RESUMO
Correlation of motif occurrences with gene expression intensity is an effective strategy for elucidating transcriptional cis-regulatory logic. Here we demonstrate that this approach can also identify cis-regulatory elements for alternative pre-mRNA splicing. Using data from a human exon microarray, we identified 56 cassette exons that exhibited higher transcript-normalized expression in muscle than in other normal adult tissues. Intron sequences flanking these exons were then analyzed to identify candidate regulatory motifs for muscle-specific alternative splicing. Correlation of motif parameters with gene-normalized exon expression levels was examined using linear regression and linear splines on RNA words and degenerate weight matrices, respectively. Our unbiased analysis uncovered multiple candidate regulatory motifs for muscle-specific splicing, many of which are phylogenetically conserved among vertebrate genomes. The most prominent downstream motifs were binding sites for Fox1- and CELF-related splicing factors, and a branchpoint-like element acuaac; pyrimidine-rich elements resembling PTB-binding sites were most significant in upstream introns. Intriguingly, our systematic study indicates a paucity of novel muscle-specific elements that are dominant in short proximal intronic regions. We propose that Fox and CELF proteins play major roles in enforcing the muscle-specific alternative splicing program, facilitating expression of unique isoforms of cytoskeletal proteins critical to muscle cell function.