CELF/RBFOX antagonism is conserved in mouse heart/muscle and influences developmental and disease splicing programs. (A) Two-tailed Fisher's exact test −log10(P-value) for the five most significant (lowest P-value) CLIP-seq peak occurrences enriched proximal to alternative exons repressed by CELF2 in mouse heart (left) or enhanced by RBFOX2 in myotubes (right) compared to unresponsive exons. Condition in which each CLIP experiment was carried out is indicated in parentheses. CLIP experiments for the CELF or RBFOX families are indicated in blue or orange, respectively. (DS) Downstream, (US) upstream. (B) RNA map of RBFOX2 CLIP-seq peaks from cardiomyocytes proximal to CELF2-repressed (red), -enhanced (green), or -unresponsive (gray) exons in the heart. (C) Venn diagram showing overlap of splicing changes in the mouse hearts that overexpressed (OE) CELF1 or CELF2 versus events altered through depletion of RBFOX2 (conditional heart KO) (P < 2.6 × 10−10, Fisher's exact test). (D) Fraction of total RBFOX2 regulated events that changed in the same direction (white) or the opposite direction (gray) of RBFOX2 regulation (top), with grayscale squares representing significance of the overlap between regulated events (bottom, −log10[P-value], Fisher's exact test). (E) Pairwise E(ΔΨ) correlation for coregulated (|E[ΔΨ]| > 20%) LSVs between indicated conditions. (T1D) Type 1 diabetes hearts. (F) Venn diagrams showing number of cassette exons that contain a splice-site proximal CLIP-seq peak for CELF1 in heart (blue), RBFOX2 in cardiomyocytes (orange), or both (purple) and are regulated across heart development (left) or dysregulated in T1D hearts (right). (G) MAJIQ ΔΨ violin plot (left) and UCSC Genome Browser view with RNA-seq tracks and CLIP-seq peaks (right) around a physiologically relevant event, the Mef2d α1/α2 mutually exclusive exons, that is dysregulated in disease and antagonistically coregulated by CELF2 (top) and RBFOX2 (bottom) (for more examples, see Supplemental Table S3; Supplemental Figs. S6, S7).
