SoxD genes are required for adult neural stem cell activation

The adult neurogenic niche in the hippocampus is maintained through activation of reversibly quiescent neural stem cells (NSCs) with radial glia-like morphology (RGLs). Here, we show that the expression of SoxD transcription factors Sox5 and Sox6 is enriched in activated RGLs. Using inducible deletion of Sox5 or Sox6 in the adult mouse brain, we show that both genes are required for RGL activation and the generation of new neurons. Conversely, Sox5 overexpression in cultured NSCs interferes with entry in quiescence. Mechanistically, expression of the proneural protein Ascl1 (a key RGL regulator) is severely downregulated in SoxD -deﬁcient RGLs, and Ascl1 transcription relies on conserved Sox motifs. Additionally, loss of Sox5 hinders the RGL activation driven by neurogenic stimuli such as environmental enrichment. Altogether, our data suggest that SoxD genes are key mediators in the transition of adult RGLs from quiescence to an activated mitotic state under physiological situations


In brief
Li et al. show that Sox5 and Sox6 transcription factors are required for neural stem cell activation and for the generation of new neurons in the adult hippocampus in mice.In that context, they also show that Sox5 and Sox6 modulate the transcription of the stem cell activator Ascl1.

INTRODUCTION
Neurogenesis, a developmental process generating functionally integrated neurons, occurs throughout life in certain areas of the mammalian brain, such as the ventricular-subventricular zone (V-SVZ) lining the lateral ventricles (Obernier and Alvarez-Buylla, 2019) and the subgranular zone (SGZ) in the dentate gyrus (DG) of the hippocampus (Bond et al., 2015;Goncalves et al., 2016).New neurons produced in the adult SGZ integrate into the adjacent granule cell layer and participate in learning and memory processes (Christian et al., 2014;Goncalves et al., 2016).Additionally, several physiological and pathological situations, such as physical exercise, task learning, environmental enrichment, and seizures, can stimulate neurogenesis in the adult DG (Rolando and Taylor, 2014).
The adult SGZ neurogenic niche is maintained through the activation of neural stem cells (NSCs) with radial glia-like morphology (RGLs).They are mostly in a reversible state of quiescence that protects cells from DNA damage and prevents depletion of the RGL population (Urba ´n et al., 2019).At any given time, a relatively small population of quiescent RGLs (qRGLs) will activate and will divide symmetrically to self-renew or asymmetrically to generate an RGL and an intermediate progenitor cell (IPC) (Pilz et al., 2018).IPCs generate neuroblasts, which exit the cell cycle to differentiate into granule neurons (GNs) (Chris-tian et al., 2014).However, active RGLs (aRGLs) in the SGZ divide mostly asymmetrically and, on average, will divide a few times before being depleted (Encinas et al., 2011;Pilz et al., 2018) or return to a temporal shallow quiescence or resting state, while most RGLs will remain in a dormant state of deep quiescence (Bottes et al., 2021;Harris et al., 2021).For that reason, a crucial aspect to understand the long-life maintenance of adult neurogenesis is to clarify the mechanisms that control the balance between quiescence and activation in the RGL population.
The transition of RGLs between quiescent and activated states is regulated by local niche signals that promote either quiescence (BMPs, Delta/Notch, and tonic-GABA) (Ehm et al., 2010;Mira et al., 2010;Song et al., 2012) or cell proliferation and neurogenesis (WNT, SHH, and IGF pathways, among others) (Bracko et al., 2012;Favaro et al., 2009;Lie et al., 2005;Nieto-Estevez et al., 2016).Moreover, few of the intrinsic factors linked to the quiescent (Hes5, p57, FOXO3, and REST) or active (TLX and Ascl1) state of NSCs in the adult DG have been identified (Morales and Mira, 2019;Shin et al., 2015).However, little is known about how RGLs integrate signals and intrinsic factors in order to transit from the quiescent to the active/proliferative state.
We show here that SoxD transcription factors are expressed predominantly in RGLs in the adult SGZ and that their expression is inhibited by the quiescence-promoting factor BMP4.Using mice with inducible conditional inactivation of Sox5 or Sox6 (Dumitriu et al., 2006;Dy et al., 2008), we show that both SoxD factors are required for Ascl1 expression in RGLs and are consequently required for RGL activation and new neuron generation in the SGZ.

Sox5 and Sox6 are expressed predominantly in RGLs in the adult dentate gyrus
To explore if Sox5 and Sox6 transcription factors (the only two SoxD family members clearly expressed in the adult DG) (Hochgerner et al., 2018) control cell proliferation during adult SGZ neurogenesis, we first characterized their expression pattern in the adult DG.For the sake of clarity, hippocampal stem cells analyzed in vivo are called, hereafter, RGLs, while in culture they are referred to as neural stem cells.
As the Sox2 + cell population includes both RGLs and IPCs with non-radial morphology (Suh et al., 2007), we used a transgenic Sox2-EGFP reporter mouse line, in which RGLs could be recognized by the radial processes of GFP-filled cells extending toward the molecular layer.We observed that the vast majority of GFP + RGLs expressed both Sox5 and Sox6 (Figure 1B).Using GFAP staining, we confirmed that around half of Sox5 + and Sox6 + cells showed radial morphology and GFAP expression (rGFAP; 48.5% ± 4.6% and 56.2% ± 6.9%, respectively; Figure 1F) and that the majority of rGFAP + cells express Sox5 or Sox6 (77.8% ± 5.5% and 73.8% ± 6.1%, respectively; Figure 1F).
To further analyze Sox5 expression in the RGL population, we used Sox2-creER T2 /Rosa26-YFP mice injected for 5 days with tamoxifen (TAM) to induce creER T2 nuclear translocation leading to YFP activation via cre-mediated recombination, as the forebrain enhancer of Sox2 drives cre-recombinase expression mostly in RGLs (Favaro et al., 2009)( Figure 1G).We observed significantly lower Sox5 expression in non-proliferating quiescent qRGLs (rGFP + MCM2 À cells; MCM2 is a marker for proliferating cells) in comparison with that found in proliferating active aRGLs (rGFP + MCM2 + cells; relative levels of 100.0 ± 2.83 versus 135.00 ± 7.90, respectively; Figures 1G and 1H).However, we did not observe any significant change in Sox6 levels when comparing qRGLs and aRGLs (Figure S1).These data indicate that SoxD genes are expressed in adult hippocampal RGLs, and at least Sox5 is enriched in activated RGLs.
In summary, Sox5 and Sox6 are co-expressed in hippocampal RGLs, with Sox5 expression being higher in aRGLs.Their expression is maintained in IPCs, but progressively lost as IPCs exit the cell cycle and differentiate into GNs (summarized in Figure 1I).
Sox5 and Sox6 are required for the activation of qRGLs in the adult DG Having shown that SoxD genes are expressed in RGLs, we next asked whether they have important functions in neurogenesis in the adult hippocampus.For that purpose, we adopted a genetic loss-of-function approach coupled to fate mapping.We generated mice harboring Sox5 and/or Sox6 conditional null alleles (Sox5 fl/fl and/or Sox6 fl/fl ) ( Dumitriu et al., 2006;Dy et al., 2008), a transgene expressing a tamoxifen-inducible form of cre-recombinase (CreER T2 ) under the control of a forebrain Sox2 enhancer (Sox2-cre-ER T2 ) ( Favaro et al., 2009), and a Rosa26floxed stop-YFP reporter allele (R26-YFP; Figure 2A) (Srinivas et al., 2001).These mice were named Sox5 icKO and/or Sox6 icKO .R26-YFP allowed fate mapping of RGL cells that had undergone selective Sox5 or Sox6 deletion (Figure 2B), and all analyses were done exclusively in YFP + cells.Sox2-CreER T2 mice carrying the R26-YFP reporter transgene, but wild-type or heterozygous for Sox5 fl or Sox6 fl , were used as controls.Two-month-old control and mutant mice were injected for 5 consecutive days with TAM and analyzed 7, 14, and 30 days after the first TAM injection (dpi; Figure 2A).
First, we evaluated the efficiency of the TAM-induced knockout at 7 dpi.Compared with control mice, there was a marked decrease in the number of Sox5 + YFP + cells in Sox5 icKO mice (53.8% ± 3.9% versus 89.3% ± 2.2% in control, considering 100% as the total number of YFP + recombined cells; p = 0.0002; Figures S2A and S2D).Similarly, the number of Sox6 + YFP + cells in Sox6 icKO mice was clearly reduced after TAM injection (33.7% ± 3.0% versus 75.9% ± 1.1% in control; p = 0.000006; Figures S2B and S2D).Although a fraction of YFP + cells had failed to recombine all Sox5 fl and Sox6 fl alleles, we observed significant alterations in the neurogenic niche, as we describe below.We discarded the possibility of analyzing double Sox5 icKO /Sox6 icKO mice, as we estimated that only 3% of recombined YFP + cells had lost both Sox5 and Sox6 expression (Figure S2E).
As Sox2 is a transcription factor essential for RGL stemness, we checked its expression and did not find any differences in YFP + cells in control, Sox5 icKO , or Sox6 icKO mice (Figures S2A-S2C).These results indicate that Sox2 expression does not depend on Sox5 or Sox6 in the adult SGZ.Moreover, at 7 dpi the proportions of RGLs among YFP + cells in Sox5 icKO or Sox6 icKO mice were similar to those in control mice (32.0%± 2.4% and 37.8% ± 5.7%, respectively, versus 38.1% ± 4.1%; Figures 2E, S3A, and S3B), indicating that despite the loss of Sox5 or Sox6, the recombined RGL pool is maintained for at least 1 week.
We next explored whether the reduction in adult RGL activation upon Sox5 deletion was accompanied by a complementary increase in qRGLs.Dividing RGLs that return to quiescence can be identified by their capacity to incorporate BrdU during DNA replication and then to retain this thymidine analog for as long as they are quiescent (BrdU-long [BrdU-L] retaining RGLs).They can be labeled in P60 mice by five consecutive BrdU injections and, upon Sox5 inactivation, be chased for 30 days (Urba ´n et al., 2016; Figure 3B).At P90, the percentage of BrdU-L-retaining cells was larger in Sox5 icKO than in control mice (1.9% ± 0.2% versus 1.2% ± 0.07%; p = 0.020; Figures 3A and 3C).Moreover, Sox5 icKO mice showed twice as many quiescent BrdU-L/rGFAP + cells as control mice (3.2% ± 0.5% versus 1.7% ± 0.3%; p = 0.029; Figures 3A and 3C).This pool of BrdU-L RGLs probably corresponds to resting RGLs with shallow quiescence (those more prone to resume cell cycle entry), in contrast with dormant RGLs in a deeper state of quiescent (Harris et al., 2021).Our data indicate that in the absence of Sox5, those resting RGLs show impaired ability to re-enter the cell cycle and remain quiescent.In summary, these data reinforce the fact that Sox5 is required for RGL transition from quiescence to activation in the adult hippocampal SGZ.

Sox5 levels are regulated by BMP4 and Sox5 overexpression is sufficient to interfere with entry in quiescence in vitro
To further prove that Sox5 is required in adult RGLs for the transition from quiescence to activation, we used an in vitro model, in which hippocampus-derived NSCs from 6-week-old mice are grown as floating Sox5-and Sox2-positive neurospheres in the presence of the EGF and FGF2 mitogens (Figure 3D).
One of the best established signals that promote entry in the quiescent state is BMP4.This factor is active both in hippocampal RGLs in vivo (Bonaguidi et al., 2008;Mira et al., 2010) and in hippocampal NSCs in vitro (Blomfield et al., 2019;Martynoga et al., 2013;Mira et al., 2010).Hippocampal neurospheres were disaggregated and seeded as adherent cells and cultured with FGF2 alone (proliferating condition) or combined with BMP4 (quiescence condition; Figure 3F).As expected, 3 times as few NSCs incorporated BrdU when cultured with FGF2 plus BMP4 as with FGF2 alone (21.2% ± 3.2% and 6.9% ± 0.4%; p = 0.0016; Figure 3F).Moreover, to validate our approach by real-time quantitative PCR (qPCR), we observed an increase in transcriptional activation of BMP4 targets such as Id2 (2.6 ± 0.5-fold, p = 0.011), Id4 (13.8 ± 3.5-fold, p = 0.0073) (Blomfield et al., 2019), Bmpr1a (4.6 ± 0.9-fold, p = 0.0087), and Bmpr1b (6.2 ± 1.8-fold, p = 0.0406) in comparison with NSCs grown in FGF2 alone (Figure 3H).Interestingly, in BMP4-induced quiescent NSCs, Sox5 and Sox6 transcript levels were significantly reduced (0.52 ± 0.07-fold and 0.23 ± 0.04-fold, respectively; p = 0.0027 and p = 0.00004; Figure 3H) in comparison with values in proliferating conditions.Similarly, at the protein level, (G) Quantitation of Sox5 protein levels in NSCs in FGF2 and in FGF2 plus BMP4 conditions expressed in arbitrary units relative to the FGF2 condition.p = 0.0001 by Student's t test (n = 300 cells).(H) qPCR analysis for the indicated transcripts in NSCs grown with FGF2 plus BMP4 with respect to those grown with FGF2 alone.Results are shown as 2 ÀDDCt normalized with respect to two housekeeping genes (Pgk1 and Gapdh) and relative to values in cells grown in FGF2 alone (dashed line on y axis = 1).A total of three or four independent neurospheres cultures were used, and three technical replicates were performed.(I) Immunohistochemistry of nucleofected NSCs (YFP + cells; green), using either pCIG or pCIG-Sox5 vectors, to analyze proliferating Ki67 + cells in each condition.(J) Quantitation of the percentage of Ki67 + cells in each indicated condition.In all graphs, data are mean value ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001 by unpaired (C) or paired (H and J) Student's t test.Scale bars represent 50 mm (A), 30 mm (E), and 80 mm (F and I).
Sox5 expression was significantly decreased in cells in BMP4induced quiescence in comparison with actively proliferating cells (relative level 86.19 ± 1.28 versus 99.57± 1.20, p = 0.001; Figure 3G).These data suggest that the BMP4 signal, which strongly promotes the quiescent state in NSCs, inhibits Sox5 and Sox6 expression at the mRNA level and consequently at the protein level as well.
Having seen that loss of Sox5 blocks the activation of RGLs in vivo, we decided to check if Sox5 was sufficient to promote NSC activation in vitro.For that purpose, we used hippocampal NSCs and overexpressed Sox5 (via a pCIG-Sox5 construct) through nucleofection, using pCIG-EGFP nucleofected cells as control.In the FGF2 condition, we observed similar number of proliferative Ki67 + cells (among EGFP-expressing cells) in both control and Sox5-overexpressing cells (38.90% ± 5.13% versus 37.70% ± 0.71%, respectively; Figures 3I and 3J), suggesting that increased levels of Sox5 did not alter NSC proliferation.However, in the FGF2 plus BMP4 condition, cells overexpressing Sox5 were more proliferative than control cells (12.03% ± 1.21% versus 4.86% ± 0.09%, respectively; p = 0.0041; Figures 3I and 3J).Thus, our data suggest that Sox5 overexpression is sufficient to interfere with the entry of NSCs into BMP4-induced quiescence.
We further investigated adult neurogenesis in animals at 14 and 30 dpi to determine the mid-term effect of Sox5 and Sox6 loss in the generation of new neurons.Although the number of Tbr2 + cells did not change much in Sox5 icKO and Sox6 icKO mice at 14 and 30 dpi (Figure S4C), the proportion of Dcx + YFP + new neurons decreased by $18% in relation to control mice 14 dpi (42.6% ± 3.1% and 41.4% ± 2.3%, respectively, versus 51.3 ± 1.7% in control; p = 0.015 and p = 0.005; Figures 4C and S4C).The reduction in the generation of new neurons was maintained at 30 dpi, when we observed a $20% decrease in the fraction of immature Dcx + YFP + neurons in Sox5 icKO and Sox6 icKO mice compared with control mice (46.4% ± 2.5% [p = 0.004] and 47.8% ± 3.4% [p = 0.02], respectively, versus 58.0% ± 2.2% in control; Figures 4A-4C).Taken together, our data demonstrate that in the adult SGZ, Sox5 and Sox6 are both required for the generation of the correct number of new granular neurons.
However, in addition to neurogenesis, new astrocytes are continuously generated from RGLs in the adult DG (Bonaguidi et al., 2011;Encinas et al., 2011), and in some mutant mice, the astrogliogenic fate is promoted at the expenses of a neurogenic one (Bonzano et al., 2018).We found that as early as 7 dpi, Sox2-creER T2 -mediated recombination in control mice could be observed in S100 + Sox5 + astrocytes in the hilus (Hi) and stratum moleculare (Smol; Figure S5A) of the hippocampus.By 30 dpi, in both Sox5 icKO and Sox6 icKO mice, the number of S100 + Sox2 + astrocytes in the YFP + population in the SGZ was similar (0.75% ± 0.75% and 1.33% ± 0.45%; Figures 4D and  4E) to that in control mice (0.92% ± 0.43%).The number of astrocytes in the YFP + population recombined in Hi and Smol was also similar in Sox5 icKO and Sox6 icKO mice (67.4% ± 2.7% and 83.8% ± 3.9%, respectively; Figures 4D and 4E) with respect to control mice (74.3% ± 5.7%).Thus, in control, Sox5 icKO , and Sox6 icKO mice, the fraction of Sox2-creER T2 -mediated recombined YFP + cells that become astrocytes in the SGZ is very low ($1%), and Sox5 or Sox6 are probably not required for the generation of astrocytes in the DG area, at least in young animals.
Finally, to discard further the option that Sox5 and Sox6 promote astrogliogenesis, we used hippocampal NSCs in culture from Sox5 icKO and Sox6 icKO mice, administered 4-OH-TAM to induce creER T2 -mediated recombination, and then forced NSCs to differentiate upon mitogen withdrawal (Figure S5B).We used bIII-tubulin and GFAP immunostaining to characterize neurons and astrocytes, respectively.Loss of Sox5 induced higher neuronal differentiation in YFP + cells with respect to YFP À cells (31.1% ± 5.9% versus 12.4% ± 2.7%, p = 0.029; n=4 ;Figures S5C and S5D).In contrast, there was a slight, although not statistically significant, decrease in astrocytic differentiation (8.2% ± 1.9% in YFP + cells versus 12.7% ± 1.2% in YFP À cells, p = 0.09; Figures S5C and S5D).Similarly, loss of Sox6 promoted neuronal differentiation more frequently in YFP + than in YFP À cells (29.3% ± 4.7% versus 15.6% ± 1.6%, respectively; p = 0.03; n = 4; Figures S5C and S5E) but did not significantly affect astrocytic differentiation (3.7% ± 1.4% in YFP + cells versus 11.6% ± 3.5% in YFP À cells, p = 0.08; Figures S5C and S5E).Thus, these data suggest that in NSCs grown in differentiating conditions, Sox5 and Sox6 are probably not essential for astrocytic differentiation and that loss of Sox5 or Sox6 could promote neuronal cell fate.This fact also reinforces the idea that in vivo, the main cause of the reduction in neurogenesis in Sox5 icKO and Sox6 icKO mice is the loss of activation of RGLs, not an alteration in neuronal cell fate commitment or in neuronal differentiation.

Sox5 and Sox6 control Ascl1 expression in RGLs
One of the most relevant factors in the control of adult RGL activation is Ascl1, a bHLH transcription factor required for the exit of RGLs from quiescence (Andersen et al., 2014).To explore a possible molecular mechanism by which Sox5 or Sox6 could control RGL activation, we analyzed Ascl1 expression in the SGZ of Sox5 icKO and Sox6 icKO mice at 14 dpi.As previously mentioned, at 14 dpi the relative number of YFP + RGLs expressing rGFAP was similar in Sox5 icKO , Sox6 icKO , and control mice (Figures S3C, S5A, and S5B).However, we observed a clear reduction in the proportion of Ascl1 + rGFAP + cells in the rGFAP + YFP + population both in Sox5 icKO and Sox6 icKO mice (7.2% ± 2.1% [p = 0.017] and 10.6% ± 2.1% [p = 0.046], respectively) with respect to those in control mice (18.6% ± 2.8%; Figures 5A and 5B).These data indicate that Sox5 and Sox6 are required for the expression of Ascl1 in RGLs in the adult hippocampal SGZ.
Nevertheless, the loss of Ascl1 expression could be either the cause or the consequence of the loss of RGL activation in Sox5 icKO and Sox6 icKO mice.To explore the possibility that Sox5 and Sox6 could directly control Ascl1 expression, we examined putative regulatory sequences near the Ascl1 genomic locus.An evolutionarily conserved region of 1,304 bp, located 6.6 kb upstream of the human ASCL1 coding region, was detected when fish, frog, chicken, opossum, rodent, and monkey Ascl1 locus sequences were compared (hg19 chr12:103345375-103346678; ECR Genome Browser; Figure 5F).An equivalent mouse Ascl1 coding region drives central nervous system reporter gene expression in transgenic mice (Verma-Kurvari et al., 1998).
Using transcriptional binding sites analysis (Regulatory VISTA, Mulan alignment, and bibliography mining), we uncovered three highly conserved putative Sox-binding sites (G/ A A/ G ACAA T/ A G/ A G/ C ) (Lee et al., 2017) within the human ASCL1 1,304 bp region (Figure 5F).First, we tested the ability of an equivalent mouse Ascl1 1,043 bp region (Enh-Ascl1-luciferase) to respond to a strong activator such as Sox2 in NSCs (Figure 5G).We observed a robust transcriptional activation in luciferase-based reporter assays of cells transfected with Sox2-pCIG with respect to control pCIG, an activity probably mediated through the conserved putative Sox-binding sites (11 ± 2.4-fold relative to pCIG1; p = 0.0004; Figure 5G).However, when we tested Sox5 and/or Sox6 factors, we observed a lot of transcriptional variation in each experiment, with half of the cases showing a moderate increase in luciferase activity driven by Enh-Ascl1 and the other half showing a reduction (Figure 5G).Thus, we concluded that the isolated 1 kb Ascl1 enhancer was not consistently activated by Sox5 or Sox6 in vitro.
For that reason, we resorted to an in vivo full-genome assay.We used a Sox6 antibody successfully described on chromatin immunoprecipitation (ChIP) (Lee et al., 2017) and performed ChIP assays in rat adult hippocampus NSCs, using an acetylated Histone3 antibody as a positive control.By real-time qPCR (Figure 5H), we found that Sox6 antibody specifically immunoprecipitated a 214 bp Ascl1 DNA fragment that comprises the three putative Sox-binding sites (0.0044 ± 0.0006 versus 0.0008 ± 0.0003 for IgG, expressed as percentage input; p = 0.0094; Figure 5H).In conclusion, SoxD factors activate Ascl1 expression in adult hippocampus NSCs, and at least Sox6 may act directly through evolutionary conserved Sox-binding sites in a neural Ascl1 enhancer.

Sox5 loss hinders RGL activation in response to environmental enrichment
Adult neurogenesis is dynamically regulated by a variety of physiological and pathological factors such as physical exercise, task learning, environmental enrichment (EE), and seizures (Rolando and Taylor, 2014).However, it is not always clear whether changes in neurogenesis are due to alterations in RGL activation or to adaptations in proliferation and survival of IPCs and immature neurons.We decided to use 8 days of exposure to EE stimuli, including a running wheel, as potent physiological stimuli of adult neurogenesis.Using Control Sox2-creER T2 /Rosa26-YFP mice injected for 5 days with TAM, we prepared groups of two males in basic cages under standard conditions (basic group) and groups of five males in enriched conditions, including larger cages with changing toys, tubes, running wheel, and bedding material (EE group; Figure 6B).At 15 dpi, we observed a clear increase in the number of YFP + RGLs relative to the YFP + population in EE mice in comparison with basic mice (54.3% ± 4.7% versus 32.1% ± 2.3%, respectively; p = 0.005; Figures 6A and  6C).Moreover, the percentage of activated RGLs (rGFAP + MCM2 + cells) was higher in EE than in basic mice (26.3% ± 1.5% versus 18.8% ± 2.7%, respectively; p = 0.046; Figures 6A and 6D).Thus, we have shown that 8 days of EE promotes RGL activation in the adult hippocampus.
Then we analyzed whether Sox5 icko animals could similarly respond to EE cues.The relative number of RGLs in recombined YFP + cells did not change in Sox5 icKO mice exposed either to the basic or the EE condition with respect to control mice (38.1% ± 11.11% versus 32.1% ± 4.7% in basic; 45.2% ± 5.3% versus 54.3% ± 8.0% in EE; Figures 6A and 6C).Regarding RGL activation, EE Sox5 icko mice did not reach the level of RGL activation observed in EE control mice (12.6% ± 0.9% versus 26.3% ± 1.5%, p = 0.001; Figures 6A and 6D).Nevertheless, EE Sox5 icko mice showed a tendency toward a 40% increase in RGLs activation with respect to Sox5 icko mice in basic cage conditions (12.6% ± 0.9% versus 8.9% ± 1.1%, p = 0.052; Figures 6A  and 6D).Using a two-way ANOVA, we observed two significant effects: a genotype effect that accounts for 52.02% of the total variance (p < 0.0001) and an environment effect accounting for 11.56% of the total variance (p = 0.0084).In summary, our data indicate that although RGLs in Sox5 icko mice show some response to EE stimuli, enriched Sox5 icko mice failed to reach the level of RGL activation observed in enriched control mice.

DISCUSSION
The potential to generate new neurons in spatially restricted brain areas during adulthood relies heavily on the regulation of the proliferative capacity of NSCs, a population predominantly in a state of reversible quiescence.Deciphering how RGLs, the resident adult NSC population, integrate physiological stimuli and local cues through their intrinsic transcriptional programs to abandon quiescence, and adjust the production of mature neurons, is essential to understand adult neurogenesis.Here, we show that SoxD proteins (Sox5 and Sox6) are required for RGL activation in the hippocampal neurogenic niche, that quiescent signals such as BMP4 modulate their transcription, and that Sox5 mutants do not exhibit the full RGL activation response to physiological neurogenic stimuli such as EE.Furthermore, Sox5 and Sox6 activate the expression of the proneural Ascl1 gene, which encodes a factor critical in adult RGL activation (Andersen et al., 2014).In summary, SoxD genes emerge as key factors for the control of adult neurogenesis.
To begin with, we have established that SoxD factors are crucial in adult RGLs to transit from a quiescent to an activated proliferative state.Loss of SoxD factors in adult RGLs, after just 7 days, reduced by more than half the ability of quiescent RGLs to engage in cell cycle progression.Conversely, increasing Sox5 levels hinders the cells ability to enter quiescence when they receive such pro-quiescent signals as BMP4.Moreover, the defect in RGL proliferation caused by SoxD loss is robust, as it lasts at least 1 month after gene deletion.We reinforced that idea by demonstrating that the RGLs that have recently undergone a round of cell division, upon losing Sox5, are unable to re-enter the cell cycle and thus remain in a quiescent state.This could also indicate that losing Sox5 increases the number of resting RGLs, which are those RGLs in a shallow quiescent state, more prone to engage in cell division (Harris et al., 2021;Urba ´n et al., 2016).However, without Sox5, those resting RGLs could not engage in cell division.
The importance of the fine control of the activation of RGLs is stressed by the fact that only 7 days after the loss of SoxD factors, there is an immediate impact on neurogenesis, observed by a reduced number in immature neurons, which do not seem to be compensated by IPC overproliferation.Furthermore, the effect is maintained for at least 1 month, when there is still a sustained reduction in new neurons in both Sox5 icKO and Sox6 icKO mice.In any case, SoxD function in RGLs does not appear to be related to cell-fate decisions, as RGLs in Sox5 icKO and Sox6 icKO mice, incapable of entering the cell cycle, do not opt for direct differentiation into astrocytes.
The SoxD family members Sox5 and Sox6 are closely related DNA-binding proteins, sharing 87% identity and acting as homodimers and heterodimers (Lefebvre, 2010).They share several common target genes (Lee et al., 2017;Liu et al., 2017) and have redundant functions when they are co-expressed in the same cell type (Stolt et al., 2006).Now we have shown that Sox5 and Sox6 share patterns of expression in the adult neurogenic niche and that both are required for RGL activation and for adult neurogenesis.
Although significant progress has been made in describing extrinsic and intrinsic cues that control RGL behavior, relatively few advances have been made in understanding how cells integrate that information.We have now shown how quiescencepromoting signals such as BMP4 (Martynoga et al., 2013;Mira et al., 2010), at least in NSCs in vitro, downregulate Sox5 and Sox6.Ascl1 protein levels are also strongly reduced in BMP4treated quiescent NSCs (Blomfield et al., 2019), and both Id4 transcription factor and E3 ubiquitin ligase Huwe1 (Urba ´n et al., 2016) regulate Ascl1 at the protein level.However, the mechanism controlling Ascl1 transcription in vivo in RGLs is not fully clear, although it involves repression by Notch signaling (Imayoshi et al., 2013;Sueda et al., 2019).Now we have unveiled another layer of Ascl1 regulation showing that Sox5 and Sox6 modulate Ascl1 expression in RGLs in vivo and that at least Sox6 may control Ascl1 transcription through an evolutionary conserved and transcriptionally active neural enhancer in Ascl1 locus.
Finally, a particularly important feature of hippocampal neurogenesis is its regulation by a variety of physiological stimuli such as EE (Rolando and Taylor, 2014).In the case of EE, acute EE stimulus could induce general cell proliferation (Chandler et al., 2020;Steiner et al., 2008), and long-term EE affects only cell survival, not proliferation (Kempermann et al., 1997;van Praag et al., 1999).We have now demonstrated that qRGLs respond to an 8 day EE paradigm entering in an active proliferative state.More important, Sox5 loss hinders the RGL response to EE physiological cues.
In conclusion, we have demonstrated how SoxD family members are required for RGL activation modulating Ascl1 expression and for the generation of new neurons in the adult hippocampus.Understanding the molecular mechanisms controlling adult NSC activation will open new approaches for regenerative medicine and treatment of brain disorders.Moreover, given the fact that SOX5 and SOX6 heterozygous inactivating variants cause neurodevelopmental diseases in humans (Lamb-Shaffer and Tolchin-Le Caignec syndromes, respectively), our findings will also shed light on the neural alterations present in those SOXopathies (Angelozzi and Lefebvre, 2019).
Limitations to the study As we achieved only partial loss of Sox5 or Sox6 in our conditional mice, we are aware of the very likely dilution of the observed phenotype due to inefficient recombination of SoxD genes.
As we did not have a good Sox5 antibody for immunoprecipitation, we could not establish Sox5 direct binding to the site occupied by Sox6 in the Ascl1 enhancer.
There is a limitation in the statistical interpretation of EE results, as there is an imbalance between Sox5 mutants in control (n = 9) versus EE (n = 3) conditions.Thus, the tendency observed in Sox5 mutants toward 40% RGL activation upon EE could be of statistical significance (p = 0.052).

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

Article
For the ChIP assays we used rat Adult Hippocampal Neural Stem and Progenitor Cells that were maintained as neurospheres in DMEM F-12 (Gibco, Thermo Fisher) adding N2 Supplement (1x) with 20 ng/ml of FGF-2 (PeproTech).

METHODS DETAILS
In vivo treatments For activation of the creER T2 recombinase, 2-to 3-months-old animals (both Control and the indicated inducible mutant mice) were administered intraperitoneally (ip) a dose of 5 mg/40gr body weight of TAM (20 mg/ml in corn oil; Sigma) for 5 consecutive days.Animals were perfused 7, 14 or 30 days post TAM injection (dpi).To examine slowly dividing RGLs, mice received 5 BrdU injections in a three days period and were sacrificed 30 days later.
For immunostaining, vibratome floating brain sections were permeabilized with 1% Triton X-100 in 0.1M Phosphate Buffer (PB) for 30 min and blocked with 10% Fetal Bovine Serum (FBS) and 0.25% Triton X-100 in 0.1M PB for 2 h at room temperature with rocking.For fixed cells, permeabilization and blocking were performed with 10% FBS in 0.25% Triton X-100 in 0.1M PB for 1 h at room temperature.Primary antibodies were prepared in incubation buffer (1% FBS, 0.25% Triton X-100 in 0.1M PB) and incubated with sections or cells overnight at 4 C. Following 3 washes, with washing buffer (0.1% Triton X-100 in 0.1M PB), sections or cells were incubated with secondary antibodies (1:1000) in incubation buffer for 2 h at room temperature.After three washes, sections or cells were incubated with bisbenzimide (1:100 in 0.1M PB) for 2 min at room temperature, and mounted using Fluoromount-G (Thermo Fisher).
To detect MCM2 protein, antigen retrieval was carried out by incubating sections in 0.15M sodium citrate at 80 C for 30min.To detect BrdU incorporation, DNA was denatured by incubating sections or cells with 2N HCL for 25min at room temperature, followed by 0.15M boric acid neutralization for 20min.When MCM2 or BrdU staining was performed combined with EGFP immunostaining, the latter (including primary and secondary antibodies) was done prior to citrate or acid treatments.

Primary cell culture nucleofection
Neurospheres were dissociated and grown as adherent cells in poly-L-ornithine (1.5 mg/ml; Sigma) and fibronectin (1 mg/ml; Invitrogen) coverslides.To induce quiescence, cells were plated at a density of 80,000 cells/MW24 well in basal media with 20 ng/ml FGF2 and after 24 h, media was replaced with basal media with 20 ng/ml FGF2 alone or in combination with with 30 ng/mL recombinant mouse BMP4 (100 ng/ ml; PeproTech).Cells were cultured for 72 h and then fixed with 4% PFA (Merck) on ice for 20min.Brdu (10 mg/ml; Roche) was added to the media for 1 h before fixation.
Sox5 coding sequence inserted into a pCIG vector containing EGFP (pCIG-Sox5) has been described before (Quiroga et al., 2015).As a control, empty pCIG vector (pCIG-EGFP) was used.2.5 3 10 6 cells from dissociated neurospheres derived from 6-weeks old wild type mice hippocampus were nucleofected with DNA plasmids at a concentration of 2.5 mg/ml using the P3 Primary Cell 4D-Nucleofector TM X Kit (Lonza, V4XP-3012) and the program #113, based on the manufacturer's instructions.After nucleofected, cells were cultured at a density of 80,000-150,000 cells/MW24 well in adherent conditions as described above.
DNA overexpression assays in Figure 5C were performed by Effectene Transfection Reagent (Qiagen) in adherent adult hippocampal NSCs in MW6 plates (280,000 cells/well) coated with polyornithine and fibronectin.Following Qiagen instructions, 3 mg of DNA plasmid was first mixed with Enhancer solution in a buffer that provides optimal salt conditions for efficient DNA condensation during 5 min.Effectene Reagent was then added and incubated for 10 min at RT.The complexes were mixed with neurospheres growth medium and added directly to adherent NSCs during 5 h.After removing the transfecting reagent cells were incubated for 24 h and then collected for total RNA extraction.

Lentiviral transduction
Adult mouse hippocampal NSCs were plated at a concentration of 60,000 cells/well in MW6 and 24 h after were transduced with lentiviral particles (5 units of multiplicity of infection; MOI) expressing three different Sox5 mouse shRNA (A to C; Locus ID, 20678; Cat# TL502111V; Origene) or one standard control scrambled shRNA (Origene).All the constructs express EGFP as a reporter.After adding lentiviral particles, plates were centrifuged for 1.5 h at 1050 rpm at RT and then viral transduction took place for 3-4 days.Cells were collected and total RNA was extracted.

Luciferase reporter assay
An evolutionarily conserved region of 1304 bp, located 6.6 kb upstream of the human ASCL1 coding region, was detected (hg19 chr12: 103,345,375-103,346,678;ATG in 103,352,023;TSS 103351452).An equivalent 1043 bp fragment of the mouse Ascl1 genomic locus (positions À8479 to À7446 from coding sequence) was cloned upstream a minimal CMV promoter-luciferase reporter gene (Vector Builder Company; Enh-Ascl1-Luciferase).Adherent NSCS cultured in MW12 plates (140,000 cells/well) were transfected using Effectene Transfection Reagent (Qiagen) with 0.5 mg Enh-Ascl1-Luciferase and a combination of 1 mg of Sox5 and/or e4 Cell Reports 38, 110313, February 1, 2022 Sox6 or Sox2 expressing constructs in vector pCIG, in combination with 50 ng of SV40-Renilla and 10 ng and CMV-Renilla constructs for normalization.After 24 h, cells were harvest and luciferase activities were measured by Dual Luciferase Reporter Assay System (Promega).
RNA extraction, cDNA synthesis and real-time quantitative PCR Cultured hippocampal NSCs (100,000cells/MW24 well) grown in adherent conditions with FGF2 or with FGF2 plus BMP4 were collected after three days of treatment and stored at À80 C. Total RNA was extracted using the QuickGene RNA tissue kit S (Kurabo) and then treated with DNAse.cDNA was synthesized with SuperScriptTM IV First-Stand Synthesis System and random hexamers (Invitrogen).Gene expression levels were measured using TaqMan Gene expression assays (Applied Biosystems) and real-time quantitative PCR (qPCR) was carried out in a 7500 PCR System using TaqMan Fast Advanced Master Mix (Applied Biosystems).Gene expression was measured relative to endogenous controls Pgk1 and Gapdh, and normalized to the expression of the Control sample in each group using the 2 ^(-DDCt) method, as indicated in the corresponding figure.At least three independent experiments were performed for each condition and samples were run in triplicates.
Chromatin immunoprecipitation (ChIP) assay 6 3 10 6 NSCs from the adult rat hippocampus (for each ChIP reaction) were cross-linked in medium containing 1% formaldehyde for 10 min at 37 C and neutralized with 0.137 M glycine for 5 min at room temperature.Next, cells were scraped off and rinsed twice with 10 ml of cold PBS containing protease inhibitors.Cells were pelleted by centrifugation for 4 min at 5000 rpm at 4 C. Cell pellets were suspended in SDS Lysis buffer (1% SDS, 1 mM EDTA, 50 mM Tris pH 8.1) and incubated for 10 min at 4 C. Lysates were sonicated using a Bioruptor Sonicator (Diagenode, BioRuptorTM UCD-200) and cellular debris was removed by spinning at 13,000 rpm for 10 min at 4 C.The supernatant fraction was diluted 5-fold in ChIP Dilution Buffer and chromatin solution was incubated overnight using Anti-acetyl-Histone H3, Anti-Sox6 and Anti-rabbit IgG after 1 h of preclearing at 4 C.The immune-complexes were incubated with Protein A agarose/Salmon Sperm DNA (Millipore), washed and eluted in elution buffer (1% SDS/0.1 M NaHCO3).Next, the cross-linking was reversed at 65 C overnight with gentle shaking.The DNA was purified by phenol-chloroform extraction followed by ethanol precipitation and was recovered in 20 ml of DNAse free water.
ChIP quantification was performed by quantitative PCR using SYBR PremixEX Taq (2x; Takara) amplifying a 214 bp fragment of the rat Ascl1 gene.ChIP-qPCR data was analyzed by % Input method based on Ct value normalized by the amount of chromatin input.The value for each sample is calculated as follows: % Input = 100*2 ^[Adjusted input -Ct (IP)].Primer set details are in key resources table.

Environmental enrichment
We used an EE protocol involving classical toys and objects, and a running wheel (Kempermann et al., 1997).Control mice were housed in groups of two in transparent Plexiglas cages (20 3 22 3 20 cm) under standard laboratory conditions.Mice subjected to EE were housed in groups of five animals in large transparent Polycarbonate cages (55 3 33 3 20 cm, Plexx Ref. 13,005).All enriched cages were equipped with different types of running wheels; different bedding material (newspapers, sheets of paper, sawdust and cloth); toys used included plastic non-edible ones as hard plastic toys, small plastic animals, gumabones (Plexx Ref. 13,110), tips, pet balls, a transparent polycarbonate mouse igloo (Plexx Ref. 13,100), polycarbonate rodent tunnels (Plexx Ref. 13,102) and others.Every other day, a set of 10-20 different toys and new bedding material was introduced into the cages and in general, the complexity of the cage environment was changed completely.

QUANTIFICATION AND STATISTICAL ANALYSIS Microscopic analysis and cell counting
All images were taken with a direct SP5 confocal microscope (Leica).Images of both left and right dorsal DG sections (À0.82 mm to À4.16 mm from bregma) were captured with a z-step of 2 mm through at least 20 mm of each 50 mm sections.Labeled cells were counted in the SGZ of every ninth of 50 mm DG sections.In wild type and Sox2-EGFP mice the analysis was done counting the number of cells (at least 300 cells/marker) that expressed a cell-type specific maker in the population of Sox5 + or Sox6 + cells and number of Sox5 + cells among SGZ cells expressing a certain cell-type marker.In Control, single mutants Sox5 icKO or Sox6 icKO and in double mutant mice we counted recombined YFP + cells that were positive for the indicated marker.In those cases, 3 to 6 sections from at least 3 to 7 mice and a minimum of 100 cells for each animal were analyzed.Counting was performed manually and blind using LAS X (Leica) software.In order to measure Sox5 immunofluorescence intensity in the RGL population, the nucleus of each RGL (characterized by YFP recombination with a radial process) was manually outlined according to bisbenzimide staining, and the average intensity of Sox5 immunostaining was measured using FIJI software.
For the experiments with adherent adult hippocampal NSCs grown in FGF2 or BMP4, six random regions in each of, at least, three coverslips for each experimental situation were photographed with a z-step of 2 mm by 40X objective.For the experiments of TAMinduced and nucleofected cells, at least 100 recombined or transfected cells were photographed with the same conditions.In all the cultures, three independent experiments were performed for each condition.For in vitro measurement of immunofluorescence intensity, an ImageJ macro was designed to first identify all the nuclei and then measure the average intensity of the selected channel for the area of each nucleus automatically, using ImageJ software.To quantify proliferation in cultured cells, ImageJ macros were designed to measure the percentage of proliferating cells (BrdU + or Ki67 + cells) among identified nucleus using ImageJ software.Differentiation markers in cultured cells after TAM administration were quantified manually.At least 100 EGFP positive cells were evaluated among EGFP + cells.

Statistical analysis
The appropriate sample size (N) was determined based on similar published data from other groups (Andersen et al., 2014;Bonzano et al., 2018), using a minimum of 3 mice per condition for in vivo experiments, and a minimum of biological triplicate for in vitro experiments.Statistical analysis and graphs were conducted with GraphPad Prism version 5 software using different tests with a significance level of at least p<0.05.Thus, two-tailed unpaired Student's t test were used for most of statistical comparisons of two conditions for in vivo experiments; paired t test for in vitro experiments where control and treatment conditions for each biological replicate were performed in parallel, in Figures 3; 5 and S5.Finally, two-way ANOVA was used in Figure 6

Figure 3 .
Figure 3. Loss of Sox5 favors quiescent state in RGLs and BMP4 inhibit Sox5 expression (A) Confocal images showing immunostaining for YFP, BrdU, and GFAP of control and Sox5 icKO mice brain sections at P90 at 30 dpi.Yellow arrows indicate recombined RGL cells that retain BrdU after 1 month (YFP + BrdU + rGFAP + cells).(B) Experimental design for BrdU label retention experiment.(C) Quantification of BrdU + cells among YFP + recombined cells and BrdU + rGFAP + YFP + cells among rGFAP + YFP + cells.Each symbol represents an independent biological replicate.(D) Scheme of adult NSC preparation grown as neurospheres.(E) Immunohistochemistry in hippocampal neurospheres for Sox2 and Sox5.(F) Neurospheres dissociated and seeded as attached cells and cultured with FGF2 alone or in combination with BMP4.A 40 min BrdU pulse was given before fixation, and BrdU incorporation was analyzed using immunohistochemistry.(G) Quantitation of Sox5 protein levels in NSCs in FGF2 and in FGF2 plus BMP4 conditions expressed in arbitrary units relative to the FGF2 condition.p = 0.0001 by Student's t test (n = 300 cells).(H) qPCR analysis for the indicated transcripts in NSCs grown with FGF2 plus BMP4 with respect to those grown with FGF2 alone.Results are shown as 2 ÀDDCt normalized with respect to two housekeeping genes (Pgk1 and Gapdh) and relative to values in cells grown in FGF2 alone (dashed line on y axis = 1).A total of three or four independent neurospheres cultures were used, and three technical replicates were performed.(I) Immunohistochemistry of nucleofected NSCs (YFP + cells; green), using either pCIG or pCIG-Sox5 vectors, to analyze proliferating Ki67 + cells in each condition.(J) Quantitation of the percentage of Ki67 + cells in each indicated condition.In all graphs, data are mean value ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001 by unpaired (C) or paired (H and J) Student's t test.Scale bars represent 50 mm (A), 30 mm (E), and 80 mm (F and I).

Figure 5 .
Figure 5. SoxD factors control Ascl1 expression in RGLs and NSCs of the hippocampal neurogenic niche (A) Confocal image showing immunostaining for YFP, Ascl1, and GFAP in the SGZ of 3-month-old control, Sox5 icKO , and Sox6 icKO mice at 14 dpi.Cyan arrows indicate recombined Ascl1 + rGFAP À cells, white arrows rGFAP + cells, and yellow arrows double rGFAP + Ascl1 + cells.(B) Quantitation of the number of rGFAP + in the population of YFP + cells and of rGFAP + Ascl1 + cells in the rGFAP + YFP + group in the indicated mice.(C) Quantitation in Sox5 icKO mice of the proportion of YFP + RGLs maintaining or losing Sox5 expression (Sox5 + or Sox5 À , respectively).The percentage of Ascl1 + cells in the subset of Sox5 + /YFP + or Sox5 À /YFP + RGLs is indicated.(D) Transcription levels for the indicated genes analyzed using qPCR in hippocampal NSCs transfected with Sox5-pCIG or Sox6-pCIG, represented as 2 ÀDDCt normalized with respect to the housekeeping gene Gapdh and relative to values in NSCs transfected with control pCIG (dashed line on y axis = 1).(E) Representative images of NSCs transduced with lentiviral particles expressing shRNA-control or shRNA-Sox5-A and GFP as a reporter protein.Graphs showing transcriptional levels by qPCR for each lentiviral transduced condition.Data are represented as 2 ÀDDCt normalized with respect to the housekeeping gene Gapdh and relative to values in NSCs transfected with shRNA-control.(F) Genomic sequence alignment comparing human ASCL1 locus with homologous D. rerio, G. gallus, X. tropicalis, M. domestica, R. norvegicus, M. musculus, and M. rhesus Ascl1 loci.Presence of conserved Sox-binding sites (red squares) within a 1,304 bp conserved enhancer, 6.6 kb upstream of the human ASCL1 coding region.(G) Quantitative analysis of transcriptional activity of Sox5, Sox6, and Sox2 on a 1 kb mouse Ascl1 enhancer in adult NSCs.Graph shows normalized luciferase activity units relative to the pCIG control.Dots in bars correspond to independent experiments.(H) Quantitation of three independent experiments of ChIP assays in rat adult hippocampal NSCs.DNA fragments for Ascl1 immunoprecipitated with Ac-H3, IgG, or anti-SOX6 antibodies were analyzed using qPCR with primers specific for a 214 bp fragment containing Sox-binding sites depicted in (F).Data are expressed as percentage of input = 100 3 2 [adjusted input À Ct (IP)] .Results are shown as mean value ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001 by unpaired Student's t test.Scale bars, 25 mm (A) and 50 mm (D).See also Figure S6.

Figure 6 .
Figure 6.Sox5 loss hinders RGL activation in response to environmental enrichment (A) Confocal images showing YFP, MCM2, and GFAP at 14 dpi in control and in Sox5 icKO mice exposed to conventional (basic) or enriched (EE) environmental conditions.White arrows indicate YFP + rGFAP + MCM2 À cells and yellow arrows YFP + rGFAP + MCM2 + cells.(B) Experimental design for control and Sox5 icKO mice in which TAM i.p. injection was performed at P60 on 5 consecutive days, and after 7 days animals were exposed to an enriched environment for 8 days, changing objects every other day.(C) Quantitation of the number of radial GFAP in YFP + cells in control and Sox5 icKO mice.(D) Quantitation of the number of rGFAP + MCM2 + in rGFAP + YFP + cells.Results are shown as mean value ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001 by unpaired Student's t test.Scale bar, 25 mm.

TABLE
d RESOURCE AVAILABILITY B Lead contact B Materials availability B Data and code availability d EXPERIMENTAL MODEL AND SUBJECT DETAILS 10 Cell Reports 38, 110313, February 1, 2022 e2 Cell Reports 38, 110313, February 1, 2022