Arid1a regulates neural stem/progenitor cell proliferation and differentiation during cortical development

Abstract Objective Neurodevelopmental diseases are common disorders caused by the disruption of essential neurodevelopmental processes. Recent human exome sequencing and genome‐wide association studies have shown that mutations in the subunits of the SWI/SNF (BAF) complex are risk factors for neurodevelopmental diseases. Clinical studies have found that ARID1A (BAF250a) is the most frequently mutated SWI/SNF gene and its mutations lead to mental retardation and microcephaly. However, the function of ARID1A in brain development and its underlying mechanisms still remain elusive. Methods The present study used Cre/loxP system to generate an Arid1a conditional knockout mouse line. Cell proliferation, cell apoptosis and cell differentiation of NSPCs were studied by immunofluorescence staining. In addition, RNA‐seq and RT‐PCR were performed to dissect the molecular mechanisms of Arid1a underlying cortical neurogenesis. Finally, rescue experiments were conducted to evaluate the effects of Neurod1 or Fezf2 overexpression on the differentiation of NSPCs in vitro. Results Conditional knockout of Arid1a reduces cortical thickness in the developing cortex. Arid1a loss of function inhibits the proliferation of radial glial cells, and increases cell death during late cortical development, and leads to dysregulated expression of genes associated with proliferation and differentiation. Overexpression of Neurod1 or Fezf2 in Arid1a cKO NSPCs rescues their neural differentiation defect in vitro. Conclusions This study demonstrates for the first time that Arid1a plays an important role in regulating the proliferation and differentiation of NSPCs during cortical development, and proposes several gene candidates that are worth to understand the pathological mechanisms and to develop novel interventions of neurodevelopment disorders caused by Arid1a mutations.


| INTRODUC TI ON
The cerebral cortex development requires complex sequential processes that have to be precisely orchestrated. 1 During the onset of cortical development, radial glial progenitor cells (RGCs), which derive from neuroepithelial cells, can divide symmetrically to expand the progenitor pool, whereas, in later stages, RGCs divide asymmetrically to directly generate neurons, but most of them indirectly give rise to neurons via intermediate progenitor cells (IPCs). 2,3 The generation of RGCs and IPCs results in the formation of two proliferative zones: the ventricular zone (VZ) and the adjacent subventricular zone (SVZ). 3 During the development of cortex, cortical layering arises in an inside-out manner as neural progenitors proliferate and differentiate into interneurons and projection neurons. 4 Disruptions in the maintenance and/or the balance between proliferation and differentiation of neural progenitors are thought to result in many neurodevelopmental disorders. [5][6][7] ATP-dependent chromatin remodeling plays important roles during cortical neurogenesis. 8 SWI/SNF complex, a class of ATPdependent chromatin remodelers, have been reported to interfere with the structure of chromatin, release of nucleosome-bound DNA, mobilization of DNA along nucleosomes and displacement of histone dimers promoting nucleosome disassembly. [9][10][11] In addition, recent genome-wide studies indicate that it is involved in cellular processes such as cell proliferation and differentiation. 12,13 ARID1A (the AT-rich interaction domain 1A, also known as BAF250a), the largest subunit of the SWI/SNF chromatin remodeling complex, has been reported that its mutations are closely related to Coffin-Siris syndrome (CSS), which is characterized by intellectual disability, growth deficiency and microcephaly. [14][15][16] In early mouse embryos, ablation of Arid1a results in developmental arrest by severely inhibiting self-renewal and promoting differentiation into primitive endoderm-like cells. 17 In the central nervous system (CNS), loss of Arid1a in neural crest cells (NCCs) leads to craniofacial defects in adult mice, including shortened snouts and low set ears, and these defects are more pronounced following homozygous, which is similar to CSS. 18 However, the biological functions and mechanisms of Arid1a in microcephaly and intellectual disability are still unknown.
Here, we generated Arid1a conditional knockout mice and found that deletion of Arid1a in forebrain neural stem/progenitor cells

| Tissues
We accurately obtain mouse embryos in the following ways: after 5 pm on the first day, we put Arid1a f/-: Emx1-cre male mice and Arid1a f/f female mice together. The vagina of the female mouse was examined at 9 am the next day. When the vaginal suppository appeared, it was E0.5. Embryonic brains of E12.5, E14.5 and E16.5 were fixed in 4% paraformaldehyde overnight and dehydrated with 30% sucrose.

| BrdU incorporation analysis
Pregnant mice were given intraperitoneal injections with 100 mg/ kg BrdU based on the weight of the mouse; the concentration of stock BrdU is 10 mg/ml (Sigma; B5002-5G). Embryonic brains were harvested 2 h after BrdU pulsing at E12.5, E14.5 or E16.5.

| Western blot analysis
The total protein of cortical tissues was extracted using RIPA buffer, and the protein concentration was defined using BCA protein assay kit (Biomed, P0012). Western blotting was conducted according to published approaches. 19 Briefly, the membranes were blocked

| RNA-seq and RT-PCR
The total RNA was extracted from E16.5 Arid1a WT or cKO forebrains according to procedures with TRIzol reagent (Invitrogen, 15596018). After quality quantification, the total RNA was con-  Table S1.

| Construction of plasmids
To generate the Neurod1-OE plasmid or Fezf2-OE plasmid, a pair of primers were annealed, and the product was inserted into the

| Lentivirus production
Lentivirus production was performed as described previously. 19 Lentiviral vector and packaging plasmid were co-transfected into 293T cells through polyethylenimine. 20 After transduction, the serum-free medium was replaced by fresh culture medium after 6 h.
The medium was collected at 48, 72 and 96 h post-transduction.
Lentivirus were then concentrated with an ultracentrifuge at 38,000 g for 2 h at 4°C and dissolved in 1 × PBS.

| Microscope imaging
Confocal images were acquired using Zeiss LSM 710 and LSM880 Fast Airyscan confocal microscopes and analysed by ZEN software.

| Statistical analysis
Experiments were conducted in at least three biological replicates for each group. Immunostaining quantification analysis was  Figure 1B). In addition, Arid1a was also expressed in Tbr2 + intermediate progenitors ( Figure 1C). Furthermore, Arid1a was also highly expressed in Tuj1 + neurons ( Figure 1D). Taken together, these results clearly demonstrated that Arid1a is widely expressed in the cortex during forebrain development, suggesting that Arid1a may play a pivotal role in regulating the development of embryonic cerebral cortex.
To examine the roles of Arid1a in cortical development, we generated Arid1a cKO mice by crossing Arid1a f/f mice with Emx1-Cre mice that drive cortex-specific Cre expression beginning at E9.5 ( Figure 1E). The results from immunofluorescence staining and Western blot showed that Arid1a was successfully deleted in the forebrain as its protein level was significantly reduced in Arid1a cKO mice ( Figure 1F-H). Moreover, the expression of Brg1, the central ATPase subunit of SWI/SNF, was significantly down-regulated after Arid1a deletion ( Figure 1I,J). However, the expression of BAF155 and BAF170, another two core subunits of SWI/SNF, was no difference between WT and cKO mice ( Figure 1I,J). Interestingly, Arid1a cKO mice had a reduced cortical thickness from E14.5 to P21 compared to WT mice ( Figure 1K,L), indicating that Arid1a loss of function did affect cortical development.

| Deletion of Arid1a results in abnormal differentiation of deep-layer cortical neurons
A previous report has shown that loss of Arid1a in hematopoietic stem cells impairs the differentiation of both myeloid and lymphoid lineages in hematopoiesis. 23 To test whether the reduced cortical thickness in Arid1a cKO mice was caused by the deficit in neural differentiation of cortical NSPCs, we firstly used the markers Tbr1 and Ctip2 to label layer VI and layer V neurons in the cortex, respectively.
At E12.5, the Arid1a cKO and WT cortices had equal numbers of layer V and layer VI neurons (Figure 2A Next, we examined cell death in the cortex of Arid1a WT and cKO mice by TUNEL assay. We observed a 5-fold increase in TUNEL + cells at E12.5 and E14.5 and a 4-fold increase in TUNEL + cells at E16.5 upon Arid1a knockout ( Figure S1A,B). These results suggested that Arid1a deficiency increased cell death during cortical development, which might be another possible cause for thinner cortex in Arid1a cKO forebrain.

| Arid1a regulates the numbers of RGCs and IPCs in the developing cortex
RGCs and IPCs are two classes of progenitors during cortical neurogenesis. In general, most RGCs divide asymmetrically to give rise to neurons via IPCs. 2 To investigate the role of Arid1a in RGCs and IPCs, we performed immunostaining of PAX6 and Tbr2 to label RGCs and IPCs, respectively. At E12.5, the number of PAX6 + cells in the cortex F I G U R E 1 Loss of Arid1a reduces cortical thickness in the developing cortex. (A) Immunofluorescence staining for Arid1a in the E16.5 cortex. Embryonic brain sections were immunostained with anti-Arid1a antibody. Scale bar, 100 µm. (B-D) Arid1a is expressed in Nestin + NSPCs, Tbr2 + intermediate progenitors and Tuj1 + neurons in the E16.5 cortex. (E) Schematic diagram for the generation of Arid1a conditional knockout mice (cKO). The mouse line with loxP sites inserted on both sides of exon 8 of Arid1a gene was crossed with the mouse line to create Arid1a cKO mice. (F, G) Western blot and quantification results showed that the Arid1a protein level was significantly reduced in the cortex of Arid1a cKO mice. (H) Immunofluorescence staining confirmed that Arid1a is almost undetectable in the cortex of Arid1a cKO mice. (I, J) Western blot and quantification demonstrated that the Brg1 protein level in the cortex of Arid1a cKO mice was significantly reduced, and BAF155 and BAF170 were almost no change. (K) The reduced cortical thickness of Arid1a cKO mice was observed at E16.5, E18.5, P0 and P7 by Nissl staining. (L) Quantification of cortical thickness at different developmental stages. WT, n = 3; cKO, n = 3. Scale bar, 50 µm. *p < 0.05, **p < 0.01, ***p < 0.001.

F I G U R E 2 A decrease in the number of deep-layer cortical neurons in the developing cortex of Arid1a cKO mice. (A, B)
Immunofluorescence staining showed that the absence of Arid1a in NSPCs did not affect the formation of cortical layers V and VI at E12.5. (C, D) At E14.5, the knockout of Arid1a led to a trend of decline in the number of deep-layer neurons. (E, F) At E16.5, the knockout of Arid1a resulted in a significant decrease in the number of deep-layer neurons, while the number of neurons in the superficial layer did not change significantly. (G, H) The number of deep-layer neurons is also reduced in Arid1a cKO mice after birth, and the number of superficial layer neurons remains unchanged. WT, n = 3; cKO, n = 3. Scale bar, 50μm. *p < 0.05, **p < 0.01, ***p < 0.001.
These results indicated that Arid1a may regulate the transformation of RGCs into IPCs and the size of the progenitor cell pool.

| Arid1a promotes the proliferation of RGCs
To examine the role of Arid1a in the proliferation of NSPCs, E12.5, E14.5 or E16.5 pregnant mice were intraperitoneal injected with bromodeoxyuridine (BrdU) to label S phase dividing cells, and ani- To further reveal the role of Arid1a in RGCs and IPCs, the proliferation of RGCs and IPCs was assessed using BrdU labelling 2 h before the pregnant mice were euthanized at E16.5. We found that the proliferation of RGCs (PAX6 + BrdU + ) cells was reduced significantly, whereas the proliferation of IPCs (Tbr2 + BrdU + ) was significantly increased in Arid1a cKO mice compared to that in WT mice ( Figure 4E-H). These data support the idea that Arid1a promotes the proliferation of RGCs but decreases the proliferation of IPCs in the VZ/SVZ at E16.5.

| Arid1a deletion leads to dysregulated expression of genes related to proliferation and differentiation of NSPCs
To further understand the molecular mechanism underlying Arid1a modulating cortical neurogenesis, we performed RNA sequencing to investigate the transcriptome differences in the cerebral cortex between Arid1a WT and cKO mice at E16.5. Genome-wide analyses  Figure 5E). Besides, RT-PCR results showed that the expression of astrocytic genes (ALDH1L1, GFAP and S100β) did not alter after Arid1a cKO ( Figure 5F). Taken together, Arid1a deletion leads to dysregulated expression of genes related to proliferation and differentiation of NSPCs.

| Overexpression of Neurod1 or Fezf2 in Arid1a cKO NSPCs rescues the neural differentiation defect in vitro
To determine the downstream targets of Arid1a, we filtered Fezf2 and Neurod1 out for further exploration from the above expressionvalidated genes associated with proliferation and differentiation of NSPCs. Fezf2 and Neurod1 are well-known regulators of neuronal differentiation. [26][27][28][29] To explore whether Arid1a directly regulated NSPCs differentiation through Neurod1 or Fezf2, we performed binding analysis with publicly available ChIP-seq data for Arid1a from mouse embryonic stem cells 30 and human embryonic stem cells. 31 ChIP-seq analysis indicates that there exist Arid1a-binding peaks on the regions of Neurod1 or Fezf2 loci, suggesting that Neurod1 or F I G U R E 4 Arid1a cKO inhibits the proliferation of RGCs in the developing cortex. (A) Representative images of BrdU (red) and Ki67 (green) immunofluorescence staining of Arid1a WT and cKO brain sections at E16.5. (B) Quantitative analysis of BrdU + Ki67 + cell numbers in the cerebral cortex of Arid1a WT and cKO mice at E16.5. (C) Representative images of PH3 (green) immunofluorescence staining on brain sections of Arid1a WT and cKO mice at E12.5, E14.5 and E16.5. (D) Quantitative analysis of the numbers of PH3-positive cells in the cerebral cortex of Arid1a WT and cKO mice at different developmental stages. (E) Representative images of BrdU (red) and PAX6 (green) immunofluorescence staining of Arid1a WT and cKO brain sections at E16.5. (F) Quantitative analysis of BrdU + PAX6 + cell numbers in the cerebral cortex of Arid1a WT and cKO mice at E16.5. (G) Representative images of BrdU (red) and Tbr2 (green) immunofluorescence staining of Arid1a WT and cKO brain sections at E16.5. (H) Quantitative analysis of BrdU + Tbr2 + cell numbers in the cerebral cortex of Arid1a WT and cKO mice at E16.5. WT: n = 3; cKO: n = 3. Scale bar, 50μm. *p < 0.05, **p < 0.01, ***p < 0.001.
Fezf2 might be the direct targets of Arid1a ( Figure S3A,B). Further analysis of accessibility peaks in Arid1a WT and KO in mouse retinal ganglion cells (RGCs) with publicly available Arid1a ATAC-seq data showed that Arid1a deletion in RGCs led to a dramatic decrease in the activity of neurogenic genes, including Neurod1 and Fezf2 ( Figure S3C). 31 However, publicly available ChIP-seq data for BRG1 from mouse E16.5 cortical neurons demonstrate that there are no

| DISCUSS ION
Neurogenesis is under the precise temporal and spatial control by many transcription factors. 32,33 Abnormal neurogenesis often results in neurodevelopmental disorders. Our study provides the first evidence that Arid1a plays an essential role in embryonic cortical neurogenesis. Conditional knockout of Arid1a decreases the genera- In Arid1a cKO mice, thinner cortex is detected beginning at E14.5. In consistent with this, the markers of deep-layer neurons, Tbr1 and Ctip2, are significantly decreased at E16.5 in Arid1a cKO mice, wherever the markers of upper-layer neurons, Brn2 and Satb2, do not significantly changed (Figure 2A-H). Considering that BAF complex promotes neuronal differentiation in late cortical development, 34 it is possible that Arid1a regulates differentiation of NSPCs along with other BAF subunits. In addition, Arid1a deficiency increases cell death in the developing cortex from E12.5 to E16.5. As most TUNEL-positive cells are located in VZ/SVZ, we speculate that RGCs and/or IPCs are prone to die in Arid1a cKO cortex, which should also contribute to the reduced thickness of Arid1a cKO cortex. We identify that Neurod1 and Fezf2 are functional downstream targets of Arid1a to regulate neural differentiation. As currently there is no commercially available ARID1A antibody for chromatin immunoprecipitation analysis of tissues, we analysed the ChIP-seq and ATAC-seq data for ARID1A deposited in public databases from embryonic stem cells 30,31 and retinal ganglion cells 35 and found that Arid1a has enrichment on Neurod1 and Fezf2 loci which support our identification ARID1A directly regulates Neurod1 and Fezf2 in the nervous system. To our surprise, our results showed that the expression of Brg1, the central ATPase subunit of SWI/SNF, was significantly down-regulated after Arid1a deletion. However, the publicly available ChIP-seq of BRG1 from E16.5 cortical neurons shows Brg1 has no enrichment on the same Neurod1 and Fezf2 loci, suggesting that Brg1 and Arid1a might have different regulatory mechanisms in neural progenitor/stem cells, while they are the main core components in SWI/SNF complex. The regulatory difference between BRG1 and ARID1A also hints that ARID1A function might be independent of the SWI/SNF complexes in neural progenitor/stem cells or in nervous system. Of course, more investigations are further needed to provide in other systems in the future.
Arid1a is a nuclear protein and widely expressed in different human tissues including brain. 36 Indeed, our study showed that Arid1a is ubiquitously expressed in all kinds of brain cells, suggesting its pivotal role not only in NSPCs but also in other cell types.
Moreover, higher expression levels of WNT/β-catenin signal pathway-associated genes such as Wnt2b, Wnt5a, Wnt8b, Nfact4 and Ror2 were observed in Arid1a cKO cortex. Given that WNT/β-catenin signalling is critical for the proper proliferation and differentiation of NSPCs during embryonic development, 37 combined single-cell and spatial transcriptomics are required to dissect the complex regulatory network of ARID1A in cortical development.
In summary, this study demonstrates for the first time that

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

AUTH O R CO NTR I B UTI O N S
C.-M.L. and X.L. involved in conception and design, collection and assembly of data, data analysis and interpretation, manuscript writing, and final approval of manuscript; S.-K.D. and P.-P.L. performed collection and assembly of data.

DATA AVA I L A B I L I T Y S TAT E M E N T
The RNA-seq datasets generated and analysed during the current study have been deposited in the NCBI Sequence Read Archive (SRA). The raw data for E16.5 forebrain RNA-seq reads are accessible through the series accession numbers PRJNA726035 (https:// www.ncbi.nlm.nih.gov/biopr oject/ PRJNA 72603 5/).