p130Cas/Cyclooxygenase-2 (Cox-2) axis sustains mesenchymal traits and predicts poor outcome in human breast cancers. (A) Left panels: extracts from LM2-4175 cells expressing scramble (Ctr shRNA), p130Cas shRNAs (p130Cas shRNA) or silenced p130Cas cells reconstituted with mouse full-length GFP. p130Cas (mFLCas) were blotted with antibodies to p130Cas, Cox-2, Snail, Slug, Twist, pSrc (pTyr416), pJnk (Thr183/Tyr185), and Vinculin as loading control. Right panels: representative images (10X magnification) of Ctr (a), p130Cas silenced (b), and p130Cas reconstituted (c) LM2-4175 cells. A 20X magnification of the squared field is shown (d, e, and f). Quantification of length/width ratio on five distinct microscope fields in each condition (**P <0.001) is shown. (B) Left panels: representative images of LM2-4175 cells treated with DMSO (a), 10 micromolar c-Src inhibitor (SU6656) (b), or 40 micromolar JNK inhibitor (SP600125) (c) for 16 h. 20X magnification of the squared field are shown below (d-f). Right panels: protein extracts were blotted with antibodies against p130Cas, Cox-2, pSrc (pTyr416), c-Src, pJnk (Thr183/Tyr185), Snail, Slug, Twist, and Vinculin as loading control. (C) In silico analysis of NKI dataset. Before analysis, the dataset was gene mean centered by subtracting the mean value for each gene across all samples of the compendium from all data points, so that in all cases expression values of each data point were reported as positive or negative depending on whether it was higher or lower than the mean value of that gene across the samples. Kaplan-Meier curves indicating survival probability of patients with tumors overexpressing p130Cas and Cox-2 (double positive), compared to tumors with an expression of p130Cas and Cox-2 lower than the mean value previously defined (non-double positive). (D) In silico analysis as in (C) on the Koo Foundation Sun Yat-Sen Cancer Center dataset.