Abstract Breast cancer is a major health problem and heterogeneity of the disease has been considered as a strong limitation to find the best therapies to cure cancer, overcome recurrences and metastases. The establishment of models that reflect tumor biology and metastatic progression is critical to develop successful new therapeutic strategies. In the breast, orthotopic xenografts currently appear as the best models to study tumor growth, metastasis and develop tools for prognosis prediction. Furthermore, mouse transplant assays have been used to assess cancer stem cell (CSC) activity and demonstrate that leukemia and many solid tumors are organized along a hierarchical model. Despite the promise of the CSC model sustained by mouse transplantation assays, the clinical relevance of xenografts studies to identify determinants of stemness able to influence clinical outcome remains challenging. In breast cancer, transcriptional programs from functionally validated CSC populations remain to be deciphered. Here, we report the establishment of a bank of primary breast tumor-derived xenografts (xenobank). We showed that the xenografts retain the main features of primary tumors, that engraftment is correlated with the presence of CSC in tumors, and that engraftment in the mouse is able to predict prognosis in patients. This suggests that CSCs may govern breast cancer prognosis. We established the gene expression profiles of functionally validated ALDEFLUOR-positive CSC populations (breast CSC-GES) and demonstrated their clinical relevance. Among 2609 patients with breast cancers, we validated that he expression of the breast CSC-GES is correlated with poor outcome and metastasis in uni-and multivariate analysis (5-year MFS was 70% CI95 [67–74] in the breast CSC-positive class and 80% (CI95 [77–83]) in the breast CSC-negative class (p = 5.5E−04 with log-rank test). Furthermore, we identified a core of 19 genes commonly expressed in breast CSC, murine embryonic, neural and hematopoietic stem cells programs and demonstrated for each gene its ability to modulate breast CSC population, being implicated in self-renewing or differentiation programs. We found that the core of genes in common between four stem cell gene expression studies (CE-BCSC-3SC) displayed an adverse prognostic impact for patients with breast cancer. The core contained genes implicated in oxidative phosphorylation, detoxification, lipid metabolism, and genomic stability, and these shared determinants of stemness influenced clinical outcome. Thus, we show ed that CSCs from orthotopically engrafted primary breast tumors have clinical and biological relevance. This functionally validated CSC population is highly correlated with survival and express genes governing main stem cell functions, substantiating a major prediction of the CSC model and opening further promises for new CSC therapies using valid preclinical models. Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P5-03-01.