASD revealed that excitatory cortical neurons are the
   main cell types preferentially affected in ASD.^[ [169]^28 ^] In this
   study, we found that Men1 deletion in excitatory neurons, but not
   inhibitory neurons, led to mice exhibiting autism‐like behaviors, which
   aligns with the aforementioned findings in ASD patients. We recently
   reported that Men1 deletion in PV interneurons induced mice to exhibit
   depressive‐like behaviors.^[ [170]^29 ^] The effects of menin on
   interneurons, especially in different types of interneurons, deserve
   thorough investigations. The regulation of FOXG1‐associated
   transcription is crucial for neuronal development and synaptic
   plasticity.^[ [171]^30 ^] The recent studies have shown that Foxg1
   plays a crucial role in regulating the specification of postmitotic
   neurons. Knocking out Foxg1 has been found to result in deficits in
   neuronal projection.^[ [172]^31 ^] Overexpression of Foxg1 attenuated
   the neuronal development defects induced by Men1 deficiency.

   The ATRX gene encodes a SWI/SNF‐like chromatin remodeling protein and
   is frequently mutated in various nervous system‐related tumors.^[
   [173]^32 ^] Deletion of Atrx in excitatory neurons selectively leads to
   working memory deficits in male mice.^[ [174]^20 , [175]^33 ^] ATRX
   interacts with EZH2 to facilitate the deposition of H3K27me3.^[
   [176]^34 ^] Furthermore, ATRX may be recruited to the heterochromatin
   histone mark, H3K9me3, either indirectly through interaction with HP1
   or via recruitment of MeCP2,^[ [177]^35 ^] or directly by binding to
   the ADD domain to H3K9me3, which is located adjacent to unmethylated
   H3K4.^[ [178]^36 ^] However, administration of menin‐MLL inhibitors
   (MI‐2 and MI‐3) did not alter Foxg1 expression, suggesting a potential
   mechanism that remains to be revealed. Our hypothesis suggests that
   Atrx recruits menin to bind to the transcriptional start region of
   Foxg1, mediating the regulation of Foxg1 expression through a
   non‐Mll1/2 dependent process involving H3K4me3. This has prompted us to
   re‐evaluate the role of the Atrx complex.

   Our work further highlights the coordination between menin and Atrx in
   mediating neuronal development. Dysfunction of these proteins can
   induce ASD‐like phenotypes in mice. Interestingly, it has been shown
   that Atrx directly interacts with Mecp2, and Mecp2 deficiency disrupts
   Atrx PCH localization.^[ [179]^36a ^] In our experiments, knockdown of
   Atrx specifically in forebrain excitatory neurons in mice led to
   impaired social memory, increased repetitive behaviors and spatial
   memory impairment. Therefore, Atrx is a potential target of
   ASD‐associated virulence gene.

   Previously, we have demonstrated that menin can regulate the activity
   of cyclin dependent kinase 5 (Cdk5) by mediating the expression of its
   activator, p35.^[ [180]^37 ^] The deficiency of p35 and Foxg1 has a
   similar impact on cognitive functions in mice. However, Foxg1
   dysregulation mainly causes ASD‐like behaviors, including social
   defects, increased repetitive behaviors, and sporadic epileptic
   encephalopathy.

   The study has certain limitations. First, we created an unnatural model
   of ASD‐like behaviors in mice by deleting Men1 specifically from
   excitatory neurons, which is distinct from human clinical cases. It
   will be necessary to explore menin‐FOXG1‐related ASD cases in the
   clinical population and elucidate the precise genetic causes of the
   symptoms. Second, numerous reports show that ATRX, menin and FOXG1 are
   all involved in the pathogenesis of cancer.^[ [181]^38 ^] Further
   investigations are needed to determine whether the ATRX‐menin‐FOXG1
   axis is an important comorbid target in FOXG1 syndrome and cancer.
   Thirdly, ATRX is a transcription repressor that may stabilize chromatin
   by inhibiting transcription or replication through the DNA damage
   pathway. However, in this study, Atrx was found to be involved in the
   transcriptional activation of Foxg1 mediated by menin. Further research
   is required to fully comprehend the regulatory mechanisms.
   Additionally, our study demonstrated that knockdown of Atrx resulted in
   the down‐regulation of Foxg1 and ASD‐like behaviors. The overexpression
   of Atrx should be performed to investigate whether the restoration of
   Atrx can rescue the physiological and behavioral change in Men1‐CcKO
   mice. However, the Atrx gene is too large to be accommodated in the
   current viral expression system. It is worth noting that the loss of
   function of Atrx is closely correlated with the ASD phenotype. We will
   investigate the effect of Atrx overexpression for an extended duration.
   In summary, our finding demonstrates that menin/Atrx regulate Foxg1
   transcription through H3K4me3 and are involved in the pathogenesis of
   FOXG1 syndrome.

4. Experimental Section

Ethics Statement

   The study was approved by the laboratory animal center at Xiamen
   University, and all experimental procedures involved were performed
   according to protocols approved by the Institutional Animal Care and
   Use Committee at Xiamen University. The approval number of the protocol
   is XMULAC20190144.

Experimental Animals

   All mice were reared on a 12/12 light/dark cycle. As described in the
   previous article.^[ [182]^37 ^] The Men1 floxed mouse strain was
   crossed with CaMK2α‐Cre mice (JAX#:005359), Nestin‐Cre mice
   (JAX#:003771), and Dlx5/6α‐Cre‐IRES‐GFP(JAX #:023724) mice
   respectively.^[ [183]^13 , [184]^39 ^] Without extra emphasis, Mice
   aged between 2 to 4 months old were selected for the experiment.
   Litter‐ and age‐matched male mice were exclusively used for behavioral
   tests, electrophysiology, and biochemistry experiment to exclude
   interfering effects derived from estrogen.

Newborn Mice were Injected with Adeno‐Associated Virus and Stereotactic
Injection of Adeno‐Associated Virus in Adult Mice

   pAAV‐EF1α‐DIO‐GFP‐WPRE (virus titer: 5.25 × 10^12/ml),
   pAAV‐EF1α‐DIO‐MEN1 ‐GFP‐WPRE (virus titer: 6.29×10^12/ml),
   pAAV‐EF1α‐DIO‐Foxg1‐T2A‐GFP‐WPRE (virus titer: 2.95 × 10^12/ml) were
   purchased from BrainVTA (WuHan, China). Additionally,
   pAAV‐CaMK2α‐EGFP‐3xFlag‐miR30‐shRNA (Atrx)‐WPRE (virus titer:
   5.14 × 10^12/ml) and pAAV‐CaMK2α‐EGFP‐3xFlag‐miR30‐shRNA (NC)‐WPRE
   (virus titer: 8.53 × 10^12/ml) were purchased from OBIO (ShangHai,
   China). All viral vectors were aliquoted and stored at −80 °C until
   used.

   pAAV‐EF1α‐DIO‐GFP‐WPRE and pAAV‐EF1α‐DIO‐MEN1‐GFP‐WPRE were injected
   into the bilateral ventricles of newborn CaMK2a‐Cre mice respectively
   at the same virus titer (virus titer: 5 × 10^12/ml, 1 ul, one side)^[
   [185]^40 ^] to overexpress MEN1 in excitatory neurons for behavioral
   assessment of the mice.

   pAAV‐CaMK2α‐EGFP‐3xFlag‐miR30‐shRNA(Atrx)‐WPRE (virus titer:
   5.14 × 10^12/ml) and pAAV‐CaMK2α‐EGFP‐3xFlag‐miR30‐shRNA(NC)‐WPRE
   (virus titer: 8.53 × 10^12/ml) were injected into the bilateral
   ventricles of newborn C57/BL6J mice respectively at the same virus
   titer (virus titer: 5 × 10^12/ml, 1 ul, one side) to further evaluate
   ASD‐like behaviors in mice after the intervention of Atrx expression.

   pAAV‐EF1α‐DIO‐GFP‐WPRE, pAAV‐EF1α‐DIO‐MEN1‐GFP‐WPRE, pAAV‐EF1α‐
   DIO‐Foxg1‐T2A‐GFP‐WPRE, pAAV‐CaMK2a‐mCherry‐T2A‐Cre were
   stereotactically injected into bilateral hippocampus of male Men1^(F/F)
   mice at the same virus titer (virus titer: 2.5 × 1012/ml, 1ul, one
   side). The specific spatial coordinates were X = ±1.7, Y = ‐2.0, and
   Z = ‐1.7 mm. To confirm the overexpression of MEN1 or Foxg1 in mouse
   brains, mice were anesthetized and sacrificed 21 days after injection.
   Subsequently, brain tissues were rapidly removed and analyzed using
   histological immunofluorescence staining and western blotting.

Primary Neurons Culture, Small Interfering RNA, Plasmids and Drug Experiments

   A mixture of primary cortical and hippocampal neurons was dissected
   from timed‐pregnant females at E16.5. Men1 knockout neurons were
   derived from Men1‐NcKO (Nestin‐Cre;Men1 ^(F/F)) mice, while Control
   neurons were derived from litter‐matched Men1^(F/F) mice. Briefly,
   brains of pups with varying genotypes were dissected. The meninges were
   removed, and the tissue was dissociated by enzymatic digestion.
   Isolated primary neurons were plated on poly‐D‐lysine coated dishes,
   and cultured in Neurobasal medium supplemented with B27 (Gibco,
   17504044), 1% penicillin/streptomycin (Invitrogen) and maintained in a
   5% CO[2] incubator at 37 °C.

   Atrx Small interfering RNA (siAtrx) were customized from Tsingke
   Biotechnology Co., Ltd. On DIV6 primary neurons, several siRNA were
   transfected with Lipofectamine 3000 (Thermo fisher, L3000150). After
   72 h, western blotting or chromatin immunoprecipitation (ChIP) was
   performed. On DIV10 primary neurons were treated with MI2 (MCE,
   HY‐15222) and MI3 (MCE, HY‐15223) at a concentration of 1 µm each.
   After 24 h, western blotting or RT‐PCR was performed. The pGL4.0‐Foxg1
   promoter‐2.5 k plasmid and the pGL4.0[LUC2] plasmids were obtained from
   Tsingke (China, Beijing). In short, the fragment of Foxg1
   promoter‐2.5 K was synthesized, according to the Mouse Genome
   Informatics was inserted into the pGL4.0[LUC2] plasmid using
   sites‐directed insertion technique. The two plasmids were severally
   transfected with Lipofectamine 3000 (Thermo fisher, L3000150) into
   primary neurons on DIV10 after 3 day's treatment of shAtrx AAV or
   shControl AAV.

Tissue and Cell Lysate Preparation, and Antibodies Used for Immunoblotting

   Cortex and hippocampus samples were prepared from at least three pairs
   of male mouse brains, including Control and Men1‐CcKO mice. Proteins
   extracted from primary neurons or organs tissue lysates were analyzed
   using western blotting. The proteins were separated on
   SDS‐polyacrylamide gel electrophoresis and probed with specific
   antibodies, including menin (rabbit, 1:1000; Abcam, ab31902), Foxg1
   (rabbit, 1:1,000; Abcam, ab196868), Atrx (rabbit, 1:1000, Bethyl
   Laboratories, A301‐045A‐T), PSD95 (rabbit, 1:1000; Cell signaling
   Tech., #2507), GluR1 (rabbit, 1:1000, Cell signaling Tech., #13185),
   GluR2 (rabbit, 1:1000, Cell signaling Tech., #13607), NR2A(rabbit,
   1:500, Sangon Biotech, D160672), NR2B (rabbit, 1:500, Sangon Biotech,
   NO. D260673), Synaptophysin (rabbit, 1:000; Cell signaling Tech.,
   #36406), Tuj1 (TUBB3) (mouse, 1:1000, Biolegend, # MMS‐435P), NeuN
   (mouse, 1:1000, Millipore, MAB377), β‐actin (mouse, 1:1,000; Cell
   Signaling Technology, #3700), GAPDH (mouse, 1:5000, Invitrogen,
   #MA5‐15738), α‐tubulin (mouse, 1:5000, abclonal, 1:5000, AC012).
   Goat‐anti‐mouse secondary antibodies and goat‐anti‐rabbit secondary
   antibody were purchased from Millipore (#AP132P, #AP124P). For
   quantification, densitometry measurements of each band (Image J) were
   first normalized to β‐actin or GAPDH or α‐tubulin, and averaged from at
   least three independent experiments. More details are provided in Data
   [186]S7 (Supporting Information).

Immunofluorescence Staining

   Immunofluorescence staining was performed as previously described.^[
   [187]^37 ^] Mouse brain sections or cultured cells were fixed in 4%
   paraformaldehyde and washed three times with PBS. Antigen retrieval was
   performed using citrate buffer (pH7.0), followed by permeabilization
   and blocking in PBS containing 0.5% Triton X‐100 and 10% normal goat
   serum at room temperature for 1 h. Brain sections or cultured cells
   were incubated with primary antibodies in the blocking buffer overnight
   at 4 °C, then probed with secondary antibodies in the same buffer for
   1 h on the following day. Samples were washed and stained with DAPI.
   Images were acquired using a Nikon confocal microscope and a Zeiss LSM
   880 Airyscan in the Core facility of Biomedical sciences at Xiamen
   University. Primary antibodies used for immunostaining include: menin
   (rabbit, 1:500; Abcam) GFP (mouse,1:1000; Santa Cruz), MAP2 (rabbit,
   1:1000; Cell Signaling Technology), Tuj1(mouse,1:1000; Biolegend),
   Foxg1 (rabbit, 1:500; Abcam, ab196868), Atrx (rabbit, 1:200, Bethyl
   Laboratories, A301‐045A‐T) et al. 488/594 donkey anti‐mouse/rabbit
   secondary antibodies (1:1000) and Mounting Medium with DAPI was
   purchased from Invitrogen. Further information on the antibodies used
   can be found in Data [188]S7 (Supporting Information).

Co‐Immunoprecipitation

   The mice were quickly killed by carbon dioxide and their brains were
   separated. Wash the brain twice with ice‐cold PBS and fully homogenize
   the tissue using the homogenate in IP buffer (Tris‐Hcl pH = 7.4 0.05m;
   Nacl 0.15 mm; EGTA 1 µm; EDTA 1 µm; Triton‐100 1%; NaPPi 0.223 g; NaVO3
   1 µm; ddH2O Up to 200 ml). Then, gently pipette up and down on ice to
   lyse the brain for 2 h (10 times every 10 min). Centrifuge at
   12,000 x g for 15 min at 4 °C and transfer all supernatant to a new
   tube. Measure the protein concentration and check the expression of
   menin by SDS‐PAGE/Immunoblotting analyses (This step is optional but
   highly recommended.). Normally 5 µg of Atrx antibodies (CST‐#14820) or
   5 µg of menin antibodies (Bethyl‐A300‐105A) were used in the lysate.
   Incubate the tube overnight at 4 °C on a rotator. The next day, use a
   large‐orifice pipette tip to transfer an adequate amount of Protein G
   (Thermo‐10004D) into a 1.5 ml tube (20–30 µl per Sample). Then,
   preclear the Protein G 3–4 times using a magnetic frame and IP buffer.
   Resuspend the Protein G in ice‐cold IP buffer (≈20 µl X (sample number
   + 1)). Add 20 µl precleaned Protein G into the lysate. Incubate the
   tube on a rotator for 2–4 h at 4 °C. After co‐incubation, aspirate the
   supernatant to remove non‐binding proteins and wash the pellet by
   resuspending it in 1 ml ice‐cold IP buffer. Resuspend the Protein
   G‐bound immune complexes in 30–40 µl 1x loading buffer, boil for
   5–10 min, and analyze them by SDS‐PAGE/Immunoblotting, mass
   spectrometry, or other methods. If the sample would be analyzed by
   SDS‐PAGE/Immunoblotting, avoid using antibodies developed in the same
   species as the antibody used for immunoprecipitation.

Liquid Chromatography‐Tandem Mass Spectrometry (LC‐MS)

   Hippocampus dissection was performed and the LC‐MS experiment was
   conducted in double‐blind conditions. Briefly, please refer to the
   Co‐IP procedure mentioned above, where Protein G‐bound immune complexes
   were analyzed by SDS‐PAGE/Immunoblotting. The glue was cut into small
   pieces with a size of 1.5 mm^3 and placed in centrifuge tube (Volume ≤
   300 uL in each tube). The glue was washed with ultra‐pure water (DDW)
   for more than three times and then decolorized with a decolorizing
   solution [40% acetonitrile (ACN), 50 mm ammonium bicarbonate (ABC) in
   H[2]O] until it became colorless. The sample was finally stored in the
   decolorizing solution. Trypsin (Promega, V5111) was added at a ratio of
   1:50 (trypsin: protein, m/m) and enzymolyzed overnight. The fractions
   were dried for LC‐MS/MS analysis using a Bruker timsTOF Pro instrument.
   The resulting raw files were imported into MaxQuant software for data
   interpretation and protein identification against the database. This
   experiment received significant support from the Core Facility of
   Biomedical Sciences at Xiamen University. The data was successfully
   submitted to ProteomicXchange via the PRIDE database, with project
   accession number: PXD042795.

Chromatin Immunoprecipitation (ChIP)

   ChIP procedures were performed followed the manufacturer's instructions
   (17‐295; Millipore). Briefly, primary neurons from different genotypes
   of mice were cultured for 8–12 days and fixed for 10 min at room
   temperature with media containing 1% formaldehyde and quenched with
   125 mm glycine for 5 min. Fixed homogenates were washed twice using
   ice‐cold PBS containing protease inhibitors. Fixed nuclei were pelleted
   at 2,000 rpm for 4 min and re‐suspended in SDS Lysis Buffer (catalog
   #20‐163), where chromatin was sheared using a SCIENTZ ultrasonic
   apparatus set to 28% power for 14 cycles of a 4.5 s sonication and a
   9.0 s resting stage on ice. The sonicated cell supernatant was diluted
   ten‐fold in ChIP dilution buffer (catalog #20‐153) and pre‐cleared
   using protein An Agarose/Salomon Sperm DNA (catalog #16‐157). Five
   percent of sample was removed before the immunoprecipitation (IP) for
   input measurements. ChIP was performed using 5 µg menin antibody
   (A300‐105A; Bethyl), 5 µg Atrx antibodies (CST‐#14820), 3ug H3K4me3
   (17‐614; Millipore), or normal rabbit IgG (H2615; Santa Cruz) antibody
   incubated overnight, followed by enrichment using Protein A Sepharose
   beads for 4 h. Beads were washed three times with four different
   buffers (low‐salt immune complex wash buffer, high‐salt immune complex
   wash buffer, and LiCl immune complex wash buffer) and one wash with TE
   (50 mm Tris HCl, 10 mm EDTA). Chromatin was eluted by agitation at
   65 °C for 20 min in TES (TE plus 1%SDS) and reverse cross‐linked
   overnight at 65 °C. Chromatin was subjected to RNase and Proteinase K
   treatment, followed by DNA purification by phenol chloroform extraction
   and ethanol precipitation. DNA pellets were re‐suspended in 10 mm Tris
   and subjected to qPCR (480; Roche). The qPCR signals were calculated
   using the following equation: ① The background value was obtained by
   subtracting the ΔCT value of Input from the ΔCT value of IP[IgG], and
   the actual value was obtained by subtracting the ΔCT value of Input
   from the ΔCT value of IP[specific antibody]. ②The background value was
   normalized by the average value of biological duplicate samples from
   the same group (defined as normalized background values) and the actual
   value of specific antibody was normalized to its corresponding
   background value (referred to as normalized actual values). ③ The fold
   change was calculated by using the formula: Fold Change = 2 raise to
   the power of (‐ normalized actual values or normalized background
   values). Primer sequences used for ChIP‐qPCR in this study are summary
   in Data [189]S7 (Supporting Information).

Behavior Tests

   Morris Water Maze: The water maze used in this study comprised a
   circular tank 120 cm in diameter with a platform filled with tap water
   at a temperature of 22 ± 2 °C. Different shapes were posted along the
   walls of the tank, which served as spatial reference cues. A camera was
   mounted above the maze to record the swimming traces in the water maze.
   During the acquisition trials, the platform was submerged 1–2 cm below
   the water surface, mice were place into the maze at one of four points
   (N, S, E, W) facing the wall of the tank. Mice could search for a
   platform for 60 s. If a mouse failed to find the platform, it was
   guided to the platform and maintained on the platform for 10 s. Four
   trials a day were conducted with at an intermission of 1 h minimum
   between trials. Escape latency which indicated spatial learning memory
   acquisition, was recorded for each trial. On day 7, the platform was
   removed and a probe test was conducted. The percentage time spent in
   each of the four quadrants and the number of target (platform) area
   crossings, mean speed, total distance was recorded.

   Fear Conditioning Test: Experiments were carried out using a fear
   conditioning system (StartFear, Panlab Harvard Apparatus). The
   apparatus consisted of a test chamber (25 cm × 30 cm × 25 cm,
   H × W × L), where the ceiling, front and back of the chamber was
   transparent, and the floor was made of a removable grid of
   stainless‐steel rods (3.2 mm diameter, 4.7 mm apart). Automated fear
   conditioning software (FREEZING, Panlab, Harvard apparatus) was used to
   control foot shocks. Video recording within the chamber monitored the
   activity of the mice throughout the procedure. On training days, mice
   were placed in the center of the grid floor and allowed to explore the
   test chamber for 120 s. Next, three consecutive vocal cues were
   administered at 80 db maintained for 30 s, and foot shocks were
   administered at 0.5 mA for 2 s, and the resulting response were
   recorded as baseline responses. Animals were removed from the chamber
   and returned to their regular habitation. One day later, each mouse was
   re‐exposed to the chamber for 600 s without any stimulus to assess
   context‐associated fear conditioning. After 1 h, mice were placed in
   bright condition and left to explore the test chamber for 120 s. A
   vocal cue of 80 db maintained for 180 s was administered, and freezing
   time within a 300 s span was recorded during the cued test. Mouse
   activity on the first day was used for baseline comparison.

   Rotarod Test: Mice were placed on a stationary rotarod (AccuRotor Rota
   Rod Tall Unit, 63 cm fall height, 30 mm diameter rotating dowel;
   Accuscan, Columbus, OH). The dowel was then accelerated at
   60 rpm min^−1, and the latency to fall (in seconds) was recorded. The
   procedure was repeated for 4 consecutive trials, which were averaged to
   give the daily latency to fall for each mouse. If an animal fell off
   the rotarod rapidly (e.g., due to inattention or slips), they were
   placed back on the rotarod for an additional trial, and the latency was
   not included in the average for the day. The entire procedure was
   repeated the following 2 days for a total of 3 days. In addition to the
   average latency across the 4 trials per day, the maximum latency to
   fall per day was also analyzed.

   Three‐Chamber Sociability Test: The test apparatus consisted of three
   chambers: a middle chamber (40 × 20 cm^2) and another two chambers
   (40 × 20 cm^2) in which wire cages containing stranger mice were
   placed. For the sociability test, the test animal was placed in the
   middle chamber for 5 min to habituate. Then, a wire cage containing a
   novel juvenile male mouse was placed in one of the side chambers, and
   an empty wire cage was placed in the other. The test animal was placed
   in the middle chamber for 10 min. Lastly, two wire cages containing two
   novel juvenile male mice were placed on both sides. The test animal was
   placed in the middle chamber for 10 min. The time spent in each chamber
   and the time spent near the wire cages were analyzed using SMART v.3.0
   software.

   Grooming Behaviors: Grooming behaviors were measured as described
   previously. In brief, to measure basal grooming behavior, mice were
   video‐recorded in the dark phase for 15 min. The measured grooming
   behaviors included face wiping, scratching/rubbing of head and ears and
   full‐body grooming. After recording the behaviors, the experimenters
   were blinded to the genotypes and manually analyzed the videotapes.^[
   [190]^41 ^]

   Nest‐Building Test: Mice received a fresh nestlet and a new standard
   housing cage with fresh bedding 1 h before the beginning of the dark
   phase. Nest status was assessed at the end of the dark phase. Nests
   were scored according to categories similar to a previously described
   protocol: untouched nestlet (0 points), partially shredded nestlet (1
   point), shredded nestlet with beginning of a nest‐like structure (2
   points), full cup‐like nest (3 points), and round enclosure (4 points).
   Then, the mice were transferred into a cage with nestlet for a 30 min
   assessment of the ability of utilizing the nesting material.^[ [191]^42
   ^]

   Marble Burying: Place one mouse into a corner of the cage containing
   marbles, being careful to place the mouse on bedding as far from
   marbles as possible, and place the filter‐top cover on the cage.
   Withhold food and water during the test to allow mouse to remain in the
   cage undisturbed for 30 min. The remove the mouse and return it to its
   home cage after test completion, taking extreme care not to move or
   dislodge the marbles in the process of removing the subject from the
   cage. Task scorers (2–3 peoples) blind to the treatment conditions or
   genotype of the mouse being tested to count the number of marbles
   buried. Score a marble as buried if two‐thirds of its surface area was
   covered by bedding. Average scores for the number of marbles buried for
   each mouse. Retrieved all 20 marbles from the bedding and disposed of
   bedding.^[ [192]^43 ^] ⑨Open field test: To explore locomotion and
   spontaneous activity, mouse behavior was characterized as they freely
   explored an open‐field plastic chamber (40 cm width × 40 cm
   length × 30 cm height) using a video tracking system (Smart 3.0). Two
   months old mice were placed in this arena for 10 min., and total
   distance and time spent in the center region (20 cm × 20 cm) was
   recorded.

RT‐PCR (Real‐Time PCR)

   Total RNA was extracted from the tissues using Trizol (Invitrogen,
   Carlsbad, CA, USA) according to manual instruction. About 60 mg of
   tissue was ground into powder by liquid nitrogen in a 1.5 ml tube and
   then homogenized for 2 min before being rested horizontally for 15 min.
   The mixture was centrifuged for 10 min at 12,000 × g at 4 °C, then the
   supernatant was transferred into a new EP tube with 0.2 ml
   chloroform/isoamyl alcohol (5:1). The mixture was vigorously shaken for
   15 s and then centrifuged at 12,000 × g for 10 min at 4 °C. After
   centrifugation, the upper aqueous phase containing RNA was transferred
   into a new tube with an equal volume of supernatant of isopropyl
   alcohol, then centrifuged at 12.000 × g for 20 min at 4 °C. After
   deserting the supernatant, the RNA pellet was washed three times with
   1 ml 75% ethanol and subsequently centrifuged at 12,000 × g for 3 min
   at 4 °C to remove residual ethanol. The pellet was air‐dry for 5–10 min
   in the biosafety cabinet. Finally, 25µl∼40µl of DEPC‐treated water was
   added to dissolve the RNA. Subsequently, total RNA was qualified and
   quantified using a NanoDrop and Agilent 2100 bioanalyzer (Thermo Fisher
   Scientific, MA, USA). Total RNA was then reverse‐transcribed in a
   reaction volume of 10 ml by the ReverTra Ace qPCR RT Kit (FSQ‐101).
   cDNA was amplified by real‐time quantitative RT‐PCR using SYBR Green
   (Roche) reagent. Samples were assayed in triplicate and GAPDH or
   β‐actin was used as an internal control. Primer sequences used for qPCR
   in this study are summary in Data [193]S7 (Supporting Information).

RNA Sequencing

   To evaluate gene expression differences between Men1‐CcKO and Men1 ^F/F
   mice, single‐end RNA libraries were prepared from hippocampus of
   Men1‐CcKO male mice and the littermate Men1 ^F/F male mice were
   sequenced using BGISEQ‐500 (BGI, Shenzhen, China) and three independent
   biological replicate samples were sequenced for each group. From above
   dataset gene expression profiles of hippocampus of mice were acquired.
   Clean reads were aligned against the Ensembl mice (mm10) genomes using
   HISAT2 software (version 2.0.4). Bowties2 was employed to clean reads
   alignment to reference sequences to calculate gene alignment (version
   2.2.5). The mapped fragments were used to calculate the expression
   levels of different transcripts by RSEM (version 1.2.12). Reads count
   information directly from the files generated by RSEM using a R
   programming language script was taken as input of DESeq2 for
   differentially expressed genes identification. Differentially expressed
   genes defined from the pairwise comparisons had to satisfy two
   selection criteria, including i) fold change larger than 1.2 and ii)
   corresponding adjusted P value less than 0.05. Gene ontology and KEGG
   pathway enrichment analysis were performed using DAVID. The sequencing
   data were deposited in the NCBI Sequence Read Archive (SRA) database
   under the accession: PRJNA826677.

   To evaluate gene expression differences between Men1‐KO and Control
   primary neurons, single‐end RNA libraries were prepared from Men1‐KO
   primary neurons and Control primary neurons which were then sequenced
   using BGI system (BGI, Shenzhen, China) and three independent
   biological replicate samples were sequenced for each group. The gene
   expression profiles of primary neurons were obtained from the above
   dataset. The sequencing data was filtered with SOAPnuke by 1) Removing
   reads containing sequencing adapter;^[ [194]^44 ^] 2) Removing reads
   whose low‐quality base ratio (base quality less than or equal to 15)
   was more than 20%; 3) Removing reads whose unknown base (‘N’ base)
   ratio was more than 5%, afterward clean reads were obtained and stored
   in FASTQ format. The subsequent analysis and data mining were performed
   on Dr. Tom Multi‐omics Data mining system. The sequencing data were
   deposited in the NCBI Sequence Read Archive (SRA) database under the
   accession: PRJNA980296.

ChIP‐Sequencing

   The ChIP DNA libraries of primary neurons were sequenced on BGISEQ‐500.
   Reads were aligned to mm10 using SOAP aligner (version 2.21t) software.
   After format conversion and sorting, duplicate reads were removed by
   rmdup tool from the SAM tools package. To identify significant binding
   site of menin, the mapped sequence reads were processed with MACS2
   against their matching control samples and only peaks with P values
   <10^−5 were kept for further analyses. A Perl script annotate Peaks.pl
   in HOMER package was used to associate ChIP peaks with nearby genes and
   visualization of the location and the shape of called peaks utilizing
   Integrative Genomics Viewer (IGV). Furthermore, all motifs were
   identified by MEME software ([195]http://meme‐suite.org/). The
   sequencing was carried out by BGI Tech (Shenzhen, China). The
   sequencing data were deposited in the NCBI Sequence Read Archive (SRA)
   database under the accession: PRJNA826745 and PRJNA980398.

CUT&Tag

   CUT&Tag was performed with Hyperactive In Situ ChIP Library Prep Kit
   for lllumina (pG‐Tn5) (TD901, Vazyme Biotech) according to the
   manufacturer's instructions. In brief, primary neurons were collected
   and bounded to Concanavalin A‐coated beads. Subsequently, cells were
   resuspended in antibody buffer and incubated with primary antibodies
   against menin antibody (A300‐105A; Bethyl) and secondary antibodies in
   order. The samples were incubated with pA‐Tn5 transposase.
   Aftertransposon activation and tagmentation, DNA was isolated,
   amplified, and purified to construct library. The library for
   sequencing was constructed and VAHTS DNAClean Beads (N411, Vazyme
   Biotech) were used for purification steps. The library was quantified
   with VAHTS Library Quantification Kit for lllumina (VazymeBiotech) and
   sequenced on a lllumina novaseq 150PE. The sequencing was carried out
   by Novogene (Novogene, Beijing, China). The raw sequence data are
   available from GEO accession number: [196]GSE261274.

Dual Luciferase Reporter Activity

   Dual‐luciferase reporter assay was performed with the Dual‐Luciferase
   Reporter Assay System (Promega, E1910). Briefly, primary neurons were
   infected with AAV (shControl AAV or shAtrx AAV) in DIV 6 for knockdown
   Atrx. After culturing until DIV 9, the PGL‐4.10 or PGL‐4.10‐Foxg1
   promoter‐2.5K plasmid was transfected with Lipofectamine 3000 (Thermo
   fisher, L3000008). 24 h later, the infected cell lysates were
   harvested, and relative luciferase activity was determined by
   normalizing the relative luciferase unit readout of the firefly
   luciferase to that of the Renilla luciferase activity measured by
   Varioskan Flash (Thermo fisher).

Statistical Analysis

   All data represent mean ± SEM. Statistical analyses were performed
   using GraphPad Prism 9, Clampfit, or SigmaStat 4 statistical software.
   For statistical significance, experiments with two groups were analyzed
   using two‐tailed Student's tests and one‐way ANOVA. Experiments with
   more than two groups were subjected to one‐way ANOVA and two‐way ANOVA
   (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001). Detailed
   statistical analysis methods for each comparison are indicated in the
   figure legends, including sample size and pre‐processing of data, etc.

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   K.Z. and L.L. contributed equally to this work. J.Z. conceptualized the
   study; K.Z., X.S., S.W., Y.S., and Y.X. prepared and maintained the
   mice; K.Z., L.L., Y.Z., and Y.C. designed and performed morphological
   analysis and biochemical assays; L.L., K.Z., Y.S., L.Z., and S.Y.
   performed behavior tests; X.S. performed electrophysiology experiments.
   K.Z., H.L., and J.Z. wrote the manuscript. Z.C. and T.Y. discussed and
   edited the manuscript. J.Z. supervised the project. All authors
   reviewed and gave final approval to the manuscript.

Supporting information

   Supporting Information
   [197]ADVS-11-2307953-s003.pdf^ (1.8MB, pdf)

   Supplemental Data 1
   [198]ADVS-11-2307953-s008.xlsx^ (5.2MB, xlsx)

   Supplemental Data 2
   [199]ADVS-11-2307953-s006.xlsx^ (6.1MB, xlsx)

   Supplemental Data 3
   [200]ADVS-11-2307953-s005.xlsx^ (50KB, xlsx)

   Supplemental Data 4
   [201]ADVS-11-2307953-s001.xlsx^ (31.5KB, xlsx)

   Supplemental Data 5
   [202]ADVS-11-2307953-s007.xlsx^ (297.7KB, xlsx)

   Supplemental Data 6
   [203]ADVS-11-2307953-s002.xlsx^ (1.8MB, xlsx)

   Supplemental Data 7
   [204]ADVS-11-2307953-s004.xlsx^ (12.7KB, xlsx)

Acknowledgements