Chromosomal translocations represent frequent events in leukemia. In t(8;21)+ acute myeloid leukemia, RUNX1 is fused to nearly the entire ETO protein, which contains four conserved nervy homology regions, NHR1-4. Furthermore RUNX1/ETO interacts with ETO-homologous proteins via NHR2, thereby multiplying NHR domain contacts. As shown recently, RUNX1/ETO retains oncogenic activity upon either deletion of the NHR3 + 4 N-CoR/SMRT interaction domain or substitution of the NHR2 tetramer domain. Thus, we aimed to clarify the specificities of the NHR domains. A C-terminally NHR3 + 4 truncated RUNX1/ETO containing a heterologous, structurally highly related non-NHR2 tetramer interface translocated into the nucleus and bound to RUNX1 consensus motifs. However, it failed to interact with ETO-homologues, repress RUNX1 targets, and transform progenitors. Surprisingly, transforming capacity was fully restored by C-terminal fusion with ETO's NHR4 zinc-finger or the repressor domain 3 of N-CoR, while other repression domains failed. With an inducible protein assembly system, we further demonstrated that NHR4 domain activity is critically required early in the establishment of progenitor cultures expressing the NHR2 exchanged truncated RUNX1/ETO. Together, we can show that NHR2 and NHR4 domains can be replaced by heterologous protein domains conferring tetramerization and repressor functions, thus showing that the NHR2 and NHR4 domain structures do not have irreplaceable functions concerning RUNX1/ETO activity for the establishment of human CD34+ cell expansion. We could resemble the function of RUNX1/ETO through modular recomposition with protein domains from RUNX1, ETO, BCR and N-CoR without any NHR2 and NHR4 sequences. As most transcriptional repressor proteins do not comprise tetramerization domains, our results provide a possible explanation as to the reason that RUNX1 is recurrently found translocated to ETO family members, which all contain tetramer together with transcriptional repressor moieties.