Treatment of solid tumors with standard chemotherapy often leads only to partial response. Thus, tumor recurrence after chemotherapy driven by tumor-reinitiating cells (TRICs) is a central problem in cancer therapy. Despite its importance, the mechanisms accounting for variable therapy response and tumor re-initiation in vivo are poorly understood. A significant problem in understanding this phenomenon in humans is the inaccessibility of matched clinical samples before and after chemotherapy from the same patient. Mouse models of cancer that closely mimic the human disease are useful tools to study the process of tumor re- initiation after chemotherapy. However, few studies have systematically utilized these models to understand chemotherapy response. Using a mouse model of human lung cancer, we have identified a subset of tumor cells that can be isolated by their cell surface markers and have an increased intrinsic resistance to chemotherapy. We utilized a Kras-driven lung tumor model crossed to a conditional transgenic reporter (tdRFP) in order to facilitate isolation of tumor cells by FACs. Cisplatin treatment of these mice leads to a dramatic decrease in the number of CD44+;tdRFP+ cells, suggesting that CD44- cells are chemotherapy resistant. Using an antibody that detects cisplatin-DNA adducts, we find that CD44- and CD44+ tumor cells have distinct DNA repair capacities. Chemoresistance in this model is associated with a dramatic increase in sphere-forming ability in vitro. In vitro sphere-forming ability is further enriched in a CD44-/CD24+ subpopulation. Thus, we have uncovered a relationship between chemoresistance and characteristics associated with cancer stem cells (cell surface marker heterogeneity, sphere formation). In this proposal, we will utilize mouse genetics, functional genomics and primary human lung cancer samples to elucidate the mechanism of chemoresistance and its relationship to the cancer stem cell phenotype in non-small cell lung cancer.