A Distinct Smoothened Mutation Causes Severe Cerebellar ...

A Distinct Smoothened Mutation Causes Severe Cerebellar Developmental Defects and Medulloblastoma in a Novel Transgenic Mouse Model Joyoti Dey,a,b Sally Ditzler,b Sue E. Knoblaugh,c Beryl A. Hatton,b Janell M. Schelter,d Michele A. Cleary,d Brig Mecham,e Lucy B. Rorke-Adams,f and James M. Olsona,b,g Molecular and Cellular Biology Program, University of Washington, Seattle, Washington, USAa; Clinical Research Divisionb and Comparative Medicine Shared Resource,c Fred Hutchinson Cancer Research Center, Seattle, Washington, USA; Merck Research Laboratories, Merck & Co., Incorporated, Boston, Massachusetts, USAd; Sage Bionetworks, Seattle, Washington, USAe; Children’s Hospital of Philadelphia, Philadelphia, Pennsylvania, USAf; and Seattle Children’s Hospital, Seattle, Washington, USAg

Deregulated developmental processes in the cerebellum cause medulloblastoma, the most common pediatric brain malignancy. About 25 to 30% of cases are caused by mutations increasing the activity of the Sonic hedgehog (Shh) pathway, a critical mitogen in cerebellar development. The proto-oncogene Smoothened (Smo) is a key transducer of the Shh pathway. Activating mutations in Smo that lead to constitutive activity of the Shh pathway have been identified in human medulloblastoma. To understand the developmental and oncogenic effects of two closely positioned point mutations in Smo, we characterized NeuroD2-SmoA2 mice and compared them to NeuroD2-SmoA1 mice. While both SmoA1 and SmoA2 transgenes cause medulloblastoma with similar frequencies and timing, SmoA2 mice have severe aberrations in cerebellar development, whereas SmoA1 mice are largely normal during development. Intriguingly, neurologic function, as measured by specific tests, is normal in the SmoA2 mice despite extensive cerebellar dysplasia. We demonstrate how two nearly contiguous point mutations in the same domain of the encoded Smo protein can produce striking phenotypic differences in cerebellar development and organization in mice.

T

he protracted phase of extensive proliferation during cerebellar development makes it vulnerable to neoplastic transformation (42). Medulloblastoma, a developmental cancer of the cerebellum, continues to be the most common pediatric brain cancer. Standard treatments result in neurocognitive impairment and adverse quality of life (12, 34). Medulloblastomas are categorized based on histological characteristics and molecular signatures (12, 41). Genetic aberrations leading to hyperactive Sonic hedgehog (Shh) signaling in granule neuron precursors (GNPs) cause 25 to 30% of medulloblastoma cases (17). The Shh pathway plays a pivotal role in cerebellar development by regulating proliferation of GNPs and foliation (7, 44). The Shh subgroup has been widely studied with numerous mouse models recapitulating the human disease (26). The overall prognosis in patients with Shh-driven medulloblastomas, however, remains intermediate (41). Within the Shh subgroup of human medulloblastoma there exists significant biological and clinical heterogeneity, the underlying molecular basis of which remains to be explored (29, 36). Leptomeningeal dissemination observed uniquely in the SmoA1 homozygous (Smo/Smo) mouse model and not other Shh-driven models (18) demonstrates disparities in pathology. Inhibition of the Shh pathway by the Smoothened (Smo) antagonist cyclopamine varies based on mutations driving hyperactive signaling (4, 5), leading to differences in therapeutic responses. Aberrations in genes outside the Shh pathway also lead to medulloblastomas with Shh signatures in mice, highlighting the widespread interactions of the Shh pathway with other networks (26). In several mouse models, medulloblastoma-prone progenitors exit the cell cycle and undergo normal neuronal differentiation, suggesting that factors in addition to initiating mutations contribute to tumorigenesis (2) and possibly tumor heterogeneity. While broad molecular classifications are important, it is nec-

4104

mcb.asm.org

Molecular and Cellular Biology

essary to investigate the unique behavior of driving mutations, since the downstream effects may be distinct. Since medulloblastoma results from developmental aberrations (25), investigation of critical milestones in cerebellar development will provide valuable insights in this area. Toward this goal, we developed the SmoA2 mouse model of medulloblastoma and carried out a comparative analysis with the existing SmoA1 model. SmoA1 (W539L) and SmoA2 (S537N) mutations, originally identified in human cancer patients (31, 45), lie in the same transmembrane domain of Smo and cause constitutive activation of the Shh pathway (40). While the SmoA1 mutation has been widely studied, very little is known about SmoA2. Through characterization of the SmoA2 model, we show striking differences between the SmoA1 and SmoA2 mutations at the molecular and cellular levels. While both mutations lead to medulloblastomas, the SmoA2 mutation uniquely causes severe defects in cerebellar development. Early in development, the two mutations lead to distinct transcriptional profiles affecting different biological processes. Despite disruptions in the cytoarchitecture thought to be critical for cerebellar function, the SmoA2 mice, intriguingly, do not display clinical signs of cerebellar malfunction.

Received 27 June 2012 Returned for modification 23 July 2012 Accepted 31 July 2012 Published ahead of print 6 August 2012 Address correspondence to James M. Olson, [email protected]. Supplemental material for this article may be found at http://mcb.asm.org/. Copyright © 2012, American Society for Microbiology. All Rights Reserved. doi:10.1128/MCB.00862-12

p. 4104 – 4115

October 2012 Volume 32 Number 20

SmoA2 Causes Cerebellar Developmental Defects

MATERIALS AND METHODS Generation of the SmoA1 and SmoA2 transgenic lines. The SmoA1 and SmoA2 transgenic mouse lines were previously described (16, 18). Both lines were generated and maintained on a C57BL/6 background. SmoA2 hemizygous and SmoA1 homozygous (Smo/Smo model) (18) mice of either sex were used for all experiments, except for the transgene copy number analysis, where the SmoA1 hemizygous line was used (16). SmoA1 and SmoA2 mutations were originally identified in human cancer cases (31, 45) and correspond to W539L and S537N, respectively, in mouse Smo (40). All mice were maintained in accordance with the NIH Guide for the Care and Use of Experimental Animals with approval from the Fred Hutchinson Cancer Research Center Institutional Animal Care and Use Committee (IR1457). Copy number determination. Transgene copy numbers were approximated using a quantitative-PCR (qPCR) approach (Platinum SYBR green; Invitrogen) based on existing methodologies (19, 24). Briefly, 10fold serial dilutions of wild-type (WT) mouse genomic DNA (ranging from 190 to 0.019 ng) were used to make standard curves to determine the efficiency and specificity of each primer pair used in the copy number analysis. Primers with 100% efficiency against the gene of interest, Smo Exon 10, and control loci, RNAseP, were used for copy number estimation. The additional primers Smo Exon 6 and control locus Gapdh were used for confirmation. The Smo primers used recognize both the endogenous Smo gene and the transgenes SmoA1 and SmoA2, which differ by a point mutation from the endogenous loci. The sequences of the primers are as follows: Smo Exon 6 FP, 5=-CGTGAGTGGCATCTGTTTTG-3=, and RP, 5=-AGTAGCCTCCCACAATAAGCAC-3=; Smo Exon 10 FP, 5=AGAGCAAGATGATCGCCAAG-3=, and RP, 5=-CCATCATGGGAGAC AGTGTG-3=; RNAseP FP, 5=-CTCCCCAAATGGAAGATGAG-3=, and RP, 5=-TATTCTACGTTCCGGTGTGG-3=; and Gapdh FP, 5=-AAATGA GAGAGGCCCAGCTAC-3=, and RP, 5=-TTATAGGAACCCGGATGGT G-3=. Subsequently, 5 ng of genomic DNA (n ⫽ 5 mice per genotype)— SmoA1 hemizygous, SmoA2 hemizygous, and WT—was amplified using the same qPCR conditions. The comparative threshold cycle (CT) method was used to calculate Smo copy numbers in SmoA1 and SmoA2 transgenic mouse genome relative to normal Smo in the WT reference genome. Mouse pathology and immunohistochemistry. Mice were anesthetized using CO2 inhalation, the cerebella were removed, and tissues were snap-frozen for RNA studies and GNP isolation or fixed in 10% paraformaldehyde for pathological examination. Tissue blocks were paraffin embedded, cut into 4-␮m sections, and stained with hematoxylin and eosin (H&E) using standard methods. Immunohistochemical analyses were carried out as follows: (i) for NeuN, a mouse-on-mouse procedure developed at the Fred Hutchinson Cancer Research Center (FHCRC) Experimental Histopathology Shared Resource was used with NeuN primary antibody (Millipore; 1:100) to create an antibody complex with a rabbit anti-mouse Fab fragment addition of adding it to the tissue, followed by Biocare Mach2 rabbit polymer; (ii) for calbindin, rabbit primary antibody (Millipore; 1:500) followed by biotinylated goat anti-rabbit secondary antibody–Vectastain Elite avidin-biotin complex (Vector Laboratories) was used; (iii) for Ki67, rat primary antibody (Dako Tec3; 1:100) followed by biotinylated goat anti-rat antibody–streptavidin-horseradish peroxidase (HRP) was used; and (iv) for S100, rabbit primary antibody (Dako; 1:400) followed by Dako Envision mouse polymer was used. Slides were developed using 3,3=-diaminobenzidine (Dako DAB Plus) reagent followed by Dako Autostainer hematoxylin counterstain. Sections were visualized with a Nikon E800 microscope, and images were captured using the CoolSnap cf color camera (FHCRC Scientific Imaging Core). qRT-PCR. For quantitative reverse transcription PCR (qRT-PCR) analyses, mRNA was isolated from either snap-frozen tissue or GNPs from postnatal day 5 (P5) SmoA1, SmoA2, and WT mice harvested using previously described techniques (22), using the RNAeasy kit (Invitrogen) followed by DNase (Qiagen) treatment and conversion to cDNA using the ABI High Capacity cDNA reverse transcription kit (Life Technologies). Reactions were set up using TaqMan Master Mix and run on an ABI


7900HT Fast Real-time PCR System. TaqMan gene expression assays (Applied Biosystems) were used for mouse Gli1, Gli2, Ptch1, Ptch2, and Smo. Data were analyzed using ABI GeneAmp SDS v2.3 software. Primers for qRT-PCR (Platinum SYBR green; Invitrogen) verification of microarray data were as follows: Bcl11b FP, 5=-ACGCGTAAAGATGAGGCCTTC3=, and RP, 5=-AAGCCATGTGTGTTCTGTGC-3=; MyoD FP, 5=-CCCG CGCTCCAACTGCTCTG-3=, and RP 5=-GGCTCGACACAGCCGCACT C-3= (a kind gift from the Stephen Tapscott laboratory); Pou4f2 FP, 5=-C GGAGAGCTTGTCTTCCAAC-3=, and RP, 5=-GCCAGCAGACTCTCAT CCA-3=; Pou4f1 FP, 5=-GACCTCAAAAAGAACGTGGTG-3=, and RP, 5=-TAAGTGTCTCTGGTCCCCTCAG-3=; Cbln4 FP, 5=-TGAGCAAC AAGACTCGCATC-3=, and RP, 5=-GTGCCACAAAGACAGATTCC-3=; Ranbp17 FP, 5=-TTAGAGCGCGCGATAATTG-3=, and RP, 5=-TCTGGG CTGTCAATCAGTTC-3=; Isl1 FP, 5=-AGATCAGCCTGCTTTTCAGC3=, and RP, 5=-ATGCTGTTGGGTGTATCTGG-3=; and Aqp1 FP, 5=-TCC TCCCTAGTCGACAATTCAC-3=, and RP, 5=-TGCAGAGTGCCAATGA TCTC-3=. All assays were run in triplicate and normalized to endogenous control mouse ␤-2 microglobulin (␤2m) or Gapdh for TaqMan assays, and Cyclophilin A (Ppia) FP, 5=-GAGCTGTTTGCAGACAAAGTTC-3=, and RP, 5=-CCCTGGCACATGAATCCTGG-3= (a kind gift from the Sunil Hingorani laboratory), were used for SYBR green assays. Western blot analysis. GNPs from P5 SmoA1, SmoA2, and WT mice were harvested by previously described techniques (22) and used for preparation of protein lysates using RIPA buffer (Millipore) with Halt protease inhibitor cocktail (Pierce Biotechnology) and Phosphatase Inhibitor Cocktail 2 (Calbiochem) and Cocktail 3 (Sigma). Equal amounts of protein from each sample (25 ␮g) were subjected to SDS-PAGE using a NuPAGE Novex 4 to 12% Bis-Tris gel (Invitrogen), transferred to a nitrocellulose membrane using the X-Cell SureLock Mini cell system (Invitrogen), and probed with Smo antibody (H-300; Santa Cruz Biotechnology; 1:100) and loading control ␤-actin (Abcam; 1:2,500). Corresponding horseradish peroxidase-conjugated secondary antibodies were obtained from Santa Cruz Biotechnology and used at 1:5,000 dilution. Proteins were detected by incubating membranes in chemiluminescent substrate (ECL kit; Pierce), followed by exposure to Kodak X-Omat scientific imaging film. Transplantation experiments. All experiments involving animals were conducted in accordance with the NIH Guide for the Care and Use of Experimental Animals with approval from the Fred Hutchinson Cancer Research Center Institutional Animal Care and Use Committee (IR1457). GNPs were harvested from P5 SmoA2 mice. Orthotopic transplants of 2-␮l cell suspensions containing 1 ⫻ 106 SmoA2 GNPs were carried out in recipient athymic nude mice (Nu/Nu; Jackson Laboratory) as previously described (37). Behavioral tests. The RotaRod test was conducted using a RotaRod (model 7650; Ugo Basile Comerio, Italy), accelerating from 3 to 30 rpm over a 5-min period. The mice were given a trial run prior to the timed run, where the latency to fall (the time it takes the mouse to fall off the rod, measured from the start of the test) was recorded. Footprint assay. The bottom of each foot was coated with nontoxic paint, and the mouse was allowed to walk through a small tunnel on white paper. Stride length (the distance between two hind paw prints) and stance width (the distance between opposite hind paw prints perpendicular to the walking trajectory) were calculated. Each mouse was given a trial run before the final run. Modified SHIRPA. SHIRPA stands for SmithKline Beecham Pharmaceuticals; Harwell, MRC Mouse Genome Centre and Mammalian Genetics Unit; Imperial College School of Medicine at St. Mary’s; Royal London Hospital, St Bartholomew’s, and the Royal London School of Medicine phenotype assessment. Our laboratory developed an abbreviated behavioral test based on SHIRPA that we called the modified SHIRPA. A single observer evaluated mice for weight, physical phenotype (tremor [1, none/ mild; 2, marked], body position [1, elongated; 2, hunched/rounded], and tail position [1, horizontally extended; 2, dragging/straub]), and behavior phenotype (grooming [1, slow/casual; 2, none/excessive], spontaneous

mcb.asm.org 4105

Dey et al.

activity [1, moderate, mouse covers all quadrants of cage; 2, slow, 1 to 3 quadrants; 3, none/darting/circling], and locomotor activity [the number of times the mouse places at least one of its paws on the side of the cage over a 2-min period]). RNA isolation and gene expression analysis. Total RNA was isolated from TRIzol lysates of each cerebellum from P5 WT, Smo/Smo, and SmoA2 cerebella (n ⫽ 3 animals per genotype) using the Promega SV-96 RNA isolation kit. Microarray analysis was performed using custom-designed Affymetrix arrays. Extracted RNA was quantified with RiboGreen RNA quantitation reagent (Invitrogen), and its quality was assessed by use of the Agilent RNA 6000 Pico kit (Agilent, Santa Clara, CA) in an Agilent 2100 Bioanalyzer (Agilent). Samples were amplified and labeled using the Ovation WB protocol (NuGen Technologies, San Carlos, CA), according to the manufacturer’s instructions. The resulting amplified cDNAs were hybridized to Affymetrix gene expression chips (Mouse Rosetta Custom Affymetrix 1.0; Affymetrix, Santa Clara, CA). Images were analyzed with Affymetrix GeneChip Operating Software (GCOS) and processed further to derive sequence-based intensities by use of the robust multiarray average (RMA) algorithm. The Rosetta Resolver system (Rosetta Biosoftware, Seattle, WA) was used to calculate fold changes and ratio P values for the differential expression of genes in each of three replicates of SmoA1 or SmoA2 mice versus a virtual pool of age-matched WT controls (n ⫽ 3). P values were calculated using the Rosetta intensitybased Affymetrix error model. Genes that were present in at least two of the three replicates for SmoA1 and SmoA2 with a P value of ⬍ 0.001 and an absolute average fold change of ⱖ2 were considered significantly differently expressed genes. For clustering analysis, we first filtered the data to remove data from probe sets whose expression levels did not significantly vary across the samples (defined as the top 50% least variable). The filtered data were then clustered using traditional hierarchical clustering. The Gene Ontology (GO) enrichment analysis was carried out using a hypergeometric test for each unique category. We selected significantly enriched categories, defined as any with a false discovery rate (FDR) of ⬍10%. Statistical tests. A Student t test was used to determine if differences observed between measurements obtained from two groups were statistically significant. Microarray data accession number. Microarray data have been deposited in the NCBI Gene Expression Omnibus (GEO) and are accessible through GEO series accession number GSE34593.

RESULTS

SmoA2 and SmoA1 mutations in the Smoothened receptor have vastly different effects on cerebellar development. SmoA2 (S537N) and SmoA1 (W539L) are activating point mutations that lead to constitutive Shh signaling and were originally identified in human cancer cases of medulloblastoma and basal cell carcinoma, respectively (31, 40, 45). In our previous studies, we described the SmoA1 transgenic mouse medulloblastoma model, which expresses the SmoA1 transgene driven by the GNP-specific fragment of the NeuroD2 (ND2) gene promoter, causing constitutive Shh signaling exclusively in the cerebellum (16, 18). In this study, we have characterized the SmoA2 transgenic mouse model with a similarly designed transgene expressing the SmoA2 mutation. Comparative histopathological analyses show striking differences in phenotypes—while SmoA1 mice have largely normal development of the cerebellum (Fig. 1G to I) similar to that of WT mice (Fig. 1A to C), SmoA2 mice have severe cerebellar malformations (Fig. 1D to F). To ensure that the phenotypes observed were not simply due to position effects at the sites of transgene insertion, we generated and evaluated multiple transgenic lines, which manifested similar phenotypes. At P5, the SmoA2 cerebellum has an extended, undefined ex-

4106

mcb.asm.org

ternal granular layer (EGL) consisting of aberrantly migrating GNPs (Fig. 1D). There is a lack of normal foliation and ectopic progenitor-like cells in the adjacent parenchyma and along the pial surface (Fig. 1D). At P14, the SmoA2 cerebellum continues to manifest extensive dysplasia with massive hypercellularity (Fig. 1E). The dysplastic regions consist of normal progenitor cells and atypical cuboidal to spindle-shaped cells with indistinct cell borders and a scant amount of eosinophilic fibrillar cytoplasm with irregularly round to fusiform nuclei, reminiscent of medulloblastoma. These features strongly suggest a primitive phenotype and possible neoplastic transformation. However, the atypical cells in the SmoA2 P14 cerebellum also have morphological similarities to the normal GNPs remaining in the outer EGL of the WT P14 cerebellum that are still undergoing migration. By P28, the WT cerebellum attains its mature size and shape (Fig. 1C). Although 100% of the SmoA2 cerebella remain dysplastic in adult mice, the cytoarchitecture is more mature than at the hypercellular P14 stage (Fig. 1F). The cells in the dysplasia are morphologically identical to the mature granule neurons in the WT cerebellum. At this stage, atypical cells, if any, are exclusively positioned along the pial surface, clearly separate from the adjacent dysplastic regions. Phenotypic differences between SmoA1 and SmoA2 mice do not result merely from differences in Smoothened expression. To determine whether the cerebellar developmental disorganization phenotype observed in SmoA2 mice in comparison to SmoA1 mice stems from differences in transgene copy numbers and the resulting expression of the SmoA1 and SmoA2 transgenes, we assessed differences at the genomic level, as well as total Smo mRNA and protein. We approximated transgene copy numbers in genomic DNA samples from SmoA1 (n ⫽ 5) and SmoA2 (n ⫽ 5) lines by comparing copies of Smo to WT mouse genomic DNA, using previously described qPCR-based methodologies (19, 24). qPCR analysis showed that SmoA1 and SmoA2 mice contain 39.1 ⫾ 7.7 (standard error of the mean [SEM]) and 6.9 ⫾ 1.5 times more copies of Smo than the WT reference, respectively (Fig. 2A and B). We investigated Smo expression by measuring total Smo mRNA and protein, that is, expression from the respective transgenes, as well as endogenous Smo. By qRT-PCR, we measured levels of total Smo and Smo-dependent canonical targets of the Shh pathway (Gli1, Gli2, Ptch1, and Ptch2) in SmoA1 and SmoA2 cerebella relative to age-matched WT controls. We accounted for the differences in cellular composition in SmoA1 and SmoA2 cerebella by using purified GNP lysates from all genotypes. We chose this specific developmental stage because (i) the phenotypes of SmoA2 and SmoA1 are distinct at P5 and (ii) GNP isolation at P5 allows accurate assessment of Smo and Shh target levels in a homogeneous cell population. Our results show that (i) although at an mRNA level, SmoA2 mice have ⬃3.8-fold higher Smo mRNA levels than SmoA1 mice, there is no meaningful difference (less than 2-fold) in the levels of activation of the Shh pathway, as is evident from similar levels of target mRNA in SmoA1 and SmoA2 mice (Fig. 2C). More importantly, at the protein level, there is no difference in total Smo protein expression between SmoA1 and SmoA2 mice (Fig. 2D). The above-mentioned results demonstrate that the phenotypic differences between SmoA1 and SmoA2 mice, with the latter being more penetrant, are not merely due to differences in transgene



FIG 1 The SmoA2 and SmoA1 mutations have vastly different effects on the development of the mouse cerebellum. H&E staining of representative sagittal sections shows development of the mouse cerebellum at P5, P14, and P28 in WT (A to C), SmoA2 (D to F), and SmoA1 (G to I) mice. (A to C) The asterisks indicate fissure formation during foliation at P5, and the dotted arrow denotes the anteroposterior (A-P) axis. At P14, formation of the molecular layer, Purkinje cell layer, and IGL is near completion, with remnants of the EGL remaining (arrow). At P28, the mature cerebellum is fully formed. (D to F) The SmoA2 cerebellum has an ill-defined expanded EGL with loss of foliation. The arrows point to ectopic clusters of cells at P5 and P14. In the P28 adult SmoA2 cerebellum, atypical cells with neoplastic morphology, if any, are concentrated along the superficial surface separated from the dysplastic region (dotted line). (G to I) The SmoA1 cerebellum closely parallels the development of the WT cerebellum, except for a slight thickening of the EGL at P14 (H, arrows). In the P28 adult SmoA1 cerebellum, neoplastic cells, if any, localize to the superficial surface of the cerebellum (I, arrows). The insets in panels A, D, and G show high-magnification views of the EGL and PC layer at P5, the inset in panel E shows a mix of atypical and normal cells at P14, and the inset in panel F shows the concentration of atypical cells along the superficial surface. Scale bars, 500 ␮m (A to I) and 25 ␮m (insets).

expression but stem from the biological differences between the two activating mutations. Medulloblastomas in the SmoA2 model develop in a dysplastic milieu. GNPs in the proliferative EGL are known to be the source of medulloblastomas caused by hyperactive Shh signaling (15, 18, 30, 35, 46). The SmoA2 mice, as shown earlier, lack an organized EGL, and the entire laminar cytoarchitecture of the cerebellum remains dysplastic throughout development (Fig. 1D to F). To understand the nature of tumor formation in these mice, we carried out comparative histopathological analyses on SmoA2 and SmoA1 tumors. The histological criteria we used for tumor definition were the same as those established in our previous study (18). In addition to our initial report of hyperplasia or lack of tumors in the SmoA2 line (16), our subsequent experiments, as described in this study, have revealed tumorigenesis.


SmoA2 mice develop cerebellar tumors in a dysplastic cerebellar milieu (Fig. 3A) compared to the SmoA1 mouse tumors, which form in an otherwise normally developed cerebellum (Fig. 3D) and are preceded by an expansion of the EGL at P14 (18) (Fig. 1H). The frank solid tumors are densely cellular, with oval to spindleshaped cells (Fig. 3B and F) arranged in undefined sheets, short streams, and bundles supported on a scant fibrovascular stroma. Multifocally neoplastic cells palisade along vessels and form numerous pseudorosettes. Normal cellular morphology (Fig. 3C and E) is distinct from that of adjacent tumor cells. Although during development the entire SmoA2 cerebellum appears hypercellular with atypical cells and other features of neoplasia (Fig. 1D to F), even in the absence of a defined EGL, the solid tumors are localized in anatomic regions along the pial surface (Fig. 3A). A natural history study to determine the clinical incidence of

mcb.asm.org 4107

Dey et al.

FIG 2 Determination of Smo transgene copy number and expression in SmoA1 and SmoA2 transgenic mice. (A) Standard curves obtained by qPCR amplification of Smo-Exon 6, Exon 10, and control loci RNaseP and Gapdh from 10-fold serial dilutions of WT mouse genomic DNA (190 to 0.019 ng). The data are plotted as cycle threshold versus log picogram DNA. The transgene copy number was confirmed by two sets of Smo and control locus primers. (B) Transgene copy number in SmoA1 and SmoA2 hemizygous mice estimated relative to endogenous Smo in the reference WT genomic DNA (n ⫽ 5 per genotype). (C) mRNA expression of canonical Shh target genes Gli1, Gli2, Ptch1, and Ptch2, as well as Smo (endogenous plus transgene), as determined by qRT-PCR analysis using mRNA from WT, SmoA1, or SmoA2 purified P5 GNP lysates (3 independent pools per genotype). The bars represent the average fold change of each target in SmoA1 and SmoA2 relative to WT controls. ␤2m was used as the endogenous control for data normalization. Each assay was run in triplicate. The asterisks denote statistically significant differences in mRNA expression between SmoA1 and SmoA2 (P ⬍ 0.01; determined by Student t test). (D) Representative Western blot image and densitometric analyses of total Smo protein from WT, SmoA1, or SmoA2 purified P5 GNP protein lysates (3 independent pools per genotype) show no difference in total Smo expression between the genotypes. The bars represent the average fold change of Smo protein relative to the WT control. ␤-Actin was used as the loading control. The error bars represent SEM.

tumor formation in SmoA2 mice (n ⫽ 235) showed 2% of the SmoA2 mice manifested clinical symptoms of tumor formation at 2 months of age, which increased to 46% by 4 months and 76% by 6 months (Fig. 3G), similar to the SmoA1 model (18). The main clinical symptoms, as previously described (18), were weight loss, protruded head as a result of tumor formation beneath the skull, hunched posture or head tilt resulting from hydrocephalus or nerve impingement, and lethargy. Histological analyses using asymptomatic SmoA2 mice showed 80% of SmoA2 mice (n ⫽ 15) have subclinical neoplastic lesions by 2 months, which increases to 100% by 4 months of age (n ⫽ 16). The presence of neoplastic features as early as P5 indicated that the SmoA2 mutation confers transformative potential on precursor cells early in development, much before the onset of clinical disease. To investigate the oncogenic potential of SmoA2 precursors, we transplanted P5 SmoA2 GNPs orthotopically into immunocompromised recipient mice. All five recipients succumbed to aggressive tumors between 20 and 30 days after transplantation (Fig. 3H), confirming that the SmoA2-expressing GNPs have neo-

4108

mcb.asm.org

plastic properties and lead to aggressive tumorigenesis. The allograft tumors, similar to the spontaneously arising tumors in the transgenic SmoA2 model, extend from the outer surface multifocally, compressing the cerebellum, and are a highly cellular, unencapsulated mass composed of two populations of cells: round cells with hyperchromatic nuclei and spindle-shaped cells with oval to elongated nuclei (Fig. 3I). The simultaneous manifestation of cerebellar developmental defects, as well as neoplastic changes, as early as P5 in the SmoA2 mice makes this a powerful model to study the temporal progression of an embryonal tumor like medulloblastoma. The SmoA2 mutation independently increases neuronal proliferation and affects neuronal migration in the cerebellum. The Shh signaling pathway plays a crucial role in proliferation of GNPs by acting as a mitogen during postnatal cerebellar development (44). We therefore characterized the proliferative status of the SmoA2 cerebellum at different developmental stages. The first developmental stage we investigated was embryonic day 15.5 (E15.5), since the EGL, consisting of proliferating GNPs, is



FIG 3 Comparative histopathology of SmoA2 and SmoA1 medulloblastomas. (A) Representative images from horizontal cerebellar sections show formation of SmoA2 tumors (arrow) in an aberrantly developed cerebellum on the superficial surface. m, meninges. (B and C) Pleomorphic neoplastic cells with multifocal pseudorosettes (B) are distinct from uniformly round granule neurons (GN) and Purkinje cells (PC) (C) in adjacent dysplasia. (D) SmoA1 tumors (arrow) form adjacent to a normally developed cerebellum also along the superficial surface. (E and F) Normal cellular morphology (E) is distinct from that of adjacent tumor cells (F). (G) Cumulative tumor incidence in SmoA2 and SmoA1 (18) mice at 2, 4, and 6 months of age, based on clinical symptoms. (H) Representative allograft tumor in athymic recipient mice (n ⫽ 5) transplanted orthotopically with P5 SmoA2 GNPs. (I) High-magnification image showing a highly cellular mass with two populations of cells, round cells with hyperchromatic nuclei (white arrows) and spindle-shaped cells with oval to elongated nuclei (black arrows). T, allograft tumor; Cbl, normal cerebellum of recipient. Scale bars, 100 ␮m.

formed by this stage (3) and the ND2 promoter driving the SmoA2 transgene is activated at ⬃E11 (23). Immunohistochemical (IHC) staining for the proliferation marker Ki67 shows an increased number of proliferating progenitors in the expanded EGL com-


pared to WT controls (Fig. 4A and B). This feature is maintained at P5 with disorganized proliferating cells, which persist at P14 in SmoA2 mice (Fig. 4F and G) as opposed to WT mice, where proliferation is high at P5 but nearly complete by P14 (Fig. 4C and D).

mcb.asm.org 4109

Dey et al.

FIG 4 The SmoA2 mutation has independent effects on neuronal proliferation and neuronal and glial organization. (A and B) Ki67 IHC identifying proliferating GNPs marks the EGL (arrow), which is expanded in SmoA2 mice compared to WT mice at E15.5. IHC analysis for Ki67 in WT cerebella at P5, P14, and P28 (C to E) compared to that in SmoA2 cerebella (F to H) shows disorganized regions of proliferating GNPs (arrows) in SmoA2 at P5 compared to the defined EGL in the WT. At P14, while a few proliferating cells remain in the outer EGL (arrow) in the WT, there is an abundance of proliferating GNPs in SmoA2. At P28, proliferation is complete in the WT, whereas Ki67-positive cells are present along the superficial surface of the SmoA2 cerebellum. (I and J) SmoA2 tumors have an abundance of Ki67-positive cells (I), which are undetectable in the dysplastic regions of the SmoA2 cerebella (J). (K to M) H&E staining of representative sagittal cerebellar sections showing that while the EGL is expanded in both the PtchF/F-Math1-Cre and SmoA2 models, the mechanisms underlying foliation and neuronal migration to form the laminar cytoarchitecture are preserved in the former as opposed to the SmoA2 cerebellum. n ⫽ 3 per group. Scale bars, 100 ␮m.

In the P28 SmoA2 cerebellum, the Ki67-positive cells remain on the outer surface of the cerebellum in the same region (Fig. 4H) where the atypical cells are localized, as shown in Fig. 1F. There are no Ki67-positive cells in the P28 WT cerebellum (Fig. 4E). The

4110

mcb.asm.org

SmoA2 tumors have extensive Ki67 staining compared to the dysplastic regions, where Ki67 is undetectable (Fig. 4I and J). Next, we investigated whether the abnormal foliation and migration observed in the SmoA2 cerebellum is a consequence of



excess GNPs generated from Shh hypersignaling that could potentially overwhelm and deregulate cell migration processes. To address this, we compared the P5 cerebella of the SmoA2 mice to an additional Shh-driven medulloblastoma model, the Patched conditional knockout mice (PtchF/F Math1-Cre) (46) (Fig. 4K, L, and M). At P5, the PtchF/F Math1-Cre mice have a vastly increased number of GNPs, which may give rise to nodular structures in some regions (46). However, neuronal migration to form the laminar architecture and mechanisms underlying foliation appear to be preserved (Fig. 4L). This shows that the neuronal disorganization in the SmoA2 cerebellum is not solely a consequence of uncontrolled GNP proliferation. SmoA2 mutation causes aberrations in neuronal and glial organization. To investigate the cellular characteristics of the cerebellar dysplasias observed in the SmoA2 mice, we identified granule neurons and Purkinje cells, as well as Bergmann glia, that are critical in GNP migration using antibodies for NeuN, Calbindin, and S100, respectively. By immunohistochemical techniques, we analyzed three distinct stages of cerebellar development, namely, P5, P14, and P28, as well as SmoA2 tumors and nontumor cerebellar dysplasia in adult mice. All three markers showed massive disruptions in the organization of granule neurons, Purkinje cells, and radial glia in SmoA2 mice (Fig. 5D to F, J to L, and P to R) compared to the WT littermate controls (Fig. 5A to C, G to I, and M to O). At P5, the SmoA2 cerebellum has an ill-defined outer region negative for NeuN and an inner region of NeuN-positive mature neurons similar to the forming internal granular layer (IGL) of the WT cerebellum (Fig. 5A and D). At P14 in the SmoA2 cerebellum, there are clusters of NeuN-positive neurons marking islands of differentiated cells amid a vast expanse of NeuN-negative cells (Fig. 5E), in stark contrast to the WT cerebellum, where differentiated NeuN-positive cells are located in the nearly complete IGL at this stage (Fig. 5B). At P28 in the SmoA2 cerebellum, a majority of the neurons are NeuN positive, representing a differentiated state with clusters of NeuN-negative cells toward the outer surface of the cerebellum (Fig. 5F), the same region where clusters of Ki67-positive proliferating cells are observed (Fig. 4H). Calbindin immunostaining shows that the Purkinje cells have severely disorganized dendritic arbors and axonal processes in the SmoA2 cerebellum, and the cell bodies fail to align in the characteristic monolayer array at all the developmental stages analyzed (Fig. 5J to L) in comparison to WT controls (Fig. 5G to I). Compared to the WT controls (Fig. 5M to O), S100 staining of the radial glia showed atypical glial tangles with entrapped granule neurons in ectopic locations in the SmoA2 mice (Fig. 5P to R). The SmoA2 tumors show a heterogeneous pattern of NeuN staining (Fig. 5S), absence of Calbindin-positive Purkinje cells (Fig. 5T), and sparse to absent S100-positive glia (Fig. 5U). The SmoA2 adult dysplasias, on the other hand, consist of NeuN-positive neuronal cells (Fig. 5V). The dysplastic regions also consist of disorganized Purkinje cells (Fig. 5W) and disorganized glial fibrils (Fig. 5X). These results demonstrate that the GNP-specific expression of SmoA2 leads to widespread disruptions in the migration and organization of other neuronal and glial cell types in the cerebellum. Severe defects in cerebellar morphology do not cause overt anomalies in motor coordination in SmoA2 mice. The classic functions of the cerebellum, which are regulation of balance and motor coordination, are thought to be dependent on its organized


laminar cytoarchitecture (11). The granule neurons carry sensory inputs to the cerebellum, while the Purkinje cells are the primary motor output from the cerebellum, relaying motor information to higher brain centers. To assess potential abnormalities in motor coordination resulting from neuroanatomic defects of the SmoA2 cerebellum, we used a modified neurobehavioral assessment tool based on SHIRPA (33) (see Materials and Methods for details). We investigated the physical phenotype (weight, tremor, body position, and tail position) and the behavioral phenotype (grooming, locomotor activity, and spontaneous motor activity) (see Movie S1 in the supplemental material). Additionally, we conducted the accelerated RotaRod assay to test fore and hind limb coordination and balance in both adult SmoA2 and age-matched WT control mice. No differences in physical phenotype were observed, and except for a decrease in locomotor activity (P ⬍ 0.01), there were no significant differences in behavior phenotype measures between WT and SmoA2 mice (Table 1). Importantly, there was no difference in RotaRod performance between SmoA2 and WT mice, demonstrating that these mice were not deficient in motor coordination and balance. We also conducted a footprint analysis using stride length and stance width measures (9) in both SmoA2 and WT mice, which demonstrated no evidence of ataxic gait, a common consequence of cerebellar dysfunction (Table 1). Together, these data reveal no overt abnormalities in cerebellar function in SmoA2 mice despite massive disorganization in the cytoarchitecture. SmoA2 and SmoA1 mutations lead to unique transcriptional profiles in the mouse cerebellum. Since the SmoA2 and SmoA1 cerebella have different cell compositions, we first determined if the cell-autonomous effects of GNP-specific SmoA2 and SmoA1 are represented in the whole cerebella. We compared, by qRTPCR, activation of the Shh pathway through its canonical targets in whole cerebellar lysates versus in purified GNPs, as determined earlier (Fig. 2C and D). Our results show similar results in whole cerebellar lysates, as well, thereby demonstrating that despite differences in cell composition, cell-autonomous effects of the mutations in GNPs were represented in whole-cerebellar analysis (Fig. 6A). Next, to assess transcriptional changes downstream of SmoA2 and SmoA1, we evaluated global gene expression profiles of P5 SmoA2, SmoA1, and WT age-matched whole cerebella. We chose this specific developmental stage because (i) the phenotypes of SmoA2 and SmoA1 are robust and distinct at P5 and (ii) at P5, GNPs undergo proliferation, migration, and differentiation, and therefore, expression profiling could capture key differences in multiple processes. Hierarchical cluster analysis showed that while the gene expression signature is unique for each group, at P5, transcriptionally, the WT and SmoA1 cerebella resemble each other more closely than either resembles SmoA2 cerebella (Fig. 6B), consistent with their phenotypes (Fig. 1A, D, and G). Compared to WT controls, we identified 106 transcripts in SmoA2 and 67 in SmoA1 that differed uniquely by an absolute change of ⱖ2-fold (P ⬍ 0.001) (Fig. 6C; see Table S2 in the supplemental material). This highlights the complex transcriptional changes induced by the SmoA2 mutation that potentially underlie the intriguing phenotype. We confirmed by qRT-PCR, using purified GNP mRNA lysates, a subset of genes that were uniquely upregulated in the SmoA2 cerebella or in both SmoA1 and SmoA2 cerebella, according to the gene expression data (Fig. 6D). Bcl11b, MyoD, Pou4f2, and Cbln4 were uniquely upregulated in SmoA2

mcb.asm.org 4111

Dey et al.

FIG 5 SmoA2-induced aberrations in neuronal and glial organization in the developing cerebellum. (A to F) NeuN IHC analysis at P5, P14, and P28 developmental stages comparing WT cerebella (A to C) to SmoA2 cerebella (D to F) showing disorganized clusters of differentiated granule neurons (GN) amid abundant NeuN-negative cells at P5 and P14 in SmoA2 cerebella. (G to L) Calbindin IHC for Purkinje cells showing monolayer organization in WT (G to I) and ectopic clusters in SmoA2 (J to L) with disorganized dendritic arbors and axonal collaterals at P5, P14, and P28. (M to R) S100 IHC analysis for radial glia comparing WT cerebella (M to O) to SmoA2 cerebella (P to R) showing disorganized, entangled glial processes in the SmoA2 cerebellum at P5, P14, and P28. (S to X) SmoA2 tumors show heterogeneous NeuN staining, absence of PC, and heterogeneous S100 staining (S to U), whereas adjacent dysplasias are composed of mature but disorganized GN, PC, and radial glia (V to X). n ⫽ 3 per group. Scale bars, 100 ␮m.

mice, and Isl1 was upregulated in both SmoA1 and SmoA2 mice, showing consistency between the microarray data using whole cerebella and the cell-autonomous effects of the transgenes in purified GNPs. To determine key biological processes that are different between SmoA2 and SmoA1 mice, we used mRNAs over- and underrepresented in SmoA2 compared to SmoA1 mice to conduct GO gene set enrichment analyses (1). The ontological categories represented by the gene sets up- or downregulated in SmoA2 com-

4112

mcb.asm.org

pared to SmoA1 mice include cell and neuronal fate specification and commitment, regulation of neuronal differentiation, neural crest cell migration, regulation of cell proliferation, neuron migration, neuronal projection, membrane anchorage, cell matrix adhesion, axonal and dendritic molecules, localization molecules, ligand gated ion channel activity, and thyroid hormone metabolism, as well as various metabolic processes. In the cell fate commitment GO category, among other transcription factors, the expression of MyoD, encoding a transcription factor that orches-



TABLE 1 Measures from neurobehavioral assessment WT Test

Measure

n

RotaRod Footprint analysis

Latency to fall (s) Stride length (cm) Stance width (cm) No. of times mouse places at least 1 paw on side of cage during a 2-min assessment

8 3 3 8

Locomotor activitya a b

b

SmoA2 Mean

SEM

n

Mean

SEM

222.4 63.2 22.5 12.1

1.04 3.84 0.64 0.07

12 4 4 12

189.8 60.4 23.3 6.5

0.59 2.13 1.03 0.02

P value ⬍ 0.01. n, number of mice in each group.

trates muscle differentiation, was particularly intriguing, since MyoD is not known to be expressed or to have a function in the mammalian brain. qRT-PCR (Fig. 6D), as well as Western blot analysis (data not shown), confirmed this ectopic expression pat-

tern. We are currently exploring the significance of this unexpected finding. In summary, the biological processes reflected by the gene expression profiles are complex, and ascertaining specific pathways remains a direction for the future. Our data demonstrate the fundamental differences between SmoA2 and SmoA1 at a molecular level that affect distinct biological processes. DISCUSSION

FIG 6 Transcriptional profiles of SmoA2 and SmoA1 P5 cerebella. (A) mRNA expression of canonical Shh target genes Gli1, Gli2, Ptch1, Ptch2, and Smo (endogenous plus transgene) as determined by qRT-PCR analysis using mRNA from WT, SmoA2, or SmoA1 P5 cerebella (n ⫽ 3 per genotype). The bars represent the average fold change of each target in SmoA2 and SmoA1 relative to WT controls. Gapdh was used as the endogenous control. Each assay was run in triplicate. (B) Hierarchical cluster analysis showing transcriptional similarity of WT and SmoA1 P5 cerebella in comparison to SmoA2 cerebella (n ⫽ 3 per genotype). (C) Venn diagram showing unique and overlapping changes in transcript expression (ⱖ2fold absolute change; P ⬍ 0.001) between P5 SmoA2 and SmoA1 cerebella. (D) Confirmation of differential expression of transcripts between P5 SmoA2 and SmoA1 whole cerebella as determined in microarray experiments and by qRTPCR in purified GNP lysates. The bars represent the average fold change of each mRNA target in SmoA2 and SmoA1 GNP lysates relative to the WT (n ⫽ 3 independent pools). Ppia was used as the endogenous control. The asterisks indicate statistically significant differences in mRNA expression between SmoA2 and SmoA1 (*, P ⬍ 0.01; **, P ⬍ 0.05; Student t test). The error bars represent SEM.


In this study, we have established a mouse model of medulloblastoma that manifests severe defects in critical pathways of cerebellar development, including neuronal proliferation, differentiation, and migration, stemming from an activating mutation, SmoA2. Our comparative analysis reveals vastly different phenotypic effects of the SmoA2 and SmoA1 mutations and thereby demonstrates the complexity of the downstream molecular pathways regulated by a single molecule. The lack of a significant difference in the levels of activation of the Shh pathway at P5 suggests that the phenotypic differences are not exclusively due to the extent of pathway activation. The similarity in global transcriptional profiles between the WT and SmoA1 mice in early development is in accordance with the phenotype of SmoA1 being indistinguishable from that of the WT at that stage, whereas the SmoA2 phenotype and transcriptional profile are unique. We have demonstrated how two activating mutations in identical domains of a single protein can cause unique changes. Identifying molecular pathways uniquely employed by SmoA2 and SmoA1 variants with cues from the GO classification remains an important future direction that may provide further insights into the mechanics of the Shh pathway in both normal development and tumorigenesis. The cell of origin of the Shh-driven subtype of medulloblastomas has been shown to belong to the granule neuron lineage (12, 15, 16, 18, 30, 35, 46). However, why the niche of the superficial surface of the cerebellum is conducive to tumorigenesis, as seen in several mouse models (18, 30, 46), including SmoA2, has yet to be understood. In the Patched heterozygous model (30), ectopic rests (discrete clusters) of preneoplastic cells that have failed to undergo proper migration reside in the niche of the EGL, whereas the vast majority of GNPs mature and organize correctly. Intriguingly, in the SmoA2 developing cerebellum, the vast majority of GNPs undergo abnormal migration and manifest neoplastic features early in development. However, the frank tumors remain confined to the superficial surface of the cerebellum. By design of the transgene, each GNP should express the SmoA2 oncogenic mutation, yet dysplasias adjacent to the tumors consist of normally differentiated granule neurons. This suggests that potential cell-extrinsic factors in the pial surface might act on a subset of SmoA2 cells arrested in migration, leading to tumor initiation. The cells that have migrated away from the EGL differentiate or regress. Alter-

mcb.asm.org 4113

Dey et al.

natively, cell-extrinsic factors could attract neoplastic cells to the pial surface by providing a favorable environment for tumor growth. The leptomeningeal membrane, known to secrete chemokines and other trophic factors (21, 39), is one potential source of such extrinsic signals. Future experiments to further characterize the environmental niche and the resident precursors will provide further insights into the development of this subtype of medulloblastoma. An interesting phenomenon observed in the SmoA2 developing cerebellum is the apparent regression of the hypercellularity observed throughout the developing cerebellum. Progenitor cells in the developing cerebellum appear to have neoplastic characteristics morphologically, as well as functionally, as shown by transplantation experiments. However, in a mature SmoA2 cerebellum, the dysplasias histologically seem to have fewer cells and to consist of mature neurons. The early neoplastic lesions are confined to the superficial surface of the cerebellum, while the rest of the cerebellum remains dysplastic. Spontaneous regression has been documented in certain cancers, such as the highly malignant stage 1V-S neuroblastoma (10, 27), the molecular basis of which is poorly understood. There are conflicting data regarding the role of apoptosis in spontaneous regression in neuroblastoma (47). We have been unable to detect any significant apoptotic cell death by immunohistochemistry for activated caspase 3 (data not shown). Whether the processes underlying the apparent regression observed in the SmoA2 cerebellum include delayed neuronal differentiation programs and caspase-dependent and possibly caspase-independent programmed cell death, like autophagic degeneration, as implicated in neuroblastoma (20), has yet to be determined. Shh signaling acting through the Smo-Gli axis has been shown to regulate foliation, with the extent of foliation proportional to the level of signaling (7, 8). However, increased cell-autonomous Shh signaling in the SmoA2 model leads to disruptions distinct from phenotypes reported in previous studies (7, 8, 16, 18, 46). We have demonstrated that the disorganized cytoarchitecture of the SmoA2 cerebellum cannot be attributed exclusively to increased proliferation of GNPs, as defects in foliation and migration are not observed in the Ptch conditional-knockout mice (46). The effects of cell-autonomous Shh signaling on neuronal migration have yet to be understood. Further investigation of the SmoA2 gene expression profile will potentially provide further insights into the role of Shh signaling in neuronal migration in cerebellar development. The stereotyped neuronal circuitry of the cerebellum, critical for its function as a motor coordination center, depends on the stereotypic arrangement and distinct morphologies of Purkinje cells, granule neurons, and the deep cerebellar neurons (6). In addition to the reiterative circuitry, it has been shown that topological gene expression patterns in Purkinje cells define longitudinal domains that are essential for proper targeting of incoming cerebellar afferents (28, 38, 43). Both the morphology and organization of Purkinje cells and granule neurons are severely disrupted in the SmoA2 cerebellum, yet the lack of detectable behavioral anomalies in SmoA2 mice suggests that the transmission of afferent and efferent signals is largely maintained. Mutant mice, such as reeler and staggerer mice, have neuroanatomic defects that are most pronounced in the cerebellum (13, 14, 32), with ectopic Purkinje cells, lack of foliation, and a reduction in the number of granule neurons. These mice display ataxia and uncoordinated

4114

mcb.asm.org

movement characteristic of cerebellar malfunctions, yet intriguingly, SmoA2 mutants with some phenotypic similarities in terms of disruptions in the laminar architecture, with more severe morphological abnormalities in multiple cell types, maintain grossly normal motor behavior. In summary, through the characterization of the unique phenotype of the SmoA2 model, we address key aspects of cerebellar development and function, as well as of medulloblastoma biology, that have yet to be understood. The SmoA2 model is therefore a valuable addition to the existing mouse models of medulloblastoma. Medulloblastoma is known to be a cancer resulting from deregulated developmental signals. The developmental phenotype of the SmoA2 model will therefore allow investigation of cerebellar developmental pathways and aberrations thereof that lead to medulloblastoma formation. ACKNOWLEDGMENTS This work was supported by NIH grants 5R01CA11456705 and 5R01CA112350. J.D. was supported by Predoctoral Developmental Biology Training Grant 5T32HD007183 from the National Institute of Child Health and Human Development. We thank Kyle Pedro and Andrew Richards for assistance with mouse colony management; Lindsey Conrad for assistance with database maintenance; Ying-tzang Tien at the University of Washington Histopathology Laboratory, Julie-Randolph Habecker and the FHCRC Experimental Histopathology Shared Resource, Julio Vazquez and David L. McDonald at FHCRC Scientific Imaging, and the Rosetta Inpharmatics Gene Expression Laboratory for profiling studies; Matthew Tudor for compiling the results for GEO submission; Andrew Strand for assistance with microarray data analysis; Michael LeBlanc for advice with statistical analysis; and Stacey Hansen, Olga Klezovitch, and Michelle Cook Sangar for critical review of the manuscript.

REFERENCES 1. Ashburner M, et al. 2000. Gene ontology: tool for the unification of biology. The Gene Ontology Consortium. Nat. Genet. 25:25–29. 2. Ayrault O, et al. 2010. Atoh1 inhibits neuronal differentiation and collaborates with Gli1 to generate medulloblastoma-initiating cells. Cancer Res. 70:5618 –5627. 3. Ben-Arie N, et al. 1997. Math1 is essential for genesis of cerebellar granule neurons. Nature 390:169 –172. 4. Berman DM, et al. 2002. Medulloblastoma growth inhibition by hedgehog pathway blockade. Science 297:1559 –1561. 5. Chen JK, Taipale J, Cooper MK, Beachy PA. 2002. Inhibition of Hedgehog signaling by direct binding of cyclopamine to Smoothened. Genes Dev. 16:2743–2748. 6. Chizhikov V, Millen KJ. 2003. Development and malformations of the cerebellum in mice. Mol. Genet. Metab. 80:54 – 65. 7. Corrales JD, Blaess S, Mahoney EM, Joyner AL. 2006. The level of sonic hedgehog signaling regulates the complexity of cerebellar foliation. Development 133:1811–1821. 8. Corrales JD, Rocco GL, Blaess S, Guo Q, Joyner AL. 2004. Spatial pattern of sonic hedgehog signaling through Gli genes during cerebellum development. Development 131:5581–5590. 9. Crawley JN. 1999. Behavioral phenotyping of transgenic and knockout mice: experimental design and evaluation of general health, sensory functions, motor abilities, and specific behavioral tests. Brain Res. 835:18 –26. 10. D’Angio GJ, Evans AE, Koop CE. 1971. Special pattern of widespread neuroblastoma with a favourable prognosis. Lancet i:1046 –1049. 11. Ghez C, Fahn S. 1985. The cerebellum, p 502–522. In Kandel ER, Schwartz JH, Principles of neural science, 2nd ed. Elsevier, New York, NY. 12. Gilbertson RJ, Ellison DW. 2008. The origins of medulloblastoma subtypes. Annu. Rev. Pathol. 3:341–365. 13. Goldowitz D, et al. 1997. Cerebellar disorganization characteristic of reeler in scrambler mutant mice despite presence of reelin. J. Neurosci. 17:8767– 8777. 14. Goldowitz D, Hamre K. 1998. The cells and molecules that make a cerebellum. Trends Neurosci. 21:375–382.



15. Goodrich LV, Milenkovic L, Higgins KM, Scott MP. 1997. Altered neural cell fates and medulloblastoma in mouse patched mutants. Science 277:1109 –1113. 16. Hallahan AR, et al. 2004. The SmoA1 mouse model reveals that notch signaling is critical for the growth and survival of sonic hedgehog-induced medulloblastomas. Cancer Res. 64:7794 –7800. 17. Hatten ME, Roussel MF. 2011. Development and cancer of the cerebellum. Trends Neurosci. 34:134 –142. 18. Hatton BA, et al. 2008. The Smo/Smo model: hedgehog-induced medulloblastoma with 90% incidence and leptomeningeal spread. Cancer Res. 68:1768 –1776. 19. Huntzicker EG, et al. 2006. Dual degradation signals control Gli protein stability and tumor formation. Genes Dev. 20:276 –281. 20. Kitanaka C, et al. 2002. Increased Ras expression and caspaseindependent neuroblastoma cell death: possible mechanism of spontaneous neuroblastoma regression. J. Natl. Cancer Inst. 94:358 –368. 21. Klein RS, et al. 2001. SDF-1 alpha induces chemotaxis and enhances Sonic hedgehog-induced proliferation of cerebellar granule cells. Development 128:1971–1981. 22. Lee HY, Greene LA, Mason CA, Manzini MC. 2009. Isolation and culture of post-natal mouse cerebellar granule neuron progenitor cells and neurons. J. Vis. Exp. 16:990. 23. Lin CH, et al. 2004. Regulation of neuroD2 expression in mouse brain. Dev. Biol. 265:234 –245. 24. Lofgren S, et al. 2008. Generation of mice transgenic for human CYP2C18 and CYP2C19: characterization of the sexually dimorphic gene and enzyme expression. Drug Metab. Dispos. 36:955–962. 25. Marino S. 2005. Medulloblastoma: developmental mechanisms out of control. Trends Mol. Med. 11:17–22. 26. Markant SL, Wechsler-Reya RJ. 2012. Personalized mice: modeling the molecular heterogeneity of medulloblastoma. Neuropathol. Appl. Neurobiol. 38:228 –240. 27. Nickerson HJ, et al. 2000. Favorable biology and outcome of stage IV-S neuroblastoma with supportive care or minimal therapy: a Children’s Cancer Group study. J. Clin. Oncol. 18:477– 486. 28. Nishida K, Flanagan JG, Nakamoto M. 2002. Domain-specific olivocerebellar projection regulated by the EphA-ephrin-A interaction. Development 129:5647–5658. 29. Northcott PA, et al. 2011. Pediatric and adult sonic hedgehog medulloblastomas are clinically and molecularly distinct. Acta Neuropathol. 122: 231–240. 30. Oliver TG, et al. 2005. Loss of patched and disruption of granule cell development in a pre-neoplastic stage of medulloblastoma. Development 132:2425–2439. 31. Reifenberger J, et al. 1998. Missense mutations in SMOH in sporadic


32.

33.

34. 35.

36. 37.

38. 39.

40. 41. 42. 43.

44. 45. 46.

47.

basal cell carcinomas of the skin and primitive neuroectodermal tumors of the central nervous system. Cancer Res. 58:1798 –1803. Rice DS, Curran T. 1999. Mutant mice with scrambled brains: understanding the signaling pathways that control cell positioning in the CNS. Genes Dev. 13:2758 –2773. Rogers DC, et al. 1997. Behavioral and functional analysis of mouse phenotype: SHIRPA, a proposed protocol for comprehensive phenotype assessment. Mamm. Genome 8:711–713. Saury JM, Emanuelson I. 2011. Cognitive consequences of the treatment of medulloblastoma among children. Pediatr. Neurol. 44:21–30. Schuller U, et al. 2008. Acquisition of granule neuron precursor identity is a critical determinant of progenitor cell competence to form Shhinduced medulloblastoma. Cancer Cell 14:123–134. Sengupta R, et al. 2012. CXCR4 activation defines a new subgroup of Sonic Hedgehog driven Medulloblastoma. Cancer Res. 72:122–132. Shu Q, et al. 2008. Direct orthotopic transplantation of fresh surgical specimen preserves CD133⫹ tumor cells in clinically relevant mouse models of medulloblastoma and glioma. Stem Cells 26:1414 –1424. Sotelo C, Chedotal A. 1997. Development of the olivocerebellar projection. Perspect. Dev. Neurobiol. 5:57– 67. Stylianopoulou F, Herbert J, Soares MB, Efstratiadis A. 1988. Expression of the insulin-like growth factor II gene in the choroid plexus and the leptomeninges of the adult rat central nervous system. Proc. Natl. Acad. Sci. U. S. A. 85:141–145. Taipale J, et al. 2000. Effects of oncogenic mutations in Smoothened and Patched can be reversed by cyclopamine. Nature 406:1005–1009. Taylor MD, et al. 2012. Molecular subgroups of medulloblastoma: the current consensus. Acta Neuropathol. 123:465– 472. Wang VY, Zoghbi HY. 2001. Genetic regulation of cerebellar development. Nat. Rev. Neurosci. 2:484 – 491. Wassef M, Cholley B, Heizmann CW, Sotelo C. 1992. Development of the olivocerebellar projection in the rat: II. Matching of the developmental compartmentations of the cerebellum and inferior olive through the projection map. J. Comp. Neurol. 323:537–550. Wechsler-Reya RJ, Scott MP. 1999. Control of neuronal precursor proliferation in the cerebellum by Sonic Hedgehog. Neuron 22:103–114. Xie J, et al. 1998. Activating Smoothened mutations in sporadic basal-cell carcinoma. Nature 391:90 –92. Yang ZJ, et al. 2008. Medulloblastoma can be initiated by deletion of Patched in lineage-restricted progenitors or stem cells. Cancer Cell 14: 135–145. Yu F, et al. 2011. Proteomics-based identification of spontaneous regression-associated proteins in neuroblastoma. J. Pediatr. Surg. 46:1948 – 1955.

mcb.asm.org 4115