Chronic myeloid leukemia (CML) is a typical model to study cancer stem cell biology. CML stem cells reside in the CD34+CD38− fraction coexisting with normal hematopoietic stem cells (HSCs). By acquisition of BCR-ABL, an oncogenic fusion transcript encoding a constitutively active tyrosine kinase, HSCs become CML stem cells and progressively outgrow normal HSCs at the stem cell niche. The majority of CML stem cells is dormant but expands their clones mainly at the myeloid progenitor stage. Therefore, the effective target of tyrosine kinase inhibitor (TKI) is proliferating CML progenitors, and a fraction of CML stem cells persist after long-term TKI treatment. Thus, CML stem cells are heterogeneous, containing a population not addicted to the BCR-ABL kinase signaling. Importantly, those residual CML stem cells express BCR-ABL at a very low level. We hypothesize that the acquisition of BCR-ABL is not sufficient for HSC to become CML stem cells, because BCR-ABL is sometimes detectable in healthy individuals. Such BCR-ABL-expressing HSCs might be pre-CML stem cells, and additional events upregulating BCR-ABL expression might be required for formation of CML stem cells.