Favorable-risk acute myeloid leukemia (AML) constitutes up to 40% of newly diagnosed AML cases. However, only 66% of younger and 33% of older patients are alive 3 years post diagnosis. Studying the biology of these entities in vivo has been inherently difficult: in fact, faithful xeno-engraftment of human AML in immunodeficient mice has been limited to higher-risk AML.

To address this limitation we hypothesized that a human myelopoiesis-supportive environment may permit xeno-engraftment of favorable-risk AML: we transplanted primary human AML with isolated NPM1 mutation and AML with inv(16) into knock-in (KI) mice with human versions of genes encoding M-CSF, IL-3, GM-CSF and Thrombopoietin. Newborn MISTRG mice (Rongvaux et al. Nat Biotechnol. 2014) were sub-lethally irradiated, intra-hepatically injected with AML blasts and analyzed 16-24 weeks post transplantation.

Flow cytometric analysis of bone marrow (BM) and blood of cytokine KI versus NSG mice revealed profound differences: Average engraftment was 36% in the BM of cytokine KI mice compared to only 10% in NSG mice across all patient samples transplanted. In line with engraftment levels, engraftment efficacy was higher in cytokine KI mice. Importantly, average engraftment in cytokine KI mice approached BM and peripheral blast count observed in human AML patients, while NSG engraftment levels were significantly lower. AML engraftment in cytokine KI mice was found to be independent of expression of the SIRPα transgene in mice expressing all four cytokines. NMP1mut AML showed a trend towards higher BM engraftment in MI(S)TRG versus NSG mice. Expression of the NPM1mutallele was more than 3-fold higher in MI(S)TRG mice compared to NSG mice. Performing targeted next generation sequencing in human and engrafted murine BM, we found comparable allelic frequency ranging from 38% to 51% for NPM1 mutation in a patient's leukemia and engrafted cytokine KI mice. Of note, even the composition of NPM1 mutation subclones was constant between patient and cytokine KI recipients.

In contrast to NPM1mut, a striking difference in quantitative inv(16) AML engraftment was observed between cytokine KI and NSG mice, both by aggregate and split-sample analysis. Serial transplantation further validated engraftment of an inv(16) leukemia-initiating cell population in cytokine KI mice. To dissect the contribution of respective single human KI cytokines to inv(16) AML propagation, we analyzed engraftment in Rag2-/-gc-/- mice carrying, in addition to a human SIRPα transgene, further KI genes for human thrombopoietin (STRG), M-CSF and Thrombopoietin (MSTRG), and M-CSF, IL-3, GM-CSF and Thrombopoietin (MI(S)TRG). STRG mice, as NSG mice, did not support engraftment. In contrast, addition of M-CSF in MSTRG mice led to robust engraftment of inv(16) AML, while addition of IL-3 and GM-CSF in MI(S)TRG mice resulted in only marginal, if any, improvement over MSTRG mice. Of note, knock-in of human M-CSF has previously been shown to be sufficiently cross-reactive to rescue the murine M-CSF knock-out phenotype (Rathinam et al. Blood 2011). We thus reasoned that the engraftment advantage observed in MSTRG mice was likely due to human M-CSF effects on blasts and not due to reduction in mouse macrophages. Together, this establishes a central role for M-CSF in maintenance of inv(16) AML.

When profiling cytokine receptor expression for TPO, M-CSF, IL-3 and GM-CSF on inv(16) AML blasts in comparison to NPM1mut AML and an aggregate of non-favorable-risk AML cases, using microarray data from an international cohort study (Wouters et al. Blood 2009), inv(16) AML showed significantly higher expression of each of the 4 receptors compared to NPM1mut AML, and significantly higher expression of M-CSFR and GM-CSFR compared to non-favorable-risk AML; the overall highest difference in expression was observed for M-CSFR. Moreover, gene set enrichment analysis demonstrated enrichment of M-CSF-inducible genes in inv(16) AML compared to NPM1mutAML and non-favorable-risk AML cases, confirming increased signaling through the M-CSF pathway.

In sum, we here describe the first faithful and highly efficacious xenotransplantation model of favorable-risk, and specifically, inv(16) AML. We further demonstrate a strong dependency of this subtype on M-CSF, a finding that might provide a basis for therapy optimization studies informed by an improved understanding of favorable-risk AML biology.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.