Identification of a Novel Chromatin Structure Associated with the Transcriptional Response to Menin Inhibitors in AML

Menin is a member of the KMT2A/B methyltransferase complex that occupies a broad region around promoters in a small subset of genes typically associated with development and cell cycle control. Menin target genes can be essential dependencies in both solid and hematopoietic cancers, particularly in KMT2A-rearranged and NPM1-mutant acute myeloid leukemia (AML), in which it drives a stemness-associated transcriptional program critical for oncogenesis. Importantly, small molecule inhibitors blocking Menin’s interaction with KMT2A/B have demonstrated remarkable efficacy against AMLs in recent phase I/II clinical trials.

The Menin-KMT2A complex occupies a select set of less than 200 active promoters in leukemia cells, and Menin inhibitors affect a subset of these Menin-KMT2A bound loci. However, the mechanisms that determine Menin-KMT2A occupancy and the transcriptional sensitivity to Menin inhibition remain unknown. Using a convolutional neuronal network approach, we show that Menin genomic occupancy is associated with the presence of large, conserved AT-rich flanks surrounding the promoter region of loci such as MEIS1, PBX3, CDK6 and others. In some cases, the AT-rich regions extend multiple kilobases into the gene body and in all cases the loci have an easily recognizable CpG island near the transcriptional start site (TSS) where KMT2A binds. These AT-rich flanks are associated with Menin and LEDGF, but not KMT2A whereas the CpG islands are bound by KMT2A, Menin and LEDGF. Remarkably, CRISPR/Cas9 deletion of the AT-rich flanks leads to loss of Menin, KMT2A, and RNAPII occupancy at the associated CpG island as well as a major decrease in gene expression. We performed Micro Capture-C (MCC) to determine if the AT-Flanks and CpG islands were in physical proximity and thus might form a loop similar to an enhancer promoter interaction. Indeed, the CpG islands and the AT-Flanks are in close physical proximity despite in some cases being separated by multiple kilobases of DNA. However, these AT-rich regions are not typical enhancers, as ATAC-seq did not demonstrate chromatin accessibility and they are generally not associated with peaks in H3K27 acetylation. Short-term treatment with Menin inhibitors disrupted the interaction between AT-rich flanks and CpG islands, while leaving the looping between other canonical enhancers and the TSS intact. Of interest, the efficacy of Menin inhibitors in suppressing the expression of Menin targets is directly correlated with the size of AT-rich flanks, and genes with particularly large AT-rich flanks, such as MEIS1 and PBX3, are highly sensitive to Menin inhibition. Furthermore, in cells, the broad genomic occupancy of Menin and KMT2A is associated with the presence of KMT2A nuclear puncta as assessed by immunofluorescence of cell lines edited with CRISPR/Cas9 to express a tagged version of the KMT2A-fusion protein from the endogenous locus. These puncta disappear rapidly upon treatment with Menin inhibitors.

In summary, we propose that AT-rich flanks are a novel type of cis-regulatory element that supports the recruitment of Menin and KMT2A to a select group of loci that are primarily involved in developmental control. The relative scarcity of this promoter architecture in the genome may help explain the targeted transcriptional effects of Menin inhibitors.