Exercise is associated with younger methylome and transcriptome profiles in human skeletal muscle

Abstract

Exercise training prevents age-related decline in muscle function. Targeting epigenetic aging is a promising actionable mechanism and late-life exercise mitigates epigenetic aging in rodent muscle. Whether exercise training can decelerate, or reverse epigenetic aging in humans is unknown. Here, we performed a powerful meta-analysis of the methylome and transcriptome of an unprecedented number of human skeletal muscle samples (n = 3176). We show that: (1) individuals with higher baseline aerobic fitness have younger epigenetic and transcriptomic profiles, (2) exercise training leads to significant shifts of epigenetic and transcriptomic patterns toward a younger profile, and (3) muscle disuse "ages" the transcriptome. Higher fitness levels were associated with attenuated differential methylation and transcription during aging. Furthermore, both epigenetic and transcriptomic profiles shifted toward a younger state after exercise training interventions, while the transcriptome shifted toward an older state after forced muscle disuse. We demonstrate that exercise training targets many of the age-related transcripts and DNA methylation loci to maintain younger methylome and transcriptome profiles, specifically in genes related to muscle structure, metabolism, and mitochondrial function. Our comprehensive analysis will inform future studies aiming to identify the best combination of therapeutics and exercise regimes to optimize longevity.

Keywords: DNA methylation; aging; cardiorespiratory fitness; exercise training; human skeletal muscle; mRNA expression; meta-analysis.

FIGURE 1

Study overview. First, we performed a large‐scale data mining exercise to gather all existing DNA methylation (DNAm) and mRNA expression microarray datasets from our own lab, our network of collaborators, and open‐access repositories (GEO, dbGAP, ArrayExpress). See Methods for inclusion criteria of datasets. Step 1: We identified age‐related changes in DNAm and mRNA expression in human skeletal muscle by meta‐analyzing 1251 samples across 16 cohorts (DNAm) and 1925 samples across 21 cohorts (mRNA). Step 2: First, we performed a cross‐sectional association between DNAm levels at age‐related CpGs or expression levels at age‐related mRNAs and VO_2max. Then, we determined whether DNAm levels at age‐related CpGs, and expression levels at age‐related mRNAs changed after exercise training, and we assessed whether expression levels at age‐related mRNAs changed after muscle disuse. Finally, we performed a series of OMIC integrations and pathway analyses to identify the molecular pathways affected by age, VO_2max, exercise training, and/or muscle disuse across both OMIC layers. Note: we only had OMIC data available at the transcriptomic level following muscle disuse. Note 2: summary statistics for the exercise‐ and disuse‐induced mRNA changes came from two recent meta‐analyses (Fanelli et al., ; Garcia et al., 2022).

FIGURE 2

Aerobic fitness and exercise training have similar associations with muscle epigenetic profiles, which contrast with those seen with age. (a) Pairwise correlation (Spearman) between the effect sizes of age, aerobic fitness (VO_2max), and exercise training at age‐related DMPs. Each dot corresponds to one of the 3168 age‐related DMPs, and the axes represent the magnitude of effect for age, VO_2max, and exercise training. (b) Unsupervised hierarchical clustering of the effect sizes of age, VO_2max, and exercise training at age‐related DMPs (ordered from the most hypomethylated to most hypermethylated with age). Note that the legend is arbitrary as effect sizes were scaled to an SD of 1 for age, VO_2max, and exercise training to be comparable.

FIGURE 3

Aerobic fitness and exercise training show similar associations with muscle transcriptomic profiles and contrast with those seen with age and muscle disuse. (a) Pairwise correlation (Spearman) between the effect sizes of age, aerobic fitness (VO_2max), exercise training, and muscle disuse at age‐related DEGs. Each dot corresponds to one of the 330 age‐related DEGs, and the axes represent the magnitude of effect for age, VO_2max, exercise training, and disuse. (b) Unsupervised hierarchical clustering of the effect sizes of age, VO_2max, exercise training, and disuse at age‐related DEGs (ordered from the most downregulated to most upregulated with age). Note that the legend is arbitrary as effect sizes were scaled to an SD of 1 for age, VO_2max, exercise training, and disuse to be comparable. (c) Comparison of the magnitude of exercise‐induced changes in gene expression between a hypothetical 20‐ and 70‐year‐old. The genes displayed are those for which “age” was a significant moderator according to the meta‐regression conducted by Amar et al. (2021) “Down DEG” = gene whose expression decreases during normal aging; “Up DEG” = gene whose expression increases during normal aging. (d) Multi‐contrast enrichment comparing the effects of age, VO_2max, exercise training, and muscle disuse at age‐related DEGs. Genes related to mitochondrial function showed clear indications of downregulation during aging and following muscle disuse while being simultaneously upregulated with higher VO_2max and following exercise training. This was visible across the Gene Ontology, Canonical Pathways, and Expression Signature of Genetic and Chemical Perturbations gene sets.

FIGURE 4

Principal component analysis of individuals from the Gene SMART cohort at age‐related DMPs, and individuals from the GSE18732 cohort at age‐related DEGs. We used principal component analysis (PCA) to reduce dimensionality and show each individual on a two‐dimensional graph. Individuals from the Gene SMART cohort (a) and from the GSE18732 cohort (b) are colored according to their baseline VO_2max levels. Lower levels of VO_2max are indicated by smaller circles in lighter colors, while higher levels are indicated by larger circles in darker reds. To objectively test the clustering of individuals according to VO_2max, we ran Pearson correlations between individual coordinates on Dimension 1, Dimension 2, or Dimension 3 and VO_2max.