Development of synthetic introns for selective elimination of U2AF1 mutant MDS

Abstract Mutations affecting the RNA splicing factors SF3B1, U2AF1, and SRSF2 are the most common class of genetic alterations in MDS. Patients with high-risk MDS have a median survival of 1-2 years and there is great need to develop novel therapies for MDS. Recent studies have engineered synthetic RNAs responsive to mutations in SF3B1 to express therapeutic payloads specifically in SF3B1 mutant cells and preferentially eliminate them. However, this approach relied on intronic sequences recognized by mutant SF3B1 and it is unclear if this strategy can be applied to higher risk mutations such as those in U2AF1, as U2AF1 is affected by both intronic and exonic sequence. To develop synthetic RNAs responsive to U2AF1 mutations, we identified endogenous genes that respond to the most common U2AF1mutations (those at the S34 residue). Analysis of RNA-seq data from 263 MDS/AML patients with U2AF1 S34 mutations or without any splicing factor mutations as well as isogenic AML cells with or without knockin of U2AF1 S34F or S34Y mutations, revealed a series of 3' splice sites consistently aberrantly spliced in U2AF1 mutant cells. This change in 3' splice site usage was validated in three mRNAs- MTA1, PRUNE1, and TRMT13- where U2AF1S34F/Y mutant MDS cells utilized a distal 3' splice site compared to U2AF1 wild-type (WT) cells. The entire endogenous intron of each of these three mRNAs and its alternative 3' splice sites were inserted to interrupt the cDNA sequence encoding the suicide gene HSV-TK. HSV-TK is inert in cells until the prodrug ganciclovir is provided which leads to cytotoxic metabolite production. Expression of the MTA1-HSV-TK construct in U2AF1S34F/Y mutant cells yielded full-length HSV-TK mRNA and protein. In contrast, U2AF1 WT cells expressed an incompletely spliced, non-functional mRNA and, as such, only U2AF1 mutant cells were killed with ganciclovir treatment while U2AF1 WT cells were unaffected. To enable massively parallel screening, optimize mutant selectivity, and allow for in vivo delivery, we next reduced the size of the natural 315 base pair MTA1 intron by ligating the minimal sequences at the 5' and 3' end to form a new 208 base pair synthetic intron. To improve mutant selectivity, we utilized SpliceAI, a deep learning model for splicing prediction. We evaluated 624 variants in silico using SpliceAI and identified intron variants that strengthen the intron-proximal 3' splice site in WT cells while reducing usage of the distal 3' splice site. Experimental validation identified a synthetic MTA1 intron allowing preferential killing of U2AF1S34F/WT and U2AF1S34Y/WT cells compared to WT cells. To further improve efficacy of the synthetic intron in mutant cells, we performed a massively parallel functional screen with 12,000 variant introns as a pool in U2AF1WT and U2AF1S34Y mutant K562 cells for unbiased identification of variants that eliminate mutant cells while leaving WT cells unaffected. We evaluated 6 of the identified introns and found one which yielded optimal ganciclovir-mediated killing in U2AF1 mutant cells while sparing WT cells. When isogenic WT or U2AF1 mutant AML cells bearing this construct were introduced into NSG mice, ganciclovir treatment improved survival of animals engrafted with U2AF1 mutant cells by a median of 20 days relative to the untreated group (p=0.0015; ganciclovir did not alter survival of mice engrafted with U2AF1 WT cells). We next evaluated lipid nanoparticles (LNPs) for in vivo delivery of U2AF1 mutant-selective synthetic introns. While LNPs are a clinically advanced RNA delivery vehicle, they yield minimal import of RNA cargo into the nucleus (the site of RNA splicing). We therefore tested delivery of plasmid DNA (allowing for nuclear import) in LNP formulations incorporating lipids that suppress STING-mediated inflammation which previously limited plasmid DNA delivery. This approach enabled nuclear expression in myeloid leukemia cells and is now being tested in vivo. This study identified a novel precision gene therapy targeting cells with MDS-associated mutations in U2AF1 by harnessing the neomorphic splicing activity of U2AF1 mutations. Moreover, the synthetic introns responsive to U2AF1 mutations created here enable future mechanistic studies to identify cis and trans factors required for mutant U2AF1 function. Finally, we investigate means for in vivo delivery of plasmid DNAs to myeloid cells which could have broader applications in gene therapy for myeloid neoplasms.