|Home | About | Journals | Submit | Contact Us | Français|
FOXO transcription factors are central regulators of longevity from worms to humans. FOXO3 – the FOXO isoform associated with exceptional human longevity – preserves adult neural stem cell pools. Here we identify FOXO3 direct targets genome-wide in primary cultures of adult neural progenitor cells (NPCs). Interestingly, FOXO3-bound sites are enriched for motifs for bHLH transcription factors and FOXO3 shares common targets with the pro-neuronal bHLH transcription factor ASCL1/MASH1 in NPCs. Analysis of the chromatin landscape reveals that FOXO3 and ASCL1 are particularly enriched at the enhancers of genes involved in neurogenic pathways. Intriguingly, FOXO3 inhibits ASCL1-dependent neurogenesis in NPCs and direct neuronal conversion in fibroblasts. FOXO3 also restrains neurogenesis in vivo. Our study identifies a genome-wide interaction between the pro-longevity transcription factor FOXO3 and the cell fate determinant ASCL1, and raises the possibility that FOXO3's ability to restrain ASCL1-dependent neurogenesis may help preserve the neural stem cell pool.
FOXO transcription factors promote longevity downstream of the insulin/IGF-1 signaling pathway from worms to mammals (Kenyon, 2010). When the insulin/IGF1 signaling pathway is inactive, FOXO transcription factors translocate to the nucleus, where they act mostly as transcriptional activators (Calnan and Brunet, 2008). Overexpression of FOXO transcription factors extends lifespan in invertebrates (Giannakou et al., 2004; Henderson and Johnson, 2001; Hwangbo et al., 2004). In humans, FOXO3, one of the four FOXO family members, has been associated with exceptional lifespan in eight cohorts of centenarians (Kenyon, 2010). Interestingly, FOXO3 has recently been found to regulate adult neural stem cells (NSCs) and hematopoietic stem cells (HSCs). FOXO3 is localized in the nucleus of NSCs and HSCs in vivo (Miyamoto et al., 2007; Paik et al., 2009; Renault et al., 2009), suggesting that FOXO3 is active in these cells. Deletion of the FOXO3 gene in mice leads to the premature depletion of NSCs and HSCs in adult animals (Miyamoto et al., 2007; Paik et al., 2009; Renault et al., 2009; Tothova et al., 2007). Combined deletion of FOXO1, FOXO3 and FOXO4 leads to a more pronounced depletion of NSCs and HSCs in vivo (Paik et al., 2009; Tothova et al., 2007), indicating that FOXO transcription factors have overlapping roles in maintaining adult stem cell pools in vivo. The ability of FOXO transcription factors to regulate adult stem cell homeostasis is likely to be crucial for tissue regeneration, and potentially organismal longevity.
While the importance of FOXO3 in stem cell maintenance is well known, the role of FOXO3 in the differentiation of stem and progenitor cells has been less explored. In vivo, the loss of FOXO transcription factors causes a premature burst in neurogenesis during development (Paik et al., 2009), raising the possibility that FOXO factors inhibit neuronal differentiation. However, the importance of FOXO3 in the neuronal differentiation of neural stem and progenitor cells, and the specific FOXO3 targets that may affect this neuronal differentiation have not been established. Given the importance of neural stem and progenitor cells for the generation of new neurons, and their potential use for regenerative medicine, understanding the mechanisms by which FOXO3 acts in cells from the neural lineage should provide important insight into the regulation of these cells and their ability to produce new neurons.
FOXO3 is well known to regulate gene expression programs that coordinate stress resistance, cell cycle progression, autophagy, and apoptosis (Salih and Brunet, 2008), but the ensemble of direct target genes of FOXO3 are still largely unknown. Although recent studies have characterized FOXO1 and FOXO3 targets in a few differentiated lineages (Lin et al., 2010; Litvak et al., 2012; Ouyang et al., 2012; Shin et al., 2012), the direct FOXO3 targets have never been characterized in a stem/progenitor context. Furthermore, the mode of action of FOXO transcription factors in specific lineages is largely unknown. Finally, whether FOXO3 shares targets with other transcription factors genome-wide has never been determined in stem/progenitor cells.
Here we use primary cultures of adult neural progenitor cells (NPCs) as a cellular system to understand FOXO3's mechanisms of action in the neural lineage. We find that FOXO3 can bind about 2000 direct targets genome-wide in these cells. FOXO3-bound genes extensively overlap with those bound by the pro-neuronal bHLH transcription factor ASCL1/MASH1 in cultured NPCs. Genome-wide analysis of the chromatin landscape in these cells reveals that FOXO3 and ASCL1 are enriched at enhancer regions. Interestingly, FOXO3 represses the expression of a subset of ASCL1 neurogenic targets and restrains neurogenesis in vitro and in vivo. FOXO3 also inhibits the ability of ASCL1 to directly reprogram fibroblasts into ‘induced neurons’ (iNs). These findings suggest that FOXO3 can act by repressing the transcriptional activity of a ‘master regulator’ of cell fate. Thus, FOXO3 may help preserve an intact pool of neural stem cells at least in part by negatively regulating neuronal differentiation.
To identify FOXO3 binding sites genome-wide in a stem/progenitor context, we used primary cultures of adult neural progenitor cells (NPCs). We chose these cells because they can differentiate into neurons, astrocytes, and oligodendrocytes (albeit with a bias towards astrocytes) (Pastrana et al., 2011), and they can be expanded in culture to provide cell numbers (10-100 million) that are compatible with chromatin immunoprecipitation followed by next generation sequencing (ChIP-seq). We performed FOXO3 ChIP-seq in these primary cultures of NPCs before and after the induction of FOXO3 nuclear localization (Figure 1A), as FOXO3 is mostly localized in the nucleus of NSCs in vivo (Figure S1A) (Paik et al., 2009; Renault et al., 2009). To induce FOXO3 nuclear localization, we transiently incubated NPCs in conditions of low growth factor signaling (Experimental Procedures). These conditions have been used previously for FOXO3 ChIP-qPCR in cultured NPCs (Renault et al., 2009) and other cellular systems (van der Vos et al., 2012). Using immunofluorescence, we verified that a tagged version of FOXO3 indeed accumulates in the nucleus of NPCs in low growth factor signaling conditions (Figure S2A-B). Following translocation to the nucleus, endogenous FOXO3 is indeed recruited to Cdkn1b (p27) – a well-known FOXO target gene – in cultured NPCs (Figure 1B). Under these experimental conditions, greater than 90% of the NPCs still express the neural progenitor markers NESTIN and SOX2 (Figure S1B-F) and retain the ability to differentiate into neurons (Figure S1G). Thus, primary cultures of adult NPCs in low growth factor signaling conditions provide a cellular system to identify FOXO3 direct targets in a stem/progenitor context.
To characterize the ensemble of FOXO3 direct targets in NPCs, we performed FOXO3 ChIP-seq. FOXO3 binding sites were identified using QuEST (Valouev et al. 2008) with a stringent false discovery rate threshold (FDR < 0.001) (Figure 1C). Prior to induction of nuclear translocation, FOXO3 was, as expected, bound to very few genomic regions (8) (Figure 1C). In contrast, following nuclear translocation, FOXO3 was bound to 4329 genomic regions (QuEST peaks), 2291 of which were within 5 kb of the transcriptional start sites of unique RefSeq genes (Figure 1C). An example of FOXO3 binding is shown in Figure 1D. The 10 FOXO3-bound genes with the highest QuEST binding scores are listed in Figure 1E (complete datasets are in Table S1 and Table S2). We verified that genes bound by FOXO3 in ChIP-seq experiments were also bound by ChIP-qPCR in wild type NPCs, but not in FOXO3−/− NPCs (Figure 1F). Thus, FOXO3 binds to approximately 2000 genes in cultured NPCs under low growth factor signaling conditions.
We next asked if FOXO3 also binds these target genes in vivo. We performed FOXO3 ChIP-qPCR on microdissected neurogenic niches (subventricular zones [SVZs]), which are enriched for neural stem and progenitor cells in the adult mouse. We found that endogenous FOXO3 was bound to several of its targets (Pot1a, Arrdc3, Akt1s1/Tbc1d17, Txnip, Hbp1, Vkorc1) in neurogenic niches (Figure 1G, Figure S2C, and Table S1). Of the genes tested (12), half validated in vivo and are likely to represent bona fide FOXO3 targets. A regression analysis showed that FOXO3 binding enrichments in vitro and in vivo were significantly correlated (Figure 1H, Pearson correlation, R = 0.94, p = 3.8 × 10-06), indicating that genes that have the highest enrichment for FOXO3 in culture are more likely to be bound by FOXO3 in vivo. Importantly, the FOXO3-bound genes that validated in vivo were indeed expressed in activated neural stem cells (NSCs) and neural progenitor cells (NPCs) freshly purified from the adult brain using a fluorescence-activated cell sorting (FACS) strategy (Pastrana et al., 2009) (Figure 1I and Figure S2D-F).
FOXO3-bound genes in cultured NPCs significantly overlapped with genes that are down-regulated in FOXO3−/− NPCs (Figure 1J, Fisher's exact test, p = 1.18 × 10-17). This observation, coupled with the fact that FOXO3 is known to act as a transcriptional activator (Calnan and Brunet, 2008; Salih and Brunet, 2008), provides further support for the relevance of the FOXO3 ChIP-seq dataset. FOXO3 could also act as a transcriptional repressor at some specific targets, although this was not statistically significant (Figure S2G). Interestingly, FOXO3 binding sites in NPCs significantly overlapped with FOXO1 and FOXO3 binding sites in other cell types (B cells, T cells, macrophages) (Figure S3A-D, Table S3), indicating that there exists a ‘core set’ of FOXO targets as well as lineage/cell-specific FOXO targets (Figure S3D, Table S3). Collectively, these results are consistent with the notion that the FOXO3 target genes identified in cultured NPCs have physiological relevance.
We asked whether FOXO3 co-occurs with other transcription factors genome-wide. De novo motif analysis of FOXO3 ChIP-seq datasets using the MEME motif finder (Bailey and Elkan, 1994) revealed that 90% of FOXO3-bound sites contained the FOXO consensus motif TGTTTAC (E value = 1.5 × 10-3175) (Figure 2A) (Furuyama et al., 2000), confirming that FOXO3 binding sites in NPCs are bona fide recruitment sites for this transcription factor. Interestingly, analysis of the 200 bp regions surrounding the FOXO3 ChIP-seq peaks revealed two other motifs that co-occur with FOXO3 binding sites: a CAGCTG motif which co-occurs at more than 30% of the binding sites (E-value = 2.5 × 10-406) (Figure 2B) and a CAGGCTG motif which co-occurs at about 30% of the binding sites (E-value = 3.1 × 10-273) (data not shown). Analysis by STAMP (Mahony and Benos, 2007) revealed that the CAGCTG motif is a specific subclass of E-box consensus sites (Powell and Jarman, 2008). These E-boxes are known to be bound by bHLH transcription factors that have been called ‘master regulators’ of cell identity, including ASCL1/MASH1, NEUROD1, NGN1, and MYOD (Figure 2C) (Guillemot, 1999; Hu et al., 2004). Such transcription factors are sufficient to promote cell differentiation and even direct reprogramming of a differentiated cell type into another (Vierbuchen and Wernig, 2011).
We focused on the bHLH transcription factor ASCL1 because of its well-characterized function in neurogenesis (Guillemot et al., 1993; Parras et al., 2004; Vierbuchen et al., 2010) and its expression in adult NSCs/NPCs (Pastrana et al., 2009). In line with previous reports, we found that the ASCL1 protein is expressed in the nucleus of a subset of neural stem and progenitor cells in the adult brain (Figure S1A) and adult NPCs in culture (Figure 2D). Ascl1 mRNA is highly expressed in activated NSCs and NPCs freshly purified from the brain whereas other bHLH transcription factor mRNAs (Neurod1,Neurog1, and Myod1) are not expressed at detectable levels in activated NSCs and NPCs (Figure 2E). Conversely, other transcription factors whose mRNA is expressed in NPCs (Hes1, E2a/Tcf3, Olig2, and Sox2) (Figure 2E) have different binding motifs and/or are not known to be involved in neurogenesis (Figure 2C). These observations raise the possibility that ASCL1 may be the bHLH transcription factor that functionally interacts with FOXO3 in NPCs.
To determine if FOXO3 and ASCL1 bind similar genes genome-wide, we performed ASCL1 ChIP-seq in NPCs (Figure 3). ASCL1 ChIP-seq revealed ~18,000 ASCL1 binding sites in NPCs, corresponding to ~6,000 unique genes (Figure 3A, Tables S4 and S5). De novo MEME motif analysis identified the presence of the ASCL1 consensus motif CAGCTG at 91.8% of the ASCL1-bound regions (E-value = 2.4 × 10-17208) (Figure 3B). The top 10 ASCL1-bound genes are included in Figure 3C (complete list and uploadable UCSC Genome Browser tracks are in Tables S4 and S5). ChIP-qPCR experiments confirmed that ASCL1 is recruited to these genomic loci in NPCs (Figure 3D) and that ASCL1 is also bound to these sites in vivo in microdissected neurogenic niches (SVZs) (Figure 3E). ASCL1 binding in primary cultures of adult NPCs significantly overlaps with ASCL1 binding in NS5 cells (neural progenitor cells derived from embryonic stem cells) as well as in the ventral telencephalon in vivo (Castro et al., 2011) (Figure S3E, Table S3). These data confirm that datasets obtained in primary cultures of adult NPCs are relevant in vivo.
Importantly, comparison of the FOXO3 and ASCL1 ChIP-seq datasets revealed that 70.9% of FOXO3-bound genes were also bound by ASCL1 in NPCs (Figure 3F). In about 35% of cases of common targets, FOXO3 and ASCL1 binding sites were indeed within 200 bp of each other (Figure 3G, upper panel), consistent with the percentage of E-box motif occurrence close to the FOXO motif (31.8%). FOXO3 and ASCL1 binding sites could also be more distant than 200 bp from one another (Figure 3G, lower panel). To test if FOXO3 and ASCL1 co-occur at some of their shared targets in the same cell, we performed sequential ChIP experiments (ASCL1 ChIP-FOXO3 reChIP) in NPCs in low growth factor signaling conditions (Figure 3H). Several genomic loci that were bound by ASCL1 (Vac14/Mtss1l, Zbtb7c, Cd276, Cbara1, Tns3, Dgkg) were co-bound by FOXO3, although there were also some loci at which FOXO3 and ASCL1 co-binding could not be detected (Ttyh2, Dll1, and Hes6) (Figure 3H). Thus, FOXO3 and ASCL1 share targets genome-wide in primary cultures of adult NPCs, and can co-occur in the same cell at some genomic loci.
FOXO3 and ASCL1 binding occurs at promoter regions, but also in intergenic regions and in introns (Figure 4A, Figure S4), which are known to contain enhancers. These observations led us to ask if FOXO3 and ASCL1 bind to bona fide enhancers, and if regions bound by both FOXO3 and ASCL1 are different from regions bound by only one of these factors. Chromatin states of enhancers have been recently defined as enriched for H3K4me1 and depleted for H3K4me3 and H3K27me3 (Heintzman et al., 2009; Rada-Iglesias et al., 2011). To characterize enhancers in primary cultures of adult NPCs, we performed ChIP-seq for histone marks H3K4me1, H3K4me3, and H3K27me3 (Figure 4B). Both FOXO3 and ASCL1 binding sites were found in genomic regions that were enriched for H3K4me1 and depleted for H3K4me3 and H3K27me3 (Figure 4C), suggesting that FOXO3 and ASCL1 bind regions with several features of enhancers. Interestingly, FOXO3/ASCL1 co-bound regions were statistically enriched for the H3K4me1 mark over regions bound by either FOXO3 or ASCL1 alone (p = 7.37 × 10-35 and p = 2.08 × 10-27, respectively) (Figure 4D). These data suggest that FOXO3 and ASCL1 are enriched at enhancers, which tend to be more lineage-specific than promoters (Heintzman et al., 2009; Rada-Iglesias et al., 2011).
To determine the functional consequence of FOXO3 and ASCL1 binding on gene expression, we asked if genes bound by both FOXO3 and ASCL1 have unique characteristics compared to genes bound by only one of these transcription factors (Figure 5A, Figure S5). PANTHER analysis revealed that the genes bound by both FOXO3 and ASCL1 were more highly enriched for the Notch and Wnt signaling pathways than genes bound by FOXO3 or ASCL1 alone (Figure 5A; Table S6). The Notch and Wnt pathways are known to regulate the balance between NSCs and more committed progenitors (Aguirre et al., 2010; Chapouton et al., 2010; Imayoshi et al., 2010; Lie et al., 2005; Oishi et al., 2004). The Notch pathway has also been shown to be critical for ASCL1-dependent neurogenesis (Castro et al., 2006), and has been implicated downstream of FOXO1 in myoblasts (Kitamura et al., 2007). Together, these observations raise the possibility that in NPCs, FOXO3 regulates ASCL1 targets involved in Notch signaling.
We focused on Dll1 and Hes6 because these genes belong to the Notch pathway and are well known targets of ASCL1 in vitro and in vivo (Castro et al., 2006). FOXO3 and ASCL1 both bind Dll1 and Hes6 at sites that are enriched for the enhancer mark H3K4me1 in cultured NPCs (Figure 5B). Interestingly, expression of FOXO3 in NPCs inhibits the ability of ASCL1 to upregulate Dll1 and Hes6 in adult NPCs (Figure 5C). FOXO3 also inhibited ASCL1-dependent expression of Dll1 and Hes6 in the embryonic stem cell-derived neural progenitor NS5 cell line (Figure 5D). FOXO3 does not appear to restrain ASCL1-dependent transcription by blocking the recruitment of ASCL1 to these genes: low growth factor signaling conditions, which induce the accumulation of endogenous FOXO3 into the nucleus, did not significantly affect ASCL1 binding (Figure S6A) and, vice versa, ASCL1 binding was not drastically altered in NPCs from FOXO3−/− mice (Figure S6B). Together, these data show that FOXO3 inhibits ASCL1-dependent induction of the Notch pathway genes Dll1 and Hes6 in cultured NPCs, and are consistent with the possibility that FOXO3 acts by preventing ASCL1-dependent transcription rather than binding.
To determine if the inhibition of ASCL1-dependent expression of neurogenic genes by FOXO3 had physiological consequences, we first tested if FOXO3 impacts ASCL1's ability to promote neurogenesis in adult NPCs. Consistent with previous reports in other cellular systems (Farah et al., 2000; Vierbuchen et al., 2010), ASCL1 expression also potently induced neurogenesis in adult NPCs (Figure 6A). Importantly, FOXO3 significantly decreased the ability of ASCL1 to induce neurogenesis (Figure 6A). However, the impact of FOXO3 on neurogenesis is more pronounced than that on gene expression, suggesting that FOXO3 could also repress neurogenesis via additional targets or mechanisms.
Expression of ASCL1 together with other pro-neural transcription factors (MYT1L and/or BRN2) was recently shown to directly reprogram fibroblasts into induced neurons (iNs) (Vierbuchen et al., 2010, Pang et al., 2011). We assessed the effect of FOXO3 on the ability of ASCL1 to directly convert mouse embryonic fibroblasts (MEFs) into iNs (Figure 6B). As expected, ASCL1 and MYTlL efficiently induced the generation of TUJ1-positive iNs in MEFs (Figure 6B). Interestingly, ectopic expression of FOXO3 significantly decreased the ability of ASCL1/MYT1L to reprogram MEFs into TUJ1-positive iNs (Figure 6B). By contrast, expression of an inactive form of FOXO3 that lacks the DNA binding domain did not significantly affect ASCL1/MYT1L-dependent iN conversion (Figure 6B). These results indicate that FOXO3 binding antagonizes ASCL1's ability to promote the direct neuronal conversion of fibroblasts.
We asked if FOXO3 impacts neurogenesis in vivo. Co-staining of adult mouse brain sections reveals that FOXO3 and ASCL1 are co-localized in neurogenic niches (SVZs) in a subset of cells that are marked with SOX2, a marker of neural stem/progenitor cells (Figure 7A and Figure S1A). FOXO3 is localized in the nucleus in a subset of these cells, though not all (Figure 7A and Figure S1A), indicating that active FOXO3 and ASCL1 co-exist in some, but not all, neural stem and/or progenitor cells in the adult brain in vivo. We next assessed neurogenesis in adult FOXO3−/− and FOXO3+/+ mice by injecting 5-ethynyl-2'-deoxyuridine (EdU) and co-staining sections of the olfactory bulb – the region where new neurons generated in the SVZ migrate to – with EdU and DCX, a marker of newly formed neurons. FOXO3 deficiency led to a significant increase in the production of new neurons in the olfactory bulbs of adult mice (Figure 7B, Student's t test, p < 0.05). Hes6 expression was slightly increased in activated NSCs and NPCs that were freshly purified by FACS from FOXO3−/− mice compared to FOXO3+/+ mice (Figure 7C, one-sided Wilcoxon exact test, p < 0.05), although Dll1 expression was not significantly affected (Figure 7C). These results are consistent with the possibility that FOXO3 may antagonize the ability of ASCL1 to induce neurogenic genes in vivo, although other targets or mechanisms may mediate the effect of FOXO3 on neurogenesis.
Our study uses ultra-high throughput sequencing technologies to identify FOXO3 and ASCL1 targets in primary cultures of adult NPCs, and reveals that FOXO3 and ASCL1 overlap genome-wide and are particularly enriched at enhancers of genes involved the Wnt and Notch signaling pathways. Our study also provides evidence that FOXO3 inhibits the ability of ASCL1 to induce neurogenesis and direct neuronal conversion. These findings have implications in other systems because bHLH master regulators are pivotal for the determination of many cell lineages. The binding of two other Forkhead transcription factors, FOXO1 and FOXH1, has been found to overlap with that of the bHLH regulator E2A in B cells and human embryonic stem cells, respectively (Lin et al., 2010; Yoon et al., 2011), although the exact bHLH transcription factor that is involved is not known. Thus, binding in close proximity to bHLH transcription factors may be a general feature of Forkhead transcription factors. As several Forkhead transcription factors are known to act as ‘pioneer factors’ to open chromatin (Zaret and Carroll, 2011), these factors may help provide an open chromatin context for lineage-specific transcription factors to activate transcription of differentiation genes. Transcription factors of the bHLH family are particularly powerful at eliciting direct reprogramming (Vierbuchen and Wernig, 2011). Our discovery that a Forkhead transcription factor and a bHLH transcription factor genomically and functionally interact also has implications for direct reprogramming. The potential applications of direct reprogramming in cellular therapies or drug screening are limited by the relatively low efficiency of this process, especially in human cells (Pang et al., 2011). Thus, identifying molecular barriers (such as FOXO) that limit direct reprogramming should be a key step in improving the efficiency of this process.
Our finding that FOXO3 inhibits ASCL1-dependent neurogenesis is consistent with the observation that inhibition of PDK1-AKT signaling – which results in FOXO activation–reduces neurogenesis in NPCs, whereas a constitutively active form of AKT or loss of PTEN – which leads to FOXO inhibition – enhances neurogenesis (Gregorian et al., 2009; Oishi et al., 2009). Interestingly, ASCL1/MASH1 protein stabilization has been shown to mediate at least in part the positive effect of AKT signaling on neurogenesis (Oishi et al., 2009). Thus, it is conceivable that the PI3K-AKT signaling pathway coordinates ASCL1 stabilization with ASCL1 activation (via inhibition of the repressive effect of FOXO3 on ASCL1) to achieve a more potent induction of neurogenesis.
The exact transcriptional mechanisms by which FOXO3 inhibits ASCL1-dependent gene expression remain to be established. FOXO3 and ASCL1 co-binding could not be detected by sequential ChIP at the Dll1 and Hes6 loci. Furthermore, FOXO3 and ASCL1 do not appear to physically interact (A.E.W. and A.B., data not shown). It remains possible that negative results in these co-binding experiments are due to technical limitations. Nevertheless, these results may also suggest a ‘displacement model’ to explain FOXO3's ability to inhibit ASCL1 dependent transcription, in which either FOXO3 or ASCL1 could bind to the Dll1 and Hes6 loci, but these transcription factors could not be bound at the same time. FOXO3 may also function by inhibiting the recruitment of ASCL1 co-activators at neurogenic genes. Alternatively, FOXO3 may recruit a repressive complex in the presence of ASCL1. Part of the effects of FOXO3 on ASCL1-dependent transcription may also be indirect: for example, FOXO3 may limit neurogenesis as a consequence of promoting stem cell quiescence. Indeed, ASCL1 is required for normal NSC proliferation (Castro et al., 2011), while also functioning as a positive regulator of neurogenesis (Parras et al., 2004). Thus, FOXO3 may inhibit neurogenesis as a consequence of preventing ASCL1-dependent proliferation of committed progenitors.
How FOXO3 and ASCL1 interact in vivo, and whether FOXO3 restrains neurogenesis in the adult brain by inhibiting ASCL1-dependent transcription, is still unclear. Furthermore, the function of ASCL1 in adult neurogenesis has not been established. We note that the expression of the ASCL1 target gene Hes6 was only slightly increased in FOXO3−/− NSCs in vivo, and that Dll1 expression was not significantly changed in these mice. Thus, other target genes may function downstream of FOXO3 to regulate neurogenesis in vivo. It will be interesting to further dissect the genetic interactions between FOXO3 and ASCL1, as well as their target genes, in stem cell homeostasis and neurogenesis in vivo.
Our data raises the possibility that FOXO3 maintains a stem/progenitor cell state in part by restraining master regulators of differentiated cell fates. Consistent with this model, loss of FOXO1, FOXO3, and FOXO4 in the brain results in increased neurogenesis, followed by NSC depletion in adulthood (Paik et al., 2009; Renault et al., 2009). Recent evidence has also implicated FOXO in inhibiting premature differentiation. In hydra, FOXO depletion leads to enhanced terminal differentiation of foot cells whereas FOXO overexpression promotes the expression of stemness genes (Boehm et al., 2012). FOXO family members have also been implicated in human embryonic stem cell (ESC) pluripotency. FOXO1 is necessary for human ESC pluripotency (Zhang et al., 2011) and FOXO4 has been shown to be required for proteasome activity, a key component of the pluripotency of these cells (Vilchez et al., 2012). FOXO4 has also been found to be necessary for the neural differentiation of human ESCs (Vilchez et al., 2013), which contrasts with our findings that adult FOXO3−/− mice exhibit increased neurogenesis and that triple FOXO knock-out mice also show enhanced neurogenesis post-natally (Paik et al., 2009). Different FOXO family members may act at different points in cellular commitment – the competence of the progenitor pool versus differentiation per se. Alternatively, pluripotent embryonic stem cells may have different requirements from more committed lineage-specific stem cells.
FOXO transcription factors are conserved regulators of longevity (Kenyon, 2010). Our discovery of target genes shared between FOXO and bHLH factors raises the exciting possibility that bHLH transcription factors and Notch signaling may, by association, also be involved in longevity. In this context, it is interesting to note that heterochronic parabiosis – the junction by blood circulation of an old mouse and a young mouse – restores Notch signaling and the regenerative capacity of muscle and neural stem cells in old mice (Conboy et al., 2005; Villeda et al., 2011). The concerted action of FOXO and ASCL1 in the balance between stem cell maintenance and differentiation might play an important role in tissue homeostasis and longevity.
Adult (12-week-old) mouse NPCs were isolated as previously described (Renault et al., 2009). Briefly, whole mouse forebrains were homogenized and incubated for 30 min in HBSS (Invitrogen) with 1 U/ml DispaseII (Roche), 250 U/ml DNaseI (Sigma) and 2.5 U/ml Papain (Worthington) at 37ºC. After mechanical dissociation, cells were purified by sequential 25% and 65% Percoll (Amersham) gradients. Cells were cultured in high growth factor signaling conditions: Neurobasal A (Invitrogen) medium supplemented with penicillin/streptomycin/glutamine (Invitrogen), 2% B27 (Invitrogen) and 20 ng/ml each of FGF2 (Peprotec) and EGF (Peprotec). To induce FOXO3 nuclear accumulation, cells were incubated in low growth factor signaling conditions: Neurobasal A medium supplemented with penicillin/streptomycin/glutamine and 2% B27 for 4 hours, followed by a 1.5 hour incubation with 20 μM LY294002 (LY, Calbiochem-Novabiochem Corp.), a specific PI3K inhibitor, to inhibit residual growth factor signaling. For differentiation, NPCs were plated at 105 cells/ml on poly-D-lysine-coated plates and incubated in differentiation conditions (Neurobasal A medium supplemented with penicillin/streptomycin/glutamine, 2% B27 supplement, and 0.5% [unless specifically noted] fetal bovine serum [Invitrogen]) for 0, 2, or 7 days and differentiation media was replaced every other day.
For ChIP-seq, primary NPCs were isolated, amplified in culture for a limited number of passages (4−9), and ChIP experiments were performed as described (Renault et al., 2009) see Extented Experimental Procedures). For FOXO3 and ASCL1 ChIP-seq, 80-100 × 106 cells were used to generate Illumina single-end libraries. For H3K4me3 and H3K27me3 ChIP-seq, 10-20 × 106 cells were used. Libraries were generated according to the manufacturer's instructions (Illumina). Twenty-five bp reads were generated on an Illumina Genome Analyzer II and subsequently mapped to the 2007 release of the mouse genome (mm9). Peak calling was performed using QuEST 2.4 (Valouev et al., 2008). The highest stringency QuEST parameters for transcription factors were used for FOXO3 and ASCL1 ChIP-seq (Fold enrichment > 50, bandwidth = 30). For histone mark ChIP-seq, Fastq files for ChIP-seq datasets were filtered using the FASTX-Tookit (http://hannonlab.cshl.edu/fastx_toolkit/index.html) and reads with phred scores less than 15 on 85% of their sequence were filtered. Histone modification sites from the filtered and aligned reads were identified using MACS (version macs2 2.0.8) (Zhang et al., 2008) at a false discovery rate threshold 0.01 with a broad setting that enables linking nearby enriched regions (linking cutoff = 0.1). RefSeq genes, genomic features and corresponding coordinates were downloaded from the UCSC Genome Browser.
For lentiviral infection, NPCs were plated on poly-D-lysine at 50,000 cells/cm2 in high growth factor signaling conditions. Sixteen hours after plating, cells were infected with lentiviral-conditioned supernatents at a 1:2 ratio (lentivirus-conditioned supernatent:NPC medium). Twenty four hours post-infection, fresh proliferating medium containing 2 Hg/ml doxycycline (Sigma) was added to induce expression of transduced genes. For gene expression analysis, total RNA was collected 6 hours post-induction and RT-qPCRs were performed as described in the Extended Experimental Procedures. For neurogenesis assays, cells were treated with 2 μg/ml doxycycline for 24 hours, then switched to differentiation conditions containing 2 μg/ml doxycycline. The media was changed after two days and samples were collected after 4 days of differentiation. For neurogenesis assays, cells were fixed and immunocytochemistry was performed as described in the Extended Experimental Procedures with the following antibodies: DCX [Santa Cruz SC-8066, 1:200], Flag [Sigma F7425, 1:1000], HA [Roche 12CA5 #11867423001, 1:200], TUJ1 [Covance PRB-435P, 1:1000]).
We thank A. Sidow, and members of the Sidow lab, A. Valouev, J. Foley, and Z. Weng, for ultra-high throughput sequencing and for providing bioinformatic tools before publication. We particularly thank A. Valouev for assistance with high throughput data analyses and B. Benayoun for help with bioinformatics analysis. We thank L. Attardi, J. Baker, A. Gitler, T. Palmer, J. Sage, A. Sidow, A. Valouev, and members of the Brunet lab for critically reading the manuscript. This work was supported by NIH grants R01 AG026648 (A.B.), P01 AG036695 (A.B. and T.A.R.), R01 MH092931 (M.W.), a California Institute for Regenerative Medicine New Faculty Award (A.B.), an Ellison Medical Foundation Senior Award (A.B.), an Ellison Medical Foundation Junior Award (M.W.), and a Grant-in-Aid U117570528 from the Medical Research Council (UK) to F.G.. M.W. is a New York Stem Cell Foundation-Robertson Investigator. A.E.W. was supported by the Cancer Biology NIH/NRSA training grant 5T32 CA09302 and by an Ellison Medical Foundation/AFAR post-doctoral fellowship. E.A.P. was supported by an NSF graduate fellowship and an NIH graduate fellowship F31 AG043232. D.S.L was supported by NDSEG and NSF graduate fellowships.
A.E.W. and A.B. conceived and planned the study, and A.E.W. performed experiments. E.A.P. performed the histone mark ChIP-seq studies (Figure 4). E.A.P. and D.S.L. performed the FACS studies in Figure S2D-F, Figure 1I and Figure 2E). D.S.L. also performed the FACS in Figure 7C. T.V. and M.W. helped with the neurogenesis and iN studies (Figure 6). N.U. and F.G. helped with gene expression in NS5 cells (Figure 5D), and with bioinformatic analysis (B.M. and F.G.). D.U. helped with analysis of the histone marks (Figure 4) and FOXO or ASCL1 overlaps (Figure S3). M.S. helped with the analysis of FOXO3 motif (Figure 2A, B). T.A.R contributed intellectually to the study and provided guidance to A.E.W.. A.E.W. and A.B. wrote the paper, and all authors provided helpful comments on the manuscript.
Accession Numbers: The Gene Expression Omnibus accession number for the data reported in this paper is GSE12345.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.