RBP-J is not required for granule neuron progenitor development and medulloblastoma initiated by Hedgehog pathway activation in the external germinal layer

Background The Notch signalling pathway plays crucial roles in neural development, functioning by preventing premature differentiation and promotion of glial cell fates. In the developing cerebellum Notch pathway components are expressed in granule neuron progenitors of the external germinal layer (EGL) but the precise function of Notch in these cells is unclear. The Hedgehog pathway is also crucial in cerebellar development, mainly via control of the cell cycle, and persistent activation of the pathways leads to the cerebellar tumour medulloblastoma. Interactions between Hedgehog and Notch have been reported in normal brain development as well as in Hedgehog pathway induced medulloblastoma but the molecular details of this interaction are not known and we investigate here the role of Notch signalling in the development of the EGL and the intersection between the two pathways in cerebellar granule neuron progenitors and in medulloblastoma. Results RBP-J is the major downstream effector of all four mammalian Notch receptors and the RBP-J conditional mouse facilitates inactivation of canonical Notch signals. Patched1 is a negative regulator of Hedgehog signalling and the Patched1 conditional mouse is widely used to activate Hedgehog signalling via Patched1 deletion in specific cell types. The conditional mouse lines were crossed with a Math1-Cre line to delete the two genes in granule neuron progenitors from embryonic day 10.5. While deletion of only Patched1 as well as Patched1 together with RBP-J leads to formation of medulloblastoma concomitant with disorganisation of cell layers, loss of RBP-J from granule neuron progenitors has no obvious effect on overall cerebellar morphology or differentiation and maturation of the different cerebellar cell types. Conclusions Our results suggest that even though Notch signalling has been shown to play important roles in cerebellar development, signalling via RBP-J is surprisingly not required in granule neuron progenitors. Furthermore, RBP-J inactivation in these cells does not influence the formation of medulloblastoma initiated by Hedgehog pathway activation. This may suggest a requirement of Notch in cerebellar development at a different developmental stage or in a different cell type than examined here - for example, in the neural stem cells of the ventricular zone. In addition, it remains a possibility that, in granule neuron progenitors, Notch may signal via an alternative pathway without the requirement for RBP-J.

Results: RBP-J is the major downstream effector of all four mammalian Notch receptors and the RBP-J conditional mouse facilitates inactivation of canonical Notch signals. Patched1 is a negative regulator of Hedgehog signalling and the Patched1 conditional mouse is widely used to activate Hedgehog signalling via Patched1 deletion in specific cell types. The conditional mouse lines were crossed with a Math1-Cre line to delete the two genes in granule neuron progenitors from embryonic day 10.5. While deletion of only Patched1 as well as Patched1 together with RBP-J leads to formation of medulloblastoma concomitant with disorganisation of cell layers, loss of RBP-J from granule neuron progenitors has no obvious effect on overall cerebellar morphology or differentiation and maturation of the different cerebellar cell types. Conclusions: Our results suggest that even though Notch signalling has been shown to play important roles in cerebellar development, signalling via RBP-J is surprisingly not required in granule neuron progenitors. Furthermore, RBP-J inactivation in these cells does not influence the formation of medulloblastoma initiated by Hedgehog pathway activation. This may suggest a requirement of Notch in cerebellar development at a different developmental stage or in a different cell type than examined here -for example, in the neural stem cells of the ventricular zone. In addition, it remains a possibility that, in granule neuron progenitors, Notch may signal via an alternative pathway without the requirement for RBP-J.

Background
The Notch signalling pathway plays crucial roles in brain development and in the cerebellum in particular. The main function of Notch is preventing (premature) differentiation of neural progenitor cells and at later stages of neural development promoting glial over neuronal cell fates [1][2][3][4][5]. In vitro analyses as well as in vivo murine models have shown that, in the cerebellum, the Notch pathway influences the development of Bergmann glia and the differentiation of granule neuron progenitors (GNPs) [6][7][8][9][10].
There are four Notch receptor paralogues  in vertebrates, all of which bind transmembrane ligands of the Delta-like and Jagged family [3,11,12]. Upon binding of ligand, the Notch receptor is proteolytically cleaved and its intracellular domain translocates to the nucleus where it forms a complex with several cofactors and the DNA binding protein RBP-J (Recombination signal binding protein for immunoglobulin kappa J region; also called CSL for CBF1 in mammalians, Suppressor of Hairless in Drosophila, Lag-1 in C. elegans), the major canonical downstream effector of the Notch pathway [3,13,14]. The complex then facilitates the expression of target genes, including basic helix-loophelix transcription factors, such as Hes1 and Hes5, which act as transcriptional repressors of proneural genes [15][16][17][18]. Possible alternative 'non-canonical' downstream pathways of Notch that are RBP-J independent have also been reported [19][20][21][22].
Of the Notch receptors, Notch1, Notch2 and Notch3 are most important for cerebellar development and the roles of each receptor vary at different developmental stages [7,23]. During granule neuron development, Notch2 is expressed predominantly in proliferating GNPs of the external germinal layer (EGL), and Notch1 in postmitotic differentiating cells in the internal granule layer (IGL) [10,24]. Of the five Notch ligands, Delta-like 3 (Dll3) is mainly expressed in GNPs while Jagged2 is expressed in differentiating and mature neurons of the IGL [25]. The specific expression patterns of Notch receptors and ligands in cerebellar development suggest diverse and cell-type-specific roles for the Notch pathway; however, detailed functions of Notch components in cerebellar cell types are not known. Accordingly, we examine here the role of canonical Notch signalling in cerebellar GNPs.
In addition to its function in normal brain development, the Notch pathway has also been implicated in tumour formation, in particular in medulloblastoma. Notch can act as an oncogene or tumour suppressor and different receptors appear to have varying tumorigenicity [24]. In various mouse models of medulloblastoma initiated by perturbation of the Hedgehog (Hh) pathway, Notch pathway components are upregulated, indicating both a potentially oncogenic role for Notch and possibly an interaction between the two pathways [26,27]. Hh signalling controls many patterning events of embryonic development via cell cycle regulation, and deregulation of the pathway often results in tumorigenesis. In the absence of Hh ligand the transmembrane receptor Patched1 (Ptc1) inhibits Smoothened (Smo) while Hh binding results in a release of the repression of Smo by Ptc1 and in a cascade of phosphorylation events [28,29]. Consequently, active Gli transcription factors translocate to the nucleus where they can activate expression of target genes such as Cyclins D1, D2 and E and N-Myc [30][31][32][33][34].
We have shown recently that the Hh and Notch pathways interact during cerebellar development in ventricular zone stem cells but this interaction remains to be defined at other stages of cerebellar development and in other cerebellar cell types [35]. In particular, even though Notch pathway components are clearly expressed in the EGL of the developing cerebellum at a time when GNPs are dividing rapidly and then differentiating under the influence of Purkinje-neuron-derived Sonic hedgehog (Shh), it is not known whether Notch signalling is actually required for appropriate formation of granule neurons.
We addressed the interaction of Notch and Hh signalling during granule neuron development and in medulloblastoma formed from GNPs by adopting a genetic approach in vivo to ablate Notch signalling in the Ptc1 conditional mouse model, which is a widely accepted and relevant model for human medulloblastoma [36][37][38]. Since the developing EGL expresses multiple components of the Notch signalling pathway, we took the approach to inactivate the common pathway effector RBP-J in order to ablate canonical Notch receptor signalling. This strategy has been previously applied in analyses of Notch signalling and has proved to be highly effective [8,35,39,40]. A Creexpressing transgenic line facilitated deletion of the two genes in Math1+ cells (Math1-Cre) of the rhombic lip and EGL [38]. Surprisingly, we found no requirement for RBP-J for the proliferation, differentiation or migration of granule neuron precursors. Also, loss of RBP-J concomitant with activation of the Hh pathway in GNPs did not block medulloblastoma formation. Our data suggest that the function of Notch signalling, as distinct from the expression of Notch signalling components during cerebellar development, is restricted to cell types other than GNPs, that is, stem cells of the ventricular zone and cerebellar cells originating from there.

Results
Math1-Cre efficiently inactivates the Patched1 and RBP-J genes in the external germinal layer Math1-Cre results in Cre expression in Math1-positive cells of the rhombic lip from embryonic day 10.5 (E10.5), and in GNPs in the EGL that have not begun to differentiate into mature neurons [38]. To validate the model used in this study, we first confirmed deletion of both the Patched1 and RBP-J alleles by Math1-Cre. GNPs isolated from P7 RBP-J lox/lox ; Math1-Cre mice show high efficiency of floxing by RT-PCR with primers flanking the excised region (exons 6 and 7; Figure 1A). Floxing of RBP-J was confirmed by Taqman and in situ analysis for RBP-J mRNA, which showed significant loss of wild-type RBP-J mRNA in mutant GNPs compared to controls ( Figure 1B,C). In addition, we assayed mRNA expression of the direct Notch target gene Hey2 by quantitative real time PCR (Additional file 1). Hey2 mRNA is significantly decreased after RBP-J deletion (P = 0.0157). The Patched1 allele was also deleted with high efficiency by Math1-Cre, as shown by RT-PCR with primers flanking the excised exon 3. Pathway activation due to Patched1 deletion was confirmed by Taqman analysis showing significant upregulation of the universal Hh target Gli1 in Ptc1 lox/lox ; Math1-Cre GNPs compared to Ptc1 lox/lox GNPs ( Figure 1D,E). To further validate our model, we performed in situ hybridisation and immunofluorescence staining to characterise the expression of Hh and Notch pathway targets in the mutants. In situ analysis for Gli1 supports Hh activation in the EGL of Ptc1 deleted as well as Ptc1 and RBP-J deleted mice ( Figure 2B,D). Interestingly, our in situ analysis also suggests a reduction of Gli1 expression in RBP-J lox/lox ; Math1-Cre cerebella, indicating that RBP-J deleted GNPs might not be able to adequately respond to Shh ligand secreted by Purkinje neurons ( Figure 2C). The canonical Notch target Hes1 (Hairy enhancer of split) is expressed in the outer EGL of control mice ( Figure 2E). However, Hes1 has also been reported to be a target of Hh signalling [41,42] and the Hes1-positive area appears expanded in response to Hh activation, with staining throughout the EGL of Ptc1 deleted brains ( Figure 2F). RBP-J deletion neither leads to loss of Hes1 nor can it overcome the increase of Hes1-positive cells in the Ptc1 deleted cerebellum ( Figure 2G,H). The canonical Notch target Hes5 is expressed in the outer EGL of control mice, with an expansion towards the inner EGL in Hh activated mice, similar to Hes1 ( Figure 2I,J). RBP-J deletion results in loss of Hes5 expression; however, some cells of Ptc1 lox/lox ;RBP-J lox/lox ;Math1-Cre EGL express Hes5, possibly due to proliferation of a small number of GNPs with incomplete deletion of RBP-J in these mutants ( Figure 2K,L). Taken together, our data confirm that Math1-Cre efficiently deletes both Patched1 and RBP-J alleles and consequently leads to GNP-specific activation of Hh and inactivation of Notch signalling, respectively.

Loss of RBP-J in GNPs does not alter cerebellar morphology
To examine the effects of Hh pathway activation and Notch inactivation on the development and differentiation of cerebellar cell types, we first analysed mutant brains at three developmental stages (E18.5, postnatal day 7 (P7) and P21) histologically by haematoxylin and eosin staining. At E18.5 all mutants are morphologically similar to control cerebellum, with a normal appearing EGL and beginning foliation ( Figure 3A-D). By P7, control cerebellum has developed ten folia and GNPs have begun to differentiate and form an IGL, which is fully developed by P21 with no EGL remaining ( Figure 3E,I). RBP-J deletion has no discernable effect on morphology throughout cerebellar development ( Figure 3G,K). Ptc1 deletion results in an increase of GNPs in the EGL and, therefore, a thickening of this region, which develops into medulloblastoma by P21. Foliation appears largely unaffected, with the correct number of lobes present, and some GNPs have been able to differentiate and migrate to form an IGL. The boundaries between the EGL, molecular layer (ML) and IGL appear disturbed, likely due to the excessive expansion of GNPs ( Figure  3F,J). Notch inhibition by RBP-J deletion does not alter this phenotype ( Figure 3H,L). In summary, GNP-specific deletion of RBP-J does not affect the morphology of the cerebellum with or without Ptc1 deletion.

RBP-J deletion does not influence the development of Hedgehog-pathway-induced medulloblastoma
Mice with GNP-specific deletion of the negative Hh regulator Ptc1 develop severe medulloblastoma by P21 in 100% of cases and this phenotype, as well as the overall morphology ( Figure 3J,L) are not altered by loss of canonical Notch signalling. We examined tumours of both genotypes (Ptc1 lox/lox ;Math1-Cre and Ptc1 lox/lox ; RBP-J lox/lox ;Math1-Cre) in more detail and observed that the tumour mass is made up of small round granular cells interspersed with stroma in both cases ( Figure 4A, B), similar to human medulloblastoma and as previously published [38]. Both also stain positively for the granule cell marker Pax6, confirming a GNP origin of the tumours ( Figure 4C,D). Staining with proliferating cell nuclear antigen (PCNA) reveals that most cells in both genotypes are actively undergoing proliferation ( Figure  4E,F). The universal Hh target Gli1 is expressed in Ptc1 deleted tumours as well as in those with additional loss of RBP-J ( Figure 4G,H), as is Hes1, a target of both the Hh and Notch pathways ( Figure 4I,J). The canonical Notch target Hes5, however, which requires RBP-J, is expressed in Hh activated tumours but lost in most cells of tumours with activated Hh and inactivated Notch signalling ( Figure 4K,L). In conclusion, even though Notch signalling is lost in the majority of cells in Ptc1 lox/lox ; RBP-J lox/lox ;Math1-Cre tumours, as confirmed by loss of Hes5 mRNA, RBP-J deletion does not appear to influence the characteristics of medulloblastoma initiated by Ptc1 deletion in GNPs.

Differentiation of cerebellar cell types does not depend on canonical Notch signalling via RBP-J
Following the characterisation of Ptc1 and RBP-J deleted medulloblastoma, we investigated cerebellar development in more detail by using immunofluorescence staining with markers specific for different cerebellar cell types. At E18.5, GNPs have migrated from the rhombic lip to form the EGL and neuronal differentiation is in progress. BetaIII-tubulin staining confirmed the presence of immature neurons migrating away from the EGL in control embryos and in all mutants ( Figure 5A-D). GNPs in the EGL proliferate in response to Shh, which is secreted by Purkinje neurons, and at P7 the peak of proliferation is reached. PCNA staining confirmed that cells in the outer EGL of the control and RBP-J deleted cerebellum are mitotically active ( Figure 5E,G). Ptc1 deletion results in an expansion of the proliferative compartment, which is not influenced by loss of Notch signalling ( Figure 5F,H). However, even when Ptc1 deletion is present, GNPs in the inner EGL can differentiate and begin to express betaIII-tubulin in addition to the granule cell marker Pax6 ( Figure 5J). The same is evident when the Notch pathway is inhibited in addition to Hh activation ( Figure  5L) and differentiation of RBP-J lox/lox ;Math1-Cre GNPs into neurons is initiated normally (Figure 5K,I). GNPs isolated from P7 to P8 cerebellum and stained for PCNA, betaIII-tubulin, NeuN (Neuronal nuclei) and by TUNEL in vitro confirm no obvious difference in proliferation, differentiation and apoptosis rates between RBP-J lox/lox controls and RBP-J lox/lox ;Math1-Cre mutants (Additional file 2A-H). Culture of GNPs for 3 days in the absence and presence of recombinant Shh-N reveals that RBP-J deletion also has no influence on the proliferative response to Shh as assayed by immunofluorescence for PCNA (Additional file 2I-L) and cell count (Additional file 2M). By P21 most cerebellar differentiation processes are complete and the layering of the mature cerebellum  is present. Mature, NeuN-positive neurons populate the IGL while Bergman glial fibres, marked by glial fibrillary acidic protein (GFAP), span the ML in parallel to each other ( Figure 5M). Notch ablation does not influence either of these cell types ( Figure 5O). Ptc1 deletion, however, results in a severe disorganisation of Bergman glial fibres as well as the presence of NeuN-positive cells in the remainder of the expanded EGL. These cells are likely to be differentiated granule neurons that were hindered from migrating into the IGL by the bulk of proliferating tumour cells ( Figure 5N). The same phenotype is evident in Hh-activated cerebella with additional RBP-J deletion ( Figure 5P). Purkinje neurons, marked by Calbindin, are organised in a single layer around the surface of the IGL with their fibres reaching into the ML in control brains as well as in Notch deficient cerebella ( Figure 5Q,S). In Hh-activated as well as Hh-activated and Notch-inactivated brains the Purkinje neuron layer is disrupted and lacks organisation of cell bodies as well as orientation of their fibres ( Figure 5R,T). Overall, the differentiation of cerebellar cell types appears unaffected by RBP-J deletion and the disorganisation of cell layers in medulloblastoma initiated by Ptc1 deletion cannot be overcome by canonical Notch inactivation.

GNP-specific deletion of RBP-J and Ptc1 does not influence the cerebellar stem and progenitor cell pool
Next we asked whether Hh activation and/or Notch inactivation in GNPs influences the stem and progenitor cell pool of developing cerebella. To examine cerebellar stem cell properties, we used Sox2, which marks neural stem cells and Bergman glia. In controls as well as all mutants, cerebellar stem cells residing in the ventricular zone (VZ) stain positive for Sox2. In addition, all genotypes show some Sox2-positive cells throughout the cerebellum, Bergman glia originating from the VZ, which are migrating towards the ML (Additional file 3A-D). In Ptc1 lox/lox ; Math1-Cre mice background staining appears increased, although there is no apparent difference in nuclear Sox2 staining (Additional file 3B). Furthermore, we utilised the neurosphere assay, a widely used tool to examine stem/ progenitor cell numbers. The number of colonies counted in this assay is indicative of the number of stem/progenitor cells in the population but it cannot distinguish between the two. No significant difference was observed in colony numbers after deletion of Ptc1 or RBP-J compared to controls after 5 days of incubation of single cell suspensions of P7 cerebella (Additional file 3E). Taken together, GNPspecific deletion of Ptc1 and RBP-J appears to have no effect on the cerebellar stem/progenitor cell pool.

Discussion
The Notch signalling pathway influences cerebellar development, in particular the differentiation of neurons and glia. Loss of the Notch1 receptor or the ligand Jagged1 in neuroepithelial cells results in premature differentiation of GNPs and defects in neuronal migration, and RBP-J plays a crucial role in the development and migration of Bergman glia [8,43,44]. Signalling from the Notch2 receptor appears to have a role opposing that of Notch1 in GNPs, promoting proliferation while inhibiting differentiation [10,24]. Several studies have also suggested a role for Notch signalling in the formation of medulloblastoma [26,27,45]. Multiple components of the Notch pathway are expressed in the EGL, so here we investigate the role of canonical Notch signalling in GNPs by deletion of the common Notch effector RBP-J in Math1+ cells and the consequences of Notch signal inactivation on the initiation and development of Hhpathway-dependent medulloblastoma.
First we confirmed the validity of our model using RT-PCR, quantitative real-time PCR and in situ analysis, showing high efficiency of floxing for both the Ptc1 and RBP-J alleles, resulting in Hh pathway upregulation and Notch inactivation in GNPs, respectively (Figures 1 and  2; Additional file 1). The overall morphology of cerebella with RBP-J deleted GNPs appeared normal, with foliation and layer formation identical to control cerebella. Hh activation by Ptc1 deletion had no effect until after birth, when GNPs in the EGL excessively proliferate and lead to medulloblastoma formation by P21 in all individuals, as we have shown previously [38] (Figure 3). Even though RBP-J deletion did not appear to alter overall cerebellar development, we then asked if it had a more subtle impact on differentiation or migration of cerebellar cell lineages. Notch signalling has been shown to be critical for neuronal and glial differentiation and migration in cerebellar development and we therefore examined the different cell types of the cerebellum using lineage-specific markers. We found that loss of RBP-J from GNPs has no effect on neuronal differentiation and migration. This finding appears to be in direct contrast to a report demonstrating defects in granule cell migration after loss of the Notch ligand Jagged1 [44]. However, deletion in the above-mentioned study led to a deficit of Bergman glia, which act as migratory scaffolds for GNPs, and it was therefore found that the GNP defect was likely secondary to the loss of Bergmann glia. Furthermore, GNP-specific RBP-J deletion cannot overcome the cerebellar disorganisation resulting from Ptc1 deletion and medulloblastoma formation (Figures 4 and 5). This raises the question of whether Notch signalling in GNPs is transduced via alternative pathways that are independent of canonical Notch signalling via RBP-J. Indeed, work by Mizutani et al. [46] indicates that, in contrast to neural stem cells, more committed neural progenitor cells may use alternative Notch pathways without a requirement for RBP-J. In the absence of Notch signals RBP-J could potentially function as a transcriptional repressor, so loss of RBP-J may have an effect even on cells that do not usually require active Notch signalling. It has been suggested that this function of RBP-J does not play a crucial role in mammals and the absence of a phenotype after RBP-J deletion in our model confirms this specifically for GNPs [13]. An alternative explanation for the lack of an effect of RBP-J deletion on GNPs is that Notch signalling may be required at an earlier time point, before commitment to the granule cell lineage. In accordance with this, inactivation of Notch signalling at early embryonic time points has a severe impact on cerebellar development and Notch has also been implicated in controlling the balance between symmetric and asymmetric stem cell division in a number of tissues, including the brain [ [3,44], and our unpublished results]. In addition, we have shown previously that RBP-J deletion in the cerebellar VZ induces an increase in progenitor cell numbers in the niche due to a loss of stem cells and a delay in differentiation [35]. Therefore, we next examined the properties of the cerebellar stem/progenitor cell niche by Sox2 staining and neurosphere assays. Neither approach indicated an effect of RBP-J deletion (Additional file 3), likely due to the specificity of Math1-Cre, which deletes in GNPs after they have left the VZ, the cerebellar stem cell niche, and committed to the granule cell lineage.
In addition to RBP-J deletion having no impact on the differentiation of cerebellar cell types and the stem/progenitor cell niche, we also found that loss of canonical Notch signalling does not influence the formation of Hhdependent medulloblastoma (Figures 3H,L, 4, and 5D,H,L, P,T). We confirmed loss of Hes5 mRNA in the majority of tumour cells and thereby excluded the possibility that the minority of cells deleted for Ptc1 but not for RBP-J may have had a growth advantage and populated the tumour mass ( Figure 4K,L). The absence of any impact of RBP-J deletion on tumour formation was surprising as several studies have noted dysregulation of Notch pathway components in both human and murine medulloblastoma, including the transcription factor targets Hes1 and Hes5, and the expression of Hes1 is associated with poor clinical outcome [24,26,27,45,47]. Medulloblastoma cell lines treated with γ-secretase inhibitors that block Notch receptor endoproteolysis display reduced growth, clonogenicity and tumorigenicity, and γ-secretase inhibitors have been proposed as a chemotherapeutic approach to treating medulloblastoma [27,48,49]. Cerebella of mice with both Ptc1 and RBP-J deletion appear identical in morphology to those with deletion of Ptc1 alone, and expression of celltype-specific markers and all mutant mice develop severe medulloblastoma by P21. This indicates that canonical Notch signalling is likely not required for the development of medulloblastoma initiated by Hh pathway activation in GNPs. However, we cannot exclude the possibility that there might be some influence of RBP-J deletion on tumour latency, and a more extensive study with a large number of mutant mice of both genotypes would be required to address this question.

Conclusions
We have shown here that canonical Notch signalling via RBP-J is not required in GNPs and Hh-pathway-dependent medulloblastoma. A crucial role for Notch signalling in cerebellar development has been shown previously and we conclude that the involvement of the Notch pathway may be restricted to the stem/progenitor cell niche and loses influence as cells commit to the granule neuron lineage. This is despite the observation that Notch pathway components such as Notch1, Notch2, Notch3, Dll3, Jagged1, Hes1, and Hes5 are expressed in the EGL, underlining the fact that it is important not to confuse detectable expression of a signalling pathway with it necessarily functioning in that tissue. Additional work is required to characterise the role of Notch in cerebellar development, in particular to define the developmental stage(s) when it is required and to identify the utilised downstream effectors.

Mouse models
All work involving mice was performed with approval and according to guidelines of the University of Queensland Animal Ethics Committee. Mouse models used were Ptc1 conditional mice [50] and RBP-J conditional mice [51] crossed with a Math1-Cre line (kindly provided by David Rowitch).

Isolation and culture of granule neuron progenitors
Cerebellar GNPs were isolated from P7 to P8 pups as described previously [38]. In brief, cerebella were dissected and cells dissociated and triturated, followed by centrifugation through a 35 to 65% percoll gradient (Amersham Biosciences, now GE Healthcare Bio-Sciences Corp., Picataway, NJ, USA) in order to segregate granule neurons from astrocytes. Isolated GNPs were washed and used for subsequent experiments.
Culture of GNPs in the presence or absence of 3 μg/ml Shh-N (R&D Systems, Minneapolis, MN, USA) was performed at a density of 5 × 10 5 cells per well in 12-well plates with poly-L-lysine (Sigma Aldrich, St Louis, MO, USA) coated coverslips in NB-B27 media (Neurobasal with 1 mM sodium pyrovate, 2 mM L-glutamine, penicillin/streptomycin, and B27 supplement, all from Invitrogen, Carlsbad, CA, USA).

RT-PCR and quantitative RT-PCR
RNA was extracted from cells using the RNeasy Mini Kit by QIAgen (Hilden, Germany) according to the manufacturer's manual. Reverse transcription was performed using the Superscript III system by Invitrogen. The following primers were used for PCR detection of the RBP-J floxed allele: forward 5' CATCTCCAAACC CTCCAAAA 3', reverse 5' GTCCAGGAAGCTC-CATCGT 3'; and the Patched1 floxed allele: forward 5'-CACCGTAAAGGAGCGTTACCTA-3', reverse 5'-TGGTTGTGGGTCTCCTCATATT-3'. Quantitave PCR was performed with Assays on Demand by Applied Biosystems for RBP-J (Mm00770450_m1), Gli1 (Mm00494645_m1), and Hey2 (Mm00469280_m1) according to the manufacturer's protocol on the ABI Prism 7000 equipment by Applied Biosystems (Austin, TX, USA). Measurements were taken in three technical replicates and data were normalized to the housekeeping gene GAPDH (assay ID 4352339E). Statistical analysis was performed using Graphpad Prism 4 (Graphpad Software, La Jolla, CA, USA) for unpaired t-tests.

Immunofluorescence, TUNEL and haematoxylin and eosin staining
Brains were dissected (after cardiac perfusion for P7 and P21 mice) and fixed in 4% paraformaldehyde overnight. Subsequently, samples were either embedded in paraffin or cryoprotected in 30% sucrose followed by embedding in OCT compound. Antigen retrieval of deparaffinised wax tissue sections or defrosted cryosections was performed by boiling in antigen unmasking solution (Vector Laboratories, Burlingame, CA, USA). Sections were blocked in 4% horse serum, 1% bovine serum albumin and 0.2% Triton-X in phosphate-buffered saline prior to primary antibody incubation overnight at 4°C. Slides were incubated with secondary antibodies for 1 h at room temperature. For immunofluorescence, a DAPI counterstain (1:10,000; Sigma Aldrich) was performed prior to mounting with Fluorescence Mounting Media (Dako, Carpentaria, CA, USA). For histological analysis, deparaffinised and rehydrated sections were stained in haematoxylin (Vector Laboratories) and eosin Y (Sigma Aldrich) and mounted in Entellan. Antibodies used were betaIII-tubulin

In situ hybridisation
In situ hybridisation was performed as previously published [30]. In summary, probes were prepared using DIG labelled probe amplification followed by phenol/chloroform extraction and precipitation. Paraffin-embedded sections (6 μm) were treated with 2 μg/ml ProK (Roche Diagnostics) in TE buffer, fixed in 4% paraformaldehyde and acetylated. Hybridisation was performed in hybridisation buffer at 64°C for Gli1 and 65°C for Hes5 and RBP-J over night. A series of saline-sodium citrate (SSC) washes was followed by blocking and washing (DIG block and wash buffer set, Roche) before incubation with anti-DIG-AP antibody (Roche). Colour reaction was performed using 3.5 mg/ml nitroblue tetrazolium (NBT) and 1.75 mg/ml 5-bromo-4-chloro-3-indolyl-phosphate (BCIP) (both Roche) in 10% polyvinylalcohol (PVA) (Sigma). Colour formation was stopped in TE solution followed by counter staining with nuclear fast red (Vector Labs), postfixation and mounting.
Probes used were Gli1 (a gift from A Joyner, New York, NY, USA), RBP-J (a gift from T Honjo, Kyoto, Japan) and Hes5 (a gift from R Kageyama, Kyoto, Japan).

Microscopy
Light and general fluorescence microscopy were performed using an Olympus BX-51 upright microscope. Confocal images were taken on a Zeiss LSM 510 META.
Stem/progenitor cell analysis P7 cerebellar cells were harvested and subsequently dissociated. For neurosphere assays, cells were plated at a density of 1 × 10 5 cells per ml in 200 μl Neurosphere assay media (Neurosphere media containing 10% Neurocult neural stem cell proliferation supplement (Stem Cell Technologies, Tullamarine, VIC), 5% bovine serum albumin (Sigma), 1% penicillin/streptomycin) containing epidermal growth factor (20 ng/ml) in a 96-well plate. For each individual the assay was set up in triplicates. The number of spheres per well was counted 5 days after plating. Statistical analysis was performed using Graphpad Prism 4 for unpaired t-tests.