|Home | About | Journals | Submit | Contact Us | Français|
Medulloblastomas that display a large cell/ anaplastic morphology and overexpress the cellular c-MYC gene are highly aggressive and carry a very poor prognosis. This so-called MYC-subgroup differs in its histopathology, gene expression profile, and clinical behavior from other forms of medulloblastoma. We generated a mouse model of MYC-subgroup medulloblastoma by transducing Trp53-null cerebellar progenitor cells with Myc. The cardinal features of these mouse medulloblastomas closely mimic those of human MYC-subgroup tumors and significantly differ from mouse models of the Sonic Hedgehog (SHH)- and WNT-disease subgroups. This mouse model should significantly accelerate understanding and treatment of the most aggressive form of medulloblastoma and infers distinct roles for MYC and MYCN in tumorigenesis.
Medulloblastoma (MB) — the most common malignant pediatric brain tumor — includes at least four clinically and molecularly distinct subgroups (Cho et al., 2011; Kool et al., 2008; Swartling et al., 2010; Thompson et al., 2006). SHH-subgroup MB most frequently results from inactivating mutations of PTCH1 (the SHH receptor) or SUFU (a downstream signal transducer). SHH signaling ultimately activates GLI family transcription factors that up-regulate pro-proliferative genes, such as MYCN, CCND1 and CCND2 (cyclins D1 and D2), and that lead to the reduced expression of inhibitors of cyclin-dependent kinases (CDKs), including p27KIP1 and p18INK4c (Roussel and Hatten, 2011). About 50% of SHH-subgroup MBs exhibit a desmoplastic / nodular histology and carry an intermediate prognosis in patients who receive contemporary surgical intervention and chemotherapy (Cho et al., 2011; Ellison et al., 2011a; Lam et al., 1999; Northcott et al., 2011; Raffel et al., 1997). In contrast, the WNT-subgroup disease has an excellent prognosis, exhibits a “classic” morphology, and is frequently triggered by mutations in the WNT pathway effector CTNNB1 (β-catenin) (Cho et al., 2011; Ellison et al., 2005; Gajjar et al., 2006; Kool et al., 2008; Northcott et al., 2011; Thompson et al., 2006). An interesting distinction between SHH- and WNT-driven MBs is their anatomic location, with SHH tumors arising laterally in the cerebellum and WNT MBs arising in the midline close to the brainstem; recent results indicate that these features reflect the different cells of origin of the two MB subgroups (Gibson et al., 2010).
Modeling both the SHH- and WNT-subgroups of MB in the mouse (Wu et al., 2011) has been instrumental in providing insights into the cellular origins of these different disease forms and in paving the way for therapeutic development (Romer et al., 2004). SHH-subgroup MBs arise within the cerebellum from committed, SHH-dependent granule neuron precursors (GNPs) (Schuller et al., 2008; Yang et al., 2008). Very recently, we demonstrated that WNT-subgroup MBs arise outside of the cerebellum from progenitor cells in the lower rhombic lip (Gibson et al., 2010). Thus, subgroups of MB are likely to reflect intrinsically different diseases with distinct origins and driver mutations.
In contrast to the SHH and WNT subgroups, very little is known about the molecular aberrations that drive two other subgroups of the disease. Non-SHH/WNT tumors include the most aggressive form of the disease (MYC-subgroup) that exhibits frequent amplification and/or overexpression of MYC, portends a dismal prognosis, and generates a high proportion of aggressive and invasive tumors with large cell/ anaplastic (LC/A) histology (Cho et al., 2011; Ellison et al., 2011a; Northcott et al., 2011; Pfister et al., 2009)
Mycn is a critical mediator of SHH signals in GNPs (Kenney et al., 2003) and is absolutely required for normal cerebellar development; however, much less is known about the function of Myc in the mouse hindbrain (Knoepfler et al., 2002; Zindy et al., 2006). Myc is not normally expressed in GNPs (Zindy et al., 2006), and overexpression of MYC and MYCN is mutually exclusive and associated with distinct subgroups of human MBs (Cho et al., 2011; Northcott et al., 2011). High-level expression and amplification of MYCN are observed across the various subgroups of human MB. Aberrant activation of Mycn expression in the developing mouse cerebellum initiates a variety of MBs including both classic and LC/A tumors (Swartling et al., 2010). In contrast, the highest levels of MYC expression and MYC amplification are found almost exclusively in the aggressive MYC-subgroup disease (Cho et al., 2011; Northcott et al., 2011). Thus, while MYCN may play a role in the pathogenesis of a variety of MBs, MYC may drive a specific aggressive subgroup of the disease. This may seem somewhat counter-intuitive, since it is widely thought that the biochemical transcriptional functions of different MYC-family genes are similar.
Here, we assessed the role of MYC and MYCN in medulloblastoma development in the absence of TRP53.
We showed previously that GNP-enriched cell isolates from the cerebella of postnatal day (P) 6–7 Cdkn2c−/−, Ptch1+/− or Cdkn2c−/−, Trp53−/− mice, but not from Cdkn2c−/− or wild type mice, generate SHH-subgroup MBs when transduced with a retroviral vector expressing Mycn but not a control virus (Zindy et al., 2007). To test if Myc might similarly transform Cdkn2c−/−, Trp53−/− GNPs, we isolated proliferating GNPs from Cdkn2c−/−, Trp53−/−, Atoh1-GFP mice, which are marked by co-expression of green fluorescent protein (GFP) (Lumpkin et al., 2003). Enrichment of GNPs showed that on average, we obtained 91.9% of GFP-positive (+) GNPs and 8.1% of GFP-negative (−) progenitor cells per preparation and found that the sorted GFP-expressing population contained 1.1 % of GFP− cells and, conversely, the GFP− population contained 1.7 % of GFP+ cells. We transduced these cells with viruses either encoding Myc and co-expressing red fluorescent protein (RFP) or expressing Mycn in lieu of Myc. Fluorescence-activated cell sorting (FACS) of Myc- and Mycn-transduced cells confirmed comparable infection efficiencies by the two retroviral vectors (45.9 ± 11.7% of GFP+/RFP+ for Myc-RFP and 51.6 ± 2.1% of GFP+/RFP+ for Mycn-RFP). Cells transduced with either Myc or Mycn (2 × 106 per mouse) were injected separately into the cerebral cortices of naïve recipient CD-1 nu/nu mice. Myc-transduced cells formed aggressive tumors that killed mice significantly faster than GNPs transduced with Mycn (median survival = 33 days for Myc versus 48 days for Mycn, p< 0.0001, Figure 1A). Immunoblotting demonstrated significant levels of ectopic Myc or Mycn protein expression within the two tumor subsets (Figures S1A and S1B). Myc-derived tumors generated from P6–7 cerebellar cells of either Cdkn2c−/−, Trp53−/− or Trp53−/− mice occurred with similar latency (median survival = 39 days for Cdkn2c−/−, Trp53−/− versus 39 days for Trp53−/−, p= 0.7096), indicating that loss of Cdkn2c was not required for Myc expression to induce MB in the absence of Trp53 (Figure S1C). Myc-tumors displayed a consistent morphology that was strikingly similar to human MBs of the MYC-subgroup (Figure 1B). Morphometric and TUNEL assays of mouse MBs revealed a much larger cell size and apoptotic rate in Myc-tumors than mouse models of the WNT (Gibson et al., 2010) or SHH-subgroups disease (Ptch-tumors, Uziel et al., 2005) that typically show a classic morphology (Figures 1B and 1C). Thus Myc-induced mouse MBs resemble the human LC/A MB phenotype reported previously (McManamy et al., 2003), and Myc and Mycn drive distinct tumors that appear to recapitulate aggressive LC/A and classic forms of human MB, respectively.
The decreased latency and LC/A morphology of Myc-generated tumors suggested that these were distinct from classic MBs induced by Mycn. We compared gene expression profiles of Myc-tumors with those of Mycn-tumors, as well as profiles generated from previously characterized mouse models of WNT- and SHH-subgroup MB (Gibson et al., 2010; Uziel et al., 2005). In addition, we compared these tumor profiles with those of FACS-sorted Atoh1-GFP-expressing GNPs obtained from the cerebellum of normal P6 mice lacking both Trp53 and p18Ink4c protein expression [Cdkn2c−/−, Trp53−/−, Atoh1-GFP] mice (designated GNPs) (Figure 2A). The transcriptome of Myc-tumors was distinct from those of the other mouse MBs (Figure 2A). Unsupervised hierarchical clustering co-segregated the gene expression profiles of mouse Mycn-tumors and with those of mouse models of SHH-subgroup disease. In contrast, transcriptomes of Myc-tumors and the mouse WNT-subgroup model formed two separate clusters. The Mycn/SHH/GNP profiles were characterized by high expression of known members of the SHH pathway and signature genes of SHH-subgroup MB (Figure 2B). This observation was confirmed by quantitative RT-PCR (qRT-PCR) including analysis of Atoh1, Gli1, Sfrp1, and Boc1 (Figures 2C and 2D). In contrast, Myc and Npr3, that are specifically expressed in the human MYC-subgroup (Northcott et al., 2011), were highly expressed in the mouse Myc-derived MBs (Figure 2B). Immunohistochemical analysis confirmed that Npr3 is expressed selectively in mouse MYC-subgroup MBs (Figure S2A). Myc-induced tumors also exhibited high levels of transcription of Prom1 and Lgr5 (Figure 2B), which are frequently expressed in stem cell-like progenitor cells (Barker et al., 2007; Lee et al., 2005; Zhu et al., 2009). In addition, high expression of Nanog and Oct4 proteins, considered to be canonical markers of embryonic pluripotency (Silva and Smith, 2008), was observed in mouse MYC-subgroup and SHH-subgroup MBs (Figure S2B–P). On the other hand, putative markers of brain tumor stem cells (Rich, 2009) exhibited various expression levels among the three distinct tumor subgroups (Figure S2Q).
As a further test of the degree to which mouse Myc-derived tumors resemble the MYC-subgroup of human MB, we compared the expression of mouse genes with orthologs previously shown to specifically distinguish the human WNT-, SHH-, and MYC-subgroup tumors (Cho et al., 2011; Northcott et al., 2011; Thompson et al., 2006). 31 of 52 orthologs (60%) that exhibited increased expression in the human MYC-subgroup were similarly up-regulated in the mouse Myc-tumors (Figure 3A). 40 of 54 orthologs (74%) with increased expression in the human WNT-subgroup were similarly increased in the mouse WNT-tumors (Figure 3B). 31 of 53 orthologs (58%) up-regulated in the human SHH subgroup were increased in the mouse Ptch-tumors (Figure 3C). Thus, mouse Myc-induced tumors express signature genes remarkably similar to those reported for human MYC-subgroup MB, while the mouse Mycn, Ptch and Trp53 tumors resemble the human SHH-subgroup, and WNT mouse tumors recapitulate human WNT-subgroup MBs, as reported previously (Gibson et al., 2010). Furthermore, cross-species comparison of human and mouse MB transcriptomes (Johnson et al., 2010), revealed a statistically significant match between mouse and human MYC-subgroups (Figure S3A). Overall, 56% of 14,261 ortholog probe pairs showed agreement in gene expression (up-regulation or down-regulation) between the human MYC-subgroup and the mouse Myc-tumors [p=0.009 and p=0.079 by permutation of the human and mouse data, respectively (Figure S3B)]. Together, these data confirm that the mouse Myc-tumors accurately model the transcriptome of human MYC-subgroup MB.
All MBs generated from neuronal progenitors purified from P7 cerebella of Cdkn2c−/−, Trp53−/−, Atoh1-GFP mice and infected with Myc-encoding retroviruses before cortical injection expressed vector-encoded RFP (Figure 1 and Figure S1D). Mycn tumors consisted mainly of GFP+/RFP+ cells (70.0 ± 8.20%, n=7) (Figure S1E). Surprisingly, Myc-induced tumors expressed little or no GFP (17.2 ± 3.67%, n=14) (Figure S1E), suggesting that these tumors arise from a small fraction (~5%) of GFP− cells in the GNP-enriched isolates or from GFP+ GNP progenitors that subsequently silence Atoh1-GFP expression during tumor formation.
To more rigorously characterize the source of mouse MYC-MBs, we further profiled our cell isolates. Comparison of gene expression profiles between GFP+, GFP− populations and Myc-engineered MBs revealed that MYC-subgroup medulloblastoma most closely matched that of GFP-negative cells (Figure S4A); however, not all genes, including Lgr5 and Npr3, were expressed in the GFP-negative sorted cell population (Figure S4B). As a first step to understand which cell population might generate MYC-subgroup MBs, we FACS-sorted GFP-positive and GFP-negative progenitor cells. These separate fractions were then transduced with Myc-encoding retroviruses at an efficiency of ~30%, and 5 × 104 cells from each transduced fraction were then separately implanted into the cortices of recipient mice. Both Myc-transduced population generated tumors in the cortex of naïve recipient animals (Table 1). Histopathological analysis of tumors revealed that all medulloblastomas derived from GFP+ FACS sorted cells infected with Myc-encoding viruses (5/8) showed LC/A characteristics (Table 1 and Figure S4C). In contrast, tumors occurring after transplants of GFP-negative cells infected with Myc-expressing retroviruses included two T cell lymphomas (2/8) as well as a range of CNS embryonal tumors – high-grade neuroepithelial tumor with dominant PNET and focal glial phenotype (3/8) and MBs with LC/A features (3/8) (Table 1). All LC/A MBs expressed low levels of Atoh1 when compared to that of normal GNPs percoll-purified from the cerebella of P7 wild type mice (Figure S4D). Comparative gene expression analysis showed that the MYC-subgroup MBs derived from Myc-transduced FACS sorted GFP-positive and GFP-negative cerebellar cell populations had similar gene profile as Myc-tumors derived from unsorted cerebellar cells (Figure 2A). These observations suggest that both GFP+ and GFP− populations contain cells that can form MYC-subgroup MBs.
Human MYC-type MBs are the most aggressive of all subgroups (Cho et al., 2011; Northcott et al., 2011), and consistent with clinical data, mouse Myc-induced tumors developed faster than Mycn-MBs (Figure 1A). One possible feature of MYC tumors that could explain their aggressiveness is that MYC-subgroup MBs contain a greater fraction of tumor-propagating cells than do tumors of other subgroups. To estimate the numbers of cells capable of initiating MBs, cells were purified from mouse MBs of the SHH- and MYC-subgroup, and different numbers of tumor cells were injected into the cortices of recipient nude mice (Figure 4A). Limited dilution experiments revealed that equal to more than 2×105 purified SHH-subgroup tumor cells were required for secondary tumor formation (Figure S5), consistent with previous reports (Read et al., 2009). Unlike SHH-subgroup MB cells, only 1 × 102 Myc-induced tumor cells were required to generate secondary tumors (5/5) (Figure 4A). Histological analysis confirmed that secondary tumors shared similar immunohistological characteristics with the parental primary tumor (Figure 4, B–K).
CD133/Prom1-positive tumor-propagating cells from human MBs expand in vitro to form “neurosphere” colonies when plated under conditions that prevent their attachment to the culture dish (Singh et al., 2004). Given that mouse MYC-subgroup tumors up-regulated Prom1, we tested whether these tumor cells could similarly form neurospheres. Purified cells from Myc-tumors were plated on an ultra low-attachment dish at a density of 5 × 104 cells/ml in culture medium supplemented at 3-day intervals with basic FGF and EGF. One week later, spheres were harvested, dissociated with trypsin, and the total number of cells per plate was enumerated. Dissociated cells were replated at the same initial density, cultured under the same conditions, and sequentially passaged multiple times. Six to seven days after their first plating, single RFP-positive Myc-derived tumor cells generated macroscopic red colonies (Figure 5A), the overall numbers of which had increased more than 20-fold (Figure 5B, passage 1). Thus, average, sphere-forming cells underwent 5–6 population doublings in the first seven day period. More cells were recovered at passage 2 and 3, after which the doubling time of the population slowed down; however a consistent proliferative rate approximately equal to 4 population doublings in succeeding 7 day intervals was maintained from passage 5 to 10 (Figure 5B). In stark contrast, spheres derived from canonical SHH-subgroup MBs that arose in Cdkn2c−/−, Ptch1+/− mice could not be serially passaged at all under the same culture conditions (Figure S6). Immunohistochemical analysis (Figure 5C) and qRT-PCR (Figure 5D) revealed that neurospheres from Myc-induced MBs expressed several markers identified in stem/progenitor cells, including Nanog, Oct4, Sox2, and Lgr5, as well as Nestin and Npr3 (Figure 5C).
To examine whether tumor spheres could be transplanted and would recapitulate primary Myc-induced MBs, we injected spheres at passage 2 and passage 6 into the cortices of recipient mice. Interestingly, 2 × 105 cells formed tumors with a similar latency as the secondary MBs generated from cells purified from Myc-engineered tumors (7/7, median latency = 22 days) (Figure 6A). Primary and secondary tumors shared the same pathology and expressed similar immunohistochemical markers (Figure 6B). Gene profiling of primary and secondary MBs confirmed that the tumors were distinct from the SHH-subgroup tumors used as a control (Figure 6C). Thus, tumor spheres recapitulate MYC- subgroup MBs after transplantation.
Smoothened inhibitors, including cyclopamine and HhAntag inhibit the proliferation, and induce the differentiation of SHH-subgroup MB in vitro and in vivo (Berman et al., 2002; Romer et al., 2004). Similarly, BMP4 antagonizes SHH signaling, induces the neuronal differentiation of GNPs (Rios et al., 2004), extinguishes protein Atoh1 expression and inhibits the proliferation of MBs of the SHH-subgroup (Zhao et al., 2008). Because human MYC-subgroup MBs lack an active SHH gene expression signature and do not contain activating mutations in SHH pathway genes, we reasoned that mouse MYC-subgroup tumors might resist Smoothened antagonists. Purified GNPs from cerebella of P7 wild type mice, from MYC-subgroup MB, and from a SHH-subgroup MB derived from Cdkn2c−/−, Trp53Fl/Fl, Nestin-Cre mice were cultured for 3 days either in the presence of cyclopamine, BMP4 or a vehicle control. Cells were then labeled with BrdU and analyzed by FACS with an antibody to BrdU. Cyclopamine or BMP4 reduced the S phase fraction of Cdkn2c−/−, Trp53Fl/Fl, Nestin-Cre “primary” tumor cells by 35 to 85% (Figure 7A). In contrast, tumor cells isolated from three independently derived mouse MYC-subgroup tumors were insensitive to cyclopamine or BMP4 treatment (Figure 7A), consistent with the finding that MYC-tumors were associated with low expression of SHH signature genes (Figure 7B). Similarly, we saw no inhibition of proliferation of neurospheres from MYC-tumors plated in the constant presence of SHH signaling inhibitors in the culture medium for two weeks (Figure 7C).
Enforced expression of Myc, but not Mycn, in concert with the loss of Trp53 in GNP-enriched mouse cerebellar progenitor cells gives rise to tumors that recapitulate the most aggressive form of human MB. These tumors exhibited neither of the characteristic gene expression signatures previously ascribed to the SHH- or the WNT-subgroups of the disease (Cho et al., 2011; Northcott et al., 2011; Thompson et al., 2006). Instead, multi-tiered analysis of data comparing the murine tumors with the human MYC-subgroup revealed convincing similarities in histology, clinical behavior, and gene expression between the tumors of both species.
Marker proteins including Prom1, Lgr5, Oct4 and Nanog, found either in embryonic or adult tissue stem cells, as well as in some “cancer stem cells”, were highly expressed as part of the defining gene expression signature of Myc-engineered tumors. Prom1 is a marker of tumorinitiating cells in many cancers, (Curley et al., 2009; Reya et al., 2001; Singh et al., 2004; Todaro et al., 2010), and Lgr5 is expressed in mitotically dividing stem cells within the colon and intestinal crypts, but not in their proliferating transient amplifying progeny (Barker et al., 2007; Zhu et al., 2009). Oct4 and Nanog, are considered markers of pluripotency, and although distinct from the “MYC expression module” (Kim et al., 2010), can collaborate with MYC in reprogramming somatic cells to induced pluripotent stem (iPS) cells (Takahashi and Yamanaka, 2006). Like cells with stem cell characteristics whose frequencies correlate with poor prognosis in other cancers (Ben-Porath et al., 2008), cells from mouse MYC-subgroup MBs could be sequentially and continuously propagated as cultured neurospheres through many ex vivo passages. These cells retained tumor-propagating potential after transplantation into the cortices of recipient mice and re-induced MBs with the same robust efficiency and defining cardinal features as the primary MYC-subgroup MBs from which they were derived. In contrast, tumor cells explanted from MBs of the SHH-subgroup tend to undergo spontaneous differentiation in culture, and neurospheres generated from these tumors rapidly lose their self-renewal capacity when sequentially passaged (Read et al., 2009; Figure S5). Given that standard front line therapies fail in children with MYC-subgroup MBs, and that such tumors in the mouse are unaffected by SHH inhibitors now being incorporated into human clinical trials, the ability to maintain tumor-propagating cells in culture from mouse MYC-subgroup MBs may prove useful in establishing a platform for identifying therapeutic drugs.
The generation of an entirely unique subgroup of MB in the mouse after Myc transduction into Trp53-deficient GNPs (irrespective of Cdkn2c loss) was unexpected. In fact, upon embarking on these experiments, a reasonable hypothesis might have been that ectopically enforced overexpression of Myc would have had the same effect as overexpression of Mycn, given that Myc-family proteins bind to the same canonical DNA consensus sequences (Grandori and Eisenman, 1997) and interact with similar dimerization partners, co-activators, and co-repressors (Blackwood and Eisenman, 1991; Grandori et al., 2000). Indeed, despite the fact that Mycn and Myc genes are differentially expressed in the hindbrain (Zindy et al., 2006), and that Mycn, but not Myc, is a target of SHH signaling (Kenney et al., 2003), genetic experiments showed that Mycn can functionally replace Myc in mouse development, proliferation and differentiation (Malynn et al., 2000) implying many interchangeable functions.
The gene expression pattern of the human SHH-subgroup MB resembles that of GNP cells (Lee et al., 2003), and that of the WNT-subgroup MB resembles cells derived from the dorsal brainstem (DBS) (Gibson et al., 2010), but the gene expression pattern of the MYC-subgroup of MB, either in mouse or humans, is quite distinct. This suggests that the latter tumors might arise from a class of MYC-responsive progenitor cells that differ from those that give rise to the other MB subgroups. Notably, although previous models of mouse MB generated by targeting cerebellar GNPs invariably yielded tumors of the SHH-subgroup, regardless of many different genetic perturbations used to initiate tumorigenesis (Wu et al, 2011), the more recent derivation of a WNT-subgroup mouse model stemmed from observations that a different group of progenitor cells expressed in the dorsal brain stem, and not cerebellar GNPs, were the most sensitive to constitutive activation of the WNT signaling pathway (Gibson et al, 2010). Therefore, one possibility is that target cells most sensitive to Myc overexpression had contaminated the highly purified Atoh1-GFP-expressing GNP population into which the Myc-RFP vector was introduced. Alternatively, enforced Myc expression in the context of Trp53 loss may have significantly altered the transcriptional program of GNPs, resulting in their trans-differentiation, loss of canonical GNP markers, and the emergence of distinctly different phenotypic features. Either scenario would account for the observation that RFP-expressing MYC-subgroup MBs no longer expressed Atoh1-GFP. By analogy to the strategy for modeling WNT-subgroup MBs (Gibson et al. 2010), the identification of the cell of origin of MYC-subgroup tumors will likely require the systematic generation of genetically-engineered animals in which Myc expression is conditionally regulated within different cell lineages.
Recent studies have more clearly identified the subgroup of human MBs that feature high MYC expression and/or amplification and that carry a dismal prognosis (Cho et al., 2011; Ellison et al., 2011b; Northcott et al., 2011). Both amplification and overexpression of MYC or MYCN have been associated with poor prognosis in human MB (Pfister et al., 2009). However, in a large cohort of children studied in the SIOP/PNET3 clinical trial, Ellison et al. (2011b) recently concluded that amplification of MYC rather than MYCN is associated with the poorest outcome. In agreement with the latter findings, Myc-induced mouse MBs were more anaplastic and aggressive than their Mycn-induced counterparts, contained a several log-fold higher fraction of tumor-propagating cells, and initiated tumors after a significantly shorter latency period.
While our finding that Mycn overexpression results in MBs of the SHH-subgroup seems contradictory to the finding published in a recent study by Swartling et al (2010), which ascribes Mycn overexpression to the production of a variety of MBs, we feel that there are distinctions in the two models that may explain this discrepancy. The Swartling model links Mycn to a Glt-1 promoter. Glt-1 is a gene that is not widely expressed by cells in the EGL of the cerebellum and it is likely that, via Glt-1, Mycn is influencing a developing cerebellar cell pool that is distinct from that of the EGL. On the contrary, in our study Mycn is transfected into a highly concentrated pool of GNP cells. It stands to reason that the introduction of a SHH pathway target member like Mycn into GNP/EGL cells, which are particularly sensitive to SHH stimulated growth, may favor the induction of SHH-subgroup MB. On the other hand, MYC, probably because it is not a direct target of the SHH pathway, appears to have a very unique and specific effect when transduced into a similar pool of cerebellar precursor cells.
Mutations of the TP53 gene are found in human LC/A MBs together with MYC amplification (Aldosari et al., 2002; Frank et al., 2004; Pfister et al., 2009). Although TP53 is frequently disrupted in LC/A MBs, its loss of function is observed only infrequently across other subgroups (Pfaff et al., 2010; Thompson et al., 2006). Under the conditions used in our mouse modeling experiments, concurrent deregulation of both Myc and Trp53 was required to induce MYC-subgroup tumors, whereas neither was effective alone. The genetic interaction between MYC and TP53 is known to be important in many different cancers. While Trp53 represses Myc expression transcriptionally, miss-regulation of Myc abrogates Trp53-mediated cell cycle arrest through the repression of inhibitors of CDKs (Hoffman and Liebermann, 2008). Overexpression of Myc also induces Trp53-dependent apoptosis (Hermeking and Eick, 1994; Wagner et al., 1994), but when accompanied by either Trp53 mutation or bi-allelic Arf deletion, Myc can readily generate immortalized tumor cells (Eischen et al., 1999; Hemann et al., 2005; Zindy et al., 1998).
Sub-lethal ionizing irradiation of Trp53-null mice at P5–P7 is sufficient to induce MBs of the SHH-subgroup with very high penetrance, implying that in this setting, the primary function of Trp53 is to eliminate rapidly proliferating GNPs that have sustained DNA damage (Uziel et al., 2005). Inactivation of Trp53 similarly accelerates MB formation in Ptch1 heterozygous mice, a context in which Arf deletion has no such effect (Wetmore et al., 2001). Although Arf is induced by high signaling thresholds conveyed by constitutively activated oncogenes, including Myc (Zindy et al., 1998), Arf is not induced by acute DNA damage (Kamijo et al., 1997). Moreover, Atoh1-expressing, proliferating GNPs normally express relatively high levels of Bmi1, a polycomb protein suppressor of the Ink4a/Arf locus (Bruggeman et al., 2005). If the cell of origin of Myc-induced tumors is not an Atoh1-expressing GNP, this begs the question of whether Arf-null mice, like those lacking Trp53, might be predisposed to MYC-subgroup MB formation. Conceivably, other mutations in the Trp53 signaling network might also substitute for inactivation of Trp53 itself. Next generation sequencing and comprehensive analysis of the methylome of primary MYC-subgroup human MBs should shed light on this issue.
In summary, the generation of a mouse model of MYC-subgroup MB is of particular importance, because it mimics the most aggressive subgroup of human MBs that remain the least responsive to therapy (Ellison et al., 2011b). The model provides an opportunity to further explore the identity of the progenitor cells from which these tumors arise and to screen for molecules that may offer improved therapeutic impact.
A detailed description of the Experimental Procedures utilized in this work can be found in the available online Supplemental Experimental Procedures
Cdkn2c−/−, Trp53−/−, Atoh1-GFP mice were generated by breeding Cdkn2c−/−; Trp53−/− animals (Uziel et al., 2005) with Atoh1-GFP transgenic mice (Lumpkin et al., 2003). Other mice used in this study were: Cdkn2c−/−; Ptch1+/− (Uziel et al., 2005), Cdkn2c−/−; Trp53Fl/Fl; Nestin-Cre, and CTNNB1+/lox (ex3); BLBP-Cre; Trp53Fl/Fl (Gibson et al., 2010). All animal work was performed under established guidelines and supervision by the St. Jude Children's Research Hospital's Institutional Animal Care and Use Committee (IACUC), as required by the United States Animal Welfare Act and NIH policy to ensure proper care and use of laboratory animals for research.
Purification of GNPs and other progenitor populations, retrovirus production, infections and orthotopic transplants were performed as previously described (Ayrault et al., 2010). Mouse stem cell virus (MSCV)-based retroviruses encode the red fluorescent protein (RFP) in cis expressed from an internal ribosomal entry site (IRES). These viruses either express RFP alone or co-express mouse Myc or Mycn cDNA inserted downstream of the MSCV-LTR. Infection efficiency was analyzed using FACS for RFP and GFP expression. In some experiments, two days after plating, infected GNPs were harvested, and transplanted into the cerebella or cortices of CD-1 nu/nu mice (Charles River Laboratories, Wilmington, MA). In other experiments, GNP-enriched cerebellar cells were sorted by FACS based on GFP expression immediately after purification. GFP+ and GFP− populations were independently infected with Myc-carrying retroviruses, and 5 × 104 infected cells from each population were transplanted into the cortices of mice 2 days after infection.
Transplantation of infected GNP-enriched cerebellar cells into the cortices or cerebellum of nude recipient mice was performed essentially as described previously (Ayrault et al., 2010). After transplant of virus-infected progenitor cells, mice were examined daily for symptoms of sickness (doming of the head, ataxia or reduced activity). In some instances, purified MB cells or cultured tumor sphere cells from MYC-tumors were injected back into the cortices of naive nu/nu recipient animals.
We analyzed progenitors purified from P7 wild type mice and tumor cells from three independently derived Myc-engineered MBs and MBs arising spontaneously in Cdkn2c−/−; Trp53Fl/Fl; Nestin-Cre mice. Purified GNPs and tumor cells were plated at 8 × 105 cells/well on a Matrigel-coated 24-well plate and grown as previously described (Zhao et al., 2008). GNPs were treated with SHH whereas tumor cells were not. Both cell populations were cultured in the presence or absence of SHH signaling inhibitors, 2.5 μM cyclopamine (LC Laboratories) or 100 ng/ml BMP4 (R&D Systems). The medium was changed every 24 hrs. Three days after initiation of culture, BrdU was added to the culture medium at a final concentration of 10 μM, and cells were harvested 2.5 hrs later. Cells that incorporated BrdU were stained with an anti-BrdU antibody using a BrdU-APC flow kit (BD bioscience), and analyzed by FACS.
For histopathology, samples of murine MBs (3 separate tumors for each genotype) were formalin-fixed and paraffin-embedded, and sectioned at 5μm thickness. For each sample, a section was stained using a standard H&E protocol; a second section was stained for apoptotic cells, using the ApopTag kit (Millipore) with peroxidase detection of TUNEL labeling. Representative images of each sample/stain combination were captured (40× original magnification, under oil immersion) and analyzed using Axiovision software (Carl Zeiss Microscopy). A single observer scored each tumor for apoptotic index (percentage of positive cells following ApopTag labeling) and for nuclear area as previously described (McManamy et al., 2003).
For immunoblotting, purified GNPs from the cerebellum of P7 wild type mice or Myc- and Mycn-engineered MB cells were lysed and proteins subjected to immunoblotting as described previously (Zindy et al., 2007)
Procedures and antibodies used for immunohistochemistry and antibodies for immunoblotting are provided in Supplemental information.
RNA from GNPs or tumor cells were subjected to hybridization using Affymetrix Mouse Genechips 430 version 2 (e.g. Figures 2 and and3).3). For comparative gene expression analysis between primary and secondary Myc-engineered tumors, we used Affymetrix Mouse Genechips HT430PM (e.g. Figure 7C). Microarray results were validated by qRT-PCR using PCR primers shown in Supplemental information.
Tumor cells were cultured under conditions described previously (Taylor et al., 2005). Briefly, cells from Myc-engineered tumors were purified and plated on an ultra low-attachment dish at 5 × 104 cells/ml or 1 × 105 cells/well per 6-well plate. Human recombinant basic FGF and EGF (Peprotech) were added into the culture medium every three days. One week later, tumor spheres were harvested, pooled, and dissociated with trypsin (Invitrogen), and total cell numbers were determined. In parallel, 1 × 105 of the dissociated cells were plated again and cultured under the same conditions. Total RNA was extracted from the reminder of the cultured cells followed by qRT-PCR analysis of selected RNAs. To test the role of SHH-antagonists, Myc-engineered tumor cells were cultured in the presence of 2.5 μM cyclopamine or 100 ng/ml BMP4.
MYC-subgroup medulloblastoma (MB) is one of the most aggressive pediatric brain tumors. This disease is resistant to combination surgery, radiotherapy, and chemotherapy and kills most affected children within three years of diagnosis. Mouse models of the SHH and WNT forms of MB have advanced understanding of the biology and treatment of these disease subgroups. In contrast, the absence of preclinical models of MYC-driven MB has limited understanding of this important tumor. Here we describe the MYC-driven mouse model of MB that accurately mimics the transcriptome, histopathology, and clinical behavior of human MYC-subgroup disease. This model should significantly advance efforts to develop therapeutic modalities and to determine the origin of this deadly childhood cancer.
We thank Dr. Robert Eisenman for helpful suggestions, Dr. Charles J. Sherr for editing of the manuscript, Dr. Frederique Zindy for mice, Shelly Wilkerson, Sarah Gayso and Jennifer Craig for excellent technical assistance, John Morris for Affymetrix microarrays, Drs. Richard Ashmun and Ann-Marie Hamilton-Easton for FACS analysis, Dr. Chris Calabrese, John Killmar and Melissa Johnson for orthotopic transplants, Pamela Johnson and Dorothy Bush for IHC of tumor tissues. This work was funded in part by NIH grant CA-096832 (MFR) and CA-21765 (RJG and MFR), the American Brain Tumor Association (TU), the Mochida Foundation (DK), the Anderson fellowship (DK), and the American Lebanese Syrian Associated Charities of St. Jude Children's Research Hospital.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Accession number Affymetrix data for mouse MBs using 430V2 and HT430PM chips can be found in the NCBI database numbers GSE33199 and GSE33200, respectively.