Divergent Skeletal Muscle Metabolomic Signatures of Different Exercise Training Modes Independently Predict Cardiometabolic Risk Factors

Abstract

We investigated the link between enhancement of SI (by hyperinsulinemic-euglycemic clamp) and muscle metabolites after 12 weeks of aerobic (high-intensity interval training [HIIT]), resistance training (RT), or combined training (CT) exercise in 52 lean healthy individuals. Muscle RNA sequencing revealed a significant association between SI after both HIIT and RT and the branched-chain amino acid (BCAA) metabolic pathway. Concurrently with increased expression and activity of branched-chain ketoacid dehydrogenase enzyme, many muscle amino metabolites, including BCAAs, glutamate, phenylalanine, aspartate, asparagine, methionine, and γ-aminobutyric acid, increased with HIIT, supporting the substantial impact of HIIT on amino acid metabolism. Short-chain C3 and C5 acylcarnitines were reduced in muscle with all three training modes, but unlike RT, both HIIT and CT increased tricarboxylic acid metabolites and cardiolipins, supporting greater mitochondrial activity with aerobic training. Conversely, RT and CT increased more plasma membrane phospholipids than HIIT, suggesting a resistance exercise effect on cellular membrane protection against environmental damage. Sex and age contributed modestly to the exercise-induced changes in metabolites and their association with cardiometabolic parameters. Integrated transcriptomic and metabolomic analyses suggest various clusters of genes and metabolites are involved in distinct effects of HIIT, RT, and CT. These distinct metabolic signatures of different exercise modes independently link each type of exercise training to improved SI and cardiometabolic risk.

Article highlights: We aimed to understand the link between skeletal muscle metabolites and cardiometabolic health after exercise training. Although aerobic, resistance, and combined exercise training each enhance muscle insulin sensitivity as well as other cardiometabolic parameters, they disparately alter amino and citric acid metabolites as well as the lipidome, linking these metabolomic changes independently to the improvement of cardiometabolic risks with each exercise training mode. These findings reveal an important layer of the unique exercise mode-dependent changes in muscle metabolism, which may eventually lead to more informed exercise prescription for improving SI.

Figure 1

Transcriptomics reveal that skeletal muscle AA metabolism is linked to enhanced SI after exercise training. A: Univariate analysis correlating gene expression with glucose R_d found enrichment of AA metabolism–related gene pathways after exercise training (independent of exercise type; n = 52). B and C: Gene set enrichment analysis shows a positive correlation between the change in BCAA metabolism–related gene expression and the change in glucose R_d induced by exercise training (independent of exercise type). D: Multivariate analysis correlating gene expression with glucose R_d after three different exercise training modes (HIIT n = 19; RT n = 18; CT n = 15). E–G: The abundance of BCKDH is increased in human skeletal muscle after 12 weeks of HIIT (n = 14) but not RT (n = 14) or CT (n = 13). Phosphorylation of BCKDH is increased after HIIT or CT but not RT. Gray lines indicate median values, and black dots indicate individual values. *P < 0.05 (significant difference in BCKDH abundance or phosphorylation after training). AU, arbitrary unit; FDR, false discovery rate; NES, normalized enrichment score.

Figure 2

Skeletal muscle concentrations of TCA cycle and amino metabolites are differentially altered by different modes of exercise training. TCA cycle intermediates are displayed with AAs and amino metabolites that can be converted to TCA intermediates and metabolized for fuel. The percent change in metabolite concentration after HIIT (A), RT (B), or CT (C) is displayed, with the larger circles indicating a greater percent change in metabolite concentration after exercise training. Dark blue and dark red circles indicate a significant (P < 0.05) increase or decrease, respectively, in metabolite concentration after exercise training. For AAs, n = 19 for HIIT, n = 18 for RT, and n = 15 for CT. For TCA intermediates, n = 16 for HIIT, n = 16 for RT, and n = 14 for CT. BAIBA, β-aminoisobutyric acid; 2-HG, 2-hydroxyglutarate; GABA, γ-aminobutyric acid; 1-MH, 1-methylhistidine; 3-MH, 3-methylhistidine.

Figure 3

Skeletal muscle acylcarnitines are altered with 12 weeks of exercise training. Free carnitines (A) and short- (B), medium-chain (C), and long-chain (D) acylcarnitines before (pre) and after (post) HIIT (n = 18), RT (n = 18), or CT (n = 15) in skeletal muscle are presented as nanogram of acylcarnitines per milligram of muscle tissue. Gray lines indicate mean values, and individual values are indicated by black dots. *P < 0.05 (significant difference in concentration after exercise training). AU, arbitrary unit; FDR, false discovery rate; N/A, not applicable; NES, normalized enrichment score.

Figure 4

Skeletal muscle ceramides are altered with 12 weeks of exercise training. Ceramides, sphingosine, sphinganine, and sphingosine-1-phosphate are plotted by lower (A), medium (B), and higher abundance (C) before (pre) and after (post) HIIT (n = 17), RT (n = 16), or CT (n = 15) in skeletal muscle. Data are presented as picomoles of ceramides per milligram of muscle tissue. Gray lines indicate mean values, and individual values are indicated by black dots. *P < 0.05 (significant difference in concentration after exercise training). BAIBA, β-aminoisobutyric acid; 2-HG, 2-hydroxyglutarate; GABA, γ-aminobutyric acid; 1-MH, 1-methylhistidine; 3-MH, 3-methylhistidine.

Figure 5

Skeletal muscle lipidome is differentially altered by HIIT, RT, and CT. A: The number of uniquely identified lipid species within each lipid class (indicated by different colors) is displayed in a pie chart. The total number of identified unique lipid species was 540. B–D: Volcano plots represent the change in skeletal muscle lipid metabolite concentration in response to HIIT (n = 18) (B), RT (n = 13) (C), or CT (n = 15) (D). Dots represent individual lipid species. Lipids with a corrected false discovery rate <0.05 and absolute log2(fold change) ≥0.25 were considered as differentially expressed and colored in the graphs. E: The numbers of lipid species that were regulated by HIIT, RT, or CT are grouped by lipid class. Lipid species were grouped within each lipid class as follows: fatty acyls, Car and FA; glycerolipids, DG and TG; phospholipids, CL, LPC, LPE, LPG, PC, PE, PG, PI, and PS; sphingolipids, Cer, GM, HexCer, and SM; and sterol lipids (not shown because none changed with exercise training), CE and Chol. Lipid species acronyms are defined in A. F: The numbers of the phospholipids CL, PE (including PE and LPE), and PC (PC and LPC) that were altered by HIIT, RT, or CT are displayed. Phospholipids that were increased by exercise training are depicted with bars going upward, and phospholipids that were reduced by exercise training are depicted with bars going downward. FC, fold change; FDR, false discovery rate.

Figure 6

Associations between cardiometabolic paramters and muscle metabolite concentrations in response to different modes of exercise training. The exercise training–induced changes in six physiological variables, including SI (glucose R_d), aerobic capacity (Vo_2max), mitochondrial protein synthesis (mitochondrial protein fractional synthesis rate [Mito FSR]), fat-free mass (FFM), strength (one repetition maximum [Rep Max]), and mitochondrial respiration (state 3 CI+II), were correlated with metabolites identified by targeted metabolomics (A) or untargeted lipidomics (B). The Spearman ρ value is presented for the correlation between each metabolite and physiological variable within age group (young [Y] and older [O]) and exercise training type (HIIT, RT, and CT), where blue indicates a positive correlation and red indicates a negative correlation.

Figure 7

Integrated analysis of targeted metabolomics and transcriptomics reveals metabolite–gene clusters associated with cardiometabolic parameters after different exercise training modes. A: Seventeen clusters of gene transcripts and targeted skeletal muscle metabolites were identified (labeled as different color modules) based on exercise-induced changes in abundance/concentration. WCNA was performed to determine which module of gene transcripts/metabolites was associated with the exercise training (HIT, RT, or CT)–induced change in SI (glucose R_d), aerobic capacity (Vo_2max), mitochondrial protein synthesis (mitochondrial protein fractional synthesis rate [Mito FSR]), fat-free mass (FFM), strength (one repetition maximum [Rep Max]), and mitochondrial respiration (state 3 CI+II) in different age groups (young [Y] and older [O]). Black dots indicate significant (P < 0.05) association between module and physiological parameter within a given group. B–F: The molecular pathways (using Reactome database), cellular compartments (using DAVID software), and metabolites of five representative modules are shown. Only pathways and cellular components with a false discovery rate–corrected P value <0.05 are shown. GABA, γ-aminobutyric acid.