Genetic alterations of SUGP1 mimic mutant-SF3B1 splice pattern in lung adenocarcinoma and other cancers

Abstract

Genes involved in 3'-splice site recognition during mRNA splicing constitute an emerging class of oncogenes. SF3B1 is the most frequently mutated splicing factor in cancer, and SF3B1 mutants corrupt branchpoint recognition leading to usage of cryptic 3'-splice sites and subsequent aberrant junctions. For a comprehensive determination of alterations leading to this splicing pattern, we performed a pan-TCGA screening for SF3B1-specific aberrant acceptor usage. While the most of aberrant 3'-splice patterns were explained by SF3B1 mutations, we also detected nine SF3B1 wild-type tumors (including five lung adenocarcinomas). Genomic profile analysis of these tumors identified somatic mutations combined with loss-of-heterozygosity in the splicing factor SUGP1 in five of these cases. Modeling of SUGP1 loss and mutations in cell lines showed that both alterations induced mutant-SF3B1-like aberrant splicing. Our study provides definitive evidence that genetic alterations of SUGP1 genocopy SF3B1 mutations in lung adenocarcinoma and other cancers.

Fig. 1. SF3B1-like pattern in the TCGA and SUGP1.

a Screening RNA-seq data in the TCGA using the SBT score. The SBT score (the occurrence of 1443 aberrant splice junctions in fastq RNA-seq) for each sample is plotted (x-axis) against the size of RNA-seq bam file (y-axis). Cases with SF3B1 hotspot (red points) and other mutations (green crosses) are indicated. The linear trend (gray) and the cutoff (red) lines for the cases further explored are shown. b Principal component analysis of the selected 456 cases (including cases with high SBT scores and control tumor cases) characterized by the splicing index (SI) in 366 cryptic 3′ss junctions selected in an unsupervised way (see “Materials and methods”). Cases with SUGP1 alterations are highlighted (magenta spots). The two first principal components, PC1 and PC2, are shown and the fraction of variance explained is indicated. The cutoff for PC1 scores to discriminate cases with SF3B1-like phenotype was set to median + 3·MAD (Median Absolute Deviation) (gray dashed line). c Mutations found in SUGP1 in all tumor types (TCGA). Missense, nonsense, frameshift, and splice mutations are indicated by green, black, red, and gray points, respectively. Mutations with loss of heterozygosity (LOH) and in heterozygous state are shown above and below the protein representation, respectively. Mutations associated with a 3′ss aberrant pattern are highlighted by magenta frames. d Effect of siRNA-mediated knockdown of SUGP1, on the aberrant splice forms of DPH5, DLST, and ARMC9 in the HEK293T cell line. Relative expression of cryptic 3′ss junction normalized to the canonical 3′ss junction was determined by quantitative RT–PCR, and effect of the different siRNA #1, 3, 6, and 21 was compared with the control (CTL) (paired t test; *p < 0.05; **p < 0.005; ***p < 0.0005). The protein knockdown was confirmed by immunoblotting with anti-SUGP1, using β-actin as a loading control. e Effect of siRNA-mediated knockdown of SUGP1, overexpression of wild-type SUGP1 or SF3B1, overexpression of SUGP1^{L515P, R625T or P636L} or SF3B1^K700E on the aberrant splice form of DPH5 in HEK293T cell line. Relative expression of cryptic 3′ss junction normalized to the canonical 3′ss junction of DPH5 was determined by quantitative RT–PCR. The results are average of three replicates and are represented as mean ± sd, and each condition is compared to the control (paired t test; *p < 0.05; **p < 0.005). The protein knockdown or overexpression was confirmed by immunoblotting with anti-Flag and anti-SUGP1 using β-actin as a loading control. f Minigene splice assay of two SF3B1^mut-sensitive 3′ss (ENOSF1, TMEM14C) and their cryptic (BP’) and canonical (BP) branchpoint mutants. Gel electrophoresis shows the different splicing processes for minigene ExonTrap constructions in SF3B1^WT cell line HEK293T with or without siRNA-mediated knockdown of SUGP1. The lower band corresponds to the usage of the canonical 3′ss. The intermediate band corresponds to the usage of the cryptic 3′ss. The upper band corresponds to the heteroduplex formation from the two products.

Fig. 2. Global similarity of splice alteration patterns in SUGP1 and SF3B1.

a P value distribution in SUGP1^alt vs. Controls (left panel), SF3B1^mut vs. Controls (right panel), and SUGP1^alt vs. SF3B1^mut (central panel) comparisons in the LUAD series performed by Wilcoxon rank test on the splicing index (SI) of 3′ss proximal aberrant junctions. b Principal component 2D plot for the set of aberrant 3′ss junctions selected independently for SUGP1^alt vs. Controls (left panel) and SF3B1^mut vs. Controls (right panel) comparisons. SUGP1^alt and SF3B1^mut cases are indicated. c Further analysis of the set of junctions from (b): 201 junctions (selection from union of 211 and 179) were classified as shared, SF3B1^mut-specific and SUGP1^alt-specific based on the maximal SI in a group. Pie diagram (left panel, counts are indicated) and distribution of aberrant junction positions (right panel, x = 0 corresponds to canonical exon start) are shown. d Hierarchical clustering and the heatmap of differentially spliced cryptic 3′ss junctions (n = 127) with high loadings, which separate SUGP1^alt and SF3B1^mut from the Control group in PCA, combined with cryptic 5′ss junctions (n = 45) and aberrant exon usage (n = 47) popped up in the comparisons but not separating mutated and control cases in the LUAD cohort (left panel). Cases are annotated according to their alterations in SF3B1 or SUGP1, 100 controls for visualization are selected randomly; type of aberrant junctions are indicated in the side row panels. For comparison, canonical junction expression is shown (increased expression of aberrant junctions is not explained by canonical junction expression; right panel).

Fig. 3. SF3B1-like splice pattern analysis in HEK293T and HAP1^SUGP1-P636L isogenic cell line.

a Relative splicing index (∆SI_max) in SUGP1-depleted (SUGP1^KD) HEK293T cells (x-axis) and HEK293T cells over-expressing SF3B1^K700E (y-axis) in different types of splicing defects: 3′ss aberration (left panel), 5′ss aberration (central panel), and aberrant exon usage (right panel). Junctions were selected in semi-unsupervised way (“Materials and methods”). Dotted lines indicate 2·MAD (Median Absolute Deviation), junctions significantly different (p value < 0.05) in Student t test comparison of mean values are marked by red (SUGP1^KD), blue (SF3B1^K700E) or violet (both) dots. b Sashimi plots showing read counts in two junctions found aberrant in experimental models SUGP1-depleted (SUGP1^KD), over-expressing SF3B1^K700E, and Controls HEK293T cells: chr16:708344-708509-708524 (left panel) and chr16:67692719-67692735-67692830 (right panel). Aberrant (red) and canonical (green) split reads counts are indicated. c Hierarchical clustering and heatmap of aberrant 3′ss junctions with ∆SI_max > 1 in either SUGP1^KD or SF3B1^K700E (74 and 49 junctions, respectively, 97 in total) HEK293T cell lines. Raw quantiles of aberrant junction split-read counts are shown. d Distances between the cryptic and canonical 3′ss for 97 junctions with ∆SI_max > 1 in either SUGP1^KD or SF3B1^K700E. The position of the canonical 3′ss is set to 0. e Hierarchical clustering and heatmap of differential splice 3′ss junctions (p value ≤ 0.05, Log2FC ≥ 1) in HAP1 and HAP1^SUGP1-P636L isogenic cell lines. Three biological replicates for each cell line (R1–R3) are indicated below the heatmap. The corresponding gene level expression heatmap is shown on the right panel for comparison (differential expression of junctions is not the consequence of differential gene expression). f siRNA-mediated knockdown of SUGP1^WT and SUGP1^P636L impact on DPH5 aberrant junction expression in HAP1 isogenic cell lines. Relative expression of cryptic 3′ss junction normalized to the canonical 3′ss junction of DPH5 was determined by quantitative RT–PCR. The results are average of three replicates and are represented as mean ± sd, and each condition is compared to the control (paired t test; **p < 0.005; ***p < 0.0005). The protein knockdown was confirmed by immunoblotting with anti-SUGP1 using β-actin as a loading control.