Bmi1 Is Required for Hedgehog Pathway-Driven Medulloblastoma Expansion^,

Abstract

Inappropriate Hedgehog (Hh) signaling underlies development of a subset of medulloblastomas, and tumors with elevated HH signaling activity express the stem cell self-renewal gene BMI1. To test whether Bmi1 is required for Hh-driven medulloblastoma development, we varied Bmi1 gene dosage in transgenic mice expressing an oncogenic Hh effector, SmoA1, driven by a glial fibrillary acidic protein (GFAP) promoter. Whereas 100% of SmoA1; Bmi1^+/+ or SmoA1;Bmi1^+/- mice examined between postnatal (P) days 14 and 26 had typical medulloblastomas (N = 29), tumors were not detected in any of the SmoA1;Bmi1^-/- animals examined (N = 6). Instead, small ectopic collections of cells were present in the region of greatest tumor load in SmoA1 animals, suggesting that medulloblastomas were initiated but failed to undergo expansion into frank tumors. Cells within these Bmi1^-/- lesions expressed SmoA1 but were largely nonproliferative, in contrast to cells in Bmi1^+/+ tumors (6.2% vs 81.9% PCNA-positive, respectively). Ectopic cells were negative for the progenitor marker nestin, strongly GFAP-positive, and highly apoptotic, relative to Bmi1^+/+ tumor cells (29.6% vs 6.3% TUNEL-positive). The alterations in proliferation and apoptosis in SmoA1;Bmi1^-/- ectopic cells are associated with reduced levels of Cyclin D1 and elevated expression of cyclin-dependent kinase inhibitor p19^Arf, two inversely regulated downstream targets of Bmi1. These data provide the first demonstration that Bmi1 is required for spontaneous de novo development of a solid tumor arising in the brain, suggest a crucial role for Bmi1-dependent, nestin-expressing progenitor cells in medulloblastoma expansion, and implicate Bmi1 as a key factor required for Hh pathway-driven tumorigenesis.

Introduction

Medulloblastoma, a primitive neuroectodermal tumor of the cerebellum, is the most common pediatric brain tumor, representing approximately 20% of newly diagnosed CNS malignancies in children [1]. Oncogenic mutations in Hedgehog (Hh) pathway components have been identified in up to 20% to 25% of human medulloblastomas [2], and aberrant activation of the Hh pathway, measured by expression of HH target genes, has been detected in a significantly higher percentage of cases [3,4]. In normal cerebellum, the Hh pathway regulates proliferation of granule neuron precursors (CGNPs), which are potential medulloblastoma progenitors that comprise the transient external granular layer (EGL) of the cerebellum [5]. During early postnatal development, CGNPs become refractory to the Shh signal for proliferation, withdraw from the cell cycle, differentiate, and migrate through the underlying molecular layer to generate the internal granular layer of neurons [6]. Oncogenic mutations leading to sustained Hh signaling in CGNPs may override the usual signals regulating the fate of these cells, resulting in their continued proliferation in medulloblastoma.

Bmi1 is a member of the Polycomb group of transcriptional repressor proteins and regulates stem cell self-renewal, partially by repressing the senescence and apoptosis-related genes p16^Ink4a and p19^Arf [7]. In developing cerebellum, Bmi1 seems to be acting downstream of the Hh pathway to control proliferation of CGNPs [4]. Thus, Bmi1-null mice develop smaller cerebella secondary to impaired production of granular neurons; however, overall cerebellar organization and cell specification and differentiation seem relatively normal, although the number of stellate cells in the molecular layer is reduced [4,8,9]. Bmi1 has also been implicated in Hh pathway-mediated self-renewal of mammary stem cells and cancer stem cells [10]. BMI1 is aberrantly expressed in a number of solid tumor types, including medulloblastoma [4], where BMI1 expression correlates with the activation of the Hh pathway. Several studies suggest a role for Bmi1 in the maintenance of cultured or grafted tumor cells [11–13], and a recent report established a requirement for Bmi1 in K-ras-driven lung tumorigenesis in mice [14]. However, a requirement for Bmi1 in the spontaneous development of medulloblastomas or other solid tumors arising in the central nervous system has not been demonstrated. In this work, we investigated the role of Bmi1 in the pathogenesis of Hh-driven medulloblastoma, using a novel, fully penetrant mouse model expressing the oncogenic Hh effector SmoA1 [3,15].

Materials and Methods

Generation of Transgenic Animals, Breeding, and Genotyping

We generated a double-transgenic mouse model of medulloblastoma using an HA-tagged, oncogenic allele of Smoothened, SmoA1 [3,15], under the control of the tetracycline responsive element (TRE). Before beginning experimental crosses, TRE-SmoA1 mice were backcrossed to C57/BL6 mice for at least five generations. When these mice were crossed with GFAP-tTA mice [16] to produce GFAP-tTA;TRE-SmoA1-HA bitransgenic mice (designated SmoA1 for the sake of simplicity), SmoA1 was expressed in glial and progenitor cells throughout the CNS and led to rapid development of medulloblastoma (Supplementary Note and Figures W1–W4). Details on generation of TRE-SmoA1HA mouse lines, breeding with GFAP-tTA and Bmi1 mutant mice, genotyping, and housing, are described in Supplementary Materials and Methods and Figure W5. Both GFAP-tTA and Bmi1 mutant mice were maintained on a C57/BL6 background.

Immunohistochemistry, In Situ Hybridization, and X-gal Staining

For immunohistochemistry and in situ hybridization, tissue was fixed overnight in 10% neutral-buffered formalin and transferred to 70% ethanol before embedding in paraffin for sectioning. Post-fixation processing and paraffin embedding were performed by Histoserv, Inc. (Bethesda, MD). To definitively confirm the absence of tumor in SmoA1;Bmi1^-/- mice, serial coronal sections were cut through the entire rostral-caudal length of the cerebella. All sections were cut at 5 µm thickness. Before immunostaining for HA, NeuN, GFAP, Nestin, or PCNA, tissues were subjected to antigen retrieval by boiling in sodium citrate buffer for 10 minutes. The antibodies used comprised HA (3F10, 1:100; Roche, Indianapolis, IN), NeuN (MAB372, 1:200; Millipore/Chemicon, Billerica, MA), GFAP (Ab-4, 1:200; Neomarkers/Thermo Fisher, Freemont, CA), Nestin (Rat-401, 1:4; Developmental Studies Hybridoma Bank, Iowa City, IA), PCNA (PC10, 1:200; Neomarkers/Thermo Fisher), and TUNEL (ApopTag In Situ Peroxidase Kit; Millipore/Chemicon). Images were obtained on an Olympus BX51 microscope, using an Olympus DP71 digital camera (Olympus, Center Valley, PA). For immunostaining of Bmi1, Cyclin D1, and p19^Arf, microwave antigen retrieval was performed for 20 minutes. The following primary antibodies were used: Bmi1 (mouse monoclonal F6, 1:50), Cyclin D1 (Sc-753, 1:100; Santa Cruz Biotechnology, Santa Cruz, CA), and p19^Arf (R562, 1:150; Abcam, Cambridge, MA). Antibodies were detected by peroxidase staining using the Powervision system (Immunologic, Duiven, the Netherlands) followed by visualization on a Zeiss Axiovert microscope (Zeiss, Thornwood, NY). In situ hybridization is described in Supplementary Materials and Methods.

For X-gal staining, tissue was fixed in 4% paraformaldehyde for 1 hour at 4°C. Tissues were stained with X-gal overnight at 37°C, then postfixed in 10% NBF, and processed as previously mentioned, with the exception of using nuclear fast red counterstaining.

Tissue Volume, PCNA, and TUNEL Measurements

Approximate tissue volumes of P21 SmoA1;Bmi1^+/+ tumors and SmoA1;Bmi1^-/- abortive lesions were measured using serial coronal sections as described in Supplementary Materials and Methods. The percentage of PCNA- or TUNEL-positive cells was determined on sections from P21 SmoA1;Bmi1^+/+ and SmoA1;Bmi1^-/- mice (N = 3 for each group), as described in Supplementary Materials and Methods.

Quantitative Real-time Reverse Transcription-Polymerase Chain Reaction

Quantitative polymerase chain reaction was performed on the Roche LightCycler 2.0, using the LightCycler FastStart DNA Master^Plus SYBR Green I kit (Roche) as per the manufacturer's instructions. RNA was isolated from wild type cerebellum or the ventrolateral region of tumor-bearing cerebellum, the region of greatest tumor burden, using the RNEasy kit (Qiagen, Valencia, CA). One microgram of total RNA was treated with DNAse I (Invitrogen, Carlsbad, CA) and reverse-transcribed using SuperScript II (Invitrogen). Of the resulting cDNA, 1/20th was used per quantitative polymerase chain reaction. Target gene expression was normalized to actin expression, and relative control cerebellum gene expression levels were set to 1. Specific primer sequences for Actin, Math1, N-Myc, CyclinD1, and CyclinD2 can be found in Supplementary Materials and Methods.

Results and Discussion

SmoA1 mice developed tumors resembling medulloblastoma as early as postnatal day 7 (P7), with 100% penetrance by P14. Early tumors were first detected as an expansion of EGL in a small ventrolateral region of the cerebellum. By P21, when the EGL is no longer detected in wild type animals, tumors frequently extended along the entire rostral-caudal length of the cerebellum when examined in serial coronal sections (Figure 1A). Tumor-bearing SmoA1 animals died within 10 to 12 weeks. Bitransgenic mice from three independently derived TRE-SmoA1HA lines developed medulloblastomas with 100% penetrance; the data reported in this study were generated using line #140.

Figure 1.

Bmi1 is required for Hh pathway-driven medulloblastoma expansion. (A) Hematoxylin and eosin staining of coronal sections from cerebella of P21 mice. (a) Tumors or (b) ectopic foci of presumed abortive tumor cells (both outlined with dotted black line) are evident external to the molecular (ML in c) in SmoA1;Bmi1^+/+ and SmoA1;Bmi1^-/- animals, respectively. Areas of higher magnification are indicated by black boxes on low power images. (B) Schematic representation of the proposed model showing cellular compartments in developing cerebellum and SmoA1-induced tumors. In transgenic mice expressing SmoA1, tumors arise within the EGL and undergo progressive expansion, leading to death within 10 to 12 weeks. The presence of a small focus of tumor-like cells in a P18 SmoA1;Bmi1^-/- mouse (Figures W1 and W2), with ectopic cells observed in a similar location in P21 SmoA1;Bmi1^-/- animals, argues that medulloblastoma formation is initiated in the absence of Bmi1, but subsequent tumor development is arrested. This is further supported by the presence of ectopic nests of neuronal cells within the molecular layer (Figure 3B) seen in other mouse models of medulloblastoma and human medulloblastoma but not normal cerebellum, in all SmoA1;Bmi1^-/- mice examined at P18 and P21 (N = 5). (C) Comparison of total volumes of medulloblastomas (SmoA1;Bmi1^+/+) and ectopic cells/abortive tumors (SmoA1;Bmi1^-/-) at P21 (N = 3). Error bars indicate range; * indicates statistically significant difference (P = .0044).

To generate SmoA1 mice on a Bmi1-deficient background, we crossed Bmi1^+/- mice with either GFAP-tTA or TRE-SmoA1HA mice, and then crossed the resulting GFAP-tTA;Bmi1^+/- and TRE-SmoA1-HA;Bmi1^+/- progeny (Figure W5). We obtained SmoA1 mice on wild type (N = 80), Bmi1^+/- (N = 17), and Bmi1^-/- (N = 6) backgrounds. In striking contrast to the complete tumor penetrance in SmoA1 wild type and SmoA1;Bmi1^+/- mice, none of the six SmoA1;Bmi1^-/- mice developed full-blown medulloblastomas when examined at P18 to P26. One of two P18 SmoA1;Bmi1^-/- mice contained both presumed abortive tumor cells, as described below, and a small focus of cells with some similarities to medulloblastoma, but with notable differences (Table 1 and Figures W6 and W7). In the additional P18 SmoA1;Bmi1^-/- mouse and three P21 SmoA1;Bmi1^-/- mice, tumors were not detected. Instead, small ectopic foci of presumed abortive tumor cells were identified external to the molecular layer (Figure 1A), a region normally devoid of cells at this stage (Figure 1A). These cells were not detected at P26. The presence of tumor-like cells in a P18 SmoA1;Bmi1^-/- mouse, populations of residual ectopic cells in P18 (N = 2) and P21 (N = 3) mice, and the absence of these cells at P26 (N = 1) implies that medulloblastoma formation is initiated in Bmi1^-/- mice, but the subsequent expansion of tumors is blocked and residual tumor cells are ultimately eliminated (Figure 1B). Our findings suggest a more stringent requirement for Bmi1 in medulloblastoma development than in leukemia, which is blocked by Bmi1 deficiency in secondary but not primary recipients [17]. This may reflect differential requirements for Bmi1 in expansion of CGNPs versus hematopoietic stem cells.

Table 1.

Histologic Diagnosis and Immunophenotyping Summary of Cerebella from SmoA1-Expressing Wild Type and Bmi1^-/- Mice.

Age	SmoA1;Bmi1+/+	SmoA1;Bmi1-/-
	P14 and older	P18 (8940)*	P18 (8941)	P21 (8021, 132-3.2, 135-3.5)	P26 (8221)
H&E, ectopic cells outside of molecular layer†	Medulloblastoma	Proliferative focus and abortive cells	Abortive cells	Abortive cells	Normal
H&E, ectopic cell clusters within molecular layer‡	+	+	+	+	-
PCNA	++++	+++¶	+	+	NA#
Cyclin D1	++++	+++	+	+	NA
Nestin	+++	+	-	-	NA
Nestin (large, round cells)§	-	++	++	+	-
NeuN	+	+	-	-	NA
GFAP	+	++	++++	++++	NA
p19	+/-	+	++	++	NA

^*Numbers in parentheses are animal identifiers.

^†Refers specifically to any unexpected cell populations outside the molecular layer.

^‡Refers to ectopic clusters of neuron-like cells in outer region of the molecular layer, underneath tumor or abortive cells.

^¶Immunohistochemistry for animal 8940 refers specifically to cells within the proliferative focus.

^#NA indicate not analyzed, because abortive tumor cells were not detected at P26.

^§At P18, these distinctive cells were detected both in ectopic regions and molecular layer (Figure W6); at P21, detected only in molecular layer.

Compared to age-matched wild type medulloblastomas, ectopic cells in P21 SmoA1;Bmi1^-/- mice had a lower nuclear-to-cytoplasmic ratio and pale eosinophilic cytoplasm, and the lesions occupied 1/140th the volume (Figure 1C): mean SmoA1;Bmi1^+/+ tumor volume was 2.69 mm³, whereas mean SmoA1;Bmi1^-/- lesion volume was 0.019 mm³ (P = .0044). Medulloblastoma cells in wild type mice and cells in ectopic foci in Bmi1^-/- mice both expressed SmoA1 (Figure 2A), indicating that impaired tumor growth in Bmi1^-/- mice cannot be attributed to lack of transgene expression. Apoptotic cells were 4.7-fold more abundant in ectopic cells than medulloblastoma (29.61% vs 6.26%, respectively, P = .028; Figure 2, A and C). Proliferation was also impaired in SmoA1;Bmi1^-/- lesions; PCNA-positive cells were reduced 13.3-fold, from 81.93% in SmoA1;Bmi1^+/+ tumors to 6.18% in ectopic lesions (P = 7.5 x 10^-6; Figure 2, A and D). These data suggest that both increased apoptosis and reduced proliferation contributed to the impaired outgrowth of putative medulloblastoma progenitors in the absence of Bmi1. Interestingly, a recent report showed that Gli1 blocks proliferation and induces apoptosis of cerebellar neural stem cells but not medulloblastoma stem cells [18], raising the possibility that Bmi1 may play a role in protecting these tumor cells from Hh pathway-driven apoptosis.

Figure 2.

Cell death and cell cycle marker expression in SmoA1;Bmi1^+/+ medulloblastomas and SmoA1;Bmi1^-/- ectopic cells. (A) Immunohistochemical staining for transgene, apoptosis, and proliferation in P21 mice, with outer edge of molecular layer, when visible, outlined with dotted black line. (a) HA-tagged SmoA1 is detected in SmoA1 tumors and ectopic lesions in SmoA1;Bmi1^-/- mice, as well as scattered cells in the molecular layer and internal granular later, but not in control (Wild Type) cerebellum. (b) TUNEL staining reveals apoptosis in both tumors and ectopic cells. (c) High proliferation rate of the tumors is evident from PCNA staining of most tumor cells, whereas only a few cells stain in the lesions of SmoA1;Bmi1^-/- mice. (B) Cell cycle marker expression in P21 mice, with outer edge of molecular layer, when visible, outlined with dotted black line. (a) Tumors express high levels of Bmi1 in nearly every cell. Bmi1 is also appreciable in cells of the molecular layer and internal granular layer of wild type cerebellum. No Bmi1 is detected in tissue from SmoA1;Bmi1^-/- mice. (b) p19^Arf is expressed in ectopic cells, and essentially undetectable within medulloblastomas. Higher magnification (inset) demonstrates nucleolar staining. (c) Cyclin D1 is broadly expressed in medulloblastomas, but is virtually absent from Bmi1-deficient lesions. (C) Apoptosis (% TUNEL-positive cells) is significantly higher in SmoA1;Bmi1^-/- lesions and proliferation (% PCNA-positive cells) is significantly lower (D) (N = 3). Error bars indicate range; * indicates significant differences at P = .028 (TUNEL) and P = 7.5 x 10^-6 (PCNA).

We next examined Bmi1 protein and cell cycle marker expression in Bmi1^+/+ and Bmi1^-/- lesions. SmoA1;Bmi1^+/+ tumors expressed Bmi1 in almost all cells, whereas no Bmi1 was detected in SmoA1;Bmi1^-/- animals (Figure 2B), as expected. Conversely, p19^Arf was detected in cells scattered throughout the ectopic foci in SmoA1;Bmi1^-/- cerebella, whereas it was weakly expressed in medulloblastomas (Figure 2B). This observation is consistent with an inhibitory role for p19^Arf in CGNP proliferation and cerebellar development [8] and suggests that Bmi1 blocks p19^Arf expression in Hh-induced medulloblastoma. Cyclin D1, a Hh target that acts downstream of p16^Ink4a [19], was weakly expressed in a small fraction of ectopic cells in SmoA1;Bmi1^-/- mice but was strongly expressed in medulloblastomas (Figure 2B). Given that Cyclin D1 is required for medulloblastoma development [20], its loss in Bmi1-deficient, SmoA1-expressing cerebellum may contribute to the failure to form tumors. Up-regulation of the CDK inhibitor p21 was not detected in Bmi1-deficient ectopic lesions (not shown).

We also examined expression of lineage markers and the stem and progenitor cell marker nestin. Compared to medulloblastomas, ectopic lesions in SmoA1;Bmi1^-/- mice contained a predominance of cells intensely positive for GFAP, but essentially no NeuN-positive neuronal cells (Figure 3, A and B). This marker profile suggests that ectopic cells in SmoA1;Bmi1^-/- mice are distinct from the preneoplastic lesions described in Ptch^lacZ/+ mice [21]. Increased expression of GFAP in residual cells raises the theoretical possibility that GFAP promoter-driven SmoA1 may be accumulating to toxic levels in some Bmi1-deficient cells, but additional studies would be required to test this idea. In both SmoA1;Bmi1^+/+ and SmoA1;Bmi1^-/- mice, clusters of postmitotic, strongly NeuN-positive neuronal cells were detected in the outer molecular layer just deep to either tumors or ectopic regions, respectively (Figure 3B). The presence of these misplaced cell aggregates, never detected in nontransgenic mice, is in keeping with the idea that there is similar initial tumorigenic response to SmoA1 in the EGL of Bmi1^-/- (Figures W1 and W2 and Table 1) and wild type mice, but it cannot be sustained in the absence of Bmi1. We also observed a striking reduction in nestin-positive cells in SmoA1;Bmi1^-/- ectopic foci (Figure 3C), indicating a depletion of progenitor cells in these presumably abortive lesions. This loss of progenitor-like cells likely contributes to the failure of nascent medulloblastomas to progressively expand in the absence of Bmi1.

Figure 3.

Lineage marker expression in SmoA1;Bmi1^+/+ medulloblastomas and SmoA1;Bmi1^-/- ectopic cells. (A–C) Immunohistochemical staining of tissue from P21 mice, with outer edge of molecular layer, when visible, outlined with dotted black line. (A) Intense immunostaining for the glial marker GFAP is seen in the ectopic cells of SmoA1;Bmi1^-/- mice. Weaker GFAP immunostaining can be appreciated in cells scattered throughout the tumor of SmoA1 mice, and in the molecular layer in all mice. (B) Weak expression of the neuronal marker NeuN is detected in scattered cells within medulloblastoma but not abortive tumors. In addition, strongly NeuN-positive clusters, representing aberrant collections of neurons, are evident within the outer molecular layers of all SmoA1;Bmi1^+/+ and five of the six SmoA1;Bmi1^-/- mice examined, but not control mice. NeuN-positive neurons are appreciable in the expected location in the internal granular layer of all mice. (C) Expression of the stem and progenitor cell marker nestin is appreciable in SmoA1 tumors but is largely undetectable in lesions in SmoA1;Bmi1^-/- mice.

In addition to our findings in P21 and P26 animals described, we also examined SmoA1;Bmi1^-/- mice at an earlier time point, as noted above. Whereas one of two P18 SmoA1;Bmi1^-/- mice was similar to P21 mice, the other P18 SmoA1;Bmi1^-/- mouse contained a small focus of cells with properties intermediate between medulloblastomas and ectopic foci seen in SmoA1;Bmi1^-/- mice at P21. This intermediate lesion was much thinner than age-matched SmoA1;Bmi1^+/+ medulloblastoma (Figure W1A). Whereas the abortive SmoA1;Bmi1^-/- tumor-like lesion was proliferative, fewer cells stained for PCNA and Cyclin D1 than in Bmi1^+/+ tumor, and expression of p19^Arf was elevated (Figure W1, B–D). Increased GFAP immunostaining was seen at the periphery of this lesion as well (Figure W2A). In keeping with what is seen at P21, there is a marked reduction in nestin staining in the Bmi1-deficient lesion and an alteration in its subcellular distribution, including the appearance of large, round, nestin-positive cells (Figure W2B). These distinctive cells, which are never seen in wild type control or SmoA1;Bmi1^+/+ mice, are detected both in regions of ectopic cells and in the molecular layer of 18-day-old and, to a lesser extent, 21-day-old animals, but are absent by postnatal day 26 (data not shown). Immunohistochemical data from all animals are summarized in Table 1.

Although we cannot formally exclude the possibility that failure of SmoA1;Bmi1^-/- mice to develop full-blown medulloblastomas is caused by a Bmi1-dependent loss of tumor progenitor cells, we do not believe this to be the case. Bmi1 deficiency impairs Shh-driven proliferation of CGNPs, the likely cell of origin for the medulloblastomas in our model. However, the cerebella of Bmi1^-/- mice develop relatively normally, despite their smaller size, with formation of the molecular, Purkinje, and internal granular layers, albeit with abnormal basket neuron arborization and reduced molecular layer cellularity [4]. Moreover, the presence of an intermediate tumor-like lesion in a SmoA1;Bmi1^-/- mouse at P18 and ectopic nests of Neu- N-positive cells in five of six SmoA1;Bmi1^-/- mice strongly argues that tumor initiation can occur. Although a major effect of Bmi1 deficiency thus seems to be at the level of medulloblastoma expansion/progression, which is in keeping with Bmi1's proposed role in lung tumorigenesis [14], Bmi1 may also be involved in medulloblastoma initiation.

Taken together, our data establish an obligatory role for Bmi1 in Hh-driven medulloblastoma pathogenesis, provide the first demonstration that Bmi1 is required for de novo development of a spontaneously arising solid tumor in the brain, and identify Bmi1 as a critical downstream effector in tumorigenesis driven by uncontrolled Hh pathway activation. Progression of SmoA1-expressing CGNPs to medulloblastoma involves many of the same molecular pathways that regulate physiologic CGNP proliferation and growth arrest, and loss of Bmi1, by disrupting these signals, prevents expansion of Hh-driven medulloblastomas. Further studies will be required to ascertain whether Bmi1 is required for other pathological or physiological responses to Hh signaling and to establish whether BMI1 will be a useful target for the prevention or treatment of human malignancy.

Supplementary Note

To examine Hh pathway activation in SmoA1-driven medulloblastomas, we performed in situ hybridization for Hh pathway components Ptch1 and Gli1 (Figure W1). Robust expression of both Ptch1 and Gli1, indicating high-level activation of the pathway, was seen in tumors. We also examined the expression of additional key genes by quantitative real-time RT-PCR. Math1, a marker for EGL cells that is seen in both mouse and human medulloblastomas [1,2], was upregulated approximately 550-fold in tumors (Figure W2A). N-Myc, a Hh target gene critical to Hh-driven cerebellum development and medulloblastoma formation [3], was up-regulated 17-fold in tumors (Figure W2B), whereas the Hh targets CyclinD1 and CyclinD2 were overexpressed 11- and 24-fold, respectively (Figure W2, C and D). CyclinD1 is also required for medulloblastoma formation in Ptch1^+/- mice [4].

To assess the location and timing of transgene activation and early tumor formation, we stained 2-day-old GFAP-tTA;TRE-lacZ mice with X-gal. We observed scattered single or small groups of X-gal-stained cells in the EGL of young mice (Figure W3). Consistent with these data, we also observed SmoA1 expression in small clusters of cells in the EGL of 7-day-old SmoA1 mice (Figure W4, A and B). By day 14, however, these small groups of cells had progressed to small masses with robust expression of SmoA1 in 100% of mice examined (Figure W4, C and D).

On the basis of lineage marker studies, Hh pathway activation studies, and overall histologic appearance, the tumors arising in SmoA1 mice resemble previously published mouse models of medulloblastoma and human medulloblastoma. In addition, SmoA1 mice develop very large tumor burden and either die spontaneously or become moribund and must be euthanized by 10 to 12 weeks, with only one SmoA1 mouse surviving for 16 weeks.

Supplementary Materials and Methods

Generation of TRE-SmoA1 Mouse Lines

To generate TRE-SmoA1 mice, we appended a sequence encoding an HA epitope tag to the 3′ end of SmoA1 and placed it under transcriptional control of the tetracycline responsive element. We ligated the resulting construct into the multiple cloning site of the p Tet-Splice vector (Invitrogen), placing it under transcriptional control of the tetracycline responsive element. Complete details of construct generation can be found elsewhere.1 We confirmed the identity of our transgenic construct by sequencing the completed construct at the University of Michigan Sequencing Core. Comparison of obtained primary sequence data with RefSeq mRNA sequence (NM_176996.3) confirmed the presence of the expected W539L mutation and also revealed the presence of an unexpected second point mutation, T425R. We excised and purified the TRE-SmoA1HA/SV40 Intron + poly adenylation fragment, which was microinjected into (C57BL/6 x SJL)F1 x (C57BL/6 x SVJ)F1 eggs by the University of Michigan Transgenic Animal Core. Three TRE-SmoA1HA mouse lines (#99, #140, and #149) were generated and maintained by backcrossing founder mice onto C57BL/6 breeders for at least five generations before experimental crosses.

Genotyping and Housing of Bmi1^-/- and TRE-SmoA1 Mice

We generated SmoA1 mice on Bmi1^+/- and Bmi1^-/- backgrounds by crossing GFAP-tTA and TRE-SmoA1HA mice separately with Bmi1^+/- mice [5] and then intercrossing the resulting GFAP-tTA;Bmi1^+/- and TRE-SmoA1HA;Bmi1^+/- progeny (Figure W5). We chose to use the GFAP promoter (in GFAP-tTA;TRE-SmoA1 mice) because GFAP-Cre-driven deletion of Rb in a p53-deficient background yielded medulloblastomas [6]; the GFAP promoter targets various progenitor cells as well as glia, allowing us to assess the consequences of Hh pathway deregulation in several cell populations in the CNS, in future studies; and a GFAP-tTA transgenic line had already been generated and characterized [7] and bitransgenic GFAP-tTA;TRE-IFNγ mice reported to yield medulloblastomas with high penetrance owing to the activation of endogenous Shh expression [8]. The rationale for a bitransgenic approach stems from the fact that conventional transgenic mice expressing SmoA1, if they develop a robust tumor phenotype, might not be capable of surviving to breed to generate lines. Pups were genotyped between P14 and P21, by which point Bmi1^-/- animals were readily distinguished by their smaller body size. Bmi1, GFAP-tTA, and TRE-SmoA1HA genotypes were confirmed by PCR. Bmi1 genotype was confirmed by multiplex PCR for both the hygromycin cassette in the Bmi1 knockout allele (5′-cgccgtgcacagggtgtcacgttgcaagac-3′ and 5′-caagccaaccacggcctccagaag-3′) [9] and the wild type Bmi1 allele (5′-ccaccacaacacctcatcac-3′ and 5′-cgggtgagctgcataaaaat-3′). SmoA1 genotype was ascertained by individual PCR for the tetracycline transactivator for GFAP-tTA (5′-ctcgcccagaagctaggtgt-3′ and 5′-ccatcgcgatgacttagt-3′ [10]) and the SV40 poly-A tail for TRE-SmoA1HA (5′-ggaactgatgaatgggagca-3′ and 5′-gggaggtgtgggaggttt-3′). In total, 90 litters were screened, comprising 417 pups from 70 of the litters, and an additional 20 litters of unrecorded size, which contained no Bmi1^-/- pups by phenotype and were therefore killed without further analysis.

Owing to impaired hematopoietic function and increased risk of infection in Bmi1^-/- mice, all animals were maintained on water supplemented with 0.1214 mg/ml polymixin B sulfate (Sigma) and 1.1 mg/ml neomycin trisulfate (Sigma). Maintenance of mouse colonies and experimental procedures were approved by the University of Michigan University Committee on the Use and Care of Animals.

In Situ Hybridization

Nonradioactive in situ hybridization for Ptch1 and Gli1 was performed essentially as described in detail elsewhere [11].

Approximate Tissue Volume Measurements

Approximate tissue volumes were calculated using a previously published protocol [4], with serial 5-µm-thick coronal sections cut through the entire rostral-caudal length of P21 SmoA1;Bmi1^+/+ and SmoA1;Bmi1^-/- cerebella (N = 3 for each group), and every third or every fourth slide stained with hematoxylin and eosin (H&E). To calculate the rostral-caudal extent of the lesions, we counted the total number of sections over which the lesions extended, measured the cross-sectional surface area on each H&E-stained section for an individual lesion using ImageJ software, and averaged these values to obtain a mean cross-sectional lesion area. We multiplied this mean area by the total rostral-caudal length of the lesion to calculate approximate volume measurements. P values were obtained using 2-tailed t test, assuming unequal variances.

Calculating Percentage of PCNA- or TUNEL-positive Tumor or Ectopic Cells

After immunostaining tissue sections from three SmoA1;Bmi1^+/+ and three SmoA1;Bmi1^-/- animals for PCNA or TUNEL as described in the Materials and Methods section, we photographed 10 random nonoverlapping high-power fields of tumor or ectopic tissue for each animal. If there were fewer than 10 high-power fields' worth of cells in a given SmoA1;Bmi1^-/- lesion, we photographed all ectopic cells present in the section. We then counted both the number of PCNA or TUNEL-positive nuclei per field and the total number of nuclei per field, and divided the number of positive nuclei by the total number of nuclei to obtain the percent positive nuclei. P values were calculated using the 2-tailed t test, assuming unequal variances.

Quantitative Real-time RT-PCR Primer Sets

Quantitative polymerase chain reaction was performed as described in the Materials and Methods section, using the following primer sets:

• Actin: 5′-tgttaccaactgggacgaca-3′ and 5′-tctcagctgtggtggtgaag-3′.

• Math1: 5′-tgcgctcactcacaaataag-3′ and 5′-taacaacacaatagtccgtgttc-3′.

• N-myc: 5′-gctgcggtcactagtgtgtc-3′ and 5′-ggagaagcctcgctcttgat-3′.

• CyclinD1 [12]: 5′-ctctggctctgtgcctttct-3′ and 5′-ccggagactcagagcaaact-3′.

• CyclinD2 [12]: 5′-ttcagcaggatgatgaagtga-3′ and 5′-gagaaggggctagcagatga-3′.