Persistent Sonic hedgehog Signaling in Adult Brain Determines Neural Stem Cell Positional Identity

Summary

Neural Stem Cells (NSCs) persist in the subventricular zone (SVZ) of the adult brain. Location within this germinal region determines the type of neuronal progeny NSCs generate, but the mechanism of adult NSC positional specification remains unknown. We show that sonic hedgehog (Shh) signaling, resulting in high gli1 levels, occurs in the ventral SVZ and is associated with the genesis of specific neuronal progeny. Shh is selectively produced by a small group of ventral forebrain neurons. Ablation of Shh decreases production of ventrally-derived neuron types, while ectopic activation of this pathway in dorsal NSCs respecifies their progeny to deep granule interneurons and calbindin-positive periglomerular cells. These results show that Shh is necessary and sufficient for the specification of adult ventral NSCs.

Introduction

Most of the neurons in the central nervous system are produced in the embryo, and for many years it was thought that this was the only source of neurons in the CNS. We now know that two populations of neurons continue to be produced in the adult: olfactory bulb interneurons and granule neurons in the dentate gyrus of the hippocampus. Neuronal production is limited to the subventricular zone (SVZ), along the walls of the lateral ventricles, and the subgranular zone, within the dentate gyrus. These two regions contain neural stem cells (NSCs) and continue to generate new neurons throughout adult life (Zhao et al., 2008) and Ihrie et al, in press). Within the SVZ, glial fibrillary acidic protein (GFAP)-expressing stem cells (type B cells) give rise to rapidly dividing transit-amplifying progeny (type C cells), which in turn generate immature neuroblasts (type A cells). These neuroblasts migrate to the olfactory bulb (OB) within a network of tangentially oriented chains that coalesce to form the rostral migratory stream (RMS) (Luskin, 1993; Lois and Alvarez-Buylla, 1994; Doetsch et al., 1999a; Doetsch et al., 1999b). Within the OB, young neurons migrate radially, complete their differentiation, and integrate into the granular and periglomerular layers.

In the mouse, the SVZ covers an area greater than six square millimeters along the rostrocaudal and dorsoventral axes (Mirzadeh et al., 2008). Neuroblasts derived from the SVZ traverse a significant distance to reach their final destination in the OB. Why are neurons derived from such an extended proliferative zone, and how does site of origin in the SVZ affect cell fate? One clue comes from recent experiments using viral or genetic lineage tracing to label specific subregions of the developing or adult SVZ (Kelsch et al., 2007; Kohwi et al., 2007; Merkle et al., 2007; Ventura and Goldman, 2007; Young et al., 2007). These results suggest that the SVZ is arranged as a mosaic; the position of stem cells within the SVZ determines the types of differentiated progeny generated. In particular, deep granule interneurons and a subpopulation of periglomerular cells arise from the ventral SVZ, while superficial granule interneurons and distinct periglomerular cells are derived from the dorsal SVZ. The molecular mechanisms responsible for this positional specification of primary progenitors in the adult SVZ are not known.

The Hedgehog (Hh) signaling pathway is central to the development and patterning of the nervous system and other organs (McMahon et al., 2003; Fuccillo et al., 2006). In mammals, this signaling pathway is initiated by one of three family members - Indian hedgehog (Ihh), Desert hedgehog (Dhh), or Sonic hedgehog (Shh). Secreted Hedgehog morphogen binds Patched at the cell surface, relieving its inhibition of the transmembrane protein Smoothened (Rohatgi et al., 2007). Smoothened triggers the activation of the Gli transcription factors. In the absence of Hh signal, the Gli3 transcription factor acts as a transcriptional repressor, while Gli2 functions primarily as a transcriptional activator upon Hh stimulation and can initiate transcription of Gli1, a constitutive transcriptional activator that indicates high levels of pathway activity (Ahn and Joyner, 2005; Palma et al., 2005; Clement et al., 2007).

The responsiveness of a cell to a given level of Hh ligand is also modulated by intrinsic expression of cell-surface proteins. The transmembrane proteins Cdo, Boc, and Gas1 are thought to positively regulate Hh pathway activation and allow cells at a greater distance from a Hh source to respond to lower levels within a Hh gradient (Tenzen et al., 2006; Allen et al., 2007; McLellan et al., 2008). Conversely, Hedgehog interacting protein (Hhip) binds and sequesters Hh ligand, and acts cooperatively with Patched as a negative-feedback mechanism to regulate pathway activation (Chuang and McMahon, 1999; Chuang et al., 2003). Studies in the developing neural tube have demonstrated that the varying levels of Hh ligand and interacting proteins present over time along the dorsoventral axis of this structure result in graded levels of activity of the Gli transcription factors, allowing a morphogen gradient to be translated into transcriptional control of neuronal identity (Jessell, 2000; Ribes and Briscoe, 2009). Shh signaling also regulates cell proliferation and fate in the developing forebrain and hindbrain (Machold et al., 2003; Corrales et al., 2004; Blaess et al., 2006; Balordi and Fishell, 2007a, b; Xu et al., 2010). Shh is important in the initial generation and proliferation of postnatal neural stem cells in the SVZ and DG, but whether it has a role in the specification of regional identity in the adult is not known (Lai et al., 2003; Balordi and Fishell, 2007b; Han et al., 2008).

Here, we demonstrate that Shh signaling has a continuous, critical role in the dorsoventral specification of adult SVZ neural stem cells. We find that despite the ubiquitous expression of Smoothened on stem cells throughout this region, Gli1 expression primarily occurs in the ventral SVZ, and is associated with particular neuronal fates. We show that Shh signaling is instructive in cell fate specification, and removal of Shh expression affects the generation of specific neuronal subtypes. Furthermore, we identify a possible source of secreted Sonic hedgehog (Shh) ligand close to the ventral SVZ: surprisingly, this source is neuronal. These results are the first identification of a signaling pathway that is sufficient to determine neuronal cell fate in adult SVZ NSCs.

Results

Gli1 Expression Occurs in the Ventral SVZ

Shh pathway members have been implicated in the development and postnatal maintenance of SVZ neural stem cells (Machold et al., 2003; Ahn and Joyner, 2005; Balordi and Fishell, 2007a, b). Remarkably, in situ hybridization for gli1, gli2, and gli3 revealed that gli1 expression is higher in the ventral SVZ (Figure 1A, 1D), while gli2 and gli3 are present both ventrally and dorsally (Figure 1B, C, E, F). Likewise, staining of brain sections from mice carrying gli1-nlacZ and ptc-lacZ reporter alleles (Goodrich et al., 1997; Bai et al., 2002) showed high levels of reporter expression in the ventral SVZ, in both the lateral and medial walls (Figure 1I, J). We also microdissected these regions from adult brains and performed qRT-PCR analysis. To confirm that the correct areas were dissected, we measured relative expression of the transcription factors Nkx2.1 and Nkx6.2, which are expressed in the ventral forebrain during development (Xu et al., 2008; Xu et al., 2010) and are present ventrally in the adult SVZ (L. Fuentealba, A. A–B., data not shown). Using qRT-PCR, we observed elevated gli1 expression in the ventral SVZ as well as the medial septum when compared to the dorsal SVZ (Figure 1K).

Figure 1. Gli1 Expression is Primarily Ventral.

(A–F) Dorsal and ventral areas shown in images A–F are indicated in the cartoon at left. In situ hybridization for the three Gli family members reveals that gli1 mRNA is present in the ventral SVZ (D), but not the dorsal SVZ (A). gli2 and gli3 mRNA are present at low levels in both the dorsal and ventral SVZ (B,C,E,F). Scale bars: 50 microns.

(G–J) LacZ staining in either Ptc-lacZ (G, I) or gli1-nlacZ (H, J) reporter strains is strongest in the ventral SVZ (I, J) and is reduced or absent in the dorsal SVZ (G, H). Red dashed lines indicate the lateral ventricle.

(K) qRT-PCR detection of Gli family members in microdissected ventral SVZ and septum recapitulates the expression observed in in situ hybridization. Values shown are relative transcript expression compared to microdissected dorsal SVZ. Nkx2.1 and Nkx6.2, which are known to be upregulated in ventral SVZ, are highly expressed in both ventral SVZ samples (blue bars) and RNA derived from the septum (red bars). gli1 is expressed more highly in the ventral SVZ and septum than in the dorsal SVZ. gli2 and gli3 transcript expression did not differ significantly between the dorsal and ventral SVZ, but both genes were expressed to higher levels in the septum. Asterisks = p < 0.05.

We next stained adult SVZ for Smoothened (Smo), an obligate component of the canonical Hh pathway. Smo protein was present throughout the SVZ in a pattern reminiscent of GFAP, which is expressed by type B cells in this region (Supplementary Figure 1A–F) (Doetsch et al., 1999a; Garcia et al., 2004; Tavazoie et al., 2008). Confocal analysis of both dorsal and ventral SVZ using two different antibodies demonstrated that Smo is expressed on a subset (~80%) of GFAP-positive cells in both subregions. This staining was not observed when the antibody was incubated with blocking peptide or when primary antibody was omitted, and was almost entirely absent in the brains of hGFAP::Cre; Smo ^fl/fl mice, where Smoothened is lost in most neural stem cells (Supplementary Figure 1G, 1H) (Han et al., 2008). Smo did not colocalize with Dcx, CD24, or EGFR, which label other cell types in the SVZ. To confirm that Smoothened is primarily expressed on stem cells, we infused the antimitotic cytosine-β-D-arabinofuranoside (Ara-C) into the brains of wild-type mice for 6 days. This treatment eliminates fast-dividing transit-amplifying (type C) cells and neuroblasts from the subventricular zone, while sparing slow-dividing stem cells (Doetsch et al., 1999b; Long et al., 2001). While a 6-day Ara-C infusion was sufficient to eliminate doublecortin-positive neuroblasts in wild-type mice, Smo expression remained after Ara-C treatment (Supplementary Figure 1J–M).

We next examined pathway activation using Gli1-CreER^T2; R26YFP mice (Ahn and Joyner, 2004), in which cells responding to high levels of Hedgehog ligand express a tamoxifen-inducible Cre recombinase. By giving tamoxifen at postnatal day 60 (P60) in adult mice, we identified SVZ stem cells in which the Hh pathway was active and followed their YFP-labelled progeny (Figure 2). We administered tamoxifen for 5 days, and examined YFP expression at 0 days, 5 days, or 30 days after the end of treatment. In contrast to the widespread expression of Smoothened, activation of Hh signaling was much more restricted. Intriguingly, pathway activity occurred in a spatially specific manner - the initial population of Hh-responsive cells was distributed along the anterior-posterior axis of this region, but largely confined to the ventral half of the SVZ (Figure 2D), an area which primarily produces deep granule interneurons within the OB (Merkle et al., 2007). This labeling was not due to limited infiltration of tamoxifen in the dorsal SVZ, as equivalent treatment of mice with ubiquitously expressed Cre-ER transgenes caused recombination throughout the SVZ ((Kuo et al., 2006) and data not shown).

Figure 2. High Hedgehog Pathway Activity is Limited to the Ventral SVZ.

(A–L) Lineage tracing of Gli1-expressing cells in the SVZ. Schematic at top indicates the experimental timeline. Immediately after 5 days of tamoxifen administration in Gli1CreER^T2; R26YFP animals, a small number of YFP/GFAP double-positive cells are present in the ventral SVZ (arrowhead and inset, D). YFP-labeled cells are absent from the dorsal SVZ, RMS, and OB (A, G, J). At 5 days after tamoxifen administration, YFP-labeled cells are present in the ventral SVZ, and some YFP labeling coincides with GFAP labeling (arrowheads and inset, E). YFP-labeled cells are also present in the dorsal SVZ (B), medial septum (right side in E), RMS (H), and the core of the OB, where doublecortin (Dcx)-expressing migrating neuroblasts enter this structure (K). One month after tamoxifen administration, there is widespread YFP labeling in the SVZ (C, F), RMS (I), and OB (L). Scale bars: 50 microns.

(M–O) Quantification of the localization of labeled granular layer neurons in Gli1CreER^T2; R26YFP animals at one month after tamoxifen administration (red bars, M). The neuronal progeny of labeled cells tend to localize in the deep OB granular layer close to the core, a phenotype that is consistent with an origin in the ventral SVZ. The distributions of cells resulting from dorsal (dark blue) or ventral (light blue) viral injection are shown for comparison. Data shown are average +/− SEM (n in figure is number of animals). This localization can be observed consistently in multiple samples (N, O). The OB core (dashed line) is located at the bottom of both images.

The initial population of YFP-labeled ventral cells were largely GFAP-positive (Figure 2D). YFP-positive migrating neuroblasts appeared at 5 days after tamoxifen treatment in the dorsal and ventral SVZ, RMS, and the core of the OB (Figure 2B, E, H, K and Supplementary Figure 2). The presence of ventrally-derived neuroblasts in the dorsal wedge is consistent with the pattern of tangential migration to the OB (Doetsch and Alvarez-Buylla, 1996). These dorsal YFP-positive cells were GFAP-negative and doublecortin-positive, confirming their identity as neuroblasts (Figure 2B and Supplementary Figure 2). At 30 days after treatment, YFP-labeled cells were present throughout the SVZ, in the RMS, and in the granular and periglomerular layers of the OB (Figure 2C, F, I, L). We also observed many YFP+ cells in the medial septum at 30 days after tamoxifen treatment (Supplementary Fig 3). These labeled septal cells were primarily S100β+ astrocyte-like cells, similar to those observed upon EGF infusion into the SVZ (Deloulme et al., 2004; Hachem et al., 2005; Gonzalez-Perez et al., 2009). We next quantified the localization of labeled progeny in the OBs of Gli1-CreER^T2; R26YFP animals at one month after tamoxifen. The labeled cells were primarily interneurons located in the deep granule layer of the OB, similar to the cells labeled by viral injection in the ventral SVZ (Figure 2M–O and Supplementary Figure 2M). The distribution of labeled cells differed significantly from distributions generated by combining dorsal and ventral injection data or from general BrdU labeling and lineage tracing in the SVZ (shown in Figure 4), confirming that Gli1 expression delineates a population of ventrally-derived interneurons.

Figure 4. Loss of Shh Alters the Production of Olfactory Interneurons.

(A–F) Treatment of ShhCreER/Shh^fl animals with tamoxifen results in the loss of Shh protein and gli1 expression. Schematic at top shows the timeline of tamoxifen and BrdU administration. At six weeks after tamoxifen administration, Shh protein is present in the septum and ventral SVZ of control animals (A, C). The SVZ is outlined by a red dashed line. In situ hybridization for gli1 shows the expected pattern of ventral expression in Shh^fl/+ animals (arrowhead, E). Shh protein is largely absent in ShhCreER/Shh^fl animals (B, D). gli1 expression is also substantially reduced or absent in ShhCreER/Shh^fl animals, indicating that pathway activity is compromised (arrowhead, F). White scale bars throughout figure: 50 microns. Black scale bars: 500 microns.

(G) The distribution of newly born granule neurons in the OB is altered in the absence of Shh. At three weeks after a one-week pulse of BrdU in Shh^fl/+ animals, labeled cells are distributed throughout the granular layers of the OB, with a greater percentage of the total present in the deep granule layers close to the core (blue bars, G). However, in ShhCreER/Shh^fl animals, BrdU-labeled cells were less frequent in the deep granule layers and more prevalent in the superficial layers of the OB. Quantification of the distribution of BrdU-labeled cells by genotype shows a statistically significant decrease in the number of deep granule cells and concomitant increase in the number of superficial granule cells (* - p = 0.01, unpaired t test). Data shown are average +/− SEM. N indicates number of biological replicates counted, with at least 100 OB granular layer cells counted per animal.

(H–J) The production of calbindin-positive periglomerular cells is decreased in the absence of Shh. At three weeks after a one-week pulse of BrdU, a subpopulation of BrdU-labeled cells in the periglomerular layer of Shh^fl/+ OBs are calbindin-positive (colabeling shown in I). This population is substantially reduced in ShhCreER/Shh^fl animals (p = 0.0033, unpaired t test).

Sonic Hedgehog Ligand is Produced in the Ventral Forebrain

We next investigated possible sources of Hh ligand in the adult brain. As previously reported (Echelard et al., 1993), we did not observe expression of Indian hedgehog or Desert hedgehog in the adult brain by in situ hybridization (data not shown). We did not observe expression of shh in the dorsal SVZ. However, shh mRNA was present in the medial septum, ventral forebrain and in infrequent cells close to the ventral SVZ (Supplementary Figure 4A–C). This was confirmed by qRT-PCR on dissected ventral SVZ and septum (Supplementary Figure 4D), and is consistent with previous reports of in situ hybridization in the rat brain (Traiffort et al., 1999). We next used a genetic approach to label the cells producing Shh and visualize cell morphology. We first examined ShhCre; R26YFP animals (Harfe et al., 2004), in which cells expressing Shh at any point in development recombine the R26YFP reporter (Supplementary Figure 4E). In addition to labeling in the cerebellum and cortex, we also observed an accumulation of YFP+ cells in the ventral forebrain.

By administering tamoxifen to ShhCreER; R26YFP animals, we induced YFP expression in cells producing Shh in the adult (Figure 3) (Harfe et al., 2004). YFP expression identified cells in the ventral forebrain, extending along the rostrocaudal axis adjacent and ventral to the SVZ. These cells primarily localized to the medial and ventral septum, the preoptic nuclei near the hypothalamus, and the bed nuclei of the stria terminalis. We also observed rare labeled cells in cortex (Garcia et al., 2010). Within the septum, YFP+ cells localized to both the horizontal and vertical limbs of the diagonal band, approximately 0.25 – 1 mm ventromedial to the SVZ (Figure 3A, 3B). A number of YFP+ cells in the bed nuclei of the stria terminalis were in close proximity to the ventral tip of the lateral ventricle (boxed area in Figure 3A, Figure 3D). We did not observe YFP-labeled cells near the dorsal SVZ, the RMS, OB, or in the choroid plexus - other sites which have been suggested to produce this ligand (Balordi and Fishell, 2007a; Angot et al., 2008). Labeled cells had the morphology of neurons, and all were colabeled by the neuronal marker protein NeuN (Figure 3H). Most YFP-labeled cells were GABAergic (GABA-positive, Figure 3I), with a small number of cholinergic (ChAT-positive), YFP-positive cells (Figure 3J). We did not observe YFP-positive dopaminergic (TH-positive) cells (data not shown). Labeled cells within the bed nucleus of the stria terminalis were of particular interest because of their close proximity to the ventral SVZ, and we examined these cells in greater detail. Computerized tracings of YFP-labeled cells highlighted thin processes that were located close to the basal side of ventral SVZ cells (Figure 3E, 3F). In order to better characterize these cells, we used whole-mount preparations of ShhCreER; R26YFP animals, allowing en-face visualization of the lateral and medial walls of the ventricles. ShhCreER; R26YFP-labeled cells had thin processes that interdigitated with the basal side of cell bodies within the SVZ (Figure 3G).

Figure 3. Sonic Hedgehog-Producing Cells Localize to the Ventral Forebrain and Can Be Labeled Via the SVZ or Lateral Ventricles.

(A–D) Colorimetric staining for YFP expression in coronal sections of ShhCreER; R26YFP animals 4 days after tamoxifen treatment. Clusters of labeled cells are visible in the medial septal nucleus (red box in left section in A, enlarged in B), the preoptic nuclei (middle section in A), and the bed nuclei of the stria terminalis (red boxes in middle section of A, enlarged in D). A small number of labeled cells approached the ventral regions of the lateral ventricles (D). We did not observe YFP-labeled processes or cell bodies in the dorsal SVZ (red box in middle section of A, enlarged in C). Scale bars throughout figure: 50 microns.

(E–G) Labeled cells in ShhCreER; R26YFP animals extend processes towards the ventral SVZ. Frontal sections of the ventral tip of the SVZ (delineated by red dashed lines) are shown in F and G with accompanying computerized tracing of processes shown in white. En-face visualization of a whole-mount specimen is shown with tracing of processes in G.

(H–J) Labeled cells in ShhCreER; R26YFP animals express the neuronal marker NeuN (H). A subset of these neurons express the neurotransmitter GABA (I) or the cholinergic cell marker choline acetyltransferase (ChAT, shown in J). No colabeling of YFP and the neuronal subtype marker tyrosine hydroxylase (TH) was observed (not shown).

(K–M) Colabeling of Shh-producing cells in the septum with Fluorogold retrograde tracer. At 4 days after tamoxifen treatment and 2 days after injection of Fluorogold into the lateral ventricle of ShhCreER; R26YFP mice, many Fluorogold-labeled cells (L) can be observed in the medial septum. YFP-positive cells are present in this region (K), and these cells are also labeled with Fluorogold (arrow, M).

(N–R) Shh protein is present in the septum and SVZ. (N) Antibody against Shh protein robustly labels the Purkinje cells of P5 cerebellum, a known source of Shh ligand. In tamoxifen-treated ShhCreER; tdTomato mice, Shh protein is found around the ventral SVZ, and is found in the cell body of an RFP-labeled neuron (arrow, O). Within the septum, Shh is present in the neuropil and in cell bodies associated with NeuN-positive nuclei (arrows, P). In the dorsal SVZ, minimal staining is observed, although infrequent punctuate staining is present (Q). In contrast, higher levels of Shh protein can be observed in and around the ventral SVZ (red dashed line, R).

To test whether these cells are actively producing Shh and might contact the subventricular zone, we pursued two strategies. First, we injected Ad:CreStopLight (Ad:CSL) virus in the ventral SVZ of Shh-Cre animals. This virus contains a CMV promoter driving the expression of the gene encoding red fluorescent protein (and a STOP sequence) flanked by loxP sequences, followed by the gene encoding green fluorescent protein (Yang and Hughes, 2001). After injecting Ad:CSL in the ventral region of the SVZ, we observed many RFP-positive cells at the injection site (Supplementary Figure 4F). We also observed a subset of GFP-positive cells located near the ventral SVZ in the bed nucleus of the stria terminalis, similar to the labeled cells identified in ShhCreER; R26YFP animals (Supplementary Figure 4G, 4H). We did not observe GFP-labeled cells after an equivalent injection in the dorsal SVZ (Supplementary Figure 4I, 4J). These ventral cells were therefore actively producing Cre recombinase under the control of the Shh promoter.

Previous studies in rodent brain have suggested that Shh may be transported anterogradely along axons and secreted distally at axon terminals (Traiffort et al., 2001). To determine if more distant Shh-producing cells might also contact the SVZ, we used the retrograde tracer Fluorogold (hydroxystilbamidine). After treating P90 ShhCreER; R26YFP animals with tamoxifen, we waited two days to allow accumulation of YFP label in the absence of injury, and injected Fluorogold into the lateral ventricle. At two days after tracer injection, we found significant Fluorogold labeling around the ventricles and in the bed nuclei of the stria terminalis, as well as many labeled cells in the medial septum and preoptic nuclei. A small number of Fluorogold-labeled cells in the septum also expressed YFP (Figure 3K–M), suggesting that Shh produced in the septum could also reach the SVZ by transport along axonal extensions. We did not observe more distant Fluorogold-labeled cells in the midbrain or hindbrain.

Finally, we performed immunohistochemical staining for Shh protein (rabbit MAb 95.9, kind gift from Genentech). We found that Shh protein was present in the septum in cell bodies associated with NeuN-positive nuclei and at low levels in the associated neuropil (Figure 3P). In the ventral SVZ, we also found Shh protein in the neuropil and in association with the apical and basal surfaces of SVZ cells (Figure 3O, 3R, and 4C). Infrequent, punctate staining for Shh protein was also observed in the dorsal SVZ (Figure 3Q), suggesting that a low level of ligand may be present in this region despite the absence of Shh-producing cells or their processes.

Removal of Shh Reduces the Production of Ventrally-Derived Olfactory Neurons

Although Shh ligand production and pathway activity both occur in the ventral forebrain, this did not necessarily indicate a requirement for this signaling in the generation of particular olfactory interneurons. We examined the requirement for Shh ligand in the adult brain by conditionally deleting Shh in mice carrying one copy of a knockin allele of ShhCreER and a floxed Shh allele in the second locus (ShhCreER/Shh^fl). Adult (P60) ShhCreER/Shh^fl mice and littermate Shh^fl/+ controls were treated with tamoxifen for 5 days to induce deletion of the functional Shh allele. After a two week period to ensure loss of Shh protein expression and tamoxifen clearance, mice were given a 1 week pulse of BrdU, followed by a 3 week chase, to label newly produced interneurons in the OB. At the end of this timecourse, we observed a decrease in Shh protein in both the septum and ventral SVZ of ShhCreER/Shh^fl mice (Figure 4A–4D). Importantly, we also observed a loss of gli1 mRNA expression in tamoxifen-treated ShhCreER/Shh^fl mice but not treated Shh^fl/+ controls, indicating that Shh pathway activity was significantly decreased in ventral SVZ (Figure 4E, 4F). As expected, the OBs of control animals had BrdU–labeled cells distributed throughout the granular layer, with 36% of cells observed in the deep granule layer (Figure 4G, Supplementary Figure 5). Similar numbers of BrdU-labeled interneurons were present in all genotypes, and label-retaining cells as well as proliferating cells were present in the SVZ for both genotypes, suggesting that stem cell self-renewal and progenitor proliferation were not grossly affected (data not shown). However, in ShhCreER/Shh^fl animals, there was a shift in the distribution of labeled cells, with 15% fewer deep granule cells present and a 25% increase in superficial granule cells in the labeled population when compared to controls (Figure 4G, p = 0.02, unpaired t test). This suggests that a subset of deep granule interneurons is lost when Shh ligand is removed from the adult brain.

We also examined the effects of Shh loss on the population of calbindin (CalB)-positive periglomerular cells normally produced by the ventral SVZ (Merkle et al., 2007). In ShhCreER/Shh^fl animals, production of new CalB+ periglomerular cells (BrdU/CalB double-positive cells) was decreased by almost 90% compared to controls (Figure 4H–J, p = 0.0033, unpaired t test). This reduction in CalB-positive cells was more pronounced than the reduction in deep granule cells. Shh signaling may be required for the generation of specific subgroups of deep granule cells, but not others, resulting in a smaller decrease in the total population of deep granule cells. However, in both populations of cells, we observed a meaningful change in the cell types generated, indicating that Shh production plays a role in the production of different types of neurons destined for the OB by ventral NSCs.

Dorsal SVZ Cells Are Intrinsically Resistant to Smo Agonist

To test whether dorsal and ventral SVZ cells are equivalently responsive to Shh pathway activation, we administered Smoothened agonist (SAG) in the cerebrospinal fluid via an intracranial osmotic pump into the lateral ventricles. SAG is a small molecule that efficiently activates the Hh pathway (Chen et al., 2002). Vehicle (saline), SAG, or the pathway antagonists cyclopamine or anti-Shh antibody (5E1) were infused for 5 days in Gli1CreER; R26YFP mice. During infusion, mice were treated with tamoxifen to acutely label cells in which the Hh pathway was active. YFP labeling in vehicle-infused mice was predominantly ventral, demonstrating that pump installation alone did not alter the pattern of Hh signaling (Figure 5A, 5E). We observed increased GFAP labeling after all pump implantations, likely due to increased numbers of reactive astrocytes. Administration of either cyclopamine or 5E1 antibody reduced the number of YFP-positive cells in the ventral SVZ, confirming that YFP labeling was dependent on pathway activation (Figure 5B, 5C, 5F, 5G). SAG infusion resulted in a dramatic increase in YFP-positive cells, both GFAP-positive and –negative, in the ventral SVZ, but did not significantly alter the pattern of YFP labeling in the dorsal SVZ (Figure 5D, 5H).

Figure 5. Dorsal SVZ Cells Are Resistant to Pathway Activation.

(A–H) Infusion of Hedgehog pathway agonists in the SVZ. Schematic at top illustrates pump cannula placement, allowing infusion of solution into the SVZ and CSF. Infusion of vehicle (sterile saline) does not alter the predominantly ventral distribution of cells labeled by the Gli1CreER^T2; R26YFP reporter (A,E). Infusion of either the small-molecule antagonist cyclopamine (B,F) or an anti-Shh antibody (C,G) results in a reduction in YFP-positive cells. Administration of the small-molecule agonist SAG causes a robust increase in YFP labeling in the ventral SVZ (H), but does not cause a similar upregulation of YFP expression in the dorsal SVZ (D). Scale bars: 50 microns.

Infusion of cyclopamine, 5E1, or SAG may also affect SVZ cell survival or proliferation, as suggested by previous experiments in which Smoothened was ablated in the SVZ (Balordi and Fishell, 2007b). Staining for the proliferation marker Ki67 indicated that large changes in proliferation did not occur during the time frame of this experiment. In both controls and SAG-infused animals, we observed small populations of YFP-positive cells in the dorsal SVZ. Most of these cells were GFAP-negative and Dcx-positive suggesting that they correspond to young migrating neurons (Figure 5 and data not shown). We cannot exclude that a small subpopulation of Gli1-expressing type B or C cells are present in dorsal regions. The regional difference in Gli1 distribution remains after agonist infusion, suggesting that additional cell-intrinsic factors may affect the ability of dorsal cells to activate the Hh pathway.

Ectopic Pathway Activation Reprograms Neuronal Progeny

While high Hedgehog pathway activity was required for the production of particular types of neuronal progeny, this observation did not necessarily indicate that Hedgehog signaling had an instructive role in cell fate. To investigate this possibility we performed targeted injections of Ad:GFAPpCre virus in SmoM2-YFP; R26R mice (Mao et al., 2006). The Ad:GFAPp-Cre virus, in which Cre recombinase expression is driven by the murine GFAP promoter, results in recombination in primary progenitors (type B cells) within this region (Merkle et al., 2007). In these animals, Cre-mediated recombination causes the expression of SmoM2, a constitutively active mutant of Smo, and activation of the Hh pathway. By injecting Ad:GFAPpCre in the dorsal SVZ of SmoM2-YFP; R26R animals, we activated the Hh pathway to high levels in a ligand-independent, cell-intrinsic fashion while simultaneously labeling these cells with β-galactosidase. This allowed us to follow the lacZ-labeled progeny of these infected stem cells. Remarkably, while expression of the SmoM2 protein is sufficient to drive rapid tumorigenesis in other contexts (Schuller et al., 2008), we did not observe abnormalities within the SVZ of injected SmoM2-YFP; R26R animals. One month after injection, lacZ-labeled cells were found at the site of injection in control and SmoM2-YFP; R26R animals (Figure 6). gli1 expression was absent in the dorsal SVZ of injected R26R mice. SmoM2-YFP; R26R animals showed a marked upregulation of gli1, but not Shh, mRNA at the site of virus injection (Supplementary Figure 6A–I), confirming that Hh signaling was active in infected cells.

Figure 6. High Pathway Activity Causes Relocalization of Neuronal Progeny.

(A–L) Ad:GFAPpCre was injected into the dorsal or ventral SVZ of R26R and SmoM2-YFP/R26R animals and brains were collected one month later to analyze labeled progeny. LacZ labeling within the SVZ (A, D, G, J) confirmed the site of virus injection. Dorsal injections in control animals primarily generated superficial labeled cells (B) that were distant from the core of the OB (indicated by a dashed line). Injections in the dorsal SVZ of SmoM2 animals gave rise to deep cells located close to the core of the OB (E), similar to those derived from ventral viral injections in either genotype (H,K). β-galactosidase-expressing progeny in both control and SmoM2 animals coexpressed the NeuN protein, a marker of mature neurons (arrows in C,F,I,L). Black scale bars: 50 microns. White scale bars: 14 microns.

(M) Quantification of the spatial distribution of labeled progeny in R26R and SmoM2-YFP; R26R animals after dorsal or ventral Ad:GFAPpCre injection. Ventral virus injection in either genotype (pale red and pale blue bars) results in cells that are mostly located close to the core of the OB. Dorsal virus injection in R26R animals (bright blue bars) results in many cells that are distant from the core. However, dorsal virus injection in SmoM2; R26R animals results in cells that are located close to the core (bright red bars), with a distribution similar to that from ventral injections. Data shown are average +/− SEM, with total number of animals indicated on graph. Asterisks = p < 0.01.

At one month after dorsal injection of Ad:GFAPpCre in control R26R animals, lacZ labeling marked a population of cells located in the superficial granular layer of the OB, consistent with previous work (Figure 6B). Remarkably, dorsal injections in SmoM2-YFP; R26R animals generated a population of labeled cells that localized to the deep granule layer of the OB (Figure 6E, 6M, and Supplementary Figure 6J), similar to the progeny resulting from injections in the ventral SVZ of R26R or SmoM2-YFP/R26R animals (Figure 6H, 6K). These labeled cells expressed NeuN (Figure 6C, 6F, 6I, 6L), suggesting that SmoM2 expression in infected neural stem cells did not block maturation, but did alter the type of progeny generated.

We next injected Ad:GFAPpCre in SmoM2-YFP; CAG animals and CAG littermates to generate progeny expressing GFP, which fills the cell and allows visualization of cell morphology (Figure 7). Ad:GFAPpCre infection of dorsal SVZ cells in SmoM2-YFP; CAG animals caused a shift in the localization of GFP-expressing progeny in the OB like that observed with the R26R reporter (Figure 7A,D). Within the SmoM2-YFP; CAG SVZ, we observed an almost 4-fold increase (p = 0.0025) in the dorsal expression of the transcription factor Pbx3a, which is normally limited to the ventral SVZ (Figure 7J) (Figure 7L, 7R). We also observed a decrease in expression of Pax6, which is normally present in the dorsal SVZ, in YFP-positive cells in SmoM2/CAG animals (Figure 7N, 7P). Within the OB, SmoM2/GFP-expressing cells were positive for NeuN and the neurotransmitter GABA (Figure 7B,C,E,F), confirming that the relocalization of progeny does not block their maturation into interneurons. The projection patterns of deep and superficial interneurons differ (Merkle et al., 2007; Whitman and Greer, 2009), so in addition to soma location, we traced the arborizations of GFP-labeled cells. The progeny of dorsally injected CAG animals were primarily superficial interneurons with dendrites that reached past the midline of the external plexiform layer of the OB (Figure 7G). After dorsal injections in SmoM2/CAG animals, labeled olfactory interneurons tended to have dendrites that contacted the inner half of the external plexiform layer, a feature that is typical of deep granule interneurons (Figure 7H). In addition to the deep granule cells that arise from the ventral SVZ, calbindin-expressing periglomerular cells are also derived from this region. Within the periglomerular cell population derived from dorsal injections, we saw a three-fold enrichment of CalB/GFP-double positive cells in SmoM2-YFP; CAG animals when compared to controls (Figure 7O, 7Q, 7S). We conclude that ectopic activation of the Hh pathway is sufficient to respecify dorsal SVZ neuronal progeny.

Figure 7. SmoM2 Expression Results in Relocalization, Altered Dendritic Projections, and Ventral Marker Expression.

(A–F) Dorsal injection of Ad:GFAPpCre in SmoM2-YFP; CAG animals (D–F) also resulted in relocalization of labeled progeny close to the core of the OB (dashed line) when compared to CAG controls (A–C). GFP-expressing progeny of both genotypes expressed NeuN (arrows, B and E) and the neurotransmitter GABA (arrows, C and F), consistent with their status as differentiated inhibitory interneurons. Scale bars: 50 microns.

(G, H). Camera lucida tracings of individual GFP-labeled cells in CAG and SmoM2; CAG animals show the projections of these cells into the external plexiform layer of the OB (delineated by gray shading). Dashed lines indicate the mitral cell layer (lower line) and periglomerular layer (upper line) of the OB. While dorsal SVZ-derived cells in CAG animals typically project to the outer half of the EPL (G), dorsal SVZ-derived cells from SmoM2; CAG animals project to the inner half of the EPL, a feature associated with a deep granule interneuron identity (H).

(I,J,L,M,N,P,R) At one month after dorsal injection of Ad:GFAPpCre in CAG animals, many GFP-labeled cells are present at the site of injection in the dorsal SVZ (I). However, the transcription factor Pbx3a is more prevalent in the ventral SVZ (J). After a similar injection in SmoM2-YFP; CAG animals, Pbx3a is present at similar frequency in the uninjected cells of the ventral SVZ (M), but is also enriched in dorsal GFP-positive cells (L). str = striatum, sep = septum. This enrichment is quantified in (R) – graph shows average +/− SEM for three independent injections per genotype (p = 0.0025, unpaired t test). In concert with the increased expression of the ventral marker Pbx3a, we also observed a loss of the dorsal marker Pax6 – while many dorsal GFP-positive cells normally express Pax6 (arrows, N), this expression is absent in SmoM2/CAG animals (P).

(O,Q,S) Within the OBs of CAG animals at one month after dorsal Ad:GFAPpCre injection, few GFP-positive periglomerular cells (O,S) expressed the marker calbindin (shown in red). However, we did observe an increase in the number of CalB/GFP double-positive cells in periglomerular cells derived from injections in SmoM2-YFP; CAG animals (Q,S). Graph shows average +/− SEM for three independent injections per genotype.

Discussion

We demonstrate that the Hedgehog pathway has a critical function in the regional specification of stem cells of the adult brain. We found that Smo is expressed by stem cells throughout the SVZ. However, high Hh pathway activation is more prevalent in the ventral SVZ and is sufficient to respecify dorsal SVZ stem cells when ectopically induced. In the forebrain, we found that only Shh, and not other Hh isoforms, are expressed. Interestingly, Shh production occurs in neurons that are located very close to the ventral subventricular zone. These results indicate that specification of subpopulations of neural stem cells and their progeny in the adult brain is actively regulated by Hh signaling.

It has previously been suggested that the Hh pathway functions in multiple cell types within the SVZ. Genetic and pharmacologic studies have argued that this pathway is important in stem cell self-renewal, the generation of transit-amplifying progeny, and neuroblast migration to the OB (Machold et al., 2003; Palma et al., 2005; Balordi and Fishell, 2007a, b). Our results indicate that astrocytes of the SVZ are the major Hh-responsive population within the SVZ: expression of Smo protein, which is required for a cell to respond to Hh ligand, is restricted to the GFAP-positive population (including type B1 cells) in this region. Intriguingly, Smo expression labels a subset of GFAP-expressing SVZ cells, suggesting either that Smo is only expressed in a subset of stem cells, or that it is a more specific marker of stem cells than GFAP, which is also expressed by some astrocytes in non-germinal regions (Garcia et al., 2010).

We searched for components of the Hh signaling pathway that might be differentially expressed in the subregions of the SVZ. We did not observe significant differences in the expression of Smo, Gli2, or Gli3 between the ventral and dorsal SVZ, but found that Gli1 expression is much higher in the ventral SVZ. Lineage tracing of these cells using Gli1-CreER^T2 mice indicates that progenitors responding to high levels of Shh are located in the ventral SVZ and primarily generate deep granule interneurons. Moreover, Hh signaling appears to be instructive in cell fate decisions: ablation of Shh expression reduced the production of ventrally-derived progeny, and ectopic activation of the pathway resulted in the respecification of dorsally-derived progeny to a ventral fate. These results are surprising, as previous studies using viral targeting and cell culture indicated that cell identity was largely encoded using a cell-intrinsic program: transplantation of dorsal cells to a ventral location, even after intervening time in culture, was not sufficient to respecify their progeny (Merkle et al., 2007).

The differences in cell fate between previous transplantation studies and the results shown here may be due to several possible mechanisms: dorsal SVZ cells may be intrinsically resistant to the Hh signal, resulting in lower pathway activation even upon exposure to equivalent levels of ligand, or a second developmental pathway may act to oppose Hh action in the dorsal region, as is the case in the developing neural tube (Liem et al., 2000). It has also recently been recognized that SVZ stem cells have a specialized apical-basal orientation within the SVZ niche (Mirzadeh et al., 2008; Shen et al., 2008; Tavazoie et al., 2008). Transplantation experiments may not allow the grafted population to integrate and adopt proper apical-basal positioning within the niche. Hh ligand may be delivered to ventral SVZ cells via specialized local contacts which are not recapitulated after transplantation. Importantly, the present results indicate that strong activation of the Shh pathway can override the intrinsic programming of dorsal neural progenitors, suggesting that reprogramming of neural stem cells for therapeutic purposes may depend on the identification of the relevant molecular signal for a desired cell type. This study also provides the first in vivo respecification of adult neural cell fate by modulation of the Hh pathway.

We identify clusters of Shh-producing neurons in the ventral forebrain, in locations that are consistent with previous studies at the RNA level. A subset of these cells, in the bed nucleus of the stria terminalis, have processes that are immediately adjacent to the ventral SVZ. In addition, some Shh-producing cells in the ventral and medial septum are able to take up retrograde tracer molecules that are injected into the lateral ventricle, suggesting that Shh ligand may also reach the ventral SVZ by anterograde transport from the septum (Traiffort et al., 2001). The localized activation of ventral SVZ stem cells, and expression of Gli1 in these cells, might be part of an adult brain regulatory mechanism to locally modulate production of specific neuronal subtypes destined for different OB circuits.

Shh is produced by Purkinje neurons in the developing cerebellum, and by cells of the floor plate in the neural tube (Ho and Scott, 2002; Fuccillo et al., 2006). In these instances, Shh signaling directs significant large-scale remodeling and patterning of developing tissue. Our results demonstrate that in the adult brain, this pathway remains active and directs the production of specific subtypes of neurons. The finding that mature neurons in the adult brain are a likely source of Shh ligand suggests that neural network activity may regulate generation of certain types of neurons within the SVZ. It remains unclear if Shh reaches ventral stem cells via diffusion, or whether more specialized contacts exist between Shh-producing neurons and stem cells. It will be of interest to study the mechanism of Hh ligand delivery to the SVZ, and to identify what circuits in the brain might activate the release of Shh ligand.

The SVZ is the largest germinal zone in the adult brain, and is capable of generating thousands of immature neurons each day. Importantly, it is organized as a mosaic, with stem cells in different locations producing different types of neurons (Hack et al., 2005; Kohwi et al., 2005; Kelsch et al., 2007; Merkle et al., 2007; Ventura and Goldman, 2007; Young et al., 2007; Alvarez-Buylla et al., 2008). These groups of olfactory interneurons differ in location and are also thought to differ functionally (Gheusi et al., 2000; Cecchi et al., 2001; Kosaka and Kosaka, 2007). Here we elucidate a molecular mechanism for the specification of a subpopulation of neural stem cells within this extensive adult germinal layer. We show that manipulation of this pathway allows stem cells to be redirected to a different fate. These studies demonstrate that adult neural stem cells, which normally produce a restricted repertoire of progeny, may be reprogrammed if the relevant specification signals are identified.

Experimental Procedures

Tamoxifen Administration - All animal procedures were carried out in accordance with institutional (IACUC) and NIH guidelines. Tamoxifen was prepared at 20 mg/mL in corn oil and administered via oral gavage. Adult mice (P60–P90) were treated with 1 mg (Gli1-CreER) or 5 mg (Shh-CreER)/day tamoxifen for 5 consecutive days and sacrificed at the specified times.

Antimitotic Treatment - Adult mice were treated with a solution of 2% cytosine-β-arabinofuranoside (Sigma) or sterile saline alone as a control. Solutions were infused for 6 days at the pial surface using a miniosmotic pump (Alzet 1007D).

Injections and Adenovirus Preparation - Ad:GFAPpCre into was injected into dorsal and ventral adult SVZ as described (Merkle et al., 2007) using 50 nl of virus. Injections of Ad:CSL were carried out using 100 nl of virus, and Fluorogold injections were carried out using 250 nL of tracer. Injections used the following coordinates: dorsal SVZ – 0.5 anterior, 3.2 lateral, 1.8 deep, needle at 45° angle; ventral SVZ – 0.5 anterior, 4.6 lateral, 3.3 deep, needle at 45° angle; lateral ventricle – 0.38 posterior, 1.0 lateral, 2.25 deep, with needle vertical. X and Y coordinates were zeroed at bregma, and the Z coordinate was zeroed at the pial surface.

Miniosmotic Pump Implantation – Miniosmotic pumps (Alzet 1007D) were assembled under sterile conditions, filled with vehicle (0.9% saline), 0.5 µM Smoothened agonist (EMD Chemicals), 5 µM cyclopamine (Sigma), or 5 µg/mL 5E1 antibody (DSHB), and allowed to equilibrate overnight at 37°C. Pump installation was performed on a stereotaxic rig as previously described.

Immunostaining - Immunostainings were carried out on methanol-fixed 10 micron frozen sections (Supplementary Figure 1) or paraformaldehyde-fixed 50 micron free floating Vibratome sections (all other stains) according to standard procedures. Primary antibodies used were goat anti-Smo (Santa Cruz, C-17), rabbit anti-Smo (MBL 2668, kind gift of J. Reiter), mouse anti-GFAP (Chemicon), rabbit anti-doublecortin (Cell Signaling), mouse anti-PSA-NCAM (Chemicon), chicken anti-GFP (Aves Labs), rabbit anti-RFP (MBL), mouse anti-β-galactosidase (Promega), anti-NG2, rabbit anti-Olig2 (kind gift of D. Rowitch), rabbit anti-S100β, mouse anti-GABA (Sigma), goat anti-ChAT (Chemicon), anti-NeuN (Chemicon), rabbit anti-tyrosine hydroxylase (Pel-Freez Biologicals), rabbit 95.9 anti-Shh (kind gift from S. Scales, Genentech), mouse anti-Pbx3a (Santa Cruz), and rabbi anti-calbindin D28k (Chemicon). The secondary antibodies used were conjugated to AlexaFluor dyes (Invitrogen/Molecular Probes). For pre-blocking of Smoothened antibody (Santa Cruz, C-17), the antibody was incubated with the accompanying blocking peptide (sc-6367P, Santa Cruz). LacZ (X-gal) staining was carried out after brief fixation in paraformaldehyde according to standard procedures.

Microscopic Analysis, Tracing, and Quantification – Fluorescent staining was visualized using a Leica SP5 confocal microscope and analyzed using NIH ImageJ. Tracings of neuronal processes from fluorescent staining were completed using the Filament module in Imaris analysis software (Bitplane). Colorimetric staining was visualized using an Olympus AX70 microscope, Retiga 2000R camera and LabVelocity software. Measurements of olfactory interneuron localization were carried out using the Measure and Label plugin in NIH ImageJ, with normalization to granular layer width carried out as described (Merkle et al., 2007). Data were quantified and analyzed using GraphPad Prism 5. Tracing of colorimetric anti-GFP staining was completed using a Nikon microscope with camera lucida adapter. Hand tracings were scanned and imported into Adobe Illustrator using the LiveTrace and LivePaint modules.

qRT-PCR – For qRT-PCR, dorsal SVZ, ventral SVZ, striatum, and septum were microdissected from 2 mm slices of unfixed brain. Dissected tissue was immediately placed in RNAlater solution (Ambion) and stored at −20°C until all brains (14 total) were dissected. RNA isolation was done with the RNEasy Mini kit (Qiagen). cDNA was synthesized using SuperScript III RT (Invitrogen), and qRT-PCR was completed using SYBR Green PCR Master Mix (Applied BioSystems) on an ABI7900HT. Data analysis and statistical tests were carried out using the REST algorithm (Pfaffl et al., 2002).

Whole mount dissection – Whole mount dissection of Shh-CreER; R26YFP brains was performed as described (Mirzadeh et al., 2008). Full methods including video are available online (Mirzadeh et al., 2010).

In Situ Hybridization – In situ hybridization was performed using standard protocols (Han et al., 2008). Antisense and sense RNA probes were labeled with digoxigenin and visualized with alkaline phosphatase-NBT/BCIP reaction (Roche).