Which genes' splicing changes with age? A genome-wide analysis of 2.8 million junction tests in GTEx v8

Abstract Background. Changes in pre-mRNA splicing are a recurrent feature of human aging, but which genes are most and least affected across tissues is not well established. Our earlier analysis of four kynurenine-pathway enzymes found age-associated shifts towards truncated isoforms, most strongly in AFMID. Here we extend the same approach to every gene in the GTEx v8 junction matrix. Methods. Junction read counts for 17,382 samples were restructured by tissue, and annotated junctions were grouped into intron clusters. In each of 49 tissues, junction usage was modelled on donor age alone with quasi-binomial regression, fitted simultaneously for all junctions; estimates matched a standard implementation (r = 1.000). False discovery rate (FDR) was controlled across 2.78 million tests. Models adjusted for sex, RNA integrity, ischaemic time and death circumstance were fitted as a secondary analysis. Results. Of 910,929 testable tissue × cluster events in 15,503 genes, 107,594 (11.8%) were associated with age at FDR < 0.05, and 10,786 genes (70%) had at least one such event. The share of age-associated events ranged from 1% to 33% between tissues, and was highest where age and ventilator-case status were most tightly linked. Among 9,320 genes testable in at least 20 tissues, splicing regulators were the most consistently age-linked: HNRNPH1, SF1, RBM39, BCLAF1 and LUC7L3 had age-associated splicing in 34–39 of 49 tissues, and regulation of mRNA splicing was the most enriched biological process (7.4-fold). At the other extreme, 584 genes, including COX4I1, DDOST and RPL13, showed no age association in any tissue despite precise estimates. Age shifted splicing away from principal isoforms: in 62% of significant events, the principal-isoform junction lost share. After adjustment, 97.7% of effects kept their direction, at a median 57% of their crude size; 3,662 genes retained an event, enriched for actin-cytoskeleton, muscle and extracellular-matrix functions. AFMID ranked at the 95th percentile genome-wide. Conclusions. Age-associated splicing is widespread but concentrated in a few hundred genes, led by the splicing machinery itself, and tends to move transcripts away from their principal isoforms. Much of its tissue-to-tissue variation in GTEx reflects how strongly age is linked to death circumstance, so crude and adjusted estimates should be read together.