Fetal Differentiation Programs Afford a Protective Barrier to NUP98 Fusion-Driven AML That Dissipates Shortly after Birth

Mutations that drive pediatric acute myeloid leukemia (AML) are often age-restricted. NUP98 rearrangements (NUP98r) cause high-risk AML primarily in early-to-mid childhood. NUP98r are rare in adult AML. Furthermore, AML almost never occurs before birth, even though underlying translocations can occur prenatally. This pattern suggests that there are developmental windows and associated molecular programs that either promote or suppress the transforming effects of driver mutations such as NUP98r. This model raises the question as to why certain mutations, such as NUP98r, associate disproportionately with AML in early/mid childhood and why fetal leukemias are exceedingly rare. Does the paucity of fetal AML simply reflect high integrity of the genome during ontogeny, or does it indicate the existence of programs that actively suppress transformation prior to birth but then dissipate after birth?

To address these questions, we developed a novel strategy to model NUP98r AML initiation during fetal, neonatal, juvenile, and adult stages of life. We generated mouse induced pluripotent stem cells (iPSCs) that give rise to chimeric mice capable of expressing diverse NUP98 fusion proteins specifically in the blood and at specific stages of ontogeny and aging. Specifically, we have engineered cells to express three common fusions – NUP98::HOXA9 (NHA9), NUP98::NSD1, and NUP98::KDM5A. In some cases, we have introduced cooperating mutations through targeted mutagenesis of the iPSCs or the chimeric blood cells. This strategy allows us to efficiently model genetically diverse pediatric AML in situ. It bypasses complex mouse breeding while preserving normal blood ontogeny.

We tested whether NHA9 initiates AML more efficiently in neonatal or juvenile mice than in fetal mice, as might be expected. We activated NHA9 at embryonic day 10.5 (E10.5) or postnatal day 21 (P21). We introduced a cooperating Flt3-internal tandem duplication (Flt3^ITD) to progenitors harvested at E18 (after fetal induction) or P28 (after postnatal induction) to test whether leukemogenic potential varies with age. NHA9-expressing juvenile progenitors gave rise to AML with high penetrance whereas fetal progenitors did not. Moreover, NHA9-expressing fetal progenitors demonstrated limited repopulating activity in transplantation assays, in contrast to juvenile progenitors which engrafted well. Leukemias that arose from juvenile cells adhered to a highly reproducible differentiation hierarchy with an isolatable leukemia stem cell population that closely resembled a population found in human NUP98r AML. Juvenile progenitors are therefore susceptible to NHA9-driven transformation, and fetal identity conveys protection against NHA9-driven leukemogenesis.

We tested whether NHA9 induces distinct patterns of transcription and epigenome organization in fetal and juvenile progenitors that could account for the discrepant leukemia susceptibilities. Specifically, we performed Cellular Indexing of Transcriptomes and Epitopes by sequencing (CITE-seq) and single-cell ATAC-sequencing (scATAC-seq) on NHA9-expressing progenitor cells after fetal or postnatal induction. Fetal NHA9 induction drove erythroid differentiation and dramatically reprogrammed progenitor epigenome, making NHA9-expressing cells transcriptionally and epigenetically discrete from normal fetal hematopoiesis. In contrast, postnatal NHA9 induction did not alter hematopoietic differentiation trajectories at either the transcriptome or epigenome levels despite anticipated changes in Hox gene expression. Thus, NHA9 superimposes ectopic self-renewal capacity and malignant potential on an otherwise normal myeloid differentiation program in a juvenile context, whereas it drives erythroid differentiation at the expense of malignant potential in a fetal context.

Our data suggest that age greatly influences the capacity of oncoproteins to drive leukemic transformation. Moreover, the data show that leukemogenesis is actively suppressed in fetal contexts as a consequence of age-specific differentiation programs rather than simply via enhanced genome fidelity. This may explain why NUP98r AML is exceedingly rare before birth, if it occurs at all.