Genetic Alterations in Mouse Medulloblastomas and Generation of Tumors De novo from Primary Cerebellar Granule Neuron Precursors

Mice lacking p53 and one or two alleles of the cyclin D–dependent kinase inhibitor p18 are prone to medulloblastoma development. The tumor frequency is increased by exposing postnatal animals to ionizing radiation at a time when their cerebella are developing. In irradiated mice engineered to express a floxed p53 allele and a Nestin-Cre transgene, tumor development can be restricted to the brain. Analysis of these animals indicated that inactivation of one or both Ink4c alleles did not affect the time of medulloblastoma onset but increased tumor invasiveness. All such tumors exhibited complete loss of function of the Patched 1 (Ptc1) gene encoding the receptor for sonic hedgehog, and many exhibited other recurrent genetic alterations, including trisomy of chromosome 6, amplification of N-Myc, modest increases in copy number of the Ccnd1 gene encoding cyclin D1, and other complex chromosomal rearrangements. In contrast, medulloblastomas arising in Ptc1 mice lacking one or both Ink4c alleles retained p53 function and exhibited only limited genomic instability. Nonetheless, complete inactivation of the wild-type Ptc1 allele was a universal event, and trisomy of chromosome 6 was again frequent. The enforced expression of N-Myc or cyclin D1 in primary cerebellar granule neuron precursors isolated from Ink4c / , p53 / mice enabled the cells to initiate medulloblastomas when injected back into the brains of immunocompromised recipient animals. These ‘‘engineered’’ tumors exhibited gene expression profiles indistinguishable from those of medulloblastomas that arose spontaneously. These results underscore the functional interplay between a network of specific genes that recurrently contribute to medulloblastoma formation. [Cancer Res 2007;67(6):2676–84]


Introduction
Medulloblastoma and supratentorial primitive neuroectodermal tumors are the most common malignant brain tumors in children.Although modern therapeutic protocols cure >60% of such children, the survivors' quality of life is usually severely compromised due to adverse effects of therapy on neurocognitive functions.Human medulloblastomas can be classified into multiple histologic variants (1), including the large cell anaplastic type that is thought to be the least curable (2).Through a combination of molecular techniques, several genes encoding proteins that act within developmental signaling pathways have been implicated in medulloblastoma development, and the contributions of some have been tested in mouse models of the disease.Alterations of genes in the sonic hedgehog (SHH)/PATCHED (PTC1) signaling pathway as well as mutations in b-CATENIN and NOTCH genes play roles in human medulloblastoma (3)(4)(5)(6)(7).In most mouse models, inactivation of p53 is a prerequisite for medulloblastoma development, whereas, in humans, TP53 mutations are infrequent and do not correlate with specific aberrations in Ptc, Notch, or h-catenin signaling (8).
During mouse development, cerebellar granule neuron progenitors (CGNPs) rapidly divide in the external granule layer (EGL) from postnatal day (P) 1 to P7, after which they exit the cell cycle and migrate inward through the Purkinje cell layer to form the internal granule layer (IGL).The latter zone is composed uniquely of postmitotic cells that extend retrograde neurites that synapse to Purkinje cell dendrites within the molecular layer of the mature organ (9,10).This entire process is terminated by P21.It is thought that a subset of medulloblastomas originates from CGNPs that fail to exit the cycle and migrate, thus yielding tumor cells that mainly arise at the cerebellar periphery (11,12).
The Ink4c gene, which encodes a polypeptide inhibitor (p18 Ink4c ) of the cyclin-dependent kinases (Cdk), Cdk4 and Cdk6, is induced transiently in CGNPs before their exit from the cell cycle (13), but the protein persists throughout subsequent phases of cerebellar patterning. 5Although loss of Ink4c alone is insufficient to trigger tumorigenesis in the mouse cerebellum, its inactivation collaborates with p53 loss, or independently, with disruption of one Ptc1 allele, to generate medulloblastoma in mice (13,14).Mice lacking both Ink4c and p53 develop medulloblastoma with low penetrance (14), but the incidence increases to 75% within 6 months when mice are irradiated at P7 (13).Ptc1 heterozygous mice that lack either one or two Ink4c alleles develop medulloblastoma with f50% incidence, also within 6 months of birth (13).The incomplete frequency of tumor formation in both mouse medulloblastoma models and the accelerating effects of ionizing radiation in Ink4c À/À , p53 À/À mice imply that additional genetic events contribute to tumor formation.
To better understand the spectrum of chromosomal alterations associated with these two mouse models, we have further analyzed the emerging medulloblastomas by use of spectral karyotyping Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/).F. Zindy  (SKY), fluorescence in situ hybridization (FISH), and array comparative genomic hybridization (CGH).Recurrent genetic alterations were identified that were either common to both mouse models of medulloblastoma or specific to one or the other.To further test the importance of these genetic events in accelerating tumor development, we introduced particular genes into purified, primary CGNPs explanted from tumor-prone mice and injected the modified cells into immunocompromised recipient animals.Our results underscore functional interactions between a group of specific genes that have also been implicated in human medulloblastoma development.

Materials and Methods
Animal husbandry.Breeding and genotyping of mice were done as reported previously (13).Where indicated, mice were irradiated at P7 using a cesium source delivering 4 Gy.Mice were housed in an American Association of Laboratory Animal Care-accredited facility in accordance with NIH Guidelines.
Spectral karyotyping.SKY was done as recommended by the manufacturer using the Applied Spectral Imaging (ASI, Vista, CA) SkyPaint kit for mouse chromosomes.Images were acquired with a fluorescence microscope equipped with an interferometer (Spectra Cube TM, ASI) and custom designed filter cube (Chroma Technologies, Rockingham, VT) and analyzed using SKY View version 2.1 software (ASI).
FISH analysis.Purified DNA from bacterial artificial chromosome (BAC) clones was labeled by nick translation using either digoxigenin dUTP or biotin dUTP.Labeled DNA was mixed with sheared genomic DNA and resuspended in buffer [50% formamide, 10% dextran sulfate, 2Â SSC (1Â SSC is 0.15 mol/L NaCl, 0.015 mol/L sodium citrate)].The probe was denatured in water at 70jC for 5 min.Slides were immersed in 70% formamide and 2Â SSC at 70jC for 2 min.After overnight hybridization at 37jC, unbound probe was removed in 50% formamide and 2Â SSC at 37jC for 4 min.Signals were detected with anti-digoxigenin FITC or Texas red avidin.Slides were counterstained using 4 ¶,6-diamidino-2-phenylindole.
Array CGH.Genomic arrays containing 6,528 mouse BAC clones (Roswell Park Cancer Institute Microarray Facility) generated by ligationmediated PCR (15) were printed in triplicate on amino-silanated glass slides (Schott Nexterion type A+) using 10K Microspot pins and a MicroGrid II TAS arrayer (BioRobotics, Ann Harbor, MI).Undigested genomic DNA was labeled and hybridized according to basic Protocol 4 (16) except that hybridizations and washes were done manually using a BioArray labeling system (Enzo, Farmingdale, NY) and mouse Cot1 DNA (Invitrogen, Carlsbad, CA).Images acquired with a Genepix 4000B scanner were analyzed with GenePix 6.0 software (Axon Corp., Union City, CA).For spots having signal-to-noise ratio >2.5 in at least one channel, the ratio was calculated from the background-subtracted median signal of the two channels.To remove intensity-specific bias, ratios were normalized on log scale with a nonlinear algorithm by applying print tip Loess function (17) using the R language-based Bioconductor6 (release 1.9).Results of triplicate replicas were combined by taking the mean of log ratios.Any BAC that had less than two replicates passing quality control steps was excluded.BAC clone mapping information 7 (August 2005 mouse Build 35) was added to the resulting ratios.The final ratio represents the relative copy number of DNA from the experimental sample and the reference control sample.
CGNP purification and retroviral infection.Purification of CGNPs from mouse cerebella (18) and proliferation studies were done as described (13).Percoll-purified CGNPs were infected during the preplating stage with murine stem cell virus-based retroviruses, either as empty vectors carrying only GFP expressed from an internal ribosomal entry site or also carrying N-Myc or cyclin D1 cDNAs.Infected CGNPs were injected into mice (see below) without prior cell culture.
Orthotopic injections.Female CD-1 nu/nu mice (Charles River Laboratories, Wilmington, MA) ages 8 to 12 weeks were anesthetized and placed into a stereotactic apparatus equipped with a Z-axis (Kopf Instruments, Tujunga, CA).A portion of the scalp was removed, a window (approximately 10 Â 5 mm) was made in the skull using a dental drill, and the pia was excised.Infected CGNPs (2 Â 10 6 to 3 Â 10 6 cells) resuspended in Matrigel (5 AL; BD BioSciences, Bedford, MA) were implanted into the cerebrum at a depth of 0.5 to 2 mm using a 25 AL Hamilton syringe with an unbeveled 30-gauge needle.The wound was covered with a sterile glass window fixed in place using tissue adhesive.Implanted animals were transferred to the intravital imaging system (Nikon, Melville, NY) for baseline measurement and later reanesthetized for additional imaging with MetaMorph Imaging software (Universal Imaging, Downingtown, PA).Measurements were restricted to a fixed region within each image throughout the course of the study.
Affymetrix GeneChip analysis.CGNPs were isolated from Percoll density gradients.Tumors were collected, flash frozen, and stored at À80jC.Total RNA was extracted using Trizol (Invitrogen) according to the manufacturer's instructions.RNA was subjected to microarray hybridization (19) to the GeneChip Mouse Expression Set (MOE) 430A and 430B (Affymetrix, Santa Clara, CA), which contains 45,000 probe sets representing >39,000 transcripts, including >34,000 mouse genes on two arrays.Signals were detected using Affymetrix scanners, and the expression value for each was calculated using Affymetrix MAS5 software.Affymetrix CEL files were normalized with the gcRMA model (Bioconductor 1.8), and only probes that were present in at least one of the sample groups were used.To identify genes that were differentially expressed among CGNPs and medulloblastomas, supervised analyses were carried out (20).ANOVA was applied to, and the nominal P value was corrected for, multiple tests with controls for the family-wise error rate.Genes were considered to be differentially expressed if the adjusted P value was <0.01.
Quantitative real-time PCR.Quantitative real-time PCR (Q-PCR) on RNA extracted from purified CGNP-like tumor cells was done as described (13), and data were analyzed with SDS version 2.0 software (ABI, Branchburg, NJ) normalized to the internal 18S rRNA level (19).

Results
Ink4c inactivation promotes anaplastic medulloblastoma.Because p53-null mice develop a spectrum of tumors that greatly shorten their life span, we generated heterozygous mice containing one ''floxed'' (FL) and one null p53 allele (p53 FL/À ) on a transgenic background, in which Cre recombinase is expressed under the control of the Nestin promoter.This procedure generated animals specifically lacking p53 in portions of the central nervous system, including CGNPs within the EGL.
Exposure of a small cohort of such animals to 4 Gy ionizing radiation at P5-P7 yielded medulloblastomas in two of five treated Ink4c +/+ mice, whereas four of four developed medulloblastoma on an Ink4c-null background (13).Because the initial numbers of mice in these groups were small, we generated larger cohorts in an attempt to more precisely determine the effects of Ink4c loss on medulloblastoma formation.In addition, tumors arising in these animals were further analyzed for specific chromosomal aberrations associated with tumor formation.Mice were irradiated at P7 and sacrificed when moribund (Fig. 1).Those that did not show signs of disease were sacrificed and necropsied at 6 months of age.Although brain tumors were induced with similar time of onset regardless of Ink4c genotype, mice lacking Ink4c alleles succumbed more readily to medulloblastoma (P values in Fig. 1).Histologic analysis revealed that the frequency of the anaplastic form of medulloblastoma was increased in animals lacking one or both Ink4c alleles (Supplementary Table S1).Thus, in this medulloblastoma model, Ink4c modulates disease pathogenesis but not tumor incidence.
Genetic alterations in medulloblastomas arising in irradiated p53 FL/À , Nestin-Cre + mice lacking Ink4c alleles.Karyotypic analysis of nine such tumors revealed that the most common event in all medulloblastomas was loss of chromosome 13, either involving the elimination of an entire chromosome, loss of a fragment, or nonreciprocal translocations of a portion of chromosome 13 with other chromosomes, including chromosomes 6, 11, 12, and 9 (Table 1).In several cases, these events were confirmed by CGH analyses (examples in Supplementary Fig. 1A  and B).Because Ptc1 resides on chromosome 13, and its loss independently predisposes to medulloblastoma development (21), which is greatly accelerated on a p53-null background (22), further studies were undertaken to determine the status of this gene in these and additional medulloblastoma cases.
Twenty-three medulloblastomas arising in irradiated p53 FL/À , Nestin-Cre + mice of various Ink4c genotypes were analyzed by FISH using a BAC clone containing the Ptc1 gene (Table 2).Eight tumors exhibited complete loss of Ptc1, 13 retained only a single copy, and 2 showed no evidence of Ptc1 deletion.CGNP-like tumor cells from seven medulloblastomas were purified on two-step Percoll gradients, and RNAs extracted from them were examined using Q-PCR with primers that detect Ptc1, Ink4c, and Math1 (the latter being highly enriched in CGNPs; ref. 13).We detected no Ptc1 transcripts in five of five representative tumors that had lost only one copy of chromosome 13 (Table 2).Similarly, no Ptc1 signal was generated from one of two tumors (no.50910 in Table 2) that retained both copies of chromosome 13.In marked contrast, Ink4c mRNA and protein were expressed in six of six tumors that retained one Ink4c allele, confirming that Ink4c, unlike Ptc1, is haploinsufficient for tumor suppression (13,23).Thus, tumors initiated in irradiated p53-null progenitor cells that lack one or two Ink4c alleles routinely inactivate Ptc1 in the course of tumor development.
The next most frequent genetic alteration detected in these medulloblastomas involved chromosome 6 (Table 1).Of nine medulloblastomas analyzed, trisomy 6 was found in three (Supplementary Fig. 1C), ins(6;2) in one of nine, and a nonreciprocal translocation with chromosome 2 in one of nine medulloblastomas.Most tumors acquired a gain of chromosome 6 or part of it, although we found one (tumor 47966) that lost an intact chromosome 6 but instead exhibited a derivative containing portions of this chromosome (Tables 1 and 2).Aberrations involving chromosome 6 were also observed in tumors independently arising in mice lacking Ptc1 and Ink4c but retaining wild-type (WT) p53 (see below), suggesting that such anomalies are characteristic of medulloblastomas arising in cells containing different predisposing genetic lesions.
In these nine tumors, we also observed rearrangements or gain of chromosome 12 in four, loss of chromosome 19 in three (Supplementary Fig. 1D), and loss of chromosome 7 in two (Table 1).Other chromosomal alterations were not consistently found.Although Pten maps to chromosome 19, we observed no increase in phospho-Akt in tumors lacking this chromosome (data not shown).SKY analysis revealed an amplification of a portion of chromosome 12 in the form of double minutes in two of these tumors, as confirmed by CGH analysis (Table 1; Supplementary Fig. 1E).FISH and CGH confirmed that N-Myc was amplified in these tumors (Supplementary Fig. 1E).Array CGH also revealed an amplicon on chromosome 7 that encodes cyclin D1, and FISH detected increased cyclin D1 copy numbers in 9 of 10 medulloblastomas so analyzed (tumor numbers indicated in footnote in Table 2; Supplementary Fig. 1F).

Cancer Research
Cancer Res 2007; 67: (6).March 15, 2007 observed in Cdk4 and Cdk2 (Fig. 2B and C).Based on these data, we designed additional experiments to explore the potential roles of cyclin D1 and N-Myc overexpression in the context of medulloblastoma development (see below).
Trisomy 6 also occurs in medulloblastomas arising in Ptc1 +/À mice lacking Ink4c.Ink4c loss also increases the incidence of medulloblastoma development in Ptc1 +/À mice, but in this setting, p53 function is retained (13).Thus, Ink4c and p53 can collaborate independently with Ptc1 to initiate disease.In contrast to the complex genotypic changes observed in medulloblastomas lacking p53 and Ink4c (Table 1), tumors arising in Ptc1 +/À mice, whether lacking Ink4c alleles or not, did not display complex karyotypic alterations (Table 3).Most such tumors were diploid and contained 40 to 41 chromosomes, implying that, as expected, retention of p53 helps to preserve genomic stability.SKY or CGH analysis revealed that none of these tumors lost chromosome 13, and FISH analysis did not detect deletions of the remaining WT Ptc1 allele in six medulloblastomas analyzed in this manner.Nonetheless, Q-PCR of purified CGNP-like tumor cells showed that Ptc1 RNA was not detectably expressed in any of these tumors, whereas Ink4c expression was retained in those medulloblastomas that arose in Ink4c +/À mice, again arguing that, in this medulloblastoma model as well, there is a strong selection for complete inactivation of Ptc1 function but not for Ink4c (13).
Distinctive clinicopathologic and molecular features of the two medulloblastoma models.Medulloblastomas initiated in irradiated p53-null progenitors tended to be much more invasive than those that retained p53 function.Tumors arising in Ink4c +/À , p53 FL/À , Nestin-Cre + mice were found in the molecular layer but invaded inward toward the IGL and the brain stem (Supplementary Fig. 2A and B), whereas those arising in Ptc1 +/À , Ink4c À/À (or Ink4c +/À ) mice were generally restricted to the periphery of the cerebellum (Supplementary Fig. 2C and D).Consequently, mice with the latter medulloblastomas followed a more indolent clinical course, such that when they developed overt signs of disease, they frequently had massive tumors but ones that were more confined to the cerebellum.
With the exception of tumors, in which N-Myc was amplified on double-minute chromosomes (Fig. 2, tumors 47966 and 50863), comparison of expression of several targets of the Shh signaling pathway by immunoblotting showed that the levels of expression of the N-Myc, cyclin D2, Cdk4, and Cdk2 proteins were generally similar in both classes of medulloblastoma (compare Fig. 2D and  B).This fits with the idea that Ptc1 was inactivated regardless of the initiating genetic lesions.In contrast, cyclin D1 and Cdk6 were more highly expressed in tumors arising in Ptc1 +/À mice that retained p53 (Fig. 2D) when compared with their levels of expression in normal P5 cerebellum (Fig. 2A, left lane) or in tumors arising in p53-null cells (Fig. 2B and C).However, FISH and CGH failed to reveal additional copies of cyclin D1 in the tumors arising in Ptc1 +/À mice, implying that increases in gene copy number did not account for the significant increase in cyclin D1 expression in this setting.
Generation of tumors from primary CGNPs.Because amplification of N-Myc and increased Ccnd1 copy number occurred in a subset of tumors that arose in irradiated p53 FL/À , Nestin-Cre + mice (Tables 1 and 2), we tested whether enforced expression of N-Myc or cyclin D1 in primary CGNPs purified from Ink4c À/À , p53 À/À tumor-prone mice could induce medulloblastoma formation.Purified CGNPs from P6 mice lacking both of these genes were infected with high titer retroviruses expressing GFP, without or together with N-Myc or Ccnd1/cyclin D1.CGNPs infected with a control vector expressing GFP alone failed to proliferate in culture in the absence of Shh and did not incorporate bromodeoxyuridine.In contrast, infection of Ink4c, p53 doubly deficient CGNPs with retroviruses encoding N-Myc or cyclin D1 stimulated them to divide under the same culture conditions (data not shown).
We then injected 2 Â 10 6 infected CGNPs into the cerebral cortex of recipient immunocompromised mice.A glass window was surgically placed over the injected area, thereby enabling us to monitor the proliferation of GFP-marked cells and to follow any tumor development in live animals in real time by use of intravital fluorescence microscopy.Medulloblastomas did not develop when bFISH analysis revealed increased copies (2-3 or 2-4) of Ccnd1 per cell.
x FISH analysis revealed amplification of N-Myc.k FISH analysis revealed normal copies (2) of Ccnd1 per cell.

Cancer Research
Cancer Res 2007; 67: (6).March 15, 2007 Ink4c/p53 doubly null cells infected with a control GFP vector were injected into the brain (none of three mice).However, cells infected with vectors encoding either N-Myc ( four of six mice) or cyclin D1 (two of five mice) generated medulloblastomas.Foci of green fluorescent cells were detected as early as 1 week after injection, and tumors necessitating sacrifice of the animals arose from 8 to 32 weeks after injection (Fig. 3A).Excised tumors analyzed histopathologically had typical characteristics of medulloblastomas, as indicated by the size and morphology of tumor cells as well as by expression of neuronal and glial markers (Fig. 3B).Enforced cyclin D1 and GFP expression was confirmed by immunostaining of tumor sections with their respective antibodies (Fig. 3B).Antibodies to N-Myc were unable to detect protein expression by immunohistochemistry, and thus, in these tumors, only GFP expression was monitored.
Using Affymetrix GeneChip analysis, we compared the gene expression profiles of the orthotopically arising medulloblastomas with those arising spontaneously in Ink4c À/À , p53 À/À mice (Fig. 3Ca, compare lanes 1 and 2 versus lanes 3-5, respectively), as well as with expression profiles generated with CGNPs purified from WT P6 pups (Fig. 3C, lanes 6-8) and P5 whole cerebellum (Fig. 3C, a lanes 9-11).All medulloblastomas were found to share similar patterns of gene expression with profiles closer to those of purified CGNPs from P5 mice than from whole P5 cerebellum.Chromosomal cytoband analysis revealed complete loss of expression of all the genes encoded by chromosome 13 represented on the array (data not shown) confirming SKY and CGH analysis of the spontaneous tumors.Together, these experiments affirm that CGNPs are target cells for medulloblastoma formation and show that N-Myc and cyclin D1 overexpression, two common events seen in human medulloblastomas, collaborate with mutations in Ptc1, Ink4c, and, in some cases, p53 to induce medulloblastomas in mice.

Discussion
Role of p53 in mouse medulloblastoma.The cerebellum is particularly sensitive to DNA damage during postnatal organogenesis.This has been amply shown in different mouse models of medulloblastoma, in which genes that regulate DNA damage signaling and repair (e.g., Ligase4, Parp-1, and Xrcc4) were deleted together with p53 (24-26).Irradiation of p53-null mice at P7 induces medulloblastomas whose aggressive nature is augmented by concomitant inactivation of Ink4c.Karyotypic and SKY analyses of these tumors revealed complex chromosomal rearrangements, including losses and gains, as well as translocations, deletions, insertions, and duplications.In contrast, medulloblastomas arising in Ptc1 +/À mice lacking one or two Ink4c alleles retain p53 function (13) and harbor fewer chromosomal rearrangements, implying that many of the genetic anomalies observed in the irradiated p53-null animals are a consequence of genomic instability but are unlikely to initiate medulloblastoma formation (27).
Medulloblastomas lacking p53 and Ink4c were more anaplastic and invasive than those initiated in mice lacking one copy of Ptc1 and Ink4c.We noted that tumors occurring in Ink4c, p53-double Figure 3. Enforced expression of N-Myc and cyclin D1 in P5 CGNPs from p53 À/À , Ink4c À/À mice induces medulloblastoma after orthotopic transplantation into the cerebral cortex.A, in vivo fluorescence imaging of tumor development from infected CGNPs implanted into the brains of immunocompromised CD-1 nu/nu recipient mice.Purified CGNPs from Ink4c À/À , p53 À/À mice were infected with retroviral vectors expressing cyclin D1 or N-Myc together with GFP or with a control virus expressing GFP alone (top ).B, histopathology of brain tumors arising after injection of Ink4c, p53 doubly deficient CGNPs overexpressing N-Myc (a-f ) compared with that in tumors arising in irradiated Ink4c À/À , p53 FL/À , Nestin-Cre + mice (g-k ).a, H&E staining of tumor tissue developing from cells injected into the cerebral cortex versus that of a cerebellar tumor arising in an irradiated Ink4c À/À , p53 FL/À , Nestin-Cre + mouse (g).b and h, H&E staining (magnification, Â400) showing the typical morphology of medulloblastoma cells.c and i, tumor cells stained strongly for class III h-tubulin (TUJ1; magnification, Â400) and showed diffuse expression of synaptophysin (magnification, Â400) (d and j ), both of which are neuronal markers.e and k, like human medulloblastomas, many mouse tumors showed glial differentiation as evidenced by sporadic GFAP expression (magnification, Â400).f, tumor cells stained positively for GFP (magnification, Â400) as opposed to adjacent normal tissue that did not.C, microarray analysis of medulloblastomas arising orthotopically (lanes 1 and 2) and spontaneously (lanes 3-5 ) and of CGNPs from WT P6 pups (lanes 6-8 ) and P5 total cerebellum (lanes 9-11 ).Microarray data were subjected to ANOVA and genes that changed >1.9-fold between the indicated groups (with adjusted P values <0.01 and maximum absolute intensity difference >32 units) were considered differentially expressed (a). Colors represent relative expression level.Red, high expression; green, low expression (see gradient).Genes were clustered based on expression pattern among the four groups.b, an expanded view of the boxed area in (a ) that contains genes overexpressed in medulloblastoma samples.The individual gene names (right ) correspond to the denoted rows in the cluster.
null or in irradiated p53 FL/À , Nestin-Cre + mice were located within the EGL but invaded inward into the IGL and toward the brain stem.In contrast, medulloblastomas arising in Ptc1 +/À , Ink4c-null mice arose in the cerebellar periphery and did not invade the IGL.Because the latter tumors were more restricted in position and less anaplastic, they tended to become very large before clinical signs of disease were overt.Therefore, we suspect that the relatively increased genetic instability of p53-null medulloblastomas, although not strictly responsible for tumor initiation, likely contributes to tumor anaplasia and to increased invasive properties.In addition, although infrequent, TP53 mutations in human medulloblastomas are associated with large cell anaplastic features (8) that are recapitulated in the Ink4c À/À , p53 À/À mouse medulloblastoma model.
Ptc1 inactivation.Although medulloblastomas initiated in the Ink4c À/À , p53 À/À and Ink4c À/À , Ptc1 +/À backgrounds exhibited karyotypes with different complexities, they shared specific genetic alterations.Remarkably, the most common event seen in spontaneously arising medulloblastomas lacking p53 with or without inactivation of Ink4c was the loss of one chromosome 13 or its translocation with other chromosomes, leading to hemizygous or homozygous loss of the Ptc1 gene.Complete loss of Ptc1 expression was evident in all medulloblastomas examined regardless of whether chromosome 13 was retained.Orthotopically derived tumors lacking p53 and Ink4c also loose chromosome 13 and Ptc1 function and similarly yield expression profiles that show activation of the Shh signaling pathway.Frequent loss of one chromosome 13 was also observed in two other mouse models that stemmed from conditional loss of p53 together with Rb or Xrcc4, respectively (26,27).These results support the idea that, at least in mice, loss of genomic stability, together with loss of normal DNA repair or of a functional retinoblastoma pathway, are themselves insufficient to induce medulloblastomas and require Ptc1 inactivation.
Although genetic analyses of mouse medulloblastomas arising either spontaneously or ''engineered'' from primary CGNPs suggest that their development can be triggered by many of the same genetic alterations found in human medulloblastomas, the uniform involvement of Ptc1 was unexpected given that only f25% human medulloblastomas show activation of the Shh signaling pathway (3,6).A survey of mutations in a cohort of 46 human medulloblastomas revealed mutually exclusive activation of the Shh/Ptc versus the Wnt/h-catenin signaling pathways with an overall incidence of f20% for each (6).This underscores the present inadequacy of mouse models in mimicking certain forms of human medulloblastoma and raises the issue of whether it will be possible to derive mouse models of human medulloblastoma, in which Ptc1 function is preserved.
Other recurrent genetic lesions.The next most common genetic event observed in medulloblastomas arising in both of our mouse models involved chromosome 6, as manifested by the gain of a whole chromosome leading to trisomy 6 or by translocations with other chromosomes.Trisomy 6 is also a frequent event in mice, in which Xrcc4 and p53 are targeted to the central nervous system (CNS; ref. 26).The genes encoding cyclin D2, cyclin E, and Fox M1 are located on this chromosome and are attractive potential candidates in contributing to tumor formation.However, the cyclin D2 gene is a direct target of the Shh/Ptc signaling pathway and is expressed at relatively high levels in the Shh-driven development of the postnatal cerebellum (28) and was therefore expected to be highly expressed in tumors, in which the pathway is constitutively activated by Ptc1 loss.Indeed, immunoblotting revealed robust expression of this protein in all tumors from Ink4ccompromised Ptc1 +/À or p53 FL/À , Nestin-Cre + mice whether they exhibited a gain of chromosome 6 or not.FISH analysis of several translocations failed to reveal common breakpoints on chromosome 6 providing no clue to the identity of the target gene(s).
Other chromosomal gains or losses were less frequent.Interestingly, three of nine p53 FL/À , Nestin-Cre + medulloblastomas exhibited monosomy 19.Loss of one chromosome 19 was also a frequent event ( four of eight tumors analyzed) in Rb FL/À , p53 FL/À , Nestin-Cre + medulloblastomas (27).Notably, two tumor suppressors, Pten and Sufu, map to mouse chromosome 19, but we saw no correlation between the loss of chromosome 19 and levels of phospho-Akt.Moreover, Pten mutations are exceedingly rare in human medulloblastomas.
In p53 FL/À , Nestin-Cre + medulloblastomas lacking Ink4c alleles, we found that several tumors exhibited three to four copies of Ccnd1, only two of which exhibited elevated cyclin D1 protein levels.Several of these tumors lacked one copy of chromosome 7 on which Ccnd1 is located, providing a possible explanation for variations in cyclin D1 expression.Strikingly, cyclin D1 protein levels were high in all medulloblastomas arising in Ptc1 +/À mice regardless of Ink4c status, but its high expression did not result from gene amplification, increased copy number, or chromosomal translocations.
We observed two medulloblastomas, in which N-Myc located on chromosome 12 was amplified on double-minute chromosomes and was highly expressed.N-Myc amplification has been observed in 5% of human medulloblastomas and is associated with the large cell/anaplastic subtype.However, overexpression of N-Myc unaccompanied by amplification occurs in 55% to 88% of such tumors (5,29,30).N-Myc is also a direct target of the Shh signaling pathway (31) and is required for proliferation of CGNPs during cerebellar development.Its conditional deletion early in CNS development leads to cerebellar microcephaly and to severe cerebellar developmental defects (32).N-Myc amplification was also a common event in medulloblastomas arising in other mouse models (26,27).Although cyclin D1 is regulated by N-Myc (28,33), we saw no correlation between high levels of N-Myc and those of cyclin D1.
Enforced expression of cyclin D1 and N-Myc in CGNPs from tumor-prone mice induces medulloblastomas.To test whether amplification or overexpression of cyclin D1 or N-Myc were important for tumor formation, we introduced them by retroviral gene transfer into purified CGNPs from genetically engineered mice lacking p53 and Ink4c and injected the modified CGNPs into the cortex of recipient mice under cranial windows.Not only did these genes provide CGNPs with a proliferative advantage in vitro, as shown also by others (33), but they also contributed to tumor formation in vivo.Marking the cells with GFP enabled us to follow cell proliferation and tumor formation in live animals in real time, thus providing a powerful method to test the contribution of candidate genes to medulloblastoma development.
Analysis of these orthotopically induced medulloblastomas by Affymetrix GeneChip Microarrays revealed that they shared similar gene expression profiles with tumors spontaneously arising in tumor prone mice.This again implies that many of the complex chromosomal translocations and alterations seen in tumors arising in irradiated p53 FL/À , Nestin-Cre + mice, with or without the loss of Ink4c alleles, has little bearing on tumor initiation.Inactivation of N-Myc or cyclin D1 not only impairs cerebellar development but also suppresses medulloblastoma formation in mouse models where the

Figure 2 .
Figure 2. Protein expression in medulloblastomas initiated in mice of different genotypes.Proteins were extracted from the cerebellum of WT mice at P5 and 1 mo (1M ) of age (A ), from medulloblastoma tissues (B and D ), or from CGNP-like tumor cells purified on Percoll gradients (C).The different initiating mutations predisposing to medulloblastoma development are noted at the top of (B-D ), with their Ink4c genotypes denoted below.Proteins (left) were detected by immunoblotting.All p53 FL/À , Nestin-Cre + mice (B and C ) were irradiated at P7. Tumor numbers discussed in the text are indicated below the lanes.Tumors 47966 and 50863 exhibited N-Myc amplification revealed by FISH and SKY.

Table 3 )
, similar to what was observed in medulloblastomas arising in mice lacking both Ink4c and p53

Table 2 .
FISH and SKY analysis of Ptc1 in medulloblastomas arising in irradiated p53 FL/À , Nestin-Cre + mice of different Ink4c genotypes *(no.indicated events/no.metaphases).cRNA isolated from purified tumor cells was analyzed by quantitative Q-PCR for Ptc1, Ink4c, and Math1 expression.