Aging-associated alternative splicing programs conserved between human and mouse tissues

Alternative splicing is a key gene regulatory process that diversifies the proteome and controls gene dosage. Previous studies have detected aging-associated splicing changes across various tissues. However, their use of bulk RNA-seq obfuscates the impacted cell-types and may confound cell-type proportion changes with cell-intrinsic ones. We present a framework that first assembles and maps alternative splicing events in appropriate single cell RNA-seq data (scRNA-seq), then applies LeafletFA, a probabilistic model that discovers coordinated splicing programs (SPs) without requiring prior knowledge of cell types or clinical information, such as age. Applying this framework to over 200,000 cells from mouse and human Smart-seq2 multi-tissue atlases, we discovered global and cell type specific aging-associated SPs. In mice, integrating SPs with gene expression significantly improved age prediction in 46 of 76 tissue-cell types tested. We identified the RNA-binding protein Snrnp70 as a key upstream mediator driving the shift from youthful to aged SPs. To test evolutionary conservation of these patterns, we employed transfer learning to map the murine splicing dictionary onto human transcriptomes. This revealed a conserved "youth" program (SP4) that is maintained in quiescent endothelial tissues but lost in high-turnover organs. We identified a conserved, age-dependent alternative 5 splice site usage in the splicing factor SRSF5 as a molecular marker of this program in both species. Our work establishes alternative splicing as a coordinated, evolutionarily conserved dimension of cellular aging and provides LeafletFA as an open-source tool for single-cell splicing analysis.