Ars2 maintains neural stem-cell identity through direct transcriptional activation of Sox2.

Fundamental questions remain unanswered about the transcriptional networks that control the identity and self-renewal of neural stem cells (NSCs), a specialized subset of astroglial cells that are endowed with stem properties and neurogenic capacity. Here we report that the zinc finger protein Ars2 (arsenite-resistance protein 2; also known as Srrt) is expressed by adult NSCs from the subventricular zone (SVZ) of mice, and that selective knockdown of Ars2 in cells expressing glial fibrillary acidic protein within the adult SVZ depletes the number of NSCs and their neurogenic capacity. These phenotypes are recapitulated in the postnatal SVZ of hGFAP-cre::Ars2(fl/fl) conditional knockout mice, but are more severe. Ex vivo assays show that Ars2 is necessary and sufficient to promote NSC self-renewal, and that it does so by positively regulating the expression of Sox2. Although plant and animal orthologues of Ars2 are known for their conserved roles in microRNA biogenesis, we unexpectedly observed that Ars2 retains its capacity to promote self-renewal in Drosha and Dicer1 knockout NSCs. Instead, chromatin immunoprecipitation revealed that Ars2 binds a specific region within the 6-kilobase NSC enhancer of Sox2. This association is RNA-independent, and the region that is bound is required for Ars2-mediated activation of Sox2. We used gel-shift analysis to refine the Sox2 region bound by Ars2 to a specific conserved DNA sequence. The importance of Sox2 as a critical downstream effector is shown by its ability to restore the self-renewal and multipotency defects of Ars2 knockout NSCs. Our findings reveal Ars2 as a new transcription factor that controls the multipotent progenitor state of NSCs through direct activation of the pluripotency factor Sox2.

Stem cells reside in most mammalian tissues throughout adult life, and contribute to normal homeostasis and repair after injury[6]. They are defined by their capacity to both self-renew and differentiate, thus perpetuating themselves whilst generating more committed daughter cells. Two major stem cell niches exist in the adult brain, within the hippocampus and the subventricular zone (SVZ). Relatively quiescent neural stem cells (NSCs) give rise to actively proliferating transit-amplifying progenitors (TAPs), which generate oligodendrocytes destined to the corpus callosum[7] and neuroblasts (NBs) that migrate rostrally and differentiate into local interneurons in the olfactory bulb (OB)[9,10]. Much remains to be understood about the mechanisms and factors that control NSC self-renewal and multipotency[13].

Mammalian Ars2 was reported as essential for cell proliferation, to be downregulated in quiescent cells, and required for accumulation of several miRNAs implicated in cellular transformation[4]. Unexpectedly, we observed that Ars2 expression in the adult SVZ did not correlate with proliferation, since 95±2% of Ars2+ cells lacked the proliferative marker Ki67. Moreover, Ars2 was present in only 7±2% of Mash1+ TAPs (Supplementary Figure 1b) and was absent from Doublecortin+ (DCX) NBs (Figure 1a); these comprise the most highly proliferative cells in the SVZ. Ars2 was also absent from GFAP+ Nestin− Sox2− astroglial cells and S100β+ mature astrocytes (Supplementary Figure 1). Instead, Ars2 was expressed by niche astrocytes, ependymal cells and by GFAP+ CD133+ stem cells[14] (Supplementary Figure 1). A hallmark of NSCs is their quiescence, reflected by their ability to retain S-phase labels such as 5-chlorodeoxyuridine (CldU) for extended periods (i.e., label retaining cells, LRCs)[15,16]. We observed expression of Ars2 in 87±3% of LRCs marked one month earlier (Figure 1a), demonstrating presence of Ars2 in this slow dividing population in vivo.

Figure 1

Ars2 maintains neural stem cells in the adult SVZ

In the adult SVZ, Ars2 (red) is not expressed by DCX+ (green) neuroblasts (a), but colocalizes with GFAP+ (green), CldU-retaining (blue) neural stem cells (NSCs) (b, b’). (c) Experimental protocol. (d) Validation of Ars2 knockdown (n=3 animals per condition). (e) Ars2 knockdown SVZs exhibited reduced GFAP+ Nestin+ NSCs, reduced NSC proliferation (f), and increased S100ß+ mature astrocytes (g); n=3 animals per condition. (h) Ars2 knockdown reduced label-retaining cells (LRCs) (n=3–7 animals per condition). (i) OB sections stained for GFP (green), CldU (red) and DAPI (blue). (j) Percentage of GFP+ cells per OB slice (n=7–11 animals per condition). (k) Reduced newborn neurons after Ars2 knockdown (n=7-11 animals per condition). (l) Immunocytochemistry for Ars2 (red) and DAPI (aqua) (top) and GFP+ transduced primary neurospheres (bottom). (m) Longterm self-renewal assay (n= 4 cultures per condition). (n) In vivo Ars2 overexpression increased neurosphere formation (n= 4 cultures per condition). (o) Ars2 deficiency reduced NSC multipotency (n= 4 cultures per condition). *, P<0.05, **, P<0.01, ***, P<0.005. Errors bars show s.e.m. Table S2 provides details of numbers of scored cells. Scale bars: a, b, 10 μm; j, 100 μm; m, 200 μm.

To assay roles of Ars2 in NSCs in vivo, we used shRNAs that suppressed endogenous Ars2 (Supplementary Figure 2c). We packaged these into GFP-expressing Mokola lentivirus, which specifically transduces astroglial cells[17] (Supplementary Figure 3). We injected these into the adult SVZ and sacrificed mice 48 hours, 5 days or 1 month later (Figure 2c). At 48 hours post-infection, shArs2-GFP+ cells exhibited 80% reduction in Ars2 mRNA relative to shControl cells (Figure 2d). Apoptosis was unaffected by shArs2, and the number of GFP+ Ki67+ cells and levels of CyclinD1 or CyclinE transcripts were also unchanged (Supplementary Figures 3, 4b, 5b). However, 5 days post-infection, shArs2 SVZs exhibited 50% reduction in the number and the proliferation rate of the GFAP+ Nestin+ NSCs (Figure 1e, f), still without change in apoptosis (Supplementary Figure 5b). Loss of NSC potential has been linked to an increase in mature astrocytes[18-20]. Accordingly, we observed a 50% increase in the number of GFP+ S100β+ cells (Figure 1g).

Figure 2

Ars2 regulates postnatal neurogenesis

(a, b) Coronal sections from P15 SVZ and OB of control and Ars2Δ/Δ mice stained with hematoxylin-eosin. SVZ deletion of Ars2 depleted NSCs and induced astrogliosis, as evidenced by loss of Sox2+ (red) and accumulation of GFAP+ (green) cells (c), loss of Nestin+ (red) cells (d), and increased S100ß+ (red) cells (e); DAPI is in blue. (f) Reduced number and proliferation of NSCs and elevation of mature astrocytes in Ars2Δ/Δ mice. (g) Strong reduction of primary neurospheres from GFAP/CD133-Cre::Ars2 electroporated pups. (h) Wholemount immunostaining for DCX (green) and Ki67 (red). (i) OB coronal sections immunostained for CB (green) and DAPI (blue). (j) Quantification of interneuron reduction in control and Ars2Δ/Δ mice. *, P<0.05, **, P<0.01, ***, P<0.005. Error bars show s.e.m from 6 animals per genotype/condition. Scale bars: a, b, 100 μm; c, d, e, 30 μm; f, g, 50 μm.

To assess LRCs, we injected shRNA-infected mice with CldU and sacrificed one month later. Strikingly, we observed ~50% decrease in transduced LRCs in shArs2 SVZs (Figure 1h), suggesting that Ars2 maintains the NSC pool. If true, this is expected to have downstream consequences on neurogenesis. Indeed, 5 days post-infection, we observed a decrease in DCX+ NBs (Supplementary Figure 6). LRCs also label post-mitotic cells that incorporated CldU just prior to cell cycle exit (such as differentiated cells and newborn OB interneurons). One month post-infection, the population of shArs2-GFP+, newly formed CldU+ OB interneurons was strongly reduced (Figure 1i-k).

We performed additional analysis using neurospheres derived from shRNA-infected SVZ. Long-term self-renewal assays revealed that depletion of Ars2 rapidly extinguished neurosphere cultures, indicating a defect in self-renewing divisions (Figure 1m). This defect was fully restored by an shRNA-resistant form of Ars2 (Figure 1m and Supplementary Figure 2d). Reciprocally, in vivo overexpression of Ars2 in wildtype mice increased neurosphere formation (Figure 1n). Multipotency of Ars2-deficient neurospheres was also affected, since the frequency of clones that generated ßIII-tubulin+ neurons and O4+ oligodendrocytes was decreased in favor of unipotent GFAP+ clones (Figure 1o). We conclude that Ars2 is required to maintain NSCs in a self-renewing and multipotent state.

We sought to confirm these shRNA results by breeding the conditional knockout allele of Ars2 (Ars2) with hGFAP-Cre[21]. hGFAP-Cre::Ars2 (i.e. Ars2Δ/Δ) mice (Supplementary Figure 7d) were born at the expected Mendelian ratios relative to wild-type and hGFAP-Cre::Ars2 littermates (used as controls). However, by postnatal day (P)15, Ars2Δ/Δ mice showed progressive growth retardation, hydrocephalus and ataxia, resulting in death between P20 and P25. Further analysis of P15 Ars2Δ/Δ mice revealed enlarged ventricles and smaller olfactory bulbs (Figure 2a, b), suggestive of a requirement of Ars2 during postnatal neurogenesis. The expression pattern of Ars2 in P15 wild-type SVZ was analogous to the adult SVZ (Supplementary Figure 7a,b), and analysis of the conditional knockout confirmed essentially complete absence of Ars2 in the SVZ (Supplementary Figure 7c). Importantly, the number of NSCs (marked by expression of Nestin, Sox2, Lex, and GFAP) was reduced by 80% in Ars2Δ/ΔSVZ, and their proliferation rate decreased 2-fold (Figure 2c,d,f). This was not due to cell death as assessed by Caspase 3 staining (Supplementary Figure 8a,b). Conversely, we observed profound astrogliosis in Ars2Δ/Δ, as assessed by GFAP and S100ß staining (Figure 2e,f).

As Ars2 is expressed in niche astrocytes and ependymal cells, in addition to NSCs, we wished to demonstrate an autonomous function of Ars2 in NSCs. We co-injected GFP+ and Split-Cre plasmids that specifically drive excision in GFAP+ CD133+ NSCs[14] into the SVZ of P0-1 Ars2 pups, and introduced them using electroporation. Five days later, we isolated GFP+ cells and plated for self-renewal assay. GFAP+ CD133+ NSCs deleted for Ars2 were strongly compromised for neurosphere generation, demonstrating a cell-autonomous requirement of Ars2 in this population (Figure 2g and Supplementary Figure 9). Consistent with decreased NSC number, Ars2Δ/Δ mice exhibited ~2.5-fold fewer DCX+ NBs (Figure 2h), although their proliferation rate was not affected (data not shown). In the OB, the frequency of Tyrosine Hydroxylase+ (TH), Calbindin+ (CB) and Calretinin+ (CR) interneurons per glomerulus were also deeply reduced in Ars2Δ/Δ (Figure 2i,j and Supplementary Figure 10). In summary, the severe defects of postnatal SVZ deleted for Ars2 solidified its requirement to maintain NSC identity.

Since Ars2 functions in miRNA biogenesis[2-5], we tested if the ability of Ars2 to promote NSC self-renewal was mediated by miRNAs. This was not the case, since overexpression of Ars2 in DicerΔ/Δ and DroshaΔ/Δ cells increased neurosphere yield (Figure 3a,b). We sought further insight by examining transcription factors known to have substantial roles in NSC self-renewal, including Hes1, Hes5 and Sox2[22-24]. Sox2, but not Hes1 mRNA, was significantly reduced 48 hr after Ars2 knockdown in vivo (Figure 3c). One month post-infection, in vivo knockdown of Ars2 decreased Hes1 and Hes5 mRNA levels by ~30%, but resulted in a more substantial 70% reduction in Sox2 (Figure 3d). Reciprocally, Sox2 increased by ~60% in neurospheres overexpressing Ars2 (Figure 3e). These effects appeared to be transcriptional in nature, since Ars2 activated a 6kb Sox2-luc reporter containing cis-regulatory sequences responsible for Sox2 expression within adult neurogenic zones[25] (Figure 3f). This was not simply due to the cancelling effects of overexpressing any self-renewal gene, because in vitro overexpression of Sox2 (Figure 3g), but not Hes5 (Figure 3h) rescued the Ars2-dependent loss in self-renewal capacity of shArs2 cells. Reciprocally, in vitro knockdown of Sox2 (Supplementary Figure 11a) caused rapid depletion of self-renewing neurospheres[26] (Figure 3i) and compromised multipotency (Supplementary Figure 11b) similar to the effects of Ars2 knockdown (Figure 1m,o). Altogether, these data suggested Sox2 as a critical downstream effector of Ars2 in NSCs.

Figure 3

Ars2 acts independently of miRNA pathway to promote NSC self-renewal via Sox2

(a) Absence of Dicer and Drosha in floxed NSC cultures following Cre treatment (n=2 experiments per condition). (b) Ectopic Ars2 promoted NSC self-renewal in Dicer and Drosha-deleted NSCs (n=3 experiments per condition). (c) qPCR measurements from purified GFP+ cells 48 h post-infection (n=3). (d) qPCR measurements of self-renewal genes from primary infected GFP+ NSCs (n=3). (e) Overexpression of Ars2 increased Sox2 mRNA (n=3). (f) Both Ars2 and Sox2 activate a Sox2 transcriptional reporter (n=3). (g,h) Percentage of neurospheres formed from NSCs infected with the indicated constructs (n=3). (i) Longterm self-renewal assay shows that Sox2 is required for NSC self-renewing divisions (n=4). *, P<0.05, **, P<0.01, ***, P<0.005. Error bars show s.e.m; ns, not significant.

To evaluate whether the ability of Ars2 to activate Sox2 expression might reflect a transcriptional role for this nuclear protein, we performed chromatin immunoprecipitation (ChIP) of Ars2 in NSCs, querying across the 6kb Sox2 promoter and the Sox2 transcription unit. Interestingly, Ars2 associated not only with its 5' UTR and 3' UTR, but was highly enriched in region 8 (−2 to −2.5 kb) of the Sox2 promoter (Figure 4a); we validated this binding pattern using an independent antibody (Supplementary Figure 12a). RNase treatment of chromatin samples eliminated UTR-associated Ars2 ChIP signals, consistent with this reflecting CBC-mediated association with capped transcripts[4]. However, binding of Ars2 to promoter region 8 was maintained, suggesting here more direct association of Ars2 with chromatin (Figure 4b). No binding was found to the Sox2 coding region (Figure 4a), or to the promoters of Hes1, Hes5[27], K14, and Myod1 (Supplementary Figure 12b-e). Chromatin association of Ars2 was cell type-dependent, since Ars2 did not bind the Sox2 enhancer in NIH3T3 cells (Figure 4a), which express high levels of Ars2 (Supplementary Figure 2a).

Figure 4

Ars2 directly activates transcription of Sox2 to mediate NSC self-renewal and multipotency

ChIP of Ars2 to the Sox2 locus (n=4 independent experiments). (b) Ars2 binding to Sox2 promoter region 8 is RNA independent (n=2). (c) Ars2 requires region 8 to activate Sox2–luc (n=4). (d) Gel-shifts of the Ars2 binding site in the Sox2 promoter using nuclear extract or in vitro-translated (IVT) Ars2. (e) Conservation of the Ars2-binding site in Sox2. (f) Long-term self-renewal assay from NSCs isolated from control and Ars2Δ/Δ mice electroporated in vivo with the indicated constructs (n=6 animals per condition). (g) Ars2Δ/Δ defect in multipotency can be rescued by in vivo expression of Sox2 (n=6 animals per condition). (h-k) Differentiated NSC colonies stained for GFAP (blue), Tuj1 (yellow), and O4 (pink). *, P<0.05, ***, P<0.005. Error bars show s.e.m.

Simple binding of Ars2 to the Sox2 enhancer might not necessarily be of functional consequence. We prepared two deletions of the 6kb Sox2-luc reporter, removing the Ars2-bound region 8 or a control region. Loss of region 8 strongly reduced Sox2-luc expression relative to the control deletion, while reciprocally, ectopic Ars2 activated the control deletion but not the version lacking the Ars2 binding site (Figure 4c). Therefore, Ars2 activates Sox2 via promoter region 8. We then incubated NSC nuclear extract with a series of overlapping 90 bp radiolabeled probes covering the ~500 bp of Sox2 region 8, and observed a specific gel-shift of subregion #8-3 (Figure 4d). This band was completely super-shifted by inclusion of Ars2 antibody, but not Myc antibody (Figure 4d). We narrowed down the Ars2 binding site, which revealed that Ars2 bound specifically to the central portion of Sox2 region #8-3 (Figure 4d). This reflected direct DNA binding activity of Ars2, since in vitro translated Ars2 recapitulated specific binding to Sox2 probe #8-3b (Figure 4d). This identified a sequence that is highly constrained across mammalian genomes (Figure 4e and Supplementary Figure 13).

To determine whether Sox2 mediates Ars2 function in vivo, we electroporated Sox2 expression construct into Ars2Δ/Δ postnatalSVZ. Strikingly, Sox2 rescued the self-renewal and multipotency defects of Ars2 knockout cells (Figure 4f-j). In contrast, NSCs derived from Ars2Δ/Δ SVZ electroporated with CyclinD1 lacked self-renewal capacity (Figure 4f) and multipotency (Figure 4g,k). This confirmed that Ars2 knockout cells cannot be rescued by driving proliferation. Instead, Ars2 confers NSC identity as a self-renewing cell type by activating Sox2.

A central goal of stem cell biology is to understand the molecular mechanisms that regulate stem cell self-renewal and multipotency. We showed that Ars2 is specifically expressed by NSCs and not by TAPs and NBs, and that it maintains the self-renewal and multipotency capacity of postnatal and adult NSCs. In this setting, Ars2 is not required for cell viability, but is instead essential for maintaining core NSC properties. Ars2 depletion or knockout decreased the NSC pool, decreased neurogenesis and strongly increased non-neurogenic astrocytes. We assigned a new molecular function for the conserved RNA factor Ars2 as a sequence-specific DNA binding protein, and a critical direct activator of Sox2 during in vivo NSC self-renewal and multipotency. More generally, in light of excitement surrounding the role of Sox2 as a core pluripotency factor in ES and iPS cells, Ars2 may conceivably regulate stem cell self-renewal in these settings as well. This possibility is bolstered by the early embryonic arrest of Ars2 knockout mice[28], which bears substantial resemblance to the Sox2 knockout[29]. These connections deserve future investigation.