A contextual genomic perspective on physical activity and its relationship to health, well being and illness

Abstract

Physical activity (PA) is one of the most fundamental traits in the animal kingdom, has pervasive health benefits, and is genetically influenced. Using data from the Million Veteran Program, we conducted genetic analyses of leisure, work and home-time PA. For leisure, we included 189,812 individuals of European ancestry (SNP-based heritability (h²) = 0.083 ± 0.005), 27,044 of African ancestry (h² = 0.034 ± 0.017) and 10,263 of Latin American ancestry (h² = 0.083 ± 0.036) in a cross-ancestry meta-analysis with UK Biobank data, identifying 70 lead variants. Leisure-time PA was genetically distinct from PA at home or work, with the latter two showing less health benefit with respect to health outcomes and lifespan. Mendelian randomization analyses showed a protective role of leisure-time PA against COVID-19 hospitalization (β = -0.067 ± 0.016; P = 2.8 × 10^-5), and with other traits including cardiovascular and respiratory system diseases, metabolic traits and aging. These findings provide new insights into the biology of PA, showing specific causal health benefits of leisure-time PA.

Fig. 1. Cross-ancestry All-PA–leisure + SSOE meta-analysis.

Annotated genes are the closest to each significant lead SNP. Different genes compared to EUR All-PA–leisure meta-analysis (Supplementary Fig. 6) are highlighted in green. The red horizontal line indicates GWS (P = 5 × 10⁻⁸) level, where P values were calculated using sample-size-based approach, two-sided tests.

Fig. 2. Traits genetically correlated with All-PA–leisure + SSOE.

a, Genetic correlations of EUR meta-analysis for All-PA–leisure + SSOE phenotype with traits of interest. We display here the traits of interest that were significant after the Benjamini–Hochberg false discovery procedure. Tests were two-sided. Black error bars represent the standard error. b, PheWAS of PA from the polygenic score derived from the EUR All-PA–leisure + SSOE meta-analysis. The red line represents the Bonferroni threshold P = 3.99 × 10⁻⁵. Up-pointing and down-pointing triangles represent positive or negative associations, respectively, which were calculated using a logistic regression model.

Fig. 3. Genetic differences among vigorous PA types.

a, Genetic correlation between each pair of vigorous PA contexts and other traits of interest. Significantly different correlations were defined as having P < 4.63 × 10⁻⁴ after multiple testing correction. An asterisk represents a significant difference between Vig-PA–leisure and Vig-PA–work, wedge represents a significant difference between Vig-PA–leisure and Vig-PA–home, and plus represents a significant difference between Vig-PA–home and Vig-PA–work. Tests were two-sided. Black error bars represent the s.e. The genetic correlation values of the vigorous PA with the other traits of interest are provided in Supplementary Tables 36–38. b, Genomic-SEM of the two-factor model, showing the loadings for the inferred traits—their standardized estimates with the s.e. in parentheses. The CFA uses a chi-square test.