DEVELO

PMENT

3929DEVELOPMENT AND DISEASE RESEARCH ARTICLE

INTRODUCTIONDuring cerebellar development, granule neuron precursors (GNPs)migrate from the rhombic lip to the outer layer of the cerebellum toform the external granule layer (EGL), where they undergo apostnatal proliferative burst before exiting the cell cycle andmigrating inward to form the mature inner granule layer.Proliferation occurs in the outer EGL, and, as post-mitotic cellsbegin to differentiate, they move inwards in the EGL beforemigrating to their mature location in the inner granule layer (IGL).Alterations in these tightly regulated events can lead to cerebellardysfunction and ataxia or to medulloblastoma, the most prevalentmalignant brain tumor of childhood (Gilbertson, 2004).

The conserved hedgehog signaling pathway plays critical roles inregulating the proliferation and differentiation of cerebellar GNPs(Wechsler-Reya and Scott, 2001). In vitro studies have shown thatsonic hedgehog (Shh), which is secreted by the Purkinje neurons ofthe cerebellum, can sustain GNP proliferation and maintain theirundifferentiated state (Wechsler-Reya and Scott, 1999). Shh acts bybinding to its receptor patched (Ptch1) on target cells. In the absenceof Shh, Ptch1 inhibits the signaling activity of smoothened (Smo),leading to the production of repressor forms of the Gli family oftranscription factors. Binding of Shh to Ptch1 blocks the inhibitoryactivity of Ptch1 on Smo, resulting in the production of activatorforms of the Gli transcription factors and increased expression ofdownstream genes (Stecca and Ruiz i Altaba, 2005).

Deregulation of the hedgehog pathway has been linked to thedevelopment of a number of cancers, including medulloblastoma(Goodrich et al., 1997; Hahn et al., 1996). Mice that areheterozygous for Ptch1 exhibit spontaneous development ofmedulloblastoma (Goodrich et al., 1997). These tumors exhibit

increased Shh signaling (Goodrich et al., 1997), and treatment ofmice with medulloblastomas with inhibitors of Shh signaling issufficient to cause tumor regression (Berman et al., 2002; Romer etal., 2004). Interestingly, recent studies revealed the presence of pre-neoplastic lesions in young adult Ptch1+/– cerebella. These pre-neoplastic lesions exhibit an expression profile intermediate betweenthose of normal proliferating GNPs and medulloblastomas (Oliveret al., 2005), suggesting that medulloblastomas in Ptch1+/– miceprobably develop from these lesions. In addition, recent studies ofhuman medulloblastomas have showed that only tumor cells positivefor the stem cell marker CD133 are capable of forming tumors inmouse xenografts (Singh et al., 2004), and these tumor initiatingcells exhibit other characteristics of neural stem cells, such as self-renewal, proliferation and the capacity to differentiate (Singh et al.,2003). These observations support a tumor stem cell model for thedevelopment of medulloblastomas.

Cyclins D1 and D2, but not D3, are upregulated in response toShh signaling in cerebellar cells (Kenney and Rowitch, 2000). Bothcyclins D1 and D2 are expressed in proliferating GNPs, and theirjoint removal results in an exceptionally hypoplastic cerebellum(Ciemerych et al., 2002), indicating that cyclins D1 and D2 playcrucial roles in the rapid postnatal expansion of the GNP population.Interestingly, although removal of either cyclin D1 or D2 alone doesnot affect in vitro proliferation of postnatal day 4 GNPs in responseto Shh signaling (Kenney and Rowitch, 2000), removal of cyclin D2alone is sufficient to cause mild cerebellar hypoplasia (Huard et al.,1999), owing to both decreased proliferation and increased apoptosisof GNPs. This result is somewhat surprising as the D-type cyclinshave very similar biochemical functions and, in general, specificeffects of individual mutations are typically seen only in tissueswhere the other D-type cyclins are not expressed at high levels(Ciemerych et al., 2002).

The cerebellar phenotype of Ccnd2–/– mice could be due to arequirement for two D-type cyclins concurrently in the proliferationof GNPs, or it could be due to differences in the expression patternsof these two cyclins in the developing cerebellum. In this report, weshow that early postnatal GNPs expressed only cyclin D1 while laterGNPs expressed both cyclins D1 and D2. Consistent with this

Loss of cyclin D1 impairs cerebellar development andsuppresses medulloblastoma formationJennifer Pogoriler1, Kathleen Millen2, Manuel Utset3 and Wei Du1,*

Medulloblastoma, the most common malignant brain tumor of childhood, is believed to derive from immature granule neuronprecursors (GNPs) that normally proliferate in the external granule layer before exiting the cell cycle and migrating to their maturelocation in the inner granule layer. In this study, we examined the expression of D type cyclins in GNPs during cerebellardevelopment and showed that GNPs in early development expressed only cyclin D1, whereas later GNPs expressed both cyclins D1and D2. Coinciding with the period of cyclin D1-only expression, Ccnd1–/– mice showed reduced proliferation of GNPs and impairedgrowth of the cerebellum. Interestingly, removal of cyclin D1 was sufficient to drastically reduce the incidence of medulloblastomain Ptch1+/– mice, despite the fact that these tumors showed upregulation of both cyclins D1 and D2. We showed that cyclin D1 hasan earlier role in tumorigenesis: in the absence of cyclin D1, the incidence and overall volume of ‘pre-neoplastic’ lesions weresignificantly decreased. We propose a model that links a role of cyclin D1 in normal GNP proliferation with its early role intumorigenesis.

KEY WORDS: Cyclin D1, Cerebellar development, Medulloblastoma, Ptch1, Mouse

Development 133, 3929-3937 (2006) doi:10.1242/dev.02556

1Ben May Institute for Cancer Research, The University of Chicago, 924 E. 57thStreet, Chicago, IL 60637, USA. 2Department of Human Genetics, The University ofChicago, 924 E. 57th Street, Chicago, IL 60637, USA. 3Department of Pathology,The University of Chicago, 924 E. 57th Street, Chicago, IL 60637, USA.

*Author for correspondence (e-mail: wdu@huggins.bsd.uchicago.edu)

Accepted 31 July 2006

DEVELO

PMENT

3930

expression data, Ccnd1–/– mice had decreased early GNPproliferation and early ataxia as a consequence of a delay inacquiring normal cerebellar function. Interestingly, removal ofcyclin D1 resulted in a significant decrease in the incidence ofmedulloblastoma in Ptch1+/– mice, as well as a decrease in theincidence and size of pre-neoplastic lesions. We suggest that cyclinD1 plays a crucial role in the proliferation of early GNPs and in theprogression of the pre-neoplastic lesions to medulloblastomas.

MATERIALS AND METHODSMiceCcnd1+/– mice were bred to obtain Ccnd1+/+ and Ccnd1–/– mice (Sicinski etal., 1995). For the righting reflex, five litters were tested daily beginning onP6. Mice were turned on their backs and observed to determine whether theycould immediately right themselves. Ptch1+/–;Ccnd1+/– mice were bred toobtain Ptch1+/– (Goodrich et al., 1997) and Ptch1+/–;Ccnd1–/– mice. All micewere maintained on a mixed C57Bl/6 � 129Sv background. Mice used forthe tumor study represent six generations beginning with the F2 generation.Mice were observed daily, and mice older than 35 weeks or demonstratingabnormal neurological symptoms were euthanized and brains were dissectedunder a dissecting microscope to examine for tumors. All mice were housedin a facility accredited by the Association for Assessment and Accreditationof Laboratory Animal Care and maintained in accordance with NIHGuidelines. The Animal Care and Use Committee at The University ofChicago approved all procedures.

ImmunohistochemistryAntibodies used: BrdU (BD Biosciences B44 1:800), cyclin D1 (Santa Cruzsc-450 1:400), cyclin D2 (Lab Vision Ab-4 1:500), NeuN (ChemiconMAB377 1:40,000), calbindin (Swant 1:6000) and phosphohistone H3(Upstate 1:16,000). Samples were fixed, dehydrated and embedded inparaffin. Sections (6 �m) were rehydrated, and endogenous peroxidaseswere blocked by incubation in 3% H2O2, 10% methanol for 30 minutes.Antigen retrieval for all antibodies was performed by boiling in 0.01 M citricacid (pH 6) for 10 minutes in the microwave. The primary antibody wasdetected using the Vector Elite ABC kit (Vector Laboratories). Slides werelightly counterstained with Hematoxylin and Eosin. Sections incubated

without primary antibody served as a negative control. For comparison ofstaining patterns within a single cerebellum, serial sections were examined.At least three pups from different litters were analyzed.

BrdU incorporation and statistic analysisTo monitor S-phase, mice were injected with BrdU (Sigma) (100 mg/kgbody weight) 1 hour prior to euthanasia. For statistical analysis of tumorincidence, the Chi-squared test was used. For comparison of proliferationand number of pre-neoplastic lesions, the Student’s t-test was used.

All developmental sections examined for size, staining patterns andproliferation were midline sagittal sections of Ccnd1+/+ and Ccnd1–/–

littermates. The midline was identified by the absence of cerebellarpeduncles and fastigial nucleus. Multiple sections have been examined butonly a single midline section was used to calculate the area. The area wasdetermined by taking pictures with a Zeiss digital camera with a knownpixel/�m ratio. The area in pixels was determined in Adobe Photoshop andconverted to �m2. For quantification of proliferation, the number of BrdUand H3-P positive EGL cells per mm of EGL was counted. BrdU-positivecells were counted on midline sections in the same location in the anteriorat P0 and at P6. As there were many fewer H3-P positive cells at P0, the H3-P positive cells over the entire length of the midline sections were countedto minimize variations. Statistical analysis used the paired two-tailedStudent’s t-test.

RESULTSCyclin D1 and cyclin D2 exhibit distinct expressionpattern in developing cerebellar GNPsTo determine to what extent cyclin D1 and D2 expression overlap inthe developing EGL, we examined their localization during lateembryonic and postnatal development when GNPs are proliferating.The earliest time point we examined was E16.5. At this point, cyclinD1 expression was primarily confined to the anteriormost 50% ofthe EGL (Fig. 1A,C) with only a few cells expressing cyclin D1 inthe posterior (Fig. 1D). By P0, high cyclin D1 expression extendedfurther posteriorly, with strong expression throughout the anterior75% of the EGL (Fig. 1E-H), and almost all nuclei in the anteriorouter EGL were positive for cyclin D1 expression (Fig. 1G). In

RESEARCH ARTICLE Development 133 (19)

Fig. 1. Cyclin D1 (brown) is expressed at high levelsin the EGL during early cerebellar development.(A-D) The E16.5 cerebellum (A, with high-power viewsof boxed areas shown in B-D) expresses cyclin D1strongly in the ventricular region (B) and anterior EGL(C), with fewer positive cells in the posterior EGL (D).(E-H) The P0 cerebellum (E, with high-power views ofboxed areas shown in F-H) shows some remainingventricular expression (F), and strong expression in theEGL extending much further posteriorly with enlargedanterior (G) and posterior (H) views of the EGL, althoughthe extreme posterior still has fewer positive cells (H).(I,J) P6 cerebellum (I, with high-power of view of boxedarea in J) shows strong staining in the outer EGLthroughout the entire AP axis. (K,L) At P12, onlybackground staining of blood vessels and meninges isseen (arrows) (K, with high-power of view of boxed areain L). Brackets indicate the width of the EGL. Nuclei arelightly stained purple with Hematoxylin and Eosin. Scalebars: 50 �m.

DEVELO

PMENT

addition, cyclin D1 also stained some cells deeper within thecerebellum and in the posterior at both E16.5 and P0 (Fig. 1D,H).By P6, cyclin D1 expression was detected throughout the entirelength of the EGL and was limited to the outer half of the EGL (Fig.1I,J). Scattered cyclin D1-positive cells were also found within theinner granule layer at this time, particularly directly adjacent toPurkinje cells (data not shown). Based on their location, these arelikely to be glial cells, but we did not confirm their identity. By P12,cyclin D1 was almost entirely absent from the remaining EGL ofwild-type cerebellum (Fig. 1K,L). With the exception of backgroundendothelial and erythrocyte staining (also seen in the absence ofprimary antibody), this antibody showed no staining in Ccnd1–/–

samples, confirming its specificity.The EGL expression of cyclin D2 was in striking contrast to that

of D1 at E16.5 and P0, even though similar strong nuclear stainingof both cyclins D1 and D2 were detected in the ventricular region ofthe midbrain (Fig. 1A,B,E,F; Fig. 2A,B,E,F) and in some cellsdeeper within the inner cerebellum (Fig. 1C,G; Fig. 2C,Garrowheads). As shown in Fig. 2C, very few cells with nuclear cyclinD2 staining were observed in the EGL at E16.5 (Fig. 2C,D, arrow),consistent with reports of absent cyclin D2 mRNA in these cellsduring embryonic proliferation (Ross et al., 1996). Similarly, veryfew cyclin D2 positive nuclei were detected in the EGL at P0 (Fig.2G,H). At P6, cyclin D2 was detected throughout the outer zone ofthe EGL in an area that was broader than that of cyclin D1 (Fig. 2I,J),although staining was not uniform: a subpopulation of cells,generally towards the middle of the EGL, expressed much strongercyclin D2 levels. Furthermore, unlike cyclin D1, nuclear cyclin D2staining was still present in some of the EGL cells by P12 (Fig.2K,L, arrows). These immunohistochemistry results were furthersupported by western blot analysis of total cerebellar extracts, whichshowed very low levels of cyclin D2 at P0 and much higher levels atP6, while high levels of cyclin D1 were present at both time points(see Fig. S1 in the supplementary material).

To determine whether cyclin D2 was upregulated to compensatefor the loss of cyclin D1, we evaluated the levels of cyclin D2 at P0in Ccnd1–/– cerebella. Comparison of Ccnd1–/– and wild-typesiblings showed no difference in the cyclin D2 staining pattern (seeFig. S2 in the supplementary material). As Ccnd1–/–;Ccnd2–/– miceshow severe cerebellar hypoplasia with only slight upregulation ofcyclin D3 (Ciemerych et al., 2002), it was unlikely that cyclin D3would be significantly upregulated in Ccnd1–/– cerebella. Consistentwith this, we were unable to detect cyclin D3 in either wild-type orCcnd1–/– GNPs (see Fig. S2 in the supplementary material).Therefore, the absence of cyclin D1 does not lead to an upregulationof cyclins D2 or D3 in the EGL.

Ccnd1–/– mouse cerebellar phenotypeGiven the absence of high levels of D-type cyclins in the EGL duringearly cerebellar development in Ccnd1–/– mice, we wonderedwhether they might have a cerebellar phenotype. Ccnd1–/– miceexhibit several symptoms such as early lethality, decreased size anda hindleg clasping reflex that have long been hypothesized to reflecta neurological defect (Sicinski et al., 1995), but no specifichistological deficiency has been reported. To test if Ccnd1–/– micemight have a mild ataxia related to their cerebellar development, weexamined the righting reflex of Ccnd1–/– pups (Aruga et al., 1998;Crawley, 1999). Wild-type pups could right themselves by P8, butCcnd1–/– pups were unable to turn over until at least P14, and insome pups the reflex was delayed until as late as P18. After thispoint, the Ccnd1–/– adults displayed no obvious ataxia, suggestingthat the cerebellar functional defect had been corrected orcompensated for.

Examination of the Ccnd1–/– cerebellum indicated no grossabnormalities in foliation or organization of cellular layers at any age(data not shown). However, cerebellar cross sections did appearsmaller than normal, so the sizes of Ccnd1–/– cerebella werecompared with those of their littermates. At early time points the

3931RESEARCH ARTICLECyclin D1 in medulloblastoma development

Fig. 2. Cyclin D2 (brown) is not expressed at highlevels in the EGL until later in cerebellardevelopment. (A-H) The E16.5 (A, with high-powerviews of boxed areas shown in B-D) and P0 (E, withhigh-power views of boxed areas shown in F-H) EGLshow little nuclear cyclin D2 staining. Enlarged viewsof the ventricular region (B,F), anterior EGL (C,G) andposterior EGL (D,H) are shown. Only a few nuclei inthe posterior EGL (arrow in D) are weakly positive, incomparison with strong staining of the ventricularzone of the midbrain and of cells deeper within thecerebellum (arrowheads in C,G). (I,J) Strong cyclin D2expression is seen in the outer half to two-thirds ofthe EGL at P6 (I, with high-power of view of boxedarea in J). (K,L) Scattered EGL cells are positive(arrows) at P12 (K, with high-power of view of boxedarea in L). Brackets indicate the width of the EGL.Nuclei are lightly stained purple with Hematoxylin andEosin. Scale bars: 50 �m.

DEVELO

PMENT

3932

cerebellum was too small to be accurately dissected and weighed, sowe estimated the size by measuring the area of a midline histologicalsection. At P0 we found no significant difference in the cross-sectional area of the midline cerebellum (Fig. 3A,D); however, byP6 the cerebellum was reduced in size to 68% of that of wild-typesiblings (Fig. 3B,E,G). This relative size difference persisted but didnot increase from P6 to P12 (Fig. 3C,F,G). By contrast, the totalbrain weight of Ccnd1–/– brains (including the cerebellum) was84% and 80% of the Ccnd1+/+ littermate weight at P6 and P12,respectively. Therefore, the cerebellum was more severely affectedthan other brain regions. In addition, the difference in the cerebellarsizes differed in its timing and severity from that of the total bodyweight of Ccnd1–/– mice, which continued to fall behind that of theirlittermates from 72% at P6 to 53% at P12 (Fig. 3H).

Much of the normal postnatal size increase of the cerebellum isdue to an increase in the granule cell population, so the normal sizedcerebellum at birth suggested that the Ccnd1–/– cerebellar size defectwas probably due to the impaired postnatal proliferation of GNPs.In Ccnd1–/– cerebella, the thickness of the EGL was not altered, butthe entire surface area was decreased. As the nuclear density wassimilar to wild type, we measured the granule neuron population bydetermining the cross-sectional area of the EGL and IGL in midlinehistological sections. This confirmed a decrease in the EGL and IGLat P6 (69% and 61% respectively) and the IGL at P12 (63%) inCcnd1–/– cerebella relative to wild-type littermates (Fig. 3I).Therefore, the phenotypically crucial period for cyclin D1 incerebellar growth is between P0 and P6, corresponding to thedevelopmental phase when cyclin D1 is the only highly expressed Dtype cyclin in GNPs.

Removal of cyclin D1 affects early granule neuronprecursor proliferationThe size defect of Ccnd2–/– cerebella was reported to be due to bothdecreased entry into S phase and increased apoptosis of GNPs(Huard et al., 1999). To determine if increased apoptosis couldcontribute to the decreased granule neuron population in Ccnd1–/–

cerebella, we performed TUNEL assays on both P0 and P6cerebella. Both wild-type and Ccnd1–/– cerebella showed very lowlevels of apoptosis, and no significant difference was seen between

them (see Fig. S3 in the supplementary material). Therefore, thereis no evidence that increased apoptosis significantly contributes tothe smaller granule neuron population in Ccnd1–/– mice, althoughwe cannot absolutely exclude this possibility.

We then examined whether Ccnd1–/– GNPs had decreasedproliferation. As the D-type cyclins are known to be involved in thetransition through G1 (Sherr and Roberts, 2004), we expected thisphase of the cell cycle to be lengthened and the percentage of cells inS phase to be decreased. To identify cells in S-phase, pups wereinjected with BrdU 1 hour prior to sacrifice. Midline sections werelabeled with anti-BrdU or anti-phospho-histone H3 (H3-P) antibodiesto measure S-phase or M phase cells, respectively. BrdU incorporationand H3-P staining were observed throughout the entire anteroposterioraxis of the midline EGL at E16.5 and P0, except in the innermost oneto two cell layers (Fig. 4A,B,E,F; data not shown). By P6 when thethickness of the EGL had greatly increased, both markers were foundonly in the outer half of the EGL (Fig. 4C,D,G,H; data not shown),consistent with known proliferation and migration patterns.The pattern of BrdU incorporation was roughly similar betweenCcnd1–/– and wild-type littermates except that BrdU incorporationwas consistently decreased in Ccnd1–/– cerebella at P0 (Fig. 4A,B,I).Quantification of S phase cells in the same anterior region of theEGL at P0 revealed that Ccnd1–/– cerebella exhibit a significantdecrease in BrdU incorporation (Fig. 4I; P0

+/+=427 cells/mm,P0

–/–=363 cells/mm; P=0.028). Interestingly, this difference was notobserved at P6, when both cyclins D1 and D2 are expressed at highlevels (Fig. 4C,D,I). There was similarly a trend in decreased mitosisin the Ccnd1–/– EGL at P0 (Fig. 4E-F,J), although this was notstatistically significant (P=0.07) probably due to the small numberof mitotic cells as a consequence of the very short length of mitosis.This modest (about 15%) decrease in BrdU incorporation couldaccount for the decrease in the granule cell population over 6 daysof rapid proliferation. For example, if early GNPs proliferate at a rateof 15 hours/cycle (Yoshioka et al., 1985), there could be 6�24/15(i.e. 9.6) rounds of GNP proliferation between P0 and P6. If eachround of cell proliferation is decreased by 15%, this could lead to thenumber of cells at P6 being decreased to as little as (1–0.15)9.6�100(i.e. 21%) of the normal number. Therefore, a small decrease inproliferation over the course of multiple cell cycles can have a

RESEARCH ARTICLE Development 133 (19)

Fig. 3. Ccnd1–/– cerebella show a postnatal sizedefect. The midline cross-sections of wild-typelittermates (A-C) are shown next to Ccnd1–/– (cycD1on figure) pups (D-F). At P0 there is no difference incerebellar size (A,D), but by P6 Ccnd1–/– cerebella aremarkedly smaller (B,E). The relative size difference ismaintained but not increased by P12 (C,F).(G) Quantification of the relative cerebellar midlinearea. (H) Total body weight of Ccnd1–/– pupscontinued to fall further behind wild-type littermatesfrom P6 to P12. (I) The size of the granule cellpopulation was similarly decreased at P6 and P12.Scale bars: 500 �m. The number of littermatesanalyzed at each time point is indicated in each chart.

DEVELO

PMENT

profound effect on the cerebellar size. Taken together, these resultsindicate that at P0, when cyclin D1 is the only highly expressed D-type cyclin, GNPs are impaired in their ability to progress into thecell cycle in the absence of cyclin D1. The observed defect in theG1/S transition of Ccnd1–/– GNPs corresponds precisely to theappearance of the cerebellar size defect seen between P0 and P6.

Delayed Purkinje cell development in the Ccnd1–/–

cerebellumPurkinje cell progenitors complete proliferation by E14.5, so thenormal size of Ccnd1–/– P0 cerebella suggested that Purkinje cellnumbers were not drastically affected; however, other mice withgranule neuron defects show abnormal Purkinje cell maturation(Goldowitz and Hamre, 1998). To determine whether Purkinjecell dendrite development was altered in Ccnd1–/– cerebella, weexamined sections from pups at different developmental stages. AtP6, Purkinje cell dendrites have not extended very far, and there islittle difference between wild-type and Ccnd1–/– pups (data notshown). At P9, when dendrites have begun to extend but have notdeveloped the more complex branches characteristic of adultcerebella, staining with calbindin showed stunted dendrites inCcnd1–/– pups (Fig. 5A,B). We further examined the dendrites byconfocal microscopy and found that, at P9, Ccnd1–/– Purkinje cellshad simplified disorganized dendritic trees (data not shown).Interestingly, adult Ccnd1–/– Purkinje cells looked normal (Fig.5C,D); therefore, the timing of the Purkinje cell abnormalitycorresponds to the period of ataxia for Ccnd1–/– pups. As Purkinjecells did not express cyclin D1 at any of the postnatal time pointsexamined, it is unlikely the observed Purkinje cell defect representsa cell-autonomous event; more likely it represents a response toabnormalities in other cell types such as the GNPs.

Comparison of Ccnd1–/– pups with siblings at P6, P12 and P15showed similar decreases in EGL thickness (see Fig. S4 in thesupplementary material), suggesting that GNPs were notaccumulating in the EGL due to migration abnormalities.Therefore, we did not expect to see abnormalities in the radial gliaalong which they migrate. Consistent with this, staining with GFAPshowed a normal density and orientation for these cells (data notshown). We conclude that the granule neuron proliferation defect ismore likely to be due to a cell-intrinsic effect of cyclin D1 than toabnormal signals from the other cells such as the Purkinje cells orglial cells.

Decreased medulloblastoma incidence inPtch1+/–;Ccnd1–/– micePtch1+/– mice, in which the Ptch1 allele has been replaced withthe lacZ gene, have been reported to show a 15-20% incidence ofspontaneous medulloblastoma (Goodrich et al., 1997; Wetmore etal., 2000). Transcriptional profiling has shown that cyclin D1 isamong the genes most robustly activated by Shh signaling inGNPs (Lee et al., 2003; Oliver et al., 2003), the cell from whichthe tumors are believed to arise, and in vitro cyclin D1 proteinlevels are more rapidly upregulated relative to other D-typecyclins in GNPs in response to Shh (Kenney and Rowitch, 2000).The observation that cyclin D1 plays a crucial role in normal earlyGNP proliferation prompted us to determine whether loss ofcyclin D1 might decrease medulloblastoma incidence in Ptch1+/–

mice.Ptch1+/–;Ccnd1+/– mice were crossed to generate both

Ptch1+/–;Ccnd1+/+ and Ptch1+/–;Ccnd1–/– mice. As had beenpreviously reported (Sicinski et al., 1995), Ccnd1–/– pups showedhigh rates of early lethality with an especially steep decline in survivalduring the first 5 weeks of life, with 54% of Ptch1+/–;Ccnd1–/– pupssurviving to 5 weeks. After this period, the rate of mortality was verylow, allowing us to examine medulloblastoma developmentin surviving adults. The peak incidence of symptomatic

3933RESEARCH ARTICLECyclin D1 in medulloblastoma development

Fig. 4. Proliferation is decreased in Ccnd1–/– cerebella duringearly development. (A-H) Representative BrdU (A-D) and H3-P (E-H)immunohistochemistry of cerebella from P0 (A,B,E,F) and P6 (C,D,G,H)pups sacrificed 1 hour after BrdU injection. (I) Quantification of BrdUincorporation in the anterior EGL of wild-type and Ccnd1–/– (cycD1 onfigure) siblings at P0 (n=7) and P6 (n=4). *P�0.03 when comparingbetween +/+ and –/– littermates. (J) Quantification of H3-P positivenuclei in the EGL at P0 (n=8, P=0.07) and P6 (n=4, P=0.22) shows atrend of decreased mitotic cells in P0 but not in P6 GNPs in the mutant.

Fig. 5. Purkinje cell dendritic development is abnormal inCcnd1–/– mice at P9. Comparison of wild-type (A,C) and Ccnd1–/– (B,D)Purkinje cells stained with anti-calbindin antibody shows that at P9(A,B), Ccnd1–/– Purkinje cell dendrites are abnormally oriented andstunted, but by 4 weeks (C,D) they look normal. Scale bars: 50 �m inA,B; 100 �m in C,D.

DEVELO

PMENT

3934

medulloblastoma in Ptch1+/– mice occurs at 16-24 weeks of age(Wetmore et al., 2000), and a similar incidence of tumors (7/23) wasfound in asymptomatic mice in this age range (Goodrich et al., 1997).All the Ptch1+/–;Ccnd1–/– mice included in our analysis (Fig. 6H)survived within or beyond this time point.

Adult mice were monitored and maintained until they showedsymptoms of illness (decreased weight, lethargy, difficulty moving,change in the shape of their head) or reached well beyond thewindow of normal tumor development. All brains were examinedeither histologically or grossly for tumors. Brains were stained withX-gal and the cerebella were assessed under the dissecting scopeor after histological sectioning (Fig. 6A-C). Out of 89Ptch1+/–;Ccnd1+/+ brains inspected, 19 were found to havemedulloblastomas (21%) (Fig. 6H). We did not find any tumors inasymptomatic mice, suggesting that by this age tumors are largeenough to produce unmistakable symptoms. By contrast, only oneof 42 (2.4%) Ptch1+/–;Ccnd1–/– mice developed a medulloblastoma(Fig. 6H). In total another five of the 42 adult Ptch1+/–;Ccnd1–/– miceshowed rapid weight loss and lethargy and were sacrificed, but noneof these had a medulloblastoma. One Ptch1+/–;Ccnd1+/+ and 3Ptch1+/–;Ccnd1–/– mice were found dead in their cages despite noprevious symptoms, and their brains could not be examined becauseof advanced autolysis.

To exclude the possibility that we did not find tumors in adultPtch1+/–;Ccnd1–/– mice because pups with tumors were selectedagainst during earlier development, we examined eightPtch1+/–;Ccnd1–/– cerebella from mice 2.5-3.0 weeks old by stainingthe brain for �-gal activity and comparing them withPtch1+/–;Ccnd1+/+ cerebella of the same age. Although we expectedonly about half of the Ptch1+/–;Ccnd1–/– mice of this age to surviveto adulthood, we could find no major abnormalities in cerebellarpatterns of �-galactosidase activity. In fact, there appeared to befewer regions of abnormal surface staining than in Ptch1+/–;Ccnd1+/+

cerebella. In conclusion, our results show that although notabsolutely essential, cyclin D1 plays an important role in Shhsignaling induced medulloblastomas.

Medulloblastomas from Ptch1+/– mice expressboth cyclins D1 and D2As the biochemical activities of the D-type cyclins appear similar, ithas been hypothesized that the requirement for a specific cyclin in aspecific tissue type is due largely to the regulation of its expression.In mouse breast tumor models, the inhibition of tumorigenesis inCcnd1–/– tissue has been linked to the ability of an oncogenicsignaling pathway to upregulate cyclin D1 specifically (Yu et al.,2001). To determine if the requirement for cyclin D1 inmedulloblastomas in Ptch1+/– mice could be explained by it being theonly D-type cyclin expressed, we assessed the levels of both cyclinsD1 and D2 by immunohistochemistry. We found that both cyclins D1and D2 were highly expressed in the eight Ptch1+/–;Ccnd1+/+ tumorsexamined (Fig. 6E,G). All tumors had a large fraction of cellsexpressing cyclins D1 (39±12.6%) and D2 (42±9.2%). BrdU stainingshowed that these tumors were also highly proliferative, with anaverage of 25% of cells showing BrdU incorporation (Fig. 6F).Interestingly, the single Ptch1+/–;Ccnd1–/– tumor that developed hada similar percentage of cyclin D2 positive cells (40%) (Fig. 6J) andwas highly proliferative (33% BrdU positive) (Fig. 6I). Thus, theabsence of cyclin D1 did not impact tumor cell proliferation at thisstage. As there was no further increase in the level of cyclin D2 in thePtch1+/–;Ccnd1–/– medulloblastoma, these observations indicate thatcyclin D1 is not required for medulloblastoma cell proliferation andsuggest that it is more likely to play a key role at an earlier stage inmedulloblastoma development.

Decreased ‘pre-neoplastic’ lesions inPtch1+/–;Ccnd1–/– miceAdult Ptch1+/– mice have been reported to maintain ectopic clustersof cells at the cerebellar surface, and these have been suggested torepresent pre-neoplastic lesions (Goodrich et al., 1997; Kim et al.,2003; Oliver et al., 2005). To determine if removal of cyclin D1affected the incidence of these pre-neoplastic lesions, we dissected3-week-old Ptch1+/–;Ccnd1+/+ and Ptch1+/–;Ccnd1–/– cerebella, andexamined sagittal sections every 200 �m for ectopic lesions. As

RESEARCH ARTICLE Development 133 (19)

Fig. 6. Medulloblastoma development issignificantly decreased in Ptch1+/–;Ccnd1–/– micecompared with Ptch1+/– mice, despite upregulationof both cyclins D1 and D2 in tumors. Tumors wereidentified by staining with X-gal (A,B): in the normalcerebellum (A), the staining is strongest in the Purkinjecell layer and inner granule layer, but medulloblastomas(B) were easily identified by their disruption of thispattern resulting in staining throughout the tumor,including at the surface of the brain.(C) Medulloblastomas were also identified histologically.(D-G) Immunohistochemistry of Ptch1+/–

medulloblastomas. (D) In contrast to mature granule cellsof the IGL, tumor cells are only weakly positive for thepostmitotic marker NeuN. Expression of both cyclins D1(E) and D2 (G), as well as BrdU incorporation (F) in thetumor but not in adjacent IGL or molecular layers.(H) Ptch1+/–;Ccnd1–/– (cycD1 on figure) mice developsignificantly fewer medulloblastomas than do Ptch1+/–

mice P=0.0105. (I,J) The medulloblastoma thatdeveloped in Ptch1+/–;Ccnd1–/– background showedsimilar patterns of BrdU incorporation (I) and cyclin D2 (J)expression. Mb, medulloblastoma; ML, molecular layer;IGL, inner granule layer. Scale bars: 500 �m in C; 100�m in D-G,I,J.

DEVELO

PMENT

expected, in the five wild-type cerebella examined the EGL wasentirely absent, and no ectopic clusters of cells were observed. Bycontrast, all Ptch1+/–;Ccnd1+/+ and some Ptch1+/–;Ccnd1–/– mice hadectopic clusters of cells above the molecular layer. We divided theseclusters into two types based on cell proliferation: one that wasproliferative (Fig. 7A-E) and the other that was not proliferative (Fig.7F-J), as determined by staining for BrdU or H3-P. Similar to maturegranule neurons in the IGL, all cells in the non-proliferative clustersexpressed high levels of the post-mitotic neural marker NeuN (Fig.7F,G), and these clusters did not express cyclin D1 or D2 (Fig. 7I,J).As we observed similar small ectopic clusters of non-proliferativegranule neurons in adult (35-week-old) Ptch1+/–;Ccnd1+/+ mice thatdid not develop medulloblastomas, these non-proliferative clustersare unlikely to represent pre-neoplastic lesions. Interestingly, nosignificant difference in the incidence of these non-proliferativelesions was observed between adult Ptch1+/–;Ccnd1+/+ andPtch1+/–;Ccnd1–/– mice (Fig. 7K).

By contrast, proliferative lesions were only detected in youngmice, and they correspond to the previously described ‘pre-neoplastic’ lesions in terms of their histology, proliferation and geneexpression properties as described below (Oliver et al., 2005). Theselesions expressed both cyclins D1 and D2 (Fig. 7D,E), and were onlyweakly positive for NeuN (Fig. 7A,B), similar to medulloblastomas(Fig. 6D). Unlike the normal EGL (Fig. 4C,D), there was noorganization into outer proliferative and inner post-mitotic layers. Insome proliferative lesions, BrdU-positive cells were found relativelyevenly throughout the entire collection of cells (not shown), whilein others (Fig. 7C) there were areas of abnormal organization withBrdU-positive cells at both the inner and outer edges of the cluster.

In these cases, comparison of serial sections showed that non-proliferative regions were more likely to express NeuN (Fig. 7B).There was also an increased number of strongly NeuN-positive cellsin the molecular layer below the proliferative lesions (compare Fig.7B and 7G), suggesting that some cells from the lesion weredifferentiating and migrating out of these clusters. Additionally,TUNEL staining revealed apoptotic cells within these lesions thatwere not detected in the normal mature granule layer (see Fig. S3 inthe supplementary material).

Each section was examined for proliferative pre-neoplasticlesions. If a lesion appeared in the same folia in two adjacentsections, it was deemed to represent a single lesion, whereas if alesion appeared distinctly anterior or posterior to a lesion in aneighboring section it probably represented a different lesion. Thesize of individual lesions varied widely, with some small lesionsappearing in only two consecutive sections (therefore extending lessthan 600 �m in the mediolateral direction), while other large lesionsspanned up to 11 sequential sections (up to 2400 �m). The volumeof a lesion was estimated based on its average cross-sectional areaand its mediolateral span. Ptch1+/–;Ccnd1–/– pups were found to havesignificantly fewer of these pre-neoplastic lesions (Fig. 7K). Theaverage number of independent lesions was approximately half thatof Ptch1+/–;Ccnd1+/+ mice, while the total volume of lesions permouse was sixfold smaller. The average size of an individual lesionwas also one quarter of that in Ptch1+/–;Ccnd1+/+ mice. Therefore,we conclude that cyclin D1 contributes to the growth of pre-neoplastic lesions in young Ptch1+/– mice and that the decrease inthe number of these cells probably accounts for the decreasedincidence of medulloblastoma in Ptch1+/–;Ccnd1–/– mice.

3935RESEARCH ARTICLECyclin D1 in medulloblastoma development

Fig. 7. Ptch1+/–;Ccnd1–/– mice develop fewer pre-neoplastic lesions. (A-E) Pre-neoplastic lesions (indicated by broken lines) express both cyclinsD1 (D) and D2 (E) and incorporate BrdU (C). (A,B) Some cells in these lesions, particularly in BrdU-negative regions, are weakly NeuN positive. Thedensity of NeuN positive cells is increased in the molecular layer below the lesion (A,B) when compared with below the mature cluster (F,G).(F-J) Clusters of mature ectopic granule cells (indicated by broken lines) do not incorporate BrdU (H) or express cyclins D1 (I) or D2 (J), and all cellsare strongly positive for the post-mitotic neural marker NeuN (F,G). Scale bars: 100 �m. (K) There is a significant decrease in the number of pre-neoplastic lesions in Ptch1+/–;Ccnd1–/– (cycD1 on figure) mice when compared with Ptch1+/– mice (*P=0.021), and a smaller total volume ofpreneoplastic lesions (**P=0.020), but no significant difference in the number of mature ectopic clusters seen in older adults. ML, molecular layer;IGL, inner granule layer.

DEVELO

PMENT

3936

DISCUSSIONEarly and late GNPs have different D type cyclinexpressionOur results demonstrate that the expression of D-type cyclins in GNPschanges over time, possibly indicative of different stages of GNPs asthey progress towards differentiation. Specifically, our analysis of theexpression of cyclins D1 and D2 revealed that GNPs at P0 expressedhigh levels of only cyclin D1, while at P6 both cyclins D1 and D2 wereexpressed. Consistent with these observations, the absence of cyclinD1 impaired the proliferation of P0 GNPs but not P6 GNPs. We havenot examined time points between P0 and P6 to determine exactlywhen cyclin D2 began to appear during cerebellar development, but itwas reported that cyclin D2 mRNA was absent from the embryoniccerebellum from E14.5 to E18.5, and was detected postnatally as earlyas P2 (Ross et al., 1996) and that Ccnd2–/– cerebella began to showphenotypes after P3 (Huard et al., 1999). These reports are consistentwith our cyclin D2 expression data and our suggestion that cyclin D2can compensate for the absence of cyclin D1 at P6.

Although both early and late GNPs can proliferate, it is possiblethat they have distinct capacities to be transformed. For example,early GNPs potentially behave more like stem cells and are thereforemore susceptible to transformation, while later GNPs have a morelimited proliferation potential because they are more committed todifferentiation. In support of this idea, it was reported that there wasa gradual decrease in the ability of irradiation to increase theincidence of medulloblastoma in Ptch1+/– pups as they aged from P1to P10: irradiation at P1 or P4 increased medulloblastoma incidenceto 81% or 51%, respectively, while by P10 irradiation had no effect(Pazzaglia et al., 2002; Pazzaglia et al., 2006). This is somewhatsurprising given that there are many more GNPs at P4 than at P1, sothis difference probably reflects a greater susceptibility of earlyGNPs to transformation.

An early role of cyclin D1 in medulloblastomadevelopmentWe propose a model that explains a role for cyclin D1 inmedulloblastoma development. Recently a pre-neoplastic stage ofmedulloblastoma was found to exhibit gene expression patternsintermediate between those of GNPs and tumor cells (Oliver et al.,2005), suggesting that medulloblastomas are derived from theselesions. The pre-neoplastic cells lacked expression of the wild-typePtch1 allele, suggesting that loss of Ptch1 expression is a rate-limiting step in the transition from normal GNPs to pre-neoplasticlesions. We suggest that genetic or epigenetic changes that activateShh signaling in either early or late GNPs could result in pre-neoplastic lesions with cells that have the capacity to not onlyproliferate but also differentiate or die. However, lesions derivedfrom early GNPs also contain a population of cells that behave morelike stem cells and continue to renew, allowing them to acquireadditional genetic and epigenetic changes, which may lead tomalignant medulloblastoma. By contrast, transformation of lateGNPs would be less likely as these cells have a decreasedproliferation potential and would require re-activation of self-renewal pathways. This model predicts that removal of cyclin D1,by limiting the proliferation of early GNPs, would reduce thenumber of lesions arising from early GNPs. This is consistent withour findings of an approximate halving in the number of independentpre-neoplastic lesions. In addition, Ccnd1–/– early GNPs may beintrinsically less competent for tumorigenesis, owing to theirimpaired ability to proliferate. In support of this, we saw a moredramatic decrease in the total volume of pre-neoplastic cells and inthe incidence of medulloblastomas in Ptch1+/–;Ccnd1–/– mice.

Interestingly, while multiple pre-neoplastic lesions per mousewere observed in young mice, only one or two lesions were observedby 6-8 weeks of age (Oliver et al., 2005). These results suggest thatthe majority of pre-neoplastic lesions will not progress to tumors. Aswe detected both apoptosis and differentiation within these pre-neoplastic lesions, it is likely that the abnormal cells in most of theselesions either differentiate or undergo apoptosis, resulting in thedisappearance of the lesions. In addition to pre-neoplastic lesions,we found that both young and adult Ptch1+/– mice had non-proliferative clusters of cells, which were abnormally localized,morphologically similar to mature granule neurons and stronglyNeuN-positive. We speculate that pre-neoplastic lesions mighttransition into these non-proliferative clusters. In support of this, thepre-neoplastic lesions often exhibited areas of non-proliferation withNeuN-positive cells.

This specific requirement for cyclin D1 during earlytumorigenesis in the Ptch1+/– medulloblastoma model contrastswith its role as described in other tumor models (Robles et al., 1998;Yu et al., 2001), which revealed an important role for cyclin D1 onlywhen cyclin D1 was the only D-type cyclin upregulated in thetumor. Furthermore, in the breast tumor model, continued cyclinD1/cdk4 kinase activity is required for tumor cell proliferation (Yuet al., 2006). By contrast, it does not seem likely that cyclin D1is required for proliferation of medulloblastoma cells asmedulloblastomas that developed in Ptch1+/–;Ccnd1+/+ miceshowed upregulation of both cyclins D1 and D2, and the onePtch1+/–;Ccnd1–/– medulloblastoma that developed did not show adecreased level of proliferation.

Other cell cycle regulators in GNPs andmedulloblastomasDuring granule cell development, a number of cell cycle regulatoryproteins, including cyclin D2 and the cell cycle inhibitors p18 andp27, are expressed in at least a partially overlapping manner, anddeletion of any one is often sufficient to create a mild cerebellarphenotype. The Cip/Kip family member p27 was reported to be theonly member of the Cip/Kip family of Cdk inhibitors expressed ingranule cell precursors, and it is detected in the inner two-thirds ofthe EGL and in the IGL at P7 (Miyazawa et al., 2000). In additionto p27, at least one member of the Ink family of Cdk inhibitors,p18ink4c, which inhibits cyclin D/cdk4 kinase activity, is expressedin the pool of proliferating granule cell precursors by P5, althoughit is absent from early EGL cells at E15.5 (Zindy et al., 2003).Given the intriguing possibility from our results thatmedulloblastoma development may involve early GNPs that arepresent at P0, it will be interesting to define more precisely thespatiotemporal expression of other cell cycle regulators and thesignaling pathways that regulate them in the GNPs at this stage. Inaddition, as cyclin D2 is not expressed in early GNPs and removalof cyclin D2 affects later GNP proliferation and apoptosis, it willbe interesting to determine if removal of cyclin D2 affects theincidence of medulloblastomas and the number or size of pre-neoplastic lesions.

This work is supported by grants from the American Cancer Society, NationalInstitute of Health and the Fletcher Scholar award from the Cancer ResearchFoundation to W.D. J.P. was supported by the Medical Scientist NationalResearch Service Award Grant and W.D. is a Scholar of the Leukemia andLymphoma Society.

Supplementary materialSupplementary material for this article is available athttp://dev.biologists.org/cgi/content/full/133/19/3929/DC1

RESEARCH ARTICLE Development 133 (19)

DEVELO

PMENT

ReferencesAruga, J., Minowa, O., Yaginuma, H., Kuno, J., Nagai, T., Noda, T. and

Mikoshiba, K. (1998). Mouse Zic1 is involved in cerebellar development. J.Neurosci. 18, 284-293.

Berman, D. M., Karhadkar, S. S., Hallahan, A. R., Pritchard, J. I., Eberhart, C.G., Watkins, D. N., Chen, J. K., Cooper, M. K., Taipale, J., Olson, J. M. et al.(2002). Medulloblastoma growth inhibition by hedgehog pathway blockade.Science 297, 1559-1561.

Ciemerych, M. A., Kenney, A. M., Sicinska, E., Kalaszczynska, I., Bronson, R.T., Rowitch, D. H., Gardner, H. and Sicinski, P. (2002). Development of miceexpressing a single D-type cyclin. Genes Dev. 16, 3277-3289.

Crawley, J. N. (1999). Behavioral phenotyping of transgenic and knockout mice:experimental design and evaluation of general health, sensory functions, motorabilities, and specific behavioral tests. Brain Res. 835, 18-26.

Gilbertson, R. J. (2004). Medulloblastoma: signalling a change in treatment.Lancet Oncol. 5, 209-218.

Goldowitz, D. and Hamre, K. (1998). The cells and molecules that make acerebellum. Trends Neurosci. 21, 375-382.

Goodrich, L. V., Milenkovic, L., Higgins, K. M. and Scott, M. P. (1997). Alteredneural cell fates and medulloblastoma in mouse patched mutants. Science 277,1109-1113.

Hahn, H., Wicking, C., Zaphiropoulous, P. G., Gailani, M. R., Shanley, S.,Chidambaram, A., Vorechovsky, I., Holmberg, E., Unden, A. B., Gillies, S.et al. (1996). Mutations of the human homolog of Drosophila patched in thenevoid basal cell carcinoma syndrome. Cell 85, 841-851.

Huard, J. M., Forster, C. C., Carter, M. L., Sicinski, P. and Ross, M. E. (1999).Cerebellar histogenesis is disturbed in mice lacking cyclin D2. Development 126,1927-1935.

Kenney, A. M. and Rowitch, D. H. (2000). Sonic hedgehog promotes G(1) cyclinexpression and sustained cell cycle progression in mammalian neuronalprecursors. Mol. Cell. Biol. 20, 9055-9067.

Kim, J. Y., Nelson, A. L., Algon, S. A., Graves, O., Sturla, L. M., Goumnerova,L. C., Rowitch, D. H., Segal, R. A. and Pomeroy, S. L. (2003).Medulloblastoma tumorigenesis diverges from cerebellar granule celldifferentiation in patched heterozygous mice. Dev. Biol. 263, 50-66.

Lee, Y., Miller, H. L., Jensen, P., Hernan, R., Connelly, M., Wetmore, C., Zindy,F., Roussel, M. F., Curran, T., Gilbertson, R. J. et al. (2003). A molecularfingerprint for medulloblastoma. Cancer Res. 63, 5428-5437.

Miyazawa, K., Himi, T., Garcia, V., Yamagishi, H., Sato, S. and Ishizaki, Y.(2000). A role for p27/Kip1 in the control of cerebellar granule cell precursorproliferation. J. Neurosci. 20, 5756-5763.

Oliver, T. G., Grasfeder, L. L., Carroll, A. L., Kaiser, C., Gillingham, C. L., Lin, S.M., Wickramasinghe, R., Scott, M. P. and Wechsler-Reya, R. J. (2003).Transcriptional profiling of the Sonic hedgehog response: a critical role for N-mycin proliferation of neuronal precursors. Proc. Natl. Acad. Sci. USA 100, 7331-7336.

Oliver, T. G., Read, T. A., Kessler, J. D., Mehmeti, A., Wells, J. F., Huynh, T. T.,Lin, S. M. and Wechsler-Reya, R. J. (2005). Loss of patched and disruption ofgranule cell development in a pre-neoplastic stage of medulloblastoma.Development 132, 2425-2439.

Pazzaglia, S., Mancuso, M., Atkinson, M. J., Tanori, M., Rebessi, S., Majo, V.D., Covelli, V., Hahn, H. and Saran, A. (2002). High incidence of

medulloblastoma following X-ray-irradiation of newborn Ptc1 heterozygousmice. Oncogene 21, 7580-7584.

Pazzaglia, S., Tanori, M., Mancuso, M., Rebessi, S., Leonardi, S., Di Majo, V.,Covelli, V., Atkinson, M. J., Hahn, H. and Saran, A. (2006). Linking DNAdamage to medulloblastoma tumorigenesis in patched heterozygous knockoutmice. Oncogene 25, 1165-1173.

Robles, A. I., Rodriguez-Puebla, M. L., Glick, A. B., Trempus, C., Hansen, L.,Sicinski, P., Tennant, R. W., Weinberg, R. A., Yuspa, S. H. and Conti, C. J.(1998). Reduced skin tumor development in cyclin D1-deficient mice highlightsthe oncogenic ras pathway in vivo. Genes Dev. 12, 2469-2474.

Romer, J. T., Kimura, H., Magdaleno, S., Sasai, K., Fuller, C., Baines, H.,Connelly, M., Stewart, C. F., Gould, S., Rubin, L. L. et al. (2004). Suppressionof the Shh pathway using a small molecule inhibitor eliminates medulloblastomain Ptc1(+/–)p53(–/–) mice. Cancer Cell 6, 229-240.

Ross, M. E., Carter, M. L. and Lee, J. H. (1996). MN20, a D2 cyclin, is transientlyexpressed in selected neural populations during embryogenesis. J. Neurosci. 16,210-219.

Sherr, C. J. and Roberts, J. M. (2004). Living with or without cyclins and cyclin-dependent kinases. Genes Dev. 18, 2699-2711.

Sicinski, P., Donaher, J. L., Parker, S. B., Li, T., Fazeli, A., Gardner, H., Haslam,S. Z., Bronson, R. T., Elledge, S. J. and Weinberg, R. A. (1995). Cyclin D1provides a link between development and oncogenesis in the retina and breast.Cell 82, 621-630.

Singh, S. K., Clarke, I. D., Terasaki, M., Bonn, V. E., Hawkins, C., Squire, J.and Dirks, P. B. (2003). Identification of a cancer stem cell in human braintumors. Cancer Res. 63, 5821-5828.

Singh, S. K., Hawkins, C., Clarke, I. D., Squire, J. A., Bayani, J., Hide, T.,Henkelman, R. M., Cusimano, M. D. and Dirks, P. B. (2004). Identification ofhuman brain tumour initiating cells. Nature 432, 396-401.

Stecca, B. and Ruiz i Altaba, A. (2005). Brain as a paradigm of organ growth:Hedgehog-Gli signaling in neural stem cells and brain tumors. J. Neurobiol. 64,476-490.

Wechsler-Reya, R. and Scott, M. P. (2001). The developmental biology of braintumors. Annu. Rev. Neurosci. 24, 385-428.

Wechsler-Reya, R. J. and Scott, M. P. (1999). Control of neuronal precursorproliferation in the cerebellum by Sonic Hedgehog. Neuron 22, 103-114.

Wetmore, C., Eberhart, D. E. and Curran, T. (2000). The normal patched allele isexpressed in medulloblastomas from mice with heterozygous germ-linemutation of patched. Cancer Res. 60, 2239-2246.

Yoshioka, H., Mino, M., Morikawa, Y., Kasubuchi, Y. and Kusunoki, T. (1985).Changes in cell proliferation kinetics in the mouse cerebellum after totalasphyxia. Pediatrics 76, 965-969.

Yu, Q., Geng, Y. and Sicinski, P. (2001). Specific protection against breast cancersby cyclin D1 ablation. Nature 411, 1017-1021.

Yu, Q., Sicinska, E., Geng, Y., Ahnstrom, M., Zagozdzon, A., Kong, Y.,Gardner, H., Kiyokawa, H., Harris, L. N., Stal, O. et al. (2006).Requirement for CDK4 kinase function in breast cancer. Cancer Cell 9, 23-32.

Zindy, F., Nilsson, L. M., Nguyen, L., Meunier, C., Smeyne, R. J., Rehg, J. E.,Eberhart, C., Sherr, C. J. and Roussel, M. F. (2003). Hemangiosarcomas,medulloblastomas, and other tumors in Ink4c/p53-null mice. Cancer Res. 63,5420-5427.

3937RESEARCH ARTICLECyclin D1 in medulloblastoma development