Genetic drivers of acute myeloid leukemia (AML) include chromosomal translocations and somatic mutations involving regulators of epigenetic, transcriptional, and cell-signaling pathways. Many of these mutations cannot be directly targeted with pharmacological agents. To circumvent the problem, a better functional characterization of the leukemic phenotype would provide additional therapeutic options. Murine in vivo models of AML, developed by introducing human AML mutations in murine hematopoietic progenitors, can recapitulate many features of the disease in a reproducible way. By coupling these models of AML with perturbations of selected genes, we can rapidly validate hypotheses regarding the growth requirements of leukemic cells in vivo. Specifically, these experiments can identify essential genes for AML even when these genes are not mutated. Hence, an experimental understanding of AML is complementary to the sequencing of human mutations.

Pooled genetic perturbation screens performed in experimental models of malignancies are an efficient approach to identify genes essential to malignant cells. However, most screens reported so far have been performed in vitro, where cell growth conditions differ significantly from the malignant milieu in vivo. Here, we report the development of experimental and statistical tools that allowed us to identify positive and negative regulators of AML growth on a genomic scale in vivo. We created Cas9-expressing murine models of AML (mAML), amenable to genetic manipulation, by overexpressing the human AML (hAML) oncogenes HOXA9 and MEIS1, or the fusion protein MLL/MLLT3, in murine myeloid progenitors isolated from a Rosa26-Cas9-knockin mouse. While MLL translocations account for approximately 3-10% of AML cases, HOXA9 and MEIS1 are overexpressed in 70% of cases. Thus, these mAML tools collectively modeled a significant fraction of hAML. Cells were transduced ex vivo with genome-wide sgRNA libraries containing 130,209 sequences then transplanted in cohorts of recipients. The depletion in vivo of sgRNAs was compared to non-targeting library controls in the reconstituted leukemia to assess statistical significance. This gave us a genome-wide functional annotation of the leukemic genome in vivo. To validate the effect of the top 1034 genes identified in the screen, we created a validation library and tested its effect on leukemic cells and normal hematopoietic progenitors to identify lead candidates.

Several genetic pathways showed greater levels of depletion in vivo compared to the in vitro arm of the screen. Among these was oxidative phosphorylation, suggesting that AML cells rely on it to a greater extent in vivo. We then focused on single genes which disruption in vivo had a much larger effect. As expected, we identified several immune regulators, partners in integrin-mediated adhesion and, interestingly, two regulators of beta-galactosylation, galactose epimerase (Gale) and beta-1,4-galactosyltransferase 1 (B4galt1). Genetically ablating this pathway, using CRISPR/Cas9, in AML cells had a statistically significant effect on survival of mice. Engraftment of AML cells was markedly reduced. Interestingly, the growth rates of leukemic or normal hematopoietic progenitors in vitro was not different between knock-out and control. Beta-galactosylation is known to post-translationally modify several proteins involved in homing and cell adhesion. Thus, we propose that altering beta-galactosylation in AML cells is a novel strategy to modulate their interaction with the bone marrow microenvironment for therapeutic benefit, which could be complementary to current approaches.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.

Sign in via your Institution