Physical inactivity contributes to type 2 diabetes (T2D), but the molecular links between exercise and metabolic improvement remain incompletely understood. We meta-analyzed genome-wide association studies of vigorous physical activity and T2D (combined n ≈ 1.95 million) and integrated eQTL/sQTL maps with single-cell and spatial transcriptomic datasets to connect genetic risk with tissues, cell types, and regulatory programs. Tissue and cell-type enrichment, colocalization, and network analyses were performed. Computational findings were further examined in male 10-week-old C57BL/6J mice with high-fat diet-induced diabetes. After 1 week of acclimatization, mice were randomly assigned to normal chow, high-fat diet, or high-fat diet plus exercise groups (n = 6 per group; high-fat diet with 60% of total energy from fat). The exercise intervention consisted of treadmill running (10 m/min for 50 min per day, 5 days per week, total 16 weeks), followed by metabolic phenotyping, skeletal muscle histology, bulk RNA sequencing, alternative splicing analysis, and RT-qPCR of Mau2 isoforms. Exercise- and T2D-associated variants showed joint enrichment in skeletal muscle and adipose eQTL/sQTL signals. Integrated single-cell analyses prioritized fibro-adipogenic progenitors and endothelial cells, and identified an extracellular matrix- and collagen-related module in fibro-adipogenic progenitors associated with both exercise and T2D. Mau2 emerged as a shared candidate gene with tissue-specific splicing signals. In diabetic mice, exercise improved glucose homeostasis and muscle fiber structure, and reduced Mau2 intron retention in skeletal muscle without changing total Mau2 expression. These findings support a multiscale framework linking exercise-responsive regulation to T2D-related tissue remodeling and splicing plasticity.