SWI/SNF Dysregulation through a Prion-like Domain Causes AML

Meningioma-1 (MN1) was first described in a sporadic t(4;22) translocation in meningioma. Rare MN1 fusions to the ETS factors ETV6 (t(12;22)) and FLI1 (cryptic) also occur in AML. t(12;22) translocations have been described in two different patterns:Type 1: MN1 exon 1, is fused to a C-terminal fragment of ETV6. The first exon codes for almost all of the protein and has been shown previously to be sufficient to induce AML in a mouse model. Type 2: The entire MN1 coding frame, including the stop codon, is fused to a portion of ETV6. This results in a fusion on a DNA level, but on a protein level only MN1 is expressed. In both cases the fusion results in very high MN1 expression, either alone (type 2) or as a fusion protein (type 1). We performed ChIP-seq for H3K4me2/H3K27ac/Med1 on cell lines and a primary patient sample, and identified a large putative enhancer within and downstream of the ETV6 locus. Type 2 translocations suggest that hijacking of this enhancer is the critical oncogenic event. Besides the rare fusions, a subgroup of AML patients have very high MN1 expression without evident fusions. We speculate that some of these patients may have cryptic enhancer fusions.

Importantly, several independent studies show that MN1 overexpression confers a poor prognosis. The survival rate 2 years after diagnosis is at only 20-30% reflecting the aggressiveness of this leukemia.

In mice MN1 overexpression induces one of the most aggressive leukemias known as a single hit. These leukemias are Hoxa9 high and transcriptionally resemble KMT2A-rearranged leukemias.

Despite its clear contribution to aggressive AML, it is not understood how MN1 functions on a molecular level. MN1 has no identified classic structural domains and lacks sequence homology with any other protein. Therefore, no predictions about structure or possible binding partners exist, and only few binding partners have been shown experimentally. This severely limits therapy options for patients and potential future drug development. Therefore, we aimed to define the MN1 interaction partner(s) and mechanism of leukemogenesis. To identify the MN1 interactome in AML we used two complementary methods, co-immunoprecipitation (CoIP) and proximity-dependent labeling (BioID), followed by Mass Spectrometry. As top hit in both screens we identified the mSWI/SNF complex, including the ATPase Smarca4, as an interactor of MN1. mSWI/SNF is a multisubunit complex with cell context- and function- dependent variable members. This complex is responsible for chromatin remodeling and plays an important role in gene expression and lineage determination. Its role in various forms of cancer is well established, where subunits are deleted, mutated, or misrecruited. We find co-sedimentation of MN1 with identified mSWI/SNF members in glycerol gradients using murine and human cells with MN1 overexpression. ChIP-seq data indicates a high overlap in DNA occupancy between MN1 and Smarca4. Using a conditional Smarca4 KO mouse model we show that Smarca4 is indispensable for MN1 driven leukemia. Together, these experiments substantiate a critical interaction between the oncogenic driver MN1 and the epigenetic modifier complex mSWI/SNF.

MN1 contains a long polyQ stretch encoded by 28 CAG repeats. Such glutamine rich regions have been recognized as domains facilitating transcriptional activation, stabilizing protein-protein interactions, and being important components in higher order complex formation. PolyQ domains belong to the family of prion-like domains, which have roles in SWI/SNF recruitment. We show that the deletion of MN1's polyQ stretch abolishes differentiation block and allows the cells to differentiate. This is reflected in poor replating efficiency in Methylcellulose assays compared to full length MN1 driven leukemia cells. In vivo, MN1s' polyQ stretch is important for AML initiation. On a molecular level, we find that polyQ deletion reduces the affinity of the mSWI/SNF complex to chromatin in comparison to full length MN1, and fails to maintain the expression of key MN1 target genes such as the later Hoxa cluster, Meis1 and Flt3.

In conclusion, our data support a model wherein MN1's oncogenic function is mediated by mSWI/SNF dysregulation, via the MN1 polyQ stretch.