Aerobic Exercise Training Rejuvenates the Human Skeletal Muscle Methylome Ten years after Breast Cancer Treatment and Survival

Many cancer survivors suffer from impairments in skeletal muscle (SkM), both in terms of reduced mass and function. Interestingly, human SkM possesses an epigenetic memory of earlier stimuli, such as exercise. Long-term retention of epigenetic changes in SkM following cancer survival and/or exercise training have not yet been studied. We therefore investigated genome-wide DNA methylation (the methylome) in SkM following a 5-month, 3/week treadmill-based aerobic training intervention in breast cancer survivors 10-14 years after diagnosis and treatment. These results were compared to breast cancer survivors who remained untrained and to age-matched controls with no history of cancer, who undertook the same training intervention. SkM biopsies were obtained before (pre) and after (post) the 5-month training period and DNA methylation was analysed using InfiniumEPIC 850K Arrays. The breast cancer survivors displayed a significant retention of increased DNA methylation (i.e., hypermethylation) at a larger number of differentially methylated positions (DMPs) compared with healthy age-matched controls pre-training. Training in cancer survivors led to an exaggerated number of DMPs with a hypermethylated signature occurring at more random non-regulatory regions across the DNA compared with training in healthy age-matched controls. However, the opposite occurred in important gene regulatory regions, where training in cancer survivors elicited a considerable reduction in methylation (i.e., hypomethylation) in 99% of the DMPs located specifically in CpG islands within promoter regions. Importantly, training was able to reverse the hypermethylation identified in cancer survivors back towards a hypomethylated signature that was observed pre-training in healthy age-matched controls at 300 (out of 881) of these island and promoter associated CpG sites. Pathway enrichment analysis identified that the training in cancer survivors evoked this predominantly hypomethylated signature in pathways associated with: Cell cycle, DNA replication and repair, transcription, translation, mTOR signalling and proteosome. Differentially methylated region (DMR) analysis also identified genes: BAG1, BTG2, CHP1, KIFC1, MKL2, MTR, PEX11B, POLD2, S100A6, SNORD104 and SPG7 as hypermethylated in breast cancer survivors with training reversing these CpG island / promoter associated DMRs towards a hypomethylated signature. Training also elicited a largely different epigenetic response in healthy individuals than that observed in cancer survivors, with very few overlapping changes. Only one gene, SIRT2, was identified as having altered methylation in cancer survivors at baseline as well as after training in both the cancer survivors and healthy controls. In conclusion, human SkM muscle retains a hypermethylated signature for as long as 10-14 years after breast cancer treatment and survival. Importantly, 5 months of aerobic training rejuvenated the SkM methylome towards signatures identified in healthy age-matched individuals in gene regulatory regions. Ethical approval: This study is part of the ongoing CAUSE-trial (Cardiovascular Survivor Exercise) initiated by the Norwegian School of Sport Sciences (NSSS) and Oslo University Hospital. The Regional Committees for Medical and Health Research Ethics (REK) approved the study (reference number 28930; 2019/1318), and the study was pre-registered in clinicaltrials.gov (NCT04307407). Participants and Experimental Design: The CAUSE-trial is a two-armed randomised controlled exercise training trial, where breast cancer survivors are randomised to either an exercise group or a control group (usual care), that also includes third arm of age-matched women with no prior history of cancer undertaking the same exercise training program. Breast cancer survivors diagnosed, with stage II-III HER2 negative breast cancer at the age of 60 years or less, between 2008 and 2012 were identified by the Cancer Registry of Norway. The study included survivors that had received anthracycline-based chemotherapy (Epirubicin), were currently exercising less than 90 minutes at moderate to high intensity per week, and that were living within driving distance from NSSS. Major exclusion criteria for the study were stage IV breast cancer diagnosis, adjuvant Trastuzumab-treatment, recurrent breast cancer or presence of secondary cancers, previous major cardiac surgery, pacemaker, chronic atrial fibrillation or any recent or uncontrolled cardiovascular disease. Participants for the healthy age-matched control group were recruited via advertisements. The participants were matched for age with the breast cancer survivors and could not have a history of any cancer and otherwise had to adhere to the same eligibility criteria as the breast cancer survivors (except for the breast cancer diagnosis). The breast cancer survivors were block randomised with a 1:1 allocation ratio to either an aerobic exercise training group (Cancer Trained) or to a control group (Cancer Untrained) for 5 months (20 weeks). The training intervention is described below. Participants in the healthy age-matched group did not undergo randomization, but all underwent the same exercise protocol as the cancer trained group (Healthy Age-Matched Trained). All participant groups had SkM biopsies taken and underwent other study-related assessments pre and post training. SkM biopsy methods are detailed below. Aerobic Exercise Training: Both the cancer survivors training group and the healthy age-matched group underwent five months of 3 times per week treadmill-based endurance training, aimed at increasing V̇O2peak. The program followed a non-linear periodization model, typically with one hard (i.e., 80-87% of HRpeak), one moderate (i.e., 70-80% of HRpeak) and one easy (i.e., 60-70% of HRpeak) session per week. Initially, all sessions were continuous in duration, but from week 5, the hard session was an interval session, with 3-5 intervals each lasting 4 minutes (i.e., 90-97% of HRpeak). After week 10, we introduced 8-minute intervals for the moderate session (i.e., 85-91% of HRpeak), progressing from 2-4 intervals throughout the intervention. Skeletal Muscle Biopsies: The biopsy procedure was conducted under local anaesthesia (Xylocaine with adrenaline, 10 mg/ml lidocaine + 5 µg/ml adrenaline, AstraZeneca, London, UK) and muscle tissue was obtained using a modified Bergström technique with suction. The tissue was quickly rinsed in physiological saline, before fat, connective tissue and blood were removed and discarded. Subsequently, the samples were quickly frozen in isopentane and cooled on dry ice or liquid nitrogen, before being transferred and stored at −80°C for later isolation of DNA. All biopsies from both pre and post training time points were taken at rest, in the morning after an overnight fast from the lateral portion of vastus lateralis muscle. Post training samples were obtained at least 72 hrs after the end of the training intervention. Pre and post biopsies were taken from the left leg in all participants. Cardiopulmonary Exercise Test and Physical Activity levels: Participants body height and mass were recorded using a stadiometer and a scale (SECA 213, Hamburg, Germany). Cardiorespiratory fitness (V̇O2peak) was assessed by a treadmill-based, symptom-limited cardiopulmonary exercise test, using a modified Balke protocol. After familiarisation with the treadmill, participants were encouraged to walk at gradually increasing exercise loads (increased inclination or speed) until voluntary exhaustion. Continuous expired air was measured breath-by-breath, using a gas and volume calibrated metabolic chart (Oxycon Pro, Jarger GmbH, Hoechberg, Germany). Objectively measured physical activity levels were recorded for seven consecutive days, using the ActiGraph™ model GT3X+ (ActiGraph LLC, Pensacola FL, USA). The average minutes spent in light and moderate-to-vigorous physical activity, using cut-off values used previously [1], with light physical activity defined as 100-1951 counts per minute and moderate-to-vigorous physical activity >1952 counts per minute. Baseline characteristics of V̇O2peak and physical activity levels for all participants can be found below. There was no significant difference in V̇O2peak or physical activity levels between the cancer groups and the healthy age-matched group prior to the start of the exercise program. Tissue Homogenization, DNA isolation, Bisulfite Conversion and Genome-Wide DNA Methylation: DNA was isolated from SkM tissue derived from a randomly selected subpopulation within each group of the larger trial (n=5 cancer trained, n=5 cancer untrained and n=6 healthy age-matched trained) at both pre and post training. These subpopulations baseline characteristics can be found here in the order of: Cancer Trained (n=5), Cancer Untrained (n=5), Healthy Age-Matched Trained (n=6), Cancer Survivors vs. Healthy Age-matched (p-value) respectively for: Age (years; SD) 60.4 (5.9), 62.0 (5.5), 58.3 (4.8), p = 0.307; Body Height (cm; SD) 167.5 (5,9), 168.1 (5.0), 169.0 (4.8), p = 0.660; Body Mass (Kg; SD) 80.5 (7.9), 74.1 (8.6), 80.0 (16.5), p = 0.670; Body Mass Index (SD) 28.8 (3.8), 26.4 (4.3), 28.0 (5.8), p = 0.861; V̇O2peak (ml/kg/min; SD) 27.2 (3.9),30.9 (6.7), 30.7 (7.7), p = 0.630; Physical activity (PA) levels (minutes; SD) for Light PA 154 (44.9),154 (38.4), 162.9 (28.3), p = 0.652; Moderate to Vigorous PA 43.3 (34.6), 41.8 (16.4), 56.2 (31.6), p = 0.362. Baseline characteristics were well balanced between the groups. There was no difference in age, height, BMI, V̇O2peak or physical activity levels between the cancer groups and the healthy age-matched group. Muscle samples were homogenized for 45 seconds at 6,000 rpm × 3 (5 min on ice in-between intervals) in lysis buffer (180 µl buffer ATL with 20 µl proteinase K) provided in the DNeasy spin column kit (Qiagen, UK) using a Roche Magnalyser instrument and homogenization tubes containing ceramic beads (Roche, UK). The DNA was then bisulfite converted using the EZ DNA Methylation Kit (Zymo Research, CA, United States) as per the manufacturer’s instructions. Infinium MethylationEPIC BeadChip Array All DNA methylation experiments were performed in accordance with Illumina manufacturer instructions for the Infinium Methylation EPIC BeadChip Array. Methods for the amplification, fragmentation, precipitation and resuspension of amplified DNA, hybridisation to EPIC beadchip, extension and staining of the bisulfite converted DNA (BCD) can be found in detail in our open access methods paper [2, 3]. EPIC BeadChips were imaged using the Illumina iScan System (Illumina, United States). DNA methylome analysis, differentially methylated Positions (DMPs), pathway enrichment analysis (KEGG pathways) and differentially methylated region (DMR) analysis: Following MethylationEPIC BeadChip arrays, raw .IDAT files were processed using Partek Genomics Suite V.7 (Partek Inc. Missouri, USA) and annotated using the MethylationEPIC_v-1-0_B4 manifest file. The mean detection p-value for all samples was 0.0004, which was well below the recommended 0.01 in the Oshlack workflow [4]. We also examined the raw intensities/signals of the probes, that demonstrated an average median methylated and unmethylated signal of 11.77 and 11.69, respectively, with an average of 11.73, which is recommended to be above 11.5 [4]. The difference between the average median methylated and average median unmethylated signal was 0.07, well below the recommended difference of less than 0.5 [4]. Note: Upon import of the data into Partek Genomics Suite we filtered out probes located in known single nucleotide polymorphisms (SNPs) and any known cross-reactive probes using previously defined SNP and cross-reactive probe lists from EPIC BeadChip 850K validation studies [5]. Although the average detection p-value for each sample across all probes was very low (on average 0.0004) we also excluded any individual probes with a detection p-value that was above 0.01 as recommended previously [4]. Furthermore, due to an all-female cohort of breast cancer survivors and age-matched healthy controls, we filtered out any probes on the male only Y chromosome (therefore the analysis included all probes on the X chromosome only). Out of a total of 865,860 probes in the EPIC array, removal of known SNPs, cross-reactive probes, those with a detection p-value above 0.01 and those on the Y chromosome resulted in 808,804 probes being taken forward for downstream analysis. Following this, background normalisation was performed via functional normalisation (with noob background correction) as previously described [6]. After functional normalisation, we also undertook quality control procedures via principal component analysis (PCA). One sample (Sample No.13_205624860016_R05C01_Cancer Untrained_Pre_Participant 10) was removed due to a larger variation than that expected within that condition (variation defined as values above 2.2 standard deviations (SDs) for that condition – depicted by ellipsoids in the PCA plots. Following normalisation and quality control procedures, we undertook differentially methylated position (DMP) analysis by converting β-values to M-values (M-value = log2(β / (1 - β)), as M-values show distributions that are more statistically valid for the differential analysis of methylation levels [7]. We then performed a two-way ANOVA for group/condition (cancer trained, cancer untrained, healthy age-matched trained) and time (pre and post), with planned contrast/pairwise comparisons of: Cancer Trained-Pre vs. Healthy Trained-Pre (to investigate the effect of Cancer on the methylome named - Cancer Alone), Cancer Trained-Post vs. Cancer Trained-Pre (to investigate the impact of training on the methylome of breast cancer survivors named - Training in Cancer), Healthy Age-Matched Trained-Post vs. Heathy Age-Matched-Pre (to investigate the impact of training on the methylome in age-matched healthy controls named- Training in Healthy), Cancer Untrained-Post vs. Cancer Untrained-Pre (to help assess if any methylation changes occur over the time of the intervention period in cancer survivors, named - Experimental Time), and finally Cancer Trained Pre vs. Cancer Untrained Pre (to assess if there were any differences in methylation between the cancer groups at baseline, named - Between Cancer Groups at Baseline). For initial discovery of CpG sites that were deemed statistically significant, DMPs with an unadjusted P value of ≤ 0.01 were accepted for downstream analysis (Kyoto Encyclopedia of Genes and Genomes/KEGG pathway, differentially methylated region/DMR analysis - see below). We then undertook CpG enrichment analysis on these DMPs within KEGG pathways [8-10] using Partek Genomics Suite and Partek Pathway software. DMR analysis was performed to identify where several CpGs were differentially methylated within a short chromosomal locations/regions, undertaken using the Bioconductor package DMRcate (DOI: 10.18129/B9.bioc.DMRcate). IMPORTANT NOTE for Methylation Array files: For submission of data to GEO it is recommended to upload all normalized data for ALL samples and ALL probes. Therefore, for the matrix signal intensities we include normalized data for all probes and all samples as the GEO submission. Therefore, it’s important to note, (as described in detail in the methods above), sample no. 13 (Sample No.13_205624860016_R05C01_Cancer Untrained_Pre_Participant 10) was removed from downstream analysis. Also, for any downstream analysis (DMP, DMR, GO / KEGG analysis etc.) we removed cross-reactive probes, probes on SNPs, probes with a detection p value of above 0.01 and due to an all-female cohort, we also removed any probes on the male only Y chromosome (therefore including all probes on the X chromosome only). We also removed 378 DMPs from downstream analysis that included those that were altered over the experimental time period (5 months) in the cancer untrained group as well as those that were significantly different between cancer groups at baseline (specific details can be found in the final publication). REFERENCES: 1. Ekelund U, Tarp J, Fagerland MW, Johannessen JS, Hansen BH, Jefferis BJ, et al. Joint associations of accelero-meter measured physical activity and sedentary time with all-cause mortality: a harmonised meta-analysis in more than 44 000 middle-aged and older individuals. Br J Sports Med. 2020; 54(24):1499-1506. 2. Seaborne RA, Strauss J, Cocks M, Shepherd S, O'Brien TD, Someren KAV, et al. Methylome of human skeletal muscle after acute & chronic resistance exercise training, detraining & retraining. Scientific Data (Nature). 2018; 5:180213. 3. Seaborne RA, Strauss J, Cocks M, Shepherd S, O'Brien TD, van Someren KA, et al. Human Skeletal Muscle Possesses an Epigenetic Memory of Hypertrophy. Scientific Reports (Nature) 2018; 8(1):1898. 4. Maksimovic J, Phipson B, and Oshlack A. A cross-package Bioconductor workflow for analysing methylation array data [version 1; referees: 3 approved, 1 approved with reservations]. F1000Research. 2016; 5(1281). 5. Pidsley R, Zotenko E, Peters TJ, Lawrence MG, Risbridger GP, Molloy P, et al. Critical evaluation of the Illumina MethylationEPIC BeadChip microarray for whole-genome DNA methylation profiling. Genome Biology. 2016; 17(1):208. 6. Maksimovic J, Gordon L, and Oshlack A. SWAN: Subset-quantile within array normalization for illumina infinium HumanMethylation450 BeadChips. Genome Biol. 2012; 13(6):R44. 7. Du P, Zhang X, Huang C-C, Jafari N, Kibbe WA, Hou L, et al. Comparison of Beta-value and M-value methods for quantifying methylation levels by microarray analysis. BMC Bioinformatics. 2010; 11(1):587. 8. Kanehisa M and Goto S. KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res. 2000; 28(1):27-30. 9. Kanehisa M, Furumichi M, Tanabe M, Sato Y, and Morishima K. KEGG: new perspectives on genomes, pathways, diseases and drugs. Nucleic Acids Res. 2017; 45(D1):D353-d361. 10. Kanehisa M, Sato Y, Kawashima M, Furumichi M, and Tanabe M. KEGG as a reference resource for gene and protein annotation. Nucleic Acids Res. 2016; 44(D1):D457-62.