SF3B1 mutations constitute a novel therapeutic target in breast cancer

Abstract

Mutations in genes encoding proteins involved in RNA splicing have been found to occur at relatively high frequencies in several tumour types including myelodysplastic syndromes, chronic lymphocytic leukaemia, uveal melanoma, and pancreatic cancer, and at lower frequencies in breast cancer. To investigate whether dysfunction in RNA splicing is implicated in the pathogenesis of breast cancer, we performed a re-analysis of published exome and whole genome sequencing data. This analysis revealed that mutations in spliceosomal component genes occurred in 5.6% of unselected breast cancers, including hotspot mutations in the SF3B1 gene, which were found in 1.8% of unselected breast cancers. SF3B1 mutations were significantly associated with ER-positive disease, AKT1 mutations, and distinct copy number alterations. Additional profiling of hotspot mutations in a panel of special histological subtypes of breast cancer showed that 16% and 6% of papillary and mucinous carcinomas of the breast harboured the SF3B1 K700E mutation. RNA sequencing identified differentially spliced events expressed in tumours with SF3B1 mutations including the protein coding genes TMEM14C, RPL31, DYNL11, UQCC, and ABCC5, and the long non-coding RNA CRNDE. Moreover, SF3B1 mutant cell lines were found to be sensitive to the SF3b complex inhibitor spliceostatin A and treatment resulted in perturbation of the splicing signature. Albeit rare, SF3B1 mutations result in alternative splicing events, and may constitute drivers and a novel therapeutic target in a subset of breast cancers.

Keywords: SF3B1; alternative splicing; breast cancer; drivers; next-generation sequencing; spliceostatin A.

Figure 1

Repertoire of mutations in genes involved in mRNA splicing in breast cancer. (A) Summary of the mutations identified in genes involved in mRNA splicing. Genes are listed on the right-hand side of the diagram and the breast cancer samples across the top. Mutations show a mutually exclusive pattern indicating functional redundancy. Mutation data retrieved from TCGA plus published studies (Nik-Zainal et al [3], Banerji et al [4], Ellis et al [5], TCGA [6], Shah et al [7], and Stephens et al [8]). (B) Distribution and frequency of SF3B1 mutations derived from re-analysis of publicly available massively parallel sequencing data from the TCGA breast cancer study and additional studies (as above). Note that mutations are clustered in the HEAT domain (exons 14–16) of the protein with a hotspot point mutation at amino acid 700 (K700E). (C) Histological and molecular status of SF3B1 mutant samples from the re-analysis. Co-mutation analysis shows PIK3CA missense mutations in 47.8% and AKT1 mutations in 17% of the samples, while mutations in other known driver genes are present at lower frequencies. Note the lack of additional significantly mutated genes or known cancer genes in the two mucinous carcinomas (curated from TCGA, http://www.cbioportal.org/public-portal/study.do?cancer_study_id=brca_tcga_pub, and the Cancer Gene Census, https://cancer.sanger.ac.uk/cosmic). IC-NST = invasive carcinoma of no special type. (D) Repertoire of somatic mutations in 19 papillary carcinomas of the breast, as defined by targeted re-sequencing of known cancer genes. The genes are listed on the left-hand side of the diagram and the breast cancer samples across the top. Two of the SF3B1 mutant tumours displayed additional mutations affecting GATA3 and MAP3K4.

Figure 2

Identification of differential splicing in SF3B1 mutant (n = 3) and wild-type (n = 11) papillary carcinomas of the breast. Plots of normalised RNA-sequencing reads for (A) CRNDE, (B) TMEM14C, and (C) RPL31 in SF3B1 wild-type (blue) and SF3B1 mutant (red) tumours. Schematic representations of the exon structures are shown above the graph, with exons represented by boxes. Differentially spliced exon bins are indicated by lighter coloured shading. Box plots representing the validation of the differential splicing event in wild-type (blue) versus mutant (red) tumours by quantitative RT-PCR.

Figure 3

SF3B1 mutant cells are sensitive to SF3b inhibition. (A) Dose–response curves of SF3B1 mutant cancer cell lines (red) or wild-type (blue) to the SF3b complex inhibitor spliceostatin A (p = 0.0118, t-test). (B) Bar plots of the relative expression of alternative spliced transcripts in ESS-1 cells upon 12 h spliceostatin A (SSA) treatment (SF30 concentration) or SF3B1 siRNA-mediated silencing relative to vehicle (DMSO) or non-targeting controls (siCON). Error bars represent the standard error of the mean of three replicates. (C) Bar plots illustrating the normalised percentage inhibition (NPI) relative to siCON negative and ubiquitin B positive controls of cells following transfection with SMARTpool siRNA of ABCC5, ANKHD1, DYNLL1, F8, RPL31, TMEM14C, UQCC, and CRNDE in SF3B1 mutant (red) or wild-type (blue) cancer cells. Error bars represent the standard error of the mean of three replicates. None of the genes showed a significant difference in cell viability between mutant and wild-type cells (p > 0.05, t-test). (D) Bar plots illustrating the normalised percentage inhibition (NPI) relative to siCON negative and ubiquitin B positive controls of cells following transfection with custom siRNA oligos against the CRNDE alternative spliced isoform (eb6) in SF3B1 mutant (red) or wild-type (blue) cancer cells. Error bars represent the standard error of the mean of three replicates. None of the genes showed a significant difference in cell viability between mutant and wild-type cells (p > 0.05, t-test). (E) Bar plot showing qPCR validation of siRNA oligos against the CRNDE alterative spliced isoform (eb6) relative to siCON.