Cancer-Associated Mutations in SF3B1 Exhibit Neomorphic Splicing Activity and Block Erythroid Differentiation

Recently, heterozygous mutations in several spliceosome genes have been observed in hematological and solid cancers, but their functional role in these diseases is not well understood. Among these, SF3B1 is the most commonly mutated spliceosome gene in myelodysplastic syndromes (MDS) and chronic lymphocytic leukemia (CLL). SF3B1 is part of the U2 complex involved in the recognition of the 3’ splice sites (3’ss) during early spliceosome assembly.

To determine the impact of SF3B1 mutations, we compared RNAseq profiles from tumor samples with SF3B1 hotspot mutations (SF3B1^MUT) or wild-type SF3B1 (SF3B1^WT) in breast cancer, melanoma, CLL and MDS. This analysis revealed significantly increased usage of aberrant 3’ss in SF3B1^MUT samples. In addition, the aberrantly spliced exons carry a proximal splice acceptor (SA) 15 to 21 nucleotides upstream of the canonical SA with a weak and short polypyrimidine tract. Using ectopic expression and allele-specific RNAi, we confirmed that mutations in SF3B1 are sufficient and required for these aberrant splicing events which suggests a neomorphic splicing activity of SF3B1^MUT. Furthermore, a common aberrant splicing profile was shared across different hotspot mutations and diseases; however, unique aberrant splicing profiles were also observed in each disease suggesting lineage and disease specific effects. In particular, gene-set enrichment analysis of aberrantly spliced and differentially expressed genes in mutant vs. wild type samples identified genes that regulate cell differentiation and epigenetics in MDS, pathways/processes known to be dysregulated in myeloid malignancies. To study the impact of SF3B1^MUT on differentiation processes, we used the well-established TF-1 model of erythroid differentiation. SF3B1^K700E (the most common mutation in MDS and CLL), SF3B1^G742D (a mutation found in CLL but not MDS patients), SF3B1^K700R (a mutation unable to induce aberrant splicing) and SF3B1^WT were overexpressed in TF-1 to study erythoid differentiation post erythropoietin (EPO) exposure. EPO treatment, as expected, induced differentiation in TF-1 cells transduced with SF3B1^WT and SF3B1^K700R. Consistent with a possible mechanism in MDS, SF3B1^K700E transduction blocked differentiation of TF-1 cells. Intriguingly, SF3B1^G742D, which is found mutated in CLL but not MDS, did not block differentiation in this myeloid differentiation model, implying that specific SF3B1 mutations and splicing aberrations have important context dependent effects. Ongoing studies comparing splicing aberrations induced by SF3B1^K700E and SF3B1^G742D in TF-1 cell differentiation will be described.

Finally, we evaluated a potent and selective modulator of SF3B1 that inhibits both canonical and neomorphic splicing activities in vitro and in vivo. The SF3B1 modulator induced tumor regression in SF3B1^MUT xenografts and increased the overall survival of animals bearing SF3B1^MUT xenografts at well tolerated doses. Taken together, our data suggest that SF3B1 mutations impair cell differentiation and that splicing modulators hold promise for the treatment of cancers with SF3B1 mutations, including CLL and MDS.