Expression data from MDA-MB-231 reference and in vitro-derived subpopulations with distinct invasive potentials

To understand the link between invasion behavior and the steps of metastasis formation, we isolated invasive subpopulations from MDA-MB-231 cells in vitro using matrigel coated boyden chambers. Whole genome transcriptional profiling was used to characterize the expression changes uniquely related to invasive abilities of these cells. In this dataset, we include expression data obtained from MDA-MB-231 invasive cells (INV cells) and those that failed to invade the coated membrane during selection (REF cells). These subpopulations have been characterized in vitro by the differences in motility and invasion, cell morphology, adhesion properties to endothelial cells or fibronectin, proliferation, and resistance to doxorubicin as well as to growth factor starvation. INV tumors xenografted in nude mice presented higher volume with more angiogenesis. In addition, when injected into blood circulation, INV cells induced more sites of metastasis in nude mice as compared to REF cells, and dramatically diminished survival by about 80%. Transcriptomic analysis showed differences in genes involved in proliferation, chemotaxis and cytoskeleton, but particularly up-regulated genes involved in negative regulation of apoptosis and down-regulated genes involved in cell adhesion or cell-cell junction. In conclusion, invasive behavior was sufficient to reveal 134 differentially expressed genes correlated with invasive cell aggressivity. 8 total samples were analyzed (4 for each condition). To determine differentially expressed genes, differences between both conditions were calculated on each day of measurement and analyzed using a two-sided paired t-test implemented in the anapuce R package (http://www.agroparistech.fr/mia/outil.html). Pairing was made to remove variation due to the day of measurement in the comparison between the two conditions. The variance was split between subgroups of genes with homogeneous variance (Delmar P, Robin S, Daudin J.-J. VarMixt: efficient variance modelling for the differential analysis of replicated gene expression data. Bioinformatics, 2005, 21: 4,502-508). Statistical raw p-values were adjusted for multiple comparisons using the Benjamini-Hochberg procedure (Benjamini Y, Hochberg Y. Controlling the false discovery rate: a practical and powerful approach to multiple testing. J. R. Statist. Soc. B, 1995, 57:1, 289-300) which controls the False Discovery Rate (FDR) after the estimation of the proportion of genes not differentially expressed using the smoother method (Storey JD, Tibshirani R, Statistical significance for genome-wide experiments. Proc. Nat. Acad. Sci. USA, 2003, 100, 9440-9445). The level of statistical significance was set at 0.05 for all the comparisons.