A Molecular Signature Defining Exercise Adaptation with Ageing and In Vivo Partial Reprogramming in Skeletal Muscle

Abstract

Exercise promotes functional improvements in aged tissues, but the extent to which it simulates partial molecular reprogramming is unknown. Using transcriptome profiling from 1) a skeletal muscle-specific in vivo Oct3/4, Klf4, Sox2, and Myc (OKSM) reprogramming-factor expression murine model, 2) an in vivo inducible muscle-specific Myc induction murine model, 3) a translatable high-volume hypertrophic exercise training approach in aged mice, and 4) human exercise muscle biopsies, we collectively defined exercise-induced genes that are common to partial reprogramming. Late-life exercise training lowered murine DNA methylation age according to several contemporary muscle-specific clocks. A comparison of the murine soleus transcriptome after late-life exercise training to the soleus transcriptome after OKSM induction revealed an overlapping signature that included higher JunB and Sun1. Also, within this signature, downregulation of specific mitochondrial and muscle-enriched genes was conserved in skeletal muscle of long-term exercise-trained humans; among these was muscle-specific Abra/Stars. Myc is the OKSM factor most induced by exercise in muscle and was elevated following exercise training in aged mice. A pulse of MYC rewired the global soleus muscle methylome, and the transcriptome after a MYC pulse partially recapitulated OKSM induction. A common signature also emerged in the murine MYC-controlled and exercise adaptation transcriptomes, including lower muscle-specific Melusin and reactive oxygen species-associated Romo1. With Myc, OKSM, and exercise training in mice as well habitual exercise in humans, the complex I accessory subunit Ndufb11 was lower; low Ndufb11 is linked to longevity in rodents. Collectively, exercise shares similarities with genetic in vivo partial reprogramming.

Graphical Abstract

Diverse forms of exercise training improve muscle function and whole-body health, even if initiated late in life. Information on conserved exercise-controlled molecular cues that underpin a younger muscle phenotype in aged muscle has potential utility in the development of anti-ageing therapies. Induction of the Yamanaka factors Oct3/4, Klf4, Sox2, and Myc are known to ameliorate ageing hallmarks. Comparison of transcriptomic data from aged exercise-trained mice and humans to muscle fibre-specific genetically driven models of epigenetic reprogramming (e.g. Yamanaka factor or Myc expression) unearthed conserved biomarkers associated with molecular age mitigation. Considering reduced biological age according to DNA methylome analysis, high-volume exercise training can be classified as an epigenetic reprogramming stimulus. Chronic exercise should be considered alongside and/or as a method to inform healthspan-extending longevity approaches such as pharmacologic and dietary interventions.

Introduction

The Yamanaka transcription factors Oct3/4, Klf4, Sox2, and Myc (OKSM) epigenetically revert somatic cells to a stem cell-like state (Takahashi & Yamanaka, 2006). Cellular reprogramming via OKSM is in part mediated by remodeling of the DNA methylome (Nishino et al., 2011; Gao et al., 2013; Lee et al., 2014; Chondronasiou et al., 2022). The expression of OKSM facilitates cellular plasticity that is characteristic of more youthful cells (Gill et al., 2022). The finding that OKSM drive molecular and cellular ageing mitigation (Ocampo et al., 2016b) prompted an interest in exploiting these factors to improve human health and lifespan (Ocampo et al., 2016a). In vivo expression of OKSM in mice combats several aspects of ageing (Ocampo et al., 2016b) including DNA methylation clock age (mDNAge) (Horvath, 2013; Olova et al., 2019), transcriptional perturbations (Chondronasiou et al., 2022), impaired tissue regeneration (Ocampo et al., 2016b; Chondronasiou et al., 2022), and mortality (Ocampo et al., 2016b; Alle et al., 2022). Phenotypes and transcriptomes associated with ameliorated ageing hallmarks are similarly observed with transient OKSM-mediated reprogramming of cells derived from aged humans (Sarkar et al., 2020; Gill et al., 2022). Potential issues with epigenetic reprogramming to alleviate ageing phenotypes are a loss of cellular identity (Olova et al., 2019; Roux et al., 2022) and neoplastic transformation (Friedmann‐Morvinski & Verma, 2014). In skeletal muscle, high levels of cell cycle inhibitors and expression of tumor-repressing genes provides terminally-differentiated muscle fibres resistance to tumorigenesis (Seely, 1980; Keckesova et al., 2017). Muscle has therefore been used as a model tissue to identify cancer-fighting strategies (Keckesova et al., 2017; Crist et al., 2022). Also due to its unique ability to resist primary and secondary tumor growth, skeletal muscle is an ideal tissue for studying the effects of epigenetic reprogramming via Yamanaka factors.

Mechanistic studies place skeletal muscle health at the center of whole-body health throughout the lifespan (Demontis et al., 2013; Rai et al., 2021). Exercise is considered the most effective and inexpensive strategy to improve muscle health and by extension whole-body health. Pharmacologic and dietary methods might extend healthspan and ameliorate ageing hallmarks as ageing progresses, but exercise training naturally elicits many of the same benefits and more (Furrer & Handschin, 2022). Despite this, exercise is often overlooked as a frontline approach to combat ageing hallmarks, restore cellular resiliency, and attenuate biological ageing (Zhang et al., 2022a). A popular drug-based therapy (metformin) used to combat ageing (Xenos et al., 2022) blunts various beneficial effects of endurance or resistance exercise training in the muscle of healthy non-diabetic humans (Konopka et al., 2019; Konopka & Miller, 2019; Walton et al., 2019). Another drug touted as “anti-ageing”, rapamycin (Blagosklonny, 2018), prevents hypertrophic muscle adaptation to loading (Bodine et al., 2001; Goodman et al., 2011). All-cause mortality and reduced quality of life are linked to reductions in muscle mass, strength, and/or power (Metter et al., 2002; Newman et al., 2006; Reid & Fielding, 2012; Li et al., 2018; de Santana et al., 2021; López-Bueno et al., 2022; Losa-Reyna et al., 2022; Winger et al., 2022). Exercise, and specifically resistance training, is a key countermeasure against ageing-associated sarcopenia (Talar et al., 2021; Grosicki et al., 2022). Preventing exercise-mediated muscle adaptation via supplementation with current anti-ageing drugs may therefore be objectionable. If exercise is a “polypill” (Fiuza-Luces et al., 2013; Hawley et al., 2021), understanding shared features of training adaptations and longevity interventions could be a prudent strategy for identifying molecular targets capable of conferring a “younger” phenotype.

To study the benefits of exercise during ageing, we applied a voluntary murine exercise training approach called progressive weighted wheel running (PoWeR) (Dungan et al., 2019; Murach et al., 2020) to aged mice (Dungan et al., 2022a). PoWeR in aged mice promotes adaptations in soleus muscle mass, cellular phenotype, and myonuclear DNA methylation reminiscent of what occurs in younger muscle with training (Wen et al., 2021; Dungan et al., 2022a). The current investigation leveraged PoWeR to identify exercise-associated genes in muscle that are linked to partial epigenetic reprogramming, a benchmark pre-clinical longevity intervention (Ocampo et al., 2016a; Ocampo et al., 2016b; Zhang et al., 2022a). We compared the in vivo gene expression landscape after PoWeR in skeletal muscle of aged mice to that of in vivo muscle-specific OKSM expression. Of OKSM, Myc mRNA and protein is highly exercise-responsive in human skeletal muscle (Trenerry et al., 2007; Broholm et al., 2011; Brook et al., 2016; Figueiredo et al., 2016; Stec et al., 2016; Townsend et al., 2016; Popov et al., 2019; Figueiredo et al., 2021). We recently reported that Myc induction specifically in muscle fibres is central to the acute mechanical loading-induced transcriptome in the plantaris of young mice (Murach et al., 2022). A corroborating result in muscle tissue was reported with frequent resistance-type training in rats (Viggars et al., 2022). Furthermore, the induction of Myc was recently linked to “biological immortality” in hydrozoans (Pascual-Torner et al., 2022). We therefore deployed a genetically driven muscle fibre-specific doxycycline-inducible model of in vivo Myc expression to explore shared features of Myc, OKSM partial reprogramming, and exercise adaptation transcriptomes in mice. We then compared these features to long-term exercise-trained human muscle biopsy transcriptomic data. Finally, we performed DNA methylome profiling after a pulse of MYC in muscle. Our findings point to new biomarkers of ageing-mitigation in skeletal muscle.

Methods

Ethical Approval

IACUCs at the University of Kentucky (UK, 2020–3535) and University of Arkansas (UA, AUP 21037) approved all animal procedures. Mice were housed in a temperature and humidity-controlled room, maintained on 12:12-h (UK) and 14:10-h(UA) light-dark cycles, and food and water were provided ad libitum throughout experimentation. All animals were sacrificed via cervical dislocation under deep anesthesia with inhaled isoflurane (UA) or lethal dosage of 0.1 ml of sodium pentobarbital injected intraperitoneally (UK).

Animals and Animal Procedures

Approximately 22-month-old female C57BL/6N mice were obtained from the Charles River Laboratories National Institute of Aging research colony; exact ages are not recorded for this colony, but the birth month is documented. 22 months in mice is estimated to correspond to ~73 human years (Dutta & Sengupta, 2016). Female mice were used for these experiments since they tend to run more in our hands. The C57BL/6N mice were used for the DNA methylation ageing experiments (mDNAge) and RNA-sequencing experiments. The mDNAge experiments were part of a related but separate study and were subcutaneously injected daily with a 0.9% NaCl sterile saline solution (Fisher Scientific, cat. 50–843-141) throughout the eight weeks of PoWeR. For all murine exercise training experiments, aged female mice were subjected to voluntary progressive weighted wheel running (PoWeR) as previously described by us (Murach et al., 2021; Dungan et al., 2022a). Mice in the PoWeR group were singly housed in cages with running wheels to allow monitoring of individual running distance using ClockLab software (Actimetrics, v6.1.01), and mice in the sedentary group were group housed in cages without running wheels. Following an introductory week with an unweighted wheel, 8 weeks of PoWeR training commenced with the following weight progression: 2g in week 1, 3g in week 2, 4g in week 3, and 5g in weeks 4–8. One-gram magnets (product no. B661, K&J Magnetics) were affixed to one side of the wheel to allow for the progressive increase in weight. Mice were approximately 24 months old upon completion of the experiment. PoWeR mDNAge mice ran 4.3±2.9 km/d (mean±standard deviation); running volume for PoWeR RNA-seq mice was reported previously (Dungan et al., 2022a). Following 8 weeks of PoWeR or 8 weeks as sedentary controls, mice were humanely euthanized by cervical dislocation under deep isoflurane anesthesia after a 24-hour wheel lock and overnight fast. Hindlimb muscles were then rapidly dissected and flash frozen.

For the Myc expression experiments, male HSA-rtTA^+/−-;TRE-Myc^+/− (HSA-Myc) were generated by crossing homozygous HSA-rtTA mice (Iwata et al., 2018) with heterozygous TRE-Myc mice (Felsher & Bishop, 1999); HSA-rtTA littermate mice were used as controls. All mice were genotyped as previously described (Iwata et al., 2018; Rutledge et al., 2020). HSA-Myc and littermate control mice were given access to doxycycline water overnight (0.5 mg/ml with 2% sucrose for 12 hours) at 2.5 months of age, and this water was replaced with un-supplemented water for 12 hours before being euthanized. Tissue was stored at −80°C until the time of RNA extraction.

DNA Methylation Age (mDNAge) Analysis

Samples were blinded until statistical analysis required unblinding. DNA was extracted from flash frozen gastrocnemius tissue by the Clock Foundation using a Qiagen DNeasy Blood & Tissue kit (cat. 69504 or 69506, Qiagen, Hilden, North Rhine-Westphalia, Germany) following standard protocols, and eluted into ultrapure, nuclease-free milliQ water. The Clock Foundation quantified the DNA and assessed purity via 260/280 absorption ratio. The isolated DNA then underwent bisulfite conversion at AKESOGen, Inc. (AKESOgen, Peachtree Corners, GA USA) using the Zymo EZ DNA methylation kit (Zymo, Irvine CA, USA). A minimum of 150 ng of total bisulfite-converted DNA was processed using a customized Infinium HD iSelect Methyl Custom 48-sample BeadChip termed the Illumina Horvath Mammalian Methylation Chip array kit (Clock Foundation, Los Angeles, CA USA), and scanned with an iScan system (Illumina, San Diego, CA, USA); samples from each treatment group were present on each chip to help eliminate chip-related artifacts. This custom methylation array targets more than 37,000 CpG methylation loci that are conserved across several mammalian species, the majority of which are found in mice (Arneson et al., 2022). Array output were normalized by the Clock Foundation using the R package SeSAMe, then values were input through an R pipeline with several clocks developed by the Horvath laboratory, including the Muscle Clock and the Developmental Muscle Clock (Figure 1A), each generated by the Horvath Laboratory using elastic net regression (Mozhui et al., 2022).

Figure 1.

Comparison of Oct3/4, Klf4, Sox2, and Myc (OKSM) expression to late-life progressive weighted wheel running (PoWeR) in mouse muscle A DNA methylation age (mDNAge) derived from clocks developed by the Horvath laboratory applied to late-life PoWeR (n=10) and sedentary control (n=9) mouse gastrocnemius muscle B Principal component analysis (PCA) plots showing DESeq2-normalized gene count transcriptome data from OKSM expression (n=2) versus control (n=2) in mouse soleus muscle fibres (Wang et al., 2021) and the late-life PoWeR (n=4) versus the sedentary (n=5) transcriptome in soleus muscle (Dungan et al., 2022a) C Chord diagram showing significantly downregulated genes common across OKSM and late-life PoWeR in the soleus D Upregulated genes common across OKSM and late-life PoWeR in the soleus E Downregulated genes common across OKSM, late-life PoWeR, and long-term exercise training. All data reported as mean ± standard deviation

Western Blotting for MYC

Western blots were carried out as previously described (Lim et al., 2022). Briefly, 20 mg of frozen quadriceps and heart muscle samples were powdered and homogenized in Laemmli buffer. Following RC/DC assay (BioRad, Hercules, CA, USA, cat. 500–0119), 40μg of total protein was subjected to SDS-PAGE using a 10% gel. Membranes were blocked with 5% of milk. Primary antibody incubation was conducted at 4C overnight using anti-MYC (D84C12 cat. 5605, Cell Signaling, Danvers, MA, USA) diluted 1:500. Secondary antibody (IRDye 800CW/680RD, LI-COR Biosciences, Lincoln, NE) was diluted 1:10,000 and membranes were imaged on LI-COR Odyssey FC using IR detection. All bands were normalized to the 45-kDa actin band of Ponceau S stain as a loading control.

RNA Isolation, Sequencing, and Analyses

Soleus RNA was isolated using TRI Reagent (Sigma-Aldrich, St Louis, MO, USA). Tissue was homogenized using beads and the Fisher Bead Mill (Fisher, Hanover Park, IL, USA). Following homogenization, RNA was isolated via phase separation by addition of bromochloropropane or chloroform then centrifugation. The aqueous phase was transferred to a new tube and further processed on columns using the Direct-zol Kit (Zymo Research, Irvine, CA, USA) (Figueiredo et al., 2021). concentration and purity were determined using a BioTek Take3 micro-volume microplate with a BioTek PowerWave XS microplate reader (BioTek Instruments Inc., Winooski, VT). Following library preparation with Poly A enrichment, RNA was sequenced by Novogene on an Illumina HiSeq using 150 bp paired-end sequencing, as we have performed previously (Wen et al., 2021; Dungan et al., 2022a).

FASTQ files (PoWeR and Myc experiments) or processed files from GEO (OKSM experiments) were aligned using STAR in Partek, then normalized and analyzed using DESeq2 (minimum read cutoff of ≤20 across all samples). P values were adjusted using the Benjamini-Hochberg false discovery rate (FDR) step-up method. Pathway over-representation analysis was performed using ConsensusPathDB (Kamburov & Herwig, 2022) with non-ordered query and up- or-downregulated genes with FDR adj. p<0.05; all KEGG and Reactome pathways reported had adjusted p values <0.05. PoWeR data were from soleus muscles of mice aged 24 months that was published previously (Dungan et al., 2022a) (GSE198652). Muscle was harvested 24 hours after the last exercise bout. OKSM data were from Wang et al. (Wang et al., 2021) (GSE148911) using male and female mice that were at least two months of age (<4.5 months old). OKSM induction in soleus muscle was 2.5 days. The human exercise data from Chapman et al. (43.5±5.2 y sedentary [n=8] and 41.8±5.9 y trained females [n=9], to match the female PoWeR mice) were already processed and presented in supplemental tables (Chapman et al., 2020). The samples were from biopsies of the vastus lateralis after a 72-hour cessation of activity and standardized dietary conditions. The online MetaMEx analysis tool from the Zierath laboratory was used to determine Yamanaka factor gene expression levels with exercise in skeletal muscle of humans (Pillon et al., 2020).

Reduced Representation Bisulfite Sequencing (RRBS)

DNA was extracted from soleus muscle using the Qiagen DNA MicroKit (Hilden, Germany). Following standard quality control, Msp1 RRBS was performed by Zymo Research (Irvine, California, USA) using 40 ng of genomic DNA. The tissue for RRBS analysis was the same as that used for RNA-seq. RRBS data were processed using MethylSig (Park et al., 2014) with a minimum cutoff of 10x coverage per CpG site in each sample, as we have previously described (Murach et al., 2021; Wen et al., 2021). Promoters were defined as within 1 kb upstream of the transcription start site (Wen et al., 2021). Differentially methylated sites were determined using a beta binomial distribution and p value <0.0005; Benjamini-Hochberg false discovery rate adjusted p values are reported for significant p value findings. Results are presented at the gene level, and site-specific CpG data are in a supplemental table.

Statistical Considerations

For mDNAge analysis, data were determined to be normal (Shapiro-Wilk test) and a two-tailed t-test was employed. The chord plot was generated using GOplot and in R version 4.1.1. The 87 genes significantly downregulated in both the OKSM versus Control comparison and the PoWeR versus Sedentary comparisons were input into gProfiler. The top KEGG, Reactome, and Wikipathway were input into GOplot for these genes (defined by the smallest adjusted p value, as calculated within the gProfiler platform), along with the GO molecular function protein binding since it contained many of the proteins not in the other top pathways. PCA plots were generated in Partek Genomic Suite. All other figures were generated using GraphPad Prism version 7.00 for Mac OS X (GraphPad Software, La Jolla, CA) and BioRender. Landscape In Silico deletion Analysis (Lisa, also called Cistrome) was run according to recommended procedures (Qin et al., 2020). In brief, differentially expressed gene lists (adj. p<0.05) from the PoWeR RNA-seq experiment were input into the online graphical user interface. The Cauchy combination p-value test was used to determine the overall influence of MYC. Data presented as mean ± standard deviation of mean unless otherwise stated.

Results

Late-life exercise training associates with lower skeletal muscle DNA methylation age using contemporary muscle-specific epigenetic clocks

To corroborate our previously-described exercise training-mediated shift in methylation age toward a younger status in muscle, female C57BL/6N mice (n=10) completed PoWeR from 22–24 months. 22-month-old mice are estimated to be equivalent to ~73 year old humans (Dutta & Sengupta, 2016). Age-matched untrained mice were sedentary controls (n=9). Muscle methylation was quantified with the Horvath Mammalian Methylation Chip that focuses on conserved cytosines (Arneson et al., 2022) and analyzed with contemporary murine mDNAge clocks (Mozhui et al., 2022). After PoWeR, skeletal muscle mDNAge was lower relative to sedentary using the Horvath Muscle Clock (−5.7 weeks, p=0.04) and Developmental Muscle Clock (−7.1 weeks, p=0.02) (Figure 1A). The magnitude of epigenetic age mitigation by late-life exercise in muscle was similar to our previous report using an independent cohort of mice and a different mDNAge analysis (8 weeks, n=5 per group, p=0.07) (Murach et al., 2021). Running volume was slightly lower in the current investigation (4.3 km/d vs 6.4 km/d). Consistently lower mDNAge with exercise motivated us to compare transcriptome profiles of a partial reprogramming stimulus (OKSM expression) versus exercise adaptation in muscle to identify a common gene expression signature.

The transcriptome resulting from OKSM expression in muscle fibres overlaps with the late-life exercise adaptation transcriptome in mouse muscle

RNA-seq data from the soleus muscle of adult mice after 2.5 days of genetic OKSM expression (Wang et al., 2021) was compared to RNA-sequencing data from the soleus of late-life PoWeR mice (Dungan et al., 2022a). The soleus adapts more substantially to PoWeR than other hindlimb muscles in young and old mice (Englund et al., 2020; Murach et al., 2020), with similar gene expression patterns regardless of age (Dungan et al., 2022a). The soleus also contains similar proportions of the primary myosin fibre types found in human vastus lateralis (Murach et al., 2019; Murach et al., 2020; Dungan et al., 2022a), further justifying our focus on this muscle.

OKSM induction and PoWeR revealed stark effects on gene expression for each intervention (Figure 1B). With OKSM, 412 genes were upregulated and 523 genes were downregulated (adj. p<0.05) (Supplemental Table 1). Pathway analysis of OKSM induction revealed repression of genes associated with oxidative phosphorylation, in agreement with the analysis from the Belmonte laboratory (Wang et al., 2021). We compared the list of genes downregulated by OKSM with those downregulated by PoWeR (377 genes, adj. p<0.05) (Supplemental Table 2). About one quarter of downregulated genes overlapped (87 genes, Figure 1C). Pathway analysis of shared downregulated genes revealed oxidative phosphorylation (KEGG, adj. p=9.77×10⁻⁵⁴), the citric acid cycle (Reactome, adj. p=8.31×10⁻⁵⁶), and electron transport chain (Wikipathways, adj. p=2.08×10⁻⁴⁴) as the top pathways (Figure 1C). The muscle-enriched genes fast skeletal muscle troponin (Tnni2) (Sheng & Jin, 2016) and striated muscle activator of Rho signaling (Stars, or Abra) (Arai et al., 2002; Wallace et al., 2012) were also lower with OKSM and PoWeR; this suggested a repression of muscle cellular identity, consistent with partial reprogramming effects. Seventeen upregulated genes were common between OKSM and PoWeR datasets (Figure 1D), including the potent muscle mass-regulator JunB (Raffaello et al., 2010) and myonuclear anchor Sun1 (Lei et al., 2009). Perhaps alteration to Sun1 could be related to myonuclear shape changes that occur with exercise training in rodents (Murach et al., 2020, Rader and Baker, 2022). Collectively, our analysis unearthed a canonical skeletal muscle gene expression signature common between OKSM and exercise adaptation.

Gene expression alterations with OKSM and exercise adaptation in mouse muscle are conserved in human muscle after prolonged exercise training

To validate our murine findings in humans we queried an RNA-sequencing dataset from muscle biopsy samples of women with at least 15 years of mixed-modality endurance training history (n=9, ~42 years old) (Chapman et al., 2020). Comparing our murine PoWeR data to long-term training in humans seemed appropriate as 8 weeks of high-volume PoWeR corresponds to ~10% of the mouse lifespan and could thus be characterized as prolonged training. Several mitochondrial-related genes downregulated by OKSM and PoWeR were significantly lower in muscle of long-term endurance-trained athletes relative to sedentary age-matched controls (COX6A2, adj. p=0.046; NDUFB11, adj. p=0.017; NDUFS7, adj. p=0.041; TOMM7, adj. p=0.028). Downregulation of mitochondrial genes with endurance training seems counterintuitive. To further explore this observation, we compared genes that were altered in female endurance trained muscle to the human MitoCarta 3.0 database (Rath et al., 2021). While 100 genes that were higher with training overlapped with MitoCarta, 52 genes that were lower also overlapped; these included NDUFs such as NDUFAF3 and NDUFS7 as well as COX6A1 and COX6A2 (Supplemental Table 3). Thus, endurance training adaptations in skeletal muscle are not characterized by unanimously elevated mitochondrial gene levels. In addition to mitochondrial genes, muscle-enriched TNNI2 and ABRA (STARS) were lower in exercise-trained human muscle (adj. p=0.002 and 0.003, respectively), consistent with OKSM expression and late-life PoWeR in mice (Figure 1E).

Myc is an exercise-responsive transcription factor in muscle that influences the gene expression landscape with PoWeR in aged mice

The OKSM genetic mouse model was designed to induce all four Yamanaka factors, but Klf4 gene expression was not increased significantly in the soleus in the OKSM experiment (Wang et al., 2021); we corroborate this in our analysis (1.05 FC OKSM vs Sedentary, adj. p=0.95). All four factors can contribute to muscle adaptation via a change in activity independent of expression levels, but partial reprogramming was likely most attributable to Oct3/4, Sox2, and Myc. Of OKSM, only MYC is significantly induced by a bout of exercise in skeletal muscle from healthy humans according to a transcriptome meta-analysis (n=64–187 subjects, acute resistance and endurance exercise) (Pillon et al., 2020) (Figure 2A). Worth noting is that the KLF4 response is quite variable but may be induced in the first few hours after endurance exercise. Using up- and downregulated differentially expressed gene lists from our PoWeR RNA-seq data in the soleus, we performed Landscape in silico deletion analysis (Lisa/Cistrome) for insight on the potential contribution of MYC in the control of gene expression (Qin et al., 2020; Murach et al., 2022). Lisa incorporates transcriptome input data with an extensive library of publicly-available transcription factor ChIP-seq and global chromatin accessibility profiles to infer transcriptional regulators. According to Lisa, MYC is in the top 2% of transcription factors (TFs) predicted to influence upregulated genes in the PoWeR-trained transcriptome of aged mice (8 of 536 TFs) (Figure 2B), and in the top 5% for downregulated genes (19 of 536) (Supplemental Table 4). MYC was the highest ranked among Yamanaka factors. MYC as a primary TF controlling exercise adaptation in muscle is consistent with human biopsy data (Popov et al., 2019). MYC transcript and protein localizes in myonuclei during developmental and hypertrophic muscle growth in muscle of young animals (Alway, 1997; Veal & Jackson, 1998; Armstrong & Esser, 2005; Murach et al., 2022). Myc was 60% higher after late-life exercise in the soleus of PoWeR versus sedentary mice (24 hours after the final exercise bout, p=0.013, adj. p=0.14, Figure 2C). We therefore sought to explore the specific effects of a pulse of MYC in muscle fibres on muscle gene expression and compare it to the effects of OKSM and exercise training.

Figure 2.

Identification of MYC-controlled genes in soleus muscle that also associate with late-life exercise and OKSM induction A Data from the MetaMEx human muscle and exercise meta-analysis tool from the Zierath laboratory (Pillon et al., 2020) showing the responsiveness of Oct3/4, Klf4, Sox2, and Myc to endurance and resistance exercise in humans. Input parameters: Male/Female, Young, Sedentary/Active/Athlete, Lean, Vastus Lateralis, Healthy, Immediate/0/3/4/5/6/8/18/24 hours. Adjusted p values are for each individual gene and condition. B Lisa analysis of transcription factors (TFs) controlling upregulated genes in the PoWeR-trained transcriptome C Myc gene expression in the soleus after 8 weeks of PoWeR from 22–24 months in mice (Dungan et al., 2022a) D Western blot showing MYC protein in skeletal muscle and heart 12 hours after removal of overnight doxycycline in biological replicate MYC and control mice, with corresponding Ponceau S stain beneath E PCA plot of full soleus DESeq2 normalized gene count transcriptome dataset from MYC (n=3) and control (n=4) mice F Telomerase complex genes controlled by MYC in the soleus (SOL) (n=3 control, n=4 MYC) and quadriceps (QUAD) (n=2 control, n=2 MYC) muscles G Upregulated genes in the soleus common to MYC and late-life PoWeR H Downregulated genes in the soleus common to MYC and late-life PoWeR I Gene expression signature common to MYC, PoWeR, and OKSM in the murine soleus. All data reported as mean ± standard deviation

Genetically controlled muscle-specific Myc induction recapitulates aspects of OKSM expression and exercise training adaptation

We generated a doxycycline-controlled muscle fibre-specific Myc expression model by crossing the HSA-rtTA mouse (Iwata et al., 2018) with the TetO-Myc mouse (Felsher & Bishop, 1999). This mouse is a tool for understanding the temporal control of MYC target genes in skeletal muscle fibres in vivo. Young adult mice had access to doxycycline in drinking water overnight resulting in Myc induction. Doxycycline was removed in the morning and soleus muscle was collected 12 hours later (n=3 HSA-rtTA littermate controls, n=4 HSA-Myc). MYC protein was elevated at 12 hours in muscle and global soleus gene expression was markedly altered at this time (Figure 2D and E). Inputting the MYC-controlled genes (adj. p<0.05) from the soleus into Lisa (Qin et al., 2020), MYC is predicted as the highest ranked transcription factor regulating induced genes (data not shown); this confirms the veracity of Lisa and agrees with our previous findings in the plantaris muscle (Murach et al., 2022). MYC induced 1001 genes and repressed 391 genes in the soleus (adj. p<0.05) (Supplemental Table 5). Ribosomal genes were the most upregulated (adj. p=3.3×10⁻²⁹, KEGG) and muscle contraction genes were downregulated (adj. p=0.0009, Reactome, Figure 2E). MYC can transiently repress cellular identity (Sullivan et al., 2022), and downregulation of muscle-enriched genes corresponds with OKSM partial reprogramming in myofibres (see Figure 1). We previously reported that a pulse of MYC downregulated Nr1d2 (Reverbβ) and upregulated Rpl3 in the plantaris muscle of the mice used here (Murach et al., 2022). These findings were corroborated in the soleus and we also report a significant repression of Rpl3l by MYC (adj. p=0.041) (Supplemental Table 5). This inverse pattern of Rpl3 (induction) and Rpl3l (repression) is indicative of a developmental- and growth-oriented gene expression program in skeletal muscle (Zhang et al.; Chaillou et al., 2016; Chaillou, 2019; Kao et al., 2021). OKSM expression versus MYC induction revealed 39 upregulated genes and 31 downregulated genes in common. Ribosome biogenesis-related genes including Akt1, Mdn1, and Wdr3 were commonly upregulated and mitochondrial genes including Apoo, Ndufa1, Ndufb11, Romo1, and Pdcd5 were commonly downregulated (Supplemental Table 6). MYC recapitulated several aspects of OKSM-mediated partial reprogramming.

To further define the gene expression profile induced by MYC, we performed RNA-seq on the quadriceps muscles from a subset of HSA-Myc mice (n=2 control and n=2 Myc). The quadriceps are considerably larger than the soleus and are also involved in murine wheel training. Quadriceps were less affected by MYC and fewer genes were altered (40 upregulated and 26 downregulated at adj. p<0.10) (Supplemental Table 7) relative to the soleus and the plantaris, the latter of which we reported previously (Murach et al., 2022). Thus, the magnitude of MYC regulation may be muscle-specific, and could depend on muscle function, fibre type distribution, and/or metabolic profile. Pathway analysis of genes altered by MYC induction in the quadriceps revealed downregulation of muscle contraction genes (e.g. various myosins and troponins) consistent with what is observed in the soleus (current study) and plantaris (Murach et al., 2022). MYC induced Dkc1, Gar1, and Nhp2 in the quadriceps which were also markedly upregulated by MYC in the soleus (Figure 2F). These genes are members of the telomerase complex and associated with telomere maintenance (Gu et al., 2008; Vulliamy et al., 2008). Telomere shortening is associated with ageing (López-Otín et al., 2013). MYC is known to activate telomerase (Wang et al., 1998) which facilitates peripheral nervous system regeneration (Ma et al., 2019). Alterations to telomerase genes further implicate MYC as a potentially age-mitigating reprogramming factor.

We next asked whether gene expression in the soleus after PoWeR in aged mice matched the MYC-controlled soleus transcriptome. Thirty-eight genes overlapped (19 up- and 19 down-regulated, adj. p<0.05, Figure 2G and H). Itgb1bp2 (Melusin), Ndufb11, and Romo1 were downregulated by MYC, PoWeR, and OKSM (Figure 2I). Repression of the mitochondrial complex I accessory subunit Ndufb11 was common to all datasets including long-term human exercise training (adj. p=0.02) (Chapman et al., 2020) (Figure 2E).

A pulse of MYC reshapes the global muscle methylome

Using the same soleus tissue from the RNA-seq experiment above, we performed RRBS to assess the global DNA methylome. At twelve hours after overnight exposure to doxycycline in HSA-Myc mice, the muscle methylome underwent global CpG remodeling (>1,000 distinct differentially methylated CpGs across promoters, exons, and introns at p<0.0005 and 10x coverage across all samples) relative to controls. MYC caused slight hypomethylation of promoter regions, and hypermethylation of exons and introns relative to controls (Figure 3A) (Supplemental Table 8). There was some agreement between promoter methylation status and gene expression mediated by MYC. Three genes (Cfap298, Marchf5, and Mettl1) had a hypomethylated promoter CpG and higher gene expression with MYC induction (Figure 3B and C). Cfap298 controls cilia motility and polarization (Jaffe et al., 2016), but its role in skeletal muscle is not understood. Marchf5 (also called March5 or Mitol) is an E3 ubiquitin ligase required for mitochondrial fission and regulates mitochondrial dynamics (Nie et al., 2021). Mettl1 is an RNA methyltransferase, which is noteworthy since MYC was recently identified as a regulator of RNA methylation (Jansson et al., 2021). Perhaps global gene expression and methylation status would align more closely after a longer and/or more frequent period of MYC induction, or if we used a more comprehensive DNA methylation analysis. We previously published low-input RRBS from soleus myonuclei of late-life PoWeR mice versus sedentary controls (Dungan et al., 2022a). There was modest agreement in gene-level regulation of promoter and exon regions when comparing the soleus MYC methylome to the myonuclear methylome after PoWeR (Supplemental Table 9). A larger number of common genes with broadly intersecting regulation was apparent in introns (Figure 3D and E). Collectively, these data suggest that MYC regulates DNA methylation in skeletal muscle and could be sufficient to induce epigenetic reprogramming, even after brief exposure.

Figure 3.

A pulse of MYC alters the DNA methylome in skeletal muscle A Soleus muscle DNA methylation in promoter, exon, and intron regions of HSA-Myc (n=4) relative to control mice (n=3), p<0.0005 and 10x coverage per CpG site for MYC methylation data B Promoter CpG DNA methylation and C gene expression of Cfap298, Mitol (Marchf5), and Mettl1 after a pulse of MYC, p<0.0005 and 10x coverage per CpG site for MYC methylation data D Common gene-level intron hypo- and E hyper-methylation with MYC induction in the soleus compared to soleus myonuclear DNA methylation after 8 weeks of late-life PoWeR (Dungan et al., 2022a).*=multiple sites of regulation in one or both conditions, but only data from one site is shown. All data reported as mean ± standard deviation

Discussion

In the current investigation, we report: 1) a biological age-mitigating effect on the epigenetic landscape by late-life exercise-training in murine skeletal muscle, 2) a common gene expression signature of partial reprogramming by OKSM and exercise training in muscle of humans and aged mice, and 3) that Myc is an exercise-responsive factor that contributes to a rewired molecular profile at the transcriptome and methylome levels.

Lower muscle epigenetic age using updated and customized mDNAge clocks with exercise training corroborates and expands on our initial report in mice (Murach et al., 2021). Studies examining the muscle methylome after chronic exercise in aged humans also support our findings (Sailani et al., 2019; Blocquiaux et al., 2021; Ruple et al., 2021). Lower mDNAge in muscle appears to be a feature of the molecular environment with exercise adaptation that could have practical consequences for biological ageing and muscle performance. Indeed, emerging evidence suggests that blood mDNAge inversely associates with grip strength and frailty in ageing humans (Verschoor et al., 2021; Peterson et al., 2022), and is lower following a dietary/physical activity intervention in postmenopausal women (Fiorito et al., 2021). Exercise training and OKSM share substantial gene expression overlap. Given the known phenotypic and functional benefits of exercise adaptation in aged animals (Dungan et al., 2022a), an altered transcriptome after brief OKSM-mediated epigenetic partial reprogramming in skeletal muscle could have functional consequences. Prior work shows that OKSM induction in muscle enhances regeneration after injury (Ocampo et al., 2016b; Wang et al., 2021), but more work is needed in the context of muscle mass, strength, quality, fatigue resistance, and senescent cell abundance with ageing. Exploration into the interaction of epigenetic reprogramming via OKSM in combination with exercise training is also warranted.

Both OKSM and late-life exercise adaption in mice as well as chronic exercise training in humans results in lower levels of Abra/Stars in muscle. Exploratory proteomics recently found that low ABRA/STARS in human skeletal muscle is strongly associated with high physical activity throughout the lifespan (Ubaida-Mohien et al., 2019). ABRA/STARS has numerous functions in muscle (Lamon et al., 2014), but knockdown in myotubes enhances insulin signaling and sensitivity (Jin et al., 2011). Since insulin resistance associates with ageing (Fink et al., 1983; Barzilai & Ferrucci, 2012), lower ABRA/STARS in muscle could be beneficial for improving muscle health throughout the lifespan. Itgb1bp2 (Melusin), Ndufb11, and Romo1 were downregulated by OKSM expression, late-life exercise training, as well as a pulse of MYC in the muscle of mice. Melusin is specific to striated muscle and increases during regeneration (Brancaccio et al. 1999). Conversely, Melusin decreases during atrophy (Vitadello et al. 2020) but its role in exercise adaptation is unclear. ROMO1 generates reactive oxygen species (Chung et al., 2006) and induces senescence (Chung et al., 2008). Downregulation of Romo1 may be beneficial in ageing muscle where senescent cells can manifest and impair adaptation (Dungan et al., 2021; Dungan et al., 2022b; Zhang et al., 2022b; Zhang et al., 2022c). ROMO1 is also a negative feedback regulator of MYC (Lee et al., 2011), so its downregulation by MYC induction seems intuitive. Repressed gene expression of the complex I accessory subunit Ndufb11 with OKSM, Myc, and exercise training in mice is conserved with chronic exercise in humans. Reduced complex I activity prevents reactive oxygen species production which enhances cellular fitness and longevity (Rodríguez-Nuevo et al., 2022). Low abundance of specific complex I components including NDUFB11 is a characteristic of tissue from long-lived animals (Miwa et al., 2014; Sahm et al., 2018). Lower Nudfb11 with chronic exercise, likely mediated by MYC, could be part of a larger adaptive response that defends trained skeletal muscle against oxidative damage (Criswell et al., 1993; Powers et al., 1999; Radak et al., 1999; Parise et al., 2005a; Parise et al., 2005b; Smuder et al., 2011). On balance, studies with ageing and exercise in muscle of humans report a disconnect between mitochondrial gene and protein abundance (Robinson et al., 2017; Tumasian III et al., 2021). Our results should be interpreted with this in mind. Future Ndufb11 gain- and loss-of-function experiments in muscle may be warranted to elucidate its potential role in combatting an aged muscle phenotype.

Myc gene and protein expression is induced in the muscle of young adult humans (Trenerry et al., 2007; Broholm et al., 2011; Brook et al., 2016; Figueiredo et al., 2016; Stec et al., 2016; Townsend et al., 2016; Popov et al., 2019; Figueiredo et al., 2021) and rodents (Whitelaw & Hesketh, 1992; Chen et al., 2002; Armstrong & Esser, 2005; Lai et al., 2010; von Walden et al., 2012; Goodman et al., 2015; Kirby et al., 2016; West et al., 2016; Murach et al., 2022) in response to loading. MYC participates in epigenetic reprogramming in concert with Oct3/4, Klf4, and Sox2 (Takahashi & Yamanaka, 2006), but can also facilitate epigenome remodeling on its own (Brenner et al., 2005; Gartel, 2006; Lin et al., 2009; Nakagawa et al., 2010; Poole et al., 2017; Pang et al., 2018; Poole et al., 2019). Consistent with work in other cell types, there was global remodeling of the muscle DNA methylome following MYC induction. We speculate this could be due to MYC’s interactions with DNA methyltransferases (DNMTs) and ten eleven translocases (TETs) (Brenner et al., 2005; Pang et al., 2018; Poole et al., 2019). It is striking that <24 h of MYC induction can change muscle DNA methylation status, but a single bout of exercise also remodels DNA methylation in human muscle tissue (Barres et al., 2012; Seaborne et al., 2018; Maasar et al., 2021). The mechanisms of methylation regulation in muscle by exercise are unclear (Small et al., 2021; Villivalam et al., 2021), but MYC could be a central factor. Recent evidence suggests that overexpression of MYC in aged oligodendrocyte progenitors is sufficient to restore regenerative remyelination in vivo (Neumann et al., 2021); this is consistent with MYC’s role in regulating in vivo sensory nerve regeneration (Ma et al., 2019). Short-term MYC expression recapitulates numerous aspects of myonuclear gene expression at the onset of rapid overload-mediated muscle hypertrophy (Murach et al., 2022). Two weeks of MYC overexpression also mimics the protein synthesis response to exercise-like high-intensity muscle contractions (Mori et al., 2020). Since MYC-controlled gene expression in muscle fibres overlaps with OKSM partial reprogramming and the exercise training-mediated transcriptomes of aged mice, we propose that pulses of Myc could serve to enhance muscle function.

A 60% induction of Myc in the soleus after late-life PoWeR is intriguing since whole-organism MYC knockdown may enhance longevity and healthspan in mice (Hofmann et al., 2015). By contrast, MYC induction in specific cell populations can attenuate cellular ageing and restore regenerative potential (Neumann et al., 2021). Myc gene and protein in muscle tissue may increase in rodents (Hofmann et al., 2014; Mobley et al., 2017) but not humans during ageing (Drummond et al., 2011; Stec et al., 2015). The behavior of MYC in muscle with ageing is likely complex and could be differentially affected in mononuclear proliferative cells versus multinuclear post-mitotic muscle fibres within muscle tissue. Single cell RNA-sequencing in muscle revealed Myc enrichment specifically in fibro/adipogenic progenitors, satellite cells, and tenocytes but not myonuclei of aged mice (Zhang et al., 2022c). Elevated Myc in satellite cells of aged mice has been described previously (Price et al., 2014). Cell type-specific upregulation of Myc may explain differing results at the tissue level. Aspects of Myc’s gene expression network, which is operative after muscle contraction when young (Popov et al., 2019; Murach et al., 2022), may be alternatively regulated after resistance training in human muscle when aged (Phillips et al., 2013). MYC gene expression in vastus lateralis muscle samples four hours after a bout of resistance exercise in ~80 year old men and women is blunted (logFC=0.64 and −0.54, respectively, adj. p>0.90) relative to ~24 year old men and women (logFC=1.84 and 1.73, respectively, adj. p<0.05, data extracted from MetaMEx) (Raue et al., 2012; Pillon et al., 2020). Blunted MYC gene and protein responsiveness to acute resistance exercise in muscle of younger, albeit still aged humans (~70 y) has also been observed (Rivas et al., 2014; Brook et al., 2016). Perhaps an attenuated MYC response in muscle fibres of very old humans and animals (Alway, 1997) explains diminished adaptability to exercise, specifically in glycolytic Type 2 muscle fibres (Slivka et al., 2008; Raue et al., 2009; Grosicki et al., 2022). We speculate that higher Myc levels in the oxidative soleus muscle after high-volume PoWeR in aged mice is favorable and could contribute to the notable cellular adaptations observed previously (Dungan et al., 2022a).

Tighter control over the duration of OKSM and Myc dosing, the timing of sampling after exercise and genetically-induced epigenetic partial reprogramming, the mode of exercise and training status, the potential effects of biological sex, and several other factors may identify a larger gene expression signature of exercise-associated reprogramming-linked genes in muscle. Importantly, understanding whether epigenetic reprogramming, specifically by MYC, induces functional consequences in skeletal muscle is essential. Such insight will help drive healthspan-extending therapeutic efforts forward and advance the ageing field beyond biomarker discovery. Since DNA demethylation alterations are required for cellular reprogramming (Simonsson & Gurdon, 2004), the precise mechanisms by which MYC reshapes the methylome and global epigenome in muscle also deserves further exploration. Ultimately, the mechanisms of muscle molecular reprogramming required to fully recapitulate a youthful phenotype remain elusive, but our data provide a roadmap for further examination of how exercise training combats aspects of ageing. Our data also implicate MYC as an exercise-induced reprogramming factor in skeletal muscle.

Supplementary Material

Supplementary Tables S1-S9

Supplemental Table 1 Gene expression in the soleus muscle of OKSM versus control mice (Wang et al., 2021, GEO GSE148911)

Supplemental Table 2 Gene expression in the mouse soleus after 8 weeks of PoWeR from 22–24 months of age versus sedentary controls (Dungan and Brightwell et al. 2022, GEO GSE198652)

Supplemental Table 3 Genes altered by long-term endurance training in human females that overlap with MitoCarta 3.0 (DEGs from Chapman et al. 2020)

Supplemental Table 4 Lisa/Cistrome analysis using up- and downregulated DEGs (adj. p<0.05) in the soleus muscle after late-life PoWeR

Supplemental Table 5 Gene expression in the mouse soleus of MYC versus controls

Supplemental Table 6 Overlap of genes altered by MYC and OKSM expression in the mouse soleus

Supplemental Table 7 Gene expression in the mouse quadriceps of MYC versus controls

Supplemental Table 8 DNA methylation in the mouse soleus muscle DNA after a pulse of MYC

Supplemental Table 9 Genes with CpGs that overlap between MYC induction in the soleus muscle and myonuclear DNA following 8 weeks of PoWeR in aged mice (Dungan and Brightwell et al. 2022)

Key Points.

• Advances in the last decade related to cellular epigenetic reprogramming (e.g. DNA methylome remodeling) toward a pluripotent state via the Yamanaka transcription factors Oct3/4, Klf4, Sox2, and Myc (OKSM) provide a window into potential mechanisms for combatting the deleterious effects of cellular ageing

• Using global gene expression analysis, we compared the effects of in vivo OKSM-mediated partial reprogramming in skeletal muscle fibres of mice to the effects of late-life murine exercise training in muscle

• Myc is the Yamanaka factor most induced by exercise in skeletal muscle, so we compared the MYC-controlled transcriptome in muscle to Yamanaka factor-mediated and exercise adaptation gene landscapes in mice and humans

• A single pulse of MYC is sufficient to remodel the muscle methylome

• We identify partial reprogramming-associated genes that are innately altered by exercise training and conserved in humans, and propose that MYC contributes to some of these responses

Biography

Ronald G. Jones III is a first-year doctoral student working with Dr. Kevin A. Murach in the Molecular Muscle Mass Regulation (M3R) Laboratory at the University of Arkansas. Ronald received his undergraduate degree in Exercise Science from Towson University in Baltimore, Maryland. He completed a master’s degree in Clinical Exercise Physiology at the University of Delaware. Ronald is a Certified Clinical Exercise Physiologist under the American College of Sports Medicine and worked in the Baltimore VA Medical Center as a Research Physiologist. His research focus includes the molecular and cellular mechanisms of skeletal muscle physiology and has a particular interest in -omics.