endent clinical cohorts [6]. Oncotype Dx has also been validated as predictive for response to chemotherapy in ER-positive, node-negative breast cancer [37]. Since both signatures selected their genes based on biomarker potential in clinical cohorts, they include some combination of the intrinsic metastatic potential of the tumor and its resistance to conventional treatments. Prosigna has also developed a prognostic signature based on the PAM50 or even if has characteristics which make it difficult to completely surgically excise. M-Sig may represent a truer measure of metastatic potential independent of treatment effect as it was derived primarily from an in vivo model system. It should be noted that M-Sig does not represent a completely “pure” metastasis signature as it was trained in a clinical cohort. Though it would have been experimentally ideal for the clinical outcomes to be independent of treatment, obviously datasets with untreated breast cancer do not exist, nor would it be ethical to collect such data. However, the relatively uniform performance of our signature between cohorts with different treatments lends evidence that M-Sig is actually predicting independent of treatment. M-Sig could be used in combination with specific treatment response signatures in order to personalize treatment. One can envision a situation where the tumor is predicted to not benefit from standard chemotherapy (low Oncotype Dx score), but has a high metastatic potential (as assessed by M-Sig), in which case the optimal treatment may involve intensified local and/or systemic therapy. Another potential use for M-Sig is in the research setting, where it could be used as an inexpensive and rapid way 10205015 to measure the effect of a particular experimental condition on all the steps in metastasis. The platform-independent nature of M-Sig is particularly advantageous, as it can be applied to any dataset without modification. In addition to providing a prognostic tool that helps identify patients with breast cancers at high risk for systemic progression, M-sig also sheds insight into the Beaucage reagent biological processes underlying metastatic spread. The M-Sig genes identified in this study are involved in both general processes critical to metastasis such as cellular adhesion, migration, proliferation, cell cycle regulation, cytokine signaling, and immune cell evasion [38], as well as canonical oncogenic pathways such as the TGF-beta [33], Jak-Stat [32], PDGF [34], Wnt [35], and Ras [36] signaling pathways. In particular, the enrichment of genes associated with TGF-beta signaling in M-Sig is of particular interest. Several key genes in the TGF-beta pathway are integral to M-Sig and are positively correlated with metastatic potential including Activin R1 (ACVRL1) and SMAD6. Additionally, the T47D cell line, known to be deficient 
in TGF-beta receptors I-II [39], demonstrated no metastatic potential in the CAM assays. Strikingly, TGF-beta receptor 3, an inhibitor of TGF-beta signaling is negatively correlated with metastatic potential, consistent with the literature [40]. Our study also identifies immune system regulation as 8 of the 16 top biological processes associated with metastatic potential. The role of the immune system in regulating metastatic potential and patient outcomes has become an area of active investigation in recent years, with many groups identifying immunophenotypes associated with favorable and unfavorable outcomes in patients with breast cancer [414]. O