Expression data from mouse neural cells and tumors

Neural stem cells (NSCs) are considered to be the cell-of-origin of brain tumor stem cells. To identify the genetic pathways responsible for the transformation of normal NSCs to brain-tumor-initiating cells, we used Sleeping Beauty (SB) transposons, to mutagenize NSCs. Mobilized SB transposons induced the immortalization of NSCs. Immortalized NSCs induced tumors upon subcutaneous transplantation in immunocompromized mice. To further classify the immortalized cells and mouse tumors, we performed Gene Set Enrichment Analysis (GSEA) using DNA microarray data. Ten immortalized NSC lines (four WT and six p53 mutant lines) and two normal NSCs were used for RNA extraction and hybridization on Affymetrix microarrays. CEL files were processed using the RMA algorithm to normalize the data. Unsupervised hierarchical clustering was applied to the normalized gene expression data across samples using the R module pvclust (Suzuki and Shimodaira, Bioinformatics 22, 1540-1542, 2006). Hierarchical clustering showed a close similarity in gene expression between all ten immortalized lines, compared with two normal NSCs. We then compared their gene expression profiles to those present in the brain transcriptome database (Cahoy et al., J Neurosci 28, 264-278, 2008). Using five gene sets that are associated with neurons, oligodendrocytes, OPCs, astrocytes and cultured astroglial cells, GSEA enrichment scores were calculated for the ten immortalized NSCs lines against two control normal NSCs. GSEA enrichment analysis was performed using GSEA software v. 2.07 (http://www.broadinstitute.org/gsea). We found a strong enrichment for genes that are differentially expressed in the cultured astroglial cells. Twelve tumors (six WT and six p53 mutant tumors) and three control subcutaneous tissues were used for RNA extraction and hybridization on Affymetrix microarrays. CEL files were processed using the RMA algorithm to normalize the data. Unsupervised hierarchical clustering was applied to the normalized gene expression data across samples using the R module pvclust (Suzuki and Shimodaira, Bioinformatics 22, 1540-1542, 2006). Hierarchical clustering of 12 tumors showed a close similarity between the tumors, compared with three control subcutaneous tissues. We next performed GSEA analysis on the tumors. Using gene sets specific to the four defined GBM subtypes (Verhaak et al., Cancer Cell 17, 98-110, 2010), a GSEA enrichment score was then calculated for the twelve tumor samples against three control subcutaneous tissue samples. GSEA enrichment analysis was performed using GSEA software v. 2.07 (http://www.broadinstitute.org/gsea). This analysis showed that the mouse tumors were significantly enriched for genes specific for the mesenchymal GBM. Immortalized cell lines and tumors were used for RNA extraction and hybridization on Affymetrix microarrays. Ten immortalized lines (Line 3, 28, 31, 59, 95, 116, 119, 120, 123, 331) were compared with two normal NSCs. Twelve tumors were generated from four lines (1, 31, 59, 123) and compared with three subcutaneous tissues (SubQ).