Partial convergence of alternative polyadenylation changes across ALS-Associated RNA-Binding proteins

Alternative polyadenylation (APA) regulates transcript 3'-end architecture, stability, localization, and translation. Although RNA-binding protein dysfunction is central to amyotrophic lateral sclerosis (ALS), whether different ALS-associated proteins cause shared APA abnormalities remains unclear. We analyzed poly(A)-site usage after TDP-43 knockdown, TDP-43 K263E expression, FUS knockout, FUS P525L expression, and MATR3 knockout. Using false-discovery rate (FDR) < 0.05 as the primary criterion, evidence varied markedly among datasets. FUS knockout produced the strongest response, with 373 significant events affecting 239 genes. Three events involving two genes remained significant after TDP-43 knockdown, and two events involving one gene remained significant in the TDP-43 K263E dataset. No events survived FDR correction after FUS P525L expression or MATR3 knockout, and no FDR-significant events or genes were shared between datasets. We therefore examined nominal events (unadjusted P < 0.05) as an explicitly exploratory layer. Nominal event overlap was enriched in 7 of 10 pairwise comparisons, and nominal gene overlap in 9 of 10. TDP-43 knockdown and TDP-43 K263E showed the strongest directional agreement: 65 nominal events were shared; 25 were directionally classified in both datasets, of which 23 were concordant. FUS knockout and FUS P525L also showed enriched nominal overlap, but without significant directional concordance or correlated ΔPAU responses. Thus, FUS loss induces a robust FDR-supported APA phenotype, whereas evidence is substantially weaker in other contexts. Cross-dataset similarities occur mainly among exploratory nominal targets and do not consistently extend to direction or magnitude. APA responses therefore show partial, context-dependent convergence rather than a universal defect.