Human non-GCIMP gioblastoma subtypes evolve from a common proneural-like precursor glioma

In order to understand the relationships between the human non-GCIMP glioblastoma subgroups, we performed computational analysis of human genomic data to predict the temporal sequence in which the driver events arise during tumorigenesis. The order of evolutionary events for non-GCIMP GBM is 1) chr 7 gain and loss of chr 10, followed by 2) CDKN2A loss and/or TP53 mutation, and 3) alterations canonical for specific subtypes such as NF1 loss or focal amplification of PDGFRα or EGFR. We then developed a computational methodology to identify the drivers of broad copy number changes, identifying PDGF-A (chr 7) and PTEN (chr 10) as driving the initial non-disjunction events. These predictions were validated using mouse modeling, showing PDGF-A is sufficient to induce PN-like gliomas that are enhanced by loss of Ink4a-Arf, Tp53 or Pten. Additional Nf1 loss converts PN to the MES subtype. Our findings suggest non-GCIMP GBMs arise as, and evolve from, a common proneural-like precursor. hGBM_sphere: Three different human GBM sphere lines expressing two different NF1-shRNAs or empty shRNA-control were generated by a lentiviral infection with the relevant pLKO.1 vectors. The total RNAs were obtained from the GBM sphere lines and the gene expression profile of the NF1-shRNA cells was then compared to the control cells. hGBM_rapamycin: Two human GBM sphere lines (TS543 and TS667) expressing the NF1-shRNAs were generated by a lentiviral infection with two different pLKO.1-NF1-shRNA vectors (two different target sequence; #2 and #5). The total RNAs were obtained from the NF1-shRNA lines treated with a 1nM rapamycin or 0.1% DMSO control for 5 hours and the gene expression profile was then compared between the both groups. Murine_gliomas: Murine brain tumors were generated by an injection of DF1 cells producing the relevant RCAS virus into newborn pups or adult mice brain. When mice presented any symptoms of disease, the mice were sacrificed and the brain tumor tissues were macroscopically dissected. Then total RNAs were extracted from the murine brain tumors or normal brain tissues and the gene expression profile was compared between various tumor types and/or normal brain tissues. In this study, two different RCAS-shNf1 (GR249 and GF249) and shp53 (R696 and mR696) was used for the generation of the murine brain tumors. Each RCAS vector has different structure but the target sequence against the Nf1 and p53 is same and similar knockdown effect was observed. Tumor grade were determined with the residual sections after a piece of tumor tissue were excised for the RNA extraction. The detail information of the experiment was described in the associated-publication. Murine_shere: Murine neurosphere lines were generated by a retroviral infection with the relevant RCAS viruses. The total RNAs were obtained from the murine neurosphere lines expressing the RCAS-shGL2, shNf1, shp53 or shNf1+shp53 and the gene expression profiles of the shNf1, shp53 or shNf1/p53 cells were then compared to the control shGL2 cells.