AUTHORS
Franchesca Gabriela Franzen, BS, Hsiangyu Hu, PhD2, Nirmalya Saha, PhD3, Yuting Yang, PhD4, Venky Basrur, PhD5, Emma George6, Lauren Lachowski7, Alexey Nesvizhskii, PhD8, Zaneta Nikolovska-Coleska, PhD9, Andrew Muntean, PhD10
AFFILIATIONS
1 University of Michigan Medical School, Ann Arbor, Michigan, United States
2 University of Michigan Medical School, Ann Arbor, Michigan, United States
3 University of Michigan Medical School, Ann Arbor, Michigan, United States
4 University of Michigan Medical School, Ann Arbor, Michigan, United States
5 University of Michigan Medical School, Ann Arbor, Michigan, United States
6 University of Michigan Medical School, Ann Arbor, Michigan, United States
7 University of Michigan Medical School, Ann Arbor, Michigan, United States
8 University of Michigan Medical School, Ann Arbor, Michigan, United States
9 University of Michigan Medical School, Ann Arbor, Michigan, United States
10 University of Michigan Medical School, Ann Arbor, Michigan, United States
ABSTRACT
Acute myeloid leukemia (AML) remains a disease of urgent unmet need, with a five-year survival rate of about 33%. Epigenetic dysregulation is a major driver of AML, as evidenced by >70% of patients harboring a mutation in an epigenetic regulator (Fennell et al. Blood 2019, Ley et al. N Engl J Med 2013). ENL (MLLT1) and AF9 (MLLT3) are highly homologous (74% sequence identity) epigenetic reader proteins that share an identical domain architecture, including a C-terminal ANC-1 homology domain (AHD) that recruits the H3K79 methyltransferase DOT1L, the super elongation complex, and PRC1 components through mutually exclusive interactions. Despite this homology, ENL is required for AML cell growth while AF9 is dispensable, an unexplained divergence in function (Erb et al. Nature 2017, Calvanese et al. Nature 2019). ENL and AF9 are also fused to MLL (KMT2A), generating MLL::ENL and MLL::AF9, which are among the most common MLL fusion partners. Analysis of genomic breakpoints reveals divergent domain retention: ~84% of MLL::ENL patients retain the full-length protein, whereas MLL::AF9 patients predominantly retain only the AHD (Hu et al. Leukemia 2023, Meyer et al. Leukemia 2018). We previously showed this pattern is biologically significant: YEATS domain deletion abolishes MLL::ENL leukemogenesis in vivo. We hypothesized that these breakpoints reflect molecular and functional AHD differences that impact oncogenesis.
To address this, we generated retroviral MLL fusion constructs comprising only the terminal 90 amino acids (AHD) of ENL or AF9 (MLL::ENL-AHD and MLL::AF9-AHD) and transduced murine lineage-negative hematopoietic progenitor cells. In vivo leukemogenesis assays showed MLL::AF9-AHD induced aggressive leukemia (median latency ~100 days), while MLL::ENL-AHD failed to induce disease. Although insufficient to generate AML, MLL::ENL-AHD formed a viable murine cell line with reduced proliferative capacity relative to MLL::AF9-AHD. RT-qPCR showed MLL::ENL-AHD had significantly reduced activation of the direct targets Meis1 and Eya1 relative to MLL::AF9-AHD, while Hoxa9 activation was comparable. TMT-based proteomic analysis corroborated this, with MLL::ENL-AHD displaying 233 downregulated and 105 upregulated proteins compared to MLL::AF9-AHD, including Meis1 and Eya1 among downregulated targets. A global protein expression analysis stratified direct MLL-fusion targets into 2 classes: Class 1 (comparably activated in MLL::ENL-AHD and MLL::AF9-AHD cells, including Hoxa9/Hoxa10) and Class 2 (selectively activated by MLL::AF9-AHD but not MLL::ENL-AHD, including Meis1 and Eya1). ChIP analysis showed H3K79me2 was comparable in Class 1 genes but reduced at Class 2 loci in MLL::ENL-AHD cells. We hypothesized this reflects differences in AHD-DOT1L binding; fluorescence polarization assays showed AF9-AHD binds DOT1L peptides with higher affinity than ENL-AHD. Structure-guided analysis (Yang et al. Protein Sci 2024) identified 3 residues unique to the AF9-AHD at the DOT1L contact interface. Mutagenesis showed AF9-L504T reduced binding affinity to ENL-AHD levels. Introducing each substitution (L504T, I538V, C548F) into MLL::AF9-AHD impaired cellular proliferative capacity. These data suggest DOT1L-binding affinity alone does not fully explain AF9-AHD oncogenic activity, implicating additional AHD-mediated interactions in leukemic transformation.
To examine this, we analyzed ChIP-seq data from MLL-AF9 cells profiling the epigenetic landscape, PRC components and the MLL::AF9 fusion protein (Neville et al. Nature Cell Biology 2026). Analysis of Class 1 and Class 2 direct MLL::fusion target loci (described above) showed similar baseline H3K79me2 and MLL::AF9 occupancy but altered H3K27me3. These data suggest PRC activity at Class 1 and Class 2 genes is unique and may influence transcriptional activation potential by ENL and AF9 AHDs.
In summary, the ENL and AF9 AHDs possess divergent oncogenic potential in MLL::ENL- and MLL::AF9-driven leukemias. This difference is manifested by unique activation of 2 classes of direct MLL::fusion gene targets that track with differential epigenetic landscapes including H3K79me2 that may be influenced by differential PRC activity. The class-specific chromatin program identified here, together with differential oncogenicity associated with ENL and AF9 AHDs suggests strategies targeting multiple mechanisms may be needed to silence these programs.
DISCLOSURES
Franchesca Gabriela Franzen, BS: No relevant conflicts of interest to declare.