In vitro pathogenicity evaluation of deep intronic variants for recessive genetic retinal diseases

Non-coding variants that alter mRNA splicing are increasingly recognized as important causes of Mendelian disorders, and deep intronic variants (DIVs) that activate cryptic exons (CEs) account for roughly 20% of inherited retinal disease (IRD) cases. However, accurately predicting whether an intronic variant is pathogenic remains challenging, especially in recessive IRDs where patients often have only one confirmed pathogenic variant.

To address this, researchers used a high-throughput splicing assay (HTSA) to measure the effects of rare deep intronic variants in recessive IRDs. They selected 640 very rare DIVs in trans position with a known mutation from 76 patients. The HTSA used a split-GFP minigene separated by an SMN1 gene intron, into which the 270 bp sequences flanking each DIV were cloned. The plasmid library was transfected into 293HEK cells, and 48 hours later total RNA was extracted. HTSA minigene RNA transcripts were then amplified by RT-PCR and sequenced, and intron sequences spliced between the two GFP exons were identified and quantified.

In the results, 98 variants activated a cryptic exon more than 100 times more in the altered oligo compared to the reference oligo and were classified as pathogenic in this experimental setting; these were validated in a longer context. Another 26 variants were classified as variants of uncertain significance (VUS). Notably, only 6 of 90 variants (6.6%) were predicted to be pathogenic by the in silico algorithm SpliceAI, indicating substantial room for improvement in prediction algorithms. The assay results confirmed a diagnosis for 50 of 78 patients (64%), and 19.4% of deep intronic variants were shown to cause CE activation.

The authors conclude that a significant portion of undiagnosed patients with recessive IRDs may carry pathogenic intronic mutations that cause disease, highlighting the value of functional splicing assays to improve molecular diagnosis when in silico tools are insufficient.