Brahma-related gene 1 has time-specific roles during brain and eye development

The development of the central nervous system (CNS) is a complex series of events. In brief, the neural tube is formed around embryonic day 8.5 (E8.5) in mice, followed by the formation of the primary vesicles (prosencephalon, mesencephalon and rhombencephalon) at E9.0 and subsequently of the secondary vesicles (telencephalon, diencephalon, mesencephalon, metencephalon and myelencephalon). In the cerebral cortex, neurogenesis is finished around E18.5, whereas cerebellar development is not finalized until postnatal day 21 (P21) (Chen et al., 2017). Spatiotemporal changes in gene expression in multipotent stem cells are the prerequisites for the development of distinct cell types that make up the CNS (Panamarova et al., 2016). Partially, this is facilitated by epigenetic regulators like the SWItch/sucrose nonfermentable [SWI/SNF, BRG1/BRM-associated factor (BAF)] chromatin remodeling complexes. These are multi-subunit complexes that use ATP hydrolysis to mobilize nucleosomes in order to regulate transcription (Kwon et al., 1994; Wang et al., 1996; Hargreaves and Crabtree, 2011). In total, about 29 genes encode for components that can be combined in a modular manner to form cell type-specific SWI/SNF complexes (Alpsoy and Dykhuizen, 2018; Mashtalir et al., 2018). For example, different SWI/SNF complex compositions have been described for embryonic stem cells (esBAF), neural progenitors (npBAF) and post-mitotic neurons (nBAF) (Lessard et al., 2007; Ho et al., 2009). Mutations in one or more subunits might lead to the expression of an altered SWI/SNF complex with aberrant functions that contribute to the development of diseases such as cancer (Dutta et al., 2017; Sen et al., 2017). Brahma-related gene 1 (Brg1, also known as Smarca4 and Baf190a) is one of the two mutually exclusive ATPase subunits of the SWI/SNF complex (Chiba et al., 1994; Wang et al., 1996). In humans, BRG1 mutations have been found in individuals suffering from intellectual disability disorders such as Coffin-Siris syndrome (Tsurusaki et al., 2012; Holsten et al., 2018; Sekiguchi et al., 2019; Li et al., 2020). Additionally, BRG1 has been proposed to play an important role in the pathogenesis of autism spectrum disorders (De Rubeis et al., 2014; Lim et al., 2017). However, mutations in this gene have also been identified in diverse tumor entities, such as rhabdoid tumors and small cell carcinoma of the ovary of the hypercalcemic type (SCCOHT), indicating a tumor-suppressive role (Hasselblatt et al., 2014; Lang and Hendricks, 2018; Holdhof et al., 2021). The presence of BRG1 mutations in these different diseases suggests a context-specific role of this transcriptional regulator. Mouse models have been used to better characterize the role of Brg1 in normal development and disease. Previous studies revealed that Brg1 is indispensable very early in embryonic development, as homozygous knockout mice die during peri-implantation stage (Bultman et al., 2000). In contrast, a heterozygous Brg1 loss results in a predisposition to mammary tumors and exencephaly (Bultman et al., 2008). Brg1 is constantly expressed during neural development and is involved in processes like self-renewal of neural stem cells (NSCs) and proper formation of all major brain structures (Machida et al., 2001; Lessard et al., 2007; Moreno et al., 2014; Holdhof et al., 2020). Additionally, the ATPase is essential for Schwann cell differentiation but less important for oligodendrocyte differentiation (Weider et al., 2012; Bischof et al., 2015). Altogether, these mouse models support the hypothesis of a context-dependent role of Brg1.

In the present study, we investigated the function of Brg1 during the different stages of embryonic CNS development. For this purpose, we generated Sox2-creER^T2::Brg1^fl/fl mice and induced the loss of Brg1 expression in Sex determining region Y (SRY) box 2 (Sox2)-positive NSCs at defined times by tamoxifen administrations. We chose the period between E6.5 and E14.5 to cover developmental stages starting immediately before neural tube formation until the peak of neurogenesis in the cerebral cortex (Martynoga et al., 2012; Chen et al., 2017). Overall, our study highlights the importance of Brg1 for brain and eye development as well as its time-specific role as a transcriptional regulator.

In order to elucidate time-specific roles of Brg1 in NSCs, we generated Sox2-creER^T2::Brg1^fl/fl mice. Here, Cre-mediated recombination of the loxP sites in the conserved ATPase/Helicase motifs of Brg1 results in an instable mRNA causing a loss of the BRG1 protein (Fig. 1A). We introduced Brg1 deficiency with administration of a single dose of tamoxifen between E6.5 and E14.5 (Fig. 1B), and analyzed their brains by Hematoxylin and Eosin staining between E14.5 and E21.5 (Fig. 1C-M).

Sox2-creER^T2::Brg1^fl/fl embryos presented morphological alterations in the brain with varying penetrance when the loss of Brg1 was induced between E7.5 and E12.5 (Fig. 1C). In contrast, induction of Brg1 deficiency at E6.5 or E14.5 did not alter normal prenatal brain development (Fig. 1C,F,G,L). In detail, all Sox2-creER^T2::Brg1^fl/fl embryos with Brg1 deletion at E7.5 developed a lesion close to the basal part of the cerebrum (Fig. 1H,I,M). Based on morphology, it resembled neural retina (NR) (Fig. 1Hii,Ii) that was altered in the embryonic eye of E18.5 mutant mice (Fig. 1I, arrow). About 63% of these mice showed additional abnormalities in the subventricular zone (SVZ), often with rosette-like structures (Fig. 1Iii,M). After Brg1 deprivation at E8.5, Sox2-creER^T2::Brg1^fl/fl embryos similarly showed features like the NR-like lesion (70%) or the involvement of the SVZ (80%) (Fig. 1M). However, in case of an induced Brg1 loss at E9.5, the alterations were mainly located in the SVZ (88%) and the cerebral cortex (12%). Furthermore, the embryos presented with hydrocephalus (12%) or with a combination of these phenotypes (Fig. 1J,K,M). The abnormalities in the SVZ usually involved rosette-like structures, and the layering of the cerebral cortex seemed disturbed, as shown in Hematoxylin and Eosin stains at E14.5 or E18.5 (Fig. 1Ji,Ki,Kii). Furthermore, the NR presented with rosettes in some animals (Fig. 1Jii). Initiation of Brg1 loss at E10.5 or E12.5 primarily resulted in a disrupted SVZ with rosette-forming structures (73% and 100% of embryos, respectively), very similar to those observed after Brg1 disruption at E9.5 (Fig. 1M).

In order to examine whether the brain regions with architectural disruptions were formed by cells that had lost Brg1, we used immunohistochemistry (IHC) to detect expression of the protein at the cellular level (Fig. S1). Overall, even though there were different morphological alterations found in the embryonic brain after loss of Brg1 at E7.5 and E9.5 in Sox2-expressing NSCs, the abundance of BRG1-negative cells (Fig. S1F-N, yellow arrow) was similar. BRG1-deficient cells were randomly scattered in the neocortex and the thalamic regions, whereas the SVZ, including the rosette-like structures, was mainly BRG1 competent. The NR-like region close to the basal cerebrum in embryos that had lost Brg1 at E7.5 harbored many BRG1-negative cells (Fig. S1J).

Taken together, Brg1 deficiency induced between E7.5 and E12.5 in Sox2-expressing cells causes architectural alterations in the embryonic brain. The type of abnormality depends on the time of Brg1 deprivation, as the NR-resembling lesion close to the basal cerebrum only occurred after Brg1 loss at E7.5 or E8.5. However, alterations in the SVZ were found after Brg1 deprivation between E7.5 and E12.5, indicating that the cells forming the SVZ are dependent on proper Brg1 expression for a longer period.

Next, we were interested in whether the morphological alterations in the brains of Sox2-creER^T2::Brg1^fl/fl embryos were specific to Brg1 or the result of a non-functional SWI/SNF complex. We generated Sox2-creER^T2::Ini1^fl/fl embryos and investigated the influence of an Ini1 deficiency induced at E7.5 or E9.5, as INI1 is another subunit of the majority of SWI/SNF complexes (Alpsoy and Dykhuizen, 2018). After induction of Ini1 deprivation at E7.5, brains appeared normal at E14.5 and E18.5 (Fig. 2A-D). We did not observe any alterations in the SVZ or the neocortex, nor did we detect any NR-resembling structure (Fig. 2Ci,Cii,Di,Dii) as we did in the brains of Sox2-creER^T2::Brg1^fl/fl embryos (Fig. 1Hii,Ii,Iii). In contrast, loss of Ini1 at E9.5 caused formation of rosette-like structures in the SVZ in 7/12 embryos (58%) that resembled those found in the SVZ of Brg1-deficient embryos (Fig. 1Ji,Kii and Fig. 2Ei,Fii). The neocortex appeared without major alterations in the analyzed animals (Fig. 2Fi). Of note, the pattern of INI1-negative cells in Sox2-creER^T2::Ini1^fl/fl embryos (Fig. S1O-Z) at E18.5 was similar to that of BRG1 deprived cells in Sox2-creER^T2::Brg1^fl/fl embryos (Fig. S1B-N).

Overall, we suggest that some morphological alterations that we observed upon loss of Brg1 in Sox2-creER^T2::Brg1^fl/fl embryos were caused by a non-functional SWI/SNF complex. However, other abnormalities, especially the NR-resembling cell accumulation, were specific to the lack of Brg1.

Subsequently, we examined whether the phenotypic alterations we observed between E14.5 and E18.5 were a direct consequence of the BRG1 protein loss. Therefore, we examined E10.5 and E12.5 embryos after Brg1 deletion at E7.5 and E12.5, respectively. Macroscopically, Sox2-creER^T2::Brg1^fl/fl embryos of both stages (Fig. 3C,Q) appeared normal in comparison with their Brg1^fl/fl littermates (Fig. 3A,O). Tamoxifen-induced Brg1 loss at E7.5 caused no obvious morphological alterations in Hematoxylin and Eosin stains (Fig. 3B,D,E,G) even though the protein was lost in a substantial number of cells (Fig. 3F,H). As the number of BRG1 negative cells was higher compared with E18.5 brains (Fig. S1), we speculated that Brg1 deficiency caused a decrease in proliferation and/or an increase in apoptosis. Therefore, we stained for phosphorylated histone H3 (pHH3) and cleaved Caspase 3 (clCasp3) to examine the number of proliferating and apoptotic cells, respectively. The proportion of pHH3-positive cells was equal in Brg1^fl/fl (Fig. 3I) and Sox2-creER^T2::Brg1^fl/fl brains (Fig. 3K,M). In contrast, we observed a significant increase in clCasp3-expressing cells in the CNS of Brg1-deprived animals (Fig. 3L) compared with controls (Fig. 3J,N). We compared proliferation and apoptosis in different brain regions and determined that the changes in apoptosis were especially pronounced in the midbrain regions (Fig. S2A-G). Furthermore, following Brg1 loss at E7.5, we observed many clCasp3-positive cells in regions harboring increased numbers of BRG1-negative cells in brains of Sox2-creER^T2::Brg1^fl/fl animals at E12.5 (Fig. S2H-O).

After loss of Brg1 at E9.5, there were no morphological differences detectable in Hematoxylin and Eosin stains of Brg1^fl/fl (Fig. 3P,S) and Sox2-creER^T2::Brg1^fl/fl brains (Fig. 3R,U). There were few BRG1-deficient cells in the ganglionic eminence but no significant alterations in proliferation or apoptosis, as indicated by quantification of pHH3-expressing (Fig. 3W,Y,AA) or clCasp3-expressing (Fig. 3X,Z,AB) cells. In conclusion, our data suggest that introducing Brg1 deficiency in Sox2-expressing cells at E7.5 results in an increase of apoptosis.

The cell accumulation attached to the basal cerebrum present in Sox2-creER^T2::Brg1^fl/fl embryos after Brg1 loss at E7.5 resembled NR in Hematoxylin and Eosin stains (Fig. 1Ii). Therefore, we compared the NR-like structure with NR of healthy controls by IHC marker expression (Fig. 4A-N). At E18.5, mainly two retinal layers are visible, the neuroblast layer (NBL) and the ganglion cell layer (GCL). Both were easily recognizable in Hematoxylin and Eosin stains of retina in healthy controls (Fig. 4B). Based on morphology, the NR-like in Sox2-creER^T2::Brg1^fl/fl embryos showed similar structures resembling NBL and GCL (Fig. 4I). Accordingly, SOX2, a marker for retinal progenitor cells (Saha et al., 2018) was found in both the NBL of controls and the NBL-resembling structure in Brg1-deprived animals (Fig. 4C,J). Similarly, expression of paired box protein 6 (PAX6), orthodenticle homeobox 2 (OTX2), oligodendrocyte transcription factor 2 (OLIG2) and neuron-specific class III beta-tubulin (TUJ1) was comparable in the NR of controls (Fig. 4D-G) and the NR-resembling structure of Brg1-deficient mice (Fig. 4K-N).

Based on this comparison, we concluded that the structure at the basal cerebrum was in fact expanded NR. Next, we investigated at which embryonic stage the enlarged NR occurred. Therefore, we examined embryonic eyes between E12.5 and E18.5 (Fig. 4O-V). First alterations became visible at E16.5 in Sox2-creER^T2::Brg1^fl/fl embryos, with the NR being present in multiple folds. Hence, we concluded that Brg1 deficiency at E7.5 results in the formation of an expanded NR, which occurs after E14.5.

In order to characterize the fate of Brg1 deficient cells, we generated Sox2-creER^T2::lslRFP^fl/fl and Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl mice. The lslRFP transgene in these mice harbors a ‘transcriptional stopper’ element flanked by loxP sites upstream of the red fluorescent protein (RFP)-encoding sequence (Luche et al., 2007). Consequently, all cells and their progeny, in which the Cre recombinase has been activated by tamoxifen, are marked by RFP expression (Fig. 5A). RFP-expressing cells were isolated using a fluorescence-activated cell sorting (FACS) and used for neurosphere assays.

Brg1 loss significantly decreased the yield of cells that were obtained from E14.5 brains (Fig. 5B). After tamoxifen exposure at E7.5 and E9.5, only 7.4% and 2.8% of cells from Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl brains were RFP positive, respectively. In comparison, 15.4% and 8.0% RFP-positive cells were isolated from Sox2-creER^T2::lslRFP^fl/fl embryos after tamoxifen treatment at E7.5 and E9.5, respectively. To validate the efficiency of the system, we performed BRG1 immunofluorescence (IF) immediately after FACS (Fig. 5C). For both time points of tamoxifen injections, only 48-86% of RFP-positive cells were negative for BRG1 (data not shown). This suggested that the efficiency of the Cre enzyme is reduced, when two floxed transgenes (here Brg1^fl/fl and lslRFP^fl/fl) are recombined simultaneously.

Next, we cultured the RFP-positive cells to investigate their ability to form neurospheres. After 7 days, neurospheres were present in the wells of all four conditions (Fig. 5D). However, in wells with cells derived from Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl brains, there were significantly fewer spheres found (Fig. 5E). After tamoxifen at E7.5, the number of neurospheres was reduced by 92% and after tamoxifen at E9.5 it was reduced by 84%. Moreover, IF staining revealed that the neurospheres contained almost exclusively BRG1-positive cells (Fig. 5F). This suggests that Brg1-expressing cells are selected in culture, whereas Brg1 deficiency hinders the formation of neurospheres. Taken together, loss of Brg1 at either E7.5 or E9.5 decreases the abundance of the respective cells at E14.5. Furthermore, these cells are not able to form neurospheres in vitro.

We performed RNA-sequencing to analyze global changes in gene expression upon loss of Brg1. We used the same experimental set-up as for the neurosphere assay and isolated the RNA immediately after FACS (Fig. 5A).

Unsupervised hierarchical clustering of significant differentially expressed genes (DEGs) defined by FDR<0.1 and log2 fold change >±0.6 revealed four distinct clusters (Fig. 5G). These clusters corresponded with the four treatment groups: Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl and Sox2-creER^T2::lslRFP^fl/fl embryos, with Brg1 disruption either at E7.5 or at E9.5. We identified 172 and 18 downregulated, as well as 115 and 97 upregulated genes upon Brg1 loss at E7.5 and E9.5, respectively (Fig. S3A,B). In order to interpret these results functionally, we performed gene ontology (GO) term enrichment analyses. Significantly enriched terms involving DEGs with a log2 fold change >±0.6 were visualized as GOchords (Fig. 5H,I). Functionally related terms were summarized as their lowest common denominator derived from the 10 most significant terms per category (i.e. ‘Molecular Function’, ‘Cellular Component’ and ‘Biological Process’). In the GOchords, the respective GO terms are shown on the right half of the circle, connected by chords to the involved DEGs, which are depicted on the left side of the circle. Loss of Brg1 at both investigated time points (E7.5 and E9.5) resulted in downregulation of the GO terms ‘cation channel complex’ (GO: 0034703) and ‘metal ion transmembrane transporter activity’ (GO: 0046873). The other GO terms downregulated upon Brg1 loss differed between the groups. Still, all downregulated GO terms were related to the function of neurons; in particular, to their ability to form synaptic cell-cell communications. This indicates that NSCs are dependent on Brg1 to give rise to functional neurons, irrespective of the time point.

Using this method to identify enriched GO terms upon Brg1 loss at E7.5 resulted in only two upregulated GO terms: ‘antioxidant activity’ (GO: 0016209) and ‘extracellular matrix structural constituent’ (GO: 00052011) (Fig. 5I). However, when Brg1 deficiency was introduced at E9.5, several GO terms were upregulated, including processes involved in cell cycle, vascular development and extracellular matrix (ECM) functions. Next, we performed Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis. The results matched to the GO term analysis, because terms like ‘GABAergic synapse’ or ‘Focal adhesion’ were altered irrespective of the time of Brg1 loss, whereas the Brg1 deficiency induced at E9.5 altered additional pathways, including ‘cell cycle’, ‘DNA replication’ or ‘PI3K-Akt signaling pathway’ (Fig. S4).

Genes encoding collagens were significantly upregulated due to Brg1 deprivation at E7.5 and at E9.5. Additionally, after Brg1 loss at E9.5, further genes encoding other ECM components such as vitronectin, fibronectin 1, heparan sulfate proteoglycan 2 (Hspg2, perlecan), laminin α4 (Lama4) and laminin γ3 (Lamc3) were significantly upregulated. Concordantly, as laminins and integrins interact (Belkin and Stepp, 2000), integrin α1 expression was also enhanced. Other upregulated DEGs involved genes such as the stem cell marker nestin, Bmp7 or Tgfβ receptor II (Tgfbr2). Consistently, clustering of KEGG pathway TGFβ signaling pathway (mmu04359) genes was according to genotype (Fig. S5A), but did not result in enhanced phosphorylated Smad1,5,8 (p-Smad 1,5,8) reactivity in IF (immunofluorescence staining) (Fig. S5B-F).

We also used the transcriptome data to examine whether and how loss of Brg1 influenced the expression of other SWI/SNF complex members (Fig. S6) present in the complex in NSCs (Lessard et al., 2007). Expression of the alternative ATPase subunit (Smarca2) was significantly (FDR<0.1) lower, irrespective of the time of Brg1 loss.

Taken together, Brg1 deprivation at either E7.5 or E9.5 results in the downregulation of genes related to the function of neurons and upregulation of genes and GO terms associated with the ECM. However, Brg1 loss at E7.5 mainly causes enhanced expression of collagen-encoding genes, whereas Brg1 deficiency at E9.5 additionally increases a variety of other ECM-related genes.

As analysis of GO term enrichment identified only two upregulated GO terms that merely included 15/115 (13%) of the DEGs upon Brg1 loss at E7.5, we took a closer look at the individual upregulated DEGs (Fig. 6A). We recognized many genes that are expressed during eye development, including fibroblast growth factor 15 (Fgf15), pituitary homeobox 2 (Pitx2), Otx2, Iroquois homeobox gene 5 (Irx5), POU domain class 4 transcription factor 1 (Pouf4f1) and LIM homeobox 9 (Lhx9) (Cheng et al., 2005; Danno et al., 2008; Gage et al., 2008; Badea and Nathans, 2011; Bharti et al., 2012; Balasubramanian et al., 2014). Fgf15, Otx2 and Lhx9 counts were increased more than twofold, Pou4f1 expression showed a threefold increase, and Irx5 and Pitx2 counts were elevated almost by a factor of four in Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl- compared with Sox2-creER^T2::lslRFP^fl/fl-derived cells.

Increased FGF signaling induces transdifferentiation of RPE cells to NR in vitro (Zhao et al., 1995). Furthermore, in a mouse model investigating RPE development, Fgf15 was specifically upregulated upon RPE to NR transdifferentiation (Bharti et al., 2012). As we isolated RFP-positive cells from the entire E14.5 brain, including the enlarged NR attached to the basal cerebrum, we hypothesize that the mutant NR was the result of a RPE to NR transdifferentiation caused by increased FGF signaling. The RPE is a single layer of cells adjacent to the outer layers of the NR (Fig. 6B). In Sox2-creER^T2::Brg1^fl/fl mice with Brg1 deletion at E7.5, the RPE was visible only in some regions (Fig. 6D) and seemed to transition into NR (white arrow), supporting the hypothesis of a transdifferentiation. BRG1- negative cells were predominantly found outside the enlarged NR in locations where the RPE is normally present (Fig. 6C,E, blue arrows). Furthermore, we found single pigmented cells close to BRG1-negative cells (Fig. 6E, blue arrowheads), indicating that BRG1 loss in the RPE might cause transdifferentiation into NR.

In order to validate whether the RPE was derived from those cells, in which we had activated the Cre recombinase at E7.5, we examined the RFP signal in the previously used fate-mapping mice (Fig. 5A). RFP-positive cells were found in the brain, the NR and the RPE at E18.5 (Fig. S7). By confirming that the RPE was also affected by the Cre-mediated recombination using the fate-mapping approach, we further support the hypothesis that the enlarged NR might result from a RPE to NR transdifferentiation.

As we suspected that increased FGF15 signaling was responsible for the formation of the enlarged NR, we performed IF demonstrating that the mutant NR expressed high amounts of FGF15 protein (Fig. 6I). Additionally, alterations in the mitogen-activated protein kinase (MAPK) signaling pathway as a potential consequence of increased FGF15 signaling were present (Fig. S8A). Phosphorylated ERK 1,2 (p-ERK 1,2) showed no activity in the NR of healthy controls (Fig. S8B) but displayed patchy reactivity in the enlarged NR (Fig. S8C). Likewise, we found changes in the MAPK signaling pathways after loss of Brg1 at E9.5 (Fig. S8A) accompanied by a high reactivity for p-ERK 1,2 in the rosette-like structures of the SVZ (Fig. S8E). This indicates that Brg1 plays a general role in the regulation of MAPK signaling.

To further support the hypothesis that the altered NR was the result of a transdifferentiation of RPE to NR, we stained for ezrin, a marker for the microvilli of the RPE (Bonilha et al., 2006) (Fig. 6G,J). Ezrin identified normal appearing RPE (Fig. 6J, green arrow), and Ezrin expression was also detectable in cells arranged in multiple layers (green arrowhead). Finally, we stained for NR2F1 (Coup-TFI), which is essential for proper eye development (Tang et al., 2010), and determined that the enlarged NR in Sox2-creER^T2::Brg1^fl/fl mice (Fig. 6K) had a similar NR2F1 expression pattern to the NR of E11.5 mice (Tang et al., 2010), suggesting that it was more immature than its E18.5 counterpart (Fig. 6H).

Based on these results, we hypothesized that Brg1 loss at E7.5 in Sox2-expressing cells resulted in the dysregulation of genes that are essential for normal eye development. Consequently, the RPE, at least parts of it, did not develop normally but instead gave rise to an immature NR.

The presence of BRG1 mutations in several different human diseases suggests a context-dependent role of this gene (Hasselblatt et al., 2014; Lang and Hendricks, 2018; Sekiguchi et al., 2019). Furthermore, different mouse models indicate that Brg1 is essential for several steps of brain ontogenesis (Lessard et al., 2007; Matsumoto et al., 2016). Here, we investigated how an early Brg1 loss in NSCs affected neural development. Furthermore, we examined whether Brg1 has time-specific roles and deleted the protein at different time points during embryonic development. Sox2 is the first intrinsic marker for neural identity during development (Adnani et al., 2018). We therefore used Sox2-creER^T2 mice to induce Brg1 deficiency between E6.5 and E14.5.

Brg1 deficiency altered embryonic brain development only when it was induced between E7.5 and E12.5, but not at E6.5 or E14.5. The deletion between E7.5 and E12.5 caused diverse and time point-dependent architectural abnormalities in the brains and eyes of Sox2-creER^T2::Brg1^fl/fl embryos. Exclusively at E7.5 and E8.5, Brg1 loss caused the formation of an enlarged NR. Between E8.5 and E10.5, Brg1 deprivation led to alterations in the neocortex in a fraction of animals. In contrast, the architecture of the SVZ was affected in the majority of animals irrespective of the time of tamoxifen-induced Brg1 deficiency and was also observed in Sox2-creER^T2::Ini1^fl/fl embryos after an Ini1 loss induced at E9.5. These observations suggest that the NSCs of the SVZ are particularly sensitive towards alterations in the SWI/SNF complex at E9.5. However, the phenotypes outside the SVZ observed between E7.5 and E12.5 in Sox2-creER^T2::Brg1^fl/fl embryos imply Brg1-specific roles in eye and brain development.

As the SVZ is one of the germinal zones harboring mainly NSCs, it seemed likely that this region was highly affected by the NSC-specific loss of Brg1. Unexpectedly, the SVZ, including the rosette-like structures, was mainly composed of BRG1-positive cells and BRG1-deficient cells were found only occasionally. This might be explained by the fact that cells essential for the integrity of the SVZ, e.g. by providing exocrine signals, were lacking. Analysis of pHH3 and clCasp3 expression 3 days after Brg1 loss at E7.5 revealed no changes in proliferation but an increase in apoptotic cells. Concordantly, using a fate-mapping approach, we show that, upon deletion of Brg1 at either E7.5 or E9.5, less BRG1-negative cells are found at E14.5. Furthermore, results from our neurosphere assay imply that Brg1-deprived NSCs are unable to survive and proliferate in vitro. Previous studies have already demonstrated that Brg1 knockdown can result in enhanced apoptosis (Deng et al., 2015; Singh et al., 2016). Therefore, we hypothesize that Brg1 loss causes an increase in apoptosis, causing a lack of precursor cells in the SVZ.

We also identified changes in gene expression responsible for the observed morphological alterations. Therefore, we did not only observe distinct changes in gene expression between Brg1-negative and Brg1-competent cells but also between cells with Brg1 loss induced either at E7.5 or at E9.5. Based on these data, we conclude that Brg1 facilitates important functions regulating transcription in Sox2-expressing NSCs. In addition, these data further support the hypothesis that the ATPase has distinct roles at different stages during early neural development. Enrichment analysis of GO terms revealed that GO terms specifically related to neuronal functions were downregulated after Brg1 loss at E7.5 or E9.5. In particular, the functionality of synaptic cell-cell interactions seemed to be disturbed upon ablation of Brg1 expression. As the morphology of the pyramidal neurons in the cortices of hGFAP-cre::Brg1^fl/fl mice was severely altered, we hypothesized that Brg1 is not only important in murine NSCs marked by hGFAP (Holdhof et al., 2020) and in Xenopus laevis neurogenesis (Seo et al., 2005), but also in Sox2-expressing murine NSCs in order to give rise to functional neurons. Furthermore, disturbances in synaptic communication due to Brg1 deficiency might explain the lack of BRG1-negative cells in the brain regions with morphological abnormalities. We hypothesize that Brg1-deficient cells transmit pathological signals to their neighboring (Brg1-competent) cells, causing the observed architectural alterations in a paracrine manner.

In addition, enrichment analysis of upregulated GO terms revealed the enhancement of terms related to the ECM, especially after Brg1 loss at E9.5. As the ECM contributes to maintaining stem cell identity, e.g. by providing a reservoir of growth factors, BMPs or TGFβ, the alterations in ECM-related terms might be another cause for the disrupted SVZ (Barros et al., 2011; Brizzi et al., 2012). We recognized that the expression of collagen-encoding genes was upregulated upon Brg1 loss at either E7.5 or E9.5. Notably, induction of Brg1 deficiency at E9.5 additionally resulted in enhanced expression of other ECM-associated genes, many of which are found in fractones, a basement membrane-resembling structure found exclusively in the SVZ (Mercier et al., 2002; Kerever et al., 2007; Mercier, 2016). Even though fractones have only been described in the adult SVZ so far, to our knowledge, there is no evidence that they are not already present in the embryonic SVZ. They provide NSCs with growth factors, supporting their stem cell function. Therefore, a misbalance in the expression of fractone-associated genes might interfere with facilitating these functions. To sum up, the dysregulation of ECM-associated genes might also contribute to the SVZ phenotype observed in Sox2-creER^T2::Brg1^fl/fl embryos, especially when Brg1 deficiency was induced at E9.5. Furthermore, the ECM also facilitates important functions in NSC differentiation, neuronal migration, the formation of axonal tracts, and the maturation and function of synapses (Barros et al., 2011). Therefore, the upregulation of ECM-associated genes and GO terms might also have an impact on these processes.

Of note, Brg1 loss at E7.5 resulted in more downregulated GO terms related to neuronal functions, whereas Brg1 deprivation at E9.5 instead caused the enhancement of GO terms, including those involved in ECM composition. In addition to the differences we observed in histology, this further highlights the time-specific role of Brg1. Still, as we observed a reduced efficiency of the Cre recombinase in Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl mice, we cannot rule out the possibility that we might have missed some DEGs contributing to the observed phenotypes.

In the developing cerebral cortex of Sox2-creER^T2::Brg1^fl/fl embryos, BRG1-deficient cells were scattered randomly across all layers. However, only in a fraction of embryos, Brg1 deprivation resulted in disrupted layering in the deeper layers (VZ/SVZ) – the site that harbors neural progenitors. Maturation of neural progenitors into neurons is accompanied by a switch in the composition of the SWI/SNF complex. This contributes to distinct changes in gene expression needed for correct differentiation and migration of neurons during cortical development (Narayanan and Tuoc, 2014; Son and Crabtree, 2014; Elsen et al., 2018; Sokpor et al., 2018). However, as only the regions where neural progenitors reside were affected in Sox2-creER^T2::Brg1^fl/fl embryos, this phenotype might not reflect issues in cortical development. It is rather another example that the knockout of Brg1 caused disruption of the NSC niche, as already discussed for the SVZ.

One of the most striking phenotypes in the Sox2-creER^T2::Brg1^fl/fl embryos was the enlarged mutant NR. It exclusively occurred after Brg1 loss between E7.5 and E8.5, and was first detectable at E16.5. The overall morphology as well as the expression patterns of marker proteins was very similar to the physiological NR of healthy littermates. Transcriptome analysis revealed that genes that are expressed during normal eye development were upregulated upon Brg1 loss at E7.5. Pouf41, for example, was one of the DEGs with the highest log2 fold change and is physiologically expressed in the embryonic NR from E12.5 onwards (Pan et al., 2005). Strikingly, Fgf15 expression was also significantly upregulated upon Brg1 loss at E7.5, and the enlarged NR showed high immunoreactivity for the FGF15 protein. In another mouse model, Fgf15 transcription was enhanced in case of RPE to NR transdifferentiation (Bharti et al., 2012). Therefore, we hypothesized that the enlarged NR was also formed as a consequence of RPE to NR transdifferentiation in our model. The results of the fate-mapping experiment revealed that both NR and RPE harbor progenies of cells, in which we had activated Cre at E7.5. In addition, presumptive NR and RPE express Sox2 (Smith et al., 2009). Consequently, we cannot rule out the possibility that the enlarged NR is (partially) the result of uncontrolled proliferative activity of the NR itself. However, as we visually observed regions in which the RPE directly transitioned into the NR, it is more likely that loss of Brg1 causes a transdifferentiation of the RPE into NR, and that Brg1 plays an essential role in the maintenance of the proper RPE.

During eye development, RPE and NR originate from the same precursors and fate is determined by several extrinsic and intrinsic factors (Fuhrmann, 2010). Balance of these factors is crucial for proper eye development, as demonstrated by several mouse models, in which dysregulation of genes, such as Mitf, β-Catenin, transcription factor AP-2α or Yes-associated protein (Yap), resulted in RPE to NR transdifferentiation (West-Mays et al., 1999; Bumsted and Barnstable, 2000; Fujimura et al., 2009; Kim et al., 2016). So far, Brg1 has not been associated with this process. However, the interaction of BRG1 itself or the SWI/SNF complex with many of the before-mentioned genes has already been reported (Laurette et al., 2015; Zhu et al., 2015; Chang et al., 2018). Additionally, Brg1 has been described to be involved in retinogenesis (Aldiri et al., 2015). Finally, we occasionally detected rosette-like structures in the retina of Sox2-creER^T2::Brg1^fl/fl embryos with Brg1 loss at E9.5. This indicates that, similar to the situation in the SVZ, certain cells that had lost Brg1 earlier in development died, which in turn disrupts cell-cell interactions and causes the formation of rosettes. In line with these findings, Brg1 has been described to be essential for eye development in zebrafish. There, a Brg1 null mutation results in a NR without proper lamination and in disruption during the differentiation of retinal cells (Gregg et al., 2003). Finally, individuals with Coffin-Siris syndrome carrying BRG1 germline mutations occasionally present with ophthalmological anomalies, including microphthalmia, strabismus or retinal dystrophy (Kosho et al., 2013; Errichiello et al., 2017; Cappuccio et al., 2019). As we did not observe any regions resembling enlarged NR in Sox2-creER^T2::Ini1^fl/fl mice, we assume that BRG1 and not the SWI/SNF complex as a whole is essential for repressing retinal fate. However, the temporal window seems very narrow, as we did not observe RPE-to-NR transdifferentiation when Brg1 loss was induced earlier than E7.5 or later than E8.5.

In conclusion, proper expression of Brg1 and Ini1 in NSCs is essential for the integrity of the SVZ. In addition, Brg1 is also essential for normal cortical development and maintenance of the RPE. As BRG1 is part of the SWI/SNF complex that regulates gene expression, these phenotypes are caused by dysregulation of cell type- and time-specific transcription.

Brg1^fl/fl, Ini1^fl/fl (JAX #004596), lslRFP^fl/fl and Sox2-creER^T2 (JAX #17593) mice have previously been generated and described (Sumi-Ichinose et al., 1997; Roberts et al., 2002; Indra et al., 2005; Luche et al., 2007; Arnold et al., 2011). They were crossed to generate Sox2-creER^T2::Brg1^fl/fl, Sox2-creER^T2::Ini1^fl/fl, Sox2-creER^T2::lslRFP^fl/fl, Brg1^fl/fl::lslRFP^fl/fl and Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl mice. Female Brg1^fl/fl, Ini1^fl/fl, lslRFP^fl/fl and Brg1^fl/fl::lslRFP^fl/fl mice were bred with male Sox2-creER^T2::Brg1^fl/fl, Sox2-creER^T2::Ini1^fl/fl, Sox2-creER^T2::lslRFP^fl/fl and Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl mice. A single dose of 1 mg tamoxifen in 50 µl corn oil was injected intraperitoneally into pregnant mice at E6.5, E7.5, E8.5, E9.5, E10.5, E12.5 or E14.5 to induce Cre recombinase activity. The observation of a post-coital plug was defined as E0.5. The fate-mapping mice were generated for better characterization of Brg1-deficient cell populations: mothers of Sox2-creER^T2::lslRFP^fl/fl (controls) and Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl (mutants) received tamoxifen at either E7.5 or E9.5. Embryos were collected by Caesarean section. Ear or tail biopsies were used for genotyping via PCR. The following primers were used to detect the cre transgene (448 bp), the Brg1 wild type (241 bp) or floxed (387 bp) allele, the Ini1 wild type (203 bp) or floxed (300 bp) allele and the lslRFP wild type (600 bp) or mutant (250 bp) allele: 5′-TCCGGGCTGCCACGACCAA-3′ (cre forward), 5′-GGCGCGGCAACACCATTTT-3′ (cre reverse), 5′-GTCATACTTATGTCATAGCC-3′ (Brg1 forward), 5′-GCCTTGTCTCAAACTGATAAG-3′ (Brg1 reverse), 5′-TAGGCACTGGACATAAGGGC-3′ (Ini1 forward), 5′-GTAACTGTCAAGAATCAATGG-3′ (Ini1 reverse), 5′-AAAGTCGCTCTGAGTTGTTAT-3′ (lslRFP forward), 5′-GCGAAGAGTTTGTCCTCAACC-3′ (lslRFP mutant reverse) and 5′-GGAGCGGGAGAAATGGATATG-3′ (lslRFP wild-type reverse). Mice were kept on a 12 h dark/light cycle and water and food was available ad libitum. Animals of both sexes were used for the experiments. The mice belonged to the species Mus musculus and were maintained on a C57Bl/6J background. The experimental procedures were approved by the Government of Hamburg, Germany (113/16, 19N099/2019) and the Government of Nordrhein-Westfalen, Germany (84-02.04.2015.A088).

For Hematoxylin and Eosin as well as immunohistochemical stains, entire embryo heads (E12.5-E18.5) or whole embryos (E10.5) were fixed in 4% formaldehyde for 24 h. Afterwards, they were dehydrated, embedded in paraffin in a frontal orientation and 4 µm sections were cut according to established protocols. Hematoxylin and Eosin stains were performed according to standard protocols. Immunohistochemical stains were prepared using the Ventana System (Roche Diagnostics) or the DCS SuperVision 2 Kit (DCS Diagnostics) according to the manufacturers' instructions. The following antibodies were used: rabbit anti-Brg1 (Abcam, ab110641, 1:25), rabbit anti-cleaved caspase 3 (cl. Casp3; Asp175; Cell Signaling, #9664, 1:100), mouse anti-Ini1 (BD Biosciences, 1:50), mouse anti-phospho-Histone H3 (pHH3, Cell Signaling, #9706, 1:200), rabbit anti-Ki67 (Abcam, ab15580, 1:100), rabbit anti-Olig2 (Millipore, AB9610, 1:200), mouse anti-Otx2 (ThermoFisher Scientific, 1H12C4B5, 1:2000), rabbit anti-Pax6 (Biolegend, 901301, 1:1000), rabbit anti-Sox2 (Abcam, ab97959, 1:200) and mouse anti-TuJ1 (Biolegend, 801201,1:200). For all images, at least n=3 embryos were analyzed.

For immunofluorescence, entire embryo heads were fixed in 4% formaldehyde for 24 h and processed according to standard protocols. For the detection of Ezrin, FGF15, NR2F1, phosphorylated ERK1/2 (p-ERK1/2) or phosphorylated SMAD1/5/8 (p-SMAD1/5/8), heads were dehydrated, embedded in paraffin wax in a frontal orientation and cut into 4 µm sections. Antigen retrieval was achieved by cooking the slides in citrate buffer (Ezrin, NR2F1 and p-SMAD1/5/8) or Tris-EDTA (FGF15 and p-ERK1/2) for 20 min. For the detection of RFP, the tissue was sequentially placed in 10%, 20% and 30% sucrose in PBS, and then placed into optimal cutting temperature compound (OCT, Sakura, #4583) in a frontal orientation. The following antibodies were used: mouse anti-Ezrin (Abcam, ab4069, 1:100), mouse anti-FGF15 (Santa Cruz, sc-514647, 1:50), mouse anti-NR2F1 (R&D system, PP-H8312-00, 1:100), rabbit anti-p-ERK1/2 (Cell Signaling, #4370, 1:100), rabbit anti-p-SMAD1/5/8 (Merck, AB3848-I, 1:50), rabbit anti-RFP (Novus Biologicals, NBP2-25157, 1:250), goat anti-mouse Alexa Fluor 555 (Cell Signaling, CST4409S, 1:500) and goat anti-rabbit Alexa Fluor 488 (Cell Signaling, CST4412S, 1:500). Nuclei were counterstained with DAPI (Roth, 1:1000 from a 1 mg/ml stock solution).

Fate-mapping mice (Sox2-creER^T2::lslRFP^fl/fl and Sox2-creER^T2::Brg1^fl/fl::lslRFP^fl/fl mice after tamoxifen at E7.5 or E9.5) were sacrificed at E14.5 and their brains isolated under a dissecting microscope. The meninges were removed and the tissue digested using Papain (20 U/ml) and DNase (400 µg/ml) for 30 min at 37°C. RFP-positive cells were sorted using the BD FACS Aria-Fusion and the software FACSDiva 8.0.1 (both BD Biosciences).

After FACS, 5×10⁵ RFP-positive cells were collected in 80 µl PBS and centrifuged for 5 min at 165 g onto a cytoslide using a cytospin centrifuge (Cytospin 4, Thermo Fisher Scientific). The cells were fixed with 4% formaldehyde for 20 min and air-dried overnight.

Immunofluorescent staining was performed according to standard protocols and by using rabbit anti-Brg1 (Abcam, ab110641, 1:25) and goat anti-rabbit Alexa Fluor 488 (Cell Signaling, CST4412S, 1:500) antibodies. Nuclei were counterstained with DAPI (Roth, 1:1000 from a 1 mg/ml stock solution).

RFP-positive cells collected by FACS were seeded in 500 µl culture medium [DMEM/F12 supplemented with 1 M HEPES buffer, N2 supplement, L-Glutamin, penicillin/streptomycin (50 U/ml), non-essential amino acids, epidermal growth factor (EGF, 20 ng/ml) and FGF (10 ng/ml)] at a concentration of 50 cells/µl. Every 3 days, fresh EGF and FGF were added to the medium. After 7 days in culture, number and size of primary neurospheres were determined using a Nikon Ti2 microscope (Nikon, Tokyo, JP) and quantified using the ImageJ software.

All statistical tests were performed using the Prism Software Version 7 (GraphPad Software). Fractions of cell counts were blindly determined by counting the total number of cells per field (identified by staining of Hematoxylin) and the number of cells that stained positive for the respective marker (pHH3 or clCasp3). Statistical significance was determined by unpaired two-tailed t-tests. The yield of RFP-positive cells after FACS as well as the numbers of neurospheres per well after 7 days in vitro was compared by unpaired two-tailed t-tests.

Total RNA of RFP-positive cells collected by FACS was isolated using TRIzol (Life Technologies) according to established protocols (Chomczynski, 1993). After isolation, the RNA integrity was analyzed with the RNA 6000 Nano Chip on an Agilent 2100 Bioanalyzer (Agilent Technologies). From total RNA, mRNA was extracted using the NEBNext Poly(A) mRNA Magnetic Isolation module (New England Biolabs) and RNA-Seq libraries were generated using the NEXTFLEX Rapid Directional qRNA-Seq Kit (Bioo Scientific) according to the manufacturer's recommendations. Concentrations of all samples were measured with a Qubit 2.0 Fluorometer (Thermo Fisher Scientific) and fragment lengths distribution of the final libraries was analyzed with the DNA High Sensitivity Chip on an Agilent 2100 Bioanalyzer (Agilent Technologies). All samples were normalized to 2 nM and pooled equimolarily. The library pool was sequenced on the NextSeq500 (Illumina) with 1×75 bp, with 14.2 to 19.7 mio reads per sample.

The quality of the raw reads was confirmed by using the FastQC tool from the Babraham Bioinformatics Institute (Cambridge, UK). Afterwards, the reads were aligned to the mouse reference genome GRCm38 employing the Spliced Transcripts Alignment to a Reference (STAR) software (v.2.6.1c) (Dobin et al., 2013). Concurrently, counts (reads/gene) were determined by the quantmode GeneCounts option. They are based on the Ensembl annotation release 95 (Yates et al., 2020). The DESeq2 package (v.1.22.2) was used to estimate DEGs between Brg1-deficient samples and their Brg1-competent controls (Love et al., 2014). The package calculated log2 fold changes, p-values and FDRs. Genes were defined as DEGs, if they had a log2 fold change >±0.6 and a FDR <0.1. Heatmaps were created with the pheatmap R package, based on the DEseq2 normalized counts. Enriched GO terms were estimated with the GAGE R package (v2.32.1). To visualize GO terms and their related DEGs, GOchords were created by using the GOChord plotting function of the R package GOplot (v1.0.2). For KEGG Pathway analysis and visualization, R packages gage (v2.36.0) (Luo et al., 2009) and ReactomePA (v1.30.0) (Yu and He, 2016) were employed.

We thank Margarethe Gregersen, Kristin Hartmann, Vanessa Thaden, Jacqueline Kolanski, Anne Reichstein, Nick Mohr, Celina Soltwedel (all University Medical Center Hamburg-Eppendorf), Arne Düsedau, Jana Hennesen and Gundula Pilnitz-Stolze (all Leibniz Institute for Experimental Virology, Hamburg) for excellent technical support. We further thank Dr Pierre Chambon (IGBMC, Illkirch-Graffenstaden, France) for providing the Brg1^fl/fl mice. Finally, we acknowledge the support of the Small animal models core facility of the Heinrich Pette Institute. The data presented here were collected as part of the PhD thesis of D.H. (Holdhof, 2020).

The peer review history is available online at https://journals.biologists.com/dev/article-lookup/doi/10.1242/dev.196147