Nos2 Inactivation Promotes the Development of Medulloblastoma in Ptch1+/− Mice by Deregulation of Gap43–Dependent Granule Cell Precursor Migration

Medulloblastoma is the most common malignant brain tumor in children. A subset of medulloblastoma originates from granule cell precursors (GCPs) of the developing cerebellum and demonstrates aberrant hedgehog signaling, typically due to inactivating mutations in the receptor PTCH1, a pathomechanism recapitulated in Ptch1+/− mice. As nitric oxide may regulate GCP proliferation and differentiation, we crossed Ptch1+/− mice with mice lacking inducible nitric oxide synthase (Nos2) to investigate a possible influence on tumorigenesis. We observed a two-fold higher medulloblastoma rate in Ptch1+/− Nos2−/− mice compared to Ptch1+/− Nos2+/+ mice. To identify the molecular mechanisms underlying this finding, we performed gene expression profiling of medulloblastomas from both genotypes, as well as normal cerebellar tissue samples of different developmental stages and genotypes. Downregulation of hedgehog target genes was observed in postnatal cerebellum from Ptch1+/+ Nos2−/− mice but not from Ptch1+/− Nos2−/− mice. The most consistent effect of Nos2 deficiency was downregulation of growth-associated protein 43 (Gap43). Functional studies in neuronal progenitor cells demonstrated nitric oxide dependence of Gap43 expression and impaired migration upon Gap43 knock-down. Both effects were confirmed in situ by immunofluorescence analyses on tissue sections of the developing cerebellum. Finally, the number of proliferating GCPs at the cerebellar periphery was decreased in Ptch1+/+ Nos2−/− mice but increased in Ptch1+/− Nos2−/ − mice relative to Ptch1+/− Nos2+/+ mice. Taken together, these results indicate that Nos2 deficiency promotes medulloblastoma development in Ptch1+/− mice through retention of proliferating GCPs in the external granular layer due to reduced Gap43 expression. This study illustrates a new role of nitric oxide signaling in cerebellar development and demonstrates that the localization of pre-neoplastic cells during morphogenesis is crucial for their malignant progression.


Introduction
Medulloblastoma (MB) is a highly malignant tumor of the cerebellum that preferentially develops in children and adolescents. Although the survival rate for standard risk MB is around 70% [1] surviving patients often suffer from neurodevelopmental and cognitive side effects of the aggressive therapy [2]. Therefore, improved understanding of the molecular pathomechanisms driving MB growth is necessary to develop less toxic and more effective treatments. Recent molecular profiling studies suggested at least four MB subtypes that are associated with distinct expression profiles, genomic aberrations and clinical features [3,4]. One of these MB subtypes is characterized by aberrant activation of the hedgehog (Hh) pathway and typically corresponds to the desmoplastic (nodular) MB variant. This subtype is supposed to develop from granule cell precursors (GCPs) of the external granular layer (EGL) [5].
The EGL is a transient germinal zone at the subpial cerebellar surface consisting of rhombic lip-derived progenitor cells that have migrated tangentially to the emerging cerebellar cortex at late stages of embryonal brain development [6]. During the early postnatal period in mouse, the morphogenic factor sonic hedgehog (Shh) is secreted by subjacent Purkinje cells and binds to patched receptors (Ptch1 and Ptch2) expressed on the GCP surface [7]. Ligand binding to Ptch1 then leads to functional de-repression of Smoh (Drosophila smoothened homolog) and subsequent activation of Gli (Glioma-associated oncogene family zinc finger) transcription factors [8]. This launches a temporally concerted gene expression pattern causing a proliferation burst and massive expansion of the GCP population during the first two postnatal weeks [7]. In particular, the direct Gli-target N-myc [9,10] and D type cyclins [11] were shown to be crucial for the growth and neoplastic transformation of GCPs [12]. In addition, the set of genes targeted by activated Gli transcription factors also include components of the canonical Hh pathway for feedback-loop regulation, such as the receptors Ptch1 and Ptch2 as well as the hedgehog-interacting protein (Hip) [10,13]. After several rounds of cell division, GCPs normally exit cell cycle and accumulate at the inner site of the EGL [14], where they start to migrate through the molecular layer (ML) and the Purkinje cell layer to form the internal granular layer (IGL) [15]. The mechanisms underlying the attenuation of the mitotic response and eventually the stop of GCP proliferation are not well understood. The most evident concepts describe extrinsic cues in gradient-based models to trigger GCP differentiation with increasing distance to the region of the outer EGL [16]. Finally, the EGL disappears at about three weeks after birth in mice.
PTCH1 was identified as a frequent target of inactivating mutations or genomic loss in sporadic MBs [17][18][19] that belong to the molecular subtype hallmarked by an aberrant activity of hedgehog signaling. The monoallelic inactivation of the Ptch1 gene in mice and thus downstream activation of the Hh pathway leads to MB development at a frequency of about 10-15% [20]. This mouse model has provided substantial insights into the pathogenesis of Hh-dependent MBs and has been used in different crossbreeding experiments to investigate tumor suppressor gene functions in this particular context [21,22].
Nitric oxide (NO) is a highly reactive gaseous molecule involved in various physiological processes ranging from vasculature modulation to neurotransmission [23,24]. NO is produced by three distinct enzyme isoforms: i) neuronal nitric oxide synthase (nNos/Nos1), ii) inducible nitric oxide synthase (iNos/Nos2), and iii) endothelial nitric oxide synthase (eNos/Nos3). Though being constitutively expressed in their respective tissue, nNos and eNos activity strongly depends on calcium [25], whereas calciumindependent iNos is primarily regulated by transcriptional induction, e.g. by inflammatory cytokines and endotoxins [26], which permits higher quantities of NO generation. The role of NO in cancer initiation and progression is heterogeneous with opposing effects in different malignancies [27]. Considering effects of tumor stroma, increased angiogenesis was reported to be associated with elevated Nos activity [28] and some immunerelated processes were found to be mediated by NO [29], including cytotoxicity of activated microglia [30]. Finally, NO released by vascular endothelial cells was reported to build a niche-like microenvironment for maintenance of glioma stem cells [31]. In the context of cerebellar development, Nos2 (inducible Nos) is initially expressed in early GCPs, whereas Nos1 (neuronal Nos) is hardly present before postnatal day 7 (Cerebellar Development Transcriptome Database [32]). Successively, Nos1 expression increases along with granule cell differentiation [33] and predominantly contributes to the common NO signaling that becomes apparent in the IGL as development proceeds [34]. Evidence has been provided that NO negatively acts on proliferation of neuronal precursors during adult neurogenesis [35]. Similarly, Ciani and colleagues demonstrated enhanced proliferation of cerebellar precursor cells upon inhibition of NO synthases [36].
Here, we report on the generation of Ptch1 +/2 Nos2 2/2 mice to investigate the impact of Nos2 on tumor development in Ptch1 hemizygous mutant mice. Interestingly, we observed an approximately two-fold increase in the incidence of spontaneous MB in Ptch1 +/2 Nos2 2/2 mice in comparison to Ptch1 +/2 Nos2 +/+ mice. To characterize the molecular pathomechanism underlying the tumor-promoting effect of Nos2 deficiency in Ptch1 +/2 mice, we performed comprehensive expression and DNA copy number profiling of MB tumors (Ptch1 +/2 Nos2 +/+ versus Ptch1 +/2 Nos2 2/2 ) as well as expression profiling of normal cerebellar tissue samples from different developmental stages and various genotypes (Ptch1 +/2 Nos2 +/+ , Ptch1 +/2 Nos2 2/2 , Ptch1 +/+ Nos2 2/2 and wild-type mice). Downregulation of the growth-associated protein 43 (Gap43) was the most striking feature in the cerebellum of Nos2-deficient mice when compared to Ptch1 +/2 Nos2 +/+ and wild-type mice. Subsequent functional analyses and results from in situ studies of GCPs in postnatal cerebellum allowed us to formulate a model for the tumor promoting role of Nos2 deficiency in Ptch1 mutant mice via deregulation of Gap43-dependent migration of GCPs.

Author Summary
Medulloblastoma is a common pediatric brain tumor, a subtype of which is driven by aberrant hedgehog pathway activation in cerebellar granule cell precursors. Although this tumor etiology has been intensively investigated in the well-established Ptch1 +/2 mouse model, knowledge is still lacking about the molecular interactions between neoplastic transformation and other developmental processes. Nitric oxide (NO) has been reported to be involved in controlling proliferation and differentiation of these cells. Therefore, inactivation of the NO-producing enzyme Nos2 in combination with the mutated Ptch1 gene should provide insight into how developmental regulation influences pathogenesis. Here, we describe a new role for NO in developing neuronal precursors of the cerebellum facilitating physiologically accurate migration via regulation of Gap43. We further demonstrate that disturbance of these processes leads to retention of granule precursor cells to the cerebellar periphery. Together with the sustained proliferation of these cells in combined Ptch1 +/2 Nos2 2/2 mice, this effect results in an increased medulloblastoma incidence relative to Ptch1 +/2 mice and demonstrates a new disease-promoting mechanism in this tumor entity.
were no obvious histopathological differences between MBs of the two genotypes.
Molecular analyses of MBs in Ptch1 +/2 Nos2 +/+ and Ptch1 +/2 Nos2 2/2 mice For an initial assessment of the molecular tumor characteristics, gene expression of hedgehog signaling pathway components were measured in 21 MBs and 24 normal (adult) cerebellar tissue samples from both Ptch1 +/2 Nos2 +/+ and Ptch1 +/2 Nos2 2/2 mice. Using quantitative real-time PCR (qRT-PCR), significant downregulation of the wild-type Ptch1 transcript and upregulation of the Shh target genes Gli1 and N-myc were generally observed in the tumor samples ( Figure S1), indicating all examined MBs to be of the same Hh-dependent molecular subtype. However, there were no significant differences for these genes between MBs of the two genotypes. Furthermore, targeted genetic analyses showed a loss of the wild-type Ptch1 allele in 10 of the 21 MBs investigated, while none of the tumors demonstrated a Tp53 mutation or N-myc amplification. The Cdkn2a/p16 INK4a locus was retained in all tumors while a single MB demonstrated a homozygous p19 ARF deletion (see Table S1 and Text S1 for details).
The expression profiling analysis revealed a total of 87 differentially regulated genes between tumors of the two genotypes (Table S2) with the vast majority (87%) showing lower transcript levels in Ptch1 +/2 Nos2 2/2 when compared to Ptch1 +/2 Nos2 +/+ mice. As expected from the initial targeted qRT-PCR measurements, there was no difference detectable concerning the activation of Hh pathway genes. Due to the important role of Nos2 during angiogenesis and cancer-associated immune response, including microglia, stromal effects need to be particularly considered in a systemic Nos2 knockout model. However, neither the set of significantly deregulated genes nor selective determination of marker expression for pericytes, vascular endothelial cells or microglia suggested any differences in the tumor stroma between the two genotypes (see Table S3 and Text S1 for details). According to the findings of Ciani and co-workers [37], reduction of NO enhances GCP proliferation through an increased expression of the proto-oncogene N-myc. Therefore, protein levels were particularly examined for differences between tumor samples from Ptch1 +/2 Nos2 +/+ and Ptch1 +/2 Nos2 2/2 mice. The results shown in Figure S2, however, revealed similar expression of Nmyc in all MBs.
Analyses of genomic copy number alterations revealed a trisomy of chromosome 6 in the majority of MBs from both groups (11/12, Figure 2A and 2B). Moreover, a small region on chromosome 13, approximately 1.5 Mb upstream of the Ptch1 gene, showed a hemizygous deletion in healthy cerebella of Ptch1-mutant mice (data not shown) but a homozygous deletion in most tumors (10/ 12). Similarly, a second small region 3.8 Mb downstream of the last Ptch1 exon exhibited a loss in 9 of 12 MBs. This suggests structural changes flanking the Ptch1 locus that likely contribute to inactivation of the wild-type allele. The frequencies of genomic aberrations showed a more heterogeneous karyotype with gross structural changes in Ptch1 +/2 Nos2 +/+ MBs when compared to Ptch1 +/2 Nos2 2/2 MBs (see Figure 2A, 2B and Text S1 for details). However, a general difference in chromosomal instability was not obvious between both genotypes. Only a small region (205.6 kb) on chromosome 14 containing the Entpd4 (ectonucleoside triphosphate diphosphohydrolase 4) gene was more frequently gained in Ptch1 +/2 Nos2 2/2 MBs (7/7) than in Ptch1 +/2 Nos2 +/+ MBs (1/5, Figure 2A). Accordingly, Entpd4 expression appeared to be specifically upregulated in expression profiles of Ptch1 +/2 Nos2 2/2 tumors. QRT-PCR validation confirmed an elevated mean expression in Ptch1 +/2 Nos2 2/2 compared to Ptch1 +/2 Nos2 +/+ MBs in those samples that overlapped with the array-CGH analysis but revealed no significant difference across the expanded tumor set ( Figure 2C). Nos2 inactivation is associated with low expression of mitotic genes in postnatal cerebellum As GCPs are considered the cells of origin for the Hh-dependent MB subtype, we examined the effect of Nos2 ablation in the context of cerebellar development. Therefore, gene expression profiles of normal cerebellar tissue samples from postnatal day 9 (P9), 6 weeks after birth (6W), and 1 year of age (1Y) were generated from wild-type, Ptch1 +/2 Nos2 +/+ , Ptch1 +/2 Nos2 2/2 , and Ptch1 +/+ Nos2 2/2 animals. While specimens of mature cerebellum (6W and 1Y) were investigated separately in 3 biological replicates per genotype and developmental stage, samples of postnatal cerebellum consisted of pooled RNA from 4-5 individuals processed in technical replicates due to limited tissue amounts.
An unsupervised hierarchical cluster analysis of transcriptome data clearly separated developing cerebellum (P9) of wild-type mice and the two Ptch1-mutated genotypes from mature cerebellum. Interestingly, P9 cerebellum of Ptch1 +/+ Nos2 2/2 mice displayed different properties highly similar to mature cerebellum ( Figure 3A). Expression profiles of MBs formed a distinct cluster clearly separated from all healthy tissue samples.
A direct comparison between gene expression profiles from Ptch1 +/+ Nos2 2/2 and wild-type P9 cerebellar tissue samples resulted in a total of 984 deregulated genes with 755 genes (76.7%) showing a decreased expression in Ptch1 +/+ Nos2 2/2 mice (Table  S4). P9 cerebellum from Ptch1 +/2 Nos2 +/+ and Ptch1 +/2 Nos2 2/2 mice revealed only 5 and 32 deregulated genes relative to wildtype, respectively (Table S5 and Table S6). This large deviation of postnatal gene expression in the Ptch1 +/+ Nos2 2/2 genotype included a set of downregulated genes that are essential for proliferation of GCPs (e.g. cyclin D1, cyclin D2 and N-myc, Figure 3B). As hedgehog signaling constitutes the main regulatory pathway for neonatal cell proliferation in GCPs of the EGL, the 984 deregulated genes were analyzed for enrichment of Gli transcription factor targets. Matching this list to a set of recently identified Gli-targets in GCPs [10] yielded a significant overrepresentation of Gli1-regulated genes (p = 0.005, chi-square test). Hence, the reduced transcript levels of these target genes suggests an attenuated hedgehog signaling in postnatal Ptch1 +/+ Nos2 2/2 cerebellum compared to wild-type (or any other genotype).
Notably, the decreased expression of Gli1-targets and proliferation-associated genes observed in Nos2-deficient cerebellar tissue was abolished upon additional inactivation of the hedgehog receptor Ptch1 (in Ptch1 +/2 Nos2 2/2 mice). Therefore, we examined the transcript levels of patched receptors themselves in more detail. While neither Ptch1 nor Ptch2 expression was changed between wild-type and Ptch1 +/2 Nos2 +/+ P9 cerebellum, a significant increase of Ptch1 and a minor increase of Ptch2 expression were observed in Ptch1 +/+ Nos2 2/2 mice relative to wild-type mice ( Figure 3C). Notably, in Ptch1 +/2 Nos2 2/2 cerebellar tissue samples, Ptch2 expression was more elevated than Ptch1. However, since Ptch2 is not capable of inhibiting smoothened (Smoh), it probably failed to take over the attenuating effect on Gli activity [38]. MB specimens from Ptch1 +/2 Nos2 +/+ versus Ptch1 +/2 Nos2 2/2 mice showed no significant difference in expression levels of either patched receptor, with Ptch2 being substantially increased over Ptch1 in both groups ( Figure 3C). These findings indicate that Nos2 deficiency leads to an upregulation of Ptch1 in GCPs, which results in a downregulation of mitotic genes and Gli-targets only in a Ptch1-wild-type background.

Impaired NO signaling reduces Gap43 transcript levels
The association of Nos2 inactivation and decreased Gap43 expression suggests a gene-regulatory function of NO signaling. In order to investigate this possible link in vitro, we used the murine cerebellar precursor cell line c17.2 and the human MB cell line D458 (see Text S2 for details). Both cell lines were treated either with the Nos inhibitor L-NAME (Nv-nitro-L-arginine methyl ester) to reduce NO levels or solvent control ( Figure S4). Relative expression of Gap43 was assessed every 24 hours by qRT-PCR. In c17.2 cells, Gap43 transcript abundance was generally low and increased with culture duration. We observed a slightly decreased expression of Gap43 upon L-NAME treatment that reached significance (p = 0.023) after 120 hours ( Figure 5A). NOS inhibition in D458 human MB cells resulted in a significant reduction of Gap43 transcript levels starting already after 72 hours with further decrease after 96 hours and 120 hours ( Figure 5B). FACS analyses of apoptosis and cell cycle excluded these observations to be attributed to secondary effects of changing cell conditions ( Figure S5). These results suggest Gap43 downregulation as a direct consequence of reduced NO levels in murine neuronal precursors and human MB cells.

Increase of Ptch1 expression and impairment of GCP migration upon knockdown of Gap43
The dependency of Gap43 expression on NO signaling suggests this gene as key mediator of the effects observed in Nos2-deficient P9 cerebellum and Ptch1 +/2 Nos2 2/2 MB, in particular, the upregulation of functional Ptch1 in Ptch1 +/+ Nos2 2/2 mice. Mishra et al. recently reported a central role of Gap43 in the polarization of developing GCPs by regulating centrosome positioning and thus defining correct orientation towards the IGL [39]. Since this is a prerequisite for directed migration, reduced levels of Gap43 in P9 cerebellar tissue may lead to retention of GCPs in the EGL. To test these hypotheses, shRNA-mediated knockdown of Gap43 was performed in c17.2 cells (see Text S1 for details). Upon knockdown of Gap43 we observed a strong inverse behavior of Ptch1 and Gap43 transcript levels ( Figure 5C). Changes in migration characteristics were assayed in a Boyden chamber using recombinant SDF-1a (CXCL12) as chemoattractant, which was reported to participate in guiding migration of embryonal GCPs in vivo [40]. Downregulation of Gap43 yielded a significant decrease in cell migration between 14% (p = 0.013) and 20% (p = 0.007) ( Figure 5D; Figure S6C). A pseudo-effect of the knockdown due to altered proliferation of the v-myc-immortalized c17.2 cells was excluded by FACS-based cell cycle analysis ( Figure  S7).

GCPs of the external granular layer show Gap43associated phenotypes
Transcriptome and functional analyses suggest that a decreased Gap43 expression accounts for Ptch1 upregulation and impairment of directed neuronal precursor migration in vitro. Accordingly, Ptch1 +/+ Nos2 2/2 P9 cerebella are supposed to increasingly retain GCPs with reduced mitotic activity in the EGL compared to wildtype and Ptch1 +/2 Nos2 +/+ mice. Moreover, the Ptch1 +/2 Nos2 2/2 genotype is also expected to exhibit retention of GCPs, but not to show any cell cycle arrest. To further verify this hypothesis in situ we performed immunofluorescent double staining of proliferating (Ki-67+) and post-mitotic GCPs on FFPE sections of postnatal cerebellum ( Figure 6A). Here, post-mitotic cells were delineated by the neuronal marker NeuN (neuronal nuclear antigen A60) [41]. At least three different regions of each mouse cerebellum were analyzed in three to four animals per genotype using confocal laser scanning microscopy. In accordance with the microarray data, averaged cell counts of wild-type and Ptch1 +/2 Nos2 +/+ mice did not show significant difference. In contrast, an increase of postmitotic GCPs (NeuN+, Ki-672) was detectable in the EGL of Ptch1 +/2 Nos2 2/2 and Ptch1 +/+ Nos2 2/2 mice ( Figure 6C). Concurrently, the ratio of dividing to non-dividing GCPs was similar in Ptch1 +/2 Nos2 2/2 , wild-type and Ptch1 +/2 Nos2 +/+ P9 cerebella but markedly decreased in Ptch1 +/+ Nos2 2/2 mice. This recapitulated the downregulation of mitotic genes observed in the expression profiles. However, the total amount of proliferating GCPs per EGL section was significantly higher in Ptch1 +/2 Nos2 2/2 mice compared to any other genotype ( Figure 6C). These results demonstrate a tissue phenotype that corresponds to the effects of reduced Gap43 in developing cerebellar neuronal precursors (in vitro). The increased accumulation of proliferating GCPs in the EGL observed in the Ptch1 +/2 Nos2 2/2 genotype supposedly leads to a larger pool of cells susceptible to neoplastic transformation and is therefore likely to promote medulloblastoma development.

Discussion
The Ptch1 +/2 MB mouse model has been intensively studied and has greatly contributed to our understanding of Hhdependent MB tumorigenesis in the context of cerebellar development. The data presented here indicate a role of Nos2 and hence NO signaling in Hh-dependent MB by demonstrating a significantly increased MB rate in Ptch1 +/2 Nos2 2/2 mice compared to Ptch1 +/2 Nos2 +/+ mice. The global genome-wide  screens performed in the present study did not reveal obvious molecular differences between MBs in Ptch1 +/2 Nos2 +/+ versus Ptch1 +/2 Nos2 2/2 animals. Assessment of genomic alterations using array-CGH identified trisomy of chromosome 6 as a recurrent feature in tumors of both genotypes. This corresponds to a recent report on MBs of the same molecular subtype with inactivated double-strand break repair proteins targeted to neuronal progenitors of p53 2/2 mice [42]. The most common loss identified in our analyses affected two small regions on chromosome 13 encompassing the Ptch1 gene and possibly indicate acquired homozygosity for the mutant allele or somatic rearrangements rather than a broad deletion of the locus. Targeted duplex PCR further confirmed loss of the functional wild-type allele to be a frequent event in these MBs. Notably, tumors of the Ptch1 +/2 Nos2 2/2 genotype showed a higher frequency of a small gain on chromosome 14. The affected Entpd4 gene encodes for an apyrase located at the internal membrane of lysosomal vacuoles and the Golgi apparatus. It preferentially catalyzes the hydrolysis of UDP to UMP [43] and thereby facilitates the inverse directed import of UDP-GlcNAc [44]. This in turn was reported to increase glycosylation of surface receptors (e.g. EGFR and PDGFR) and foster cell growth [45]. According to the microarray and qRT-PCR expression data, Entpd4 transcript levels were indeed increased in tumors with this chromosomal gain. However, this effect did not turn out to be Nos2-dependent in an expanded sample set. Consequently, Entpd4 likely plays a role in MB pathogenesis but is not directly linked to loss of Nos2.
The examination of tumor-relevant changes in developing cerebellum as a consequence of impaired Nos2 activity and hence NO signaling surprisingly revealed a decreased proliferation of GCPs in the cerebellum of Ptch1 +/+ Nos2 2/2 mice. The concurrent upregulation of Ptch1 and the significant enrichment of downregulated Gli1-target genes strongly suggest that this effect is a consequence of reduced hedgehog signaling. Moreover, this phenotype was completely abrogated by a concomitant Ptch1 mutation. The slight increase of Ptch2 in Ptch1 +/2 Nos2 2/2 cells points to a compensatory effect and further supports the notion of an inhibitory function of Nos2 loss on the hedgehog pathway in postnatal cerebellum. Since neither a Smoh-regulating domain [38] nor a function for cell cycle arrest through seizing cyclin B1 [46] were reported for Ptch2, its upregulation may be insufficient for preventing MB induction. In contrast to these observations, Ciani et al. demonstrated that proliferation of cultured GCPs increased upon withdrawal of NO and that this effect was showing intensity histograms of the EGL and ML area (upper panel), as well as intensity distribution of the different genotypes with subtracted wildtype signal (lower panel). Black arrow indicates a common peak of nucleus background signal. EGL: external granule layer, ML: molecular layer, and IGL: internal granule layer. doi:10.1371/journal.pgen.1002572.g004 mediated by augmented N-myc levels [37]. However, N-myc was not differentially expressed between Ptch1 +/2 Nos2 2/2 and Ptch1 +/2 Nos2 +/+ MBs of our series. A possible explanation for this discrepancy might be an unrecognized heterogeneity in the isolated cerebellar cell population used in the Ciani study. Since eNos and nNos are known to attenuate the mitotic activity of subventricular neuronal stem cells [47,48] Nos inhibitor treatment possibly resulted in a selective growth advantage over GCPs. Downregulation of Gap43 was the only feature observed in Nos2-deficient versus Nos2proficient postnatal cerebella irrespective of the Ptch1 status. This difference was also conserved between Ptch1 +/2 Nos2 2/2 and Ptch1 +/2 Nos2 +/+ MBs, and particularly visible in outer regions of the molecular layer, where maturating GCPs of the EGL develop contact forming projections prior to radial migration. Other studies already suggested a link between Gap43 mRNA levels and NO signaling due to co-induction of nNos and Gap43 during axon regeneration and reactive synaptogenesis following injury of spinal motoneurons [49] and sensory neurons [50]. Furthermore, a downregulation of Gap43 was reported after silencing of soluble guanylate cyclase subunits, the central elements of cGMPmediated NO signaling [51]. Finally, the present study demonstrates Gap43 downregulation to be a consequence of NO withdrawal in neuronal progenitors and MB cells. A possible mechanism for this regulation refers to decreased protein levels of the poly(U)-binding and degradation factor AUF1 upon NOdependent cGMP production [52]. AUF-proteins generally bind to AU-rich elements of the 39UTR (untranslated region) of coding transcripts and associate with proteins of the ELAV-like family to control gene expression via mRNA decay [53]. Tsai et al. demonstrated that Gap43 mRNA levels are post-transcriptionally regulated during neuronal differentiation and that elements of the 39UTR confer transcript instability, which is abolished upon TPA treatment (inter alia inducing NOS2) [54]. At the same time, Chung et al. demonstrated that indeed ELAV-like family member HuD was binding to 39UTR regions of GAP43 [55]. Taken together, NO accumulation possibly decreases cellular levels of mRNAdestabilizing AUF1 protein and thus might contribute to a high transcript abundance of Gap43.
Gap43 is a membrane-anchored protein at the cytoplasmic side of neuronal cell projections and found to be highly expressed during development of the CNS [56]. It is particularly localized in axonal growth cones and participates in the coordination of extrinsic stimuli and intrinsic cell remodeling [57] by regulating cytoskeleton dynamics [58]. Granule cell (GC) migration follows a sequence of tangential and radial movements controlled by successive formation of leading projections [59]. As maturating GCPs exit cell cycle, positioning of the centrosome determines the site of axon growth cone emergence and thus neuronal polarity including localization of such projections [60]. This defines the structural orientation of GCPs in terms of directing its dendrite to descend across the molecular and Purkinje cell layers to populate the IGL. However, centrosome positioning and therefore accurate polarization of GCPs require phosphorylated Gap43 to bind to the centrosome-associated microtubule-organizing center [61]. Hence, inaccurate GCP migration was observed in Gap43 2/2 animals [39], a finding that is in full agreement with our data from the functional Gap43 knockdown assays. Downregulation of Gap43 in Nos2-deficient P9 cerebellum therefore likely mediates the retention of GCPs observed in FFPE sections. Accordingly, NO/ cGMP signaling was demonstrated to be crucial for accurate migration of the neuronal precursor cell line NT2 [62]. Furthermore, slice culture experiments of neonatal cerebella (P9) exhibited a substantial reduction of proliferation and migration of maturating granule cells to the IGL upon application of NO synthases inhibitors [63]. The elevation of Ptch1 levels upon Gap43 reduction in vitro fits to the data by Shen et al. who reported an upregulation of Ptch1 gene expression in inner EGL regions of Gap43 2/2 mice compared to wild-type animals. Moreover, cultured Gap43-deficient GCPs show decreased proliferation in response to administered recombinant Shh protein [64]. A possible regulatory link was recently provided as the activation of the hedgehog signaling component Smoh was found to depend on PI4P (phosphatidylinositol 4-phosphate) levels that immediately increase when Shh binds to Ptch1 or when functional Ptch1 is absent [65]. The authors further showed that imbalanced conversion of the precursor molecule PI into PI4P influences hedgehog pathway activity. Alternatively, the production of PI4P can also result from a specific dephosphorylation of PI(4,5)P2 [66]. In this context, Gap43 protein was recently demonstrated to build oligomeric structures in the plasma membrane which sequester specifically PI(4,5)P2 [67]. A similar finding has been reported earlier showing that GAP43 participates in the accumulation of plasmalemma rafts, which promoted retention of PI(4,5)P2 [68]. The amount of Gap43 associated with the plasma membrane therefore possibly modulates the utilization of PI(4,5)P2, including its conversion into PI4P, which in turn directly affects hedgehog signaling through Smoh activation. However, the effective impact on downstream Gli-targets would still be difficult to conclude regarding the multitude of responses to Shh, including negative feedback regulation [13]. Further studies applying depletion and enrichment of specific phosphatidyl derivatives and selective silencing of hedgehog pathway elements will be necessary to elucidate the molecular nature of this proposed signaling axis.
The increased accumulation of mitotic granule cells at the EGL seen in the combined Ptch1 +/2 Nos2 2/2 genotype supposedly gives a special clue to MB induction. In contrast to the classical view of neonatal EGL organization, which describes radial migration of granule cells to follow a proliferation stop, more and more evidence arises showing that cell cycle arrest is not a prerequisite for migration but rather occurs during a temporally coordinated interplay of gene expression patterns [69]. This corroborates our data shown in Figure 6 (white arrows), where proliferation is still maintained in migrating granule cells and even in cells of the IGL of wild-type cerebellum. Regulation of such expression patterns is largely dependent on Shh stimuli being most intensive in the EGL [70], as well as gradients of other soluble factors such as Bmps, which were reported to account for a regulatory environment along the transition through the cerebellar layers [71]. Further evidence for a niche-like-concept was provided by Choi et al. in Bdnf 2/2 mice that displayed a severe retardation of GCP migration [16]. The authors could demonstrate that mitotic activity of maturating GCPs was significantly enhanced when cells were retained in the EGL and declined with increasing distance from outer EGL regions. Therefore, the accumulation of GCPs in the EGL in combination with the insensitivity to Ptch1-mediated cell cycle arrest in Ptch1 +/2 Nos2 2/2 mice provide a growth advantage and increase the number of putative transformation targets over Ptch1 +/2 Nos2 +/+ mice ( Figure 7C).
In conclusion, the following picture emerged from our data: Homozygous deletion of Nos2 leads to a reduction of basic NO levels in immature GCPs of the EGL during postnatal development of the cerebellum. This reduction causes a downregulation of Gap43 expression, which results in an increased expression of Ptch1 and impaired directed migration of maturating GCPs. As a consequence, undifferentiated granule cell precursors exit cell cycle and are retained at the EGL ( Figure 7B). In case of an additional heterozygous Ptch1 mutation, upregulation of this receptor does not suffice to exert the anti-proliferative stimulus following Gap43 decrease, which results in an increased fraction of continuously dividing cells in the EGL ( Figure 7C). As reduced migration towards the IGL further leads to a withdrawal of growth-limiting signals, expansion of the GCP population is additionally supported. Finally, this advances medulloblastoma development in Ptch1 +/2 Nos2 2/2 mice compared to Ptch1 +/2 Nos2 +/+ mice. The mechanism described here illustrates a new tumor-promoting concept in MB showing that the localization of pre-neoplastic cells within the developing cerebellum is important for pathogenesis.

Microarray-based genomic and expression profiling
Total RNA of tumor specimens and normal cerebellar tissue samples was isolated via CsCl density gradient centrifugation [73] and assessed for integrity using an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, USA). For gene expression microarrays, linear amplification of mRNA and labeling of cDNA were conducted on samples and Mouse Universal Reference RNA (Stratagene, La Jolla, USA) according to the TAcKLE protocol [74]. Both were combined for two-color hybridizations with each sample being performed as two replicates of inverse dye orientation. Global gene expression profiling was performed on self-printed oligonucleotide microarrays. Further details of microarray production and hybridization are described in Text S2. Genomic DNA of tumor specimens was isolated from the interphase of the CsCl gradient by ethanol precipitation, proteinase K digest, and phenol/chloroform extraction. DNA samples were monitored for purity and adequate fragment size using spectrophotometric measurements and gel electrophoresis. Array-based comparative genomic hybridization (array-CGH, matrix-CGH, [75]) was performed on Mouse Genome CGH 244 k Microarrays (Agilent). Cy5-labeled tumor DNA was combined with corresponding Cy3-labeled reference (wild-type genomic DNA) to receive either sex-matched sample pairs or pairs of different gender for internal negative or positive control. Sample preparation, microarray hybridization, and washing procedures were carried out as described in the manufacturer's protocol. Microarray data are available in GEO (http://www-ncbi.nlm.nih. gov/geo), under accession number GSE29201.

qRT-PCR analyses
Total RNA isolated from cell culture samples using the RNeasy Mini Kit (Qiagen, Hilden, Germany) or RNA from tissue specimens was subjected to oligo(dT)-primed reverse transcription. QRT-PCR measurements were conducted in an ABI PRISM 7900HT thermal cycler (Applied Biosystems, Foster City, USA) using the SYBR green reaction and detection system (ABgene, Epsom, UK). For relative quantification mean ratios were calculated between genes of interest and a set of five housekeeping genes (Table S7) according to the Pfaffl method [76].
The expression levels of Ptch1, Gli1, N-myc, and Nos2 were determined by real-time reverse transcription PCR analysis using the ABI PRISM 5700 system (Applied Biosystems) as reported before [73]. For these experiments, the mRNA expression level of mitochondrial ribosomal protein L32 (Mrpl32) served as housekeeping reference. All primer sequences are depicted in Table S8.

Cell culture experiments
Inhibition of NO synthases was performed in c17.2 and D458 cells which were seeded at densities of 2610 5 and 4610 5 cells per well in 12-well plates, respectively. Cells were daily treated with either 1 mM of the inhibitor L-NAME or 16 PBS as solvent control. For knockdown experiments of Gap43, c17.2 cells were grown in a 12-well plate to 80% confluency and transfected with 2 mg of pLKO.1-puro vector that contained either shRNA constructs targeting Gap43, shRNA against GFP, or non-target shRNA as a control (Sigma-Aldrich, St. Louis, USA) using 9 ml FuGene HD reagent (Roche, Basel, Switzerland). Transfection was repeated 2 times each after 8 hours and subjected to selection conditions (1 mg/ml puromycin) for 24 hours. Subsequently, cells were trypsinized, adjusted to 4610 5 cells/ml and seeded into the inserts of a Costar Polycarbonate Membrane Transwell plate (8 mm pores, Corning, USA). After 24 hours cells were either harvested for gene expression and protein analyses or 0.1 mg/ml recombinant SDF-1a was applied to the lower compartment for migration assays. Following 12 hours of incubation, cells at the bottom of the insert membrane were methanol-fixed, hematoxylin-stained, and counted.
Quantification of Gap43 staining was performed for areas of interest using Image J software (NIH). Numbers of dividing and non-dividing cells in the EGL of postnatal cerebellar tissue sections were counted manually and normalized to the corresponding length of the EGL edge. Cell counts for each region were averaged across three sections, each with 10-20 mm distance in z-axis, and per individual.

Statistical analyses
Kaplan-Meier survival plots were calculated for a total of 1167 mice, including 315 wild-type mice, 412 Ptch1 +/+ Nos2 2/2 mice, 215 Ptch1 +/2 Nos2 +/+ mice and 221 Ptch1 +/2 Nos2 2/2 mice. MBfree survival was plotted using the GraphPad Prism 5 software (GraphPad, La Jolla, USA). The logrank test was applied to compare survival (tumor occurrence) of the different genotypes. For comparisons of differences of means between two groups of replicates, p-value calculations were performed using an unpaired, two-tailed t-test, unless indicated otherwise. Calculated error bars represent the standard error of the mean (SEM). For details on microarray statistics please refer to Text S2. Figure S1 Expression of Ptch1, Gli1, and N-myc in medulloblastomas. (A) Expression of Ptch1 is significantly downregulated in medulloblastomas of both genotypes relative to normal adult cerebellum. (B-C) Gli1 and N-myc transcripts are significantly upregulated relative to the same control cerebella. Shown are mRNA expression levels (means with standard deviation) of 8 Ptch1 +/2 Nos2 +/+ and 13 Ptch1 +/2 Nos2 2/2 MBs as well as 12 normal cerebellum samples (NCB) of each genotype. All data are based on qRT-PCR results calculated relative to the reference gene MrpL32 and normalized to mouse universal reference RNA (Stratagene). Significant differences are indicated (*p,0.05, **p,0.01, ***p,0.001, Mann-Whitney U test or ANOVA test, respectively). (TIF) Figure S2 N-myc protein is not differentially expressed between Ptch1 +/2 Nos2 +/+ and Ptch1 +/2 Nos2 2/2 medulloblastomas. (A) Western blot analysis of N-myc protein expression in separated protein extracts from tumor specimens of three Ptch1 +/2 Nos2 +/+ and five Ptch1 +/2 Nos2 2/2 mice, as well as healthy adult cerebellum of wild-type, Ptch1 +/2 Nos2 +/+ and Ptch1 +/2 Nos2 2/2 . Depicted is stained for N-myc and a-tubulin (housekeeping protein). (B) Quantification of N-myc protein bands normalized to a-tubulin, p = 0.34, Dmean = 20.14960.144. (TIF) Figure S3 Validation of candidate genes that were observed to be differentially expressed between Ptch1 +/2 Nos2 +/+ and Ptch1 +/2 Nos2 2/2 tumors in the microarray data. Linear expression values were obtained from qRT-PCR measurements on the expanded sample set and indicate mRNA expression against a pool of housekeeping genes normalized to mouse universal reference RNA