Modelling the Progression of a Preleukemic Stage to Overt Leukemia in Children with Down Syndrome

Myeloid leukemia of Down syndrome (ML-DS) is a tractable human model of acute myeloid leukemia. A preleukemia phase, transient abnormal myelopoiesis (TAM) and silent TAM, occurs in 28% of neonates with Down Syndrome (Roberts et al. Blood 2013). TAM is caused by trisomy 21 and acquired mutations in GATA1 that result in a N-terminal truncated protein, GATA1s, in hematopoietic stem and progenitor cells (HSPCs) of fetal origin. ML-DS evolves from TAM by acquisition of additional genetic lesions. The nature of these lesions and the mechanism of transformation are incompletely understood.

We performed exome sequencing and targeted resequencing of 141 ML-DS and 111 TAM patients to characterize the evolving mutational landscape from TAM to ML-DS. On average 1.6 acquired mutations were detected in ML-DS (in addition to GATA1 mutations), significantly more than in TAM (0.4 mutations per sample). Additional anticipated loss-of-function mutations acquired in ML-DS mainly affected cohesin components including CTCF (43% of patients), PRC2 components (13%), KANSL1 and other epigenetic regulators (14%). Conversely, anticipated gain-of-function mutations were most prevalent in signaling pathways, e.g. JAK kinases, MPL, KIT and RAS family members (40%). Importantly, we detected a novel recurrent hotspot mutation in 4% of patients (6/141 cases) in CSF2RB encoding the IL3-, IL5-, GM-CSF-receptor common beta chain. To test if the A455D/T variant in the CSF2RB transmembrane domain is a putative oncogenic driver, we ectopically expressed CSF2RB^A455D in TF1 cells. Cells expressing CSF2RB^A455D exhibited cytokine independent growth and STAT5 autonomous phosphorylation. In a CD34⁺-HSPC megakaryocytic differentiation assay, CSF2RB^A455D blocked terminal megakaryocytic differentiation whilst increasing proliferation by 30-fold (P=0.046). Moreover, the median survival of NSG mice transplanted with CSF2RB^A455DTF1 cells was shortened by 30 days compared to wild type TF1 cells (23 days compared to 53 days, P=0.0097).

To experimentally test the potential of loss-of-function mutations to transform TAM to ML-DS, we performed an in vivo murine isogenic transplantation screen using Gata1s expressing fetal hematopoietic cells from Cas9-knockin mice. We tested variants in 22 genes, recurrently detected in ML-DS, with a pool of prevalidated gRNAs. This resulted in short latency (n=18 mice; median survival 36 days) and high penetrance (100%) leukemia. Leukemia was not detected in mice infected with control gRNAs. Leukemias had a typical ML-DS megakaryoblastic phenotype (CD117⁺ and CD41a⁺). Amplicon sequencing revealed on average 2.9 gRNAs per leukemia and high representation (61% of all leukemias) of gRNAs directed to the tumor suppressor Trp53, which was alone sufficient to induce leukemia with 100% penetrance. When excluding the Trp53 gRNA from pools, leukemic cells from moribund mice contained gRNAs against negative regulators of the RAS and JAK-STAT signaling cascade, such as Nf1, Cbl and Sh2b3 (70% of the mice), Ezh2, Asxl1, Kdm6a,Bcor and other epigenetic modifiers (85%) or Ctcf (15%), closely resembling the mutational landscape of ML-DS. In contrast to ML-DS, gRNAs targeting cohesion components, such as Rad21 and Stag2, were not present in any of the leukemias.

In summary, we performed the largest genetic analysis of transforming events in ML-DS that cooperate with trisomy 21 and GATA1s and uncovered a previously undescribed activating mutation in CSR2B. We experimentally validated many of the loss-of-function mutations in a novel murine fetal leukemia assay for ML-DS. The field is now well-placed to study mechanisms of oncogenic cooperativity and identify novel therapeutic approaches for this leukemia.