Sonic Hedgehog is expressed by hilar mossy cells and regulates cellular survival and neurogenesis in the adult hippocampus

Sonic hedgehog (Shh) is a multifunctional signaling protein governing pattern formation, proliferation and cell survival during embryogenesis. In the adult brain, Shh has neurotrophic function and is implicated in hippocampal neurogenesis but the cellular source of Shh in the hippocampus remains ill defined. Here, we utilize a gene expression tracer allele of Shh (Shh-nlacZ) which allowed the identification of a subpopulation of hilar neurons known as mossy cells (MCs) as a prominent and dynamic source of Shh within the dentate gyrus. AAV-Cre mediated ablation of Shh in the adult dentate gyrus led to a marked degeneration of MCs. Conversely, chemical stimulation of hippocampal neurons using the epileptogenic agent kainic acid (KA) increased the number of Shh+ MCs indicating that the expression of Shh by MCs confers a survival advantage during the response to excitotoxic insults. In addition, ablation of Shh in the adult dentate gyrus led to increased neural precursor cell proliferation and their migration into the subgranular cell layer demonstrating that MCs-generated Shh is a key modulator of hippocampal neurogenesis.

Graded Sonic Hedgehog (Shh) signaling is a crucial regulator of cell proliferation, cell fate determination and migration, leading to cell diversification and congruent growth during early nervous system development 1 . In the adult brain, the Shh receptor Patched (Ptc) is expressed by neural stem cells ("B"-cells) and rapidly amplifying ("C"-cells) within the neurogenic niches of the subventricular zone (SVZ, forebrain) and subgranular zone (SGZ, hippocampus) 2 . Consistently, Shh is critically involved in B-cell maintenance and -profileration and in C-cell fate determination in the neurogenic niches of the adult forebrain and the hippocampus [3][4][5] . Whether neurogenic activity is controlled by physiological needs remains an active area of research. Variable Shh signaling strength within the germinal niche can determine the rate of neurogenesis and the type of cells being produced 2 . A critical step in investigating whether neurogenic outcome could be adapted to need is therefore the identification of the cellular source of Shh and the determination whether Shh expression is variable. The relevant cellular source of Shh for adult hippocampal neurogenesis, however, remains ill defined.
Shh was found to be expressed in calretinin positive neurons (CR + ) of the hilus in the dorsal DG but not in the ventral DG in the early post-natal brain at P15 6 . The deletion of Shh from these CR + cells was associated with a significant decrease in proliferation and the number neuronal stem cells (NSCs) 6 . Whether these neurons express Shh in the adult hippocampus has not been studied. In contrast, immunohistochemical analysis has suggested that pyramidal neurons 7 or astrocytes 8 might express Shh in the adult hippocampus. However, the failure to detect Shh mRNA in the hippocampus by in situ hybridization early studies, led some authors to propose that Shh could originate outside of the hippocampus. Thus, the protein would be produced by neurons in the basal forebrain cholinergic nucleus VDB 9,10 where Shh transcription is abundant and anterogradely transported to the SGZ via the fimbria-fornix pathway 3 .
The difficulties associated with the identification of Shh cellular sources in the hippocampus might stem from the fact that Shh is a secreted protein. The presence of axonal transport signals in the Shh mRNA and protein sequence 11 and the release of Shh from axons as well as from the somato-dendritic compartment 12 , yielding low and difficult to detect concentrations of both Shh mRNA and protein in the soma of Shh producing neurons. Furthermore, the Shh protein may accumulate in target cells that could easily be misidentified as sources 12 . We therefore re-examined the expression of Shh within the hippocampus using a sensitive Shh gene expression tracer allele which marks nuclei of Shh expressing cells by nuclear targeted lacZ and allows selective identification of cells in which the Shh locus is transcriptionally active. This reporter was used previously to discover that mesencephalic dopamine neurons are a significant source of Shh throughout adulthood in the forebrain 13 .
Mossy cells (MCs) constitutes a major population of CR + neurons in the dentate gyrus (DG) of the hippocampus 14 . Extensive research has been performed to characterize MCs, but many of their functional and morphological properties remain elusive 15 . MCs are usually described as glutamatergic neurons that may exert feed-forward inhibition onto granular cells (GC) through GABAergic neurons 16,17 . However, no consensus has been reached as to whether the net effect of mossy cells on GCs is excitatory or inhibitory 15,18,19 . Many investigators assume that thorny excrescences define MCs, but there are spiny hilar cells without thorns that have the same physiological characteristics as 'thorny' MCs. Furthermore, MCs vary in their expression of neurochemical markers such as calretinin which is expressed in ventral but not dorsal mossy cells in mice (for review 15 ).
Mossy cells could be implicated in SGZ neurogenesis driving glutamate and GABA transmission at different phases of granular cell development, but few studies have investigated specific interactions between MCs and neurogenesis in the adult brain 15 . Recently, Yeh et al. 20 reported that MCs may control NSC quiescence through glutamatergic and GABAergic signaling. However, the notion that MCs could deliver Shh onto the NSCs as a possible activity-dependent regulatory mechanism of neurogenesis has not been explored so far.
Using a Shh-nlacZ genetic reporter 13 we demonstrate here that Shh is expressed by most hilar MCs in the adult brain of mice. We find that Shh is expressed by most MCs and that these cells co-express GABA and glutamatergic markers. Shh expression reduces excitotoxicity of MCs in response to kainate induced epilepsy. Conversely, genetic ablation of Shh from hilar cells results in decreased numbers of MCs but increased migration of newly born neuronal precursor cells into the granular cell layer. Together, our results suggest that Shh expression in adult MCs serves as a neuro-protectant for MCs, as a chemo attractant for immature neuronal precursor cells that ectopically migrate to the hilus to become CR + cells during induced excitotoxicity, and as an inhibitor of neuronal cell fates that home to the granular cell layer.

Results
calretinin expressing GABAergic neurons are the source of Shh in the hippocampus. To identify the cells that produce Shh in the DG, we first visualized expression of Shh in the adult brain using mice homozygous for a gene expression tracer allele of Shh (Shh-nlacZ +/+ ) 13 . Here Shh and nucleus-targeted LacZ is transcribed into a bi-cistronic mRNA from the endogenous Shh locus such that all cells that express Shh are also marked by nuclear localized betaGal allowing sensitive chromogenic and fluorescent immunohistochemical analysis of Shh expression with single cell resolution. The use of anti-beta galactosidase antibodies in combination with cell type specific markers demonstrated that Shh is not produced by oligodendrocytes (PLP-EGFP, Fig. 1G and Table 1), astrocytes (GFAP, Fig. 1H and Table 1) or immature cells (Nestin, Fig. 1I and Table 1) but only by mature neurons (NeuN, Fig. 1D,E and Table 1). Most nlacZ + neurons co-express GAD-65/67 ( Fig. 1B,C, Table 1) and form a pattern similar to that seen by recent in situ hybridization for Shh (Fig. 1F, Allen Atlas 21 ). We next investigated which sub-type of GABAergic neuron would express Shh. We stained brain sections from animals carrying the Shh-nlacZ allele with antibodies for parvalbumin (PV), somatostatin (STT), neuropeptide Y (NPY), and calretinin (CR), which are GABAergic neuronal subtypes present in the DG (Fig. 2) and as revealed by in situ hybridization (Allen Atlas 21 ). We found that among all the cells that express the Shh-nlacZ tracer allele, 98.1 ± 8.4% co-expressed CR and, conversely, among cells expressing CR, 72.4 ± 3.9% expressed the Shh-nlacZ tracer (Fig. 2E,F, Table 1). The CR + Shh-nlacZ + cells exhibited a multipolar morphology containing large polygonal somata (diameter ≈ 20 μm) with abundant primary axodendritic arborizations forming a dense network within the hilus. We did not find any other GABAergic neuronal subtypes among the cells expressing Shh. cR + neurons that express Shh in the adult hippocampus are mossy cells (MC). We next determined whether the cell population described here were mossy cells (MCs), a prominent subpopulation of CR + cells with elusive function. We found that cells expressing Shh-nlacZ co-localize with glutamate receptor 2/3 (GluR2/3) expression, a marker for MCs (Fig. 3C, Table 1). Among the GluR2/3+ neurons, 41.7% were Shh-nlacZ+ (Table 1, columm-5), and among Shh-nlacZ+, 92.4% were Glut R 2/3+ (Table 1, column-7). Further, over 90% of the cells expressing Shh-nlacZ are labeled for both CR and GlutR2/3 markers (Table 1, column-7). Consistent with the immunohistochemical staining, the GluR2 mRNA in situ hybridization image cropped from the Allen atlas 21 (Fig. 3A) resembles the distribution of Shh-nlacZ+ or CR+ cells in the hilus. Further, the majority of α-GluR2/3 stained soma colocalized with GABA in the hilus (Fig. 3D-F). Consistently, double-color fluorescence in situ hybridization (FISH) for CR (Calb2-IRES-Cre) and GAD1, show that most hilar CR + cells (≈75%) express GAD-1 mRNA (Fig. 3B) (Allen Atlas 21 ). Therefore, based on their distribution, morphologic features and staining for CR, GluR2/3, these results reveal that MCs remain as a prominent source of Shh in the adult hippocampus. The prevailing view is that CR and GluR2/3 are markers for mossy cells while MCs are thought to be non-immunoreactive to GABA markers. To further test the notion that MCs may express GABA, we decided to use another mossy cell-specific marker named calcitonin gene-related peptide (CGRP) 15,22,23 . We found that about 80% of the CGRP + cells colocalize with GABA (Supplemental Fig. S2). As the three MC markers (CR, GluR2/3 and CGRP) highly colocalize with GABA, it follows that mossy cells also express GABA even if these cells are not functionally GABAergic (see below).
www.nature.com/scientificreports www.nature.com/scientificreports/ Mossy cells that produce Shh are resistant to KA toxicity. We next sought to evaluate whether MCs expressing Shh were endowed with greater resistance to kainic acid (KA), a neurotoxic and epileptogenic agent, compared to GABAergic cells that do not express Shh. Shh-nlacZ mice were injected with increasing doses of KA (IP 5 mg/kg/h) to a maximum of 35 mg/kg until status epilepticus was reached (Racine's stage 4/5) 24 . The behavioral assessment of the animals during the second week after KA injections revealed increased motor activity, exaggerated grooming, stereotypes and epileptiform jumping, shaking and forelimb clonus (Hyperexcitability or Racine's stage 3) (Fig. 4K-N). These observations indicated that at this point the animals have not achieved full development of the epileptic phenotype but rather were engaged in an epileptogenic process. To characterize this stage, histological changes were studied in these animals and to avoid confounding effects derived of the stereotaxic surgery and electrode implantation, the electrophysiological effects of the KA injection were studied in a separated group of mice subject to the same procedure. The behavioral changes shown in Fig. 4 parallel the increased neural activity in the hippocampus as shown by EEG recordings (Supplemental Fig. S1). The EEG Figure 1. Shh is expressed locally in the hilus of the DG by GABAergic neurons in the adult brain (P60) (for quantification see Table 1). (A) From horizontal slices of the brain, we examine the whole ventral portion of the dentate gyrus. The horizontal brain slices yield transverse sections of the ventral portion of the hippocampus. also showed typical interictal activity consistent in high-amplitude bursting and hyper-synchronized spiking in KA-injected animals, which are specific abnormalities associated with the epileptogenic stage.
Two weeks after KA injections animals were euthanized and their brains were processed for immunohistochemistry (IHC). We stained for nlacZ and GABAergic markers in the hippocampus. The level of Shh expression per cell as measured by the stained surface area and fluorescence intensity of the nlacZ immunostaining did not differ between the groups [surface area stained by nlacZ for vehicle, 63.2 ± 6.4 and for KA, 60. To distinguish these possibilities, we focused on CR + cells and determined the morphological characteristics and locations of CR + Shh − and CR + Shh + neurons in the hippocampus of untreated animals. As quantified above, a minority of CR + cells do not express Shh (Fig. 4J). These Shhcells are located in the hilus where they make up 14% of all CR + cells and in the SGZ where they make up 35% of all CR + cells (Fig. 5). These two CR + Shh − groups differ from Shh + CR + neurons by morphology. CR + Shh − neurons exhibited small round somata (diameter 10.7 ± 0.3 μm, surface area ≈ 150 μm 2 ; Fig. 5C) and small dendritic arborizations, while the CR+ Shh + neurons possessed a larger somata (diameter 20.9 ± 0.7 μm, surface area ≈300 μm 2 ) and abundant primary axodendritic arborizations. KA treatment increased the number of both large CR + Shh + multipolar neurons and small round CR + Shh − cells in the hilus (Fig. 5A,B E,F) but decreased the number of small CR + Shh − neurons in the SGZ (Fig. 5G). These findings are in line with previous results that revealed that the number of hilar CR + cells increased 2.5 times in mice injected with intra-DG KA 25 .
We tested next whether the increase of CR + cells in the hilus might have been caused by increased neurogenesis. In agreement with previous reports [26][27][28] , however, we did not find increased numbers of cells expressing the mitosis marker phosphorylated Histone 2B (pH2B) (Fig. 6A-H,M), or that had incorporated the nucleotide analog BrdU (Fig. 6I-L,N) in the SGZ suggesting that KA treatment did not induce an increase in the rate of neurogenesis. Therefore, in the absence of increased proliferation, our observations suggest that KA might induce increased migration of immature, small CR + cells from the SGZ into the hilus where they mature into large CR + Shh + neurons. This possibility is consistent with the previous finding of ectopic migration of newborn cells from the SGZ to the hilus in the KA model 29 and with the increase in CR + neurons following KA administration 25 . in the SGZ on neuronal progenitors 3,4 . Calretinin based chromogenic staining (Fig. 7) showed profuse reciprocal innervation between CR + cells (Fig. 7B), as described previously by Gulyás et al. 30 , as well as CR + innervation of both SGZ (Fig. 7I) and hilar progenitors 20 (Fig. 7J). Interestingly, hilar CR + neurons innervate immature CR + cells in the SGZ (Fig. 7K-L). Because of this pattern of reciprocal innervation observed between CR + neurons, we wanted to investigate whether hilar CR + neurons were able to perceive Shh signaling and therefore might not only act as a source but also as a target for Shh signaling. Staining of CR + cells in the hilus (Fig. 7C,D) for Ptc (Fig. 7E) and Smo (Fig. 7F) revealed no co-expression with Shh-nlacZ. In contrast, oligodendrocytes (PLP+ cells) did not express Shh-nlacZ (Table 1), but Ptc suggesting that these cells are likely local recipients of Shh signaling in the This analysis leads to conclude that Shh is exclusively express in mossy cells that express CR and GAD. % Cells labeled for a cell type that express lacZ (column 5) and, conversely, lacZ + cells that express a cell type label (column 7) were quantified.  Table 1) in the hilus shows no colocalization. (K, 25X, bars 50 µm). In situ hybridization images were obtained from Allen Mouse Brain Atlas (http://mouse. brain-map.org) 21  www.nature.com/scientificreports www.nature.com/scientificreports/ mice to induce the ablation of the conditional Shh-nlacZ allele. (Fig. 9AA,BB). In this mouse line, Cre activity removes exons 2 and 3 of the Shh gene and the nlacZ marker producing a Shh null allele and allowing for convenient identification and quantification of cells with ablated Shh. At day 45 following virus injection, animals were euthanized, and their brains perfused for histological analysis. Quantification of viral expressed GFP revealed that the virus was expressed in CR + cells in the central hilus (Fig. 9G,H,K,L). We observed a reduction in CR staining intensity inversely proportional to GFP expression (Fig. 9CC.G-T). The expression of nlacZ in the DG was almost completely abolished among infected cells as compared to controls (Fig. 10A,B and E), while nlacZ expression in the cerebellum was unchanged (Fig. 10C,D and F), revealing the anatomical selectivity and efficiency of the viral infection mediated ablation of Shh. As expected, there was a marked reduction of the Shh-protein detected  21 , which resemble the expression of Shh-nlacZ and CR + cells (bar 100 m) (mouse P57) (Bar = 100 µm). (B) Double fluorescent in situ hybridization for GAD1 and CR (http://mouse.brain-map.org) 21 showing colocalization of the two mRNAs (CR/GAD-1 = 71.8 ± 5.2%; GAD-1/CR = 50.2 ± 2.4%,estimated from 2 mice, P56) (Bar = 45 µm). (C) IHC staining for GluR2/3 and Shh-nlacZ showing an orthogonal view to demonstrate colocalization (see Table 1, columm-7). (100X, Bar = 20 µm). For the cell numbers marked by GluR 2/3 and LacZ see Table 1 www.nature.com/scientificreports www.nature.com/scientificreports/ by immunohistochemical staining in the AAV injected mice compared to control mice injected with vehicle in the same hemisphere (Fig. 11B,F and T). Further, and consistent with reduced Shh signaling strength, we found that the expression level of Ptc protein, a transcriptional target of Shh signaling 31 , was downregulated by 80% (Fig. 11M,Q and S). Quantification of CR + cells showed a pronounced, 60%, reduction in CR + cell numbers (Fig. 11C,G and R). The average surface area of hilar CR + cells was also drastically reduced from 285.2 µm 2 to 60.8 µm 2 (Fig. 11Y). Taken together, these results indicate that viral Cre mediated ablation of Shh abolished most of the hilar Shh signaling, which, in turn, critically compromised survival and/or marker expression of hilar CR + cells.
Ablation of Shh in cR + neurons results in increased SGZ proliferation and neurogenesis. We next tested whether the ablation of Shh from hilar MCs would impact neurogenesis. Contrary to our expectations, the decrease in Shh expression in the hilus increased SGZ proliferation as shown by EdU incorporation (Fig. 11K,O and X) and Ki67 staining (Fig. 11L,P and ZZ). Quantification of GFAP + cells in the hilus shows a moderate decrease in the number of GFAP + cells in the virus-injected hilus leading to altered GFAP + scaffolding www.nature.com/scientificreports www.nature.com/scientificreports/ in the SGZ (Fig. 11J,N and V) and reduced numbers of GFAP + , EdU + NSCs (Fig. 12A-P and Q). Importantly, there was an increase in the number of small CR + cells located in the SGZ (Fig. 12S,T and W). These small triangular CR + cells in the SGZ and in the inner granular cell layer have been described as new born or immature neurons that transitorily express calretinin during maturation 32,33 . Consistently, we found an increase in the numbers of DCX + cells (Fig. 12U,V and X), a marker for immature neurons 34 . Noteworthy, these newly formed CR + and DCX + cells were mostly located towards the granular cell layer (GCL) and the molecular layer (ML) (Fig. 12T,V)  www.nature.com/scientificreports www.nature.com/scientificreports/ forming a pattern distinctly different from the increase in hilar CR + cells observed during KA-induced epileptiform activity (Fig. 5B). Together, we find that the hilar ablation of Shh results in an increased production of neuronal precursor cells that migrate preferentially into the GCL while the KA induced up regulation of Shh expression is associated with migration of immature CR + cells into the hilus from the GCL.

Discussion
Here we identified hilar MCs as a prominent source of Shh within the adult hippocampus. Our data indicates that Shh signaling originating from MCs is critical for their survival upon excitotoxic insults and the modulation of dentate gyrus neurogenesis.
These studies were enabled by a sensitive gene expression tracer allele for Shh which was produced by homologous recombination resulting in the continuous bicistronic transcription of Shh and nlacZ genes under the control of the endogenous Shh promoter 13 . The faithfulness of this recombinant allele during development and in the adult brain was extensively verified 13,[35][36][37][38][39][40] . A recent study by Ortega et al. 41 confirms the pattern of Shh expression from the wt allele and from the gene expression tracer allele in the nigro-striatal system. Likewise, in the cerebellum, Shh-nlacZ was consistently expressed in large GABAergic neurons known as Purkinje cells (Fig. 13A), in agreement with previous reports 42 , further corroborating the fidelity of the Shh-nlacZ gene expression tracer allele. Using this gene expression tracer allele we found that among Shh-nlacZ + cells in the hilus, more than 90% co-expressed calretinin (CR) and GluR 2/3, the neurochemical signature of mossy cells. Among all CR + neurons, only the large multipolar neurons in the hilus, which traditionally have been considered mossy cells 14 , expressed Shh.  Table 1) suggesting separated cellular sources (63X, 25 µm). (F) Shh-nlacZ (red arrow) and Smo (yellow arrow) do not co-express in the DG (for quantification see Table 1) suggesting independent cellular sources (63X, 25 µm). (G) Ptc-1 is expressed by olidodendrocytes (PLP/Shh-nlacZ, 92.1 ± 4.7%, n = 3 × 10 slices) that do not co-express Shh (see www.nature.com/scientificreports www.nature.com/scientificreports/ www.nature.com/scientificreports www.nature.com/scientificreports/ Mossy cells are the only excitatory cell type in the hippocampus that can be considered a type of "feedback interneuron" 43 . In fact, some authors have cataloged MCs as excitatory interneurons 44,45 . As an interneuron, a single MC receives convergent inputs from granular cells through mossy fibers, sending back short but highly distributed signals to a proportionally greater number of granule cells. We also show here that MCs express CR and innervate profusely other MCs -displaying a pattern of short reciprocal MC to MC projections (Fig. 7B). These results agree with previous EM findings 46 , and are in line with the finding that reciprocal connections are a distinctive feature of calretinin + neurons 30 . While there is no conclusive proof for Mossy to Mossy cell innervation, our data are consistent with results from Wenzel et al. 46 who argued for cross innervation by collaterals. Although many MCs extend their projection to the DG inner molecular layer, MCs form a plexus of exuberant axon collaterals within the hilus, which has been recognized in anatomical 47 and electrophysiological 16,48,49 studies. These results also show that MCs are immunoreactive to GABA and GAD proteins, which does not necessarily imply that the cells are functionally GABAergic but, rather that mossy cells express GABA like other glutamatergic neurons [50][51][52] .
The neural cell damage in Shh-expressing cells and massive cell loss induced by Shh ablation revealed that Shh expression is critical for the long-term survival of MCs. Conversely, Shh + signaling was increased during the KA-induced hyperactivity. This upregulation of Shh and a concomitant increased expression of the Shh effector Smo, conferred a survival advantage to CR + neurons expressing Shh, since these cells increased in numbers. Exaggerated neural activity, inflammation and persistent oxidative stress may lead to Shh/Smo upregulation as observed during amphetamine treatment 53 , facial nerve axotomy 54 , hypoxia 55 or ischemic stroke 56 . Nevertheless, the upregulation of Shh during KA-induced neural hyperactivity did not seem to affect SGZ proliferation even when Smo signal was upregulated in the SGZ (Fig. 8B,K). Bragina et al. 57 showed that rising Shh signaling by increasing Smo agonist (SAG) concentrations from 1 nM to 5 nM decreased proliferation from significantly high levels (at 1 nM) back to control levels (at 5 nM) in hippocampal cell cultures. Further, when Shh concentrations www.nature.com/scientificreports www.nature.com/scientificreports/ increased from 10 nM to 50 nM significant cell death was observed in a cerebellar primary culture. Thus, it is possible that Shh's effects on proliferation in vivo are modulated in narrow time dependent concentration ranges. Shh may be a factor that maintains or stimulates proliferation at low concentration levels but an increase over some concentration limits or for longer periods of time may lead to inhibition.
The association between Shh upregulation and decreased -instead of increased-neurogenesis agree with previous studies in the SVZ. Thus, activation of the Shh pathway, through the conditional ablation of Ptc, led to a dramatic decrease in neurogenesis in the SVZ. The neurogenesis blockade was related to a shift in NSC division mode from asymmetric to symmetric 58 . Sustained activation of the Shh pathway using this strategy also induced a progressive increase in NSC numbers in the quiescent state and a marked reduction of the activated NSC pool, leading to an almost complete exhaustion of neurogenesis 59 . Our results in the SGZ agree with the studies in the SVZ. Collectively, these studies suggest that the response of Shh-producing neurons is adapted to preserve NSCs under changing conditions in the neurogenic niches. Under normal conditions, basal levels of Shh would act as a trophic agent for quiescent-NSCs allowing them to homeostatic be converted to activated-NCSs and increasing the production of intermediary progenitors. Under excitotoxic conditions, Shh is upregulated (e.g. by mossy cells in the hilus), which lead to inhibition of the NSCs activation process and a decline in neurogenesis. This produces the accumulation of quiescent NSCs 59 . Quiescent NSCs are more resistant to toxic conditions because replicate slowly 58 . The consequence of this Shh-concentration dependent effect is that NSCs are preserved in the neurogenic niches.
During the second week post KA-injection we observed a hyperexcitability syndrome characterized by stereotypes, forelimb clonus (Racine's stage 3) and electrographic abnormalities such as high-amplitude bursting and hyper-synchronized spiking. During this epileptogenic process, the MCs survived and there was an upregulation of Shh and Smo. In more severe epilepsy, i.e. when KA is directly administered into the DG and epilepsy stages 4/5 become persistent, MC death might occur, in which case, a fall in Shh levels as well as an increase in neurogenesis could be expected 60 . However, while some authors have considered MCs "vulnerable", others find that the loss of both MCs and interneurons is equivalent in epilepsy 61 . Furthermore, in humans with epilepsy, Seress et al. 62 found that MCs were present in the hilus of the DG when most pyramidal neurons of the CA1 and CA3 areas were lost.
The neurogenic response to neurotoxic insults is biphasic: the immediate response is an increase in neurogenesis which is followed, after a variable time span, dependent on experimental conditions and species, by a decline in neurogenesis. Low levels of proliferation and neurogenesis persist chronically. In the KA model, several groups www.nature.com/scientificreports www.nature.com/scientificreports/ using intracerebral KA administration have showed that proliferation increases during the first 3 days after the KA injections. By the end of the first week after the injection, there is a fall in the levels of SGZ proliferation. After 7 days of the injection, the levels of proliferation/neurogenesis are normal or reduced [26][27][28]60 , while a deficiency in proliferation and neurogenesis persist chronically 26,27,60 . In addition, the initial increase in proliferation after the KA injection have been associated with expansion of the glial pool rather than the neuronal pool 26,60 . In these KA-studies, Shh levels were not measured at different stages of the experiments and further investigations are required to establish whether there is a consistent relationship between Shh levels and neurogenesis in this model.
Conversely, the dramatic reduction in Shh protein levels after the ablation of the Shh gene leads to neurodegeneration, atrophy and death of MCs and to an increased proliferation and neurogenesis in the SGZ. Thus, in the SGZ adjacent to hilar CR cells, where the ablation took place, we found significant increases in EdU, Ki67 (Fig. 11), and immature neuron markers (CR and DCX) in the SGZ (Fig. 12). The number of GFAP + EdU + radial glia cells were diminished, and this could be interpreted as a depletion of NSCs due to the massive and prolonged proliferative response 60 (Fig. 12). Interestingly, chronic ablation of MCs transiently increases the activation of NSCs (at 15 days after ablation), leading to a NSCs depletion at 42 days after ablation 20 , which supports the present finding that NSCs are depleted 45 days after the initial CRE induced gene ablation.
In the SVZ of the adult brain, it has been shown that Shh exerts both positive and negative regulatory actions on neurogenesis through Smo and Gli3, respectively. NSC proliferation and neurogenesis appear to be dominated in the SVZ by Gli3R repressor activity. Thus, removing Gli3 in Smo conditional mutants largely rescues neurogenesis, while, expression of a constitutive GLI3(R) abrogates neurogenesis 63 . If a similar signaling mechanism would take place in the SGZ, the downregulation of Shh could lead to decreased repressor activity and increased proliferation. Molecular interactions of this type may help explain why the effects of a Shh deletion (in Shh sources) may greatly differ from the effects of a Smo deletion (in Shh target cells) 6 . www.nature.com/scientificreports www.nature.com/scientificreports/ The present studies are the first to identify the cells that synthetize Shh in the ventral adult hippocampus, an area where most MCs are located 14 . From horizontal brain section, we visualized the areas where the hippocampus was cross sectioned to accurately localize DG sub-regions (see Fig. 1A and methods). Li et al. 6 , using a Shh-gfpcre transgenic, followed the Shh lineage and found that CR neurons were actively producing Shh in the hippocampus at P15. MCs were recognized as CR + cells in the dorsal but not the ventral DG at P15. However, these authors did not assess whether these putative MCs were positive for either GABA or glutamate markers, thus leaving their identity ambiguous. Several authors have found that mossy cells are CR + only in the ventral DG and not in the dorsal DG 14,25,32 , while at dorsal DG levels, calretinin immunoreactivity was limited largely to a subpopulation of interneurons 64 . Liu et al. 32 identified in the hilus many CR immunoreactive cells as GABAergic www.nature.com/scientificreports www.nature.com/scientificreports/ and some that express both GABA and glutamate markers. Our results show that cells expressing CR, GluR2/3 or CGRP also co-stained for α-GAD and α-GABA antibodies. Among lacZ + cells, over 90% were positive for GAD, CR, GluR2/3 or CGRP (Table 1, columm-7). Further, a high GABA-CR colocalization can be observed in double in situ hybridization (FISH) images from 2 mice (P57) published in the Allan Atlas 21 . From these images, we estimated that among GAD-1 + cells, about 70% were positive for CR (Fig. 3B). These observations support our finding using IHC that most hilar CR + cells in the ventral hilus are GAD1 + (GAD67) (Figs 4 and 5). There are many factors related to IHC technique that can account for the discrepancy with other groups. An advantage of our methodology was the strong, selectively nuclear, immune-reactivity for lacZ, These lacZ + cells were immunoreactive within the cytoplasm to α-GAD antibodies at conventional concentrations (1:500 to 1:1000) and using several commercial brands including α-GAD65/67 (ABcam), α-GAD67 and α-GAD65 (Santa Cruz) and α-GABA (Sigma-Aldrich). In addition, we showed that MCs expressing nlacZ strongly co-localized with GluR2/3 (Fig. 3C,G) and CGRP (Supplemental Fig. S2). GluR2/3 and CGRP cytoplasmic signals highly co-localized with GABA markers in the hilus (Fig. 3D-F,H; Supplemental Fig. S2). Collectively, these observations suggest that MCs express both glutamatergic and GABAergic markers like medium spiny neurons (MSNs) in the striatum 65 or the granular cells in the dentate gyrus 50 .
In addition to the well-known effects of Shh on NSCs in the SGZ, Shh could be delivered to the GCL via axonal transport by Shh expressing MCs and could influence cell differentiation/maturation and the continuous integration of granular cells produced by neurogenesis into the hippocampal circuits. MCs have recently been implicated in place fields neural processing 66 , spatial exploration and pattern separation 67 and in generating a special type of LTP 68 . Our findings will prompt follow up studies that will seek to clarify whether hippocampal learning and memory processes might depend on Shh expression by MCs and whether Shh might regulate structural plasticity in hippocampal circuitry.
In conclusion, we have found that Shh expression in MCs is dynamic and can be upregulated during epileptogenesis to enhance survival. Furthermore, we show that Shh expression by MCs is indispensable for their survival. Finally, Shh ablation led to reactive proliferation of immature neurons and depletion of NSCs suggesting that MC derived Shh has multiple regulatory functions during neurogenesis. As MCs are implicated in learning and memory, the modulation of Shh could be a key strategy to modulate plasticity and neurogenesis. This suggests that the modulation of Shh expression in MCs might be a potential approach to generate therapies aimed at protecting neural circuits in epilepsy and excitotoxicity and treating cognitive and emotional disorders in which altered neurogenesis is a major component. www.nature.com/scientificreports www.nature.com/scientificreports/ Methods Mouse strains. The Shh-nlacZ allele was generated by homologous recombination in ES cells by Dr. A.
Kottmann and T. Jessell. Additional details of the allele construction are described in Gonzalez-Reyes et al. 13 .
Transgenic mice expressing the oligodendrocyte reporter PLP-EGFP were generated by Dr. W. Macklin and described in Mallon et al. 69 .
Animals were housed no more than 5 adult animals per cage and maintained in a SPF room under light (12-h light/12-h dark cycle), temperature and humidity controlled conditions. For acute in vivo experiments, adult male animals were used (range from P60 to P90). All experimental procedures performed in this study followed the NIH animal use guidelines and were approved by the Institutional Animal Care and Use Committee (IACUC) at Case Western Reserve University.
KA injections. The dose of KA was fractioned (5 mg/kg IP every 30 min) until a maximum of 35 mg/kg; mice were scored using a Racine's severity scale 24 . Only mice that reach status epilepticus during administration of KA were selected for histological studies. The histological studies were performed 2 weeks after the KA injections. BRDU injections. Two doses of BRDU (30 mg/kg each) separated by three hours-intervals were injected (IP) seven days before euthanasia. BRDU staining procedures are described below.
For BRDU staining the procedure was the same as above but the slices were pretreated with citrate buffer and heated to 99 °C in a water bath before blocking 71 . Chromogenic immunohistochemistry was performed using the above-mentioned primary antibodies and developed using betazoid DAB, vulcan fast red and hematoxylin (BioCare Medical Kit).
Cells were quantified from 10 horizontal sections of 40 µm thickness each that were spaced 160 μm apart (4-section interval) covering the entire the dorsal-ventral-and anterior to posterior-extent of the hilus. The horizontal slices allowed to sample the volume represented in Fig. 1. The number of cells was estimated using a stereological approach 13 by two independent observers blinded to the experimental groups. Fluorescent colocalization was demonstrated using Z-stack reconstructions of the whole cell allowing orthogonal views at 63X (Zeiss LSM 510 microscope) or 100X (Olympus FV-1000 microscope) magnification. Cells were counted exhaustively from images with a full view of the hilar region (25X). To count double-labeled cells we took Z-stacks at 1μm-steps across the section and generate a Z-projection (vertically superimposed) into a single x-y plane (FV10-ASW software). The projected image was electronically amplified 4 times (to 100X) on the computer screen. Using the image J cell-counting tool only cells with clear morphology and delineated nucleus were counted and marked to avoid repetitions. Co-localization was reported as average ± SEM of cells that co-express two markers for every 100 cells expressing one of the markers (Table 1). To determine optical density, the channel corresponding to the protein was selected. Using image J, the integrated optical density (IOD) and the cell surface area (CSA) were measured and the IOD/CSA ratio was calculated for each immune positive cell. This value was then average and normalized to controls.
Data analysis and statistics. Student's t-tests were used to compare either two repeated (paired-t-test) or independent (unpaired-t-test) measures. Test results and number of animals per group are reported in the figures or in the figure legends. A statistical significance criterion of α = 0.05 was used for all tests. Results are shown as mean ± standard error (SEM). Statistical test results and group sizes are reported in the figure legends.