Robustness of gene expression profiling in glioma specimen samplings and derived cell lines

One of the most promising applications of microarrays is class distinction through gene expression profiling as a diagnostic tool. However, as there is apparent spatial heterogeneity in the morphology of cancer cells within a tumor, it is unclear if tumor sampling can be applied and yield consistent signals. In this report, we examined six brain tumors, four glioblastoma, and two oligodendroglioma biopsies. The six brain tumor tissues from two distinct different classes were dissected in four distinct areas and gene expression was profiled using microarrays. We used hierarchical clustering to compare the variability of gene expression profiles between spatially distinct biopsies of the same tumor as compared to the variability between tumors of the same histologic group. We conclude that, in general, repeat spatially distinct samples are not needed for microarray experiments and the gene expression signatures are robust across the tumor. Predominantly, variation was much greater between samples from different patients than from the multiple samplings of given tumor. Further, we compared biopsy expression profiles to the cell lines derived from those tissues. In general, the tumor cell lines vary greatly from the parental tissues and cluster more strongly with each other than the parental tissue. We select and examine the set of genes altered in expression to allow adaptation to cell culture.