Recurrent training rejuvenates and enhances transcriptome and methylome responses in young and older human muscle

Abstract

Background

The interaction between the muscle methylome and transcriptome is understudied during ageing and periods of resistance training in young, but especially older adults. More information is needed on the role of retained methylome training adaptations in muscle memory to understand muscle phenotypical and molecular restoration after inactivity or disuse.

Methods

We measured CpG methylation (microarray) and RNA expression (RNA sequencing) in young (n = 5; age = 22 ± 2 years) and older (n = 6; age = 65 ± 5 years) vastus lateralis muscle samples, taken at baseline, after 12 weeks of resistance training, after training interruption (2 weeks of leg immobilization in young men, 12 weeks of detraining in older men) and after 12 weeks of retraining to identify muscle memory‐related adaptations and rejuvenating effects of training.

Results

We report that of the 427 differentially expressed genes with advanced age (FDR < 0.1), 71% contained differentially methylated (dm)CpGs in older versus young muscle (FDR < 0.1, M‐value difference >0.4). The more dmCpGs within a gene, the clearer the inverse methylation–expression relationship. Around 73% of the age‐related dmCpGs approached younger methylation levels when older muscle was trained for 12 weeks. A second resistance training period after training cessation increased the number of hypomethylated CpGs and upregulated genes in both young and older muscle. We found indication for an epi‐memory within pro‐proliferating AMOTL1 in young muscle and mechanosensing‐related VCL in older muscle. For the first time, we integrate muscle methylome and transcriptome data in relation to both ageing and training‐induced/inactivity‐induced responses and identify focal adhesion as an important pathway herein.

Conclusions

This preliminary evidence indicates that previously trained muscle is more responsive to training than untrained muscle at methylome and transcriptome level and recurrent resistance training can partially restore ageing‐induced methylome alterations.

Introduction

Age-related skeletal muscle decay is the consequence of primary and secondary ageing.¹ The first is described as the inevitable deterioration of cellular structures and biological functions with little chance of reversal and inter-individual differences caused by genetic factors.² Secondary ageing, on the other hand, is the result of harmful lifestyle choices and environmental factors. These harmful factors can influence epigenetic mechanisms, defined as events that regulate gene expression without changing the DNA sequence. One of the most studied epigenetic events is DNA methylation,³ which mostly refers to CpG methylation and implies the binding of a methyl group to a 5′-cytosin-phosphate-guanine-3′ dinucleotide (CpG). Hypermethylation of CpG-dense promoter regions of actively transcribed genes has been associated with transcriptional repression, also known as ‘gene silencing’.⁴ The methylome of ageing tissues is characterized by global CpG hypomethylation, loci-specific differential methylation and more divergence between DNA methylation patterns of different individuals with the end result leading to aberrant gene expression, reactivation of transposable elements and genomic instability.^5,6 Age-related methylation changes in muscle tissue, on the other hand, appear distinct compared with other tissues.⁷ Recently, two genome-wide DNA methylation analyses identified 5518⁸ and 17 994⁹ differentially methylated CpGs (dmCpGs) between old and young skeletal muscle with most of them hypermethylated in older adults. The integration between age-related muscle methylome and transcriptome changes remains yet to be studied and is a crucial step in understanding the molecular mechanism behind muscle ageing.

As DNA methylation modifications are reversible, adopting a healthy lifestyle could potentially restore the age-associated damage to the muscle methylome and promoting physical activity is one strategy to do this. For instance, endurance exercise interventions are associated with altered methylation levels in genes involved in metabolism, cardiovascular adaptations, inflammatory and immunological processes, and muscle remodelling in young and middle-aged muscle.^10–13 More recently, methylome modifications comparing change of direction to straight line running also identified a more explicit hypomethylated signature for change of direction running in genes within protein binding, MAPK, AMPK, insulin, and axon guidance pathways.¹⁴ These beneficial methylome adaptations could even be passed on to later generations, as has been observed in mice.¹⁵ In older muscle, lifelong physical activity induced promoter hypomethylation of genes involved in energy metabolism, myogenesis, contractile properties, and oxidative stress.¹⁶ The impact of resistance training (RT) on the muscle methylome is less documented in literature, especially in elderly, but it has been suggested that RT has a distinct effect on the methylation status of, among other, growth-related genes.^13,17 Understanding the impact of RT on the muscle methylome is important, as resistance exercises can contribute to regaining muscle mass and strength and should be part of the exercise programme designed to counteract muscle ageing.

Nevertheless, lifelong training without training interruptions seems utopian as training can be occasionally ceased for various reasons. Fortunately, training-related muscle adaptations, and especially strength gains, are preserved longer than the initial training period and are often more rapidly regained following retraining.^18–20 In addition, the basal and acute RT-induced transcription of key genes involved in muscular adaptations appears to be influenced by a previous RT period.²¹ This so-called muscle memory has been partially attributed to motor learning¹⁹ and linked to the retention of myonuclei during muscle atrophy,²² as well as to lasting epigenetic modifications.¹⁷ The slow reversibility of methylation has been, for instance, seen when methylation changes in young skeletal muscle following 5 days of high-fat diet only partially and non-significantly reversed after 6 to 8 weeks.²³ The epigenetic retention of prior external stimuli in a way that re-encountering these stimuli later in life is associated with an enhanced muscle phenotypical response compared with the initial response (i.e. epi-memory²⁴), has been described in relation to metabolic (e.g. obesity²⁵), catabolic (e.g. TNF-α²⁶), and anabolic (e.g. RT¹⁷) stimuli. As DNA methylation modifications are transient in nature after one bout of exercise,²⁷ the question remains how long beneficial methylation alterations due to an extended period of RT are retained and whether multiple re-encounters are necessary to accumulate and stabilize these changes. Recently, a number of genes were identified in young skeletal muscle with a possible ‘epi-memory’.^17,28 These genes displayed hypomethylation after 7 weeks of RT and continued to be hypomethylated during 7 weeks of detraining, leading to enhanced gene expression following retraining. The number of hypomethylated dmCpGs following retraining was also twice the number compared with following training, which coincided with larger muscle growth.¹⁷ Our knowledge on epi-memory is, however, still rudimentary and identifying genes that constitute the epi-memory might provide crucial insights for the future development of therapies that promote muscle regain after inactivity.

Here, we use a training, immobilization and retraining resistance exercise protocol in young men to identify muscle memory-related genes at muscle methylome and transcriptome level. In addition, we test the rejuvenating effect of RT on the older skeletal muscle methylome and whether these beneficial methylome adaptations are maintained during detraining and promote an enhanced transcriptional response during retraining.

Methods

Subjects and experimental design

Nine older men, divided in an exercise group (n = 6, age: 66 ± 5 years, BMI: 26 ± 3 kg/m²; mean ± SD) and control group (n = 3, age: 70 ± 4 years, BMI: 26 ± 2 kg/m²), and five younger men (age: 22 ± 2 years, BMI: 24 ± 3 kg/m²) gave written consent for this study after being fully informed. All men were non-smoking, untrained, healthy individuals (exclusion factors described in detail elsewhere²⁹). The experimental design of the study is displayed in Figure 1. One younger subject withdrew after the first training period for reasons unrelated to this investigation. Baseline and post-training samples from this subject are included in the analysis. The study protocol was approved by the Medical Ethics Committee of the Catholic University of Leuven (S59380) and is in accordance with the Declaration of Helsinki.

[IMAGE OMITTED. SEE PDF]

Resistance training programme

A 12 week supervised RT programme was conducted as previously described.²⁹ Briefly, three times a week, on non-consecutive days, subjects performed eight exercises targeting all major muscle groups of the upper and lower body with load intensity gradually increasing every 4 weeks (65% to 80% of one repetition maximum). The training load was increased for each exercise when the final set comprised more repetitions than the prescribed number. The vastus lateralis muscle was specifically trained with a Signature Series Plate Loaded Linear Leg Press and Insignia Series Leg Extension machine (Life Fitness, Barendrecht, the Netherlands). The same RT programme was used for training and retraining phases with starting loads adapted to the physical state at that time.

Training interruption

Young subjects

The right leg of the young subjects was immobilized with the knee in nearly full extension for 2 weeks using a full leg cast (foot included). Subjects were instructed to walk with crutches during the immobilization phase. To prevent deep vein thrombosis, subjects were screened for risk factors and anticoagulant enoxaparin (40 mg per 0.4 mL per day, Clexane®, Sanofi Belgium, Diegem, Belgium) was administered daily via subcutaneous injection into the abdominal skinfold.

Older subjects in the exercise group

The training programme was interrupted by a detraining phase. Subjects were instructed not to engage in any structured strength or endurance exercises for 12 weeks and to resume normal physical activities from before the start of the study. To verify this, subjects weekly filled in a physical activity diary.

Skeletal muscle biopsy

Muscle biopsies were collected at baseline (left leg), ~72 h after the final RT session of the first training period (left leg), at the end of week 12 from the detraining period (older subjects, right leg) or immediately after cast removal (younger subjects, right leg) and ~72 h after the final RT session of the retraining period (right leg). The older control subjects provided samples at the post-training (right leg) and post-detraining (left leg) time point of the exercise group. Samples were taken under local anaesthesia from the mid-portion of the m. vastus lateralis with a 5 mm Bergström needle,³⁰ immediately cleaned of fat and connective tissue and frozen in liquid nitrogen before being stored at −80°C for further analysis.²⁹ Subjects were asked not to participate in moderate to vigorous physical activity and not to consume alcohol in the 24 h before biopsy collection. Food intake during the 72 h before biopsy collection was comparable between each time point within each subject.

DNA extraction and Illumina methylation assay

Genomic DNA was extracted from 15 to 20 mg muscle tissue with the proteinase K method, which uses proteinase K and lysis buffer. DNA was quantitated with the Qubit 2.0 Fluorometer (Invitrogen). DNA samples were transferred to the Genomics Core facility (UZ/KULeuven, Belgium) to measure DNA methylation level at over 850 000 CpGs using the Illumina Infinium MethylationEPIC BeadChip.

Pre-processing DNA methylation data

IDAT-files, containing raw intensity data from the EPIC BeadChip, were pre-processed in R (version 3.6.1).³¹ Firstly, the R/Bioconductor package minfi v1.30.0³¹ was used to read the IDAT files and perform quality control. None of the samples contained more than 10% of sites with detection P-value > 0.05,³² and all samples were considered good quality samples based upon clustering the log median intensities of raw methylated and unmethylated values. Background correction was performed with Noob dye-bias normalization,³³ implemented in minfi. Next, β-values, defined as the ratio of the methylated probe intensity and the overall intensity, with β = 0 indicating sites are completely unmethylated and β = 1 expressing fully methylated sites,³⁴ were extracted from the normalized data. These β-values were subsequently normalized with BMIQ,³⁵ implemented in R/Bioconductor package wateRmelon v1.28.0,³⁶ to correct for the different design of type I and type II probes in the EPIC BeadChip, as suggested by Liu et al.³⁷ Finally, we filtered out 357 probes that had failed in 20% of the samples based on detection P-value > 0.05,³² 20 655 probes that had failed to hybridize based on a bead count <3 in 5% of the samples,³² 97 253 probes that are cross-reactive or overlap with genetic variants based on the h19 genome³⁸ and 206 probes with unknown genomic location. Ultimately, 747 741 probes were considered for downstream analyses, which was performed on M-values, the log2 ratio of the intensities of methylated probe versus unmethylated probe.³⁹ Principal component analysis (PCA) plots were able to cluster young and older samples based on age, sample ID and timing in the intervention (Supporting Information, Figures S1 and S2). To correct for unwanted sources of variation, such as batch effects and cell type heterogeneity within muscle tissue, we used the package sva v.3.32.1⁴⁰ to perform surrogate variable analysis (SVA) on the M-values of each subject group, while protecting the information coming from the intervention (variable of interest, included in the full model) and the paired subject IDs (adjustment variable, included in the null and full model). The number of surrogate variables per group was estimated with the default ‘be’ approach.⁴¹ No SVA was performed when comparing young to older muscle methylome, as the function could not handle a design where individuals were nested within groups.

Differential DNA methylation analysis

Differential DNA methylation analysis was performed in Partek Genomics Suite V.7.0 (Partek Inc., St. Louis, MO, USA). To identify age-related dmCpGs, we applied a mixed model with methylation M-values as response variable, ‘time’ (baseline, post-training) and ‘group’ (younger subjects, older subjects of the exercise group) as fixed factors and ‘subject ID’ as random factor. We specifically interpreted the planned contrasts ‘younger versus older subjects at baseline’ and ‘younger subjects at baseline versus older subjects at post-training’. A CpG list was generated using Benjamini-Hochberg's⁴² false discovery rate (FDR) < 0.1 and difference in M-values > 0.4³⁹ as cut-off rules. A histogram of P-values for all 747 741 tests for the planned contrast ‘younger versus older subjects at baseline’ can be found in Data S1a. To determine within-group differences over time, three separate mixed models were fitted (one for each subject group). Methylation M-values were entered into the model as response variable, ‘time’ (all available time points) and surrogate variables as fixed factors and ‘subject ID’ as random factor. Contrasts were specified to study the different combinations of the various time points. As an example, the histogram of P-values for all 747 741 tests for the planned contrast ‘casting versus retraining in young subjects’ can be found in Data S4. Because only one CpG remained after FDR-control and this part of the study is considered explorative in nature,⁴³ statistical significance was determined at P < 0.01 with a M-value difference threshold of 0.4. Probes were annotated based on ‘HumanMethylation850’ reference, ‘MethylationEPIC_v-1-0_B4’ annotation file, and hg19 genome build. Overlapping genes and gene regions were determined based on Ensembl Transcripts data release 75. ‘Promoter regions’ are defined in this study as the combination of TSS1500 (1500–201 base pairs upstream from the transcription start site), TSS200 (200–0 base pairs upstream from the transcription start site), and the first exon (5′UTR + the first coding sequence). ‘Intragenic’ will be used as the overlapping term for all exons (except for the first coding sequence), introns, regions between the transcription start and end site of non-coding transcripts and 3′UTR regions. ‘Intergenic regions’ are regions between genes, not including the promoter region. The relationship to CpG islands (island, shore, shelf, and open sea) was determined with the UCSC genome browser. We furthermore performed pathway enrichment via a Fisher's Exact test in Partek Pathway (Partek Inc., St. Louis, MO, USA) based on the KEGG pathway database. We restricted the analysis to pathways of minimal 10 and maximal 500 genes, as small pathways are redundant with larger pathways and larger pathways overly general (e.g. pathways in cancer), inflating their statistical significance.⁴⁴ Finally, to find overlapping CpGs, genes and pathways following each intervention phase, we used the Venn diagram tool available via . This tool was also consulted to find overlap between dmCpGs of the present study and comparable studies. Pearson correlations tests and Fisher's exact tests were respectively used to identify correlations and associations between the overlapping datasets (with significance level set at P < 0.05).

RNA extraction and sequencing

Total RNA was homogenized from 15 to 20 mg of muscle tissue with silicon beads in 1 mL Trizol reagent (Invitrogen) using a FastPrep-24 5G (MP Biomedicals) and extracted with phase separation reagent 1-Bromo-3-chloropropane. RNA concentration and purity (A260/230) were determined with SimpliNano (Biochrom). RNA integrity was assessed with the Agilent 2100 Bioanalyzer System and was at least 7.7 in all samples. At the Genomics Core facility, sequencing libraries were prepared with the QuantSeq 3′ mRNA-Seq Library Prep Kit FWD (Lexogen) and sequenced on a HiSeq 4000 Sequencing System (Illumina) using the HiSeq3000/4000 SR Cluster and SBS Kit (Illumina).

Pre-processing RNA-sequencing data

Quality control of raw reads was performed with FastQC v0.11.5. Adapters were filtered with ea-utils v1.2.2.18. Splice-aware alignment was performed with the CRAN package Star v2.6.1b against the human reference genome hg19 based on annotations from Ensembl release 75. The number of allowed mismatches was two. Reads that mapped to more than one site to the reference genome were discarded. The minimal score of alignment quality to be included in count analysis was 10. Resulting SAM and BAM alignment files were handled with Samtools v0.1.19.24.⁴⁵ Quantification of reads per gene was performed with HT-Seq count v0.5.3p3.⁴⁶ Further pre-processing was performed in R (version 3.6.1).⁴⁷ Data quality assessment and count-based differential expression analysis of the 56 638 sequenced genes were carried out with R/Bioconductor package DESeq2 v1.26.⁴⁸ By inspecting heatmaps of the count matrix and PCA plots, we detected and excluded one outlier within the older exercise group (post-retraining time point) (Figure S3). The sva package (svaseq-function) was used in a similar way as described earlier to detect unwanted variation in the normalized counts. Low count genes were filtered with DESeq2's independent filtering technique, leaving 23 296, 24 661, and 19 211 genes left for expression analysis in the young exercise, older exercise, and older control groups, respectively. Results were generated on normalized counts, corrected for library size.

Differential RNA expression analysis

To test for any difference over the multiple time points, a likelihood ratio test was performed via DESeq2, including ‘time’, surrogate variables and ‘subject ID’ in the full model and surrogate variables and ‘subject ID’ in the reduced model. Pairwise comparisons between time points were analysed with Wald tests. Reported P-values were adjusted for multiple testing with the Benjamini–Hochberg procedure⁴² after assigning weights using Independent Hypothesis Weighting (IHW package v1.6.0⁴⁹). Differentially expressed genes (DEGs) are reported if FDR < 0.1 and the average read count ≥ 5. Overlap between methylation and expression data was determined with the Venn diagram tool (cfr. supra).

Comparison with the MetaMEx meta-analysis

We ran the top 100 DEGs of each analysis in MetaMEx,⁵⁰ a tool combining transcriptome data from more than 90 human studies that allows to meta-analyse changes in single genes across exercise and disuse studies. The training and immobilization data of the younger men was, respectively, cross-referenced with RT studies of 10 to 20 weeks (n = 58)^51–56 (Data S4h) and immobilization/bed rest studies of 5 to 21 days (n = 42)^10,57–60 (Data S4i) in healthy, sedentary/active, lean/overweight, younger men. MetaMEx does not include retraining studies. Therefore, we report genes that display the same expression response as RT studies in MetaMEx or the opposite expression response compared with disuse studies in MetaMEx (Data S4j). We cross-referenced the (re-)training data of the older men with RT studies of 12 to 24 weeks in healthy, sedentary and active, lean or overweight, elderly men (n = 66)^{51,53,55,61–63} (Data S2h and S2j). MetaMEx does not include detraining studies; instead, the detraining data of the older men are compared with a 5 day bed rest study in elderly men (n = 10)⁵⁷ (Data S2i).

Results

More hypomethylated CpGs and upregulated genes in previously trained young muscle

A first training period affected methylation levels of 3623 CpGs in young skeletal muscle (P < 0.01, M-value difference > 0.4) with an approximate 50/50-distribution of hypomethylated versus hypermethylated CpGs (1822 vs. 1801) (Figure 2A, Data S4a). The dmCpGs were not overrepresented or underrepresented in any gene region or structure (Figure 2B), thereby mirroring the distribution of the CpGs in the whole genome (Data S4d). Interestingly, within promoter-associated CpG islands, a higher number of CpGs were hypomethylated (1326) compared with hypermethylated (909), whereas in other areas, the number of hypomethylated versus hypermethylated CpGs was more equally distributed (Figure 2B). At transcriptional level, 11 genes were upregulated, and 16 genes were downregulated following training (FDR < 0.1, average read count ≥ 5) (Figure 2A, Data S4h). Ten of these 27 genes were differentially expressed in the same direction as comparable training studies (cfr. Supra: Methods) included in the metaMEx meta-analysis (Data S4h, 10 showing the same change, 0 showing the opposite change, 13 showing no change, and 4 not found in MetaMEx).

[IMAGE OMITTED. SEE PDF]

The 2 week casting period reversed methylation levels in 716 of the 3623 dmCpGs following training (P < 0.01, M-value difference > 0.4) (Figure 2A, Data S4b). Using the same threshold criteria, an additional 7627 CpGs unique CpGs were differentially methylated following the 2 weeks of immobilization compared with post-training with the majority being hypermethylated (5138). This immobilization-induced hypermethylation was also reflected within the different gene regions and structures (Figure 2B). Two hundred seventy-four genes were upregulated, and 560 genes were downregulated following immobilization compared with post-training (FDR < 0.1, average read count ≥ 5) (Figure 2A, Data S4i). Immobilization has a clear and distinct effect on the muscle transcriptome, as the top 100 most DEGs following immobilization in the present study corresponded largely to what is found in other disuse studies of the MetaMEx meta-analysis (Data S4i, 82 showing the same change, 1 showing the opposite change, 14 showing no change, and 3 not found in MetaMEx). Most genes within this top 100 were downregulated following immobilization and related to metabolism (49) or muscle structure and contraction (17) (Data S4i).

A second 12 week training period had a clear hypomethylating effect as 5150 of the 6074 dmCpGs displayed lower methylation levels following retraining compared with post-immobilization (P < 0.01, M-value difference > 0.4) (Figure 2A, Data S4c). Retraining reversed methylation levels in 1264 of the 8343 dmCpGs following immobilization (Figure 2A). Furthermore, merely 151 CpGs were differentially methylated in the same direction following retraining as following training (Figure 2A). This pattern of similar training–retraining methylation changes is higher than opposite patterns of methylation changes (65 CpGs), as visualised in the alternative patterns in Data S4a (Venn diagrams). In both CpG-rich and CpG-poor gene areas, the frequency of hypomethylated CpGs was enhanced compared with following training (Figure 2B). A similar enhanced retraining response was reported earlier in young men after 7 weeks detraining.¹⁷ The enhanced retraining response was also reflected at transcriptional level with 431 genes upregulated and 188 downregulated (FDR < 0.1, average read count ≥ 5) (Figure 2A, Data S4j). Retraining reversed expression levels in 463 of the 834 immobilization-induced transcriptional dysregulated genes. Restoration was observed in, among other, the mitochondrial respiratory chain (16 members of the NADH ubiquinone oxidoreductase complex, 8 members of the ubiquinol-cytochrome c reductase complex, 13 members of the cytochrome c oxidase complex, and 13 members of the ATP synthase complex), in 11 myosin light and heavy chain genes, in 8 genes of the mitochondrial ribosome complex, in 11 ribosomal protein-coding genes, and in 12 genes encoding for solute carrier membrane transport proteins (Data S4i-j). Finally, of the top 100 most DEGs following retraining, 32 genes displayed transcriptional changes in the same direction as comparable RT studies (3 in opposite direction, 62 no change) of the metaMEx database and 76 genes in the opposite direction compared with metaMEx disuse studies (1 in opposite direction, 20 no change) (Data S4j). Most genes within this top 100 were upregulated and related to metabolism (36), muscle structure and contraction (15) or muscle remodelling (10) (Data S4j). This is the first study describing the retraining-related recovery of human muscle after disuse-induced atrophy in relation to the transcriptome and methylome. In rats, denervation-induced muscle atrophy decreased methylation in MYOG, TRIM63, FBXO32, and CHRNA1 together with increased expression that was returned to control levels following 7 days of recovery.⁶⁴ In accordance, TRIM63, FBXO32, and CHRNA1 were significantly upregulated following immobilization (FDR ≤ 0.067) in the present study and retraining reversed the RNA expression of all three genes to untrained and trained levels. We could not link this finding to any methylation changes.

Finally, pathway analysis on the methylome data identified 66, 79, and 12 enriched pathways (FDR < 0.1) following training, casting, and retraining, respectively (Table 1, Data S4e-g), of which 7 enriched during the whole intervention period. The overlapping pathways contained nearly the same number of hypomethylated and hypermethylated CpGs following training, whereas the frequency of hypermethylated CpGs clearly increased and decreased following casting and retraining, respectively (Figure 2C). The pathway analysis thereby confirmed that retraining induced more hypomethylation in comparison with training.

Table 1 Top 3 enriched pathways per analysis

	ES	FDR	DMG	Genes in pathway	dmCpGs hypo	dmCpGs hyper
Young muscle following traininga
Longevity regulating	11.1	0.005	27	89	9	30
Axon guidance	9.7	0.009	43	182	24	23
Regulation of actin cytoskeleton	9.4	0.009	48	213	35	23
Young muscle following castinga
Focal adhesion	17.8	<0.001	89	199	35	80
Prostate cancer	13.0	<0.001	48	99	17	45
Rap1 signalling	12.1	<0.001	84	206	38	69
Young muscle following retraininga
MAPK signalling	12.2	0.002	91	296	104	9
Focal adhesion	8.5	0.031	61	199	61	12
Endocytosis	8.1	0.032	72	247	75	16
Young versus older muscle
Focal adhesion	37.4	<0.001	161	199	442	377
Axon guidance	31.3	<0.001	145	182	356	447
Oxytocin signalling	30.7	<0.001	125	153	372	314
Older muscle following traininga
Axon guidance	12.9	<0.001	54	182	43	24
Cholinergic synapse	11.0	0.003	36	112	27	20
Calcium signalling	9.7	0.005	51	187	37	31
Older muscle following detraininga
Focal adhesion	19.6	<0.001	85	199	40	69
MAPK signalling	10.7	0.004	102	296	41	88
Endometrial cancer	9.6	0.004	28	59	13	23
Older muscle following retraininga
Platelet activation	14.3	<0.001	42	124	36	14
Phospholipase D signalling	13.9	<0.001	47	147	40	16
Non-small cell lung cancer	12.2	<0.001	26	67	21	11

Ageing-related hypermethylation in promoter-associated CpG islands and inverse expression levels

Comparing young to older muscle samples at baseline, we identified 50 828 age-related dmCpGs (FDR < 0.1, M-value difference > 0.4), of which 28 443 were hypomethylated and 22 385 were hypermethylated in older muscle tissue (Data S1a). Two earlier genome-wide methylation studies^8,9 that compared young and older human muscle samples found a clear global trend towards hypermethylated aged muscle. We could only confirm this trend in promoter-associated CpG islands, where more than twice as many hypermethylated CpGs (2761) were found compared with hypomethylated ones (1202). In comparison, other regions contained 23 to 45% more hypomethylated CpGs (Figure 3A). Promoter-associated dmCpGs in CpG islands were underrepresented compared with CpGs in the whole genome (Figure 3B, Data S1b), which is a confirmation of the results of Zykovich et al.,⁸ but in contrast to the findings of Turner et al.,⁹ where most dmCpGs were located in CpG islands. Finally, we found a significant overlap of 2334 CpGs between the age-related dmCpGs of the present study and the age-related dmCpGs of Zykovich et al. (P = 2.2e-16) (Figure 3C, Data S1d). Interestingly, the age-related differences in β-values of the overlapping dmCpGs was nearly identical (r = 0.94, P < 0.001) (Figure 3D). These overlapping dmCpGs could, therefore, be a representation of an epigenetic clock⁵ common between individuals, whereas the dmCpGs uniquely found in each study represent the epigenetic drift⁵ between individuals due to stochastic factors and the specific environmental conditions of the population under investigation. To test this assumption, the 2334 overlapping dmCpGs were compared with the 200 CpGs that were recently selected to create a muscle-specific epigenetic clock.⁶⁵ We found a significant overlap of 12 CpGs (P = 2.6e-12) (Figure 3C, Data S1d).

[IMAGE OMITTED. SEE PDF]

Pathway analysis identified 151 enriched pathways (FDR < 0.1) (top 3 in Table 1), belonging to ‘organismal systems’ (52, e.g. axon guidance), ‘human diseases’ (44 of which 18 related to cancer), ‘environmental information processing’ (24, e.g. PI3K-Akt signalling pathway), ‘metabolism’ (16), ‘cellular processes’ (13, e.g. focal adhesion), and ‘genetic information processing’ (2) (Data S1c).

In addition, ageing affected expression levels of 427 genes (FDR < 0.1), of which 183 were downregulated and 244 upregulated in older muscle. Remarkably, 71% of these 427 DEGs contained at least one promoter-associated or intragenic dmCpG (FDR < 0.1, M-value difference > 0.4) (Data S1e). We selected all DEGs with dmCpGs methylated in the same direction (i.e. all sites hypomethylated or all sites hypermethylated within a specific gene region) (Figure 4A, DEGs with differential methylation in promoter regions reported in Data S1f). Most of the hypomethylated genes displayed increased expression, whereas expression of hypermethylated genes was generally downregulated. This was independent of the gene region and especially true for genes with more dmCpGs. However, many DEGs included both hypomethylated and hypermethylated CpGs (Data S1g), which confirms that the relationship between methylation and expression is more nuanced than a simple inverse relation.⁶⁶ Exploring the top 100 differentially methylated and expressed genes between young and old, we found indications of impaired processes important for muscle functioning and remodelling in older muscle (Table A1).

[IMAGE OMITTED. SEE PDF]

Training has rejuvenating effects on the older muscle methylome

Twelve weeks of training in the older men rejuvenated their muscle methylome, as 37 339 of 50 828 age-related dmCpGs at baseline were no longer significantly different between trained older muscle and untrained young muscle (FDR ≥ 0.1 or M-value difference ≤ 0.4) (Data S1h). Of these 37 346 CpGs, we selected the ones that changed significantly with training (606 CpGs with P < 0.01, M-value difference > 0.4). Based on these 606 CpGs, cluster analysis effectively clustered older post-training samples with younger baseline samples, instead of with older baseline samples (Figure 5). Pathway analysis linked four pathways to these CpGs (FDR < 0.1): ‘cortisol synthesis and secretion’ (ES = 9.6, FDR < 0.001), ‘cAMP signalling pathway’ (ES = 7.4, FDR = 0.067), ‘adrenergic signalling in cardiomyocytes’ (ES = 6.8, FDR = 0.067) and ‘aldosterone synthesis and secretion’ (ES = 6.75, FDR = 0.067). Most dmCpGs in these four pathways were hypermethylated in older compared with young muscle at baseline with a hyper/hypo ratio (i.e. number of hypermethylated CpGs/number of hypomethylated CpGs) between 2.7 and 11. Training had the opposite effect, directing the methylation level towards younger levels.

[IMAGE OMITTED. SEE PDF]

In total, 2657 CpGs were hypomethylated and 2163 CpGs were hypermethylated after training compared with baseline (P < 0.01, M-value difference > 0.4) (Figure 6A, Data S2a). In comparison, 12 weeks of normal daily living changed the methylation status of merely 255 CpGs in the older control group (P < 0.01, M-value difference > 0.4) (Figure 6A, Data S3a). With respect to CpG islands and shores in these comparisons, training had a hypomethylating effect (Figure 6B). Comparable with the results in younger men, the dmCpGs were not overrepresented or underrepresented in any region (Data S2d). Of the 20 DEGs following training, only 1 (TM4SF1) displayed similar transcriptional changes following training as comparable training studies in older adults in the metaMEx meta-analysis (Data S2h, 1 showing the same change, 0 showing the opposite change, 17 showing no change, and 2 not found in MetaMEx). Against expectations, there were more genes downregulated than upregulated following a first training period (Figure 6A). Four of these genes (PPDPFL, BBC3, PER1, and MCL1) were upregulated in older versus young muscle (FDR ≤ 0.068) and subsequently rejuvenated (i.e. downregulated) following training (FDR ≤ 0.038) (Data S1e and S2h). Twelve weeks of detraining, subsequently reverted the training effect and increased expression levels again in PPDPFL, PER1, and MCL1 (FDR ≤ 0.040) (Data S2i). In addition, detraining reversed training adaptations in RNA expression of seven more genes (Data S2i), as well as reverting methylation levels of 628 of the 4820 dmCpGs following training (Figure 6A, Data S2b). Globally, detraining mainly induced hypermethylation compared with post-training and more genes were downregulated (54) compared with upregulated (40) (Figure 6A,B). Eleven of these 94 DEGs were transcriptionally altered in the same direction as a 5 day bed rest study in elderly men,⁵⁷ as displayed by the MetaMEx tool (Data S2i). On the other hand, RNA expression of 10 genes changed in the opposite direction as this disuse study. In addition, 43 upregulated and 28 downregulated genes in our study did not show significant expression changes by bed-rest in MetaMEx (Data S2i). These findings indicate that detraining and bed-rest might have a different effect on certain genes.

[IMAGE OMITTED. SEE PDF]

Similar to the results in young men, retraining hypomethylated more CpGs compared with training (3077 vs. 2657) (Figure 6A), which was observed across all gene regions and structures (Figure 6B). Pathway analysis identified 48, 66 and 107 enriched pathways (FDR < 0.1) following training, detraining, and retraining, respectively (Table 1, Data S2e-g). The majority of the pathways influenced by training were also differentially methylated following detraining and/or retraining, whereas 34 pathways were uniquely affected by retraining (Figure 6C), suggesting that the methylome was more responsive to retraining. We found 18 pathways in common between each intervention period (Figure 6C). In the majority of the overlapping pathways, the number of hypomethylated CpGs further increased following retraining compared with training (Figure 6D). This could be indicative for an epi-memory within these pathways. Furthermore, these 18 pathways were also enriched when comparing the young and older muscle methylome at baseline (Figure 6E). Only 12 genes were transcriptionally altered following retraining (Figure 6A, Data S2j). However, half of them were also picked up by comparable training studies of the MetaMEx meta-analysis and are involved in inflammation (IL17D), muscle growth inhibition (PLN), muscle contractions (TPM3), mechanosensing (ANKRD2), muscle structure (MYL5), and energy metabolism (CARNS1) (FDR ≤ 0.091) (Data S2j). Four other upregulated genes, not picked up by the meta-analysis, are involved in myogenic differentiation (SOX11), energy metabolism (AK1, PGK1), and muscle development and functioning (PDLIM7) (FDR ≤ 0.097) (Data S2j).

Genes indicative for an epi-memory in young and older skeletal muscle

Of the differentially expressed genes across all intervention phases in young and older muscle (FDR < 0.1, average read count ≥ 5), 11 to 33% contained at least one dmCpG in the promoter or intragenic region (P < 0.01, M-value difference > 0.4) (Figure 7). Detailed information on these genes, including plots, is available in Data S2l-m and S4l-m. The interaction between the transcriptome and methylome was stronger following retraining compared with training. This might indicate that other epigenetic mechanisms were responsible for the alterations in chromatin accessibility during the first training period, whereas methylation was used during retraining to stabilize these changes.

[IMAGE OMITTED. SEE PDF]

We were specifically interested in genes constituting the epi-memory. Epi-memory genes can be defined as genes displaying an enhanced transcriptional response to retraining compared with training that could be linked to retained methylation modifications from the first training period (genes with retained methylation levels over time displayed in Data S2k and S4k). In young muscle, one CpG (cg15758225) within the open sea area of intron 3 of AMOTL1 became more methylated following training (P = 0.006), which persisted during immobilization (P < 0.001) and retraining (P < 0.001) (Figure 7). As response to the second training period and potentially the hypermethylated state of cg15758225 pre-retraining, AMOTL1 expression levels decreased (FDR = 0.012). This might indicate an epi-memory within AMOTL1 and we found confirmation in the muscle biopsy data from Seaborne et al.,¹⁷ as the hypermethylated state of two CpGs (cg13554187 and cg06447552) within AMOTL1 following loading, was retained following unloading and for cg13554187 also following reloading. The investigators found several retraining-upregulated genes themselves indicative for an epi-memory in young muscle (AXIN1, TRAF1, GRIK2, and CAMK4 in Seaborne et al.¹⁷ and FLNB, MYH9, SRGAP1, SRGN, and ZMIZ1 in Turner et al.²⁸). Despite finding one to three hypomethylated CpGs following retraining in FLNB, MYH9, ZMIZ1, GRIK2, and CAMK4, we did not find the expected increased expression levels in these genes. We did find several genes with a trend towards enhanced expression levels following retraining and a correlated methylation pattern in young muscle (e.g. NEXN, SERF2, and PLCL2) (Data S4m). Seaborne et al.¹⁷ also reported on a list of genes (UBR5, RPL35a, C12orf50, BICC1, ZFP2, HEG1, PLA2G16, SETD3, and ODF2) presenting a different pattern of epigenetic memory with hypomethylation responses by loading, no retention of methylation during detraining, followed by a stronger hypomethylation and associated increased gene expression during reloading. Specifically, the role of UBR5 in muscle atrophy and hypertrophy was further confirmed by in vivo and in vitro studies in mice and rats by the Sharples group.^67,68 Although not reaching the criteria (P < 0,01, M-value difference >0.4) a pattern indicating hypomethylation by training (P = 0.02), hypermethylation by casting (P = 0.01) and again hypomethylation by retraining (P = 0.017) in cg21230538 in the young muscle was supportive of the role of epigenetic changes in UBR5 due to chronic training and unloading. This pattern was however not found in older muscle or in other CpGs in this gene region.

In older men, we identified VCL as a gene with retained epigenetic modifications (Figure 7). Promoter-associated and intron-associated cg15758225, located in open sea area, was less methylated after training, detraining and retraining compared with baseline (all P < 0.001). This methylation change might have contributed to the trend towards an enhanced expression level following retraining (FDR = 0.136, P = 0.002). In addition, we found several other genes with a trend towards enhanced expression levels following retraining and a correlated methylation pattern in older muscle (e.g. SOX11, PDLIM7, and PGK1) (Data S2m).

Discussion

In this study, we identified CpG methylation as a potential upstream mechanism that can explain transcriptional changes in skeletal muscle tissue associated with ageing and training. Interestingly, recurrent RT was accompanied with increasing numbers of hypomethylated CpGs and upregulated genes in young and older skeletal muscle and rejuvenated methylation levels of several CpGs altered by ageing. In addition, we found indications of an epi-memory within AMOTL1 in young muscle and VCL in older muscle.

There have been many studies trying to reveal the age-associated human muscle transcriptome signature,^69–72 but literature lacks the integration of the mechanisms behind these transcriptional changes. We discovered that of the 427 DEGs between young and old, 71% contained at least one dmCpG. This confirms that DNA methylation might be an important regulator of gene expression in muscle tissue. Furthermore, we found that the more dmCpGs methylated in the same direction, the clearer the inverse relationship between DNA methylation and RNA expression and this was regardless of the location of the dmCpG within the gene. It has been reported that a threshold of three CpGs is required to effectively inhibit transcription, with more CpGs and closer proximity to the promoter region associated with a higher percentage of gene repression.⁷³ This explains why some but not all methylome adaptations over time mediated gene expression alterations in the present study, something that has been observed before in blood tissue.^74,75 In addition, the relationship between methylation and expression is more complex than a simple inverse correlation, as evidenced by the combination of both hypomethylated and hypermethylated CpGs within the different gene regions and structures of some genes. The complex transcriptome–methylome relationship has been described before,⁶⁶ but lacked tissue specificity. Therefore, we warrant further investigation into the interplay between methylation and other epigenetic mechanisms in muscle tissue (e.g. histone deacetylation, which works closely together with DNA methylation to silence genes⁷³) and their role in the different gene regions and structures. Interestingly, older muscle tissue displayed signs of impaired myogenesis, mitochondrial bioenergetics, autophagy and glucose uptake, as well as indications of muscle atrophy and inflammageing at methylome and transcriptome level. These signs are closely related to cellular senescence,⁷⁶ one of the hallmarks of ageing,⁷⁷ and might contribute to age-related anabolic resistance and impaired skeletal muscle growth in response to an anabolic stimulus.^78,79 In accordance, we reported earlier that a first training period was unable to induce significant muscle growth in these same older men.²⁹ In the present study, we were able to link this impaired muscle growth to ageing-induced molecular dysregulation, both at transcriptional and epigenetic level.

Remarkably, 73% of the age-related dmCpGs in untrained older muscle were no longer differentially methylated following 12 weeks of RT compared with young untrained muscle. The ability of resistance exercises to reverse age-related impairments has been extensively described at the muscle phenotypic level,^80–82 proteome level,⁵³ transcriptome level,^55,61 and other hallmarks of ageing.⁸³ For the first time, we confirm this finding at methylome level. We are aware of only one study⁵³ that investigated the effect of RT on the older muscle methylome. However, the investigators were unable to find methylation changes after 12 weeks of RT, probably due to the low sample size, the strict FDR adjustment and the limited number of CpGs tested. We hypothesized that training would reverse the ageing-induced hypermethylation, predominantly found in older muscle tissue in literature,^8,9 based on the hypomethylating effect of aerobic exercise,^10–13 (lifelong) physical activity,^16,84 and RT in young to middle-aged adults.^13,85 However, both in young and in older muscle tissues, we observed that a first training period induced both hypomethylation and hypermethylation. That said, mainly training-induced hypomethylation was found in CpG islands and shores, possibly counteracting the ageing-induced hypermethylation in promoter-associated CpG islands of the older muscle samples. At transcriptional level, training altered 20 genes in older men. Interestingly, PPDPFL, BBC3, PER1, and MCL1 were upregulated in older versus young muscle, downregulated following training, and upregulated again during detraining in older men. The role of PPDPFL in muscular training adaptations is currently unknown. The other three genes, on the other hand, are known to be involved in apoptosis (BBC3, BCL2 Binding Component 3, also known as PUMA),⁸⁶ (MCL1, MCL1 Apoptosis Regulator, BCL2 Family Member),⁸⁷ and circadian clock-regulated physiological processes (e.g. muscle hypertrophy) (PER1, Period Circadian Regulator 1).⁸⁸

Both young and older muscles responded better to a second training period. This was reflected by the higher number of hypomethylated CpGs and upregulated genes following retraining compared to training. In addition, retraining induced more hypomethylation than training did in the 7 overlapping pathways between training, immobilization and retraining in young men and in the 18 overlapping pathways between training, detraining and retraining in older men. Interestingly, these 18 pathways were also enriched in the age-related analysis, which supports the fact that chronic RT may act on several ageing-related pathways. Accordingly, the indications of inflammageing at baseline in older muscle tissue were only improved by the second training period, and not by the first. We observed nine hypomethylated dmCpGs in proximity to the promoter region of the pro-inflammatory cytokine Interleukin 17D (IL17D) when comparing older to young muscle, which was associated with elevated IL17D RNA expression in older muscle. IL17D, which also activates the acute inflammatory response mediated by NF-κB in the first hours after intense RT,⁸⁹ was successfully downregulated following retraining compared to post-detraining, whereas the decline in expression was not significant following the first training period. In addition, we found one hypermethylated dmCpG within IL17D following retraining. In agreement, NFKB Inhibitor Alpha (NFKBIA), inhibitor of NF-κB signalling, was downregulated following 12 weeks training. NFKBIA downregulation could be indicative for a permanent state of inflammation in older exercise-stimulated muscles. Surprisingly, NFKBIA expression levels returned to baseline levels during detraining and remained stable during retraining. These results indicate that previously trained older muscle may be less susceptible to exercise-induced chronic inflammation and recurrent RT may in fact improve age-related low-grade inflammation, as has been reported before.⁹⁰

Most interestingly, we identified two genes, indicative for an epi-memory within young and older muscle. In young men, the epi-memory pattern was reflected within the Angiomotin Like 1 (AMOTL1) gene. Upon activation of satellite cells, AMOTL1 activates YAP1 which triggers pro-proliferating genes and inhibits differentiation of myotubes.⁹¹ The decreased expression following retraining might signal myotube differentiation. Importantly, this result was confirmed when investigating the data of Seaborne et al.¹⁷ In older men, an increased expression level in Vinculin (VCL) following retraining was linked to the retained hypomethylated state of one CpG during training, detraining and retraining. In striated muscle, vinculin is a structural component of costameres,⁹² which connects sarcomeres to the cell membrane to stabilize myofibres during contraction and relaxation. The gene, furthermore, potentiates mechanosensing^93,94 and is upregulated in skeletal muscle tissue following chronic stimulation and disuse.⁹⁵ Most interestingly, VCL is part of the focal adhesion pathway, which was enriched during both training, detraining and retraining in older men, as well as during training, immobilization and retraining in younger men and the age-related comparison between young and old. Focal adhesion translates the (exercise-induced) cytoskeletal stress signals into cell growth by activating a cascade of signalling pathways.⁹⁶ The importance of the focal adhesion pathway in our results might indicate that the rapid regain of training adaptations is related to improved mechanosensing signalling during retraining. To support our results, we investigated other genes related to the mechanosensing pathway and found a trend towards enhanced gene expression and several hypomethylated CpGs following retraining in mechanosensitive transcriptional cofactor-encoding WWTR1, stretch-activated ion channel-encoding PIEZO2 and extracellular-matrix associated COL1A1 (Figure A1).

Finally, we would like to address several limitations and future perspectives. First of all, sample sizes were small. This limited the statistical power and especially the intervention-related analyses might have been prone to type I error. Findings are consequently exploratory and preliminary in nature and should be validated in future studies. On the other hand, findings that were found significant at FDR level of 0.1 despite the limited sample size, should be considered highly interesting. To increase power, a combined analysis of both young and old muscle samples could have been performed for the first training period, as the time*group interaction term was non-significant in the mixed model (data not shown). However, we preferred to test specific hypotheses and contrasts towards the ability of strength training to decrease the large set of age-related dmCpGs between young and old at baseline. Secondly, due to the different unloading modalities (detraining versus casting), comparison of the unloading and retraining phases between young and old was limited. We chose not to apply both situations to each group, because of the complex study design and the risks associated with immobilization in elderly. Nevertheless, by using both unloading modalities, we were able to give additional insights in the differences between a more invasive type of disuse versus long-term training cessation. Thirdly, CpGs may overlap with different regions of different gene transcripts belonging to one gene and it is not clear how each gene transcript is influenced by differential methylation, as RNA expression was only analysed at gene level. Fourthly, although we did perform SVA, we did not directly correct for cell type heterogeneity and fibre type differences between muscle samples. This might have confounded the analysis.⁹⁷ Finally, while we do integrate methylome and transcriptome data, we currently lack the translation to the proteome level, which should be the focus in future research. Also needed are similar analyses at single-cell or -nucleus level and the integration of other epigenetic mechanisms to unravel the precise role of epigenetics in the muscle memory phenomenon.

To conclude, these results indicate an enhanced transcriptional training response in previously trained young and older muscle, partially explained by retained and novel methylation adaptations. This supports that we might benefit from earlier training periods and that RT is an effective strategy to rejuvenate our methylome.

Acknowledgements

This study was supported by a research project grant (FWO G.0898.15) from Research Foundations Flanders. The authors thank Dr. Lingxiao He for sharing his DNA methylation processing script, Mr. Alvaro Cortes Calabuig for initial help with the processing of RNA expression data, Ms. Vanessa Brys for her help with the bioanalyzer. Image acknowledgement: image of the muscle and DNA helix in Figure 1 made by Freepik from .

Ethical approval

The study protocol was approved by the Medical Ethics Committee of the Catholic University of Leuven (S59380). Our work conformed to the standards set by the latest revision of the Declaration of Helsinki (except for registration in a database) and to the ethical guidelines for authorship and publishing set out by the Journal of Cachexia, Sarcopenia and Muscle.⁹⁸

Conflict of interest

The authors report no conflict of interest.

Appendix

APPENDIX - A

Table A1 Important muscle-functioning genes differentially expressed and methylated in older muscle

Gene	Function		Log2fc	SE	FDR	dmCpG		Old	Young	FDR
β	SE	β	SE
Impaired myogenesis
MYF6a	Homeostatic role by repressing hypertrophy-related genes	Up	0.96	0.17	<0.001	cg02939404	Hypo	0.34	0.02	0.53	0.03	0.065
				cg08352786	Hypo	0.33	0.01	0.45	0.02	0.045
				cg15166296	Hypo	0.24	0.01	0.35	0.02	0.050
					cg15757530	Hypo	0.42	0.03	0.65	0.03	0.040
					cg17159916	Hypo	0.39	0.01	0.58	0.04	0.067
						cg20171297	Hypo	0.19	0.01	0.33	0.01	0.042
						cg25178519	Hypo	0.45	0.02	0.67	0.02	0.035
						cg26711820	Hypo	0.31	0.01	0.40	0.01	0.063
JAK2	Inhibitor of satellite cell function via JAK/STAT	Up	1.11	0.28	0.009	cg02104682	Hypo	0.35	0.01	0.46	0.02	0.046
IGF2a	Autocrine and paracrine growth stimulator	Down	−1.28	0.34	0.011	cg14890224	Hyper	0.75	0.01	0.69	0.01	0.046
				cg17274401	Hyper	0.07	0.01	0.03	0.00	0.062
				cg20728696	Hyper	0.13	0.02	0.06	0.00	0.090
					cg21728792	Hyper	0.93	0.01	0.86	0.01	0.051
					cg01351425	Hypo	0.32	0.01	0.42	0.02	0.064
						cg01921126	Hypo	0.32	0.02	0.48	0.02	0.053
						cg08686462	Hypo	0.22	0.01	0.33	0.01	0.035
						cg10650127	Hypo	0.36	0.01	0.55	0.04	0.076
						cg11005826	Hypo	0.32	0.01	0.40	0.01	0.049
						cg26516510	Hypo	0.17	0.01	0.30	0.03	0.044
						cg27263998	Hypo	0.20	0.01	0.38	0.02	0.035
						cg27331871	Hypo	0.06	0.00	0.10	0.01	0.100
IGFBP2	IGFs modulator and myoblast differentiation stimulator via Akt phosphorylation	Down	−0.90	0.20	0.002	cg06571246	Hyper	0.78	0.01	0.69	0.01	0.060
Impaired mitochondrial bioenergetics
PERM1a	Regulator of mitochondrial biogenesis and oxidative function	Down	−0.90	0.23	0.007	cg00380574	Hypo	0.29	0.01	0.39	0.01	0.035
				cg08843537	Hypo	0.53	0.01	0.63	0.02	0.043
				cg10201533	Hypo	0.23	0.01	0.31	0.01	0.050
					cg10381813	Hypo	0.24	0.01	0.34	0.01	0.035
					cg11143486	Hypo	0.38	0.01	0.46	0.01	0.081
						cg11782208	Hypo	0.39	0.01	0.49	0.01	0.039
						cg15526941	Hypo	0.20	0.01	0.25	0.01	0.078
						cg20869814	Hypo	0.31	0.01	0.39	0.01	0.072
						cg22583825	Hypo	0.18	0.01	0.24	0.00	0.046
						cg22599150	Hypo	0.36	0.00	0.43	0.01	0.044
						cg22699004	Hypo	0.29	0.02	0.42	0.01	0.073
NOS1	Regulator of mitochondrial bioenergetics, excitation-contraction coupling, muscle force generation, auto-regulation of blood flow, calcium homeostasis and myogenesis	Down	−0.61	0.16	0.008	cg01207748	Hyper	0.79	0.01	0.71	0.01	0.066
				cg01686420	Hyper	0.92	0.01	0.84	0.02	0.044
				cg02500231	Hyper	0.69	0.02	0.58	0.02	0.068
				cg02595280	Hyper	0.54	0.02	0.38	0.02	0.036
				cg03019507	Hyper	0.65	0.01	0.57	0.01	0.091
				cg05923508	Hyper	0.60	0.03	0.43	0.04	0.080
				cg07846732	Hyper	0.45	0.02	0.34	0.02	0.071
				cg10914558	Hyper	0.61	0.01	0.50	0.01	0.036
					cg17119907	Hyper	0.60	0.02	0.51	0.02	0.094
					cg18243853	Hyper	0.43	0.01	0.33	0.01	0.051
						cg18947446	Hyper	0.50	0.02	0.41	0.01	0.052
						cg19945354	Hyper	0.67	0.01	0.56	0.03	0.080
						cg20358453	Hyper	0.72	0.01	0.61	0.02	0.045
						cg20840426	Hyper	0.66	0.02	0.55	0.02	0.068
						cg07748838	Hypo	0.40	0.02	0.50	0.02	0.060
						cg13023936	Hypo	0.29	0.02	0.42	0.02	0.049
SAMM50a	Role in structural integrity of mitochondria outer membrane via mitochondrial intermembrane space bridging	Down	−0.56	0.15	0.012	cg03373887	Hypo	0.28	0.01	0.37	0.01	0.035
				cg07494691	Hypo	0.43	0.02	0.51	0.01	0.087
				cg15212821	Hypo	0.40	0.02	0.53	0.03	0.041
				cg17161504	Hypo	0.35	0.01	0.49	0.02	0.037
				cg19172487	Hypo	0.29	0.01	0.40	0.01	0.037
COQ10Aa	Intra-mitochondrial coenzyme in the respiratory chain	Down	−0.70	0.19	0.016	cg03319773	Hypo	0.69	0.01	0.78	0.02	0.095
				cg11631125	Hypo	0.34	0.02	0.55	0.02	0.035
				cg16272957	Hypo	0.32	0.01	0.40	0.01	0.039
				cg19969040	Hypo	0.45	0.01	0.56	0.01	0.035
					cg21147422	Hypo	0.51	0.01	0.65	0.01	0.035
					cg23196303	Hypo	0.47	0.01	0.57	0.01	0.035
Impaired glucose uptake
ERBB2	Role in homeostasis of glucose via NRG1/ERBB2	Up	0.99	0.22	0.001	cg00459816	Hypo	0.29	0.01	0.37	0.01	0.045
				cg04048710	Hypo	0.42	0.01	0.59	0.02	0.035
				cg13236904	Hypo	0.34	0.02	0.44	0.02	0.051
				cg15227682	Hypo	0.54	0.01	0.70	0.02	0.035
				cg20498199	Hypo	0.89	0.01	0.94	0.01	0.057
Impaired autophagy
MAP 1 LC3Aa	Autophagy marker of autophagosome formation	Down	−0.86	0.20	0.003	cg07486199	Hyper	0.67	0.01	0.56	0.02	0.047
				cg15112783	Hyper	0.75	0.01	0.63	0.02	0.039
Signs of atrophy and impaired structural integrity
AMPD3a	Regulator of protein degradation via ATP energetics	Up	1.76	0.44	0.012	cg01116568	Hypo	0.31	0.01	0.44	0.02	0.035
				cg02082273	Hypo	0.37	0.01	0.49	0.02	0.044
				cg02973178	Hypo	0.29	0.01	0.40	0.02	0.044
					cg07268332	Hypo	0.30	0.01	0.42	0.02	0.054
					cg07329251	Hypo	0.26	0.01	0.34	0.01	0.043
						cg12069276	Hypo	0.34	0.02	0.46	0.02	0.058
						cg16118534	Hypo	0.46	0.01	0.57	0.01	0.038
						cg18010131	Hypo	0.32	0.01	0.44	0.01	0.036
						cg18678492	Hypo	0.34	0.01	0.42	0.01	0.057
						cg19132462	Hypo	0.27	0.01	0.38	0.01	0.040
						cg23341488	Hypo	0.60	0.01	0.51	0.01	0.039
						cg26489413	Hypo	0.33	0.02	0.44	0.02	0.043
						cg24933917	Hyper	0.78	0.02	0.70	0.01	0.095
COL4A3	Structural component of basement membranes	Down	−1.38	0.37	0.016	cg05715709	Hypo	0.54	0.02	0.67	0.01	0.043
				cg24260952	Hypo	0.23	0.01	0.33	0.02	0.039
				cg27243623	Hypo	0.34	0.01	0.45	0.02	0.050
Inflammageing
IL17Da	Stimulator of inflammatory cytokine production	Up	1.63	0.33	<0.001	cg01552711	Hypo	0.25	0.01	0.35	0.02	0.044
				cg02439964	Hypo	0.31	0.02	0.45	0.03	0.044
				cg08371706	Hypo	0.59	0.02	0.68	0.02	0.055
					cg11809038	Hypo	0.37	0.02	0.47	0.02	0.071
					cg12475590	Hypo	0.36	0.02	0.45	0.02	0.089
					cg14133014	Hypo	0.52	0.01	0.60	0.01	0.040
						cg23098661	Hypo	0.31	0.01	0.39	0.01	0.047
						cg24495327	Hypo	0.17	0.01	0.32	0.03	0.035
						cg25710235	Hypo	0.31	0.03	0.44	0.01	0.042
HMGB1a	Initiator of inflammatory response via innate immune system activation	Up	0.85	0.20	0.003	cg01504447	Hypo	0.24	0.02	0.36	0.03	0.057
				cg05818394	Hypo	0.19	0.01	0.27	0.01	0.074

[IMAGE OMITTED. SEE PDF]

References

Cartee GD, Hepple RT, Bamman MM, Zierath JR. Exercise promotes healthy aging of skeletal muscle. Cell Metabolism 2016;23:1034–1047.

Garatachea N, Lucía A. Genes and the ageing muscle: a review on genetic association studies. Age (Dordrecht, Netherlands) 2013;35:207–233.

Greenberg MVC, Bourc'his D. The diverse roles of DNA methylation in mammalian development and disease. Nature Reviews Molecular Cell Biology 2019;20:590–607.

Weber M, Hellmann I, Stadler MB, Ramos L, Pääbo S, Rebhan M, et al. Distribution, silencing potential and evolutionary impact of promoter DNA methylation in the human genome. Nature Genetics 2007;39:457–466.

Jones MJ, Goodman SJ, Kobor MS. DNA methylation and healthy human aging. Aging Cell 2015;14:924–932.

Pal S, Tyler JK. Epigenetics and aging. Science Advances 2016;2: [eLocator: e1600584].

Day K, Waite LL, Thalacker‐mercer A, West A, Bamman MM, Brooks JD, et al. Differential DNA methylation with age displays both common and dynamic features across human tissues that are influenced by CpG landscape. Genome Biology 2013;14:R102.

Zykovich A, Hubbard A, Flynn JM, Tarnopolsky M, Fraga MF, Kerksick C, et al. Genome‐wide DNA methylation changes with age in disease‐free human skeletal muscle. Aging cell 2014;13:360–366.

Turner DC, Gorski PP, Maasar MF, Seaborne P, Baumert P, Brown AD, et al. DNA methylation across the genome in aged human skeletal muscle tissue and muscle stem cells: the role of HOX genes and physical activity. Scientific Reports 2020;10:15360.

Alibegovic AC, Sonne MP, Højbjerre L, Jacobsen S, Nilsson E, Færch K, et al. Insulin resistance induced by physical inactivity is associated with multiple transcriptional changes in skeletal muscle in young men. American journal of physiology Endocrinology and metabolism 2010;299:E752–E763.

Nitert MD, Dayeh T, Volkov P, Elgzyri T, Hall E, Nilsson E, et al. Impact of an exercise intervention on DNA methylation in skeletal muscle from first‐degree relatives of patients with type 2 diabetes. Diabetes 2012;61:3322–3332.

Lindholm ME, Marabita F, Gomez‐Cabrero D, Rundqvist H, Ekström TJ, Tegnér J, et al. An integrative analysis reveals coordinated reprogramming of the epigenome and the transcriptome in human skeletal muscle after training. Epigenetics 2014;9:37–41.

Rowlands DS, Page RA, Sukala WR, Giri M, Ghimbovschi SD, Hayat I, et al. Multi‐omic integrated networks connect DNA methylation and miRNA with skeletal muscle plasticity to chronic exercise in type 2 diabetic obesity. Physiological Genomics 2014;46:747–765.

Maasar MF, Turner DC, Gorski PP, Seaborne RA, Strauss JA, Shepherd SO, et al. The comparative methylome and transcriptome after change of direction compared to straight line running exercise in human skeletal muscle. Frontiers in Physiology 2021;12.

Laker RC, Lillard TS, Okutsu M, Zhang M, Hoehn KL, Connelly JJ, et al. Exercise prevents maternal high‐fat diet‐induced hypermethylation of the Pgc‐1a gene and age‐dependent metabolic dysfunction in the offspring. Diabetes 2014;63:1605–1611.

Sailani MR, Halling JF, Møller HD, Lee H, Plomgaard P, Pilegaard H, et al. Lifelong physical activity is associated with promoter hypomethylation of genes involved in metabolism, myogenesis, contractile properties and oxidative stress resistance in aged human skeletal muscle. Scientific Reports 2019;9:3272.

Seaborne RA, Strauss J, Cocks M, Shepherd S, O'Brien TD, van Someren KA, et al. Human skeletal muscle possesses an epigenetic memory of hypertrophy. Scientific Reports 2018;8:1898.

Staron RS, Leonardi MJ, Karapondo DL, Malicky ES, Falkel JE, Hagerman FC, et al. Strength and skeletal muscle adaptations in heavy‐resistance‐trained women after detraining and retraining. Journal of Applied Physiology 1991;70:631–640.

Taaffe DR, Marcus R. Dynamic muscle strength alterations to detraining and retraining in elderly men. Clinical physiology (Oxford, England) 1997;17:311–324.

Henwood TR, Taaffe DR. Detraining and retraining in older adults following long‐term muscle power or muscle strength specific training. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences 2008;63:751–758.

Moberg M, Lindholm ME, Reitzner SM, Ekblom B, Sundberg C‐J, Psilander N. Exercise induces different molecular responses in trained and untrained human muscle. Medicine & Science in Sports & Exercise 2020; [DOI: https://dx.doi.org/10.1249/mss.0000000000002310]

Jones PA. Functions of DNA methylation: islands, start sites, gene bodies and beyond. Nature Reviews Genetics 2012;13:484–492.

Seaborne RA, Hughes DC, Turner DC, Owens DJ, Baehr LM, Gorski P, et al. UBR5 is a novel E3 ubiquitin ligase involved in skeletal muscle hypertrophy and recovery from atrophy. Journal of Physiology 2019;597.

Hughes DC, Turner DC, Baehr LM, Seaborne RA, Viggars M, Jarvis JC, et al. Knockdown of the E3 ubiquitin ligase UBR5 and its role in skeletal muscle anabolism. American Journal of Physiology ‐ Cell Physiology 2021;320.

Zahn JM, Sonu R, Vogel H, Crane E, Mazan‐Mamczarz K, Rabkin R, et al. Transcriptional profiling of aging in human muscle reveals a common aging signature. PLoS Genetics 2006;2:1058–1069.

Giresi PG, Stevenson EJ, Theilhaber J, Koncarevic A, Parkington J, Fielding RA, et al. Identification of a molecular signature of sarcopenia. Physiological Genomics 2005;21:253–263.

Su J, Ekman C, Oskolkov N, Lahti L, Ström K, Brazma A, et al. A novel atlas of gene expression in human skeletal muscle reveals molecular changes associated with aging. Skeletal Muscle 2015;5:35.

Shafiee G, Asgari Y, Soltani A, Larijani B, Heshmat R. Identification of candidate genes and proteins in aging skeletal muscle (sarcopenia) using gene expression and structural analysis. PeerJ 2018;6: [eLocator: e5239].

Curradi M, Izzo A, Badaracco G, Landsberger N. Molecular mechanisms of gene silencing mediated by DNA methylation. Molecular and Cellular Biology 2002;22:3157–3173.

Yuan T, Jiao Y, de Jong S, Ophoff RA, Beck S, Teschendorff AE. An integrative multi‐scale analysis of the dynamic DNA methylation landscape in aging. PLoS Genetics 2015;11:1–21.

Steegenga WT, Boekschoten MV, Lute C, Hooiveld GJ, de Groot PJ, Morris TJ, et al. Genome‐wide age‐related changes in DNA methylation and gene expression in human PBMCs. Age 2014;36:1523–1540.

Ferrucci L, Fabbri E. Inflammageing: chronic inflammation in ageing, cardiovascular disease, and frailty. Nature Reviews Cardiology 2018;15:505–522.

López‐otín C, Blasco MA, Partridge L, Serrano M, Kroemer G. The hallmarks of aging longevity. Cell 2013;153:1194–1217.

Mero AA, Hulmi JJ, Salmijärvi H, Katajavuori M, Haverinen M, Holviala J, et al. Resistance training induced increase in muscle fiber size in young and older men. European Journal of Applied Physiology 2013;113:641–650.

Tieland M, Trouwborst I, Clark BC. Skeletal muscle performance and ageing. Journal of Cachexia, Sarcopenia and Muscle 2018;9:3–19.

Fragala MS, Cadore EL, Dorgo S, Izquierdo M, Kraemer WJ, Peterson MD, et al. Resistance training for older adults: Position statement from the national strength and conditioning association. Journal of strength and conditioning research 2019;33:2019–2052.

Candow D, Chilibeck P, Abeysekara S, Zello G. Short‐term heavy resistance training eliminates age‐related deficits in muscle mass and strength in healthy older males. Journal of Strength and Conditioning Research 2011;25:326–333.

McLeod M, Breen L, Hamilton DL, Philp A. Live strong and prosper: the importance of skeletal muscle strength for healthy ageing. Biogerontology 2016;17:497–510.

Rebelo‐Marques A, Lages ADS, Andrade R, Ribeiro CF, Mota‐Pinto A, Carrilho F, et al. Aging hallmarks: the benefits of physical exercise. Frontiers in Endocrinology 2018;9:258.

Turner DC, Gorski PP, Maasar MF, Seaborne RA, Baumert P, Brown AD, et al. DNA methylation across the genome in aged human skeletal muscle tissue and muscle‐derived cells: the role of HOX genes and physical activity. Scientific Reports 2020;10.

Lund J, Rustan AC, Løvsletten NG, Mudry JM, Langleite TM, Feng YZ, et al. Exercise in vivo marks human myotubes in vitro: training‐induced increase in lipid metabolism. PLoS ONE 2017;12: [eLocator: e0175441].

Harford TJ, Kliment G, Shukla GC, Weyman CM. The muscle regulatory transcription factor MyoD participates with p53 to directly increase the expression of the pro‐apoptotic Bcl2 family member PUMA. Apoptosis 2017;22:1532–1542.

Yang T, Buchan HL, Townsend KJ, Craig RW. MCL‐1, a member of the BCL‐2 family, is induced rapidly in response to signals for cell differentiation or death, but not to signals for cell proliferation. Journal of Cellular Physiology 1996;166:523–536.

Popov DV, Makhnovskii PA, Kurochkina NS, Lysenko EA, Vepkhvadze TF, Vinogradova OL. Intensity‐dependent gene expression after aerobic exercise in endurance‐trained skeletal muscle. Biology of Sport 2018;35:277–289.

Vella L, Caldow MK, Larsen AE, Tassoni D, Gatta PAD, Gran P, et al. Resistance exercise increases NF‐κB activity in human skeletal muscle. American Journal of Physiology ‐ Regulatory Integrative and Comparative Physiology 2012;302:667–673.

Calle MC, Fernandez ML. Effects of resistance training on the inflammatory response. Nutrition Research and Practice 2010;4:259–269.

Li L, Fan C. A CREB‐MPP7‐AMOT regulatory axis controls muscle stem cell expansion and self‐renewal competence. Cell Reports 2017;21:1253–1266.

Pardo JV, D'Angelo Siliciano J, Craig SW. A vinculin‐containing cortical lattice in skeletal muscle: transverse lattice elements (‘costameres’) mark sites of attachment between myofibrils and sarcolemma. Proceedings of the National Academy of Sciences of the United States of America 1983;80:1008–1012.

Rindom E, Vissing K. Mechanosensitive molecular networks involved in transducing resistance exercise‐signals into muscle protein accretion. Frontiers in Physiology 2016;7:547.

Chorev DS, Volberg T, Livne A, Eisenstein M, Martins B, Kam Z, et al. Conformational states during vinculin unlocking differentially regulate focal adhesion properties. Scientific Reports 2018;8:2693.

Rezvani M, Ornatsky O, Connor M, Eisenberg H, Hood D. Dystrophin, vinculin, and aciculin in skeletal muscle subject to chronic use and disuse. Med Sci Sports Exerc 1996;28:79–84.

Graham ZA, Gallagher PM, Cardozo CP. Focal adhesion kinase and its role in skeletal muscle. J Muscle Res Cell Motil 2015;36:305–315.

Taylor DL, Jackson AU, Narisu N, Hemani G, Erdos MR, Chines PS, et al. Integrative analysis of gene expression, DNA methylation, physiological traits, and genetic variation in human skeletal muscle. Proceedings of the National Academy of Sciences of the United States of America 2019;166:10883–10888.

von Haehling S, Morley JE, Coats AJS, Anker SD. Ethical guidelines for publishing in the journal of cachexia, sarcopenia and muscle: update 2017. Journal of Cachexia, Sarcopenia and Muscle 2017;8:1081–1083.