DNA methylation across the genome in aged human skeletal muscle tissue and stem cells: The role of HOX genes and physical activity

Skeletal muscle tissue demonstrates global hypermethylation with aging. However, methylome changes across the time-course of differentiation in aged human muscle derived stem cells, and larger coverage arrays in aged muscle tissue have not been undertaken. Using 850K DNA methylation arrays we compared the methylomes of young (27 ± 4.4 years) and aged (83 ± 4 years) human skeletal muscle and that of young/aged muscle stem cells over several time points of differentiation (0, 72 hours, 7, 10 days). Aged muscle tissue was hypermethylated compared with young tissue, enriched for; ‘pathways-in-cancer’ (including; focal adhesion, MAPK signaling, PI3K-Akt-mTOR signaling, p53 signaling, Jak-STAT signaling, TGF-beta and notch signaling), ‘rap1-signaling’, ‘axon-guidance’ and ‘hippo-signalling’. Aged muscle stem cells also demonstrated a hypermethylated profile in pathways; ‘axon-guidance’, ‘adherens-junction’ and ‘calcium-signaling’, particularly at later timepoints of myotube formation, corresponding with reduced morphological differentiation and reductions in MyoD/Myogenin gene expression compared with young cells. While young cells showed little alteration in DNA methylation during differentiation, aged cells demonstrated extensive and significantly altered DNA methylation, particularly at 7 days of differentiation and most notably in the ‘focal adhesion’ and ‘PI3K-AKT signalling’ pathways. While the methylomes were vastly different between muscle tissue and isolated muscle stem cells, we identified a small number of CpG sites showing a hypermethylated state with age, in both muscle and tissue and stem cells (on genes KIF15, DYRK2, FHL2, MRPS33, ABCA17P). Most notably, differential methylation analysis of chromosomal regions identified three locations containing enrichment of 6-8 CpGs in the HOX family of genes altered with age. With HOXD10, HOXD9, HOXD8, HOXA3, HOXC9, HOXB1, HOXB3, HOXC-AS2 and HOXC10 all hypermethylated in aged tissue. In aged cells the same HOX genes (and additionally HOXC-AS3) displayed the most variable methylation at 7 days of differentiation versus young cells, with HOXD8, HOXC9, HOXB1 and HOXC-AS3 hypermethylated and HOXC10 and HOXC-AS2 hypomethylated. We also determined that there was an inverse relationship between DNA methylation and gene expression for HOXB1, HOXA3 and HOXC-AS3. Finally, increased physical activity in young adults was associated with oppositely regulating HOXB1 and HOXA3 methylation compared with age. Overall, we demonstrate that a considerable number of HOX genes are differentially epigenetically regulated in aged human skeletal muscle and muscle stem cells and increased physical activity may help prevent age-related epigenetic changes in these HOX genes.


75
Maintaining skeletal muscle mass and function into older age is fundamental for human health-span and 76 quality of life 1 . Five to ten percent of older humans have sarcopenia 2, 3 , which is characterized by reductions 77 in muscle mass and strength 4 . This loss of muscle mass and strength leads to frailty, increased incidence of 78 falls, hospitalization and morbidity 4,5,6,7,8,9,10,11 . Annual costs of fragility are estimated to be 39/32 billion 79 (Euros/USD) for European and USA fragility fractures respectively, with the cost of sarcopenia estimated to 80 be £2 billion in the UK 12 . With an ageing population, these costs are likely to increase with time.

82
A primary hallmark of ageing is the alteration of the epigenetic landscape. Epigenetics encompasses the 83 interaction between lifestyle/environmental factors and modifications to DNA and histones, without changes 84 to the inherited DNA sequence 13,14 . DNA methylation is the most studied epigenetic modification and 85 involves the addition of a covalent methyl group to the 5' position of the pyrimidine ring of a cytosine (5mC).

86
Increased methylation (hypermethylation) to cytosine-guanine (C-G) pairings (CpG sites), especially in CpG-87 rich regions such as gene promoters, typically leads to reduced capacity for the transcriptional apparatus to 88 bind to these regions, suppressing gene expression 14 . Methylated CpG islands in promoters also leads to a 89 tight compaction of adjacent chromatin via the recruitment of chromatin modifying protein/protein-90 complexes, further silencing gene expression. In contrast, reduced methylation (hypomethylation) provides 91 a more favorable DNA landscape for the transcriptional apparatus to bind to these regions, as well as more 92 'relaxed' chromatin, enabling gene expression to occur.

94
DNA methylation in aged skeletal muscle occurs at tissue-specific genes 15 . However, aged muscle also has 95 the smallest overlap with other aged tissue types, suggesting skeletal muscle is unique in comparison with 96 other tissues in its epigenetic aging processes 15 . Indeed, it has recently been demonstrated that the 97 methylation status of approximately 200 CpG sites can accurately predict chronological age in skeletal muscle 98 tissue 16 . But that this muscle 'clock' only shares 16 of these CpG's with the original 353 CpG pan-tissue 99 Horvath clock 16,17 . Further, using DNA methylation arrays with coverage of ~450,000 CpG sites 18 , Zykovich 100 et al. demonstrated that compared with young human skeletal muscle, aged skeletal muscle is 101 hypermethylated across the genome. Moreover, our group has demonstrated that mouse skeletal muscle 102 stem cells exposed to a high dose of inflammatory stress in early proliferative life retained hypermethylation 103 of MyoD (a muscle-specific regulatory factor) 30 population doublings later 19 . This suggests that inflamed 104 proliferative aging in muscle stem cells leads to a retained accumulation of DNA methylation. Finally, lifelong 105 physical activity 20 , endurance and resistance exercise have been associated with predominantly 106 hypomethylation of the genome in young skeletal muscle 21,22 . This contrasts with the hypermethylation 107 observed with aging, suggesting that exercise may reverse some age-related changes in DNA methylation. 108 109 4 Skeletal muscle fibers are post-mitotic as they contain terminally differentiated/fused nuclei (myonuclei); 110 thus, repair and regeneration of skeletal muscle tissue is mediated by a separate population of resident stem 111 cells (satellite cells) that can divide. Once activated, satellite cells proliferate and migrate to the site of injury 112 to differentiate and fuse with the existing fibers to enable repair. Target gene analysis showed altered DNA 113 methylation during differentiation of muscle cells into myotubes in-vitro 23 . This included altered methylation 114 of MyoD 24 , Myogenin 25 and Six1 26 . While muscle stem cells derived from aged individuals display similar 115 proliferative capacity and time to senescence as young adult cells 27, 28 , they do have impaired differentiation 116 and fusion into myotubes 29,30,31,32,33,34,35,36,37,38,39,40,41,42,43,44,45 . However, a small number of studies did 117 not find an effect of age on the differentiation capacity of isolated cells 27,46,47 Seaborne et al. (2018). This is because we used this baseline tissue to derive 139 cells (detailed below) for DNA methylation analysis of stem cell experiments in the present study. For older 140 adults (n = 5, 2 men/3 women, 83 ± 4 years), tissue biopsies were obtained during elective orthopedic 141 surgeries from University Hospitals of the North Midlands, from the vastus lateralis (knee surgery, n = 2) or 142 gluteus medius muscles (hip surgery, n = 3), under consent and ethical approval 18/WM/0187. DNA and RNA 143 were isolated from these young and aged tissue samples. DNA samples from all 9 young and 5 aged adults 144 were analysed for DNA methylation arrays (detailed below), and a subset were analysed for gene expression 145 5 (young n = 4, aged n = 5). Primary skeletal muscle cells were derived from a subset of young adult and aged 146 tissue samples, and isolated as per our previous work 21,22,48,49,50,51 . Briefly, approximately 100 mg biopsy 147 tissue was immediately (~10-30 mins) transferred to a sterile class II biological safety cabinet in pre-cooled 148 (4˚C) transfer media (Hams F-10, 2% hi-FBS, 100 U/ml penicillin, 100 μg/ml streptomycin and 2.5 µg/ml 149 amphotericin-B). Any visible connective and adipose tissue were removed using sterile scalpels and muscle 150 tissue was thoroughly washed 2 × in sterile PBS (containing 100 U/ml penicillin, 100 µg/ml streptomycin, 2.5 151 µg/ml amphotericin-B). PBS was removed and the muscle tissue was minced in the presence of 0.05% 152 Trypsin/0.02% EDTA and all contents (tissue and trypsin) were transferred to a magnetic stirring platform at 153 37°C for 10 minutes. The addition of 0.05% Trypsin/0.02% EDTA and titration was repeated on any remaining 154 tissue. The supernatant was collected from both procedures and horse serum (HS) was added at 10% of the 155 total supernatant volume to neutralize the trypsin. The supernatant was centrifuged at 340 g for 5 minutes

166
at the same time of day (approx. 08.00-11.00) to minimize the impact of circadian oscillation 52 . All 167 experiments were carried out below passage 10 to prevent senescence. We undertook methylation arrays 168 on DNA isolated from: 0 h young (n = 7), 0 h aged (n = 4), 72 h young (n = 7), 72 h aged (n = 4), 7 d young (n 169 = 6), 7d aged (n = 3), 10 d young (n = 2) and 10 d aged (n= 4). Gene expression was analysed using young (n 170 = 4, 7 d) and aged (n = 3, 7 d) cells. It's worth noting that we had n= 3-4 for most conditions, however in the 171 10 d young cells condition, the DNA did not pass QA/QC for the arrays, and we had no cells left for these 172 participants. Therefore, unfortunately we could only run n = 2 for this single condition. Therefore, results for 173 this condition should be viewed with this caveat in mind. A schematic of experimental design can be found 174 in Suppl.  cultures. Briefly, approximately 2 × 10 4 cells were seeded onto 3 × wells of a 6-well plate and were incubated 180 for 24 h in GM. Existing media was removed and cells were washed 3 × in PBS before fixation using the graded 181 6 methanol/acetone method (50:25:25 TBS:methanol:acetone for 15 minutes followed by 50:50 182 methanol:acetone) after which cells were permeabilised in 0.2% Triton X-100 and blocked in 5% goat serum 183 (Sigma-Aldrich, UK) in TBS for 30 minutes. Cells were washed 3 × in TBS and incubated overnight (4 ˚C) in 300 184 μl of TBS (with 2% goat serum and 0.2% Triton X-100) containing primary anti-desmin antibody (1:50; 185 ab15200, Abcam, UK). After overnight incubation, cells were washed 3 × in TBS and incubated at RT for 3 h 186 in 300 μl of secondary antibody solution (TBS, 2% goat serum and 0.2% Triton X-100) containing the 187 secondary antibody, anti-rabbit TRITC (1:75; T6778, Sigma-Aldrich, UK) to counterstain desmin. Finally, cells 188 were washed again 3 × in TBS, prior to counterstaining nuclei using 300 μl of DAPI solution at a concentration 189 of (300 nM; D1306, Thermo Fisher Scientific, UK) for 30 minutes. Immunostained cells were then visualised 190 using a fluorescent microscope (Nikon, Eclipse Ti-S, Japan or Olympus IX83, Japan) and imaged using     Figure 2a), which is well below the 224 recommended 0.01 in the Oshlack workflow 53 . We also produced density plots of the raw intensities/signals 225 of the probes (Suppl. Figure 2b). These demonstrated that all methylated and unmethylated signals were 226 over 11.5 (mean was 11.52 and median was 11.8), and the difference between median methylation and 227 median unmethylated signal was 0.56 53 . Upon import of the data into Partek Genomics Suite we removed 228 probes that spanned X and Y chromosomes from the analysis due to having both males and females in the 229 study design, and although the average detection p-value for each samples was on average very low (no 230 higher than 0.0023) we also excluded any individual probes with a detection p-value that was above 0.01 as 231 recommended in 53 . Out of a total of 865,860 probes, removal of the X and Y chromosome probes and those 232 with a detection p-value above 0.01 reduced the probe number to 846,233 (removed 19,627 probes). We 233 also filtered out probes located in known single-nucleotide polymorphisms (SNPs) and any known cross-   analyses, e.g. > 100,000 differentially methylated CpG sites were identified at FDR ≤ 0.05, so we also show 255 results at a more stringent FDR of ≤ 0.01 or 0.001, or at FDR of ≤ 0.05 and 'a change (difference in M-value) 256 in methylation greater than 2'. These shorter lists of CpGs contain the most significant sites to enable sensible 257 pathway analysis. We specify when a more stringent FDR than 0.05 was used in the results text. We then 258 undertook CpG enrichment analysis on these differentially methylated CpG lists within gene ontology and 259 KEGG pathways 57, 58, 59 using Partek Genomics Suite and Partek Pathway. We also undertook differentially

277
(Qiagen, Manchester, UK). Reverse transcription was initiated with a hold at 50˚C for 10 minutes (cDNA 278 synthesis) and a 5-minute hold at 95˚C (transcriptase inactivation and initial denaturation), before 40-50 PCR 279 cycles of; 95˚C for 10 sec (denaturation) followed by 60˚C for 30 sec (annealing and extension). Primer 280 sequences for genes of interest and reference genes are included in Table 1. All genes demonstrated no 281 unintended targets via BLAST search and yielded a single peak after melt curve analysis conducted after the 282 PCR step above. All relative gene expression was quantified using the comparative Ct ( ∆∆ Ct) method 60 . For 283 human cell differentiation analysis via measurement of myoD and myogenin, a pooled mean Ct for the 0 h 284 young adult control samples were used as the calibrator when comparing aged vs. young adult cells. This 285 approach demonstrated a low % variation in Ct value of 9.5 and 8.5% for myoD and myogenin, respectively.

286
For HOX gene analysis between young and aged tissue and for the 7 d aged cells vs. 7 d young adult cells, the 287 mean Ct of the young adult cells were used as the calibrator. The average, standard deviation and variations 288 in Ct value for the B2M reference gene demonstrated low variation across all samples (mean ± SD, 13.12 ± 9 0.98, 7.45% variation) for the analysis of myoD and myogenin. For HOX gene analysis, the RPL13a reference 290 gene variation was low in the human tissue (17.77 ± 1.71, 9.6% variation) and stem cell (15.51 ± 0.59, 3.82% 291 variation) experiments. The average PCR efficiencies of myoD and myogenin were comparable (94.69 ± 8.9%, 292 9.4% variation) with the reference gene B2M (89.45 ± 3.76%, 4.2% variation). The average PCR efficiencies 293 of the all genes of interest in the tissue analysis of the HOX genes were also comparable (90.87 ± 3.17%, 294 3.39% variation) with the human reference gene RPL13a (92 ± 2.67%, 2.9% variation). Similarly, for the cell 295 analysis, HOX genes of interest efficiencies were comparable (89.59 ± 4.41%, 4.93% variation 4.93%) with the 296 reference gene RPL13a (89.57 ± 3.55%, 3.97% variation). Statistical analysis for HOX genes was performed 297 using t-tests (aged tissue vs. young tissue and 7d aged versus 7d young).

299
Physical Activity and DNA methylation 300 The human association study involved 30 physically active and endurance-oriented men of Eastern European 301 descent (32.9 ± 9.9 years). The study was approved by the Ethics Committee of the Physiological Section of

308
Bisulfite conversion of genomic DNA was performed using the EpiMark® Bisulfite Conversion Kit in 309 accordance with the manufacturer's instructions. In the same analysis as described above for the aged muscle 310 tissue and stem cell data, methylome of the vastus lateralis in the physically active men was evaluated using 311 the Infinium MethylationEPIC 850K BeadChip Array (Illumina, USA) and imaged and scanned using the 312 Illumina iScan® System (Illumina, United States). Also as in the above analysis we filtered out probes located 313 in known single-nucleotide polymorphisms (SNPs) and any known cross-reactive probes using previously 314 defined SNP and cross-reactive probe lists 61 . Low quality probes were also filtered, as with the above 315 analysis. The final analyses included 796,180 of 868,565 probes. Data were normalised using the same 316 functional normalisation (with noob background correction), as in the aged tissue and stem cell data, and as 317 previously described 55 . Any outliers were interrogated via PCA as above, however all samples passed the QC 318 and therefore there were no outliers. The methylation level of each CpG-site after normalization and filtering 319 processes was represented as a β-value ranging from 0 (unmethylated) to 1 (fully methylated) in order to  Figure 3b), and 'molecular function' (Suppl. Figure 3c). Within 341 the GO terms that included the search term 'muscle', the most significantly enriched was 'regulation of

531
Distinguishing differentiation-specific CpG sites in aged cells versus those altered as a consequence of age 532 alone 533 The data above suggest that aged cells demonstrate hypermethylation versus young adult cells across all 534 stages of differentiation, and that aged cells significantly altered their methylation profiles at 7 days of 535 differentiation. Therefore, we conducted further analysis on the overlap of DMPs within aged cells at 7 d of 536 differentiation (from the 'time' analysis above) and those that were changed as a consequence of age at 7 537 days (aged cells at 7 days versus young cells at 7 d). This enabled the identification of which methylation sites 538 were altered, but also shared in both aged cell differentiation alone and as a consequence of age. Or, 539 alternatively the sites that were simply changed with age and not differentiation process and vice versa.

555
Tissue and cell PCA plots demonstrated that methylation of even late differentiated cells was vastly different 556 to tissue, suggesting that cell versus tissue samples were comprised of vastly different methylation profiles 557 (Suppl. Figure 8). Therefore, in order to compare if there were any sites similarly altered in the cells that were 558 also altered in the tissue with age, we overlapped the DMP lists from the tissue and cell analysis described 559 above. Six CpG's that were identified in the 6,828 significantly differentially methylated CpG list from the 560 aged versus young tissue analysis, as well as highlighted in the 2,719 list 'age' cells analysis (Figure 4). These

602
File 9c). A Venn diagram analysis depicted the overlap in common gene symbols for these three gene lists 603 (Figure 5a, Suppl. File 9d). It was also demonstrated that the majority of these HOX genes were 604 hypermethylated in aged tissue (Figure 5b; Suppl. File. 9a). In the cell analysis across all time, these HOX 605 genes also displayed the most varied methylation at 7 days of differentiation in aged cells versus young cells, 606 therefore confirming the varied temporal profile in methylation described above at 7 d was also the case for 607 these HOX genes. Finally, when SOM-profiling these 9 HOX genes by symbol (17 CpGs as some HOX genes 608 contained more than one CpG site), over the time-course of differentiation based on the main effect for 'age' HOXC10 (note this Venn diagram analysis is by 'gene symbol' not 'probe cg' (CpG site), as some HOX genes also had more than 1 CpG 622 per gene symbol, full CpG lists are located in Suppl. Figure 9 a,b,c,d). b. All HOX family genes by CpG site (cg probe) differentially 623 methylated in aged compared with young skeletal muscle tissue, predominantly all demonstrating hypermethylation. c. SOM profiling 624 depicting the temporal regulation of DNA methylation in aged muscle stem cells as they differentiate in CpGs located amongst the 625 HOX family of genes (depicting the 9 HOX genes altered, by gene symbol, in both the tissue and cells). The majority of these HOX CpG 626 sites were differentially methylated at 7 days of differentiation in the aged cells.

20 628
We next analysed the gene expression of the HOX genes that changed at the methylation level in both the 629 tissue and cell analysis (by gene symbol-HOXD8, HOXA3, HOXC9, HOXB1, HOXB3, HOXC-AS2 and HOXC10).

630
Interestingly, in aged tissue, despite displaying hypermethylation in all these HOX genes versus young tissue, 631 the genes were not suppressed at the gene expression level, which may have been expected, yet all elevated 632 (Suppl. Figure 9). However, in the aged cells when analysing these genes at the expression level at 7 days of 633 differentiation (including HOXD8, HOXA3, HOXC9, HOXB1, HOXB3, HOXC-AS2 and HOXC10, as well as HOXC-634 AS3 identified in the above in the cell analysis only) (Figure 6), we identified that there was significantly 635 reduced gene expression in gene HOXB1 (Figure 6), that was inversely related with increased HOXB1 636 methylation. Where in the 5,524 DMP list at 7 d in aged cells versus 7 d young adult cells (Suppl. File 5o), we 637 previously identified that a region of the HOXB1 located in Chr17:46607104-46608568, contained 8 CpG's 638 that were hypermethylated, as well as HOXB1 Cg's: cg04904318, cg02497558, cg22660933, cg10558129 639 being identified as hypermethylated in the 2,719 significant main effect for 'age' cell CpG list above (Suppl.

640
File 3b). There was also significantly increased gene expression for HOXC-AS3, with this gene identified 641 earlier in the analysis as being having reduced (hypo)methylation. Indeed, hypomethylation occurred in 6 642 CpG's in a region upstream of the HOXC10 gene (Chr12:54376020-54377868, 1849 bp) within the HOXC-AS3 643 gene. Interestingly, HOXC10 also demonstrated an average increase in gene expression, however, it was not 644 statistically significant (Figure 6). Finally, HOXA3 also demonstrated significantly reduced expression ( Figure   645 6) with corresponding hypermethylation (at 10 not 7 days of differentiation) in aged cells (see Figure 5c).

21
Effect of physical activity on methylation status of HOX family genes 651 Next, given that aging generally hypermethylated the genome, we tested the hypothesis that increasing 652 physical activity may oppositely regulate DNA methylation and be associated with increasing 653 hypomethylation in these HOX genes. We thus performed a multiple regression analysis using methylation 654 data and the level of physical activity of 30 endurance-trained men. As in the above analysis, we also found 655 that CpG-sites associated with physical activity (P<0.05) were significantly enriched with HOX genes (137 of 656 1219 CpG-sites, Fisher's exact test OR=1.7, P=2.3*10-8, Suppl. File 10a). Where we determined that highly 657 active men had hypomethylated HOXB1 (cg10558129, P=5.2*10 -4 ), HOXA3 (cg16406967, P=0.03), HOXD12 658 (cg17512364, P=0.008) and HOXC4 gene (cg13826247, P=0.014) compared to less active men (adjusted for 659 age and muscle fiber composition) on the same sites identified above in the aging data. Furthermore, we

671
In the present study we first aimed to investigate the methylome in aged skeletal muscle tissue and 672 differentiating primary muscle stem cells compared with young adults, in order to identify important 673 epigenetically regulated genes in both aging skeletal muscle tissue and muscle derived stem cells. As with 674 previous studies 18, 27 , and by using more recent, higher coverage array technology, we identified that aged 675 skeletal muscle tissue demonstrated a considerably hypermethylated profile compared with young adult 676 tissue. We also demonstrated that these hypermethylated profiles in aged tissue were enriched in gene 677 ontology pathways including, 'regulation of muscle system process' and KEGG pathways 'pathways in cancer', 678 a pathway that incorporates previously well described molecular pathways in the regulation of skeletal 679 muscle such as; focal adhesion, MAPK signaling, PI3K-Akt-mTOR signaling, p53 signaling, Jak-STAT signaling, 680 TGF-beta and Notch signaling, as well as the other significantly enriched pathways of 'rap1 signaling', 'axon 681 guidance', and 'hippo signaling'. This was also the first study to profile DNA methylation over the entire time-

731
Importantly, in both tissue and stem cell analysis, we also identified that the homeobox (HOX) family of genes 732 were significantly enriched in differentially methylated region analysis, showing several (e.g. 6-8) CpGs to be 733 methylated within chromosomal regions on these genes in aged compared with young adults. In particular,

734
we identified: HOXC10 (just upstream of HOXC6) and HOXB1 as having several CpGs differentially 735 methylated. Therefore, closer analysis of all HOX gene associated CpG changes across both tissue and cell 736 differentiation data identified that CpG's located within: HOXD10, HOXD9, HOXD8, HOXA3, HOXC9, HOXB1, 737 HOXB3, HOXC-AS2 and HOXC10 were all significantly differentially methylated across these analyses. In aged 738 tissue the majority of these HOX genes were hypermethylated. In the cell analysis, these HOX genes displayed average, yet not significant increase in HOXC10 gene expression) in aged muscle cells at 7 days of 775 differentiation. Given the data above in bone and cancer, HOXC-AS3 upregulation appears to be pro-growth 776 and linked with expression of HOXC10, therefore their increase in the current study maybe hypothesised to 777 be a co-operative and compensatory drive to maintain aged muscle. However, this hypothesis is speculative, 778 and more work needs to be conducted as to the role of HOXC10 and HOXC-AS3 and their potential 779 cooperative mechanisms of action in aged skeletal muscle.

781
We also identified that HOXB1 was hypermethylated with increased gene expression in aged cells at 7 days.

782
HOXB1 has been demonstrated to be hypermethylated in inflamed muscle of children with Juvenile 783 Dermatomyositis (JDM) 79 . This is interesting given aged skeletal muscle is known to be chronically inflamed profiles and we also removed sex (X and Y) chromosome probes from the analysis. Furthermore, the muscle 795 tissue and cells were isolated from two sites, the gluteus medius and vastus lateralis in the aged samples 796 compared with only the vastus lateralis in the young adult samples. This is important to note as the gluteus 797 medius and vastus lateralis have different fibre type proportions. Therefore, the data should be viewed with 798 that caveat in mind. Despite this, the changes in methylation of the sites observed were identified across 799 samples from both biopsy sites, suggesting that these changes occurred in aged individuals across muscle 800 types.

802
Finally, with aging evoking a hypermethylated signature in tissue and aged muscle derived stem cells, it was 803 also interesting to speculate that physical exercise, that has been shown to hypomethylate the genome 20, 21, 804 22 , could therefore be 'anti-ageing' at the epigenetic level. Indeed, this hypothesis was supported indirectly 805 in the present study and by previous literature, where the aged tissue analysis in the present study identified 806 the top significantly enriched KEGG pathway as, 'pathways in cancer'. A pathway that incorporates well 807 known pathways important in skeletal muscle, including: Focal adhesion, MAPK, PI3K-Akt, mTOR, p53 808 signaling, Jak-STAT, TGF-beta and Notch signaling. Where, this pathway was also the top enriched 809 hypomethylated pathway after acute and chronic resistance exercise 22 . While, perhaps the significance of 810 these larger pathways such as 'pathways in cancer' can be inflated in methylation analysis 82 , and therefore 811 should be viewed with some caution. This data perhaps suggests that exercise (resistance exercise) could 812 perhaps reverse the hypermethylated profiles in these pathways in aged muscle. Therefore, in the present 813 study, given that we identified the HOX family of genes to be extensively differentially methylated in aged 814 tissue and stem cells, we went on to determine that increasing physical activity levels (endurance exercise) 815 in healthy young adults was associated with larger reductions in HOXB1 and HOXA3 methylation 816 (hypomethylation). This was opposite to the changes we observed with age in both muscle tissue and stem 817 cells, that demonstrated increased hypermethylation with age. This also provided evidence to suggest that 818 increased physical activity could perhaps reverse the age-related epigenetic changes in the HOX genes.

819
Importantly, HOXA3 has been shown to be hypomethylated on multiple sites after resistance exercise 820 training, with retention of hypomethylation for site (cg12434681) during detraining into retraining 21 .

821
Suggesting, that this HOX gene is also hypomethylated with exercise training and possesses an epigenetic 822 memory from earlier exercise training as previously described by our group 60 . Unpublished work by our 823 group also suggests that acute sprint exercise in human muscle can hypomethylated the same HOXA3 CpG 824 site (cg00431187) that was oppositely hypermethylated in aging but demonstrated hypomethylation in 825 individuals that have higher physical activity level. However, more research into the effect of exercise in an 826 aged population and the changes in HOX methylation status will be required in the future to confirm these 827 findings. Our overarching main results from this study are summarised schematically in Figure 7.