Aneuploidy and structural aberrations of chromosome 21 (Hsa21) are most frequent in acute myeloid leukemia. However, it remained largely unknown, why leukemic blasts select for amplifications of Hsa21 or parts of it. We used myeloid leukemia associated with Down syndrome (ML-DS) as a model to understand the oncogenic properties of trisomy 21. We performed a CRISPR-Cas9 loss-of-function screening targeting the 218 currently annotated coding genes on Hsa21, which revealed a strong and specific dependency of ML-DS on RUNX1. Interestingly, the short isoform of RUNX1 (RUNX1A) was upregulated in leukemic blasts from ML-DS patients, leading to disequilibrium of the different RUNX1 isoforms. Modulation of the equilibrium in human CD34+ hematopoietic stem and progenitor cells (HSPCs) and ML-DS patient blasts showed that high levels of RUNX1A interfered with megakaryocytic differentiation. Functional studies supported the interplay of the Gata1s mutation – pathognomonic for ML-DS – with RUNX1A overexpression resulting in proliferation and accumulation of immature megakaryocytic progenitors in vitro and ML-DS-like leukemia in vivo. Molecularly, RUNX1A replaces RUNX1C from its endogenous binding sites and alters GATA1 chromatin occupancy, thereby synergizing with GATA1s in inducing oncogenic programs and perturbing normal differentiation, which could be reverted upon restoring the normal RUNX1A:RUNX1C equilibrium. By providing strong evidence for the crucial role of RUNX1A in the initiation of a leukemic state in ML-DS, our study highlights the importance of alternative splicing in leukemogenesis and paves the way towards development of specific and targeted therapies in ML-DS, but also other types of leukemia with numerical changes of Hsa21 or RUNX1 isoform disequilibrium.