Evidence that cancer is a stem-cell disease comes mainly from studies of acute myeloid leukaemia (AML) in humans. Most leukaemic cells do not proliferate and the leukaemia must be sustained by rare, self-renewing leukaemic stem cells (LSCs), which, instead of becoming more specialized, retain the ability to divide. John Dick and colleagues have now further characterized LSCs and the LSC hierarchy.

Previous experiments in a non-obese diabetic/severe combined immunodeficiency (NOD/SCID) mouse model showed that normal human stem cells (HSCs) and the cells capable of initiating AML in these mice — SCID leukaemia-initiating cells (SL-ICs) — have many of the same cell differentiation markers. Isolated SL-ICs derived from patients with AML can generate AML grafts in mice that resemble the original patient's disease. The authors hypothesized that LSCs originated from the HSC pool, rather than from a committed clonogenic progenitor pool as previously thought.

Normal HSCs vary in their repopulation capacity — they yield both short- and long-term repopulating cells — so Dick and colleagues investigated whether the LSC pool is similarly complex. Using lentivirus-vector-mediated clonal tracking of patient samples engrafted to NOD/SCID mice, they found that the tracked SL-ICs could be divided into clones that occurred transiently during the engraftment process (short-term SL-ICs) and those that contributed stably throughout the experiment (long-term SL-ICs).

So, what underlies the functional heterogeneity of SL-ICs? It has been suggested that SL-ICs have increased self-renewal capacity compared with normal HSCs. To study this further, the authors used a serial transplantation technique — they injected bone marrow from an engrafted NOD/SCID mouse into two secondary recipient mice and then injected bone marrow from one of the secondary mice into two tertiary mice. SL-ICs were efficiently transduced from mouse to mouse. Some clones from primary mice were present in one or both of the secondary recipients, proving that long-term SL-ICs must have undergone self-renewal in the primary mouse. Of the 169 clones that were unique to the primary mouse, many were not found in either of the secondary mice — indicating that these are short-term clones of reduced self-renewal capacity. Other clones that persisted in the primary mice then only occurred transiently in secondary mice, indicating that short-term clones can be generated from long-term clones. Fewer clones were seen after tertiary transplants, indicating that only a restricted population of SL-ICs have extensive long-term self-renewal potential. In addition, a few clones were detected in secondary mice that were not seen in primary mice. The authors hypothesize that these cells might be largely quiescent in the primary mice, but then differentiate into a rapidly expanding clone, or that although the cell-cycling speed remains slow, the proliferating pool becomes detectable with subsequent divisions.

The similar complexity of the LSC and HSC compartments indicates that initiating leukaemogenic events of AML probably occur in HSCs, with subsequent alterations in either stem cells or downstream progenitors altering the self-renewal potential and resulting in LSCs. Standard therapies for AML target proliferating cells; the presence of quiescent cells that can later proliferate illustrates the shortcomings of this approach. These data highlight the need for therapies targetting the long-term repopulating LSCs, which are responsible for aggressively driving the growth of AML.