Review

1 65 2 Exercise-associated DNA methylation change in 66 3 67 4 skeletal muscle and the importance of imprinted 68 5 69 Q16 genes: a bioinformatics meta-analysis 70 Q2¶7 71 Q3¶8 William M Brown 72 ¶9 73 10 74 ▸ Additional material is 11 ABSTRACT processes of development, with an emphasis on 75 published online only. To view Background Epigenetics is the study of processes— interactions among genes and between genes and 12 please visit the journal online 76 — the environment. In contrast, Nanney3 used ‘epi- 13 (http://dx.doi.org/10.1136/ beyond DNA sequence alteration producing heritable 77 fi ’ 14 bjsports-2014-094073). characteristics. For example, DNA methylation modi es genetic in 1958 to describe a system of cellular 78 gene expression without altering the nucleotide heredity that was not based on DNA sequence. 15 Correspondence to 79 16 Dr William M Brown, sequence. A well-studied DNA methylation-based Most molecular biologists use the term epigen- 80 17 Department of Sport Science phenomenon is genomic imprinting (ie, genotype- etic to mean the study of heritable changes in gene 81 and Physical Activity, Faculty of 18 independent parent-of-origin effects). expression or cellular phenotype caused by 82 Education and Sport, Health Objective We aimed to elucidate: (1) the effect of mechanisms other than changes in the underlying 19 Epigenetics and Ageing 83 DNA sequence—hence the name epi- (Greek for 20 Laboratory (HEAL), Institute of exercise on DNA methylation and (2) the role of 84 Sport and Physical Activity imprinted genes in skeletal muscle gene networks ‘over’, ‘above’ or ‘outer’) -genetics. Examples of Q7 21 fi 85 22 Research, University of (ie, gene group functional pro ling analyses). such changes are DNA methylation and histone 86 Bedfordshire, Polhill Avenue, fi fi 23 Design Gene ontology (ie, gene product elucidation)/ modi cations ( gure 1), both of which can regulate Q887 Bedford, Bedfordshire, MK41 meta-analysis. gene expression without altering the underlying 24 9EA, UK; 88 Q425 [email protected] Data sources 26 skeletal muscle and 86 imprinted DNA sequence. One well-studied epigenetic phe- 89 fi ¶26 genes were subjected to g:Pro ler ontology analysis. nomenon based on DNA methylation in mammals 90 Accepted 18 February 2015 4 27 Meta-analysis assessed exercise-associated DNA (and notably humans) is genomic imprinting. One 91 28 methylation change. purpose of this review is to introduce this term to 92 Data extraction g:Profiler found four muscle gene the sports and exercise medicine/physiotherapy 29 fi 93 30 networks with imprinted loci. Meta-analysis identi ed 16 community and explain its importance. 94 31 articles (387 genes/1580 individuals) associated with 95 32 exercise. Age, method, sample size, sex and tissue Genomic imprinting 96 33 variation could elevate effect size bias. Genomic imprinting is defined as genotype-independ- 97 34 Data synthesis Skeletal muscle gene networks ent parent-of-origin gene expression. Specifically, for 98 Q635 including imprinted genes reported. Exercise-associated most genes we inherit two working parental copies. 99 36 effect sizes were calculated by gene. Age, method, However, in the case of imprinted genes, an epigen- 100 37 sample size, sex and tissue variation were moderators. etic tag (via DNA methylation) is placed on either the 101 38 Results Six imprinted loci (RB1, MEG3, UBE3A, maternal or paternal copy rendering the other 102 PLAGL1, SGCE, INS) were important for muscle gene 39 fi inactive. Such parent-of-origin gene expression is 103 40 networks, while meta-analysis uncovered ve exercise- mediated by epigenetic modifications which differ 104 41 associated imprinted loci (KCNQ1, MEG3, GRB10, between the two parentally derived chromosomes.5 105 42 L3MBTL1, PLAGL1). DNA methylation decreased with Approximately 1% of the human genome is 106 43 exercise (60% of loci). Exercise-associated DNA imprinted.67Despite their rarity, imprinted genes are 107 44 methylation change was stronger among older people of great medical importance. Studies of metabolic 108 45 (ie, age accounted for 30% of the variation). Among growth and neurodevelopmental disorders have 109 46 older people, genes exhibiting DNA methylation shown that imprinted genes are absolutely essential 110 – 47 decreases were part of a microRNA-regulated gene for healthy development.8 13 Furthermore, epigenetic 111 48 network functioning to suppress cancer. — 112 fi dysregulation in imprinted genes which often have 49 Conclusions Imprinted genes were identi ed in growth-enhancing and tumour suppressor functions 113 50 skeletal muscle gene networks and exercise-associated —predict disease and cancer outcomes. 114 51 DNA methylation change. Exercise-associated DNA Once epigenetic mechanisms emerged during 115 fi ‘ 52 methylation modi cation could rewind the epigenetic mammalian evolution (eg, genomic imprinting), a 116 ’ 53 clock as we age. source of environmental information (ie, 117 54 Trial registration number CRD42014009800. parent-of-origin of a gene) was transmitted transge- 118 55 nerationally. Imprinting machinery (eg, DNA 119 56 methylation, see figure 1—a type of silencer at a 120 57 INTRODUCTION gene’s promoter—places marks on a gene when the 121 58 British developmental biologist Sir Conrad H gametes are produced) allowed for biased gene 122 59 Waddington introduced the term ‘epigenetics’ as a expression in the subsequent generation (ie, mono- 123 60 science of development from genotype to pheno- allelic gene expression). 124 To cite: Brown WM. Br J 1 61 Sports Med Published Online type. However, the term ‘epigenetics’ had an inde- The realisation that the environment has pro- 125 62 First: [please include Day pendent origin and meaning, which led to a found influences on the epigenome has led to a 126 63 Month Year] doi:10.1136/ conflation of terms.2 Recall Waddington’s use of strong hypothesis that exercise can also affect DNA 127 64 bjsports-2014-094073 the term ‘epigenetics’ to refer to the causal methylation and have long-term health outcomes. 128

Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 1 Review

129 193 130 194 131 195 132 196 133 197 134 198 135 199 136 200 137 201 138 202 139 203 140 204 141 205 142 206 143 207 144 208 145 209 146 210 147 211 148 212 149 213 150 214 151 215 152 216 153 217 154 218 155 219 156 220 157 221 158 222 159 223 160 224 161 Figure 1 DNA methylation: DNA methylation occurs when methyl groups ‘tag’ DNA and activate or repress gene expression. DNA methyltransferase 225 162 is a catalyst, which transfers methyl groups to DNA. Silencing of a gene activity can occur if the hydrogen (H) molecule of cytosine (C) is replaced by a 226 163 methyl (Me) group at a gene’s promoter. Histones: Histones are proteins around which DNA can wind for compaction and histone modification can 227 fi 164 regulate gene activity. Both epigenetic processes (ie, DNA methylation and histone modi cations) affect health resulting in cancer, autoimmune 228 fi 165 disease, neurological disorders or diabetes. Image modi ed from the National Institutes of Health, Benjamin I. Laufer and Forluvoft. 229 166 230 167 Specifically, we propose that the same mechanism that controls methylation outside the promoter region (eg, gene body) is 231 168 genomic imprinting in mammals (ie, DNA methylation) allows sometimes associated with increased gene expression. In the 232 169 for phenotypic modification and the possibility of multiple case of genomic imprinting, hypermethylation of one of the 233 170 sources of environmental information (eg, exercise, nutrition) to two parental alleles leads to monoallelic expression (conceptu- 234 171 be transmitted to the next generation. Muscle and nerve cells ally similar to gene-dosage reduction in X-inactivation, see 235 172 share the property of responding to electrochemical/environ- figure 2). DNA methylation has been implicated in cancer, neu- 236 15 173 mental stimuli and thus are ideal epigenetic interfaces for trans- rodevelopmental disorders and autoimmune diseases. Thus, if 237 174 generational phenotypic modification, which may explain in exercise can influence DNA methylation, it may be the mechan- 238 175 part why imprinted genes are particularly involved in neural ism that underpins the lower cancer rate in those who are phys- 239 176 development. ically active. 240 177 241 178 DNA methylation DNA methylation, health and diverse disease states 242 179 DNA methylation refers to the adding of a methyl group on a Cancer cells are characterised by a global loss of DNA methyla- 243 180 cytosine base. It occurs primarily in the context of CpG dinu- tion among growth enhancers; and the coordinated acquisition 244 181 cleotides, which cluster in regions called CpG islands.14 ‘CpG’ of hypermethylation at the CpG islands of tumour suppressor 245 182 refers to regions of DNA where a cytosine nucleotide appears genes. Global hypomethylation occurs primarily at parasitic 246 183 next to guanine nucleotide interconnected by phosphate. When DNA regions of the genome. For example, the LINE family 247 184 cytosines in CpG dinucleotides are methylated to form member L1 is hypomethylated in a variety of cancers, such as 248 185 5-methylcytosine, a gene can be turned off. CpG dinucleotides those of the breast and colon.16 249 186 are rare in mammals (∼1%), but ∼50% of gene promoters are Neurological disorders are also associated with epigenetic dys- 250 187 linked to CpG islands which are often unmethylated in healthy regulation (ie, reversed patterns of a normal DNA methylation 251 188 cells. Cells become methylated in a tissue-specific and age- profile). Specifically, dysregulation of DNA methylation occurs 252 189 specific manner during development. Where DNA methylation in several neurological diseases, giving rise to hypermethylated 253 190 occurs can be critical for its effect. DNA methylation (ie, adding and hypomethylated CpG sites. FMR1 promoter hypermethyla- 254 191 methyl groups to a cytosine base, figure 1) at a gene’s promoter tion occurs among individuals diagnosed with Fragile X syn- 255 192 is linked to silencing (ie, less gene expression); in contrast, DNA drome. Rett syndrome, an X-linked neurological disease, is 256

2 Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 Review

257 321 258 322 259 323 260 324 261 325 262 326 263 327 264 328 265 329 266 330 267 331 268 332 269 333 270 334 271 335 272 336 273 337 274 338 275 339 276 340 277 341 278 342 279 343 280 344 281 345 282 346 283 347 284 348 285 349 286 350 287 351 288 352 289 353 290 354 291 355 292 356 293 357 294 358 295 359 296 360 Figure 2 (A) Genomic imprinting: Genomic imprinting is parent-of-origin (but genotype independent) gene expression. When males and females 297 361 produce gametes (ie, sperm or eggs) an epigenetic mark (eg, DNA methylation, which silences one of the parental alleles) is placed on the DNA to 298 indicate parent-of-origin. Regardless of sex of offspring imprinted genes affect growth and neural development differentially by parent-of-origin. 362 299 Once the child produces its own gametes the imprints are erased and new parent-of-origin marks are established. Imprinted genes are rare but have 363 300 profound effects on growth and neurodevelopment. (B) X-inactivation: X-inactivation is a process by which one parental copy of the X chromosome 364 301 in women is randomly deactivated. X-inactivation prevents females from having twice as much X chromosome gene production as males (which 365 302 only have one copy of the X chromosome). Once the X chromosome is deactivated, it remains silent throughout the cell’s lifetime. Compared with 366 303 transcriptionally active X chromosome, the inactive X has higher levels of DNA methylation, which is associated with gene silencing. One difference 367 304 between imprinting and X-inactivation is the former is not random process with respect to which parental allele is epigenetically silenced. 368 305 369 306 370 18 307 caused by point mutations in MECP2, which encodes a methyl phenotypes, suggesting that imprinting performs a regulatory 371 17 308 binding protein and is proposed to be a gene silencer. function during ontogeny (eg, regulation of other genes). 372 309 DNA methylation dysregulation is associated with auto- The current paper argues that imprinted genes are important 373 310 immune disease. For example, the Immunodeficiency for skeletal muscle development and their phenotypic effects 374 311 Centromeric Instability and Facial Anomalies (ICFA) syndrome reflect an underlying ancestral tug of war between parental 375 312 is caused by heterozygous mutations in DNMT3B. Individuals genomes over offspring growth and developmental trajectories. 376 313 with ICFA show DNA hypomethylation among Alu repeats. 377 314 Interestingly, despite patients with ICFA having normal global Imprinted genes and skeletal muscle gene networks: 378 315 DNA methylation profiles, key developmental regulatory and growth suppression and enhancement 379 316 immune function genes show loci-specific epigenetic Imprinted genes are genes whereby an epigenetic mark is laid 380 15 317 dysregulation. down during gametogenesis, indicating a key environmental 381 318 Whether it is cancers, neurological disorders or autoimmune source of information, the parental origin of a particular gene. 382 319 disease, DNA methylation and imprinted genes are emerging There are few imprinted genes in the human genome, but they 383 320 causal factors. Some imprinted genes are involved in multiple are often associated with growth, neural functioning and 384

Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 3 Review

385 behaviour. Parental antagonism theory19 is currently the best receptor expression or intracellular signalling). Since transcrip- 449 386 theory for the phenotypic effects of growth regulators (eg, the tion precedes translation, it is often at the level of the ‘transcrip- 450 387 paternal genome within offspring fosters growth at a cost to the tome’ that adaptations can be tracked at the molecular level. 451 388 maternal genome, while the maternal genome attempts to min- Subtle changes in gene transcription occur through epigenetic 452 389 imise these costs by suppressing growth). regulatory machinery. A variety of epigenetic mechanisms allow 453 390 Haig’s model19 proposes that imprinting evolved as a result for transcriptional activation and specification of cell identity, 454 391 of opposing fitness interests of parental genomes. For example, maintaining homoeostasis and responding to environmental 455 392 in polygamous species, patrigenes (genes expressed within off- conditions. These epigenetic mechanisms encompass DNA 456 393 spring inherited from the father) favour fetal growth at the methylation, post-translational histone modifications and 457 394 expense of depleting maternal resources and disadvantaging microRNA (figure 3). Much of the previous research on envir- 458 395 future offspring. Meanwhile, matrigenes (genes expressed onmental epigenomics involves nutrition; however, exercise 459 396 within offspring inherited from the mother) will oppose the physiology is coming to the forefront. There is evidence that 460 397 paternal effect and conserve resources to optimise inclusive DNA methylation can change due to short bouts of exercise (eg, 461 398 fitness of the mother and future offspring. Haig’s theory19 pre- exercising to exhaustion26) and longer, more sustained exercise 462 399 dicts that paternally expressed imprinted genes will often regimens (eg, 6 months of controlled walking27). For example, a 463 400 promote growth, while maternally expressed genes will have high-intensity interval walking regimen increased DNA methyla- 464 401 opposite effects to reduce costs on matrilineal inclusive fitness. tion of the proinflammatory gene ASC (apoptosis-associated 465 402 IGF2—an imprinted gene and paternally expressed in speck-like protein containing a caspase recruitment domain) 466 403 humans—regulates muscle development. IGF2 is upregulated among older adults nearly to the levels of healthy younger 467 404 early in MyoD-induced in myocyte differentiation and IGF2 adults.28 Among breast cancer sufferers, DNA methylation 468 405 inhibition leads to reduced expression of MyoD target genes, changes in L3MBTL1 (an imprinted and possible tumour sup- 469 406 which suggests that IGF2 is essential for amplifying and main- pressor gene) due to exercise (eg, brisk walking on treadmill) 470 407 taining MyoD efficacy.20 IGF2’s role as a paternally derived skel- has been demonstrated.29 471 408 etal muscle growth enhancer is consistent with the theoretical 472 409 orientation of this paper. Exercise epigenetics: research questions 473 410 Germane to sports medicine, some imprinting disorders affect What is the average effect of exercise on changes in DNA 474 411 muscle growth. For example, Angelman (AS) and Prader-Willi methylation and does age, research design, sample size, sex or 475 412 syndromes (PWS) are imprinting disorders affecting muscle tissue heterogeneity influence the size of the effect? Are 476 413 development and health. PWS is caused by an overexpression of imprinted genes implicated in skeletal muscle gene networks 477 414 maternal genes on chromosome 15, while AS is due to an overex- and exercise-associated DNA methylation changes in humans? 478 415 pression of paternal genes. Muscle biopsies of 11 PWS children Since imprinted genes regulate adiposity, energy expenditure 479 416 have been investigated using histochemical and morphometric 480 21 and glucose homoeostasis, it was hypothesised that imprinted 417 methods. The phenotypic abnormalities included (A) fibre size genes will be involved in human skeletal muscle gene networks 481 418 variation of both type 1 and 2 fibres, (B) type 2 fibre atrophy, (C) and targets of exercise-associated DNA methylation change. To 482 419 increased numbers of type 2C fibres and (D) decreased numbers test these hypotheses, gene ontology and meta-analytic method- 483 420 of type 2B fibre. This finding is consistent with the overexpres- ologies were utilised. 484 421 sion of maternal genes suppressing skeletal muscle growth. In 485 422 addition to their low muscle tone, PWS individuals experience 486 423 chronic hunger, potentially leading to overeating and obesity. METHODS 487 424 DNA methylation and imprinted loci influence muscle hyper- A human skeletal muscle gene network: testing the 488 425 trophy—extremely muscled hindquarters—in callipyge sheep.22 importance of imprinted genes 489 426 Hypomethylation of Clpg1 causes muscle hypertrophy, in part due It is hypothesised that human skeletal muscle growth is regu- 490 427 to the overexpression of Dlk1,22 a paternally expressed imprinted lated by imprinted genes. A gene ontology networking web 491 fi 30 31 428 gene associated with muscle precursor cell (myoblast) differenti- server called g:Pro ler was used to assess the functional 492 429 ation.23 DNA demethylation promotes skeletal myotube matur- involvement of imprinted genes for skeletal muscle gene net- 493 430 ation.24 Early experiments in the 1970s showed that DNA works. First, human skeletal muscle genes were determined 494 fi 32 33 431 methyltransferase inhibitors (eg, 5-azacytidine) induced transdif- using the Xavier laboratory gene enrichment pro ler 495 25 fi 432 fi ( gure 4). To test the hypothesis that imprinted genes are impli- 496 ferentiation of broblasts into myoblasts. More recently, in 30 31 24 fi 433 C2C12 culture, Hupkes et al noticed that on treatment with the cated in skeletal muscle gene networks, g:Pro ler was used. 497 434 methylation inhibitor (ie, 5-azacytidine), myotubes spontaneously Human imprinted genes (29 maternally expressed and 57 pater- 498 435 acquired repetitive membrane activity, intracellular calcium transi- nally expressed) were selected from http://www.geneimprint. 499 436 ents and contractility. Hupkes et al24 suggested that DNA methyla- com. Imprinted genes were combined with 26 skeletal muscle 500 fi 437 tion may pose an epigenetic barrier to C2C12 myotubes reaching genes ( gure 4). 501 438 maturity. However, when imprinted genes are involved in skeletal 502 439 muscle development, the so-called ‘DNA methylation barrier’ will Meta-analysis on exercise-associated DNA methylation 503 440 likely be parent-of-origin dependent. Beyond the distinct possibil- change 504 441 ity that imprinted genes coordinate mammalian skeletal muscle This meta-analysis was limited to English as no foreign language 505 442 development (eg, regulating skeletal muscle gene networks), it results were found. Only published papers measuring DNA 506 443 remains to be investigated whether the DNA methylation of methylation and exercise in humans were used. The following 507 444 imprinted loci are responsive to human exercise. search strings were entered into PubMed (1968 to 2 May 2014): 508 445 (1) ‘DNA methylation and exercise’; and (4) ‘DNA methylation 509 446 Exercise epigenetics and DNA methylation and physical activity (human-only)’. Search results finalised by 510 447 Traditionally, exercise biologists envision biological systems the author (figure 5). Both PRISMA (see online supplementary 511 448 changing by the regulation of protein synthesis (eg, alteration of file 1) and MOOSE (see online supplementary file 2) guidelines 512

4 Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 Review

513 577 514 578 515 579 516 580 517 581 518 582 519 583 520 584 521 585 522 586 523 587 524 588 525 589 526 590 527 591 528 592 529 593 530 594 531 595 532 596 533 597 534 598 535 599 536 600 537 Figure 3 MicroRNAs (miRNAs): A miRNA is a small (ie, approximately 19 to 25 nucleotides in length) non-coding RNA molecule functioning to 601 538 silence RNA and involved in the post-transcriptionally modification of gene expression. miRNA should not be confused with messenger RNA. 602 539 miRNAs are another important molecular epigenetic regulator. miRNAs can result in small, but important reductions in physiologically relevant gene 603 540 expression by blocking translation. Modified from Kelvin Song. 604 541 605 542 606 543 followed. One author did not respond to a direction of effect actin cytoskeleton to the extracellular matrix). Epsilon sarco- 607 544 query. Reference lists did not yield additional articles. glycan is more broadly expressed (ie, not just restricted to 608 545 striated muscle). Mutations in this gene are associated with 609 546 RESULTS the myoclonus-dystonia syndrome and it is imprinted (pref- 610 34 547 Imprinted genes and skeletal muscle gene networks erentially expressed from the paternal copy). 611 548 Table 1 shows the imprinted genes associated with skeletal 3. Insulin (INS): As seen in table 1, INS is part of a human 612 549 muscle gene networks. Six imprinted genes (ie, three maternally phenotype gene network called ‘motor delay’ (HP: 001270). 613 550 expressed genes RB1, MEG3 and UBE3A and three paternally The INS gene encodes for proinsulin (a prohormone precur- 614 551 expressed genes INS, PLAGL1 and SGCE) were revealed to be sor to insulin), which is post-translationally cleaved into 615 552 part of the gene networks of highly enriched skeletal muscle three peptides. Binding of insulin to the insulin receptor 616 553 loci. Considering the rarity of imprinted genes in the human, (INSR) stimulates glucose uptake. A multitude of mutant 617 554 this is biologically significant. Below is a description of each alleles with phenotypic effects have been identified. Notably, 618 555 imprinted gene involved. INS-IGF2, a read-through gene, aligns to the INS gene, 619 556 whereby INS is at the 5’ region and IGF2—an extremely 620 557 Paternally expressed genes linked to muscle-related well-studied growth regulatory imprinted gene—is at the 3’ 621 34 558 phenotypes region. 622 559 1. Pleomorphic adenoma gene-like 1 (PLAGL1): As seen in 623 560 table 1, PLAGL1 is part of two gene ontology networks: Maternally expressed genes linked to muscle-related 624 561 ‘muscle organ development’ (GO: 0007517) and ‘skeletal phenotypes 625 562 muscle tissue development’ (GO: 0007519). PLAGL1 1. Retinoblastoma 1 (RB1): As seen in table 1, RB1 is part of 626 563 encodes a zinc finger protein with transactivational and two gene ontology networks: ‘muscle organ development’ 627 564 DNA-binding functions. PLAGL1 has antiproliferative prop- (GO:0007517) and ‘skeletal muscle tissue development’ 628 565 erties making it a candidate for functioning as a tumour sup- (GO:0007519). RB1 encodes a protein that negative regu- 629 566 pressor gene. Overexpression of this gene during fetal lates cell cycle and was the first tumour suppressor gene dis- 630 567 development underlies transient neonatal diabetes mellitus covered. The encoded protein maintains overall chromatin 631 568 (TNDM). In most tissues (eg, skeletal muscle), PLAGL1 structure. Defects in RB1 cause childhood retinoblastoma 632 34 569 appears to be expressed from the paternal allele.34 (RB), bladder cancer and osteogenic sarcoma. 633 570 2. Sarcoglycan, epsilon (SGCE): As seen in table 1, SGCE is 2. Maternally expressed 3 non-protein coding (MEG3): As seen 634 571 part of a gene ontology network called ‘muscle organ devel- in table 1, MEG3 is part of two gene ontology networks: 635 572 opment’ (GO: 0007517) and a human phenotype gene ‘muscle organ development’ (GO: 0007517) and ‘skeletal 636 573 network called ‘abnormality of the musculature of the neck’ muscle tissue development’ (GO: 0007519). MEG3 is 637 574 (HP: 0011006). SGCE encodes the epsilon member of the expressed in many healthy tissues, but expression is lost in 638 575 sarcoglycan family (ie, transmembrane proteins which are multiple cancer cell lines of various tissue types. Notably, 639 576 part of the dystrophin-glycoprotein complex linking the MEG3 suppresses tumour cell proliferation in vitro and 640

Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 5 Review

641 Figure 4 Enrichment profile for 705 642 selected human skeletal muscle genes 706 643 (defined as genes that have high 707 644 enrichment scores within tissue and 708 645 between tissues). Embryonic stem cells, 709 646 t cells, b cells and myeloid gene 710 647 loadings have been removed from heat 711 32 33 648 map. 712 649 713 650 714 651 715 652 716 653 717 654 718 655 719 656 720 657 721 658 722 659 723 660 724 661 725 662 726 663 727 664 728 665 729 666 730 667 731 668 732 669 733 670 734 671 735 672 736 673 737 674 738 675 739 676 740 677 741 678 742 679 743 680 744 681 745 682 746 683 747 684 748 685 749 686 750 687 751 688 752 689 753 690 754 691 755 692 756 693 757 694 758 695 759 696 760 697 761 698 762 699 763 700 764 701 765 702 766 703 767 704 768

6 Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 Review

769 833 770 834 771 835 772 836 773 837 774 838 775 839 776 840 777 841 778 842 779 843 780 844 781 845 782 846 783 847 784 848 785 849 786 850 787 851 788 852 789 853 790 854 791 855 792 856 793 857 794 858 795 859 796 860 797 861 798 862 799 863 800 864 801 865 802 866 803 867 804 868 805 869 806 870 807 871 808 872 809 873 fl 810 Figure 5 PRISMA ow diagram for meta-analysis component of paper. Search terms used were ‘DNA methylation exercise’ or ‘DNA methylation 874 ’ 811 physical activity . Human in vivo studies only (May 2014). 875 812 876 813 877 814 interacts with tumour suppressor p53. Deleting MEG3 Two-hundred and eighty-seven genetic elements (out of 478) 878 815 enhances angiogenesis in vivo. Many studies show that showed significantly decreased DNA methylation after exercise 879 34 816 MEG3 is a long non-coding RNA tumour suppressor. (binomial test p<0.001). Online supplementary table S2 reports 880 817 3. Ubiquitin protein ligase E3A (UBE3A): As seen in table 1, the 478 genetic elements showing exercise-associated DNA 881 818 UBE3A is part of a human phenotype gene network called methylation change, 5 of which are imprinted genes (maternally 882 819 ‘motor delay’ (HP:001270). UBE3A encodes an E3 expressed (GRB10 , KCNQ1, MEG3) and paternally expressed 883 820 ubiquitin-protein ligase. This imprinted gene is maternally (PLAGL1, L3MBTL)). Despite appearing like a small percentage 884 821 expressed in the brain and most likely biallelically expressed of imprinted loci, this is much higher than the expected number 885 822 in skeletal muscle. Maternally inherited deletion of this gene of 1–2 (ie, assuming there are ∼90 imprinted genes in a human 886 823 causes a neurodevelopmental disorder AS which is charac- genome containing ∼22 300 genes): binomial test p<0.03. All 887 824 terised by severe motor and intellectual impairments, ataxia, imprinted genes showed a decrease in DNA methylation after 888 825 hypotonia, epilepsy and absence of speech. UBE3A’s protein exercise, except for GRB10 and KCNQ1 (adipose tissue only). 889 826 in part causes ubiquitination and proteolysis of tumour Table 2 provides location, ontology and growth-related effects 890 34 827 protein p53. for the imprinted genes showing exercise-associated DNA 891 828 methylation change. Unfortunately, it was not possible to deter- 892 829 Meta-analysis of exercise-associated DNA methylation mine if exercise-associated DNA methylation change in the 893 830 change imprinted loci was near differentially methylated regions 894 831 To determine if there is a directional bias in exercise-associated (DMRs) as each imprinted gene in table 2 has clinically relevant 895 832 methylation change, a binomial signed test was conducted. single-nucleotide polymorphisms. Among the five imprinted 896

Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 7 Review

897 961 Table 1 Imprinted genes networked with human skeletal muscle genes and network function 898 962 899 Function ID p Value Genes PAT MAT 963 900 964 Muscle organ development GO: 0007517 0.0005 PLAGL1, SGCE, RB1, MEG3, MYL1, NEB, ACTA1, PLAGL1, SGCE RB1, MEG3 901 965 TTN, MYLPF, RYR1, CASQ1 902 966 Skeletal muscle tissue development GO: 0007519 0.05 RB1, MEG3, PLAGL1, ACTA1, MYLPF, RYR1, CASQ1 PLAGL1 RB1, MEG3 903 967 Motor delay HP: 001270 0.03 UBE3A, INS, NDN, NEB, ACTA1, TTN, TNNT1, TPM3, RYR1 INS UBE3A 904 968 Abnormality of the musculature of the neck HP: 0011006 0.05 SGCE, NEB, ACTA1, TTN, TPM3 SGCE – 905 969 Networks determined by g:Profiler.30 31 Static url: http://tinyurl.com/imprinting-skm. 906 970 GO, gene ontology; HP, human phenotype; MAT, maternally expressed gene; PAT, paternally expressed gene. 907 971 908 972 909 973 genes revealed by the meta-analytic search in table 2, DNA was 95% CI of the mean 2.45 to 3.63). However, the reverse was 910 974 extracted from diverse tissues (ie, skeletal muscle, adipose and true for people under 40 years of age. Specifically, among 911 975 blood). younger people (people less than 40 years of age), the effect 912 976 size was significantly smaller (p<0.05) when DNA methylation 913 977 increased with exercise (0.90; 95% CI of the mean 0.83 to 914 Degree of exercise-associated DNA methylation change: 978 0.97) compared with when DNA methylation decreased with 915 effect of age and confounded factors 979 exercise (1.00, 95% CI of the mean 0.95 to 1.05). 916 The effect size of exercise on DNA methylation for the 478 980 To elucidate the possible function of this interaction among 917 genetic elements (387 were unique genes) across 16 different 981 older people, the genes showing increases and decreases after 918 publications and 1580 people—see online supplementary table 982 exercise were exposed to g:Profiler30 31 for ontology analysis by 919 S2)—is large (mean Cohen’s d=1.20±1.20; 95% CI of the 983 age. As seen in table 3, among older people the genes that 920 mean 1.10 to 1.31) and significantly different from a test value 984 increased in DNA methylation after exercise were associated 921 of zero (ie, no effect of exercise on DNA methylation): one- 985 with growth regulation (GO:0022603), and the genes that 922 sample t(477)=22.77, p<0.001. Analysis of covariance 986 become less methylated after exercise are targets of a putative 923 (ANCOVA) revealed that the effect size of exercise-associated 987 tumour suppressing microRNA miR-519B. Two imprinted genes 924 DNA methylation change was significantly greater for people 988 (L3MBTL1, PLAGL1)—both of which are tumour suppressors 925 over 40 years of age (Cohen’s d=2.89±1.97) compared with 989 —are associated with miR-519b’s microRNA network. 926 those under 40 years of age (Cohen’s d=0.90±.51): (F(1,471) 990 2 Among younger individuals, a microRNA-regulated gene 927 =197.26, p<0.001, partial η =0.30). In this model, the effect 991 network involved in stem cell activity was implicated. 928 of age was independent of research design (experimental 992 Specifically, as seen in table 4, genes that increased in DNA 929 designs’ larger effect size, p<0.001), sample size (smaller 993 methylation among younger people were part of a 930 studies’ larger effect sizes, p<0.001), sex (larger effect size 994 microRNA-regulated (hsa-miR-130b*) gene network that sup- 931 among females than males, p<0.03) and tissue specificity (larger 995 presses stem cell activity. Statistically significant (all p values 932 effect sizes in tissue with more cell types, p<0.001). The fact 996 <0.04) gene networks were uncovered for the genes that 933 that sample size and effect size are significantly and negatively 997 decreased in DNA methylation after exercise among younger 934 correlated suggests the presence of publication bias: r (477)= 998 people. Specifically, these gene networks are important for the 935 −0.11; p<0.02. Sample size has been included as a covariate in 999 biological processes of the extracellular matrix, skeletal muscle 936 analyses. 1000 and cartilage development (table 4). 937 Online supplementary table S2 shows that most of the genes 1001 938 decreased in DNA methylation percentage after exercise (238/ 1002 939 387 different genes). ANCOVA (sample size controlled) found a Tissue heterogeneity 1003 940 direction of change by age interaction: F(1,381)=20.14, Tissue type is a moderator of the degree of exercise-associated 1004 2 941 p<0.001, partial η =0.05. Specifically, among older people DNA methylation change. To further elucidate this apparent 1005 942 (people older than 40 years of age), the effect size was signifi- moderator, a one-way ANCOVA (controlling for sample size) 1006 943 cantly larger (p<0.05) when DNA methylation increased with was conducted and found significant: F(4,471)=137.03, 1007 2 944 exercise (3.85; 95% CI of the mean 3.32 to 4.38) compared p<0.001, partial η =0.54 (figure 6A). As seen in figure 6A, 1008 945 with when DNA methylation decreased with exercise (3.04, exercise-associated DNA methylation change was greater in 1009 946 1010 947 1011 948 Table 2 Imprinted genes that showed DNA methylation changes associated with exercise 1012 949 1013 Gene Ontology Expression Chromosome Start End Growth 950 1014 951 PLAGL1 Cell differentiation skeletal muscle; apoptosis Paternal 6 144 328 445 144 328 885 Enhancer 1015 952 GRB10 Insulin receptor pathway (negative regulation) Maternal skeletal muscle γ 7 50 850 662 50 851 107 Suppressor 1016 953 splice variant 1017 954 KCNQ1 Cardiovascular system development; negative regulation of insulin Maternal 11 2 465 914 2 870 339 Suppressor 1018 955 secretion; gene silencing 1019 956 MEG3 Negative regulation of angiogenesis; cell proliferation, positive Maternal 14 101 293 947 101 294 390 Suppressor 1020 regulation of skeletal muscle fibre development 957 1021 L3MBTL Regulation of megakaryocyte differentiation Paternal 20 42 142 508 42 142 820 Enhancer 958 1022 48 959 Function, parent-of-origin, differentially methylated region (DMR) location and hypothesised growth effects. Location of DMRs courtesy of Randy Jirtle. For KCNQ1, the DMR resides in 1023 intron 10. Genome Reference Consortium Human Build 37 (GRCh37). 960 1024

8 Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 Review

1025 1089 Table 3 Genes that showed exercise-associated increases (http://tinyurl.com/methincreased) and decreases (http://tinyurl.com/methdecreased) in 1026 1090 DNA methylation among older people 1027 1091 1028 Change Function ID p Value Genes 1092 1029 1093 Increased DNA methylation after Regulation of anatomical structure GO:0022603 0.018 CXCL10,DCC, PPP2R3A, RASA1,SULF1,TMEM100, 1030 1094 exercise morphogenesis WNT7A 1031 1095 Decreased DNA methylation after microRNA 519 inhibits cell proliferation MI:hsa-miR-519b-3p 0.016 GAB1,L3MBTL1, PLAGL1, WNK3,BCL2L11,CACNA2D3 1032 exercise and decreases tumour growth 1096 1033 1097 Gene networks courtesy g:Profiler.30 31 1034 GO, gene ontology; MI, computationally predicted microRNA target sites from the MicroCosm database (formerly miRBase). 1098 1035 1099 1036 1100 1037 1101 35 1038 blood samples compared with all tissue types (ie, buccal and H19 regulates muscle development. H19 is also known as 1102 1039 salvia, breast and adipose, skeletal muscle and gastric tumour). ASM for ‘adult skeletal muscle’. H19 is a negative regulator of 1103 35 1040 For three out of five tissue types, the effect sizes are large. For prenatal growth and bovine muscle development. H19 is 1104 1041 buccal cells and saliva and gastric tumours, the effect sizes were maternally expressed at high levels in embryonic and fetal skel- 1105 1042 medium. etal muscle and is located closely downstream of paternally 1106 1043 expressed IGF2 performing antagonistic functions (ie, growth 1107 1044 Exercise type enhancement). 1108 1045 Exercise type is a potential moderator of the degree of DNA Beyond H19, other imprinted genes studied in non-human 1109 1046 methylation change. To further elucidate the effect of exercise animals were absent from the human skeletal muscle gene net- 1110 1047 type on the degree of DNA methylation change, a one-way works, such as DLK1, which is a well-known imprinted gene 1111 1048 ANCOVA (controlling for sample size) was conducted and and paternally expressed muscle growth enhancer. In mouse 1112 2 1049 found significant: F(2,453)=126.11, p<0.001, partial η =0.36 skeletal muscle cultures, the genetic ablation of Delta-like 1113 1050 (figure 6B). As seen in figure 6B, DNA methylation change was 1homolog (Dlk1) causes reductions in skeletal muscle mass, in 1114 1051 significantly greater among those engaged in Tai Chi and part due to myofiber number loss and myosin heavy chain IIB 1115 36 1052 walking compared with those engaged in cycling. Walking and gene expression. GRB10 was another imprinted gene missing 1116 1053 Tai Chi were not significantly different from one another from g:Profiler’s human skeletal muscle gene networks. In mice, 1117 1054 (p=0.25). Regardless of the specific type of exercise, effect sizes Grb10 has a tissue-dependent imprinting status (ie, paternally 1118 1055 were large. expressed in the brain and maternally expressed in muscle, see 1119 37 38 1056 Garfield et al for its links to behaviour). Holt et al have 1120 1057 DISCUSSION found evidence that when Grb10 is deleted, hypermuscularity 1121 1058 Imprinted genes and skeletal muscle gene networks overgrowth occurs, suggesting that maternal gene expression 1122 1059 As predicted, imprinted genes were implicated in skeletal muscle functions to suppress muscle growth. The same pattern occurs 1123 39 1060 gene networks. Table 1 is consistent with the hypothesis that in human skeletal muscle. 1124 1061 parental genomes act simultaneously to suppress and promote Considering the rarity of imprinted genes, it is remarkable 1125 1062 skeletal muscle growth. The argument is that for skeletal muscle that six imprinted genes were discovered as part of skeletal 1126 1063 the maternal genes in table 1 (RB1, MEG3 and UBE3A) suppress muscle gene networks. Imprinted genes most likely repress, 1127 1064 growth, while paternal genes (INS, PLAGL1 and SGCE) maintain and induce muscle-specific transcription during myo- 1128 1065 perform antagonistic functions (ie, growth enhancement). genesis. Future studies should investigate epigenomic antagon- 1129 1066 Some imprinted genes expected based on non-human animal isms between paternally and maternally derived genes during 1130 1067 research were conspicuously absent from the human skeletal myogenesis, as opposed to assuming that decreased methylation 1131 1068 muscle gene networks. For example, in sheep, imprinted gene invariably leads to growth. Owing to the importance of 1132 1069 1133 1070 1134 1071 Table 4 Genes that showed exercise-associated increases and decreases (http://tinyurl.com/methdecreases-young) in DNA methylation among 1135 1072 younger people 1136 1073 1137 Change Function ID p Value Genes 1074 1138 1075 Increased DNA methylation Regulation of stem cell activity MI: hsa-miR-130b* 0.008 SNCG,NCOA6, MRPS26, SPINT4,HDACC3,ESR2,TSTD1,RGS6,FHL1, 1139 1076 after exercise ANO2 1140 1077 Increased DNA methylation Negative regulation of cell cycle GO:0045786 <0.02 FHL1,RB1,RPTOr,ZFHX3,CAB39,CDK9,HDAC3,MED25, PSMC5, 1141 1078 after exercise BRCA1, RUNX3 1142 1079 Decreased DNA methylation Extracellular structure GO:0043062, <0.02 LTBP4, COL15A1, COL18A1,COL4A1,LAMA2,NID1, NRXN1, 1143 1080 after exercise organisation; extracellular matrix GO:0030198 OLFML2A,PTK2,SFRP2,SULF2,COMP,FBLN2 1144 organisation 1081 1145 Decreased DNA methylation System development; skeletal GO:0048731, <0.04 PRKG1,ALDH1A2, ANK3,ANKS1A, BATF,CAMK2B, CASP8,CD74, 1082 1146 after exercise system development; cartilage GO:0001501, CENPF,CHL1, COL15A1, COL18A1,COL4A1,DKK3,EHD1,EYA1,FLG, 1083 development GO:0051216 HDAC9,HYAL2,IL7,IPMK,KCNQ1, KERA,LAMA2, LFNG,LILRB1, LMO4, 1147 1084 LY6D, MBNL1,MEF2A, MEPE,MITF,NFIB, ND1,NNMT, NPR2,NRXN1, 1148 1085 PLXND1, PTK2,RUNX1,SCIN,SERPINI1,SFRP2,SIX6,SLC35D1, SULF2, 1149 TRPV4, TTLL7,CHRDL2, COMP,PAX6, RUNX3,SIM1 1086 1150 30 31 1087 Gene networks courtesy g:Profiler. 1151 GO, gene ontology; MI, computationally predicted microRNA target sites from the MicroCosm database (formerly miRBase). 1088 1152

Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 9 Review

1153 1217 1154 1218 1155 1219 1156 1220 1157 1221 1158 1222 1159 1223 1160 1224 1161 1225 1162 1226 1163 1227 1164 1228 1165 1229 1166 1230 1167 1231 1168 1232 1169 1233 1170 1234 1171 1235 1172 1236 Figure 6 (A/B) Effect of tissue and exercise variation on exercise-associated DNA methylation change. Error bars represent Bonferroni-corrected 1173 1237 95% CIs. 1174 1238 1175 1239 1176 1240 1177 imprinted genes for skeletal muscle development (table 1), it DNA methylation were part of the cancer-suppressing 1241 1178 was hypothesised that imprinted genes would be implicated in microRNA gene network. This strongly suggests that exercise 1242 1179 exercise-associated DNA methylation changes. These findings may have a protective function among older people, perhaps 1243 1180 are discussed below. shielding them to a degree from the age-related diseases and 1244 1181 decline. 1245 1182 Exercise-associated DNA methylation change It is notable that two of the six genes that decreased in DNA 1246 1183 Human exercise has medium to large effect sizes on the DNA methylation among older people after exercise (ie, L3MBTL1, 1247 1184 methylation of genes extracted from different tissues across sex PLAGL1) are imprinted targets of tumour suppressor 1248 1185 and lifespan. The effect sizes are strong among older people microRNA miR-519b. Considering that imprinted genes (eg, 1249 1186 above and beyond the independent effects of research design, hypermethylation of tumour suppressors) are often associated 1250 1187 sample size, sex and tissue type. Publication bias is possible and with diverse forms of cancer, this is both a biologically and 1251 1188 difficult to rule out. However, the correlation between effect medically important finding. Exercise-associated decreases in 1252 1189 size and sample size was small (r=−0.11) and driven in part by L3MBTL1 DNA methylation are associated with decreased mor- 1253 29 1190 two studies, both of which were tightly controlled molecular tality among patients with breast cancer. Exercise-associated 1254 1191 exercise physiology experiments with small sample sizes and decreases in CACNA2D3 (also included in the microRNA 1255 1192 large effect size. The large effect size in these studies could be network reported here) DNA methylation may help reduce 1256 40 1193 due to methodological rigour. Nonetheless, sample size and gastric tumorigenesis. These findings suggest that, at least for 1257 1194 other moderating factors were included in the analyses, suggest- older people, exercise could have a protective effect against a 1258 1195 ing that exercise has substantial effects on DNA methylation. variety of cancers in both sexes. More experimental approaches 1259 1196 Since this work was limited to published studies from western (ie, a mouse transfection model) likewise suggest that miR-519b 1260 41 1197 cultures, file drawer artefacts are a potential source of bias. suppresses breast cancer. The ability of the miR-519b network 1261 42 1198 Future work will need to investigate cultural and geographical to inhibit cell proliferation and decrease tumour growth 1262 1199 effects, which could bias the findings. Given the plasticity of makes it potentially an epigenetically labile network for clinical 1263 1200 human development and population genetic variation, we may researchers interested in exercise. In contrast, among younger 1264 1201 expect regional variation in the size of exercise-associated DNA people (less than 40 years of age), a microRNA-regulated 1265 1202 methylation change. (hsa-miR-130b*) gene network, which functions to suppress 1266 1203 As expected, imprinted genes—a DNA methylation-based stem cell proliferation, increased in DNA methylation after 1267 1204 transgenerational epigenetic phenomenon—are responsive to exercise. 1268 1205 exercise exposure. Recall the skeletal muscle gene network ana- The fact that exercise-associated DNA methylation change 1269 1206 lyses revealed both maternally expressed (RB1, MEG3, UBE3A) was stronger in older compared with younger people indicates 1270 1207 and paternally expressed (PLAGL1, SGCE, INS) imprinted that exercise could alter an organism’s ‘epigenetic age’ by 1271 Q91208 genes likely played growth regulatory functions. Likewise, the warding off senescence. Why would exercise have more pro- 1272 1209¶ meta-analysis imprinted genes showed changes in DNA methyla- found effects on older compared with younger epigenomes? 1273 1210 tion associated with exercise. Specifically, maternally expressed One possibility is that as organisms age, epigenetic errors accu- 1274 1211 (GRB10 , KCNQ1, MEG3) and paternally expressed (L3MBTL1, mulate, and because there are more to correct or reset, older 1275 1212 PLAGL1) genes were represented in the meta-analysis. Sixty per people experience greater (and more positive) DNA methylation 1276 1213 cent of the 478 genetic elements uncovered in the meta-analysis change compared with younger people (who have experienced 1277 1214 showed decreased DNA methylation after physical exercise. fewer epigenetic errors). Since physically active grandparents 1278 1215 Among older people, the genes that increased in DNA methyla- were probably a characteristic of the majority of human 1279 1216 tion were involved in growth, while the genes that decreased in evolution, the pronounced age effect could also be an example 1280

10 Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 Review

1281 of an age-dependent adaptive epigenetic response to antagonis- TNDM locus, as well as at the following imprinted loci IGF2R, 1345 1282 tic pleiotropy. Antagonistic pleiotropy is the hypothesis that DIRAS3 and PEG1. Future work in humans and other animals 1346 1283 genes with multiple effects can be beneficial at younger ages and should be able to develop a more comprehensive list of 1347 1284 costly later in life. In the context of growth effects, older indivi- imprinted genes regulating skeletal muscle and associated with 1348 1285 duals’ tumour suppressor genes were becoming demethylated exercise. One reason that some imprinted genes may be missed 1349 1286 (possibly more expressed), but growth-related loci were becom- from these analyses could be due to the fact that imprinted 1350 1287 ing methylated (presumably less expressed). Thus, rebuilding genes are often involved in neural systems,18 which, unlike skel- 1351 1288 new tissue would become increasingly costly for older people etal muscle, cannot be extracted from healthy human 1352 1289 relative to younger people. To reiterate, stem cells and tissue participants. 1353 1290 regeneration adaptive gene complexes in younger people may Future work needs to be conducted to test whether or not 1354 1291 well be costly among older age groups due to cancer prolifer- imprinted DMRs48 are modified by exercise. Once again, given 1355 1292 ation and premiums on cancer suppression as we age. Please the relevance of imprinted genes for human cancers, one long- 1356 1293 note that reverse antagonistic pleiotropic effect may be operat- standing conundrum in medicine could be resolved. Specifically, 1357 1294 ing here, that is, growth suppression among younger individuals why does exercise treatment appear to reduce the incidence of 1358 1295 may be particularly costly (eg, inability to regenerate or develop cancers? One answer is that tumour suppressor genes are ‘reacti- 1359 1296 costly secondary sexual characteristics) relative to older postre- vated’ at promoters on long-term exercise treatment and the 1360 1297 productive individuals. The effect of age was independent of corresponding reduction in DNA methylation.29 Given these 1361 1298 tissue type effects. Specifically, it was found that possible medical benefits, future research should look at the 1362 1299 exercise-associated DNA methylation changes are greater in relationship between exercise stress and regulation of imprinted 1363 1300 blood compared with breast, adipose and skeletal muscle tissues. genes in order to understand more fully the underlying 1364 1301 This latter result is consistent with the hypothesis that epigenetic mechanisms. 1365 1302 dysregulation in more heterogeneous samples (eg, blood)—com- It is worth noting that exercise-associated DNA methylation 1366 1303 pared with a single tissue sample—may be a better proxy for changes for imprinted genes occurred only in studies where par- 1367 1304 the accumulation of environmental stress as we age. Beyond ticipants were exposed to longer term exercise (ie, 6 months) as 1368 1305 repeated exposure of physical exercise across the lifespan being opposed to short bouts of exercise. This interpretation should 1369 1306 important for a healthy epigenome, in utero maternal epigenetic be taken with caution as it is biased by the fact that fewer genes 1370 1307 effects are likely to be important. were studied in the acute study by Barrès et al26. Specifically, 1371 1308 There is recent work in mice showing that maternal exercise Barrès et al26 selected genes from a previous study of DNA 1372 1309 during pregnancy can reverse the deleterious epigenetic effects methylation in patients with type 2 diabetes, while the long- 1373 1310 of poor maternal diet on newborn pups’ Pgc-1a.43 The benefi- term exercise studies revealing the imprinted genes (see online 1374 1311 cial effects of maternal exercise during pregnancy on Pgc-1a supplementary table 2) screened many more genes using 1375 1312 DNA methylation levels in the next generation suggest that a Illumina’s Infinium HumanMethylation450 BeadChip (San 1376 1313 transgenerational mechanism exists for long-lasting epigenetic Diego, California, USA). 1377 1314 changes and is consistent with a fetal origins of disease 1378 1315 approach.44 There is no evidence of transgenerational DNA CONCLUSIONS 1379 1316 methylation effects of exercise in humans.44 However, it is note- Modern epigenomics helps to end nature-nurture debates over 1380 1317 worthy from the meta-analysis that GRB10 (γ isoform)—mater- health and disease. The genome is sensitive to the environment 1381 1318 nally expressed in human fetal skeletal muscle39—showed and environmental information is encoded into the epigenome 1382 1319 greater exercise-associated DNA methylation change among transgenerationally (eg, imprinted genes). Rather than argue 1383 1320 those without a type 2 diabetes family history. Relaxation of the which is more important, nature or nurture, we can now 1384 1321 growth-inhibiting effects of maternal genes could be dependent measure the interface between the two directly. Measuring the 1385 1322 on genetic, ecological conditions or fetal exposure to maternal interface between genes and the environment (eg, DNA methy- 1386 1323 exercise. For example, individuals from families exhibiting more lation) will have ramifications for health and human disease due 1387 1324 sedentary behaviour (ie, characterised by a history of type 2 dia- to an ageing and an increasingly physically inactive population. 1388 1325 betes) have less silencing of maternally expressed GRB10 in skel- Given the increasing amount of research from multiple inde- 1389 – – 1326 etal muscle. Such differentially epigenetic responses of GRB10 ’s pendent laboratories26 29 49 40 indicating that human exercise 1390 1327 γ isoform depending on a family history of type 2 diabetes has varied associations and effects on DNA methylation, it is a 1391 1328 would be extremely interesting if reliable. A powerful interface reasonable hypothesis that long-term exercise throughout the 1392 1329 between family history and offspring epigenetic could be sig- lifespan (or exposure during sensitive periods of in utero devel- 1393 1330 nalled during gestation. Specifically, maternal exercise during opment) could have profound effects on the epigenome.44 1394 1331 gestation could produce dose-dependent epigenetic responses in Future work should determine the optimal exercise types, 1395 1332 offspring.43 timing and duration for ameliorating epigenetic-based disease 1396 1333 Despite using two different methods (ie, g-Profiler gene ontol- outcomes. The strength of exercise-associated DNA methylation 1397 1334 ogy network analysis vs meta-analysis), multiple imprinted loci change could be an overestimate due to publication bias (eg, 1398 1335 appear to be missing. This raises the distinct possibility that add- unpublished studies not included). Genetic background could 1399 1336 itional imprinted genes will be found to be associated with affect the associations reported here and was not ruled out. 1400 1337 muscle adaptation and exercise adaptation in humans. For Future work should sample from monozygotic twins reared 1401 1338 example, in newborns with transient neonatal diabetes, the loss together and apart to elucidate the importance of genetic back- 1402 1339 of an epigenetic mark at the TNDM locus on chromosome ground. To rule out publication bias, a collection of unpublished 1403 1340 6q24 in the mesodermal lineage causes abdominal muscle hypo- exercise epigenetics papers will need to be collected and 1404 1341 plasia, the so-called prune belly sequence.45 46 When Laborie analysed. 1405 1342 et al47 investigated a family with prune belly that included one Uncovering epigenetic biomarkers is likely to be more clinic- 1406 1343 discordant set of MZ twins, the twin with prune belly (relative ally relevant than looking for ‘disease genes’ because epigenetic 1407 1344 to the normal co-twin) had extensive loss of methylation at the changes can be reversed and also since disease variants are 1408

Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 11 Review

1409 expected to be at low frequency due to the power of natural 4 Killian JK, Nolan CM, Wylie AA, et al. Divergent evolution in M6P/IGF2R imprinting 1473 – 1410 selection to remove deleterious genetic variants from a popula- from the Jurassic to the Quaternary. Hum Mol Genet 2001;10:1721 8. 1474 5 Abramowitz LK, Bartolomei MS. Genomic imprinting: recognition and marking of 1411 tion. Techniques are being developed to remove epigenetic 1475 imprinted loci. Curr Opin Genet Dev 2012;22:72–8. 1412 marks (ie, DNA demethylation), which can radically change 6 Morison IM, Paton CJ, Cleverley SD. The imprinted gene and parent-of-origin effect 1476 1413 disease phenotypes (eg, tumour progression). Exercise medicine database. Nucleic Acids Res 2001;29:275–6. 1477 1414 should work alongside clinical epigenetics to investigate how 7 Jirtle R. [Internet]. Gene Imprint; c2012. http://geneimprint.com/site/ 1478 genes-by-species 1415 exercise shapes the human epigenome. An applied goal would 1479 8 Eggermann T. Silver-Russell and Beckwith-Wiedemann syndromes: opposite (epi) 1416 be to adaptively decouple chronological and epigenetic age by mutations in 11p15 result in opposite clinical pictures. Horm Res 2009;2:30–5. 1480 1417 using exercise interventions. The analyses presented here 9 Kong A, Steinthorsdottir V, Masson G, et al. Parental origin of sequence variants 1481 1418 suggest that exercise-associated DNA methylation change associated with complex diseases. Nature 2009;462:868–74. 1482 1419 reduces epigenetic age (eg, cancer reduction29 40). Controlled 10 Buiting K. Prader-Willi syndrome and Angelman syndrome. Am J Med Genet C 1483 Semin Med Genet 2010;154C:365–76. 1420 1484 exercise interventions could help the ageing epigenome, espe- 11 Monk D, Arnaud P, Frost J, et al. Reciprocal imprinting of human GRB10 in 1421 cially among older hospitalised patients. For example, one study placental trophoblast and brain: evolutionary conservation of reversed allelic 1485 1422 of older people (aged 58–90 years) with cerebrovascular disease expression. Hum Mol Genet 2009;18:3066–74. 1486 1423 suggests that physical function improvements during hospitalisa- 12 Mackay DJ, Temple IK. Transient neonatal diabetes mellitus type 1. Am J Med 1487 59 Genet C Semin Med Genet 2010;154C:335–42. 1424 tion covary with subtelomeric methylation of long telomeres. 1488 13 Small KS, Hedman AK, Grundberg E, et al. Identification of an imprinted master 1425 If exercise alters the epigenome to reduce age-related disease trans regulator at the KLF14 locus related to multiple metabolic phenotypes. Nat 1489 1426 outcomes, it could be a relatively inexpensive treatment option Genet 2011;43:561–4. 1490 1427 within hospital environments. In conclusion, human studies in 14 Deaton AM, Bird A. CpG islands and the regulation of transcription. Genes Dev 1491 – 1428 exercise epigenetics are required not only because of the impact 2011;25:1010 22. 1492 15 Portela A, Esteller M. Epigenetic modifications and human disease. Nat Biotechnol 1429 on health of ageing populations, but also because key epigenetic 1493 2010;10:1057–68. 1430 elements (ie, imprinted genes) responsible for regulating adipos- 16 Cruickshanks HA, Tufarelli C. Isolation of cancer-specific chimeric transcripts 1494 1431 ity, energy expenditure, glucose homoeostasis and hunger are induced by hypomethylation of the LINE-1 antisense promoter. Genomics 1495 1432 differentially imprinted (or read differently) between mice and 2009;6:397–406. 1496 17 Hite KC, Adams VH, Hansen JC. Recent advances in MeCP2 structure and function. 1433 man. 1497 Biochem Cell Biol. 2009;87:219–27. 1434 18 Dent CL, Isles AR. Brain-expressed imprinted genes and adult behaviour: the 1498 1435 example of Nesp and Grb10. Mamm Genome 2014;25:87–93. 1499 1436 What is already known on this subject 19 Haig D. The kinship theory of genomic imprinting. Annu Rev Ecol Syst 1500 1437 2000;31:9–32. 1501 20 Wilson EM, Rotwein P. Control of MyoD function during initiation of muscle 1438 1502 fi differentiation by an autocrine signaling pathway activated by insulin-like growth 1439 Recent empirical studies suggest that physical exercise modi es 1503 fi factor-II. J Biol Chem 2006;281:29962–71. 1440 the human epigenome. Speci cally, DNA methylation—an 21 Sone S. Muscle histochemistry in the Prader-Willi syndrome. Brain Dev Jpn 1504 1441 important regulator of gene expression and correlate of diverse 1994;16:183–8. 1505 1442 disease states—is altered by physical activity. No systematic 22 Murphy SK, Nolan CM, Huang Z, et al. Callipyge mutation affects gene expression 1506 review has been conducted to elucidate these effects and in cis: a potential role for chromatin structure. Genome Res 2006;16:340–6. 1443 23 Bassel-Duby R, Olson EN. Signaling pathways in skeletal muscle remodeling. Annu 1507 1444 associations. Rev Biochem 2006;75:19–37. 1508 1445 24 Hupkes M, Jonsson MK, Scheenen WJ, et al. Epigenetics: DNA demethylation 1509 1446 promotes skeletal myotube maturation. FASEB J 2011;25:3861–72. 1510 25 Taylor SM, Jones PA. Multiple new phenotypes induced in 10T1/2 and 3T3 cells 1447 1511 treated with 5-Azacytidine. Cell 1979;17:771–9. 1448 What this study adds 26 Barrès R, Yan J, Egan B, et al. Acute exercise remodels promoter methylation in 1512 1449 human skeletal muscle. Cell Metab 2012;15:405–11. 1513 1450 27 Rönn T, Volkov P, Davegårdh C, et al. A six months exercise intervention influences 1514 This study isolates imprinted genes—known to be important for 1451 the genome-wide DNA methylation pattern in human adipose tissue. PLoS Genet 1515 health and disease—as important for muscle growth and 2013;9:e1003572. 1452 1516 clinical targets of exercise. Further, older people received 28 Nakajima K, Takeoka M, Mori M, et al. Exercise effects on methylation of ASC 1453 gene. Int J Sports Med 2010;31:671–5. 1517 significant benefits from exercise in terms of the adaptive 1454 29 Zeng H, Irwin ML, Lu L, et al. Physical activity and breast cancer survival: an 1518 epigenetic regulation of tumour suppressor genes. 1455 epigenetic link through reduced methylation of a tumor suppressor gene L3MBTL1. 1519 Breast Cancer Res Treat 2012;133:127–35. 1456 1520 30 Reimand J, Kull M, Peterson H, et al. g:Profiler a web-based toolset for functional 1457 profiling of gene lists from large-scale experiments. Nucleic Acids Res 2007;35: 1521 1458 Twitter Follow William Brown at @coevolve W193–200. 1522 1459 31 Reimand J, Arak T, Vilo J. g:Profiler—a web server for functional interpretation of 1523 Acknowledgements Thanks to Randy Jirtle, John Brewer, Thomas Chalk, Javaid – 1460 gene lists. Nucleic Acids Res 2011;39:W307 315. 1524 Kashani and Louis Metcalf for discussions of the ideas presented here. Alex 32 Benita Y, Cao Z, Giallourakis C, et al. Gene enrichment profiles reveal T-cell 1461 Forcellati (BSc), Ryan Lenton and Meishan Raal-Nunes (MSc) helped compile development, differentiation, and lineage-specific transcription factors including 1525 1462 information provided in online supplementary table S2. WMB’s research time was ZBTB25 as a novel NF-AT repressor. Blood 2010;115:5376–84. 1526 fi 1463 supported by the European Of ce of Aerospace Research & Development 33 Xavier RJ. [Internet]. Enrichment Profiler; c2013. http://xavierlab2.mgh.harvard.edu/ 1527 FA8655-10-1-3037. fi 1464 EnrichmentPro ler/ 1528 34 Xiang R, Ghanipoor-Samami M, Johns WH, et al. Maternal and paternal genomes 1465 Competing interests None. 1529 differentially affect myofibre characteristics and muscle weights of bovine fetuses at 1466 Provenance and peer review Not commissioned; externally peer reviewed. midgestation. PLoS ONE 2013;8:e53402. 1530 1467 35 RefSeq [Internet]. NCBI; c2014. http://www.ncbi.nlm.nih.gov/refseq/ 1531 1468 36 Waddell JN, Zhang P, Wen Y, et al. Dlk1 is necessary for proper skeletal muscle 1532 – Q51469 REFERENCES development and regeneration. PLoS ONE 2010;5:e15055 0. 1533 1 Waddington CH. The epigenotype. Endeavour 1942;1:18–20. 37 Garfield AS, Cowley M, Smith FM, et al. Distinct physiological and behavioural 1470¶ 1534 2 Haig D. The (dual) origin of epigenetics. Cold Spring Harb Symp Quant Biol functions for parental alleles of imprinted Grb10. Nature 2011;469:534–8. 1471 2004;69:67–70. 38 Holt LJ, Turner N, Mokbel N, et al. Grb10 regulates the development of fiber 1535 1472 3 Nanney DL. Epigenetic control systems. Proc Natl Acad Sci USA 1958;44:712–17. number in skeletal muscle. FASEB J 2012;26:3658–69. 1536

12 Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 Review

1537 39 Blagitko N, Mergenthaler S, Schulz U, et al. Human GRB10 is imprinted and 50 Bryan AD, Magnan RE, Hooper AE, et al. Physical activity and differential 1601 1538 expressed from the paternal and maternal allele in a highly tissue- and methylation of breast cancer genes assayed from saliva: a preliminary investigation. 1602 isoform-specific fashion. Hum Mol Genet 2000;9:1587–95. Ann Behav Med 2013;45:89–98. 1539 1603 40 Yuasa Y, Nagasaki H, Akiyama Y, et al. DNA methylation status is inversely 51 Coyle YM, Xie XJ, Lewis CM, et al. Role of physical activity in modulating breast 1540 correlated with green tea intake and physical activity in gastric cancer patients. Int J cancer risk as defined by APC and RASSF1A promoter hypermethylation in 1604 1541 Cancer 2009;124:2677–82. nonmalignant breast tissue. Cancer Epidemiol Biomarkers Prev 2007;16:192–6. 1605 1542 41 Navarro P, Ramkissoon SH, Shah S, et al. An indirect role for oncomir-519b in the 52 Ren H, Collins V, Clarke SJ, et al. Epigenetic changes in response to tai chi practice: 1606 1543 expression of truncated neurokinin-1 in breast cancer cells. Exp Cell Res a pilot investigation of DNA methylation marks. Evid Based Complement Altern 1607 2012;318:2604–15. Med 2012;2012:841810. 1544 42 Abdelmohsen K, Srikantan S, Tominaga K, et al. Growth inhibition by miR-519 via 53 White AJ, Sandler DP, Bolick SC, et al. Recreational and household physical activity 1608 1545 multiple p21-inducing pathways. Mol Cell Biol 2012;32:2530–48. at different time points and DNA global methylation. Eur J Cancer 1609 1546 43 Laker RC, Lillard TS, Okutsu M, et al. Exercise prevents maternal high-fat 2013;49:2199–206. 1610 1547 diet-induced hypermethylation of the Pgc-1a gene and age-dependent metabolic 54 Zhang FF, Cardarelli R, Carroll J, et al. Physical activity and global genomic DNA 1611 dysfunction in the offspring. Diabetes 2012;63:1605–11. methylation in a cancer-free population. Epigenetics 2011;6:293–9. 1548 1612 44 Chalk TEW, Brown WM. Exercise epigenetics and the fetal origins of disease. 55 Zhang FF, Santella RM, Wolff M, et al. White blood cell global methylation and IL-6 1549 Epigenomics 2014;6:469–72. promoter methylation in association with diet and lifestyle risk factors in a 1613 1550 45 Mackay DJ, Hahnemann JM, Boonen SE, et al. Epimutation of the TND locus and cancer-free population. Epigenetics 2012;7:606–14. 1614 1551 the Beckwith Wiedemann syndrome centromeric locus in individuals with transient 56 Luttropp K, Nordfors L, Ekströ TJ, et al. Physical activity is associated with 1615 – 1552 néonatal diabetes mellitus. Hum Genet 2006;119:179 84. decreased global DNA methylation in Swedish older individuals. Scand J Clin Lab 1616 46 Mackay DJG, Boonen SE, Clayton-Smith J, et al. A maternal hypomethylation Invest 2013;73:184–5. 1553 syndrome presenting as transient neonatal diabetes mellitus. Hum Genet 57 Gomes MV, Toffoli LV, Arruda DW, et al. Age-related changes in the global DNA 1617 1554 2006;120:262–9. methylation profile of leukocytes are linked to nutrition but are not associated with 1618 1555 47 Laborie LB, Mackay DJ, Temple IK, et al. DNA hypomethylation, transient neonatal the MTHFR C677T genotype or to functional capacities. PLoS ONE 2012;72: 1619 1556 diabetes, and prune belly sequence in one of two identical twins. Eur J Pediatr e52570. 1620 2010;169:207–13. 58 Morabia A, Zhang FF, Kappil MA, et al. Biologic and epigenetic impact of 1557 1621 48 Mancini-DiNardo D, Steele SJ, Ingram RS, et al. A differentially methylated region commuting to work by car or using public transportation: a case-control study. Prev 1558 within the gene Kcnq1 functions as an imprinted promoter and silencer. Hum Mol Med 2012;54:229–33. 1622 1559 Genet 2003;12:283–94. 59 Maeda T, Oyama J, Higuchi Y, et al. The physical ability of Japanese female elderly 1623 1560 49 Nitert MD, Dayeh T, Volkov P, et al. Impact of an exercise intervention on DNA with cerebrovascular disease correlates with the telomere length and subtelomeric 1624 fi 1561 methylation in skeletal muscle from rst-degree relatives of patients with type 2 methylation status in their peripheral blood leukocytes. Gerontology 1625 diabetes. Diabetes 2012;61:3322–32. 2011;57:137–43. 1562 1626 1563 1627 1564 1628 1565 1629 1566 1630 1567 1631 1568 1632 1569 1633 1570 1634 1571 1635 1572 1636 1573 1637 1574 1638 1575 1639 1576 1640 1577 1641 1578 1642 1579 1643 1580 1644 1581 1645 1582 1646 1583 1647 1584 1648 1585 1649 1586 1650 1587 1651 1588 1652 1589 1653 1590 1654 1591 1655 1592 1656 1593 1657 1594 1658 1595 1659 1596 1660 1597 1661 1598 1662 1599 1663 1600 1664

Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 13 Review

1665 1729 1666 1730 1667 1731 1668 1732 1669 1733 1670 1734 1671 1735 1672 1736 1673 1737 1674 1738 1675 1739 1676 1740 1677 1741 1678 1742 1679 1743 1680 1744 1681 1745 1682 1746 1683 1747 1684 1748 1685 1749 1686 1750 1687 1751 1688 1752 1689 1753 1690 1754 1691 1755 1692 1756 1693 1757 1694 1758 1695 1759 1696 1760 1697 1761 1698 1762 1699 1763 1700 1764 1701 1765 1702 1766 1703 1767 1704 1768 1705 1769 1706 1770 1707 1771 1708 1772 1709 1773 1710 1774 1711 1775 1712 1776 1713 1777 1714 1778 1715 1779 1716 1780 1717 1781 1718 1782 1719 1783 1720 1784 1721 1785 1722 1786 1723 1787 1724 1788 1725 1789 1726 1790 1727 ONLINE USE ONLY 1791 1728 1792

14 Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 Review

1793 1857 1794 1858 1795 1859 1796 1860 1797 1861 1798 1862 1799 1863 1800 1864 1801 1865 1802 1866 1803 1867 1804 1868 1805 1869 1806 1870 1807 1871 1808 1872 1809 1873 1810 1874 1811 1875 1812 1876 1813 1877 1814 1878 1815 1879 1816 1880 1817 1881 1818 1882 1819 1883 1820 1884 1821 1885 1822 1886 1823 1887 1824 1888 1825 1889 1826 1890 1827 1891 1828 1892 1829 1893 1830 1894 1831 1895 1832 1896 1833 1897 1834 1898 1835 1899 1836 1900 1837 1901 1838 1902 1839 1903 1840 1904 1841 1905 1842 1906 1843 1907 1844 1908 1845 1909 1846 1910 1847 1911 1848 1912 1849 1913 1850 1914 1851 1915 1852 1916 1853 1917 1854 1918 1855 ONLINE USE ONLY 1919 1856 1920

Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 15 Review

1921 1985 1922 1986 1923 1987 1924 1988 1925 1989 1926 1990 1927 1991 1928 1992 1929 1993 1930 1994 1931 1995 1932 1996 1933 1997 1934 1998 1935 1999 1936 2000 1937 2001 1938 2002 1939 2003 1940 2004 1941 2005 1942 2006 1943 2007 1944 2008 1945 2009 1946 2010 1947 2011 1948 2012 1949 2013 1950 2014 1951 2015 1952 2016 1953 2017 1954 2018 1955 2019 1956 2020 1957 2021 1958 2022 1959 2023 1960 2024 1961 2025 1962 2026 1963 2027 1964 2028 1965 2029 1966 2030 1967 2031 1968 2032 1969 2033 1970 2034 1971 2035 1972 2036 1973 2037 1974 2038 1975 2039 1976 2040 1977 2041 1978 2042 1979 2043 1980 2044 1981 2045 1982 2046 1983 ONLINE USE ONLY 2047 1984 2048

16 Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 Review

2049 2113 2050 2114 2051 2115 2052 2116 2053 2117 2054 2118 2055 2119 2056 2120 2057 2121 2058 2122 2059 2123 2060 2124 2061 2125 2062 2126 2063 2127 2064 2128 2065 2129 2066 2130 2067 2131 2068 2132 2069 2133 2070 2134 2071 2135 2072 2136 2073 2137 2074 2138 2075 2139 2076 2140 2077 2141 2078 2142 2079 2143 2080 2144 2081 2145 2082 2146 2083 2147 2084 2148 2085 2149 2086 2150 2087 2151 2088 2152 2089 2153 2090 2154 2091 2155 2092 2156 2093 2157 2094 2158 2095 2159 2096 2160 2097 2161 2098 2162 2099 2163 2100 2164 2101 2165 2102 2166 2103 2167 2104 2168 2105 2169 2106 2170 2107 2171 2108 2172 2109 2173 2110 2174 2111 ONLINE USE ONLY 2175 2112 2176

Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073 17 Review

2177 2241 2178 2242 2179 2243 2180 2244 2181 2245 2182 2246 2183 2247 2184 2248 2185 2249 2186 2250 2187 2251 2188 2252 2189 2253 2190 2254 2191 2255 2192 2256 2193 2257 2194 2258 2195 2259 2196 2260 2197 2261 2198 2262 2199 2263 2200 2264 2201 2265 2202 2266 2203 2267 2204 2268 2205 2269 2206 2270 2207 2271 2208 2272 2209 2273 2210 2274 2211 2275 2212 2276 2213 2277 2214 2278 2215 2279 2216 2280 2217 2281 2218 2282 2219 2283 2220 2284 2221 2285 2222 2286 2223 2287 2224 2288 2225 2289 2226 2290 2227 2291 2228 2292 2229 2293 2230 2294 2231 2295 2232 2296 2233 2297 2234 2298 2235 2299 2236 2300 2237 2301 2238 2302 2239 ONLINE USE ONLY 2303 2240 2304

18 Brown WM. Br J Sports Med 2015;0:1–18. doi:10.1136/bjsports-2014-094073