Phase Separation Contributes to Pathogenicity for Nonsense Mediated Decay-Escaping Variant Alleles

Premature termination codons (PTCs) usually trigger nonsense-mediated decay (NMD), but PTCs in the final exon or within 50 base pairs of the last exon-exon junction escape surveillance and can yield stable truncated or altered proteins whose disease mechanisms remain poorly defined. Screening pathogenic and likely pathogenic ClinVar variants, we identified 134 autosomal dominant disease genes enriched for NMD-escape frameshift (N=94) or stopgain (N=64) variants. At the gene level, these genes were indistinguishable from length-matched controls in dosage sensitivity, interaction-network centrality and tissue specificity; only the stopgain class showed greater overlap between its NMD-escape regions and protein-protein interaction interfaces (log-odds ratio +1.22). Comparing pathogenic with benign variants in the same genes, pathogenic alleles truncated further upstream and therefore spanned more post-translational modification sites, nuclear localization signals, protein-protein interaction interfaces and disordered regions; after adjusting for truncation position, pathogenic stopgain variants still overlapped interaction interfaces and Pfam domains more often than benign ones. The two classes remodelled the C-terminus differently: pathogenic frameshift alleles generated novel tails that were less positively charged with a higher phase-separating disordered-region fraction in their divergent segment (q = 0.028), whereas pathogenic stopgain alleles removed longer, more disordered and interaction-prone C-terminal sequence, leaving a truncated protein enriched in arginine and positive charge. NMD-escape disease genes were over-represented among neurodevelopmental-disorder risk genes (35 observed versus 25 expected; odds ratio 1.6, P = 1.3 x 10-{superscript 2}), and an NMD-escape truncating variant in RARB (microphthalmia, syndromic 12; MIM 615524) formed aberrant condensates in cells. Together, these results define NMD-escape alleles as a mechanistically distinct class of dominant disease variants in which remodelling of interaction-prone, phase-separation-competent C-terminal sequence, rather than transcript loss, drives pathogenicity.