A Transcriptional Mechanism Integrating Inputs from Extracellular Signals to Activate Hippocampal Stem Cells

1Adult stem cells maintain tissue function and integrity throughout the lifetime of an organism. They produce mature progenies to replace short-lived cells and repair tissue damage while maintaining their numbers through self-renewing divisions. Many tissue stem cells are relatively quiescent, which delays their attrition and minimizes the accumulation of deleterious mutations. The transit of stem cells between quiescent and activated states is not well understood in most systems. Elucidating the mechanisms that control the activation of tissue stem cells is an important goal in stem cell biology.

2A variety of extracellular signals present in stem cell niches have been shown to influence the activity of tissue stem cells. For example, BMP signaling induces quiescence, while Wnts promote proliferation of skin and blood stem cells. However, the cell-intrinsic mechanisms that mediate the activity of extrinsic signals and promote stem cell quiescence or proliferation are poorly characterized. Niche signals might act by inducing the expression or activity of transcription factors that in turn regulate the large number of genes differentially expressed between quiescent and active stem cells. Transcription factors have indeed been shown to regulate stem cell activity in various tissues by controlling their proliferation, survival, or differentiation. However, it is not known in most instances how these factors are regulated.

3In the adult mammalian nervous system, neural stem cells (NSCs) are found mostly in two regions of the anterior brain, the dentate gyrus (DG) of the hippocampus and the ventricular-subventricular zone (V-SVZ) lining the lateral ventricles, where stem cells produce new neurons that integrate into neuronal circuits of the hippocampus and olfactory bulb, respectively. Most adult NSCs are quiescent and rest in G0, with only a small fraction progressing through the cell cycle at any time. NSC divisions result in the generation of transit-amplifying cells or intermediate progenitor cells (IPCs) that undergo a limited number of rapid divisions before they exit the cell cycle and differentiate into neurons. Clonal analysis in the adult mouse hippocampus in vivo has provided evidence that hippocampal NSCs, also called radial glia-like cells (RGLs), are multipotent and can generate both neurons and astrocytes, and that they use two modes of divisions to self-renew. Some RGLs divide asymmetrically to generate a new RGL and an IPC or an astrocyte, while others divide symmetrically into two new RGLs.

4A particularly important feature of hippocampal neurogenesis is its regulation by a variety of physiological stimuli. Neurogenesis in the hippocampus declines sharply with age, due in part to a reduction of the fraction of RGLs that divide, and it is suppressed by stress and depression. Conversely, an enriched environment, task learning, or seizures stimulate hippocampal neurogenesis, in part by stimulating RGL divisions. Some of the extracellular signals that regulate RGL activity have been identified. In particular, the BMP and Notch signaling pathways maintain RGLs in a quiescent state, while the Wnt and IGF-1 pathways, among others, promote RGL divisions and stimulate neurogenesis. Little is known, however, of how the activity of physiological stimuli or extrinsic signals is transduced inside RGLs to control their divisions. The orphan nuclear receptor Tlx is required to maintain RGLs in proliferation, but how Tlx expression and activity are regulated has not been addressed.

5The proneural transcription factor achaete-scute homolog 1 (Ascl1/Mash1) is an important regulator of neurogenesis in the embryonic nervous system. It is expressed by dividing progenitors and promotes their proliferation, specification, and differentiation into neurons. Moreover, ectopic expression of Ascl1 can reprogram various cell types into neurons. Ascl1 is also expressed in the DG and V-SVZ of the adult rodent brain, but its function there has not been examined. Ascl1 adult expression is mostly confined to IPCs, but recent genetic lineage-tracing experiments have established that it is also present in self-renewing stem cells in both the V-SVZ and hippocampus. Consistent with this finding, Ascl1 was found expressed by a small subset of cycling stem cells in both neurogenic zones. Here we show that Ascl1 expression is rapidly induced by neurogenic signals in hippocampal RGLs, and that Ascl1 has a crucial role in RGL activation in both DG and V-SVZ. Ascl1 is specifically expressed in activated adult stem cells and is specifically required for the exit of stem cells from quiescence.

6To study the function of Ascl1 in hippocampal neurogenesis, we first characterized its expression in the adult hippocampus. Labeling of 2-month-old mouse brains with a monoclonal antibody against Ascl1 showed that in the hippocampus, Ascl1-expressing cells are restricted to the subgranular zone (SGZ) of the DG (Figure 1A). Double labeling for Ascl1 and the cell proliferation marker Ki67 showed that most Ascl1-expressing cells are proliferating (Figure 1A). Double labeling for Ascl1 and markers of progenitor cells of the DG neurogenic lineage showed that Ascl1 is expressed by three distinct progenitor cell populations. It is expressed by a small subset of RGLs, identified by their radial morphology and expression of GFAP. Most Ascl1-positive RGLs are activated, as 83.3% ± 16.7% of them express the cell cycle and cell activation marker MCM2 (Figures 1B and 1C). Ascl1-positive RGLs represent 35.6% ± 3.2% of activated (MCM2+) RGLs and 2.0% ± 0.7% of all RGLs (Figure 1C). Ascl1 is also expressed by nonradial GFAP+ cells in the SGZ, which are also considered to be hippocampal NSCs, and by IPCs, characterized by their SGZ location, proliferative state, and lack of GFAP expression (Figure 1D). Ascl1-positive IPCs represent 13.5% ± 4.1% of all IPCs and are subdivided into two subsets that differ in the expression of the IPC marker Tbr2. Ascl1 is not expressed by more mature cells in the lineage, including doublecortin (DCX)+ neuroblasts and NeuN+ granule neurons (Figure S1; data not shown). Together, these data agree with previous reports showing that Ascl1 expression is restricted to the earliest stages of the neurogenic lineage of the adult DG, including proliferating RGLs and early IPCs, and that it is downregulated before IPCs begin to express neuronal markers and exit the cell cycle.

7The fact that Ascl1 expression is restricted to activated RGLs suggested that this factor might be induced by signals that promote RGL activity and neurogenesis in the hippocampus. To address this possibility, we examined Ascl1 expression in mice treated with the ionotropic glutamate receptor agonist kainic acid (KA), a neurogenic molecule that induces progenitor divisions in the DG. A single injection of KA in 8- to 9-week-old wild-type (WT) mice produced, as expected, a robust increase in the number of MCM2+ RGLs in the DG, which became detectable 2 days after injection (Figure S2). Remarkably, KA induced Ascl1 expression in RGLs with more rapid kinetics, as the number of Ascl1+ RGLs was already increased 24 hr after injection of KA (Figure S2). Thus, a significant fraction of quiescent (MCM2−) RGLs expressed Ascl1 at 24 hr after injection (7.8% ± 1.3% Ascl1+ MCM2− RGLs in KA-injected mice and 1.4% ± 0.9% in saline-injected mice at 24 hr; Figures 2A, 2B, and S2). The number of Ascl1+ quiescent RGLs decreased, while the number of activated RGLs increased at 2 days and 4 days (Figure S2). Together, these data indicate that KA administration induces Ascl1 expression in RGLs that are still quiescent. Therefore, Ascl1 induction precedes RGL activation.

8To determine whether antineurogenic stimuli also regulate Ascl1 expression in the DG, we examined Notch signaling. Deletion of the Notch pathway component RBPJk has been shown to transiently induce the proliferation of hippocampal RGLs, followed later by a depletion of RGLs. We deleted RBPJk from RGLs by injecting tamoxifen in 3-month-old mice carrying an inducible RBPJk-mutant allele (Glast::CreERT2; RBPJkloxP/loxP; Rosa26R-stop-YFP mice, called thereafter RBPJk cKO mice). Examination of the hippocampus 7 days later revealed a dramatic increase in the numbers of MCM2+ activated RGLs and of Ascl1-expressing RGLs in RBPJk cKO mice compared with control mice, demonstrating that loss of Notch signaling stimulates both the proliferation of RGLs and the expression of Ascl1 (Figures 2C and 2D). A fraction of RGLs in RBPJk cKO DGs expressed Ascl1, and not MCM2 (3.60% ± 1.50% in RBPJk cKO versus 0.25% ± 0.25% in control mice; Figure 2D), suggesting that like KA administration, inactivation of the Notch pathway in RGLs sequentially promotes Ascl1 expression and quiescence exit. We also examined the effect of voluntary exercise on RGL proliferation and expression of Ascl1 and found that it did not significantly increase proliferation of hippocampal RGLs, but only that of IPCs, as previously reported. Together, our results demonstrate that neurogenic stimuli rapidly induce Ascl1 expression in quiescent RGLs, which in turn suggests that this factor might be implicated in RGL activation.

9To directly address the role of Ascl1 in the activation of DG RGLs, we generated triple-transgenic mice that were homozygous for a conditional mutant allele of Ascl1 and also carried the Glast-CreERT2 allele to delete Ascl1 in RGLs in a tamoxifen-dependent manner and the Rosa26-floxed stop-YFP reporter transgene to identify cells having undergone Cre-mediated recombination by their expression of YFP. Administration of tamoxifen for 5 days to postnatal day 60 (P60) triple-transgenic mice and control mice (carrying Glast-CreERT2 and Rosa26-floxed stop-YFP, but WT for Ascl1) resulted in widespread induction of YFP in radial GFAP+, Nestin+ RGLs (Figures 3A, 3B, and data not shown). However, examination by immunolabeling revealed that a fraction of SGZ cells that expressed YFP and had therefore recombined the Rosa26-floxed stop-YFP locus also expressed Ascl1 and had therefore not recombined the Ascl1flox locus, indicating that recombination at the two loci was partially uncoupled. We also examined triple-transgenic mice carrying a different conditional mutant allele of Ascl1, in which a PGK promoter-neo cassette remained inserted on the 3′ side of the Ascl1 locus (Ascl1neoflox mice; Figures 3C and S3A). Interestingly, even without tamoxifen-induced recombination, Ascl1 RNA and protein expressions were significantly reduced in the DG of these mice compared with WT mice (Figures 3D and S3B), suggesting that Ascl1neoflox is a hypomorphic allele. Analysis of Ascl1neoflox mice at P10 did not reveal any overt morphological defect of the DG, and the rate of RGL proliferation was similar to that found in WT mice, indicating that the hypomorphic allele of Ascl1 does not result in a developmental defect in the DG (Figure S3). Tamoxifen administration to these mice at P60–P64 resulted in undetectable Ascl1 expression in the DG at P90 (Ascl1neocKO mice; Figures 3C, 3D, S3A, and S3B). We therefore used Ascl1neocKO mice in the rest of this study to examine the effect of loss of Ascl1 on hippocampal neurogenesis, and we used Ascl1neoflox mice to examine the effect of a reduced expression of Ascl1.

10To determine whether Ascl1 deletion has an impact on DG RGLs, we administered tamoxifen at P60–P64 and examined Ascl1neocKO mice 1 month later, at P90. Double labeling for YFP and for MCM2, Ki67, or bromodeoxyuridine (BrdU) after a 2 hr pulse revealed a near-complete absence of proliferating YFP+ cells in the SGZ of Ascl1neocKO mice (Ascl1neocKO mice versus WT mice, 19.65 ± 19.65 versus 4,635 ± 921 YFP+ MCM2+ cells; 0 versus 2,875 ± 815 YFP+ KI67+ cells; 9.8 ± 4.9 versus 471.6 ± 115.5 YFP+ BrdU+ cells), while proliferating cells were present in Ascl1neoflox mice, albeit in reduced numbers (2,233 ± 632 MCM2+ cells; 1,225 ± 298 KI67+ cells; 142.8 ± 10.2 BrdU+ cells; Figures 3E–3H). Moreover, triple labeling for YFP, GFAP, and MCM2 to mark activated stem cells or Ki67 to mark proliferating stem cells demonstrated a complete absence of recombined RGLs that were activated or proliferating in Ascl1neocKO mice, while activated and cycling RGLs were present, but less numerous, in Ascl1neoflox mice than in WT mice (Ascl1neocKO versus Ascl1neoflox versus WT mice, 0 versus 126.3 ± 26.3 versus 281.2 ± 70.1 MCM2+ cells; 0 versus 66.0 ± 29.5 versus 161.2 ± 48.5 KI67+ cells; Figures 3I–3K; see also Figures S3D–S3G for measures of proliferation in the DG of Ascl1flox and Ascl1cKO mice). Rarely dividing RGLs, characterized by their capacity to retain BrdU, were labeled in tamoxifen-injected P90 mice by 10 days of BrdU administration followed by 20 days of chase (Figure 3L). BrdU label-retaining cells were present in WT mice and to a lesser extent in Ascl1neoflox mice, but were again completely absent in Ascl1neocKO mice (0 cells in Ascl1neocKO mice; 9.7 ± 4.9 cells in Ascl1neoflox mice; 43.8 ± 6.2 cells in WT mice; Figure 3L). We could thus demonstrate by several independent methods the complete inability of RGLs to exit quiescence and divide in the absence of Ascl1, and therefore establish that Ascl1 is essential for activation of RGLs in the adult hippocampus. Interestingly, Ascl1 deletion had no significant effect on the rate of RGL proliferation in the postnatal DG. When Ascl1 was deleted by tamoxifen administration in Ascl1neocKO mice at P7, RGLs continued to proliferate at P10 at a rate that was not significantly different from that seen in WT mice (Figure S3H). Therefore, the absolute requirement of Ascl1 for RGL activity is specific to the adult DG.

11Dividing RGLs in the adult hippocampus generate IPCs that proliferate before producing postmitotic granule neurons. Since RGLs require Ascl1 to divide, the production of IPCs and their neuronal progeny might also depend on Ascl1 function. The absence of Ki67+, MCM2+, and BrdU+ cells in the SGZ of Ascl1neocKO mice (Figures 3F–3H) already suggested that IPCs are indeed missing in these mice. We further examined neurogenesis by double labeling the hippocampus of Ascl1neocKO mice for YFP and for Tbr2 to mark IPCs, for DCX to mark neuroblasts, and for NeuN to mark granule neurons. No YFP+ cells expressed these markers in the DG of Ascl1neocKO mice at P90, demonstrating that no new IPCs or granule neurons were produced in these mice (Figures S3I–S3K; data not shown). Ascl1 is therefore absolutely required for the generation of IPCs and for neurogenesis in the hippocampus.

12We also examined neurogenesis in the V-SVZ of Ascl1neocKO mice to determine whether the role of Ascl1 in adult RGLs extends to the other main neurogenic region of the adult rodent brain. Analyzing the expression of GFAP, DCX, EGFR, and GFP in Ascl1neocKO and WT mice showed that deletion of Ascl1 results in a severe decrease in the fraction of RGLs of the V-SVZ that are activated (GFAP+ EGFR+) and proliferate (BrdU label retaining), and in a severe reduction in the production of DCX+ neuroblasts (Figures 4 and S4). Therefore, Ascl1 is essential for NSC activation and proliferation and for neurogenesis in the two main neurogenic regions of the adult brain.

13Ascl1 is expressed in proliferating RGLs, and the loss of Ascl1 results in an arrest of RGL proliferation, suggesting that this gene is required in RGLs to promote their divisions. However, Ascl1 is also expressed in some IPCs, raising the alternative possibility that Ascl1 is primarily required for the generation and/or division of IPCs, and that the arrest of RGL divisions is a secondary consequence of the loss of IPCs. In particular, a loss of IPCs might disrupt Notch signaling in the SGZ, resulting in a transient increase in RGL proliferation, followed later by a reduction of proliferation due to RGL exhaustion. To address this possibility, RGL proliferation was monitored in Ascl1neocKO mice just 4 days after the beginning of tamoxifen administration at P60. RGLs had already stopped dividing in P64 Ascl1neocKO mice, indicating that the effect of Ascl1 deletion on RGL proliferation is rapid and therefore likely direct (Figures 5A and 5B). Moreover, although RGLs in the DG of P64 Ascl1neocKO mice are mostly YFP+ and have therefore recombined (Figures 3B and 5A), WT Tbr2+ IPCs and DCX+ neuroblasts that were produced by RGLs before tamoxifen administration at P60–P64 are still present (Figure 5C). This suggests that the RGL proliferation defect is not the consequence of a loss of Ascl1-mutant IPCs and neuroblasts, i.e., Ascl1 is required cell autonomously for RGL divisions.

14To more rigorously address the cell autonomy or noncell autonomy of Ascl1 function, we performed a mosaic analysis. We activated CreERT2 in a fraction of RGLs with only one injection of tamoxifen at P60, resulting 1 month later in an intermingling of recombined YFP-positive RGLs and nonrecombined YFP-negative RGLs and their progenies (Figure 5D). In WT mice that had received a single tamoxifen injection, a fraction of recombined YFP+ RGLs was proliferating (Figures 5E and 5F). In contrast, recombined YFP+ RGLs did not proliferate in mosaic Ascl1neocKO and Ascl1cKO mice, irrespective of the recombination efficiency (Figures 5E, 5F, and S5). Therefore, the proliferation defect of Ascl1-mutant RGLs cannot be rescued by the presence of nearby WT cells, demonstrating that Ascl1 is required cell autonomously in RGLs for their divisions.

15Next, we examined the phenotype of the cells remaining in the SGZ after Ascl1 deletion. Antibody labeling of the DG of P90 Ascl1neocKO mice revealed that these cells retain the typical radial morphology of RGLs and maintain expression of the RGL markers GFAP, Nestin, and Sox2 and do not express the astrocytic marker S100β or the oligodendrocyte progenitor marker Olig2 (Figures 6A, 6B, and S6A–S6C). RT-PCR analysis showed that expression of p16INK4a/Cdkn2a was not elevated in the DG of Ascl1neocKO mice, suggesting that RGLs in these mice do not become senescent. Moreover, RGLs in Ascl1neocKO maintained normal levels of the cyclin-dependent kinase inhibitor p57Kip2, which is required for quiescence of hippocampal RGLs. These data, together with the lack of MCM2 expression in Ascl1-deficient RGLs (Figure 3J), show that Ascl1 is not required for the maintenance of RGLs, but specifically for their activation, and that loss of Ascl1 keeps RGLs in an inactive and undifferentiated state.

16Although Ascl1-deficient RGLs are unable to proliferate when in a steady state, they might still be able to respond to a potent neurogenic stimulus such as KA. We therefore injected tamoxifen in Ascl1neocKO mice at P60 to delete Ascl1, then injected KA at P86 and analyzed the DG at P90. KA failed to activate RGLs in Ascl1neocKO mice, while it strongly activated them in WT mice (Figures 6E and 6F). Therefore, loss of Ascl1 in RGLs results in an inactive state that cannot be reversed by KA stimulation. We also confirmed the inactivated state of Ascl1-deficient RGLs by performing an in vitro neurosphere assay in the presence of mitogens. The DG was dissected from 7- to 8-week-old mice and dissociated, and single-cell suspensions were cultured at clonal density in the presence of FGF2 and EGF. Cultures of WT DG cells produced large numbers of primary neurospheres that generated secondary neurospheres when passaged. In contrast, Ascl1neocKO DG cell cultures produced very few neurospheres, and antibody labeling showed that these neurospheres maintained expression of Ascl1 and therefore originated from cells that had escaped recombination (Figures 6G, 6H, and S6E). Therefore, Ascl1-deficient RGLs are quiescent and unable to respond to mitogens and divide in vitro.

17We next examined whether the inactive state of Ascl1-deficient RGLs was stable over a longer period. Analysis of Ascl1neocKO mice 5 months after Ascl1 deletion, at P210, showed that RGLs retained a radial morphology and GFAP expression and remained Ki67 negative (Figure 6I). Interestingly, while the total number of RGLs in the DG of WT mice decreased considerably between P65 and P90 and decreased further at P210 (22,116 ± 1,681 at P65, 14,453 ± 1,021 at P90, and 12,516 ± 1,063 at P210), their number in Ascl1neocKO mice did not change between P65 and P90 and decreased only slightly and nonsignificantly at P210 (24,533 ± 1,399 at P65, 26,098 ± 1,913 at P90, and 20,764 ± 395 at P210; Figure 6J). As the age-related attrition of hippocampal RGLs is thought to result from non-self-renewing divisions, the maintenance of RGL numbers in older Ascl1neocKO mice supports the finding that RGLs do not divide in these mice.

18To identify target genes that mediate the proliferative role of Ascl1 in hippocampal stem cells, we examined the genome-wide binding of Ascl1 in adult hippocampus-derived NSCs by chromatin immunoprecipitation-sequencing (ChIP-seq) with an anti-serum against Ascl1 (Figures 7 and S7). Ascl1 was bound to 7,826 high-confidence sites in the genome of AH-NSCs (Figure 7A), a majority of which mapped to enhancers previously identified in cultured NSCs (Figure 7B; ). Moreover, a large fraction of enhancers active in proliferating NSCs featured an Ascl1-binding event in AH-NSCs (Figure 7C).

19We then established a list of the genes associated with an Ascl1-binding event in the AH-NSC genome (Table S1) and searched for candidate direct targets of Ascl1, i.e., genes that are both bound by Ascl1 and misregulated in RGLs of Ascl1neocKO. We reasoned that since Ascl1 is required for NSC proliferation in both DG and V-SVZ, it might regulate some of the same genes in the two tissues. We therefore intersected our list of Ascl1-bound genes with a recently published list of V-SVZ genes enriched in activated NSCs compared with quiescent NSCs (Figure S7; Table S2). The 250 genes that are both bound by Ascl1 in AH-NSCs and enriched in activated V-SVZ NSCs include, in addition to known Ascl1 targets such as the Notch ligands Dll1 and Dll3, several cell-cycle regulators such as E2f1, Ccnd2, Cdc6, and Skp2. To establish whether these genes are regulated by Ascl1 in DG RGLs, we used fluorescence-activated cell sorting to purify YFP+ cells from the DG of Ascl1neocKO and WT mice 4 days after tamoxifen treatment, and we analyzed gene expression by qRT-PCR. We found that the cyclin Ccnd2 and the ribonucleotide reductase subunit Rrm2 were significantly downregulated when Ascl1 was deleted in RGLs, while E2f1 and Cdc6 expression were not detectable in either WT or mutant cells, and expression of Skp2 and non-cell-cycle genes such as Fbl were not significantly reduced (Figure 7E). Together, these data demonstrate that Ascl1 controls the proliferation of hippocampal RGLs by directly activating the expression of Ccnd2, Rrm2, and possibly additional cell-cycle genes.

20Stem cells in adult tissues respond to environmental signals by adjusting the production of mature cells to the needs of the tissue. Deciphering the pathways that link physiological stimuli to NSC activity requires characterization of the intrinsic machinery that controls stem cell activation. We show in this study that signals regulating hippocampal stem cell activity control the expression of Ascl1, and that this factor has an essential role in stem cell activation.

21Ascl1 expression in the adult brain has often been described as being restricted to IPCs in the SVZ and DG. We found, however, that Ascl1 is already expressed in proliferating RGLs in the hippocampus, in agreement with earlier studies. Ascl1 expression is then presumably maintained by the IPCs that are produced when RGLs divide. Most Ascl1+ RGLs are activated, but only about a third of activated RGLs express Ascl1 detectably, which might be due to an oscillation of Ascl1 expression in activated RGLs as in embryonic neural progenitor cells.

22The results of our experiments with KA-injected mice suggest that neurogenic stimuli such as neuronal activity promote stem cell activation in the hippocampus by inducing Ascl1 expression in quiescent RGLs. The essential role of Ascl1 in the activation of adult NSCs suggests that other neurogenic signals controlling this step, including Wnt signals in the hippocampus and VEGF in the V-SVZ, might also act by inducing Ascl1 expression. Our results with RBPJk conditional mutant mice also suggest that antineurogenic stimuli, including the Notch pathway, suppress stem cell activity by repressing Ascl1 expression in RGLs. The mechanism by which the Notch-RBPJk pathway represses Ascl1 expression can be inferred from studies in the embryonic brain and in cancer cells, which have shown that the Notch-induced Hes factors directly repress the Ascl1 gene. Moreover, antineurogenic signals may also target Ascl1 protein activity. The maintenance of hippocampal stem cell quiescence by BMPs may involve an inactivation of Ascl1 protein by BMP effectors Id proteins, which are known to block Ascl1 activity by preventing its dimerization with E proteins. FoxO3, which acts downstream of the insulin/IGF-1 signaling pathway to maintain hippocampal stem cell quiescence, shares many targets with Ascl1 and may also function by antagonizing Ascl1 function. Inactivation of Ascl1 protein in RGLs that have begun to transcribe the Ascl1 gene may be important to slow down the transition to an active state or to accelerate the return of active RGLs to quiescence.

23Mice carrying the hypomorphic allele Ascl1neoflox express Ascl1 in the DG at a reduced level and retain only a small fraction of the proliferating RGLs found in WT mice (Figure 3). The finding that a change in Ascl1 expression level translates into a change in the fraction of RGLs that proliferate suggests that extrinsic signals may fine-tune the rate of hippocampal neurogenesis by modulating the expression level of Ascl1, as shown for PDK1/Akt signaling, which regulates Ascl1 protein stability in the embryonic brain.

24A cell that exits quiescence and transits from the G0 to the early G1 phase of the cell cycle assembles a prereplication complex that contains minichromosome maintenance proteins, including MCM2. MCM2 expression therefore marks not only cycling cells, but also activated cells that have exited the quiescent state but not yet re-entered the cell cycle. The glutamate receptor agonist KA induces the expression of Ascl1 before that of MCM2, indicating that Ascl1 induction is one of the first steps in the pathway through which neuronal activity promotes the activation of quiescent RGLs. Since MCM2 is not expressed in Ascl1-deficient RGLs even after KA stimulation, Ascl1 is also absolutely required for the quiescence exit of hippocampal RGLs.

25Because Ascl1-deficient RGLs do not exit the quiescent state, it is not possible to ascertain whether Ascl1 also regulates the cell-cycle progression of RGLs. However, Ascl1 promotes the proliferation of progenitor cells in the embryonic brain (which do not enter quiescence), and it directly regulates the expression of cell-cycle regulatory genes in AH-NSCs, including Ccnd2 and Rrm2, suggesting that it promotes the divisions of hippocampal RGLs in addition to their activation. Ascl1 has previously been shown to promote cell proliferation in cancer cells and in the injured zebrafish retina by regulating Wnt-signaling genes. In glioblastoma cancer stem cells (GBM CSCs), Ascl1 has been shown to bind a site near the Wnt antagonist gene Dkk1, whose regulation mediates Ascl1 activity in these cells. In AH-NSCs, however, Ascl1 does not bind this site and many other sites bound in GBM CSCs, and reciprocally, many Ascl1-bound loci were found in AH-NSCs, and not in GBM CSCs, including sites near the cell-cycle regulators E2f1, Ccna1, Ccnd2, and Skp2 (data not shown), suggesting that Ascl1 controls the proliferation of hippocampal RGLs by regulating different genes from those regulated in cancer cells and the injured retina.

26Transcriptomic and genetic studies suggest that NSCs undergo profound changes in their oxygen and lipid metabolisms and cell adhesion properties when they exit quiescence. Further characterization of Ascl1 target genes in RGLs should elucidate whether Ascl1 directly controls these physiological changes in addition to promoting cell-cycle re-entry and cell-cycle progression. Ascl1 inactivation also blocks RGL activity in the V-SVZ. Whether Ascl1 acts through the same downstream mechanisms in the two adult neurogenic regions remains to be addressed.

27In contrast with the complete lack of RGL activity in the adult DG in the absence of Ascl1, RGL proliferation was not significantly affected by Ascl1 deletion in the early postnatal DG, supporting our earlier finding that Ascl1 is not required for DG morphogenesis during embryonic development. Interestingly, cell proliferation in the DG becomes also increasingly dependent on the Ascl1 target Ccnd2 between early postnatal and adult stages. Therefore, a switch in the genetic control of DG RGL proliferation occurs during the first few weeks of life, with the activation of a mitogenic pathway involving Ascl1 and Ccnd2.

28Ascl1-deficient RGLs remain in Ascl1neocKO mice for at least 5 months without dividing, differentiating, or dying. The permanent cell-cycle arrest can be readily explained by the cell-autonomous role of Ascl1 in RGL activation. The lack of astrocyte differentiation may also be partially explained by the absence of proliferation, as astrogenesis is normally coupled to RGL divisions. However, astrocytes can also be generated by direct differentiation of RGLs without cell division. In the embryonic brain, activation of Notch signaling by ligands presented by IPCs and young neurons switches neurogenic progenitors to an astrocytic fate. In the hippocampal SGZ, Ascl1 deficiency greatly reduces the expression of the Notch effectors Hes1 and Hes5 (Figure S6D), likely due to the elimination of Notch-ligand-presenting IPCs and young neurons. The resulting decrease in Notch activity may thus block the astrocytic differentiation of Ascl1-deficient RGLs.

29Numerous transcription factors have been shown to stimulate the self-renewal of stem cells in tissues such as the blood and the skin. These factors often act by regulating multiple aspects of the biology of the stem cells; for example, by suppressing their differentiation, senescence, or apoptosis. Only a few factors, such as Gata3 in hematopoietic stem cells and Runx1 in hair follicle stem cells, have been proposed to primarily regulate adult stem cell divisions. In the brain, the orphan nuclear receptor Tlx promotes hippocampal NSC proliferation through induction of Wnt7a and repression of p21/WAF1. However, how Tlx is regulated is currently not known. The function of Ascl1 of controlling adult stem cell activity in response to environmental signals is therefore so far unique in the adult brain. It is akin to that of MyoD in muscle satellite stem cells, which is expressed shortly after quiescent satellite cells have been activated, and which in turn induces the expression of the component of the prereplication complex Cdc6. Identifying the molecular pathways that control Ascl1 expression at transcriptional and posttranscriptional levels will be important in order to learn how to manipulate hippocampal neurogenesis for therapeutic purposes.

30Mice were housed, bred, and treated according to the guidelines approved by the Home Office under the Animal (Scientific Procedures) Act 1986. All experimental procedures involving mice have been approved by the Animal Welfare and Ethical Review Panel of the National Institute for Medical Research. RBPJkloxp/loxp animals were generated as previously described and bred to Glast::CreERT2 BAC transgenic mice. Ascl1neoflox/neoflox mice, in which exon 1 of the Ascl1 gene is flanked by loxP sites, were bred with Glast-CreERT2 knockin mice and with Rosa26-floxed stop-YFP reporter mice to generate Glast-CreERT2; Ascl1neoflox; R26 YFP mice, which are heterozygous for Glast-CreERT2 and homozygous for Ascl1neoflox and R26 YFP. Both Glast::CreERT2 and Glast-CreERT2 lines target both radial and horizontal astrocytes in the DG. In order to remove the PGK promoter-neo cassette from the Ascl1 locus, Ascl1neoflox animals were crossed with actβ-Flp mice (The Jackson Laboratory).

31For activation of the CreERT2 recombinase, P60 animals were administered intraperitoneally (i.p.) 4-hydroxytamoxifen (TAM) for 5 consecutive days. For mosaic experiments, P60 animals received a single TAM injection at the same concentration. All animals including WT and Ascl1neoflox mice received TAM injections. To examine proliferating progenitors, mice received a single i.p. injection of BrdU 2 hr prior to tissue collection. To examine slowly dividing RGLs, mice received 5 daily BrdU injections followed by 5 consecutive days of BrdU-containing drinking water. Mice were sacrificed 20 days later. Male mice received KA as a single i.p. injection and were monitored for 90 min after KA injection. Animals that did not display rearing and falling were sacrificed 1, 2, or 4 days later and processed as described below.

32Animals were transcardially perfused with saline followed by 4% paraformaldehyde (PFA). Brains were postfixed with 4% PFA for 2 hr at 4°C and sectioned coronally at 40 μm with a vibratome. The immunofluorescence procedure and the primary and secondary antibodies are described in Supplemental Information.

33Labeled cells were counted in every ninth 40 μm section through the entire rostrocaudal length of the DG (−0.82 mm to −4.24 mm from bregma). Counted cells were divided by the number of z planes counted to obtain the number of cells per 1 μm, and then multiplied by the total length of the DG. To count RGLs (GFAP+ radial cells), cells were deemed radial if the cell body clearly associated with a DAPI-positive nucleus was located in the SGZ and had a single radial process extending through at least two-thirds of the granule layer. In all figures, the cell numbers counted in WT and Ascl1neocKO mice are numbers of YFP+ marker+ double-labeled cells, while the numbers counted in Ascl1neoflox mice are for Ki67+ cells only, since YFP is not expressed in these mice.

34Coronal sections 14 μm long were cut from fresh-frozen brains in OCT with a cryostat and placed on slides. The SGZ of WT, Ascl1neoflox, and Ascl1neocKO mice was excised by a PLAM laser-capture microdissection system (Zeiss) and collected in an adhesive cap. RNA from microdissected tissue was extracted and purified using Arcturus Pico Pure RNA Isolation Kit (Applied Biosystems) and reverse transcribed using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems). Gene expression was detected using TaqMan Gene Expression Assays (Applied Biosystems). The protocol used for fluorescent-activated cell sorting is described in Supplemental Information.

35Statistical analyses were conducted using a two-sample t test with equal variance in Prism software. Values represent mean values ± SD.

36Clonal primary and secondary neurosphere cultures were performed from dissociated DG dissected from 7- to 8-week-old mice as described. The number of neurospheres per well was counted 10 days after plating.

37For chromatin immunoprecipitation, adult hippocampus-derived NSCs were fixed and processed as described and immunoprecipitated using a rabbit anti-Ascl1 antibody (Abcam, ab74065, 4.5 μg per ChIP sample). DNA libraries were prepared and sequences analyzed as described. A total of 13.5 million nonredundant reads were used to call peaks, and only peaks with an FDR-corrected q value ≤ 1 × 10−5 were used for the analysis. p300 and H3K27ac data and active enhancer definitions in NSCs were from ).