Identification of variables determining the engraftment potential of human acute myeloid leukemia in the immunodeficient NOD/SCID human chimera model

Among a variety of immunodeficient mouse strains the non-obese diabetic (NOD)/LtSz scid/scid strain appears to be most useful in allowing the engraftment of human AML. However, the large variability in ability to engraft and the levels of engraftment reached have not been explained. To address these issues we have investigated the NOD/SCID repopulating ability of 27 newly diagnosed AML samples. Patients were selected for the absence of internal tandem duplications in the Flt3 gene as we previously reported this mutation to be associated with an enhanced engraftment potential in this model. We observed that secondary AML (n = 6) had a significantly increased level of engraftment when compared to primary AML (n = 21, median levels 73.3% for secondary AML vs 8.94% for primary AML, P = 0.01). Within the primary AML, a significantly higher engraftment was observed in the FAB class MO than in FAB classes M2, M4 and M5. Within primary AML, samples of patients who failed to respond to the initial therapy gave rise to a higher level of engraftment than samples of patients who did respond to therapy. A similar observation of an increased engraftment correlating with a poorer patient prognosis could be made when applying cytogenetic risk stratification. However, within the primary AML the most important clinical parameter correlating with the level of engraftment appeared to be the patient's WBC count at diagnosis (P = 0.0000). Covariate analysis with the WBC count as a covariate could also fully explain the differences observed in the cytogenetic risk groups, or on the basis of the initial therapy response. Although large differences could be observed, the ability to engraft the NOD/SCID mice was not linked to either the autonomous or cytokine-induced proliferation in vitro. As the leukemic cobblestone area-forming cell frequencies also revealed no correlation with repopulation in the NOD/SCID model, we consider it very likely that the level of engraftment reflects the in vivo proliferative ability of the AML samples assayed rather than the number of leukemia-initiating cells infused into the NOD/SCID mice. Phenotypic analysis based on the expression of CD33, CD34 and CD38 before and after passage in NOD/SCID showed that in 10 out of 16 samples investigated phenotypes were different. In summary, in addition to the Flt3 internal tandem duplications we have identified a series of clinical parameters that determine the NOD/SCID repopulating ability of AML samples, whilst our data strongly suggest that AML in NOD/SCID does not reflect the leukemic process in the patient.