The concept of cancer stem cells has revolutionized our current vision of cancer development and was validated in solid tumors and cancers of the primitive hematopoietic compartment. Proof of the principle is still lacking, however, in malignancies of differentiated B-cells. We review here the current literature, which nevertheless suggests hierarchical organizations of the tumor clone for mostly incurable B-cell cancers such as multiple myeloma, lymphomas and B-chronic lymphocytic leukemia. We propose two models accounting for cancer stem cells in these contexts: a "top-to-bottom" clonal hierarchy from memory B-cells and a "bottom-to-top" model of clonal reprogramming. Selection pressure on the growing tumor can drive such reprogramming and increase its genetic diversity.