The Transcriptome Heterogeneity of MLL-Fusion ALL Is Driven By Fusion Partners Via Distinct Chromatin Binding

Chromosome rearrangements involving the Mixed Lineage Leukemia (MLL) gene on chromosome 11q23 account for 15-20% of acute lymphoid leukemia (ALL) and confer poor prognosis. Such rearrangements generate the MLL-fusion proteins, in which the N-terminus of the MLL protein mediating chromatin interactions is fused with one of more than 70 different partner proteins. Proteins that are frequently involved in MLL translocations, including AF4, ENL, AF9 and AF10, were identified as components of the super elongation complex (SEC) or DOT1L complex (an H3K79 histone methyltransferase). Based on these observations, a consensus model of MLL-fusion leukemogenesis has been proposed, which suggests that all fusion proteins bind to the targets of wildtype MLL and lead to the aberrant transcriptional elongation and H3K79 methylation via the recruitment of SEC or DOT1L complex and thus the uncontrolled activation of the target genes. Therefore, regardless of the nature of fusion partners, all MLL-fusion proteins work in a similar fashion by dysregulating the same pathways.

Our group has successfully established xenograft models of MLL-AF4 and MLL-AF9 B-ALL using human CD34+ hematopoietic stem and progenitor cells transduced with FLAG-tagged MLL-Af4 or MLL-AF9, which faithfully recapitulate the clinical features of the disease. We generated MLL-Af4 and -AF9 ALL using matched units of human CD34+ cells to directly test the consensus model. Interestingly, although having the same genetic background, the immunophenotype of the two ALL are unique, with CD34 expressed only in MLL-Af4 but not -AF9 cells. The transcriptomes of the two ALL were analyzed by RNAseq and dysregulated genes were defined by comparison with the transcriptome of normal pro-B cells (p≤0.05, fold-change≥2). Strikingly, only 40% of MLL-Af4-regulated genes overlap with those of MLL-AF9. This transcriptome heterogeneity is mirrored in clinical samples, where the gene signature generated from our model leukemia can be utilized to accurately classify patient samples in unsupervised hierarchical clustering analysis, with MLL-AF4 patient samples readily distinguishable from MLL-AF9 samples.

To identify the mechanisms accounting for this heterogeneity, we performed ChIP-seq analysis using anti-FLAG antibody to compare the chromatin occupancy of MLL-Af4 and MLL-AF9 in our model ALL cells. The MLL-Af4 ChIP-seq signal displayed a clear correlation with those of published MLL-AF4 ChIP-seq datasets from patient-derived cell lines, in the range of 70-90%, highlighting the faithfulness of our model. Surprisingly, MLL-Af4 shows a distinct genome-wide distribution compared to MLL-AF9, with only 20% of MLL-Af4 peaks and 35% of MLL-AF9 peaks overlapping. In contrast to MLL-Af4 which predominantly binds to promoter regions, MLL-AF9 has a relatively greater enrichment at intra- and inter-genic regions. Intriguingly, MLL-AF9 tends to bind at repetitive sequences in introns, suggesting these repetitive sequences may serve as regulatory elements for gene expression. Integration with RNAseq data reveals a significant association between differentially-expressed MLL-Af4 and -AF9 targets and specific chromatin binding of different MLL-fusion proteins. This data demonstrates that chromatin binding is not solely controlled by the MLL portion of the fusion protein and that differential target recognition of different fusion proteins is one molecular mechanism driving gene expression heterogeneity.

To test whether distinct co-factor recruitment by MLL-fusions adds another layer regulating gene expression heterogeneity beyond DNA binding, we purified the core complexes of MLL-Af4 and -AF9 from ALL cells by anti-FLAG immunoprecipitation and analyzed by mass spectrometry. These experiments identified both common and fusion-specific interacting proteins. MLL-Af4 showed a higher affinity with SEC component EAF2 but a lower affinity with DOT1L compared to MLL-AF9, suggesting that MLL-fusions have distinct associations with complex components which may achieve differential gene regulation.

In summary, our data question the consensus model of MLL-fusion leukemia and emphasize that MLL-fusion ALL is a heterogeneous disease. These findings have important implications for therapy development as each MLL-fusion leukemia could have its own Achilles' heel and customized therapy may need to be introduced for each type of disease.