Heterogeneity in the AML stem cell pool

To examine AML and stem cell diversity, Heuser and colleagues develop a novel murine model that closely mimics aggressive human AML and demonstrate an essential role of Stat5 in leukemic stem cell renewal.

It would be simpler to develop novel molecularly targeted therapies if the molecular genetic abnormalities in acute myeloid leukemia (AML) cells were simple and consistent. Sadly, they are not. Diverse molecular defects account for this disease in humans. In the most aggressive cases, the chromosomal and genetic abnormalities are highly complex. This challenging state of affairs is further complicated by the fact that molecular events involving an oncogene do not always mean that the oncogene is playing a dominant role in the leukemogenic process. Reliable functional studies on AML cells and, more importantly, the leukemic stem cell (LSC) first identified by Griffin and Lowenberg¹  are required to validate the cause-and-effect relationship between the disease and specific genetic alterations. Furthermore, they are required to develop truly targeted therapies. In this issue of Blood, Heuser et al present and validate a model accomplishing these goals. Using this model in assessing leukemogenesis, they find that leukemic cells produce results that can differ substantially than when using a strategy that quantifies effects specifically on leukemia-initiating cells (LIC)²  (see figure).

Specifically, the authors ectopically overexpressed oncogenes commonly mutated in human AML (MN1; ND13 or HOXA9) and injected them into recipient animals. The leukemogenic potential of the transduced cells was quantified in the 1- (MN1) and 2-oncogene (MN1 + ND13) models using limiting dilutions of the transplanted cells. The prevalence of recipient leukemias at limiting dilutions and the frequency of LIC were substantially greater in the 2-oncogene–expressing cells compared with the one. The importance of using limiting-dilution analyses was demonstrated by subsequent experiments that showed that when nonlimiting dilutions (a large number of bulk AML cells) of the 1- and 2-oncogene cells were injected, there was no difference in latency of leukemia development. Thus, studies in standard mouse models, which do not employ quantification using limiting-dilution assays, may mask potential differences in disease phenotypes, particularly in studies that test the effects of somatic mutations on the function of leukemic stem cells. Interestingly, passaging of transduced cells ex vivo (6 days) or in vivo (serial passaging) proportionally increased LIC frequency in the 2-oncogene cells more than in the 1-oncogene cell. The authors hypothesize this results from increased LSC self-renewal capacity of the 2-oncogene cells. However, they have not ruled out the alternative hypothesis that 2-oncogene cells rapidly acquire additional genetic alterations that increase LIC frequency and expansion capacity, new functions that might not be attributable solely to the transduced genes. The use of cell-cycle phase–\E specific agents ex vivo or other related experiments using this model could possibly distinguish one mechanism from the other.

The development of heterogenous AML (and LSC) models allows characterization of corresponding molecular changes that mediate disease relapse and aggressiveness. The correlation between STAT hyperactivation and AML has been well documented.^3,4  In the different 2-oncogene models (MN1 + ND13 or MN1 + HOXA9), Heuser et al found that cell growth was stimulated by GM-CSF, which, in turn, induced hyperphosphorylation of Stat5 and Stat1. To examine the requirement of these 2 genes in LSC expansion and proliferation, genetic models (Stat1^−/− and Stat5^−/− mice) were used. Stat5b deficiency in the 2-oncogene–expressing cells (MN1 + HOXA9), and to a lesser extent Stat1 deficiency, suppressed both GM-CSF–induced growth proliferation in vitro and LIC frequency in vivo. Thus, Stat5 up-regulation appears to enhance LSC self-renewal and provides further rationale for utilization of STAT5 inhibitors in treatment of human AML, as has been proposed by a variety of groups. It should be noted that although LIC frequency and proliferation of the 2-oncogene–expressing cells was significantly reduced in Stat5^−/− mice, the cells remained leukemogenic, demonstrating that targeting of the STAT5 pathway will likely be insufficient to treat AML. Clearly, in light of the recently reported role of Stat5 in maintaining quiescence of normal hematopoietic stem cells,⁵  trials of such agents should be attended by careful companion studies seeking to validate effects on the target cell population and the nontarget population as well.

Murine models of AML (and human malignancies in general) have been limited by problems: either the genetic alterations used to initiate the malignancies in mice are not commonly found in human cancers or inactivation of homologs of human cancer-causing genes in mice does not replicate the disease phenotype. However, in this manuscript, the 2-oncogene model is indeed relevant to human AML. The group found that these same 2 oncogenes (MN1 and HOXA9) were concomitantly altered in the most aggressive and most cytogenetically unstable human AMLs and also linked this subset of AML to the highest levels of STAT activation. The group has developed an important and relevant model of poor-prognosis AML. It should ultimately allow direct testing of rationally designed agents that will effectively target LSCs without negatively influencing nonleukemic stem cells and progenitors.