Improved Patient-Derived Xenograft Model for Acute Myeloid Leukemia

We previously demonstrated that the NSG (NOD-Scid-IL2Rgc^null) xenotransplantation is an improved model for human AML samples, allowing us to better understand and characterize AML biology, especially in the context of drug therapy studies (Sanchez et al., Leukemia 2009). However, we observed that approximately half of AML patients’ samples either did not engraft in NSG mice (based on <0.1% human blasts in mouse bone marrow) or showed low (0.1 to 1% blasts) and highly variable engraftment. Recently, NSG mice expressing human SCF, GM-CSF and IL-3 transgenes (NSG-S) have been reported to enhance the engraftment of normal hematopoietic stem cells and primary AML cells, although only a few AML patients were evaluated (Wunderlich M et al, Leukemia 2010). This report describes a comprehensive paired analysis of engraftment of AML samples in NSG and NSG-S mice. T-cell depleted AML cells (5 -10 x 10⁶ per mouse) were injected intravenously in sub-lethally irradiated mice (n=5/AML sample). Leukemia engraftment was assessed up to 16 weeks after injection in peripheral blood (PB), spleen (SPL) and bone marrow (BM) based on the percentage and absolute number of human leukemic blasts (huCD45⁺CD33^+/-CD3^-) in each tissue. Samples from 71 AML patients, representing all FAB and prognosis groups, were injected in NSG mice and only 35 samples (49%) engrafted based on human blasts >0.5% in mouse BM. From these 35 NSG-engrafting samples, 14 were also injected in NSG-S mice and all of them engrafted. Leukemic burden was significantly (p ≤ 0.05) increased in NSG-S versus NSG mice: 39±21% vs 22±23% BM blasts, 21±15% vs 7±10% SPL blasts, 2,732±6,488 vs 141±221 blasts/ml PB. Interestingly, the dramatic increase in peripheral blast count observed in NSG-S mice provides new opportunities to use PB to monitor drug response for the many patient samples that show no or very low peripheral engraftment in NSG mice. However, for 7 of these 14 NSG-engrafting AML samples, the use of NSG-S mice as recipients was associated with rapid engraftment, excessive leukemic burden, anemia, weight loss and lethargy requiring early sacrifice and leading to shorter overall survival (54±26 days in NSG-S vs >90 days in NSG). Out of the 36 patients’ samples that failed to engraft in NSG mice, 19 were tested for engraftment in NSG-S mice. Remarkably, 14 out 19 (74%) samples engrafted (17±16% BM blasts, 8±12% SPL, and 1,418±4,609/ml PB blasts at Day 77 post-transplant) and the kinetics of engraftment were slower compared to AML samples that can engraft in both mouse strains. These results suggest that the presence of human SCF, GM-CSF and IL-3 in NSG-S is sufficient to rescue leukemia-initiating cells for most AML samples that fail to engraft in NSG mice. Only 5 out of 33 samples (15%) failed to engraft in NSG and NSG-S mice, indicating that the NSG-S BM microenvironment remains suboptimal for a small minority of AML samples. We are investigating if NSG-S engraftment is correlated to CD116, CD117, CD123 expression, cytogenetics, mutations, and prognosis. Overall, our results show that NSG-S mice represent a significant improvement over previous patient-derived xenograft models since they can (1) accelerate and enhance leukemic engraftment compared to NSG mice, and (2) support engraftment for 85% of our AML patients, making this model particularly useful for pre-clinical studies.