Loss of the tumor suppressor gene PTEN exerts diverse outcomes on cancer in different developmental contexts. To gain insight into the effect of its loss on outcomes in the brain, we conditionally inactivated the murine Pten gene in neonatal neural stem/progenitor cells. Pten inactivation created an abnormal perivascular proliferative niche in the cerebellum that persisted in adult animals but did not progress to malignancy. Proliferating cells showed undifferentiated morphology and expressed the progenitor marker Nestin but not Math1, a marker of committed granule neuron progenitors. Codeletion of Pten and Trp53 resulted in fully penetrant medulloblastoma originating from the perivascular niche, which exhibited abnormal blood vessel networks and advanced neuronal differentiation of tumor cells. EdU pulse-chase experiments demonstrated a perivascular cancer stem cell population in Pten/Trp53 double mutant medulloblastomas. Genetic analyses revealed recurrent somatic inactivations of the tumor suppressor gene Ptch1 and a recapitulation of the sonic hedgehog subgroup of human medulloblastomas. Overall, our results showed that PTEN acts to prevent the proliferation of a progenitor niche in postnatal cerebellum predisposed to oncogenic induction of medulloblastoma. Cancer Res; 77(1); 123–33. ©2016 AACR.
PTEN, one of the most commonly mutated genes in human cancer, is the major negative regulator of the PI3–kinase signaling cascade and engages multiple downstream mechanisms as a tumor suppressor (1). Importantly, PTEN also plays diverse roles in normal development. Within the brain, loss of PTEN can cause multiple complex phenotypes, including megalencephaly, seizures, autism, or mental retardation (2). PTEN mutations also occur in brain tumors, including glioblastoma and medulloblastoma (3–7). Here, we identified a novel function for Pten in regulating progenitor cells in the cerebellum.
Cells in the cerebellum arise from two germinal zones: the ventricular zone and the external granule layer (EGL). The ventricular zone is the embryonic germinal neuroepithelium that generates deep cerebellar nuclei, Purkinje cells, and other GABAergic neurons in the cerebellum. At birth, the ventricular germinal neuroepithelium disappears, and proliferative precursors in the rhombic lip emigrate along the pial surface to form the EGL. The EGL proliferates most during early postnatal days and produces all the cells that differentiate to become internal granule neurons (8). Sonic hedgehog (Shh) is expressed by Purkinje cells and binds to its receptor, Ptch1, on granule cell precursors in the EGL, promoting proliferation and preventing differentiation (9, 10). PI3K and Shh signaling pathways can synergize to promote granule cell progenitor proliferation in vitro (11). Pten, the major negative regulator of PI3K signaling, is also required for normal cerebellar development. Deletion of Pten in EGL granule progenitors causes ectopic positioning and hypertrophy of cerebellar granule cells (12–14).
Aside from the two well-characterized germinal zones in the developing cerebellum, multipotent neural stem cells have also been isolated from the cerebellum postnatally and are thought to be located in the white matter tracts (15), possibly arising from embryonic ventricular zone progenitors that emigrate into the white matter tracts (16). The identity and function of this pool of neural stem/progenitor cells are largely unknown.
Medulloblastoma is the most common malignant brain tumor of childhood. It is a collection of clinically and molecularly heterogeneous embryonal tumors and can be divided into four molecular groups defined by gene expression or DNA methylation signatures: SHH, WNT, Group 3, and Group 4 (17). The SHH group of medulloblastomas contains tumors carrying mutations in PTCH1 or SUFU, tumor suppressor genes that negatively regulate SHH signaling, as well as tumors without detectable mutations in these components, but with similar gene expression signatures. The SHH subgroup of medulloblastoma is thought to originate from the EGL precursors, which is supported by Ptch1-mutant mouse models (18–21). The Wnt subgroup of medulloblastoma is likely to develop from cells of the dorsal brainstem (22). Expression of c-Myc and loss of p53 function in granule neuron progenitor cells or in mouse cerebellar stem cells, defined by expression of Prominin 1 and absence of neuronal and glial lineage markers, can drive medulloblastomas that resemble human MYC-driven Group 3 medulloblastomas (23, 24). It is unclear how many different cell types may serve as cells of origin for Group 3 tumors, and the cell of origin for Group 4 medulloblastoma is not known.
Cancer-initiating cells, or cancer stem cells, are the subpopulation in the primary tumor that are capable of self-renewing, differentiating, and regenerating the original tumor when transplanted in vivo. Cancer-initiating cells have now been isolated from human cancers of the blood, breast, brain, skin, bone, and prostate (25). They are hypothesized to sustain tumor growth and to be responsible for relapse and metastasis. Cancer stem cells can also be isolated from human and mouse medulloblastomas (21, 26) and have been shown to reside in a perivascular niche (27). The perivascular cell population was found to account for the regrowth of Ptch1-mutant medulloblastoma after irradiation-induced apoptosis (28). However, whether there is a perivascular population of cells in the wild-type brain that is susceptible to oncogenic transformation and whether this population could serve as cells of origin for medulloblastoma are unknown.
In the current study, we inactivated Pten in neural stem/progenitor cells in neonatal mice and revealed a novel perivascular proliferative niche in the cerebellum. Combined deletion of Pten and Trp53 generated medulloblastomas with complete penetrance. Pten;Trp53-null medulloblastomas acquired somatic mutations in Ptch1 and had gene expression signatures that were similar to the human SHH subgroup of medulloblastoma. Paradoxically, Pten deletion induced abnormal proliferation of perivascular progenitor cells but was associated with extensive neuronal differentiation of bulk tumor cells in the context of medulloblastomas. Thus, Pten deletion may influence the expansion of progenitor cells that can serve as cells of origin for medulloblastoma and may also contribute to the differentiation state of tumor cells.
Materials and Methods
For Cre activity mapping, Nestin-creERT2 transgenic mice (29) were bred with R26R-EYFP (30) mice to generate Nestin-creERT2;R26R-EYFP mice. Cre activity was detected by EYFP immunofluorescence staining on frozen tissue sections. Nestin-creERT2;PtenloxP/loxP (PtencKO) mice were generated by crossing Nestin-creERT2 mice to PtenloxP mice (31), a gift from Tak Mak (University of Toronto, Toronto, Canada). Nestin-creERT2;PtenloxP/loxP;Trp53loxP/loxP (PtencKO;Trp53cKO) mice were generated by breeding Nestin-creERT2;PtenloxP/loxP with Trp53loxP mice (32). For all analyses including Nestin-creERT2, the transgene was hemizygous to avoid possible variation from cre transgene dosage. All procedures were reviewed and approved by the Animal Care and Use Committee at St. Jude Children's Research Hospital (Memphis, TN).
Genotyping PCR assays
PCR was used to genotype mice. Primers used were:
cre, forward primer: 5′-AGCGATCGCTGCCAGGAT-3′; reverse primer: 5′-ACCAGCGTTTTCGTTCTGCC-3′; expected product: 157 base pairs.
Pten, forward primer: 5′-TTATCTGGATCAACTTTGGGCC-3′; reverse primer: 5′-TCCCACCAATGAACAAACAGT-3′; expected products: 146 (wild-type) and 244 (floxed allele) base pairs.
Trp53, forward primer: 5′-CACAAAAACAGGTTAAACCCAG-3′; reverse primer: 5′-AGCACATAGGAGGCAGAGAC-3′; expected products: 288 (wild-type) and 370 (floxed allele) base pairs.
Induction of Cre activity
Tamoxifen (Sigma) was dissolved at 5 mg/mL at 37°C in corn oil (Sigma), sterilized by filtering, and stored at 4°C in the dark for up to 10 days. Pups were injected intraperitoneally daily with 3 mg tamoxifen/40 g body weight on P0 and P1. An equivalent volume of sterile filtered corn oil alone was used for vehicle injections. Two injections in the same mouse were separated by 24 hours.
5-ethynyl-2′-deoxyuridine injection and detection
5-ethynyl-2′-deoxyuridine (EdU; Invitrogen) was dissolved at 2.5 mg/mL in sterile PBS and stored at −20°C. EdU was injected intraperitoneally at a dose of 5 μg/g body weight every 2 hours for 5 injections. Fluorescent detection of EdU was performed using Click-iT EdU Cell Proliferation Assay Kit (Invitrogen).
IHC and immunofluorescence
Mice were anesthetized and perfused transcardially with 1× PBS, followed by 4% paraformaldehyde (PFA) in PBS. Following dissection, tissues were postfixed overnight in 4% PFA in PBS at 4°C and then equilibrated in 25% sucrose in PBS for an additional 24 hours at 4°C. Tissues were embedded in TBS embedding media (Triangle Biomedical Sciences) on dry ice and cut into 12-μm thick cryosections. Tissue slides were equilibrated at room temperature for 20 minutes and then washed three times in PBS prior to staining. For paraffin sections, mice were processed the same way as above. Following perfusion, tissues were postfixed for an additional 24 hours in 4% PFA, embedded in paraffin, and cut into 5-μm sections. Primary antibodies used for immunostaining were against GFP (1:1,000, Invitrogen A6455; 1:2,000, Abcam #13970), Pten (1:100, Cell Signaling Technology #9559), p-Akt S473 (1:50, Cell Signaling Technology #9271), Gfap (1:200, Sigma G3893), Ki67 (1:5000, Novocastra NCL-Ki67p), CD34 (1:100, BD Pharmingen #553731), CD45 (1:400, BD Pharmingen #553076), CD133 (1:100, eBioscience #12-1331), Nestin (1:100, Chemicon MAB353), PCNA (1:500, Santa Cruz Biotechnology #sc-56), and Math1 (1:200, Abcam, #ab105497). For IHC, microwave antigen retrieval was performed for all antibodies, and biotinylated secondary antibodies were used in conjunction with horseradish peroxidase–conjugated streptavidin (Elite ABC, Vector Laboratories). Color development was conducted with substrates NovaRed, DAB or VIP (Vector Laboratories) and counterstaining with hematoxylin or methyl green (Vector Laboratories), respectively. For immunofluorescence, Alexa Fluor 488, 647 (Invitrogen) and cyanine 3, 5 (Jackson ImmunoResearch)–conjugated secondary antibodies were employed along with Vectashield mounting media containing DAPI (Vector Laboratories).
Quantification of tumor vasculature
Representative hematoxylin and eosin (H&E) fields were taken from medulloblastomas of both Pten;Trp53 dKO and Trp53cKO (4 tumors/group). Images are analyzed by ImageJ (version 1.50i) “Image → Adjust → Threshold Color” and "Analyze → Measure" tools to quantify area of open blood vessels represented as percentage per field.
Comparative genomic hybridization and gene expression microarray
Genomic DNA was prepared by scraping tumor tissues from multiple deparaffinized sections, followed by proteinase K digestion and phenol–chloroform extraction. DNA quality was verified by gel electrophoresis. Control DNA was isolated from the spleens and livers of wild-type FVB mice. A total of 18 Pten;Trp53 dKO tumors were analyzed using the Roswell Park Cancer Institute (RPCI) 6.5K mouse BAC array comparative genomic hybridization (aCGH) as described previously (33).
Total RNA was extracted from frozen tumor tissue from Pten; Trp53 dKO using TRIzol reagent (Invitrogen). RNA integrity was verified on all samples using the Agilent 2100 Bioanalyzer (Agilent Technologies). Gene expression profiles were generated using the Affymetrix Mouse Genome 430 v2.0 chip.
aCGH and gene expression array data are deposited in Gene Expression Omnibus under accession number GSE40106.
Gene expression analysis
Double knockout (dKO) tumors resemble the human SHH medulloblastomas. The top 25 genes most highly expressed in each of the four subgroups were derived as gene sets, as reported by Kawauchi and colleagues (23) and used for gene set enrichment analysis (34). Comparing six Pten;Tp53 dKO tumors with three Pten cKO knockout cerebellum samples induced at P5, three P5 cerebellum samples and three Pten cKO samples induced at P30, only the gene set highly expressed in human SHH subgroup is significantly enriched in the dKO tumors (P = 0.082, FDR = 0.0767). Gene sets for the other three medulloblastoma subgroups were not significantly enriched in dKO tumors (FDR > 0.25).
Ptch1 sequence analysis
Total RNA was extracted from snap-frozen medulloblastoma samples using TRIzol (Invitrogen). cDNA was synthesized from tumor RNA by reverse transcription using random hexamer primers and SuperScript II Reverse Transcriptase (Invitrogen). After cDNA synthesis, Ptch1 sequencing analysis was done as describe by Lee and colleagues (35).
Induction of Nestin-creERT2 in neonatal mice directed extensive cre activity in cerebellum
Tamoxifen injection of adult Nestin-creERT2 mice induces cre activity in neural stem and progenitor cells (29). We bred these mice with R26R-LacZ or R26R-EYFP Cre reporter mice to map the Cre activity induced by tamoxifen administration in neonatal mice. Consistent with the fact that the developing brain has a stronger expression of Nestin and more numerous neural stem/progenitors, induction of Cre activity induced by Nestin-creERT2 at postnatal days 0 and 1 (P0 and P1) was more robust than observed in adult brain, with strongest activities in the neural stem/progenitor cell niches including the subventricular zone of the lateral ventricle (SVZ; ref. 29). The bulk of the cerebellum develops postnatally, and tamoxifen injection at P0 and P1 induced extensive Nestin-creERT2 activity in the cerebellum, including the EGL, internal granule layer (IGL), and white matter tracts very effectively (Supplementary Fig. S1A–S1D). Cre activity was also detected by reporter expression in NG2-expressing glial progenitor cells adjacent to the blood vessel, but absent in CD34-expressing vascular endothelial cells.
Pten deletion causes cerebellar disorganization and hypertrophy
To determine the effect of broad Pten loss on the developing brain, we induced deletion of Pten in Nestin-CreERT2;PtenloxP/loxP mice by tamoxifen injection at P0 and P1 (hereafter PtencKO mice). PtencKO mice appeared normal until 2 to 3 months of age and then began exhibiting lethargy, increasing macrocephaly, and progressive neurologic abnormalities, including ataxia and seizure, requiring euthanasia with a median survival of 153 days. One hundred percent of the mice (24/24) induced at birth were affected, whereas none of the uninduced PtenloxP/loxP mice or tamoxifen-treated wild-type littermates showed these deficits. In addition to the previously described dramatic expansion of the SVZ and rostral migratory stream (29), brains from PtencKO mice showed profound hydrocephalus and dramatic increases in overall size, especially in the cerebellum (Fig. 1A), where foliation was visible, but the organization of cell layers was disrupted. The molecular layer (ML) was dramatically thickened with many Pten-deficient ectopic granule cells. The majority of Purkinje cells expressed Pten, but were not well aligned at the interface of the IGL and ML. The IGL was thinner than wild-type control and composed of both Pten-positive and -negative granule cells (Fig. 1B). The size of the Pten-deficient granule cells was larger than the Pten-positive granule cells in aging mice (not shown). All of these abnormalities mimic human Lhermitte–Duclos disease and are in agreement with previous findings but with a more severe manifestation compared with other Pten conditional knockout models in which Pten deletion was driven by different GFAP-cre mice (12, 13).
Pten inactivation induced novel perivascular proliferative niches in the cerebellum
In addition to the cerebellar disorganization and hypertrophy, we identified multiple perivascular proliferative niches in PtencKO cerebellum, which were composed of hyperchromatic cells with progenitor-like morphology (20/24 mice; Fig. 2). Previous models in which Pten conditional deletion was driven by GFAP-cre showed extensive deletion in cerebellar granule cells, but no evidence of hyperproliferative lesions. In these models, cre activity was limited to the majority of cerebellar granule neurons, as well as more variable activity in astrocytes, granule neurons of the dentate gyrus, and subsets of neurons in the cerebral cortex, and minimal activity in the subventricular zone despite the known expression of Gfap in the stem cells of the subventricular zone. The cre activity in GFAP-cre mice is not completely overlapping with endogenous Gfap expression, presumably due to integration site effects of the transgene (12, 13, 36). Nestin is endogenously expressed in a broader range of stem and progenitor cells than Gfap, including all cells within the subventricular zone, and the Nestin-creER driver used here showed strong induction of cre activity in these stem/progenitor compartments (32) as well as granule neurons and perivascular cells in the cerebellum (Supplementary Fig. S1D). The perivascular hyperplastic lesions varied in size from as few as several cells to hundreds of cells and were observed as early as 3 months of age but were not found in 30-day-old mice. These hyperplastic lesions grew surrounding CD34+ blood vessels and had a high percentage of Ki67+ cells (Fig. 2). The cells did not express CD45 and, therefore, were not extravasated lymphocytes (Supplementary Fig. S2). The lesions were distributed throughout cerebellum, but occurred more frequently in anterior and lateral cerebellar lobules and in areas around the pia between lobules. They expressed Nestin and CD133 (Fig. 2), which mark neural stem/progenitor cells or brain tumor–initiating cells (26, 37), but did not express Math1, a marker for granule cell progenitors of the EGL (Supplementary Fig. S2). They also expressed Gfap, which labels neural stem cells in the SVZ in addition to astrocytes (Fig. 2; ref. 38). The perivascular precursor-like cells are Pten null and p-Akt positive, consistent with cells that underwent Cre-mediated recombination (Fig. 2). These perivascular hyperplastic lesions persisted but did not develop into tumors in mice up to 10 months of age. Although Pten has been shown to influence stem cell self-renewal (39, 40), this is the first example of Pten loss inducing a novel progenitor niche.
Concurrent Pten and Trp53 deletion induced medulloblastoma with extensive vascularization and highly differentiated tumor cells
TP53 is one of the most frequently mutated genes in human cancer and is also recurrently mutated in different types of brain tumors (7, 17). Therefore, we asked whether Trp53 deficiency would cooperate with Pten deletion to transform the perivascular proliferative niche into a full-blown cerebellar malignancy. Deletion of Trp53 alone by tamoxifen induction of Nestin-creERT2;Trp53loxP/loxP (Trp53cKO) mice at P0 and P1 resulted in a median onset of morbidity of 135 days, with 47% of mice developing medulloblastoma and the remainder developing non-CNS tumors or other types of brain tumors (Fig. 3A). Combined deletion of Pten and Trp53 accelerated tumor onset and increased the incidence of medulloblastoma. One hundred percent (26/26) of induced Nestin-creERT2;PtenloxP/loxP;Trp53loxP/loxP (hereafter dKO mice) developed medulloblastomas with a median onset of morbidity of 63 days of age (Fig. 3A). All dKO medulloblastomas had two striking features that distinguished them from Trp53cKO medulloblastomas, extensive and abnormal vascularization, and highly differentiated tumor cells surrounding a perivascular tumor stem cell niche (Figs. 3B–D and 4). Vascularization of dKO medulloblastomas was significantly more extensive (P < 0.0009) and included abnormal blood vessels with dilated and angular shapes. Blood clots were often seen in the lumen of abnormal blood vessels (Fig. 3C; Supplementary Fig. S3). Pten immunohistochemical staining showed that the vast majority of the tumor mass was devoid of Pten, with only blood vessel endothelial cells retaining Pten (Fig. 3D). In addition to cerebellar tumors, dKO mice also developed gliomas in the forebrain (18/26) and brain stem (5/26).
Both dKO and Trp53cKO medulloblastomas were positive for class III β-tubulin, a marker for neurons. However, the dKO medulloblastomas exhibited an advanced neuronal immunophenotype, with the majority of tumor cells expressing NeuN and GabaARα6, markers that are normally expressed in mature cerebellar granule neurons that were not expressed in Trp53cKO medulloblastomas (Fig. 4). Thus, Pten loss drove abnormal angiogenesis and extensive neuronal differentiation in dKO medulloblastomas.
Early proliferative lesions showed that Pten;Trp53 dKO medulloblastoma initiates from a perivascular niche
To investigate where the dKO medulloblastoma originates, and how it relates to the perivascular proliferative lesions in PtencKO cerebellum, we analyzed dKO cerebella at P21, P31, and P51, earlier than the median onset of morbidity at P63. Proliferation associated with the development of wild-type cerebellum is completed by P21, with only rare Ki67+ proliferative cells detected under the pial surface or in the white matter tracts. In contrast, the dKO cerebellum at P21 showed disrupted lamination and abundant Ki67+ proliferative cells throughout the cerebellum, which were concentrated around blood vessels either under the pial surface or in the parenchyma (Fig. 5). By P31, the number of Ki67+ cells dramatically increased. The hyperchromatic cell clusters around blood vessels were easily visible by H&E staining (Fig. 5). Although the dKO mice were not symptomatic by P51, there were sizable tumor masses present in the cerebellum. The perivascular growth pattern of tumor cells was maintained at the edges of medulloblastoma where tumor cells were invading brain parenchyma (Fig. 5). All of the perivascular lesions were negative for Pten IHC, consistent with cells that have undergone Cre-mediated recombination (Supplementary Fig. S4). Thus, the early perivascular proliferative lesions in the dKO cerebellum were reminiscent of those in PtencKO mice. dKO medulloblastomas emerged from these perivascular niches over time.
dKO medulloblastomas arose from a perivascular tumor stem cell niche
To investigate whether the proliferating perivascular cells gave rise to the nonproliferative NeuN+ medulloblastoma tumor bulk, we traced the fate of proliferative tumor cells 2 hours or 10 days after an EdU pulse in P50 dKO mice. Two hours after an EdU injection, EdU+ cells and NeuN+ cells were mutually exclusive, consistent with nonproliferating NeuN+ tumor cells (Fig. 6A). However, 10 days after an EdU injection, a significant proportion of EdU+ cells were also NeuN+, marking proliferating cells that had differentiated (Fig. 6B; Supplementary Fig. S5A). There were also EdU+ cells that were positive for Ki67 but negative for NeuN. These EdU+ Ki67+ cells were usually less intensely stained for EdU, consistent with dilution of incorporated EdU with multiple rounds of cell division. These data demonstrate that the dKO medulloblastoma has its own lineage hierarchy in which a subpopulation of proliferative tumor cells exit the cell cycle and differentiate along the granule cell lineage, consistent with descriptions of cancer stem cells (25). Strikingly, many, but not all, of the proliferating tumor cells expressed Math1, consistent with SHH subgroup medulloblastoma (Supplementary Fig. S5B).
Pten;Trp53 dKO medulloblastomas acquired somatic inactivation of Ptch1
Tumorigenesis is considered a multistep process involving initiation, promotion, and progression. To investigate which other secondary genetic alterations in addition to the initiating Pten and Trp53 mutations were involved in the medulloblastoma formation, we used aCGH to compare genomic DNA from the tumors and reference genomic DNA from liver or spleen. All of the 18 medulloblastomas showed loss of chromosome 13, on which Ptch1 is located (Fig. 7A). In addition, 5 of the tumors had focal homozygous deletion of the BAC probe encompassing Ptch1 (Supplementary Fig. S6). We also performed RT-PCR and direct sequencing to analyze Ptch1 mRNA and found mutations of Ptch1 in two medulloblastoma samples that had hemizygous deletion of Ptch1 (Fig. 7B). These data support the hypothesis that loss of heterozygosity on chromosome 13 was associated with Ptch1 tumor suppressor loss of function. Seventy-two percent (13/18) and 100% (18/18) of medulloblastomas also showed loss of the entire or part of chromosomes 7 and 16, respectively (Fig. 7A). No recurrent focal homozygous deletion on chromosome 7 or 16 could be identified.
Pten;Trp53 dKO medulloblastomas showed expression signatures of SHH subgroup medulloblastoma
Human medulloblastoma is classified into four molecular groups based on gene expression profiles; SHH, WNT, Group 3, and Group 4 (17).
We compared gene expression profiles of dKO medulloblastomas with the expression signatures of the four subgroups of human medulloblastoma using gene set enrichment analysis (34). Pten;Trp53 dKO medulloblastomas were significantly similar to the SHH subgroup (P = 0.082, FDR = 0.0766) and were not significantly similar to the other three subgroups of medulloblastoma (FDR > 0.25).
In this study, we used an inducible Cre driven by Nestin-regulatory elements to delete Pten in a wide spectrum of progenitor cells in the neonatal cerebellum, when the brain is still undergoing active development. We recapitulated the Pten-deficient phenotypes of neuronal hypertrophy and disrupted cerebellar lamination found in previous studies and confirmed that Pten regulates neuronal size and positioning (12, 13, 36). The more severe phenotypes of cerebellar disorganization and hypertrophy found in the current model compared with previous studies are likely caused by the more potent and ubiquitous Cre activity in neural stem/progenitor cells driven by the Nestin promoter than by the Gfap promoter used previously (12, 13, 36).
Unexpectedly, neonatal deletion of Pten revealed a novel perivascular proliferative progenitor niche in the cerebellum that was not found in previous Pten conditional knockout models. Deletion of Pten driven by the same Nestin-creERT2 induced at P30 did not generate this phenotype, suggesting the presence of an unknown perivascular germinal niche in the developing cerebellum that is normally regulated by Pten but regresses or diminishes upon completion of development. Combined deletion of Pten and Tp53 transformed this proliferative niche into medulloblastomas with expression of markers of terminal granule neuron differentiation. The proliferative niche in Pten-cKO cerebellum shared several features with Nestin-expressing progenitors (NEP) of the EGL (41). The vast majority of granule neurons in the cerebellum are derived from Math1-expressing granule neuron progenitors (GNP). However, 3% to 5% of EGL cells are NEPs, which express Nestin, but not Math1, and are committed to the granule lineage. These cells are quiescent postnatally and normally undetectable by P21; however, they were more efficiently transformed into sonic hedgehog subgroup medulloblastomas than GNPs in response to Ptch1 deletion (41). However, perivascular proliferative lesions in Pten cKO cerebella were frequently also found within the IGL or molecular layers, clearly isolated from the pial surface where EGL progenitors would reside, or from the white matter tracts, where another population of postnatal cerebellar stem cells has been identified (15). Neural stem cells of the SVZ also reside in a vascular niche (42–44). In Pten cKO cerebella, Pten deficiency may have expanded the location of the previously identified population of Nestin+ progenitors of the EGL or may have revealed a novel population of perivascular progenitor cells not yet described.
Our study underlines the importance of Pten regulation in the neural stem/progenitor cells in the cerebellum, indicating that Pten is required for silencing or withdrawing the neural stem cells upon the completion of development under physiologic conditions. In the nervous system, Pten negatively regulates proliferation and self-renewal of embryonic neural stem cells (39) and induces premature differentiation of neuroblasts in the subventricular zone of mice from neonatal ages through adulthood (29). Here, we report the first example of Pten loss resulting in the presence of a normally undetectable ectopic niche with ongoing progenitor cell proliferation in adult mice.
The outcome of Pten deletion varies significantly in the stem cell niche of different tissues. Importantly, the cerebellar perivascular hyperplasia in Pten cKO cerebellum did not show progressive expansion up to 10 months of age and was never found to undergo neoplastic transformation. This is consistent with our finding that Pten deletion is not sufficient to initiate tumorigenesis in neural stem/progenitor cells in the forebrain (29) and is in contrast to Pten function in other stem cells. For instance, deletion of Pten causes short-term expansion, but long-term decline of adult hematopoietic stem cells, and is sufficient to induce eventual emergence of leukemia (40). Intriguingly, Pten deletion has no effect on hematopoietic stem cells in neonatal mice, demonstrating another example of the striking developmental dependence of Pten function (45).
PTEN loss of function or PI3K-activating mutations contribute to multiple molecular groups of human medulloblastoma, and in some cases cooccur with TP53 or PTCH1 mutations (3–6). Proliferation of human medulloblastoma cell lines was decreased by a PI3K inhibitor, and medulloblastomas showed reduced PTEN expression (46). PI3K antagonists also inhibit tumor growth of a mouse model of Group 3 medulloblastoma (47). In the postnatal cerebellum, proliferating cells in Pten cKO mice constituted a premalignant niche that could be transformed to medulloblastoma with complete penetrance when combined with Tp53 deletion. The recurrent somatic loss of Ptch1 in the Pten;Tp53 dKO medulloblastomas is consistent with the cooperative effects of deregulated PI3K signaling and Ptch1 deletion inducing medulloblastoma in other mouse models. Mouse studies have shown cooperative effects of Shh pathway activation, driven by Ptch1 loss or SmoM2 expression, combined with PI3K activation, driven by PI3K or activated Akt expression, or Pten deletion, in generating medulloblastomas (28, 48–50). Combined inhibition of Shh and PI3K pathways synergize to inhibit tumor growth in a mouse model driven by combined deletion of Pten and Ptch1 (51).
Pten;Trp53 dKO medulloblastomas have a very well-defined organization of cancer stem/progenitor cells maintained in a perivascular niche that give rise to nonproliferative bulk tumor cells expressing markers of fully differentiated granule neurons. This spatial organization and extent of differentiation distinguishes them from other mouse medulloblastoma models associated with Ptch1 loss (18, 20, 49, 52–54). Cancer stem/progenitor cells are maintained in a perivascular niche and give rise to nonproliferative bulk tumor cells expressing markers of fully differentiated granule neurons. This is consistent with a previous study showing that perivascular Nestin-expressing tumor cells were responsible for the recurrence of Ptch1 and Pten dKO medulloblastoma after irradiation (28) and that combined Pten deletion and SHH pathway activation induced medulloblastoma with extensive nodularity in which bulk tumor cells showed differentiated morphology (28, 50). This association between Pten loss and differentiation is consistent with the premature differentiation of neuroblasts caused by Pten deletion in the subventricular zone (29).
The contribution of Pten loss to tumor phenotype also varies markedly between brain tumor types. PTEN and TP53 are commonly comutated in human glioblastomas (7), and combined deletion of Pten and Trp53 driven by induction of GFAP-creER in adult brain resulted in high-grade astrocytomas (33). These tumors did not display the abnormal vasculature or the spatially organized hierarchy of perivascular progenitor cells and highly differentiated bulk tumor cells consistently observed in Pten; Trp53 dKO medulloblastomas. Thus, in keeping with the highly context-dependent outcomes of Pten loss (2), there is not a simple association between Pten loss and tumor phenotype, even among different tumor types within the brain.
Mouse models of spontaneous medulloblastoma seem to overrepresent the SHH subgroup of human disease in that the majority of mouse medulloblastomas harbor Ptch1 mutation (20, 49, 53, 55), especially when involving Trp53 mutation. Our Pten and Trp53 dKO medulloblastoma also spontaneously developed Ptch1 mutations as well as other genomic abnormalities. One hundred percent penetrance of medulloblastoma and Ptch1 mutation demonstrates strong cooperation between PI3K, P53, and SHH pathways. Although the subgroup of SHH medulloblastomas are shown to be able to arise from the EGL (18–20, 41), our study indicates a potential alternative origin from Nestin-positive cells within a perivascular niche that may contribute to potential heterogeneity among human SHH medulloblastoma. This niche may give rise to human medulloblastomas of the SHH molecular subgroup that can arise in the midline of the cerebellum, and not on the cerebellar cortical surface where the EGL is located.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Conception and design: G. Zhu, S.J. Baker
Development of methodology: G. Zhu, X. Zhu
Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): G. Zhu, S.L. Rankin, J.D. Larson, L.M.L. Chow, S.J. Baker
Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): G. Zhu, J.D. Larson, C. Qu, J. Zhang, D.W. Ellison, S.J. Baker
Writing, review, and/or revision of the manuscript: G. Zhu, L.M.L. Chow, D.W. Ellison, S.J. Baker
Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): G. Zhu, J.D. Larson, X. Zhu, S.J. Baker
Study supervision: S.J. Baker
We thank Dr. Michael Wang for assistance with array figures, Kristen Cox for genotyping, and the Hartwell Center for Biotechnology for array experiments.
This work was supported by NIH grant P01 CA096832 (S.J. Baker, J. Zhang, and D.W. Ellison), the St. Jude Cancer Center support grant P30 CA021765, and ALSAC of St. Jude Children's Research Hospital.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.