henhaobio
Sign in

bioRxivPhysiologyPreprint

Exercise induces time-dependent but not sex-specific transcriptomic changes in healthy human skeletal muscle

Dale F. TayloriD, Elizabeth G. ReismaniD, ANDREW P. GARNHAMiD, Romain BarrèsiD, Nolan J. HoffmaniD, John Alan HawleyiD, Nikeisha J. CaruanaiD, David John BishopiD

PDFDOICodeData

Abstract

Elucidating the time-dependent transcriptional response of skeletal muscle to exercise is essential for uncovering the molecular mechanisms that drive its health-promoting effects. However, previous studies have been limited by a small number of muscle biopsies, often collected at arbitrary time points post-exercise, and in predominantly male subjects. Using the most comprehensive skeletal muscle biopsy time course following a single session of exercise in both males and females, we identified over 16,600 differentially expressed genes (DEGs), including more than 7,000 novel exercise-responsive genes. Of these DEGs, the magnitude of differential expression in 60% of genes was influenced by cardiorespiratory fitness. Although most mitochondrial genes were differentially expressed after exercise, the majority were downregulated at 24-48 hours, suggesting that mitochondrial protein expression may be regulated by post-transcriptional regulation. Despite 1,193 genes showing sex-specific expression at baseline, exercise-induced gene expression differences were minimal between males and females, suggesting that when cardiorespiratory fitness and exercise stimulus are matched, skeletal muscle adaptations are similar between sexes. To enhance data accessibility, we created an interactive Shiny app (https://BishopLab.shinyapps.io/EXERgene/) that allows users to investigate specific genes and interrogate potential mechanisms of skeletal muscle adaptation to exercise.

In-depth summary

Claude Opus 5.5 · xhigh effort · 7 Nov 2026 · from the full text of version 1 (19 Aug 2025) · CC BY 4.0

Nine vastus lateralis biopsies from before exercise to 48 h after one high-intensity interval session reveal 16,625 exercise-responsive genes; only four respond differently between males and females.

As the authors describe it, earlier exercise transcriptome studies biopsied muscle at two or three arbitrary early time points, mostly in males, so many responsive genes were likely missed or mischaracterised and the assumed sex difference in response had little direct evidence. This dataset adds nine biopsies through +48 h in males and females matched for fitness and relative intensity, extending the list of exercise-responsive genes by more than 7,000 and testing sex effects directly.

Key results

  1. ShownSampling through +48 h shows that nearly every gene quantified in vastus lateralis responds to a single interval session, including more than 7,000 genes not reported as exercise-responsive in prior studies.

    16,625 of 17,345 analysed genes (96%) were differentially expressed genes (DEGs) vs PRE (pre-exercise) at one or more time points (adjusted p < 0.05, sexes pooled), covering 81% of previously reported exercise DEGs. Participant subsampling (100 iterations) indicated the larger cohort accounted for less than 15% of newly identified genes. · An emphasis on early post-exercise time points…; Statistical analysis; Figure legends

  2. ShownExpression changes keep accumulating for a day after exercise, downregulation outweighs upregulation from +3 h onward, and the apparent direction of well-studied genes such as PPARGC1A depends on when the biopsy is taken.

    Up- and downregulated DEG numbers grew at each time point until +24 h and fell slightly at +48 h; from +3 h through +48 h, downregulated genes outnumbered upregulated ones at every time point; over 4,100 genes went up at some times and down at others. NR4A3, the most upregulated gene, peaked at +3 h and returned to baseline by +48 h; PPARGC1A rose at +3 to +9 h, was not DE at +12 h, and fell at +24 and +48 h. · Exercise-induced gene expression is temporal and…

  3. SuggestedLower cardiorespiratory fitness (maximal oxygen uptake, VO2max) goes with larger exercise-induced expression changes, mainly at +24 and +48 h, with ribosome biogenesis and (at +12 h) mitochondrial gene expression among the most affected processes.

    VO2max ranged 18.4 to 40.0 mL•min^-1•kg^-1. DEGs whose log2 fold change correlated with VO2max (linear correlation, unadjusted p < 0.05) went from 3 at MID to 5,803 at +24 h and 1,733 at +48 h, mostly with larger changes at lower fitness; ribosomal biogenesis was the top enrichment at +24 and +48 h. Approximately 60% of all DEGs correlated with VO2max at one or more time points or a gene feature. · Higher cardiorespiratory fitness blunts the…; Minimal effect of sex on exercise-induced gene…; Statistical analysis

  4. SuggestedMost mitochondrial genes respond to exercise, but mainly by downregulation at +24 and +48 h, with members of the same pathway spread over six temporal clusters; the authors infer substantial post-transcriptional control of mitochondrial adaptation.

    1,037 of 1,136 MitoCarta3.0 genes (91%) were differentially expressed; 565 (50% of all mitochondrial genes) had not been reported as exercise-responsive. Upregulated mitochondrial DEGs peaked at +12 h, downregulated ones at +24 and +48 h, where they formed the majority. Six mfuzz clusters split every pathway; mRNA showed no consistent match to training-induced protein changes (Granata et al.). · Mitochondrial gene expression is primarily…; Figure legends

  5. ShownAlthough 1,193 genes differ by sex at rest, the exercise response is almost the same in fitness-matched males and females; the authors attribute sex-only DEGs largely to lower per-sex statistical power.

    At baseline 1,193 genes (6.9%) differed by sex (675 higher in males, 518 in females), over-representing X (10.6% of detected X genes) and Y (100%). After exercise 14,662 DEGs were shared, 810 male-only and 987 female-only, the sex-only ones with smaller mean log2 fold changes. Directly comparing responses, only SNX30, DNMBP, DAAM2-AS1 and CTIF of 17,345 genes differed; AUC analysis found no sex differences. · Minimal effect of sex on exercise-induced gene…; Figure legends

Methods, data and reproducibility

Bulk RNA-seq (DNBseq, 100 bp paired-end) of vastus lateralis biopsies at nine time points, PRE (pre-exercise) and MID through +48 h, around one cycling session of high-intensity interval exercise (HIIE). edgeR/limma voomLmFit tested change vs PRE; responses were also correlated with VO2max, mitochondrial genes soft-clustered (mfuzz) and cell types deconvolved (MuSiC).

  • Vastus lateralis biopsies from healthy untrained adults (18-40 y) at PRE, MID, +0, +3, +6, +9, +12, +24 and +48 h around one HIIE session · human · n = 20 males and 20 females; 353 samples sequenced
  • Tabula Sapiens single-cell RNA-seq atlas, used as reference for MuSiC cell type deconvolution · human (Tabula Sapiens is a human cell atlas)
  • Human Protein Atlas skeletal muscle protein-coding transcripts, used for coverage comparison · human (Human Protein Atlas) · n = 12,945 protein-encoding transcripts
  • ENCODE skeletal muscle transcriptome, used for coverage comparison
  • MitoCarta3.0 mitochondrial gene inventory · n = 1,136 genes
  • Published single-session exercise transcriptome studies and meta-analyses (including Pillon et al. and Amar et al.) used to label genes as previously reported
  • Granata et al. training-induced muscle protein abundance (PH vs BL comparison)
  • Lundell et al. 2020 circadian expression data for CLOCK, BMAL1, PER1 and CRY1
Methods & tools
DNBseq (BGI), SOAPnuke v2.1.7, STAR v2.7.11b, featureCounts v2.1.1, edgeR v4.4.2, limma v3.62.1, enrichR / clusterProfiler, mfuzz v2.66.0, MuSiC v1.0.0, lme4 / emmeans, pracma (AUC), R Shiny
Software & environment
R v4.4.3; edgeR 4.4.2; limma 3.62.1; clusterProfiler 4.14.6; enrichR 3.4; mfuzz 2.66.0; MuSiC 1.0.0; lme4 1.1-37; emmeans 1.11.2; pracma 2.4.4; SOAPnuke 2.1.7; STAR 2.7.11b; featureCounts 2.1.1; reference genome GRCh38.p14.
Reproducibility gaps
No accession for raw or processed RNA-seq data is printed; access is through the EXERgene Shiny app and supplementary tables. The code link is a GitHub user page, not a named repository.
Limitations (6)
  • No non-exercise control group, so circadian or time-of-day expression changes cannot be separated from the exercise response (PER1 and CRY1 appeared phase-shifted). · stated by the authors
  • Repeated biopsies may themselves trigger inflammatory or repair transcription; switching legs between +3 h and +6 h slightly reduced the 'regulation of skeletal muscle regeneration' response, and KLF5 and COL6A1 changes may reflect biopsy repair. · stated by the authors
  • Bulk homogenised muscle cannot resolve cell types; deconvolution found lower fast (+9 h) and slow (+24, +48 h) muscle cell proportions in females, which may explain female-only inflammation and circulation enrichments. · stated by the authors
  • Young, healthy, untrained participants, unstandardised menstrual phase and a single HIIE protocol limit transfer to other ages, health states, and modalities such as resistance, moderate continuous or sprint interval training. · stated by the authors
  • VO2max correlations use a plain p < 0.05 with no stated multiple-testing correction across more than 16,600 genes and several time points, so counts such as 5,803 correlated genes and the approximately 60% figure likely include many false positives. · noted in this summary
  • 'Not previously reported' gene counts rest on comparison with prior studies analysed with their own pipelines, thresholds, exercise modes and cohorts, so part of the more than 7,000 may reflect methodological differences rather than biology. · noted in this summary

Read this if you plan biopsy timing for exercise studies, want a densely sampled human muscle exercise transcriptome with sex and fitness metadata, or question whether exercise-induced mRNA predicts mitochondrial protein change.

Citations
3

Reproducibility ledger

Nobody has reported trying to reproduce this paper yet. Did you replicate an experiment or run the code? Your report saves the next lab a week.

Sign in to file a reproducibility report.