The identification of cancer stem cells (CSCs) has fundamentally changed the understanding of tumor biology. CSCs drive tumor initiation and long term tumor growth, can cause relapse and are essential for metastatic colonization. These features place cancer stem cells in the focus for cancer therapy. Recently, great advances have been made in the field of cancer adoptive immunotherapy using engineered T cells. Combining both fields of cancer research to generate immune therapies directly targeting CSCs for increased efficacy and long-term anti-tumor effects would generate great hope for cancer patients. Herein, proof of concept is made that chimeric antigen receptor (CARs) T cells killing breast CSCs are capable of eliminating established metastases. Killing of the CSCs modifies the immune environment resulting in the loss of regulatory T cells and in the induction of an endogenous immune response directed against the tumor that is indispensable for treatment success. Antibody mediated ablation of the CAR T cells induced a rapid re-appearance of CSCs and re-establishment of an immunosuppressive environment in the tumor creating a need for continuous targeting of the CSCs. I demonstrate that endogenous T cell responses are less effective against CSCs than the rest of the tumor. Furthermore, proof is made across multiple tumor models, that activated immune cells and inflammation induce dedifferentiation of bulk tumor cells which regain stem-like phenotypes. This observation further emphasizes the importance of directly targeting the CSCs. Studying the interplay between CSCs and the immune system requires syngeneic, immune competent murine models. This greatly limits the transgenic markers that one can use to visualize, track and target CSCs. Herein, we developed different tolerant mouse models to study breast cancer stem cells in vivo. 4T1 breast adenocarcinoma cells were engineered to express hCEA or GFP-based model surface antigens under the control a stem cell specific promoter. These cells, injected into syngeneic model antigen-tolerant mice established lung metastases which were eliminated by CAR T cell therapy. To further validate this approach using endogenous CSC markers, a genetic cancer model (MMTV-PyMT) has been developed, where stem cell Identification is based on CD90 allelic differentiation. Overall, this work highlights an important interplay between CSCs and the immune system. My results suggest that a feedback loop may be created were immune attack induces CSCs, which are immune privileged and escape killing from endogenous T cells. Subsequently, CSCs induce and control an immunosuppressive environment leading to immune escape. This provides a strong rationale to develop CSC-targeted therapies to be administered in combination with immunotherapies.