Context-dependent effects of mutations on complex splicing decisions

Alternative pre-mRNA splicing is an essential step in human gene regulation, and mutation-induced aberrant splicing is frequently found in diseases and therapy resistance. Splicing regulation is highly dependent on sequence and cellular context, posing a challenge to predict outcomes of splicing-related disease mutations. Here, we use kinetic modeling to derive the underlying quantitative principles, describing splice site competition and downstream effects on a wide spectrum of splice isoforms. Employing statistical learning on a large-scale mutagenesis dataset for CD19 mRNA splicing, our model quantitatively describes the generation of 93 RNA isoforms. It takes into account various splicing events such as cassette exon skipping, intron retention, and extensive alternative 3 and 5 splice site usage, which are implicated in CART therapy resistance in leukemia. Beyond CD19, by analyzing genome-wide RNA sequencing data and large-scale screening of synthetic splicing decisions, we find that splice site distance is an important parameter controlling the switch from canonical to alternative/cryptic splice site usage, as mutations located in between two nearby splice sites show a pronounced directionality, regulating up- and downstream effectors in opposite direction. Taken together, our work demonstrates that a quantitative description of splice site competition provides insights into context-dependent, complex isoform changes and cryptic splice site activation in health and disease.