The E592K variant of SF3B1 creates unique RNA missplicing and associates with high-risk MDS without ring sideroblasts

Abstract

Among the most common genetic alterations in myelodysplastic syndromes (MDS) are mutations in the spliceosome gene SF3B1. Such mutations induce specific RNA missplicing events, directly promote ring sideroblast (RS) formation, and generally associate with a more favorable prognosis. However, not all SF3B1 mutations are the same, and little is known about how distinct hotspots influence disease. Here, we report that the E592K variant of SF3B1 associates with high-risk disease features in MDS, including a lack of RS, increased myeloblasts, a distinct comutation pattern, and a lack of favorable survival seen with other SF3B1 mutations. Moreover, compared with other hot spot SF3B1 mutations, E592K induces a unique RNA missplicing pattern, retains an interaction with the splicing factor SUGP1, and preserves normal RNA splicing of the sideroblastic anemia genes TMEM14C and ABCB7. These data have implications for our understanding of the functional diversity of spliceosome mutations, as well as the pathobiology, classification, prognosis, and management of SF3B1-mutant MDS.

Graphical abstract

Figure 1.

Distribution of exon 13-16 SF3B1 mutations within WHO 2016 classifications of MDS and AML. Amino acid positions are shown along the x-axis, and individual variants are counted along the y-axis according to the legend above the graphs. EB, excess blasts.

Figure 2.

Asymmetric RNA missplicing by distinct SF3B1 mutation hotspots. HEK293T cells were transfected with constructs expressing FLAG-SF3B1 variants. Top row is western blotting with anti-SF3B1 antibody, showing FLAG-SF3B1 and endogenous SF3B1 at similar levels. End point PCR used isoform-competitive primers, with arrows for canonical (blue), cryptic (red), and heteroduplex (green) forms. Cryptic vs canonical UQCC1 was quantified as a ratio between 2 separate isoform-specific quantitative PCRs.

Figure 3.

Clinical parameters of MDS patients with the E592K variant of SF3B1. Compared with cases with exon 14-16 mutations, patients with E592K have (A) higher fractions of MDS-SLD/MLD, MDS-EB, and AML, (B) higher IPSS-M, (C) lower platelets, (D) nearly-absent RS, and (E) lower overall survival which is retained in (F) lower IPSS-M score groups. Outcomes are similarly poor for E592K and exon 14-16 patients in (G) higher IPSS-M score groups.

Figure 4.

Comutation landscape of MDS with E592K or exon 14-16 SF3B1 mutations. All cases of SF3B1-mutant MDS that sequenced at least the 24 additional genes shown were included. Fisher exact test P values and Benjamini linear multiple testing q values for FDR < 0.05 were as follows: ∗∗∗∗ASXL1 = 1.13E-16 and 2.37E-15; ∗∗∗∗RUNX1 = 2.06E-10 and 2.16E-09; ∗∗∗∗STAG2 = 5.03E-08 and 3.52E-07; and ∗NRAS = 4.61E-03 and 2.42E-02.

Figure 5.

E592K induces unique RNA missplicing events. TF1 cells transduced with different SF3B1 variants were analyzed by RNA-seq, with (A) highest-scoring ΔPSIs shown. (B) Western blot with anti-SF3B1 antibody. (C) End point PCR/quantitative PCR validation of K700E–specific missplicing events. (D) Validation of E592K-specific events. (E) RNA-seq reads from TF1 cells showing the cryptic event in EZH2. (F) Orthogonal validation of missplicing in heterozygous SF3B1 knockin hiPSCs. PTC, premature termination codon.

Figure 6.

The E592K variant preserves association of SF3B1 with SUGP1. HEK293T cells were transfected with His6-FLAG-SF3B1 variants and subjected to affinity purification with anti-DYKDDDDK (FLAG) antibody. (A) Silver-stained protein gel, with arrow pointing to the size of SUGP1, which is decreased in K700E but not E592K eluate. (B) Western blot showing decreased SUGP1 in K700E, but not E592K, eluate. PHF5A is present with all SF3B1 variants. (C) Reprobing with anti-SF3B1 shows native and His6-FLAG–tagged protein levels. E, E592K; K, K700E; KD, kilodaltons; M, marker; Vec, vector only; W, wild type SF3B1.

Figure 7.

E592K exhibits unique RNA missplicing in primary MDS samples. RNA-seq read distribution in the MLL cohort shows that E592K exhibits (A) canonical TMEM14C and ABCB7 missplicing and (B) cryptic RAVER2, NUTM2B-AS1, and EZH2 missplicing. (C) Distinct TMEM14C and RAVER2 missplicing was validated in marrow CD34⁺ cells from an independent patient by end point PCR.