Fig 2: SRSF6 specifically recognizes GAA motifs in exons.(A) SRSF6 binding sites in CDS frequently display AG-rich pentamers, contrasting the uridine-rich pentamers in introns, most prominently UUUUU. Scatter plots compare pentamer frequency within the 9-nt binding sites and in flanking 20-nt windows for SRSF6 binding sites in introns and CDS. Two most enriched pentamers of clusters 1 and 2 derived from hierarchical clustering of pentamer profiles are colored (see (B) and Fig S2A). (B) GAAGA and AAGAA enrich around SRSF6 binding sites in the CDS, whereas UUUUU marks binding site centers. Heat map shows cluster 1 from hierarchical clustering of pentamer profiles. Two most enriched pentamers are labeled and colored as in (A). Clusters 2 and 3 with uridine-rich and other pentamers are shown in Fig S2A and B. (C) SRSF6 positions towards end of motif-enriched stretches. Metaprofile shows pentamer frequencies in 201-nt window around SRSF6 binding sites. Two most enriched pentamers of three clusters from hierarchical clustering of pentamer profiles are shown. (D) Binding site strength increases with the number of GAA triplets. Boxplot (bottom) shows the distribution of binding site strengths (log2-transformed PureCLIP score) for binding sites with a given number of GAA triplets within 30 nt from binding site center. Reverse complement UUC was used as control. Box represents quartiles, center line denotes 50th percentile, and whiskers extend to most extreme data points within 1.5× interquartile range. The bar chart (top) gives number of binding sites in each category. (E) Two or more GAA triplets in direct sequence are associated with increased binding site strength. Boxplot shows the distribution of binding site strengths (log2-transformed PureCLIP score) for binding sites with no or one triplet (GAA or UCC) compared with two or more triplets in direct sequence or with 1-nt or 2-nt gaps. Visualization as in (D). (F) Motif enrichment analysis using DREME (Bailey, 2011) detected a purine-rich motif, reinforcing the role of GAA regions in SRSF6 binding. The motif is present at 25,148 SRSF6 binding sites.

Fig 3: SRSF6 binding on alternative and constitutive exons defines the splicing outcome in EndoC-βH1 cells.(A) SRSF6 knockdown (KD) affects the inclusion of 1,212 cassette exons (CEs). Volcano plot shows log10-transformed adjusted P-values against differences in “percent spliced-in” (ΔPSI) of CEs. Exons with significantly increased or decreased inclusion are colored (ΔPSI > 0.05, adjusted P-value < 0.05), according to reanalysis of our previously published RNA-seq data of EndoC-βH1 upon SRSF6 KD (n = 5) (Juan-Mateu et al, 2018). (B) Most SRSF6-regulated CEs are directly associated with an SRSF6 binding site. Venn diagram depicts number and percentage of significantly regulated exons that harbor an SRSF6 binding site in the alternatively spliced region. Numbers in brackets at arrowheads specify exons with significantly increased and decreased inclusion in the overlap. (C) SRSF6 RNA splicing maps. SRSF6 shows more cross-link events on exons with decreased inclusion after SRSF6 KD (blue, top), suggesting that SRSF6 binding is required for their inclusion, whereas exons with increased inclusion after SRSF6 KD (orange, bottom) display more SRSF6 on the flanking constitutive exons. Metaprofile depicts fraction of exons with cross-link events at a given position within 100-nt on either side of indicated 3′ and 5′ splice sites. For comparison, mean and standard deviation are shown for a PSI-matched background distribution of unchanged exons (light blue and light orange lines, respectively; see Fig S3C–E), resulting from randomly sampled unchanged exons in EndoC-βH1 cells with comparable exon inclusion levels under control conditions. Only positive z-scores with adjusted P-value < 0.05 are shown. (D) Up-regulation upon SRSF6 KD predominantly affects alternative exons, whereas down-regulation is almost equally observed for constitutive exons. The bar chart shows the contribution of different exon categories to exons with significantly increased or decreased inclusion upon SRSF6 KD according to Exon Ontology analysis (Tranchevent et al, 2017). All remaining CEs (Unchanged) are shown as control.

Fig 4: SRSF6 cross-link events around SRSF6-regulated cassette exons.(A, B) Metaprofiles (top) and heat maps (bottom) represent the SRSF6 cross link signal on exons with significantly decreased (n = 975) (A) or increased (n = 237) (B) inclusion upon SRSF6 KD in EndoC-βH1 cells. Type 1 (T1D) and type 2 (T2D) diabetes susceptibility genes are labeled on the side. Metaprofiles depict fraction of exons with cross-link events (merged replicates) at a given position within 100 nt on either side of indicated 3′ and 5′ splice sites. Heat maps depict individual positions with cross-link evens in the respective exons. Rows are sorted by exon length and iCLIP coverage. Positions missing for a particular exon are set to 0 and thus are included in the metaprofile above.

Fig 5: Predicted impact of SRSF6 on diabetes susceptibility genes expression.(A, B) SRSF6 KD affects expression of nine T1D (A) and 81 T2D (B) susceptibility genes in EndoC-βH1 cells. Some of these differentially expressed genes also harbor SRSF6-regulated splicing events. Bar charts depict log2-transformed fold-changes of differentially expressed diabetes susceptibility genes (adjusted P-value < 0.001). Genes with additional alternatively spliced exons are colored. (C) Proteins encoded by SRSF6-regulated diabetes susceptibility genes form a protein–protein interaction network which bridges functionally related proteins, including further diabetes susceptibility genes, namely TCF12, CDKN1B, CCND1, INS, and IRS. Protein–protein interaction network was obtained using STRING version 11.0 (https://string-db.org/), using as input 29 T1D and T2D susceptibility genes with significant alternative splicing and/or differential expression (with absolute log2-transformed fold-change >1) after SRSF6 KD.