A potent few. Breast cancer cells with stem cell–like markers formed a large tumor (bottom left and mouse’s left side), whereas injections of other breast cancer cells didn’t grow (top left and mouse’s right side). leukemia-initiating cells from the various AML subtypes and found that all belonged to that same CD34+/CD38–class. When put into NOD/SCID mice, however, each cell type produced a leukemia identical to that in the patient from which it had originally been isolated. All in all, Dick concludes, it’s likely that the initial mutations that gave rise to the leukemias arose in normal stem cells, causing them to take the wrong developmental pathway.

Stanford’s Weissman has come to a similar conclusion by tracing a common genetic abnormality leading to AML. This is the socalled 8; 21 chromosomal translocation, in which the AML gene on chromosome 21 is fused to the ETO gene on chromosome 8, thus leading to production of an abnormal hybrid protein that alters gene expression. Weissman found the translocation both in stem cells and in more mature blood cells of several different types in AML patients. That suggests that the translocation originated in a stem cell that then gave rise to the different mature cells as well as replicas of itself. But even though the translocation is necessary for this type of AML to develop, Weissman says it is not sufficient. Stem cells that were isolated from patients in remission and carried the abnormality, he and his colleagues found, differentiated normally in culture. This indicates that additional mutations, possibly occurring at a later developmental stage, are necessary to produce leukemia. Unlike Dick, who sees the original hematopoietic stem cell as potentially leukemic, Weissman thinks that the leukemia stem cell may be one or more steps down the differentiation pathway. The cells of solid tumors are harder to isolate and study than those of the hematopoietic system, but recent evidence suggests that solid tumors also contain stemlike cells. Some of it comes from research on breast cancer by Michael Clarke and his colleagues at the University of Michigan Medical School in Ann Arbor. They, too, used the NOD/SCID mouse to assay for cells able to initiate cancer, in this case injecting human breast cancer cells into mouse mammary glands. The Michigan team found that, as with AML, only a small fraction of breast cancer cells form tumors in the NOD/SCID mouse, as reported in the 1 April issue of the Proceed-