(A) The cerebellar anlagen at E14. The VZ and RL serve as mitotic niches at this age. The EGL is formed by proliferating CGNP that have left the RL. (B) At P7, the developing CB is divided into 10 lobes with defined cortical layers. (C-E) Sagittal sections at E14 in hGFAP-cre mice showing co-localization of Nestin, cre and Pax6 proteins in progenitor cells of the VZ and rRL. (F) Olig2 expression is largely restricted to dorsal rRL. (G, H) Fate mapping experiments (hGFAP-cre × ROSA26-eYFP^Fl/Fl mice) reveals YFP in the EGL and a subset of RL cells that express Olig2 (arrows). (I) Olig2⁺ RL precursors co-label with the CGNP marker Pax6; some single Pax6⁺ precursors in the VZ were observed (arrows). (J) Summary of cells in the RL region at E14 showing relative positions of radial glia (yellow/green), Pax6(blue) ⁺, Olig2(red)⁺ and Pax6⁺Olig2(purple)⁺ cells. BG, Bergmann glia; CP, choroid plexus; GNP, granule neuron precursors; ML, molecular layer; PL, Purkinje cell layer.

Targeting hGFAP⁺ and Olig2⁺ precursor cells with SmoM2 results in MB (A-H). While neoplastic transformation of hGFAP⁺ cells affects the entire cerebellar cortex (A-D), MBs from Olig2-expressing precursors are restricted to posterior cerebellar regions (E-H). Although hGFAP-cre and Olig2-cre are broadly expressed in all parts of the CNS, neoplastic lesions due to SmoM2 are restricted to the CB. Forebrain sections including the subventricular zone (SVZ) in adult animals appear normal (I-L). Transplantation of Olig2-tva-cre:SmoM2 tumor cells into SCID mice gives rise tumors both in hindbrain (M,N) and forebrain (O,P) regions. A, E and I-P are frontal, B-D and F-H are sagittal sections stained with H&E. C and D are high power magnifications from B, so are G and H from F. Asterisks mark lateral ventricles in I, K and O and the IV^th ventricle in M. BS, brainstem.

(A) Principal component analysis of global gene expression in conditional MB models [Math1(green)-, Tlx3(magenta)-, hGFAP(red)-, Olig2(blue)-cre:SmoM2 and Ptc^+/- (turquoise)] and samples from normal littermate control CB (black circles) are mapped onto normal cerebellar developmental space as defined by a 7000 member, rank-normalized gene set (Kho et al., 2004), from developing CB at stages P1, 3, 5, 7, 10, 15, 21, 30, 50 and 60 (orange numbers). Tumors were related to early rather than late developmental stages of the CB as we have reported for human MB (Kho et al., 2004). (B) Immunohistochemical analysis confirms all murine tumors exhibit a similar immunohistochemical staining pattern to each other as well as human MB with respect to standard neuronal and glial neuropathological markers. Results are representative of at least two tumor samples.

(A-D) OLIG2 expression in human classic and desmoplastic MB and percentage of tumors of each class that contain OLIG⁺ cells (also see Suppl. Tab 1). (E, F) Combined FISH and immunohistochemical analysis reveals the presence of tumor specific C-MYC genomic aberrations (Herms et al., 2000) within OLIG2⁺ cells (yellow nuclei) of human MB (n=2). C-MYC aberrations were identical in OLIG2⁺ and OLIG2^- tumor cells. Note combined copy number gain (red signals, 5′ C-MYC) and structural aberration (loss of green signals, 3′ C-MYC), as detected using break apart probe set. Normal MYC loci show merged colors as (focal yellow signals). (G, H) OLIG2 co-localizes with NEUN, and the proliferation marker Ki67. (I) Scheme for experiments using RCAS-GFP infection of Olig2-tva-cre:SmoM2 tumor cells. (J-M) Double-fluorescence images show expression of GFP that co-localizes with immunoreactivity using antibodies against Olig2 and GC lineage markers. Note that while Olig2-tva-cre:SmoM2 express YFP the signal is very weak/undetectable compared with GFP encoded by RCAS virus, as shown (J-M). (N) Summary of findings at Day 6 in GFP-labeled cells. (O-Q) Analysis of acutely disassociated Olig2-tva-cre:SmoM2 tumor (Day 0). Of Olig2+ cells, 66% co-labeled with Pax6 and 12% with NeuN, such cells represented 4% and 1% of total cells in the tumor, respectively.

(A)Olig2-tva-cre allele and genotyping of the respective mutant mice using PCR to generate wild type 429bp and mutant (cre) 446bp products. (B-I) Fate mapping of Olig2-expressing cells with ROSA26-eYFP^Fl/Fl reporter. Co-localization of YFP and Olig2 proteins in sagittal sections of P7 cerebella reveals that YFP is expressed in oligodendrocytes within the cerebellar WM (yellow arrows in C) as well as in occasional cells within the EGL (white arrows). YFP-Pax6 labeling confirms the granule lineage character of YFP-expressing EGL cells (D, E, arrows). The IGL of P21 cerebella demonstrates lobe-specific differences in Olig2 fate mapping. (G) The IGL of Lobe II contains YFP⁺ cells that predominantly co-express Olig2 (arrow and cell shown, inset) but not the GC marker Zic. In contrast, the IGL of Lobe X (H) contains numerous YFP/Zic double-positive cells (arrow and inset). (I) Mean percentage (+/-SEM) of Olig2⁺YFP⁺ and Zic1⁺YFP⁺ double-positive cells in the IGL of lobe II and X. (J-M) In utero fate mapping of Olig2-expressing cells by RCAS-GFP virus injection into the rRL of E15.5 Olig2-tva-cre mice. Analysis of the CB at P7 reveals GFP to be expressed in Olig2⁺ (K) WM cells, as well as Pax6⁺ (L) and Zic⁺ (M) cells of the IGL in Lobe X.

(A-H) Cre expression driven by Math1 or Tlx3 promoter sequences is normally restricted to granule lineage cell as shown by crosses with Rosa26-eYFP conditional reporter mice. (C, D, G) GC fate-mapped by Tlx3-cre are restricted to Lobes VI-IX. (F, H) In contrast, Calbindin-expressing PN do not derive from Math1⁺ or Tlx3⁺ precursors. (I-L) Both Math1⁺ and Tlx3⁺ granule neurons give rise to MB after activation of SmoM2, and Tlx3-cre driven tumors obeyed posterior restriction consistent with fate mapping (compare arrows in C, K).

(A, C) In Math1-cre:SmoM2 animals, morphologic evidence hyperplasia is predominantly observed in the EGL, but not RL or VZ, at P0 and is more apparent by P7. (B, D) Greatly increased numbers of Olig2⁺ cells were observed in hyperplastic regions of the EGL--but not RL--of P0 and P7 Math1-cre:SmoM2 mice. (E, F) Expression of cre proteins in Math1-cre mice co-localized with Olig2 at E18 but not P7. (G) Olig2 co-expression was detected in only ∼2% of pH3⁺ cells in tumors. (H) In contrast, ∼98% of pH3⁺ cells co-labeled with cre proteins. (I-K) Phenotype of acutely disassociated cells and (insets) cells cultured from MB for four weeks. While most of dissociated Math1-cre:SmoM2 tumor cells express the GC markers Pax6 or (inset) NeuN (I), some express Olig2 (J) and Gfap (K). SmoM2 is fused to YFP, which therefore marks tumor cells. (L) Cell counting reveals that more than 90% of cells expressing Gfap, Olig2 and Pax6 are tumor cells, as shown by co-expression of YFP. (M, N) Immunohistochemistry of tumor sections demonstrates similar results. (O) 81% of Olig2⁺ and (P) 100% of Gfap⁺ cells in tumors expressed the CGNP marker Pax6.

(A) Scheme showing P7 neuronal progeny deriving from hGFAP- and Olig2-expressing progenitors of the embryonic CB. hGFAP⁺ cells produce the majority of granule cell precursors (GNP) as well as astrocytes (Glia) and IN. While some Olig2⁺ cells derive from hGFAP⁺ cells (black dashed arrow), additional sources are possible. Olig2⁺ cells of the rRL produce a subset of GNP in cerebellar lobes IX and X, and also give rise to IN and oligodendrocytes (Glia) in all lobes (red arrows). The blue dashed arrow indicates the possibility of GNP de-differentiation in the context of Hh-induced MB. (B) Activation of SmoM2 in GNP using hGFAP- or Math1-cre produced diffuse MB involving most cerebellar lobes and the vermis, while focal, posterior-lateral MB was produced from the Olig2-cre driver; Tlx3-cre produced tumors with posterior restriction. (C) Possible cellular pathways of tumorigenesis. Introduction of an oncogene (red arrowheads) into multipotent (A, giving rise to distinct progeny types B1, B2, B3), unipotent (C2) or even differentiated (D2) subtypes of a tumor-competent lineage (e.g., GNP of CB) might result in a uniform tumor phenotype providing only one stage (C2) is capable of rapid expansion in response to the signal. This seems the simplest model to fit our collective findings; however, tumor propagation from an A or B2 cell cannot be ruled out. The later possibility of de-differentiation from D2 > C2 is speculative and not supported by the present work.