The BAF complex interacts with Pax6 in adult neural progenitors to establish a neurogenic cross-regulatory transcriptional network.
Journal: 2014/December - Cell Stem Cell
ISSN: 1875-9777
Abstract:
Numerous transcriptional regulators of neurogenesis have been identified in the developing and adult brain, but how neurogenic fate is programmed at the epigenetic level remains poorly defined. Here, we report that the transcription factor Pax6 directly interacts with the Brg1-containing BAF complex in adult neural progenitors. Deletion of either Brg1 or Pax6 in the subependymal zone (SEZ) causes the progeny of adult neural stem cells to convert to the ependymal lineage within the SEZ while migrating neuroblasts convert to different glial lineages en route to or in the olfactory bulb (OB). Genome-wide analyses reveal that the majority of genes downregulated in the Brg1 null SEZ and OB contain Pax6 binding sites and are also downregulated in Pax6 null SEZ and OB. Downstream of the Pax6-BAF complex, we find that Sox11, Nfib, and Pou3f4 form a transcriptional cross-regulatory network that drives neurogenesis and can convert postnatal glia into neurons. Taken together, elements of our work identify a tripartite effector network activated by Pax6-BAF that programs neuronal fate.
Relations:
Content
Citations
(65)
References
(64)
Chemicals
(6)
Genes
(6)
Organisms
(2)
Processes
(2)
Anatomy
(3)
Affiliates
(1)
Similar articles
Articles by the same authors
Discussion board
Cell Stem Cell 13(4): 403-418

Essential role of BAF complex interacting with Pax6 in establishment of a core cross-regulatory neurogenic network

+13 authors

Introduction

Cell fate specification includes multiple steps in restricting progenitor potential and directing the expression of genes that elicit a lineage-specific program. According to the “master regulator concept” a single gene expressed in competent tissue is capable of inducing the expression of the entire lineage specific transcriptional cascade, resulting in final and complete fate commitment (Baker, 2001; Stoffers et al., 1997). However, the molecular mechanisms by which these master regulators work are not clear. It is unknown for most lineages if master regulators directly control multiple genes executing lineage decision and differentiation of cells further along the lineage, or only control a few downstream effector genes, as e.g. proposed by the concept of terminal selector genes (Hobert, 2011). The transcription factor Pax5, a master regulator of B cell fate in the hematopoietic system, promotes B cell differentiation and maintenance of B cell fate (Medvedovic et al., 2011) exploiting the epigenetic machinery to either shut down expression of genes of other lineages or activate genes of the B cell lineage (McManus et al., 2011). Thus, Pax5 directly regulates many effector genes and stabilizes the lineage decision using the epigenetic machinery. Other master regulators have relatively weak activation capacity themselves (e.g. Hnf3b and Sry transcription factors), but have the capacity to open chromatin enabling binding and transactivation by other transcription factors executing the lineage decision (Zaret and Carroll, 2011).

These mechanisms remain largely open for neural fate decisions. Key transcriptional regulators of neurogenic fate have been identified in the developing and adult brain, such as Pax6, Neurog2, Ascl1, SoxC and Dlx2 (Bergsland et al., 2011; Bergsland et al., 2006; Hack et al., 2005; Haubst et al., 2004; Heins et al., 2002; Mu et al., 2012; Nieto et al., 2001; Petryniak et al., 2007; Schuurmans et al., 2004). However, it is still largely unknown how these transcription factors act at the molecular level to direct neuronal fate decisions. While ChIP experiments and mouse mutant analysis for Pax6, Ascl1, SoxC (Sox4 and Sox11) and Neurog2 revealed binding and regulation of genes involved in proliferation as well as neurogenesis (Asami et al., 2011; Bergsland et al., 2011; Bergsland et al., 2006; Castro et al., 2011; Sansom et al., 2009) the key molecular targets allowing these factors to orchestrate neurogenesis and neuronal reprogramming have not yet been identified.

In direct reprogramming, it appears that the number of factors necessary for fate conversion is inversely correlated to the distance of their lineage relation. If cells share a common lineage, such as glia with neurons, they can be converted with one factor (Davis et al., 1987; Heinrich et al., 2010; Heins et al., 2002), while turning more distinctly related cells, such as fibroblasts, into neurons requires several factors (Caiazzo et al., 2011; Vierbuchen et al., 2010; Yoo et al., 2011). This may reflect the state of the chromatin in the initial cell and reinforces the importance of transcription factors in chromatin restructuring and opening of new binding sites for fate conversion (Lalmansingh et al., 2012; Siersbaek et al., 2011; Zaret and Carroll, 2011). However, virtually nothing is known about the interaction of key neurogenic factors with the chromatin modifying machinery. Several chromatin remodeling factors have been found to be important during neural development, including members of the BAF complex, CHD and ISWI complexes (Engelen et al., 2011; Ho et al., 2009; Yip et al., 2012). For example, conditional deletion of the ATPase subunits, Snf2l or Brg1 of the ISWI or BAF chromatin remodeling complexes reduced neural stem cell proliferation and self-renewal in the developing forebrain (Lessard et al., 2007; Matsumoto et al., 2006; Yip et al., 2012). These chromatin-remodeling complexes appear necessary to regulate expression of pro-proliferative genes, such as Foxg1 (Yip et al., 2012) and Notch and Sonic Hedgehog pathway genes (Lessard et al., 2007). Interestingly, whether the BAF complex regulates proliferation or neuronal differentiation (Wu et al., 2007), depends on its subunit composition (Ho et al., 2009). In mammals, BAF complexes contain the ATP-ase subunits Brg1 or Brahma which are mutually exclusive and essential for remodeling activity and up to 12 Brg1/Brm-associated factor (BAF) subunits. Depending on specific BAF complex subunits, this complex is involved in neural stem cell maintenance or neuronal differentiation (Ho et al., 2009; Singhal et al., 2010). However, it is not known how these specific complexes are targeted to and regulate the respective and distinct targets and to which extent they may interact with specific transcription factors.

To shed light on the molecular mechanisms underlying the function of key neurogenic master regulators in fate specification and conversion, we choose to search for Pax6 interactors. Pax6 is not only a master regulator in eye development (Baker, 2001) and in neurogenesis (Brill et al., 2008; Gotz et al., 1998; Hack et al., 2005; Haubst et al., 2004; Heins et al., 2002; Stoykova and Gruss, 1994; Stoykova et al., 2000), but has also been found to localize to heterochromatin-rich territories (Elvenes et al., 2010) consistent with the suggestion that Pax6 may act as a pioneer transcription factor (Zaret and Carroll, 2011). Indeed, Pax6 regulates neurogenesis in the adult brain (Hack et al., 2005) and is sufficient to reprogram glia into neurons in vitro (Berninger et al., 2007; Heins et al., 2002) and in vivo (Buffo et al., 2005). Understanding how Pax6 exerts its neurogenic function is therefore of crucial interest to reveal the basic principles of endogenous and enforced neurogenesis.

Results

Transcription factor Pax6 interacts with BAF chromatin remodeling complex in neurogenic progenitors

In order to understand the mechanisms underlying Pax6-mediated neurogenesis, we purified Pax6-containing complexes from neural stem cells expressing Pax6 (Suppl. Fig. 1A) and used mass spectrometry to examine their composition. Pax6-complexes were purified by either Pax6 antibody (Pax6-IP, Fig. 1A) or FLAG antibody from neural stem cells stably expressing FLAG-tagged Pax6 (FLAG-Pax6-IP, Suppl. Fig. 1B). In either case, multiple subunits of the BAF complex were present in the Pax6 samples. The interaction of Pax6 with the BAF complex was confirmed by western blot (WB) detection of Brg1 and other subunits of the BAF complex in Pax6 immunoprecipitations (Fig. 1A). Thus, Pax6 physically interacts with Brg1-containing BAF chromatin remodeling complexes in neural stem cells. To examine this in the brain, we prepared nuclear extracts from the core of the adult mouse olfactory bulb (OB) which is enriched in Pax6+ neuroblasts (Hack et al., 2005; Brill et al., 2008). Immunoprecipitation with Pax6 antibody followed by WB for Brg1 (Fig. 1A) confirmed the interaction of Brg1-containing complexes with Pax6 in the OB.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f1.jpg
Loss of Brg1 function in neuronal progenitors renders them into glial cells

(A) Western blot depicting the direct interaction of Pax6 and members of BAF complex in the crude protein extract from NS5 cells and the OB. (B-E) Micrographs depicting co-expression of Pax6 and Brg1 in DCX+ neuroblasts (blue arrow in D and E) in the SEZ (D) and the OB (E). D and E are magnifications of area boxed in B and C, respectively. (F, G) Representative micrographs of the olfactory bulbs of Brg1 cKO (G) and corresponding control (F) 28 days after tamoxifen-induced recombination. (H-K) Micrographs depicting the immunoreactivity of recombined cells for DCX and NG2 in Brg1 cKO (I, K) and its WT sibling (H, J). J and K are magnifications of area boxed in D and E, respectively (L-M) Micrographs depicting the diverse morphology of NG2-positive cells generated from the Brg1-deficient neural progenitors. (N) Histogram depicting the total number of NG2 positive cells in different OB layers 28 days after recombination. Data are shown as mean ± SEM and n(animals analyzed)≥5. *-p≤0.05; ***-p≤0.005. (O) Pie charts illustrating the identity of recombined cells in the OB of Brg1 cKO and age-matched siblings 28 days after tamoxifen-induced recombination. Data are shown as mean and n(animals analyzed)≥5. Scale bars: 100 μm in B,C, F and G; 50 μm in L-M′; 20 μm in D, E, H-K. Abbreviations: ctx-cerebral cortex; cc-corpus callosum; RMS-rostral migratory stream; SEZ-subependymal zone; St-striatum; lv-lateral ventricle; GL-glomerular layer; GCL-granule cell layer; EPl-external plexiform layer; dGCL-deep granule cell layer and sGCL- superficial granule cell layer.

To examine co-localization of Pax6 and Brg1 at the cellular level we performed immuno-staining for Brg1 in the adult brain, which showed a broad expression of Brg1 in neurons and astrocytes throughout the brain, while surprisingly weak immunoreactivity for Brg1 was detectable in white matter (WM) where mostly oligodendrocytes and their precursors reside (Fig. 1B; Suppl. Fig. 1C). In the regions of adult neurogenesis, we focused on the region generating OB interneurons where Pax6 has been shown to regulate neurogenesis (Hack et al., 2005; Brill et al. 2008). These adult-generated neurons derive from neural stem cells (NSCs) located in the subependymal zone (SEZ) that generate via transient-amplifying progenitors (TAPs) neuroblasts migrating through the rostral migratory stream (RMS) towards the OB (Ming and Song, 2012). Brg1-immunoreactive nuclei were detected in GFAP+ astrocytes and stem cells as well as in Ascl1+ TAPs and Doublecortin (DCX)+ neuroblasts (Suppl. Fig. 1D,E and data not shown) also expressing BAF53a and BAF45a (Suppl. Fig. 1F,G; but not BAF45b,53b in H,I), subunits characterizing the neural progenitor specific BAF complexes (Lessard et al., 2007). Conversely, Pax6 immunoreactivity is largely restricted to neuroblasts in the SEZ and RMS, where it colocalizes with Brg1 (Fig. 1D-E). Together with the immunoprecipitation experiments, these data suggest that Pax6 interacts with the neural progenitor-specific, Brg1-containing BAF chromatin remodeling complex in neuroblasts in vivo.

Loss of Brg1 in adult NSCs converts olfactory bulb neurogenesis into gliogenesis

To test the functional relevance of the observed Pax6-BAF interaction, we ablated Brg1 in the Brg1fl mouse line (Indra et al., 2005; Matsumoto et al., 2006) by tamoxifen (TM) inducible Cre-based excision in Glast mice, mediating genomic recombination in astrocytes and NSCs (Mori et al., 2006; Ninkovic et al., 2007). As these mice were also crossed with the CAG-CAT-GFP reporter line (Nakamura et al., 2006), this allowed visualization of the recombined cells. By 9 days post TM administration (9dpt) 95 % of reporter positive cells in Glast/Brg1 mice (further referred as Brg1 cKO) were no longer Brg1-immunopositive, while virtually all reporter positive cells were Brg1+ in Glast/Brg1 or Glast/Brg1 mice (further referred as Brg1 controls) (Suppl. Fig. 2A-C).

While no altered neurogenesis was detectable in the SEZ, RMS or OB of Brg1 cKO mice 9 dpt (Suppl. Fig. 2D), 28 dpt the number of recombined cells was significantly reduced in the OB, the final destination of adult generated neurons (Fig. 1F,G). As this may be due to cell death, we examined activated caspase3, an indicator of programmed cell death. Indeed, the number of activated caspase3+ cells was significantly increased in the OB and RMS but not in the SEZ of Brg1 cKO mice compared to controls (Suppl. Fig. 2E and data not shown). This suggests that cell death is initiated at later stages in the neuroblasts when they migrate along the RMS. Accordingly, we also observed an increase in GFP-positive cells located just beside the RMS (Suppl. Fig. 2F-H) with a morphology reminiscent of oligodendrocyte progenitor cells (OPCs). Staining for the transcription factor Olig2 (data not shown) and proteoglycan NG2 labeling OPCs (Dimou et al., 2008) confirmed the strikingly higher number of Brg1cKO GFP+ OPCs (Fig. 1I,K-O; 2000x increase) while virtually no GFP+ OPCs were detectable in the OB of control animals (Fig. 1H,J). Notably, GFP+ OPCs were present only at the end of the RMS, the core of the OB and the deep granule cell layer (GCL), but were absent from the glomerular layer (GL) in Brg1 cKO mice (Fig. 1N). The newly generated OPCs displayed a variety of morphologies with different level of cellular complexity (Fig. 1K-M) indicative of different stages in the oligodendrocyte lineage. However, even 2 months after TM, no GFP+ cells had matured into oligodendrocytes immunoreactive for GST-Π (Suppl. Fig. 2I), consistent with the behavior of OPCs in the GM (Dimou et al., 2008). Interestingly, other glial cells - like GFAP+ astrocytes (Fig. 1O and Suppl. Fig. 2J-K′) and marker-negative cells (Suppl. Fig. 2L,L′, cells were immunoprobed for more than 20 antigens indicative of astroglial, neuronal, oligodendroglial, endothelial, microglial lineage as well as fibroblasts, see material and methods) were also increased in Brg1cKO OB, reflecting a broader conversion towards gliogenesis and some cells failing to adopt any coherent cell identity after loss of Brg1 (Fig. 1O and Suppl. Fig. 2L). Conversely, cells of the neuronal lineage, labeled with DCX (neuroblasts and young neurons) or NeuN (recently matured neurons), were strongly reduced in number to less than a third (Fig. 1O and Suppl. Fig. 2M), with the only exception of neurons in the superficial GCL (Suppl. Fig. 2M), indicating that this neuronal sub-lineage is selectively spared by Brg1 depletion and accounts for almost half of the remaining neurons. Taken together, inducible, cell-specific deletion of Brg1 in vivo converts adult OB neurogenesis to gliogenesis starting in the RMS and is accompanied by increased cell death.

GLAST-mediated recombination is not restricted to the SEZ and RMS, the origin of adult OB neurogenesis, but also depletes Brg1 in astrocytes throughout the brain. Therefore, it is possible that some astrocytes in the OB and RMS may be converted to OPCs. While no GFP+ OPCs were detected outside the OB (e.g. cerebellum or dentate gyrus, data not shown), it remains possible that specifically astrocytes in the OB would be more easily converted to OPCs. To examine the fate of cells originating in the SEZ directly, we injected DsRed-expressing MLV-based retroviral vectors 9 dpt into the SEZ of GLAST//Brg1//CAT-CAG-GFP animals in order to label Brg1cKO progenitors (DsRed+/GFP+) and their WT cellular counterparts (not recombined and hence GFP-, but DsRed+) already in the SEZ (Suppl. Fig. 3A,B). When we analyzed the identity of labeled cells in the OB 7 days later, most of the DsRed+, GFP- control cells (more than 90%) were DCX+ neuroblasts, as is normally the case (Suppl. Fig. 3B). However, only a minority (40%) of the Brg1-deficient cells (DsRed+,GFP+) originating in the SEZ had acquired a neuroblast identity (Suppl. Fig 3B) and a significant proportion expressed Olig2 (Suppl. Fig. B). As this proportion was similar to the GFP+, DsRed- (38% for DCX and 36 % for Olig2), we conclude that there is no major additional source of cells contributing to the OPCs in the OB of Brg1 cKO mice. Thus, most cells from the SEZ fail to complete their neurogenic fate upon Brg1 deletion and convert to gliogenesis.

Viral vector injection causes an injury in the SEZ, which may affect the lineage progression. To address this possibility and verify the above findings, we deleted Brg1 with the Nestin-CreER line (Lagace et al., 2007) mediating recombination exclusively in nestin+ cells of the SEZ but not in the nestin- parenchymal glia. Consistent with our findings with the Glast mice shown above, mice that lack Brg1 in nestin+ cells and their progeny had significantly fewer GFP+ cells reaching the OB and acquiring a neuronal identity compared to control mice 30 dpt (Suppl. Fig. 3C-D). Similar to the Glast-mediated deletion of Brg1, these cells expressed Olig2. Taken together, multiple independent experimental approaches confirm that Brg1 is an essential component of SEZ-derived neurogenesis, and that the absence of Brg1 causes the conversion of neuroblasts in the RMS and OB to the glial lineage.

Niche-dependent gliogenesis elicited by loss of Brg1

As our fate mapping experiments demonstrated that the reporter+ glia in the OB were converted from neuroblasts originating from the SEZ, we examined cells in the SEZ of Brg1 cKO mice. However, even 28dpt (17 days after loss of Brg1 protein), numbers of DCX+ neuroblasts amongst the recombined (GFP+) cells in the SEZ did not significantly differ between Brg1cKO and control mice (Suppl. Fig. 3E,F), suggesting that conversion to gliogenesis occurs only when neuronal cells exit the SEZ. This strikingly late fate conversion may be due to a powerful role of the niche environment or other intrinsic mechanisms sufficient to stabilize neuronal fate for a period of time (Beckervordersandforth et al., 2010; Colak et al., 2008; Lim et al., 2006). To distinguish between these two possibilities, we isolated progenitors from the SEZ of Brg1cKO or control mice 9dpt and cultured them in a primary SEZ culture system maintaining neurogenesis (Costa et al., 2011; Ortega et al., 2011). Consistent with the in vivo analysis of neurogenesis in the SEZ (Suppl. Fig. 3E), we did not observe any difference in cell composition of the SEZ in Brg1 cKO and respective controls after isolating cells 7 dpt and analyzing them 4h after plating (Suppl. Fig. 3G). However, cells lacking Brg1 were not able to proceed further along the neurogenic lineage also after 7 days in vitro, as only 9% of them were DCX+ neuroblasts in contrast to 60% of control cells (Fig. 2A-D). Intriguingly, Brg1-deficient cells now had acquired different glial identities, with 15% NG2+ OPCs, 22% GFAP+ astrocytes and 49% S100b+, CD24+ ependyma-like cells (Fig. 2C, D and Suppl. Fig. 3H).

An external file that holds a picture, illustration, etc.
Object name is nihms595448f2.jpg
Loss of Brg1 results in full conversion to gliogenesis in vitro

(A-C) Micrographs depicting the immunoreactivity of GFP+ adult neural stem cell progeny derived from either Brg1 cKO or age matching controls (10 days after TM induction) for DCX (neuroblasts) or NG2 (OPCs) after 7 days in vitro. (D) Pie charts depicting the composition of the adult neural stem cell progeny following Brg1 depletion (lower pie chart) or age matching controls (upper panel) after 7days in vitro. Data are shown as mean and n(animals analyzed)≥7. (E-F) Representative trees illustrating the predominant behavior of the Brg1 KO or control neural progenitors in vitro followed by continuous live imaging at the single cell level (see also Suppl. Movies). (G) Box-chart showing the cell cycle length (defined as the time between two divisions) of control and Brg1-deficieng progenitors. (H) Histogram depicting the proportion of clones (control or Brg1-deficient) containing at least one dead cell. Data are shown as mean (3 independent experiments) ± SEM. Scale bars: 100 μm in A, B and C.

These observations prompted us to ask if ependymal cells are generated also in the SEZ in vivo after Brg1 deletion. As both Glast and nestin are expressed in ependymal cells (Beckervordersandforth et al., 2010; Lagace et al., 2007), they are also reporter+ in the above mouse lines and their possible increase is difficult to detect. To overcome this limitation, we used the recently developed Split-Cre technology that specifically mediates recombination in SEZ NSCs using two halves of Cre driven by GFAP and P2 prominin promoters simultaneously active in NSCs (Beckervordersandforth et al., 2010). Stereotactic injection of 2 lentiviruses encoding each half of Cre under the respective GFAP or P2 promoter into Brg1//CAG-CAT-GFP or Brg1//CAG-CAT-GFP mice allows to selectively delete Brg1 in NSCs and to follow their progeny by the GFP reporter. The majority of GFP+ cells were neuroblasts 60 days after virus injection into control mice with very few ependymal cells labeled (Suppl. Fig. 3I-K), while the progeny of Brg1-deficient NSCs contained 30% ependymal cells (Suppl. Fig. 3I-K). Taken together, the local niche not only influences the selection of glial fate subtype after Brg1 deletion, but also contributes to maintenance of some cells as neuroblasts even in the absence of Brg1, as virtually all cells convert to gliogenesis outside this neurogenic niche in vitro.

Mode of fate conversion determined by continuous single cell live imaging in vitro

As Brg1-deficient cells in vitro largely convert into the same cell types as in vivo, we used single cell continuous live imaging for 7 days in vitro followed by postimaging immunostaining to discriminate whether the conversion of Brg1 cKO cells from neuro- to gliogenesis occurs by selective cell death, selective proliferation or a true fate conversion (Suppl. Movies 1-3; Fig. 2E-I). In agreement with the population based analysis, 93% of lineage trees of single control cells and their entire progeny as observed by live imaging contained only neuronal cells and only a small fraction of trees, the NSC progeny, had generated both neurons and glia (Fig. 2 E,G). In contrast, when we observed Brg1 cKO cells most of the trees contained either NG2 glia or ependyma-like cells, but only a minority had generated neurons only (Fig. 2F). Interestingly, cells giving rise to ependyma-like cells were the only lineage that initially perfermed a series of fast symmetric proliferative divisions (Fig. 2F). Other than this, cell cycle length did not differ between control and Brg1 cKO cells (Fig. 2G), nor could we observe any difference in regard to cell death that was rather rare in both control and Brg1 cKO cells (Fig. 2E,F,H). Thus, Brg1 cKO cells convert to glial lineages by fate change rather than selective cell death or proliferation. Interestingly, some Brg1 cKO cells generate neurons and NG2+ OPCs, while cells generating ependymal progeny did not give rise to any other glial or neuronal cells, suggesting that fate conversion occurs distinctly into different glial lineages. Moreover, as most of the control and Brg1 cKO cells at the start of these cultures are neuroblasts (63±4 %, Suppl. Fig. 2D), and no selective cell death of Brg1 cKO cells was observed, these experiments demonstrate that many Brg1 cKO neuroblasts directly convert to glial lineages.

Loss of Brg1 leads to down-regulation of Pax6 targets in adult SEZ and OB

Given the surprisingly specific defects in neurogenesis after deletion of Brg1, we used genome-wide expression analysis to determine which genes may be involved in this phenotype as well as to clarify additional gene regulation effects indicative of a further phenotype we may have missed so far. Towards this aim, we isolated the SEZ and core of the OB (containing neuroblasts entering the OB) from Brg1 cKO and control mice 10 dpt, just 1 day after loss of Brg1 protein. Genome-wide expression profiling by microarray (GST 1.1 gene array, Affymetrix, USA) revealed 244 significantly regulated genes (change in the expression level >1.2 or <0.8 and p<0.05) in the OB (Suppl. Table 1) and 136 genes in the SEZ (Suppl. Table 2 and Suppl. Fig. 4A) and qPCR on independent samples confirmed the reliability of this analysis (Fig. 3A). The majority of genes altered in their expression were down-regulated consistent with a role of Brg1-containing chromatin remodeling complexes in positively regulating gene transcription. Importantly, most of the significant gene ontology (GO) terms were related to neurogenesis, synaptic transmission, axonogenesis and cell migration, consistent with defective neurogenesis (Suppl. Fig. 4B,C). The other major terms were related to DNA and RNA metabolism and cell replication and cell cycle, suggestive of defects in cell cycle regulation in Brg1cKO SEZ and OB and possibly linked to the very specific cell division pattern of ependymal cells generation we observed using live imaging (Fig. 2F). Interestingly, 64% of the down-regulated genes in SEZ (Suppl. Fig. 4E) are expressed in NSCs and their progeny, as identified in our previous transcriptome analysis (Beckervordersandforth et al., 2010 and Suppl. Fig. 4D-F). Most intriguingly, 87% of the down-regulated genes had predicted Pax6 binding sites, a significant enrichment in regard to the expected frequency (63% expected, Genomatix Matinspector) (Fig. 3B). This suggests that most genes down-regulated in Brg1 cKO may be regulated by Pax6 in conjunction with a Brg1-containing BAF complex.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f3.jpg
Brg1 deletion results in down-regulation of Pax6 target genes and is phenocopied by Pax6 deletion

(A) Histogram depicting the comparison of a specific gene set miss-regulated after Brg1 deletion and measured by microarray and qPCR on independent samples. Data are shown as mean and n(animals analyzed)≥3. (B) Venn diagram depicting predicted Pax6 binding sites in the promoters of genes deregulated following Brg1 deletion. (C, D) Representative micrographs of the olfactory bulb of Pax6 cKO (C) and control (D) 60 days after tamoxifen-induced recombination. (E, F) Micrographs depicting the immunoreactivity of recombined, GFP+ cells in Pax6 cKO (E) and its control sibling (F) for DCX (neuroblasts) and NG2 (OPCs). (G) Histograms depicting the total number of NG2 positive cells in different OB layers 60 days after recombination. Data are shown as mean ± SEM and n(animals analyzed)≥7. *-p≤0.05; ***-p≤0.005. (H) Pie charts illustrating the identity of recombined cells in the OB of Pax6 cKO and age-matching sibling 60 days after tamoxifen-induced recombination. Data are shown as mean and n(animals analyzed)≥3. Scale bars: 100 μm in C and D; 20 μm in E and F. Abbreviations: RMS-rostral migratory stream; GL-glomerular layer; EPl-external plexiform layer; dGCL-deep granule cell layer and sGCL-superficial granule cell layer.

Deletion of Pax6 phenocopies Brg1 deletion and converts adult SEZ neurogenesis to gliogenesis

In order to test the above suggestion directly in vivo, we crossed the floxed Pax6 mice (Ashery-Padan et al., 2000) with GLAST mice to delete Pax6 as before Brg1. While Pax6 protein was more stable than Brg1 and disappeared only 21 dpt (Suppl. Fig. 5 A-C), the phenotype emerging thereafter was remarkably similar to the phenotype observed in Brg1 cKO mice. As in the Brg1 cKO mice, we observed fewer numbers of GFP+ cells reaching the OB in the Pax6 cKO mice compared to controls (Fig. 3C,D), and most of them no longer differentiated along the neuronal lineage (DCX+/NeuN+ 45%, compared to 82% in controls, Fig. 3E-H), but rather converted to glial identities (Fig. 3G-H) in ratios similar to the Brg1 cKO cells (Fig. 1H-O). Thus, consistent with the interaction of Pax6 with Brg1-containing BAF complex in vitro and in vivo, deletion of either of these proteins results in the same phenotype with severe defects in OB neurogenesis implying a key role of this transcriptional complex in regulating neurogenesis.

Pax6 and catalytically active Brg1 are both essential for forced neurogenesis

The above findings prompted us to examine to which extent Pax6-induced neurogenesis would also require the presence of Brg1. Towards this end, we prepared neurosphere cells from the adult SEZ of Brg1 fl/fl animals and transduced them in vitro with viral vectors encoding Cre to obtain Brg1-deficient cells (Fig. 4A-D). Two days after transduction, Brg1 protein was virtually absent from the Cre-transduced cells, while 95% of control virus transduced cells were still Brg1+ (data not shown). Consistent with the low endogenous neurogenic potential of these cells, 15% of control virus (GFP only) infected cells had differentiated into DCX+ neuroblasts (Fig. 4A-D), whereas Cre-transduced, Brg1cKO cells largely failed to generate any DCX+ neurons (Fig. 4B,D), in agreement with the above observed results that Brg1 is necessary for endogenous neurogenesis. Brg1 is also essential for neurogenesis elicited by Pax6 over-expression. Pax6 over-expression did not induce neurogenesis in Brg1 cKO cells (co-transduced with CreIRESGFP virus) in contrast to the Ctrl situation where Pax6 transduction induced more than 80% DCX+ neurons (Fig. 4D). Importantly, the full-length form of Brg1, but not the ATPase deficient form (Brg1KS), restored both endogenous and Pax6-induced neurogenesis (Fig. 4E), demonstrating that Brg1 needs to be catalytically active to mediate neurogenesis. The requirement of Brg1 for neurogenesis is rather specific for Pax6, as transduction with the neurogenic factor Ngn2 resulted in very efficient induction of neurogenesis to over 90% even in the absence of Brg1 (Fig. 4C, D). Thus, Pax6 and Brg1 are not only required for endogenous OB neurogenesis, but also neurogenesis forced by Pax6 over-expression requires the presence of catalytically active Brg1 and hence chromatin remodeling activity.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f4.jpg
Pax6 requires catalytically active Brg1 for its neurogenic function

(A-C) Micrographs depicting immunoreactivity of Brg1-deficient cells (green) derived from Brg1 neurospheres for the neuroblast marker DCX after over-expression of control vector expressing GFP (A), Pax6 (B) and Ngn2 (C) 7 days after transduction. (D, E) Histograms depicting the proportion neurons generated from Brg1cKO and control neurosphere cells 7 days after expression of Pax6, Ngn2, Brg1 and Brg1KS (ATPase deficient form). Note that only catalytically active Brg1 restores the neurogenic function of Pax6. (F) Histogram depicting down-regulation of genes following Brg1 deletion (open bars) and Pax6 deletion (grey bars) measured by qPCR. Data are shown as ratio of average gene expression in the mutant and age-matching WT normalized to the expression of GAPDH housekeeping gene. (G) Venn diagram depicting transcription factors down-regulated after Brg1 deletion and harboring Pax6 binding site in their promoters. (H) Histogram showing the down-regulation of Pou3f4, Sox11 and Nfib in the OB and SEZ following Brg1 and Pax6 deletion. (I) Histogram depicting the induction of the expression of Pou3f4 and Sox11 after forced Pax6 expression in neurosphere cells measured 24 hours after transduction. (J, K) Micrographs depicting deletion of Brg1 in neurosphere cells 36h after nucleofection of control (J) or Cre (K) encoding plasmid. (L) Histogram depicting the induction of the expression of Pou3f4, Sox11 and Nfib after forced Pax6 overexpression in control or Brg1-deficient neurosphere cells measured 24 hours after Pax6 nucleofection. (M) Histogram depicting the relative expression of Pou3f4, Sox11 and Nfib in the acutely isolated, FACS purified population enriched in neural stem cells and their progeny. Scale bars: 100 μm in A, B, C, J and K. Data in D, E, H, I, and L are shown as mean ± SEM and n(animals analyzed)≥3 and in M as mean n(animals analyzed)≥3. **-p≤0.01; ***-p≤0.005.

Identification of a neurogenic transcriptional network downstream of Pax6 and Brg1

Given the similarity of the phenotypes after deletion of either Pax6 or Brg1 in adult NSCs of the SEZ and functional interaction in forced neurogenesis, we next asked to which extent this is also reflected at the transcriptional level. First, we examined randomly selected genes, found to be down-regulated in the core of the OB or the SEZ of Brg1 cKO mice, in the OB and SEZ of Pax6 cKO mice. Consistent with the similarity at the phenotypic level, 90% of these genes down-regulated after loss of Brg1 were also down-regulated after loss of Pax6 (Fig. 4F and data not shown), in agreement with the observation that the majority of these genes possess Pax6-binding sites. This provided us with an opportunity to search for genes implementing neurogenesis as this is deficient in both these mutants. We therefore searched for transcriptional regulators down-regulated upon Brg1 depletion with a common regulatory motif including Pax6. This was the case for 7 of the 11 down-regulated transcription factors (Fig. 4G; FrameWorker, Genomatix, Germany) and only for 4 of these a specific consensus DNA binding sequence was described (Sox11, Sox4, Pou3f4 and Nfib). Indeed, all these are expressed in the SEZ and RMS (Suppl. Fig. 5G-J) and their expression is down-regulated in both Brg1cKO and Pax6 cKO OB (Fig. 4H). This reduction was relevant at protein levels; for example Nfib was present in virtually all neuroblasts in the RMS, but upon loss of Pax6 (GFP+ cells in Pax6 cKO 60 dpt) only 17% of recombined, GFP+ neuroblasts expressed Nfib (Suppl. Fig. 5K-M).

Pax6 was also sufficient to induce Sox11 and Pou3f2/4 expression within 24hours in neurosphere-derived cells (Fig. 4I) in a Brg1-dependent manner (Fig. 4J-L). Interestingly, the expression of Sox11, Nfib and Pou3f4 was highest in neuroblasts expressing the highest levels of Pax6 (Fig. 4M). Thus, the regulation and expression of these transcription factors is consistent with a role downstream of the Pax6-Brg1 complex in neuroblasts.

Most importantly, these downstream transcription factors were also predicted to cross-regulate each other, thus potentially forming a self-sustaining corss-regulatory network critical for neurogenic fate maintenance in the adult brain. To test this we first examined if each of these factors could indeed induce/increase expression of the respective others. Indeed, Sox11 over-expression in adult neurosphere-derived cells increased mRNA for Nfib, Pou3f4 and Pou3f2 over-expression increased Sox11, Pou3f4 and Nfib, but none of these increased Pax6 mRNA levels (Fig. 4I), consistent with the concept of a downstream cross-regulatory transcriptional network. Chromatin-immunoprecipitation (ChIP) followed by quantitative PCR (qPCR) further demonstrated not only Pax6 binding in the promoter regions of Sox11, Nfib and Pou3f4, but also binding of the respective other members of this network (Fig. 5A-C). Moreover, ChIP-seq experiments identified Brg1 binding in the promoters of Sox11, Pou3f4 and Nfib (Fig. 5D) in NSCs isolated from E10.5 embryos. These data further support the concept that Pax6 binds in close interaction with a Brg1-containing BAF complex in the promoter regions of each member of this cross-regulatory network.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f5.jpg
Identification of the “minimal amplifying loop” downstream of Pax6-Brg1 complex necessary for neurogenesis

(A-C) Histograms depicting binding of Pax6, Pou3f4 and Nfib in the promoters of Sox11 (A), Nfib (B) and Pou3f4 (C). X axis indicates the position in the promoters in kb starting from the TSS and underlined position contained predicted Pax6 binding site. Data are shown as mean ± SEM and n(independent experiments)≥5 (D) Genome browser snapshot encompassing segments of DNA spanning the Sox11, Nfib and Pou3f4 loci with the Brg1 ChiP-seq signals. (H) Scheme depicting cross-regulatory loop containing Pou3f4, Sox11 and Nfib and regulated by Pax6-Brg1 complex.

As a further test for Sox11, Pou3f4 and Nfib acting as an cross-regulatory neurogenic network initiated by Pax6 interacting with Brg1-containing chromatin remodeling complexes, we examined to which extent neuronal specification and differentiation genes down-regulated after Brg1 depletion contain a regulatory motif composed of SoxC, Pou3f4 and Nfib binding sites. Interestingly, a total of 65% of genes down-regulated in the Brg1 cKO have binding sites for Nfib, Sox11 and Pou3f4 with conserved distance and orientation (25%) or a regulatory module containing at least two of them (40%, Suppl. Fig. 6A). Sox11, Nfib and Pou3f4 themselves are amongst these regulatory module-containing genes (Suppl. Fig. 6A), further supporting the validity of this motif analysis as these bindings have been confirmed by ChIP-qPCR (Fig. 5A-C). Other than these, however, the set of SoxC/Pou3f4/Nfib binding genes does not comprise further regulatory transcription factors but rather effector molecules involved in neuronal migration and differentiation or interaction to extracellular matrix. These data therefore suggest that Sox11, Pou3f4 and Nfib act as a cross-regulatory transcriptional network downstream of Pax6/Brg1-containing complexes regulating a multitude of effector genes involved in neuronal differentiation and migration, thereby implementing and stabilizing the initial fate (Fig. 5E).

The function of BAF-Pax6 complex is necessary to maintain the core network in the neuroblasts

The above concept suggests an effector network stabilizing neuronal fate in neuroblasts or late stages of TAPs, when Pax6 and the downstream effectors are detectable by immunostaining. Moreover, upon deletion of either Brg1 or Pax6 conversion to gliogenesis occurred in the RMS and core of the OB, i.e. regions mostly composed of neuroblasts. However, we had also observed ependymal cells in the SEZ emerging after Brg1 deletion in NSCs, consistent with a possible direct conversion of NSCs into this fate upon Brg1 deletion. To further test fate conversion after Brg1 deletion at later stages in the lineage, we used MLV-based retroviruses to transduce only fast dividing TAPs and neuroblasts with Cre as previously shown (Colak et al., 2008). Transduced cells were analyzed 21 day after stereotactic virus injection, allowing sufficient time for the transduced progenitors to differentiate in the OB. While 90% of control virus infected cells had arrived in the OB and differentiated into DCX or NeuN+ neurons with the typical morphology of granule cells (Fig. 6A,B,E-G and Suppl. Fig. 6B), cells transduced with the Cre-containing virus were mostly located outside the OB next to the RMS in the cortical WM or the striatum (Fig. 6C,E,F). Accordingly, most Cre transduced cells were Olig2+ or NG2+ glia located outside the SEZ and the RMS (Fig. 6 F,H,I) and only few expressed DCX mostly located in the superficial GCL, a neuronal population spared by Brg1 deletion described above (Fig. 6F and data not shown). Interestingly, Olig2 expression already started in cells within the RMS (Fig. 6 H,I), suggesting that the fate conversion upon Brg1 deletion starts in the RMS in agreement with transcriptional changes in neuroblast genes (Fig. 3A and Suppl. Fig 4). Interestingly, a significant (40%) proportion of the Cre-transduced Brg1 cKO cells remained in the SEZ even 21 days after the transduction. These were non-proliferative (Ki67-negative, data not shown) astrocytes (GFAP+) or CD24+ ependymal cells at the ventricular lumen (Fig. 6K-F). Importantly, the latter were never observed amongst control transduced cells (Fig. 6F), indicating that Brg1 deletion in fast proliferating cells integrating the MLV-virus results in conversion to ependymal cells in the SEZ, i.e. the reverse of ependymal cell-neuroblast conversion after stroke injury (Carlen et al., 2009). Taken together, these experiments the Pax6-BAF complex is required at later stages in the NSC lineage for neuroblast fate maintenance and its absence in these progenitors results in conversion to the glial lineage as they fail to up-regulate the cross-regulatory effector network.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f6.jpg
Brg1 function is necessary in the TAPs and neuroblasts

(A-D) Composite images of cells transduced with control (A,B) and Cre-expressing virus (C, D) 21 day after stereotactic injections of the virus in the SEZ of Brg1 floxed animals. B and D are images of cells settled in the OB. (E) Scheme depicting the position of control virus (green dots) and Cre virus (red dots) transduced cells 21 days after stereotactic injections. (F) Histogram depicting the distribution of control and Brg1-deficient cells in the forebrain 21 days after stereotactic injection. Data are shown as mean (3 amimals) ± SEM. Note that most of the cells deficient for the Brg1 function in the rostral forebrain reside outside of the RMS. (G-L) Micrographs depicting the identity of control virus (G) and Cre virus (H-L) transduced cells 21 days after transduction. (F) Pies depicting the identity of transduced cells in the OB, rostral forebrain and SEZ 21 days after viral transduction. Scale bars: 100 μm in A-D and 20 μm in G-L. Abbreviations: ctx-cerebral cortex; cc-corpus callosum; RMS-rostral migratory stream; SEZ-subependymal zone; St-striatum; lv-lateral ventricle; GCL-granule cell layer; cor-core of the OB.

A minimal neurogenic network is sufficient for forced neurogenesis and independent of Brg1

If Sox11, Nfib and Pou3f4 can indeed function as a neurogenic effector network as suggested by the above loss-of-function experiments, these factors should also be sufficient to replace Pax6 in forced neurogenesis in gain-of-function experiments and act independent of Brg1, as expected for factors downstream of Pax6 and Brg1. We tested these predictions first in adult neurosphere-derived cells which already express Nfib (Suppl. Fig. 7A,B), by introducing Sox11 and/or Pou3f4/Pou3f2. Indeed, transfection with Sox11 increased the proportion of transduced cells differentiating into neurons 6-fold (30% DCX+ cells; 5% after transfection with control dsRed plasmid, see also (Haslinger et al., 2009; Mu et al., 2012)) and Pou3f2 (as well as Pou3f4, data not shown) was even more efficient by instructing about 12-fold more neurons (about 60% DCX+ cells; Fig. 7A). Co-transfection of both, Sox11 and Pou3f2 elicited neurogenesis in 75% of transduced cells, a proportion not significantly different from the neurogenesis elicited by Pax6 over-expression (Fig. 4D). Moreover, Sox11 and Pou3f2 could still induce and enhance neurogenesis in the absence of Brg1 after co-transduction of Cre into Brg1fl/fl neurosphere cells (Fig. 7A), in pronounced difference to Pax6 (Fig. 4D). However, their function was critically dependent upon the presence of the other members of the core regulatory network, as genetic deletion of both Sox4 and Sox11 simultaneously or knock-down of Pou3f2, Pou3f4 or Nfib (Suppl. Fig. 7C,D) significantly reduced neurogenesis of neurosphere cells (Fig. 7B). Likewise, deletion of Sox4 and Sox11 or knock-down of Pou3f2 interferes with forced neurogenesis upon transduction with Pax6, Sox11 or Pou3f2 (Fig. 7C). These data further substantiate the concept of the cross-regulatory effector network, as the other two members (endogenously expressed Nfib or overexpressed Sox11 or Pou3f2) were not sufficient to instruct neurogenesis in the absence of the third member. We therefore conclude that the cross-regulatory transcriptional network of SoxC, Pou3f and Nfib network is sufficient and necessary to achieve equal levels of neurogenesis in the absence of Brg1, consistent with its function downstream of this initiator complex.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f7.jpg
Cross-regulatory loop genes induce neurogenesis from glia cells independent of Pax6-Brg1 complex

(A, B) Histograms depicting the proportion of neurons generated after forced expression (A)or loss of function (B) of cross-regulatory loop genes in cells derived from neurospheres after 7 day in vitro. (C) Histogram depicting the proportion of neurons generated after forced expression of Pax6 or core regulatory network members in SoxC- and Pou3f2-deficient neurosphere cells after 7 days in vitro. (D, E) Micrographs depicting immunoreactivity for neuronal (DCX) and astrocyte (GFAP) marker of cells derived from glia cells enriched for OPC after over-expression of Pax6 (D) and control vector expressing GFP (C) for 7 days in vitro. (E) Histogram depicting the proportion neurons generated from the postnatal glia after the over-expression of Pax6 or its downstream targets. Data in A, B, C and F are shown as mean ± SEM and n(independent experiments)≥7. **-p≤0.01; ***-p≤0.005. Scale bars: 100 μm.

Lastly, we examined to which extent this network may have a broader relevance for reprogramming postnatal glial cells, which would not generate neurons endogenously. Mixed glial cultures from the postnatal cerebral cortex were cultured for 7 days and infected with MLV-based viral vectors encoding for Pax6 or Sox11 or Pou3f2 (as again Nfib was found to be expressed in these cells endogenously) and examined the transduced cells 7 days later. As expected, virtually no DCX+ neurons were observed among control virus infected cells (Fig 7D,E F), while Pax6 was sufficient to instruct neurogenesis in 40% of all transduced cells (Fig. 7E, F). Strikingly, the combination of Sox11 and Pou3f2 was at least as efficient instructing the majority of glial cells towards neurogenesis (Fig. 7F) demonstrating that these factors are indeed able to instruct neuronal differentiation also in glial cells.

Transcription factor Pax6 interacts with BAF chromatin remodeling complex in neurogenic progenitors

In order to understand the mechanisms underlying Pax6-mediated neurogenesis, we purified Pax6-containing complexes from neural stem cells expressing Pax6 (Suppl. Fig. 1A) and used mass spectrometry to examine their composition. Pax6-complexes were purified by either Pax6 antibody (Pax6-IP, Fig. 1A) or FLAG antibody from neural stem cells stably expressing FLAG-tagged Pax6 (FLAG-Pax6-IP, Suppl. Fig. 1B). In either case, multiple subunits of the BAF complex were present in the Pax6 samples. The interaction of Pax6 with the BAF complex was confirmed by western blot (WB) detection of Brg1 and other subunits of the BAF complex in Pax6 immunoprecipitations (Fig. 1A). Thus, Pax6 physically interacts with Brg1-containing BAF chromatin remodeling complexes in neural stem cells. To examine this in the brain, we prepared nuclear extracts from the core of the adult mouse olfactory bulb (OB) which is enriched in Pax6+ neuroblasts (Hack et al., 2005; Brill et al., 2008). Immunoprecipitation with Pax6 antibody followed by WB for Brg1 (Fig. 1A) confirmed the interaction of Brg1-containing complexes with Pax6 in the OB.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f1.jpg
Loss of Brg1 function in neuronal progenitors renders them into glial cells

(A) Western blot depicting the direct interaction of Pax6 and members of BAF complex in the crude protein extract from NS5 cells and the OB. (B-E) Micrographs depicting co-expression of Pax6 and Brg1 in DCX+ neuroblasts (blue arrow in D and E) in the SEZ (D) and the OB (E). D and E are magnifications of area boxed in B and C, respectively. (F, G) Representative micrographs of the olfactory bulbs of Brg1 cKO (G) and corresponding control (F) 28 days after tamoxifen-induced recombination. (H-K) Micrographs depicting the immunoreactivity of recombined cells for DCX and NG2 in Brg1 cKO (I, K) and its WT sibling (H, J). J and K are magnifications of area boxed in D and E, respectively (L-M) Micrographs depicting the diverse morphology of NG2-positive cells generated from the Brg1-deficient neural progenitors. (N) Histogram depicting the total number of NG2 positive cells in different OB layers 28 days after recombination. Data are shown as mean ± SEM and n(animals analyzed)≥5. *-p≤0.05; ***-p≤0.005. (O) Pie charts illustrating the identity of recombined cells in the OB of Brg1 cKO and age-matched siblings 28 days after tamoxifen-induced recombination. Data are shown as mean and n(animals analyzed)≥5. Scale bars: 100 μm in B,C, F and G; 50 μm in L-M′; 20 μm in D, E, H-K. Abbreviations: ctx-cerebral cortex; cc-corpus callosum; RMS-rostral migratory stream; SEZ-subependymal zone; St-striatum; lv-lateral ventricle; GL-glomerular layer; GCL-granule cell layer; EPl-external plexiform layer; dGCL-deep granule cell layer and sGCL- superficial granule cell layer.

To examine co-localization of Pax6 and Brg1 at the cellular level we performed immuno-staining for Brg1 in the adult brain, which showed a broad expression of Brg1 in neurons and astrocytes throughout the brain, while surprisingly weak immunoreactivity for Brg1 was detectable in white matter (WM) where mostly oligodendrocytes and their precursors reside (Fig. 1B; Suppl. Fig. 1C). In the regions of adult neurogenesis, we focused on the region generating OB interneurons where Pax6 has been shown to regulate neurogenesis (Hack et al., 2005; Brill et al. 2008). These adult-generated neurons derive from neural stem cells (NSCs) located in the subependymal zone (SEZ) that generate via transient-amplifying progenitors (TAPs) neuroblasts migrating through the rostral migratory stream (RMS) towards the OB (Ming and Song, 2012). Brg1-immunoreactive nuclei were detected in GFAP+ astrocytes and stem cells as well as in Ascl1+ TAPs and Doublecortin (DCX)+ neuroblasts (Suppl. Fig. 1D,E and data not shown) also expressing BAF53a and BAF45a (Suppl. Fig. 1F,G; but not BAF45b,53b in H,I), subunits characterizing the neural progenitor specific BAF complexes (Lessard et al., 2007). Conversely, Pax6 immunoreactivity is largely restricted to neuroblasts in the SEZ and RMS, where it colocalizes with Brg1 (Fig. 1D-E). Together with the immunoprecipitation experiments, these data suggest that Pax6 interacts with the neural progenitor-specific, Brg1-containing BAF chromatin remodeling complex in neuroblasts in vivo.

Loss of Brg1 in adult NSCs converts olfactory bulb neurogenesis into gliogenesis

To test the functional relevance of the observed Pax6-BAF interaction, we ablated Brg1 in the Brg1fl mouse line (Indra et al., 2005; Matsumoto et al., 2006) by tamoxifen (TM) inducible Cre-based excision in Glast mice, mediating genomic recombination in astrocytes and NSCs (Mori et al., 2006; Ninkovic et al., 2007). As these mice were also crossed with the CAG-CAT-GFP reporter line (Nakamura et al., 2006), this allowed visualization of the recombined cells. By 9 days post TM administration (9dpt) 95 % of reporter positive cells in Glast/Brg1 mice (further referred as Brg1 cKO) were no longer Brg1-immunopositive, while virtually all reporter positive cells were Brg1+ in Glast/Brg1 or Glast/Brg1 mice (further referred as Brg1 controls) (Suppl. Fig. 2A-C).

While no altered neurogenesis was detectable in the SEZ, RMS or OB of Brg1 cKO mice 9 dpt (Suppl. Fig. 2D), 28 dpt the number of recombined cells was significantly reduced in the OB, the final destination of adult generated neurons (Fig. 1F,G). As this may be due to cell death, we examined activated caspase3, an indicator of programmed cell death. Indeed, the number of activated caspase3+ cells was significantly increased in the OB and RMS but not in the SEZ of Brg1 cKO mice compared to controls (Suppl. Fig. 2E and data not shown). This suggests that cell death is initiated at later stages in the neuroblasts when they migrate along the RMS. Accordingly, we also observed an increase in GFP-positive cells located just beside the RMS (Suppl. Fig. 2F-H) with a morphology reminiscent of oligodendrocyte progenitor cells (OPCs). Staining for the transcription factor Olig2 (data not shown) and proteoglycan NG2 labeling OPCs (Dimou et al., 2008) confirmed the strikingly higher number of Brg1cKO GFP+ OPCs (Fig. 1I,K-O; 2000x increase) while virtually no GFP+ OPCs were detectable in the OB of control animals (Fig. 1H,J). Notably, GFP+ OPCs were present only at the end of the RMS, the core of the OB and the deep granule cell layer (GCL), but were absent from the glomerular layer (GL) in Brg1 cKO mice (Fig. 1N). The newly generated OPCs displayed a variety of morphologies with different level of cellular complexity (Fig. 1K-M) indicative of different stages in the oligodendrocyte lineage. However, even 2 months after TM, no GFP+ cells had matured into oligodendrocytes immunoreactive for GST-Π (Suppl. Fig. 2I), consistent with the behavior of OPCs in the GM (Dimou et al., 2008). Interestingly, other glial cells - like GFAP+ astrocytes (Fig. 1O and Suppl. Fig. 2J-K′) and marker-negative cells (Suppl. Fig. 2L,L′, cells were immunoprobed for more than 20 antigens indicative of astroglial, neuronal, oligodendroglial, endothelial, microglial lineage as well as fibroblasts, see material and methods) were also increased in Brg1cKO OB, reflecting a broader conversion towards gliogenesis and some cells failing to adopt any coherent cell identity after loss of Brg1 (Fig. 1O and Suppl. Fig. 2L). Conversely, cells of the neuronal lineage, labeled with DCX (neuroblasts and young neurons) or NeuN (recently matured neurons), were strongly reduced in number to less than a third (Fig. 1O and Suppl. Fig. 2M), with the only exception of neurons in the superficial GCL (Suppl. Fig. 2M), indicating that this neuronal sub-lineage is selectively spared by Brg1 depletion and accounts for almost half of the remaining neurons. Taken together, inducible, cell-specific deletion of Brg1 in vivo converts adult OB neurogenesis to gliogenesis starting in the RMS and is accompanied by increased cell death.

GLAST-mediated recombination is not restricted to the SEZ and RMS, the origin of adult OB neurogenesis, but also depletes Brg1 in astrocytes throughout the brain. Therefore, it is possible that some astrocytes in the OB and RMS may be converted to OPCs. While no GFP+ OPCs were detected outside the OB (e.g. cerebellum or dentate gyrus, data not shown), it remains possible that specifically astrocytes in the OB would be more easily converted to OPCs. To examine the fate of cells originating in the SEZ directly, we injected DsRed-expressing MLV-based retroviral vectors 9 dpt into the SEZ of GLAST//Brg1//CAT-CAG-GFP animals in order to label Brg1cKO progenitors (DsRed+/GFP+) and their WT cellular counterparts (not recombined and hence GFP-, but DsRed+) already in the SEZ (Suppl. Fig. 3A,B). When we analyzed the identity of labeled cells in the OB 7 days later, most of the DsRed+, GFP- control cells (more than 90%) were DCX+ neuroblasts, as is normally the case (Suppl. Fig. 3B). However, only a minority (40%) of the Brg1-deficient cells (DsRed+,GFP+) originating in the SEZ had acquired a neuroblast identity (Suppl. Fig 3B) and a significant proportion expressed Olig2 (Suppl. Fig. B). As this proportion was similar to the GFP+, DsRed- (38% for DCX and 36 % for Olig2), we conclude that there is no major additional source of cells contributing to the OPCs in the OB of Brg1 cKO mice. Thus, most cells from the SEZ fail to complete their neurogenic fate upon Brg1 deletion and convert to gliogenesis.

Viral vector injection causes an injury in the SEZ, which may affect the lineage progression. To address this possibility and verify the above findings, we deleted Brg1 with the Nestin-CreER line (Lagace et al., 2007) mediating recombination exclusively in nestin+ cells of the SEZ but not in the nestin- parenchymal glia. Consistent with our findings with the Glast mice shown above, mice that lack Brg1 in nestin+ cells and their progeny had significantly fewer GFP+ cells reaching the OB and acquiring a neuronal identity compared to control mice 30 dpt (Suppl. Fig. 3C-D). Similar to the Glast-mediated deletion of Brg1, these cells expressed Olig2. Taken together, multiple independent experimental approaches confirm that Brg1 is an essential component of SEZ-derived neurogenesis, and that the absence of Brg1 causes the conversion of neuroblasts in the RMS and OB to the glial lineage.

Niche-dependent gliogenesis elicited by loss of Brg1

As our fate mapping experiments demonstrated that the reporter+ glia in the OB were converted from neuroblasts originating from the SEZ, we examined cells in the SEZ of Brg1 cKO mice. However, even 28dpt (17 days after loss of Brg1 protein), numbers of DCX+ neuroblasts amongst the recombined (GFP+) cells in the SEZ did not significantly differ between Brg1cKO and control mice (Suppl. Fig. 3E,F), suggesting that conversion to gliogenesis occurs only when neuronal cells exit the SEZ. This strikingly late fate conversion may be due to a powerful role of the niche environment or other intrinsic mechanisms sufficient to stabilize neuronal fate for a period of time (Beckervordersandforth et al., 2010; Colak et al., 2008; Lim et al., 2006). To distinguish between these two possibilities, we isolated progenitors from the SEZ of Brg1cKO or control mice 9dpt and cultured them in a primary SEZ culture system maintaining neurogenesis (Costa et al., 2011; Ortega et al., 2011). Consistent with the in vivo analysis of neurogenesis in the SEZ (Suppl. Fig. 3E), we did not observe any difference in cell composition of the SEZ in Brg1 cKO and respective controls after isolating cells 7 dpt and analyzing them 4h after plating (Suppl. Fig. 3G). However, cells lacking Brg1 were not able to proceed further along the neurogenic lineage also after 7 days in vitro, as only 9% of them were DCX+ neuroblasts in contrast to 60% of control cells (Fig. 2A-D). Intriguingly, Brg1-deficient cells now had acquired different glial identities, with 15% NG2+ OPCs, 22% GFAP+ astrocytes and 49% S100b+, CD24+ ependyma-like cells (Fig. 2C, D and Suppl. Fig. 3H).

An external file that holds a picture, illustration, etc.
Object name is nihms595448f2.jpg
Loss of Brg1 results in full conversion to gliogenesis in vitro

(A-C) Micrographs depicting the immunoreactivity of GFP+ adult neural stem cell progeny derived from either Brg1 cKO or age matching controls (10 days after TM induction) for DCX (neuroblasts) or NG2 (OPCs) after 7 days in vitro. (D) Pie charts depicting the composition of the adult neural stem cell progeny following Brg1 depletion (lower pie chart) or age matching controls (upper panel) after 7days in vitro. Data are shown as mean and n(animals analyzed)≥7. (E-F) Representative trees illustrating the predominant behavior of the Brg1 KO or control neural progenitors in vitro followed by continuous live imaging at the single cell level (see also Suppl. Movies). (G) Box-chart showing the cell cycle length (defined as the time between two divisions) of control and Brg1-deficieng progenitors. (H) Histogram depicting the proportion of clones (control or Brg1-deficient) containing at least one dead cell. Data are shown as mean (3 independent experiments) ± SEM. Scale bars: 100 μm in A, B and C.

These observations prompted us to ask if ependymal cells are generated also in the SEZ in vivo after Brg1 deletion. As both Glast and nestin are expressed in ependymal cells (Beckervordersandforth et al., 2010; Lagace et al., 2007), they are also reporter+ in the above mouse lines and their possible increase is difficult to detect. To overcome this limitation, we used the recently developed Split-Cre technology that specifically mediates recombination in SEZ NSCs using two halves of Cre driven by GFAP and P2 prominin promoters simultaneously active in NSCs (Beckervordersandforth et al., 2010). Stereotactic injection of 2 lentiviruses encoding each half of Cre under the respective GFAP or P2 promoter into Brg1//CAG-CAT-GFP or Brg1//CAG-CAT-GFP mice allows to selectively delete Brg1 in NSCs and to follow their progeny by the GFP reporter. The majority of GFP+ cells were neuroblasts 60 days after virus injection into control mice with very few ependymal cells labeled (Suppl. Fig. 3I-K), while the progeny of Brg1-deficient NSCs contained 30% ependymal cells (Suppl. Fig. 3I-K). Taken together, the local niche not only influences the selection of glial fate subtype after Brg1 deletion, but also contributes to maintenance of some cells as neuroblasts even in the absence of Brg1, as virtually all cells convert to gliogenesis outside this neurogenic niche in vitro.

Mode of fate conversion determined by continuous single cell live imaging in vitro

As Brg1-deficient cells in vitro largely convert into the same cell types as in vivo, we used single cell continuous live imaging for 7 days in vitro followed by postimaging immunostaining to discriminate whether the conversion of Brg1 cKO cells from neuro- to gliogenesis occurs by selective cell death, selective proliferation or a true fate conversion (Suppl. Movies 1-3; Fig. 2E-I). In agreement with the population based analysis, 93% of lineage trees of single control cells and their entire progeny as observed by live imaging contained only neuronal cells and only a small fraction of trees, the NSC progeny, had generated both neurons and glia (Fig. 2 E,G). In contrast, when we observed Brg1 cKO cells most of the trees contained either NG2 glia or ependyma-like cells, but only a minority had generated neurons only (Fig. 2F). Interestingly, cells giving rise to ependyma-like cells were the only lineage that initially perfermed a series of fast symmetric proliferative divisions (Fig. 2F). Other than this, cell cycle length did not differ between control and Brg1 cKO cells (Fig. 2G), nor could we observe any difference in regard to cell death that was rather rare in both control and Brg1 cKO cells (Fig. 2E,F,H). Thus, Brg1 cKO cells convert to glial lineages by fate change rather than selective cell death or proliferation. Interestingly, some Brg1 cKO cells generate neurons and NG2+ OPCs, while cells generating ependymal progeny did not give rise to any other glial or neuronal cells, suggesting that fate conversion occurs distinctly into different glial lineages. Moreover, as most of the control and Brg1 cKO cells at the start of these cultures are neuroblasts (63±4 %, Suppl. Fig. 2D), and no selective cell death of Brg1 cKO cells was observed, these experiments demonstrate that many Brg1 cKO neuroblasts directly convert to glial lineages.

Loss of Brg1 leads to down-regulation of Pax6 targets in adult SEZ and OB

Given the surprisingly specific defects in neurogenesis after deletion of Brg1, we used genome-wide expression analysis to determine which genes may be involved in this phenotype as well as to clarify additional gene regulation effects indicative of a further phenotype we may have missed so far. Towards this aim, we isolated the SEZ and core of the OB (containing neuroblasts entering the OB) from Brg1 cKO and control mice 10 dpt, just 1 day after loss of Brg1 protein. Genome-wide expression profiling by microarray (GST 1.1 gene array, Affymetrix, USA) revealed 244 significantly regulated genes (change in the expression level >1.2 or <0.8 and p<0.05) in the OB (Suppl. Table 1) and 136 genes in the SEZ (Suppl. Table 2 and Suppl. Fig. 4A) and qPCR on independent samples confirmed the reliability of this analysis (Fig. 3A). The majority of genes altered in their expression were down-regulated consistent with a role of Brg1-containing chromatin remodeling complexes in positively regulating gene transcription. Importantly, most of the significant gene ontology (GO) terms were related to neurogenesis, synaptic transmission, axonogenesis and cell migration, consistent with defective neurogenesis (Suppl. Fig. 4B,C). The other major terms were related to DNA and RNA metabolism and cell replication and cell cycle, suggestive of defects in cell cycle regulation in Brg1cKO SEZ and OB and possibly linked to the very specific cell division pattern of ependymal cells generation we observed using live imaging (Fig. 2F). Interestingly, 64% of the down-regulated genes in SEZ (Suppl. Fig. 4E) are expressed in NSCs and their progeny, as identified in our previous transcriptome analysis (Beckervordersandforth et al., 2010 and Suppl. Fig. 4D-F). Most intriguingly, 87% of the down-regulated genes had predicted Pax6 binding sites, a significant enrichment in regard to the expected frequency (63% expected, Genomatix Matinspector) (Fig. 3B). This suggests that most genes down-regulated in Brg1 cKO may be regulated by Pax6 in conjunction with a Brg1-containing BAF complex.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f3.jpg
Brg1 deletion results in down-regulation of Pax6 target genes and is phenocopied by Pax6 deletion

(A) Histogram depicting the comparison of a specific gene set miss-regulated after Brg1 deletion and measured by microarray and qPCR on independent samples. Data are shown as mean and n(animals analyzed)≥3. (B) Venn diagram depicting predicted Pax6 binding sites in the promoters of genes deregulated following Brg1 deletion. (C, D) Representative micrographs of the olfactory bulb of Pax6 cKO (C) and control (D) 60 days after tamoxifen-induced recombination. (E, F) Micrographs depicting the immunoreactivity of recombined, GFP+ cells in Pax6 cKO (E) and its control sibling (F) for DCX (neuroblasts) and NG2 (OPCs). (G) Histograms depicting the total number of NG2 positive cells in different OB layers 60 days after recombination. Data are shown as mean ± SEM and n(animals analyzed)≥7. *-p≤0.05; ***-p≤0.005. (H) Pie charts illustrating the identity of recombined cells in the OB of Pax6 cKO and age-matching sibling 60 days after tamoxifen-induced recombination. Data are shown as mean and n(animals analyzed)≥3. Scale bars: 100 μm in C and D; 20 μm in E and F. Abbreviations: RMS-rostral migratory stream; GL-glomerular layer; EPl-external plexiform layer; dGCL-deep granule cell layer and sGCL-superficial granule cell layer.

Deletion of Pax6 phenocopies Brg1 deletion and converts adult SEZ neurogenesis to gliogenesis

In order to test the above suggestion directly in vivo, we crossed the floxed Pax6 mice (Ashery-Padan et al., 2000) with GLAST mice to delete Pax6 as before Brg1. While Pax6 protein was more stable than Brg1 and disappeared only 21 dpt (Suppl. Fig. 5 A-C), the phenotype emerging thereafter was remarkably similar to the phenotype observed in Brg1 cKO mice. As in the Brg1 cKO mice, we observed fewer numbers of GFP+ cells reaching the OB in the Pax6 cKO mice compared to controls (Fig. 3C,D), and most of them no longer differentiated along the neuronal lineage (DCX+/NeuN+ 45%, compared to 82% in controls, Fig. 3E-H), but rather converted to glial identities (Fig. 3G-H) in ratios similar to the Brg1 cKO cells (Fig. 1H-O). Thus, consistent with the interaction of Pax6 with Brg1-containing BAF complex in vitro and in vivo, deletion of either of these proteins results in the same phenotype with severe defects in OB neurogenesis implying a key role of this transcriptional complex in regulating neurogenesis.

Pax6 and catalytically active Brg1 are both essential for forced neurogenesis

The above findings prompted us to examine to which extent Pax6-induced neurogenesis would also require the presence of Brg1. Towards this end, we prepared neurosphere cells from the adult SEZ of Brg1 fl/fl animals and transduced them in vitro with viral vectors encoding Cre to obtain Brg1-deficient cells (Fig. 4A-D). Two days after transduction, Brg1 protein was virtually absent from the Cre-transduced cells, while 95% of control virus transduced cells were still Brg1+ (data not shown). Consistent with the low endogenous neurogenic potential of these cells, 15% of control virus (GFP only) infected cells had differentiated into DCX+ neuroblasts (Fig. 4A-D), whereas Cre-transduced, Brg1cKO cells largely failed to generate any DCX+ neurons (Fig. 4B,D), in agreement with the above observed results that Brg1 is necessary for endogenous neurogenesis. Brg1 is also essential for neurogenesis elicited by Pax6 over-expression. Pax6 over-expression did not induce neurogenesis in Brg1 cKO cells (co-transduced with CreIRESGFP virus) in contrast to the Ctrl situation where Pax6 transduction induced more than 80% DCX+ neurons (Fig. 4D). Importantly, the full-length form of Brg1, but not the ATPase deficient form (Brg1KS), restored both endogenous and Pax6-induced neurogenesis (Fig. 4E), demonstrating that Brg1 needs to be catalytically active to mediate neurogenesis. The requirement of Brg1 for neurogenesis is rather specific for Pax6, as transduction with the neurogenic factor Ngn2 resulted in very efficient induction of neurogenesis to over 90% even in the absence of Brg1 (Fig. 4C, D). Thus, Pax6 and Brg1 are not only required for endogenous OB neurogenesis, but also neurogenesis forced by Pax6 over-expression requires the presence of catalytically active Brg1 and hence chromatin remodeling activity.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f4.jpg
Pax6 requires catalytically active Brg1 for its neurogenic function

(A-C) Micrographs depicting immunoreactivity of Brg1-deficient cells (green) derived from Brg1 neurospheres for the neuroblast marker DCX after over-expression of control vector expressing GFP (A), Pax6 (B) and Ngn2 (C) 7 days after transduction. (D, E) Histograms depicting the proportion neurons generated from Brg1cKO and control neurosphere cells 7 days after expression of Pax6, Ngn2, Brg1 and Brg1KS (ATPase deficient form). Note that only catalytically active Brg1 restores the neurogenic function of Pax6. (F) Histogram depicting down-regulation of genes following Brg1 deletion (open bars) and Pax6 deletion (grey bars) measured by qPCR. Data are shown as ratio of average gene expression in the mutant and age-matching WT normalized to the expression of GAPDH housekeeping gene. (G) Venn diagram depicting transcription factors down-regulated after Brg1 deletion and harboring Pax6 binding site in their promoters. (H) Histogram showing the down-regulation of Pou3f4, Sox11 and Nfib in the OB and SEZ following Brg1 and Pax6 deletion. (I) Histogram depicting the induction of the expression of Pou3f4 and Sox11 after forced Pax6 expression in neurosphere cells measured 24 hours after transduction. (J, K) Micrographs depicting deletion of Brg1 in neurosphere cells 36h after nucleofection of control (J) or Cre (K) encoding plasmid. (L) Histogram depicting the induction of the expression of Pou3f4, Sox11 and Nfib after forced Pax6 overexpression in control or Brg1-deficient neurosphere cells measured 24 hours after Pax6 nucleofection. (M) Histogram depicting the relative expression of Pou3f4, Sox11 and Nfib in the acutely isolated, FACS purified population enriched in neural stem cells and their progeny. Scale bars: 100 μm in A, B, C, J and K. Data in D, E, H, I, and L are shown as mean ± SEM and n(animals analyzed)≥3 and in M as mean n(animals analyzed)≥3. **-p≤0.01; ***-p≤0.005.

Identification of a neurogenic transcriptional network downstream of Pax6 and Brg1

Given the similarity of the phenotypes after deletion of either Pax6 or Brg1 in adult NSCs of the SEZ and functional interaction in forced neurogenesis, we next asked to which extent this is also reflected at the transcriptional level. First, we examined randomly selected genes, found to be down-regulated in the core of the OB or the SEZ of Brg1 cKO mice, in the OB and SEZ of Pax6 cKO mice. Consistent with the similarity at the phenotypic level, 90% of these genes down-regulated after loss of Brg1 were also down-regulated after loss of Pax6 (Fig. 4F and data not shown), in agreement with the observation that the majority of these genes possess Pax6-binding sites. This provided us with an opportunity to search for genes implementing neurogenesis as this is deficient in both these mutants. We therefore searched for transcriptional regulators down-regulated upon Brg1 depletion with a common regulatory motif including Pax6. This was the case for 7 of the 11 down-regulated transcription factors (Fig. 4G; FrameWorker, Genomatix, Germany) and only for 4 of these a specific consensus DNA binding sequence was described (Sox11, Sox4, Pou3f4 and Nfib). Indeed, all these are expressed in the SEZ and RMS (Suppl. Fig. 5G-J) and their expression is down-regulated in both Brg1cKO and Pax6 cKO OB (Fig. 4H). This reduction was relevant at protein levels; for example Nfib was present in virtually all neuroblasts in the RMS, but upon loss of Pax6 (GFP+ cells in Pax6 cKO 60 dpt) only 17% of recombined, GFP+ neuroblasts expressed Nfib (Suppl. Fig. 5K-M).

Pax6 was also sufficient to induce Sox11 and Pou3f2/4 expression within 24hours in neurosphere-derived cells (Fig. 4I) in a Brg1-dependent manner (Fig. 4J-L). Interestingly, the expression of Sox11, Nfib and Pou3f4 was highest in neuroblasts expressing the highest levels of Pax6 (Fig. 4M). Thus, the regulation and expression of these transcription factors is consistent with a role downstream of the Pax6-Brg1 complex in neuroblasts.

Most importantly, these downstream transcription factors were also predicted to cross-regulate each other, thus potentially forming a self-sustaining corss-regulatory network critical for neurogenic fate maintenance in the adult brain. To test this we first examined if each of these factors could indeed induce/increase expression of the respective others. Indeed, Sox11 over-expression in adult neurosphere-derived cells increased mRNA for Nfib, Pou3f4 and Pou3f2 over-expression increased Sox11, Pou3f4 and Nfib, but none of these increased Pax6 mRNA levels (Fig. 4I), consistent with the concept of a downstream cross-regulatory transcriptional network. Chromatin-immunoprecipitation (ChIP) followed by quantitative PCR (qPCR) further demonstrated not only Pax6 binding in the promoter regions of Sox11, Nfib and Pou3f4, but also binding of the respective other members of this network (Fig. 5A-C). Moreover, ChIP-seq experiments identified Brg1 binding in the promoters of Sox11, Pou3f4 and Nfib (Fig. 5D) in NSCs isolated from E10.5 embryos. These data further support the concept that Pax6 binds in close interaction with a Brg1-containing BAF complex in the promoter regions of each member of this cross-regulatory network.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f5.jpg
Identification of the “minimal amplifying loop” downstream of Pax6-Brg1 complex necessary for neurogenesis

(A-C) Histograms depicting binding of Pax6, Pou3f4 and Nfib in the promoters of Sox11 (A), Nfib (B) and Pou3f4 (C). X axis indicates the position in the promoters in kb starting from the TSS and underlined position contained predicted Pax6 binding site. Data are shown as mean ± SEM and n(independent experiments)≥5 (D) Genome browser snapshot encompassing segments of DNA spanning the Sox11, Nfib and Pou3f4 loci with the Brg1 ChiP-seq signals. (H) Scheme depicting cross-regulatory loop containing Pou3f4, Sox11 and Nfib and regulated by Pax6-Brg1 complex.

As a further test for Sox11, Pou3f4 and Nfib acting as an cross-regulatory neurogenic network initiated by Pax6 interacting with Brg1-containing chromatin remodeling complexes, we examined to which extent neuronal specification and differentiation genes down-regulated after Brg1 depletion contain a regulatory motif composed of SoxC, Pou3f4 and Nfib binding sites. Interestingly, a total of 65% of genes down-regulated in the Brg1 cKO have binding sites for Nfib, Sox11 and Pou3f4 with conserved distance and orientation (25%) or a regulatory module containing at least two of them (40%, Suppl. Fig. 6A). Sox11, Nfib and Pou3f4 themselves are amongst these regulatory module-containing genes (Suppl. Fig. 6A), further supporting the validity of this motif analysis as these bindings have been confirmed by ChIP-qPCR (Fig. 5A-C). Other than these, however, the set of SoxC/Pou3f4/Nfib binding genes does not comprise further regulatory transcription factors but rather effector molecules involved in neuronal migration and differentiation or interaction to extracellular matrix. These data therefore suggest that Sox11, Pou3f4 and Nfib act as a cross-regulatory transcriptional network downstream of Pax6/Brg1-containing complexes regulating a multitude of effector genes involved in neuronal differentiation and migration, thereby implementing and stabilizing the initial fate (Fig. 5E).

The function of BAF-Pax6 complex is necessary to maintain the core network in the neuroblasts

The above concept suggests an effector network stabilizing neuronal fate in neuroblasts or late stages of TAPs, when Pax6 and the downstream effectors are detectable by immunostaining. Moreover, upon deletion of either Brg1 or Pax6 conversion to gliogenesis occurred in the RMS and core of the OB, i.e. regions mostly composed of neuroblasts. However, we had also observed ependymal cells in the SEZ emerging after Brg1 deletion in NSCs, consistent with a possible direct conversion of NSCs into this fate upon Brg1 deletion. To further test fate conversion after Brg1 deletion at later stages in the lineage, we used MLV-based retroviruses to transduce only fast dividing TAPs and neuroblasts with Cre as previously shown (Colak et al., 2008). Transduced cells were analyzed 21 day after stereotactic virus injection, allowing sufficient time for the transduced progenitors to differentiate in the OB. While 90% of control virus infected cells had arrived in the OB and differentiated into DCX or NeuN+ neurons with the typical morphology of granule cells (Fig. 6A,B,E-G and Suppl. Fig. 6B), cells transduced with the Cre-containing virus were mostly located outside the OB next to the RMS in the cortical WM or the striatum (Fig. 6C,E,F). Accordingly, most Cre transduced cells were Olig2+ or NG2+ glia located outside the SEZ and the RMS (Fig. 6 F,H,I) and only few expressed DCX mostly located in the superficial GCL, a neuronal population spared by Brg1 deletion described above (Fig. 6F and data not shown). Interestingly, Olig2 expression already started in cells within the RMS (Fig. 6 H,I), suggesting that the fate conversion upon Brg1 deletion starts in the RMS in agreement with transcriptional changes in neuroblast genes (Fig. 3A and Suppl. Fig 4). Interestingly, a significant (40%) proportion of the Cre-transduced Brg1 cKO cells remained in the SEZ even 21 days after the transduction. These were non-proliferative (Ki67-negative, data not shown) astrocytes (GFAP+) or CD24+ ependymal cells at the ventricular lumen (Fig. 6K-F). Importantly, the latter were never observed amongst control transduced cells (Fig. 6F), indicating that Brg1 deletion in fast proliferating cells integrating the MLV-virus results in conversion to ependymal cells in the SEZ, i.e. the reverse of ependymal cell-neuroblast conversion after stroke injury (Carlen et al., 2009). Taken together, these experiments the Pax6-BAF complex is required at later stages in the NSC lineage for neuroblast fate maintenance and its absence in these progenitors results in conversion to the glial lineage as they fail to up-regulate the cross-regulatory effector network.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f6.jpg
Brg1 function is necessary in the TAPs and neuroblasts

(A-D) Composite images of cells transduced with control (A,B) and Cre-expressing virus (C, D) 21 day after stereotactic injections of the virus in the SEZ of Brg1 floxed animals. B and D are images of cells settled in the OB. (E) Scheme depicting the position of control virus (green dots) and Cre virus (red dots) transduced cells 21 days after stereotactic injections. (F) Histogram depicting the distribution of control and Brg1-deficient cells in the forebrain 21 days after stereotactic injection. Data are shown as mean (3 amimals) ± SEM. Note that most of the cells deficient for the Brg1 function in the rostral forebrain reside outside of the RMS. (G-L) Micrographs depicting the identity of control virus (G) and Cre virus (H-L) transduced cells 21 days after transduction. (F) Pies depicting the identity of transduced cells in the OB, rostral forebrain and SEZ 21 days after viral transduction. Scale bars: 100 μm in A-D and 20 μm in G-L. Abbreviations: ctx-cerebral cortex; cc-corpus callosum; RMS-rostral migratory stream; SEZ-subependymal zone; St-striatum; lv-lateral ventricle; GCL-granule cell layer; cor-core of the OB.

A minimal neurogenic network is sufficient for forced neurogenesis and independent of Brg1

If Sox11, Nfib and Pou3f4 can indeed function as a neurogenic effector network as suggested by the above loss-of-function experiments, these factors should also be sufficient to replace Pax6 in forced neurogenesis in gain-of-function experiments and act independent of Brg1, as expected for factors downstream of Pax6 and Brg1. We tested these predictions first in adult neurosphere-derived cells which already express Nfib (Suppl. Fig. 7A,B), by introducing Sox11 and/or Pou3f4/Pou3f2. Indeed, transfection with Sox11 increased the proportion of transduced cells differentiating into neurons 6-fold (30% DCX+ cells; 5% after transfection with control dsRed plasmid, see also (Haslinger et al., 2009; Mu et al., 2012)) and Pou3f2 (as well as Pou3f4, data not shown) was even more efficient by instructing about 12-fold more neurons (about 60% DCX+ cells; Fig. 7A). Co-transfection of both, Sox11 and Pou3f2 elicited neurogenesis in 75% of transduced cells, a proportion not significantly different from the neurogenesis elicited by Pax6 over-expression (Fig. 4D). Moreover, Sox11 and Pou3f2 could still induce and enhance neurogenesis in the absence of Brg1 after co-transduction of Cre into Brg1fl/fl neurosphere cells (Fig. 7A), in pronounced difference to Pax6 (Fig. 4D). However, their function was critically dependent upon the presence of the other members of the core regulatory network, as genetic deletion of both Sox4 and Sox11 simultaneously or knock-down of Pou3f2, Pou3f4 or Nfib (Suppl. Fig. 7C,D) significantly reduced neurogenesis of neurosphere cells (Fig. 7B). Likewise, deletion of Sox4 and Sox11 or knock-down of Pou3f2 interferes with forced neurogenesis upon transduction with Pax6, Sox11 or Pou3f2 (Fig. 7C). These data further substantiate the concept of the cross-regulatory effector network, as the other two members (endogenously expressed Nfib or overexpressed Sox11 or Pou3f2) were not sufficient to instruct neurogenesis in the absence of the third member. We therefore conclude that the cross-regulatory transcriptional network of SoxC, Pou3f and Nfib network is sufficient and necessary to achieve equal levels of neurogenesis in the absence of Brg1, consistent with its function downstream of this initiator complex.

An external file that holds a picture, illustration, etc.
Object name is nihms595448f7.jpg
Cross-regulatory loop genes induce neurogenesis from glia cells independent of Pax6-Brg1 complex

(A, B) Histograms depicting the proportion of neurons generated after forced expression (A)or loss of function (B) of cross-regulatory loop genes in cells derived from neurospheres after 7 day in vitro. (C) Histogram depicting the proportion of neurons generated after forced expression of Pax6 or core regulatory network members in SoxC- and Pou3f2-deficient neurosphere cells after 7 days in vitro. (D, E) Micrographs depicting immunoreactivity for neuronal (DCX) and astrocyte (GFAP) marker of cells derived from glia cells enriched for OPC after over-expression of Pax6 (D) and control vector expressing GFP (C) for 7 days in vitro. (E) Histogram depicting the proportion neurons generated from the postnatal glia after the over-expression of Pax6 or its downstream targets. Data in A, B, C and F are shown as mean ± SEM and n(independent experiments)≥7. **-p≤0.01; ***-p≤0.005. Scale bars: 100 μm.

Lastly, we examined to which extent this network may have a broader relevance for reprogramming postnatal glial cells, which would not generate neurons endogenously. Mixed glial cultures from the postnatal cerebral cortex were cultured for 7 days and infected with MLV-based viral vectors encoding for Pax6 or Sox11 or Pou3f2 (as again Nfib was found to be expressed in these cells endogenously) and examined the transduced cells 7 days later. As expected, virtually no DCX+ neurons were observed among control virus infected cells (Fig 7D,E F), while Pax6 was sufficient to instruct neurogenesis in 40% of all transduced cells (Fig. 7E, F). Strikingly, the combination of Sox11 and Pou3f2 was at least as efficient instructing the majority of glial cells towards neurogenesis (Fig. 7F) demonstrating that these factors are indeed able to instruct neuronal differentiation also in glial cells.

Discussion

Here, we demonstrate a role of chromatin remodeling via the BAF complex in conjunction with a specific neurogenic fate determinant, Pax6, in neuronal fate commitment in the adult mouse brain in vivo. We showed a physical interaction between Pax6 and a Brg1-containing BAF complex and the functional requirement of either factor for adult neurogenesis. If only one of these is missing, neuroblasts originating in the SEZ differentiate into glial cells, either in the SEZ, en-route to the OB or even after completing the normal migration reaching the OB. This drastic phenotype was observed after inducible deletion of either Brg1 or Pax6 in SEZ NSCs or by deletion targeting fast proliferating SEZ progenitors, i.e. largely neuroblasts (Colak et al., 2008). Intriguingly, the conversion occurs largely to NG2 glia in the OB and parenchymal regions surrounding the RMS, while many ependymal cells are generated within the SEZ. This is of particular interest in the light of signals inhibiting oligodendrogliogenesis in this region (Colak et al., 2008) and in regard to the finding that even fast proliferating cells infected with the MLV-Cre virus can differentiate into ependymal cells when not able to complete their neurogenic fate.

We conclude that these common defects in neurogenesis are due the absence of Pax6 interacting with a Brg1-containing BAF chromatin-remodeling complex in neuroblasts. Moreover, we did not observe direct transcriptional cross-regulation of Pax6 and Brg1 and showed that neurogenesis forced by other neurogenic transcription factors, such as Dlx2 (data not shown) or Neurog2, is not affected by loss of Brg1, while loss of Brg1 affects with a striking specificity Pax6-mediated neurogenesis. The specificity of this function is particularly intriguing, as Brm, the other ATPase subunit of BAF complexes, is expressed at even higher levels than Brg1 in neuroblasts (Suppl. Fig. 7E,F), but is obviously not able to compensate for the loss of Brg1. Thus, not only individual BAF subunits (Lessard et al., 2007; Yoo et al., 2009) convey specificity to a BAF complex, but also the respective ATPase subunit. This concept is further substantiated by specific defects upon Brg1 deletion in glial cell differentiation (Limpert et al., 2013; Weider et al., 2012). Interestingly, Brg1 plays an important role in oligodendrocyte differentiation in early postnatal development and interacts with the transcription factor Olig2 in this context (Yu et al., 2013), but its virtual absence in Olig2+ cells in the adult forebrain suggests an intriguing difference between the molecular mechanisms acting at these different stages.

Indeed, the specific function of Pax6-Brg1/BAF complex in neurogenesis also appears to be specific for the adult SEZ lineage as the phenotypes caused by Brg1 or Pax6 deletion in the developing telencephalon are rather different from each other (Gotz et al., 1998; Haubst et al., 2004; Heins et al., 2002; Lessard et al., 2007; Matsumoto et al., 2006). Despite the co-expression of Pax6 and Brg1 and neurogenic function of Pax6 in this region (Gotz et al., 1998; Haubst et al., 2004; Heins et al., 2002), deletion of Brg1 at early developmental stages does not affect neurogenic fate acquisition, but rather affects stem and progenitor cell proliferation (Matsumoto et al., 2006). This raises the intriguing suggestion that this interaction may have specific functions in adult neurogenesis, which are not required in the developing brain. We therefore propose that mechanisms for neuronal fate stabilization are particularly relevant in niches where gliogenesis is more prevalent. This concept is further substantiated by the relevance of these factors in reprogramming glial cells into neurons.

Maintenance of neurogenic fate in the adult brain by a cross-regulatory neurogenic network

One of the major differences between the developing and adult brain is the prevalent neurogenesis in the former while gliogenesis largely dominates in the latter. Indeed, transplantation of many neural stem cells into the adult brain parenchyma results in their conversion to gliogenesis, while they readily generate neurons in the developing brain (for review see (Ninkovic and Gotz, 2013)). Thus, neuroblasts face particular challenges in the adult brain, not to convert to gliogenesis. Moreover, while the stem cells in the developing telencephalon, the radial glial cells, express high protein levels of neurogenic factors such as Pax6, adult NSCs express these only at lower mRNA levels and need to reach high protein levels when progressing further along the lineage (Beckervordersandforth et al., 2010; Feng et al., 2013). Indeed, our analysis shows that these two challenges can be overcome by Pax6 interacting with a Brg1/BAF chromatin remodeling complex and activating a cross-regulatory neurogenic transcriptional network required for stabilizing neurogenic fate in the adult brain. First, we have shown that lack of either Pax6 or Brg1 results in conversion of SEZ-derived cells towards different glial cells, depending on their local environment as described above. But even after isolation in vitro, SEZ cells lacking Brg1 or Pax6 can no longer complete a neurogenic fate suggesting that their intrinsic neurogenic factors inherited from aNSCs are not sufficient to further progress along the neurogenic lineage outside this niche. This demonstrates the critical role of Pax6-Brg1/BAF complex activity in increasing the levels of neurogenic fate determinants (as also seen in their down-regulation in genome-wide expression analysis in these mutants). Notably, this requires a catalytically active form of Brg1 demonstrating that chromatin remodeling is indeed critical for this step.

Here we discovered the molecular basis of Pax6/Brg1 complex mediated neurogenic fate maintenance by enhancing the expression of neurogenic transcription factors forming cross-regulatory network. These transcription factors cross-regulate each other and regulate neuronal effector genes. Interestingly, loss of Pax6 or Brg1 resulted in reduced expression of only few transcriptional regulators consisting of the epigenetic regulators Chd7, Ezh2 and the transcription factors, Sox11, Sox4, Nfib and Pou3f4. While the members of the polycomb repressor complex may play key roles in repressing alternative fates, such as glial lineages (Hirabayashi et al., 2009; Pereira et al., 2010), we demonstrated a positive role for the transcription factors Sox11/4, Nfib and Pou3f in executing and stabilizing neuronal fate.

Importantly, each member of this cross-regulatory network consisting of SoxC, Pou3f and Nfi factors is critical for neurogenesis. Previous work has already shown that SoxC factors are necessary for neurogenesis in various regions and lineages (Bergsland et al., 2006; Mu et al., 2012), and their regulation depends also on a chromatin remodeler interacting with Sox2 and Chd7 (Feng et al., 2013). Interestingly, Chd7 is also down-regulated upon Pax6 or Brg1 deletion, further supporting the concept that Pax6/BAF complex acts as an up-stream regulator of these. In addition we demonstrate here that also lowering the levels of Pou3f or Nfib interfere with neurogenesis, demonstrating the key role of each member of the cross-regulatory network to achieve full neurogenesis. This cross-regulatory network then activates genes mediating neuronal differentiation, as a large proportion of neuronal differentiation genes possess a motif for either all 3 or at least 2 of these factors, such as DCX, Tubb2b, CD24 and Prokr2 (Prosser et al., 2007).

Thus, our data suggest the following sequence of events (Suppl. Fig. 6C). Initiating neurogenic fate is mediated by factors such as Pax6 with a pioneering function that allows for altering the chromatin state via recruitment of a chromatin remodeling complex, such as the BAF complex. This Pax6-BAF complex then activates a cross-regulatory transcriptional effector network sufficient to maintain the high expression of genes involved in neuronal differentiation and thereby executing the lineage decision. At this later stage, lineage commitment can occur independent of Brg1 as demonstrated by expressing SoxC and Pou3f in cells lacking Brg1. Notably, genes controlled by the cross-regulatory network have a Pax6 binding site in addition to the Sox-Nfi-Pou3f regulatory motif compatible with the idea that Pax6-Brg1 complex might be important for making these loci accessible for high level expression that can then, once Nfi, Pou3f and SoxC are expressed at sufficiently high levels, be exerted independent of the initial role of Pax6 and Brg1 (Suppl. Fig. 6C). Indeed, the deletion of Brg1 in more committed TAPs and neuroblasts resulted in the switch to the glia cells, adopting either NG2 glia or ependymal glial fates. Consistent with this model, reduction of some members of this cross-regulatory network results in defects of adult SEZ neurogenesis as detailed above. At yet later stages, as the final layer in neuronal fate determination terminal selector genes then mediate specific neuronal subtype and neurotransmitter identity (Hobert, 2011; Kratsios et al., 2011). Taken together, this work has unravelled a highly specific interaction of Pax6 with a chromatin remodeling complex, consistent with the role of pioneer transcription factors (Zaret and Carroll, 2011) and further elucidated the molecular logic of neurogenesis and neuronal fate stabilization by igniting a cross-regulatory effector network.

Maintenance of neurogenic fate in the adult brain by a cross-regulatory neurogenic network

One of the major differences between the developing and adult brain is the prevalent neurogenesis in the former while gliogenesis largely dominates in the latter. Indeed, transplantation of many neural stem cells into the adult brain parenchyma results in their conversion to gliogenesis, while they readily generate neurons in the developing brain (for review see (Ninkovic and Gotz, 2013)). Thus, neuroblasts face particular challenges in the adult brain, not to convert to gliogenesis. Moreover, while the stem cells in the developing telencephalon, the radial glial cells, express high protein levels of neurogenic factors such as Pax6, adult NSCs express these only at lower mRNA levels and need to reach high protein levels when progressing further along the lineage (Beckervordersandforth et al., 2010; Feng et al., 2013). Indeed, our analysis shows that these two challenges can be overcome by Pax6 interacting with a Brg1/BAF chromatin remodeling complex and activating a cross-regulatory neurogenic transcriptional network required for stabilizing neurogenic fate in the adult brain. First, we have shown that lack of either Pax6 or Brg1 results in conversion of SEZ-derived cells towards different glial cells, depending on their local environment as described above. But even after isolation in vitro, SEZ cells lacking Brg1 or Pax6 can no longer complete a neurogenic fate suggesting that their intrinsic neurogenic factors inherited from aNSCs are not sufficient to further progress along the neurogenic lineage outside this niche. This demonstrates the critical role of Pax6-Brg1/BAF complex activity in increasing the levels of neurogenic fate determinants (as also seen in their down-regulation in genome-wide expression analysis in these mutants). Notably, this requires a catalytically active form of Brg1 demonstrating that chromatin remodeling is indeed critical for this step.

Here we discovered the molecular basis of Pax6/Brg1 complex mediated neurogenic fate maintenance by enhancing the expression of neurogenic transcription factors forming cross-regulatory network. These transcription factors cross-regulate each other and regulate neuronal effector genes. Interestingly, loss of Pax6 or Brg1 resulted in reduced expression of only few transcriptional regulators consisting of the epigenetic regulators Chd7, Ezh2 and the transcription factors, Sox11, Sox4, Nfib and Pou3f4. While the members of the polycomb repressor complex may play key roles in repressing alternative fates, such as glial lineages (Hirabayashi et al., 2009; Pereira et al., 2010), we demonstrated a positive role for the transcription factors Sox11/4, Nfib and Pou3f in executing and stabilizing neuronal fate.

Importantly, each member of this cross-regulatory network consisting of SoxC, Pou3f and Nfi factors is critical for neurogenesis. Previous work has already shown that SoxC factors are necessary for neurogenesis in various regions and lineages (Bergsland et al., 2006; Mu et al., 2012), and their regulation depends also on a chromatin remodeler interacting with Sox2 and Chd7 (Feng et al., 2013). Interestingly, Chd7 is also down-regulated upon Pax6 or Brg1 deletion, further supporting the concept that Pax6/BAF complex acts as an up-stream regulator of these. In addition we demonstrate here that also lowering the levels of Pou3f or Nfib interfere with neurogenesis, demonstrating the key role of each member of the cross-regulatory network to achieve full neurogenesis. This cross-regulatory network then activates genes mediating neuronal differentiation, as a large proportion of neuronal differentiation genes possess a motif for either all 3 or at least 2 of these factors, such as DCX, Tubb2b, CD24 and Prokr2 (Prosser et al., 2007).

Thus, our data suggest the following sequence of events (Suppl. Fig. 6C). Initiating neurogenic fate is mediated by factors such as Pax6 with a pioneering function that allows for altering the chromatin state via recruitment of a chromatin remodeling complex, such as the BAF complex. This Pax6-BAF complex then activates a cross-regulatory transcriptional effector network sufficient to maintain the high expression of genes involved in neuronal differentiation and thereby executing the lineage decision. At this later stage, lineage commitment can occur independent of Brg1 as demonstrated by expressing SoxC and Pou3f in cells lacking Brg1. Notably, genes controlled by the cross-regulatory network have a Pax6 binding site in addition to the Sox-Nfi-Pou3f regulatory motif compatible with the idea that Pax6-Brg1 complex might be important for making these loci accessible for high level expression that can then, once Nfi, Pou3f and SoxC are expressed at sufficiently high levels, be exerted independent of the initial role of Pax6 and Brg1 (Suppl. Fig. 6C). Indeed, the deletion of Brg1 in more committed TAPs and neuroblasts resulted in the switch to the glia cells, adopting either NG2 glia or ependymal glial fates. Consistent with this model, reduction of some members of this cross-regulatory network results in defects of adult SEZ neurogenesis as detailed above. At yet later stages, as the final layer in neuronal fate determination terminal selector genes then mediate specific neuronal subtype and neurotransmitter identity (Hobert, 2011; Kratsios et al., 2011). Taken together, this work has unravelled a highly specific interaction of Pax6 with a chromatin remodeling complex, consistent with the role of pioneer transcription factors (Zaret and Carroll, 2011) and further elucidated the molecular logic of neurogenesis and neuronal fate stabilization by igniting a cross-regulatory effector network.

Experimental Procedures

Detailed experimental procedures are available in the supplementary material on line.

Supplementary Material

Suppl. Figure1. Expression of Brg1 in the adult brain

A) Micrographs depicting the expression of Pax6 in undifferentiated NS5 neural stem cells. (B) Table depicting the Mascot score and number of peptides for BAF complex proteins purified by Pax6-immunoprecipitation (Pax6-IP or FLAG-Pax6-IP) of the crude protein extract from NS5 cells. Mascot score and the number of peptides of the corresponding control purification are between brackets. When no brackets are indicated, the protein was not observed in the control. (C) Western (C-E) Micrographs depicting the expression of Brg1, an ATP-ase unit of SWI/SNF complex, in the adult SEZ astrocytes marked by the expression of GFAP (C, D) and neuroblasts marked by the expression of DCX (E). Note the absence of Brg1 signal in the white matter (WM, C), an area reach in oligodendrocytes and their precursors. (F-E) Micrographs depicting the expression of BAF subunits in the SEZ containing fast dividing progenitors labeled with short pulse BrdU. Scale bars: 100 μm in A, C; 50 μm in D-I. Abbreviations: ctx-cerebral cortex; WM-white matter; RMS-rostral migratory stream; SEZ-subependymal zone; St-striatum; lv-lateral ventricle.

Suppl. Figure2. Loss of Brg1 function in neuronal progenitors renders them into glial cells.

(A, B) Representative micrographs of the SEZ in Brg1 cKO animals (B) and control animals (A) 9 days after tamoxifen(TM)-induced recombination illustrating the loss of Brg1 protein. (C) Histogram depicting the proportion of Brg1+ cells amongst the recombined. GFP+ cells as a function of time after the TM induction. (D) Histograms depicting the proportion of DCX+ neuroblasts amongst all recombined cells in the SEZ (left) and the OB (right) 9 dpt in brg1 cKO animals and their siblings. (E) Histogram showing the number of Caspase 3-positive cells in Brg1 cKO and control animals 28 days after TM induction. (F) Histogram depicting the distribution of the recombined cells 28 days after recombination in brg1 cKO and control animals. (G. H) (H) Micrograph illustrating the distribution of recombined cells in the SEZ and the RMS in Brg1cKO (H) and control (G) animals 28 days after recombination. Boxed area is magnified in the upper corner. (I)) Pies illustrating the identity of recombined cells in the OB of Brg1 cKO and control animals 60 days after tamoxifen-induced recombination. (J-L) Micrographs depicting GFAP+ (J, K) and marker negative glia cells (L) in the OB of Brg1 cKO (K, l) and control (J) animals 1 month after TM induction. (M) Histograms depicting number of neuroblasts in different OB layers 28 days after recombination. Scale bars: 100 μm in A and B; 20 μm in inlets in A and B, J-L′ and 200 μm in G, H. Abbreviations: RMS-rostral migratory stream; SEZ-subependymal zone; St-striatum; lv-lateral ventricle. Data in D, G and I are shown as mean ± SEM, n(animals analyzed)≥3 and in C and J as mean n(animals analyzed)≥3. **-p≤0.01; ***-p≤0.005.

Suppl. Figure3. Brg1-deficient neuronal progenitors generate both NG2 glia and ependymal cells.

(A, A′) Micrographs depicting the immunoreactivity of Brg1-deficient (GFP-positive), SEZ-derived (RFP-positive) cells for the early neuronal marker DCX 7 in the core of the OB days after viral labeling. A′ is magnification of area boxed in A. (B) Histogram showing the proportion of DCX+ and olig2+ cells in the OB generated by Brg1-deficient and SEZ progenitors with intact Brg1 function. (C) Micrographs depicting the immunoreactivity of Nestin-CreERT2-positive adult neural stem cell progeny for DCX in the Brg1 cKO (lower panel) and control (upper panel) animals 30 days after TM induction. (D) Histogram depicting the composition of the progeny generated by the Brg1-deficient and WT Nestin-CreERT2 positive adult neural stem cell 30 days after recombination. (E) Histogram showing the proportion of DCX-positive recombined cells in the SEZ of Brg1 cKO (generated using Glast) and control animals 28 days after TM administration. (F) Histogram showing the density of BrdU-positive cells in the SEZ of Brg1 cKO (generated using Glast) and control animals 28 days after TM administration. (G) Pies depicting cellular composition in the SEZ in Brg1 cKO and the control animals 7 days after recombination. (H) Micrograph depicting immunoreactivity of ependymal cells derived from Brg1-deficent progenitors for CD24 and S100β after 7 days in vitro. (I, J) Micrographs depicting the immunoreactivity of neural cells progeny labeled with split-Cre technology in WT (I) and Brg1 cKO (J) animals for neuronal marker DCX and glia marker GFAP 60 days after labeling. (K) Historgram depicting the composition of the progeny generated by the Brg1-deficient and WT neural stem cell 60 days after Split-Cre mediated recombination. Scale bars: 100 μm in A; 20 μm in A′, H, I and J; 200 μm in C. Data in B, D, E, F and J are shown as mean ± SEM and n(animals analyzed)≥. **-p≤0.01; ***-p≤0.005.

Suppl. Figure4. Brg1 regulates a set of neuronal genes in the adult neurogenic niches.

Heat maps demonstrating statistical significantly (p<0.05) deregulated genes in the SEZ or the OB ten days after Brg1 deletion (green indicates low and red high expression levels; scale bar in log2). (B-C) Histograms illustrating GO term enrichment analyses of deregulated genes after loss of Brg1 function in the OB (B) and SEZ (C) Shown are significantly (p<0.05) enriched terms and terms highlighted in red were observed in the analysis of both tissues. (D-F) Venn diagrams depicting the overlap between gene sets deregulated in Brg1 cKO and gene sets expressed in the purified populations enriched for adult neural stem cells and their progeny.

Suppl. Figure5. Pax6 with Brg1 co-regulate the set of genes necessary for both endogenous and induced neurogenesis.

(A, B) Representative micrographs depicting the immunoreactivity for Pax6 in the SEZ of Pax6 cKO (B) and control (A) animals 28 days after tamoxifen-induced recombination. (C) Histogram depicting number of Pax6 positive cells following recombination in Pax6 cKO and control animals. (D) Histogram showing the proportion of DCX-positive cells amongst all recombined cells in the SEZ of Pax6 cKO and control animals 60 days after TM administration. (E-F) Micrographs depicting neurogenesis in the DG of Pax6 cKO animals. (G-J) Micrographs depicting ISH signal for Pou3f4 (H), Nfib (I). Sox11 (J) and Sox4 (J) in the adult brain. Images from the Allen Brain Atlas (http://mouse.brain-map.org/). (K-L) Micrographs depicting immunoreactivity of recombined cells for Nfib in the SEZ (K) and RMS (L) in the Pax6 cKO 60 days after TM administration. (M) Histogram showing the proportion of Nfib-positive cells amongst Pax6-deficient, recombined neuroblasts and control, non-recombined neuroblasts. Data in C, D and M are shown as mean ± SEM and n(animals analyzed)≥4. **-p≤0.01. Scale bars: 100 μm in A, B, E, F and 20 μm in K and L.

Suppl. Figure6. Regulation of neurogenic genes in the SEZ. (A) Scheme depicting the regulation of the representative set of genes by the cross-regulatory network activated by the Pax6-BAF complex. (B) Micrograph depicting the morphology of superficial GCL generated from the MLV-retrovirus transduced SEZ progenitors 1 month after the stereotactic injection. Scale bar 100 μm. (C) Model depicting the changes in chromatin structures of neurogenic genes during the differentiation.

Suppl. Figure7. Nfib is expressed in neurosphere derived astrocytes. (A-B) Micrographs depicting immunoreactivity for Nfib in neurosphere derived cells. B is maginification of area boxed in A. (C, D) Histograms depicting the efficiency of Pou3f4 (C) and Nfib (D) knock-down using esiRNAs in the neurosphere cells 36 h after transduction. (E) Dot-plot representing populations sorted from the adult SEZ using FACS for the cell-type specific surface antigens (also see (Fischer et al., 2011)). (F) Histogram showing the expression of ATP-ase units of SWI/SNF complex in purified cells from the neuronal and oligodendrogenic lineage. Scale bars: 100 μm in A and 20 μm in B.

Suppl. Figure1. Expression of Brg1 in the adult brain

A) Micrographs depicting the expression of Pax6 in undifferentiated NS5 neural stem cells. (B) Table depicting the Mascot score and number of peptides for BAF complex proteins purified by Pax6-immunoprecipitation (Pax6-IP or FLAG-Pax6-IP) of the crude protein extract from NS5 cells. Mascot score and the number of peptides of the corresponding control purification are between brackets. When no brackets are indicated, the protein was not observed in the control. (C) Western (C-E) Micrographs depicting the expression of Brg1, an ATP-ase unit of SWI/SNF complex, in the adult SEZ astrocytes marked by the expression of GFAP (C, D) and neuroblasts marked by the expression of DCX (E). Note the absence of Brg1 signal in the white matter (WM, C), an area reach in oligodendrocytes and their precursors. (F-E) Micrographs depicting the expression of BAF subunits in the SEZ containing fast dividing progenitors labeled with short pulse BrdU. Scale bars: 100 μm in A, C; 50 μm in D-I. Abbreviations: ctx-cerebral cortex; WM-white matter; RMS-rostral migratory stream; SEZ-subependymal zone; St-striatum; lv-lateral ventricle.

Suppl. Figure2. Loss of Brg1 function in neuronal progenitors renders them into glial cells.

(A, B) Representative micrographs of the SEZ in Brg1 cKO animals (B) and control animals (A) 9 days after tamoxifen(TM)-induced recombination illustrating the loss of Brg1 protein. (C) Histogram depicting the proportion of Brg1+ cells amongst the recombined. GFP+ cells as a function of time after the TM induction. (D) Histograms depicting the proportion of DCX+ neuroblasts amongst all recombined cells in the SEZ (left) and the OB (right) 9 dpt in brg1 cKO animals and their siblings. (E) Histogram showing the number of Caspase 3-positive cells in Brg1 cKO and control animals 28 days after TM induction. (F) Histogram depicting the distribution of the recombined cells 28 days after recombination in brg1 cKO and control animals. (G. H) (H) Micrograph illustrating the distribution of recombined cells in the SEZ and the RMS in Brg1cKO (H) and control (G) animals 28 days after recombination. Boxed area is magnified in the upper corner. (I)) Pies illustrating the identity of recombined cells in the OB of Brg1 cKO and control animals 60 days after tamoxifen-induced recombination. (J-L) Micrographs depicting GFAP+ (J, K) and marker negative glia cells (L) in the OB of Brg1 cKO (K, l) and control (J) animals 1 month after TM induction. (M) Histograms depicting number of neuroblasts in different OB layers 28 days after recombination. Scale bars: 100 μm in A and B; 20 μm in inlets in A and B, J-L′ and 200 μm in G, H. Abbreviations: RMS-rostral migratory stream; SEZ-subependymal zone; St-striatum; lv-lateral ventricle. Data in D, G and I are shown as mean ± SEM, n(animals analyzed)≥3 and in C and J as mean n(animals analyzed)≥3. **-p≤0.01; ***-p≤0.005.

Suppl. Figure3. Brg1-deficient neuronal progenitors generate both NG2 glia and ependymal cells.

(A, A′) Micrographs depicting the immunoreactivity of Brg1-deficient (GFP-positive), SEZ-derived (RFP-positive) cells for the early neuronal marker DCX 7 in the core of the OB days after viral labeling. A′ is magnification of area boxed in A. (B) Histogram showing the proportion of DCX+ and olig2+ cells in the OB generated by Brg1-deficient and SEZ progenitors with intact Brg1 function. (C) Micrographs depicting the immunoreactivity of Nestin-CreERT2-positive adult neural stem cell progeny for DCX in the Brg1 cKO (lower panel) and control (upper panel) animals 30 days after TM induction. (D) Histogram depicting the composition of the progeny generated by the Brg1-deficient and WT Nestin-CreERT2 positive adult neural stem cell 30 days after recombination. (E) Histogram showing the proportion of DCX-positive recombined cells in the SEZ of Brg1 cKO (generated using Glast) and control animals 28 days after TM administration. (F) Histogram showing the density of BrdU-positive cells in the SEZ of Brg1 cKO (generated using Glast) and control animals 28 days after TM administration. (G) Pies depicting cellular composition in the SEZ in Brg1 cKO and the control animals 7 days after recombination. (H) Micrograph depicting immunoreactivity of ependymal cells derived from Brg1-deficent progenitors for CD24 and S100β after 7 days in vitro. (I, J) Micrographs depicting the immunoreactivity of neural cells progeny labeled with split-Cre technology in WT (I) and Brg1 cKO (J) animals for neuronal marker DCX and glia marker GFAP 60 days after labeling. (K) Historgram depicting the composition of the progeny generated by the Brg1-deficient and WT neural stem cell 60 days after Split-Cre mediated recombination. Scale bars: 100 μm in A; 20 μm in A′, H, I and J; 200 μm in C. Data in B, D, E, F and J are shown as mean ± SEM and n(animals analyzed)≥. **-p≤0.01; ***-p≤0.005.

Suppl. Figure4. Brg1 regulates a set of neuronal genes in the adult neurogenic niches.

Heat maps demonstrating statistical significantly (p<0.05) deregulated genes in the SEZ or the OB ten days after Brg1 deletion (green indicates low and red high expression levels; scale bar in log2). (B-C) Histograms illustrating GO term enrichment analyses of deregulated genes after loss of Brg1 function in the OB (B) and SEZ (C) Shown are significantly (p<0.05) enriched terms and terms highlighted in red were observed in the analysis of both tissues. (D-F) Venn diagrams depicting the overlap between gene sets deregulated in Brg1 cKO and gene sets expressed in the purified populations enriched for adult neural stem cells and their progeny.

Suppl. Figure5. Pax6 with Brg1 co-regulate the set of genes necessary for both endogenous and induced neurogenesis.

(A, B) Representative micrographs depicting the immunoreactivity for Pax6 in the SEZ of Pax6 cKO (B) and control (A) animals 28 days after tamoxifen-induced recombination. (C) Histogram depicting number of Pax6 positive cells following recombination in Pax6 cKO and control animals. (D) Histogram showing the proportion of DCX-positive cells amongst all recombined cells in the SEZ of Pax6 cKO and control animals 60 days after TM administration. (E-F) Micrographs depicting neurogenesis in the DG of Pax6 cKO animals. (G-J) Micrographs depicting ISH signal for Pou3f4 (H), Nfib (I). Sox11 (J) and Sox4 (J) in the adult brain. Images from the Allen Brain Atlas (http://mouse.brain-map.org/). (K-L) Micrographs depicting immunoreactivity of recombined cells for Nfib in the SEZ (K) and RMS (L) in the Pax6 cKO 60 days after TM administration. (M) Histogram showing the proportion of Nfib-positive cells amongst Pax6-deficient, recombined neuroblasts and control, non-recombined neuroblasts. Data in C, D and M are shown as mean ± SEM and n(animals analyzed)≥4. **-p≤0.01. Scale bars: 100 μm in A, B, E, F and 20 μm in K and L.

Suppl. Figure6. Regulation of neurogenic genes in the SEZ. (A) Scheme depicting the regulation of the representative set of genes by the cross-regulatory network activated by the Pax6-BAF complex. (B) Micrograph depicting the morphology of superficial GCL generated from the MLV-retrovirus transduced SEZ progenitors 1 month after the stereotactic injection. Scale bar 100 μm. (C) Model depicting the changes in chromatin structures of neurogenic genes during the differentiation.

Suppl. Figure7. Nfib is expressed in neurosphere derived astrocytes. (A-B) Micrographs depicting immunoreactivity for Nfib in neurosphere derived cells. B is maginification of area boxed in A. (C, D) Histograms depicting the efficiency of Pou3f4 (C) and Nfib (D) knock-down using esiRNAs in the neurosphere cells 36 h after transduction. (E) Dot-plot representing populations sorted from the adult SEZ using FACS for the cell-type specific surface antigens (also see (Fischer et al., 2011)). (F) Histogram showing the expression of ATP-ase units of SWI/SNF complex in purified cells from the neuronal and oligodendrogenic lineage. Scale bars: 100 μm in A and 20 μm in B.

Click here to view.(491K, pdf)

Acknowledgments

We want to thank Peter Becker, Ales Cvekl and Maren Eckey for help with establishment of ChIP experiments and Karel Bezstarosti for help with the mass spectrometry analysis. We are also thankful to Timucin Özturk, Emily Violette-Baumgardt and Angelika Weisser for excellent technical help, Alexandra Lepier for the viral vector production, Jelena Ninkovic for the help with statistical analysis and Leda Dimou for critical reading of the manuscript. We also gratefully acknowledge funding to JN from the German Research foundation (DFG) by the SFB 870 and SPP “Integrative Analysis of Olfaction” and to MG the German Research foundation (DFG) by the SFB 870, the Leibniz Prize and the Munich Cluster for Systems Neurology (EXC 1010 SyNergy).

Helmholtz Centre Munich German Research Centre for Environmental Health (GmbH), Institute for Stem Cell Research, Ingolstädter Landstr.1, 85764 Neuherberg/Munich, Germany
Physiological Genomics, Medical Faculty, University of Munich, Schillerstr. 46, 80633 Munich, Germany
Dept. of Cell Biology, Erasmus MC, Dr. Molewaterplein 50. 3015 GE Rotterdam, The Netherlands
Dept. anatomy and Cell Biology, Heidelberg University, Im Neuenheimer Feld 307 and Interdisciplanry Center Neurosciences (IZN), Heidelberg, Germany
Technical University Munich, Center of Life and Food Sciences Weihenstephan, Freising, Germany
Helmholtz Centre Munich German Research Center for Environmental Health (GmbH), Institute of Developmental Genetics, Neuherberg, Germany, Institute of Biochemistry, Emil Fischer Center, University Erlangen-Nürnberg, Erlangen, Germany
Department of Psychiatry, UT Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX, 75390-9070 USA
Howard Hughes Medical Institute, Beckman Center B211, 279 Campus Drive, Stanford University School of Medicine, Stanford, CA 94305-5323, USA
Department of Physiology and Developmental Biology, UT Southwestern Medical Center, 5323 Harry Hines Blvd., Dallas, TX, USA
Department of Cell Biology (NC10) Cleveland Clinic Lerner Research Institute, 9500 Euclid Avenue, Cleveland, Ohio 44195,USA
Proteomics Center, Erasmus MC, Rotterdam, Dr. Molewaterplein 50. 3015 GE Rotterdam, The Netherlands
Institut de Génétique et de Biologie Moléculaire et Cellulaire (IGBMC), INSERM U964 / CNRS UMR 7104 / Université de Strasbourg, France
Helmholtz Centre Munich German Research Center for Environmental Health (GmbH), Institute of Experimental Genetics, Neuherberg, Germany
Corresponding author: ed.nehcneum-ztlohmleh@zteog.aneladgam

Abstract

The molecular mechanisms of neurogenic fate determination are of particular importance in light of the need to regenerate neurons. Here we define the mechanisms of installing neurogenic fate by the transcription factor Pax6 acting together with the Brg1-containing BAF chromatin remodeling complex. We show that Pax6 physically interacts with Brg1-containing BAF complex and genetic deletion of either Pax6 or Brg1, in the neural stem cells in the adult mouse subependymal zone results in a strikingly similar fate conversion from neuronal progenitors to glia. The Pax6-BAF complex drives neurogenesis by directly activating transcription factors Sox11, Nfib and Pou3f4, which form a cross-regulatory network that maintains neurogenic fate downstream of the Pax6-BAF complex in neuroblasts. Our work identifies a novel concept of stratification in neural fate commitment with a strikingly specific role of the Pax6-BAF complex in initiating a cross-regulatory network essential for maintenance of the neurogenic lineage in the adult brain.

Keywords: Chromatin, Fate determinants, Neurogenesis, Fate conversion
Abstract
Milestones
Collaboration tool especially designed for Life Science professionals.Drag-and-drop any entity to your messages.