Does muscle remember an earlier strength block?
In 8 untrained men some DNA marks of a 7-week block outlasted 7 weeks off; a second block built more lean mass (7 finished). One group, no control: order, skill and load are not separated.
Main caveat: No control. One group, one order: the reloading block always followed the loading block.
Seaborne, Sharples and colleagues asked whether adult human muscle keeps an epigenetic memory of earlier growth. Eight healthy, previously untrained men consented; one withdrew at week 17 of 21, so lean mass, strength and the last biopsies are from 7 (the study is often quoted as 8).
After a familiarisation week and one acute bout, the men did 7 weeks of supervised whole-body resistance training three times a week (loading), 7 weeks without resistance training (unloading) and the same 7-week programme again (reloading).
Lower-limb lean mass rose 6.5% with loading (19.47 to 20.74 kg), fell 4.6% with unloading to 19.83 kg (not different from baseline) and after reloading was 12.4% above baseline (21.85 kg); a paired comparison that allowed for the lean mass kept through unloading still found the second gain larger than the first (P = 0.022).
Isometric knee-extension torque rose 9.3% with loading (not significant), fell 8.3% with unloading and was 18% above baseline after reloading (P = 0.015).
Across the methylome, more sites were hypomethylated than hypermethylated after loading (9,153 vs 8,212 at P < 0.05), a similar number of sites were hypomethylated against baseline after unloading (8,891; the count does not say they were the same sites) and about twice as many after reloading (18,816).
Four genes (AXIN1, GRIK2, CAMK4, TRAF1) were hypomethylated after loading and stayed lower through unloading and reloading — the authors' "memory" profile, although the unloading and reloading values were not significant — while their expression rose stepwise; another cluster (UBR5, RPL35a, SETD3, PLA2G16 and others) returned to baseline in unloading and responded most on reloading.
The authors conclude that human muscle possesses an epigenetic memory of earlier hypertrophy. Our reading: the first human measurement that some methylation changes from a training block outlast the loss of the muscle it built.
The design cannot say that the earlier block caused the larger second gain: every man did the second block after the first, having learned the lifts and lifting a higher average load, and no group did the second block alone.
A human strength trial with an untrained control limb found no faster regain (REF-1360); a later interval-training study with the same senior author found marks kept and no physiological memory (REF-1427).
Protocol
Design: single group, within-subject, 21 weeks plus a familiarisation week; NHS West Midlands Black Country Research Ethics Committee approval (16/WM/0103). No control group and no untrained limb.
Participants: n = 8 healthy, previously untrained men (27.6 ± 2.4 yr, 82.5 ± 6.0 kg, 178.1 ± 2.8 cm; means ± SEM), physician-screened; one withdrew at week 17 of 21 for unrelated reasons, so methylation n = 8 at baseline, after the acute bout, loading and unloading and n = 7 after reloading; lean mass and torque n = 7.
Familiarisation and starting loads: one week with no or low load; because the men were unaccustomed to lifting, the starting load for 4 sets of 8-10 repetitions was set individually from technique, range of motion and their own feedback.
Acute bout: one lower-limb session 3-4 days later (total load 8,223 ± 284 kg), biopsy 30 min after.
Loading (7 weeks): 60-min sessions Monday, Wednesday and Friday; Monday and Friday lower limb (behind-head squat, leg press, leg extension, leg curl, Nordic curls, weighted lunges, calf raises), Wednesday upper body (bench press, shoulder press, lat pull-down, dumbbell row, triceps cable extension); 4 sets of 10 repetitions, about 90-120 s between sets and 3 min between exercises; load raised about 5-10% once 3 sets of 10 were completed (a progressive volume model); a trainer at every session.
Unloading (7 weeks): told to return to habitual pre-intervention exercise and do no resistance training; compliance by regular verbal contact, not measured.
Reloading (7 weeks): the same programme (same exercises on the same days, sets, repetitions and rests). Weekly volume load rose through both blocks; the average load was significantly higher in reloading than in loading (P = 0.043).
Measures: whole-body DEXA (Hologic; one operator) for lower-limb lean mass at baseline and after each phase; isometric knee-extension maximal voluntary contraction (Biodex) for peak torque; conchotome biopsies of the vastus lateralis at baseline, 30 min after the acute bout and after each phase, away from earlier incisions; Infinium MethylationEPIC array, SWAN normalisation, two outlier samples removed after principal component analysis (which ones is not stated in the main text; Supplementary Figure 1 was not read); rt-qRT-PCR for 48 genes chosen from the most frequently changed sites.
Analysis: lean mass and torque by one-way repeated-measures ANOVA (n = 7) and a paired t-test of the reloading gain (from unloading) against the loading gain (from baseline); methylation by ANOVA in Partek Genomics Suite (the Methods mention a Bonferroni correction without saying what it covered; the counts of changed sites use P < 0.05 and the clustering an unadjusted P < 0.05); gene-expression clusters by MANOVA and ANOVA; Pearson correlations of expression with lean-mass change; mean ± SEM.
Funding and interests: "funded by a PhD studentship for Robert A. Seaborne by the Doctoral Training Alliance UK/LJMU/Keele University awarded via Adam P. Sharples (PI). Genome-wide methylation/gene expression studies were funded by a GlaxoSmithKline grant awarded to Adam P. Sharples (PI)"; "The authors declare that they have no competing interests". Sharples conceived the experiments and, with Seaborne, designed them, analysed all the data and wrote the manuscript.
Key Results
Lower-limb lean mass, torque and methylation across the phases (mean ± SEM; lean mass and torque n = 7):
| Measure | After loading (7 wk) | After unloading (7 wk) | After reloading (7 wk) |
|---|---|---|---|
| Lower-limb lean mass (baseline 19.47 ± 1.01 kg) | 20.74 ± 1.11 kg, +6.5% (± 1.0%) vs baseline, P = 0.013 | 19.83 ± 1.06 kg, −4.6% ± 0.6% vs loading (P = 0.02); not different from baseline | 21.85 ± 2.78 kg, +12.4 ± 1.3% vs baseline, P = 0.001; +5.9 ± 1.0% vs loading (P = 0.005) |
| Isometric knee-extension peak torque (baseline 296.2 ± 22.1 Nm) | 324.5 ± 27.3 Nm, +9.3 ± 3.5% (not significant) | −8.3 ± 2.8% vs loading | 349.6 ± 27.7 Nm, +18 ± 3.6% vs baseline (P = 0.015) |
| CpG sites changed vs baseline at P < 0.05 (hypo / hyper) | 17,365 (9,153 / 8,212) | 8,891 / 8,638 | 27,155 (18,816 / 8,339) |
Second block against the first, allowing for what was kept: the gain from unloading to reloading was larger than the gain from baseline to loading (paired t-test, P = 0.022).From the group means that is about +10.2% (19.83 to 21.85 kg) against +6.5% (19.47 to 20.74 kg) (our arithmetic); the headline 12.4% is measured from the original baseline and so includes the lean mass kept through unloading (about 1.8%, our arithmetic). For torque the main text reports changes only against baseline; no test of the second gain against the first is given there.
One acute session: 30 min after the first bout, 10,284 sites were hypomethylated and 7,600 hypermethylated (17,884 at P < 0.05).
"Retained" profile (AXIN1, GRIK2, CAMK4, TRAF1): methylation 0.78 ± 0.09 of baseline after loading (P = 0.036), 0.84 ± 0.09 after unloading and 0.83 ± 0.05 after reloading, "albeit non-significantly" for the last two; expression 1.19 ± 0.08 after loading, 1.58 ± 0.13 after unloading (P = 0.001 vs baseline) and 1.79 ± 0.09 after reloading (P < 0.0001) — expression rose most during the 7 weeks without training. These genes' expression did not correlate with lean mass.
Larger second response without retention (UBR5, RPL35a, SETD3, PLA2G16 and others): UBR5 expression 1.65 ± 0.4 after loading (not significant), 0.82 ± 0.27 after unloading and 1.84 ± 0.5 after reloading (P = 0.035 vs unloading, P = 0.07 vs baseline); UBR5 CpG methylation 0.87 ± 0.03 after loading (not significant), 1.27 ± 0.02 after unloading (P = 0.013, above baseline, not retained low) and 0.78 ± 0.02 after reloading (P = 0.039).PLA2G16 expression 1.60 ± 0.18 after reloading (P = 0.026 vs baseline; P = 0.067 vs loading).
Expression and lean mass: across conditions, expression of RPL35a (R = 0.6, P = 0.014), UBR5 (R = 0.58, P = 0.018), SETD3 (R = 0.57, P = 0.013), PLA2G16 (R = 0.55, P = 0.027) and HEG1 (R = 0.53, P = 0.05) correlated with the percentage change in lean leg mass.
Genes changed by one session: BICC1, GRIK2, TRAF1 and STAG1 were hypomethylated 30 min after the first bout and still hypomethylated after loading and reloading, with expression highest after reloading.
Mechanisms
Measured: lean mass and torque rose, fell and rose again with the three phases; the methylome shifted toward hypomethylation after one session and after each block; some sites changed by the first block stayed changed when lean mass had returned to baseline; the expression of the genes picked from those sites tracked their methylation inversely, and some of them correlated with lean mass across conditions within the same 7-8 men.
Authors' interpretation: earlier hypertrophy leaves growth-related genes hypomethylated and "switched on"; the hypomethylation is kept through unloading and lets those genes respond more on reloading, so human muscle "possesses an epigenetic memory of earlier acute and chronic anabolic stimuli". UBR5, RPL35a, SETD3, PLA2G16 and HEG1 are proposed as new hypertrophy genes, and the acute-bout genes as early biomarkers of later growth.
Inference (what the design can and cannot separate): it shows within-person change over time and that some marks outlasted the muscle they came with.It cannot attribute the larger second gain to the first block, because the second block always came second: the men knew the lifts and their loads (the first block's starting loads were set from technique and feedback in men new to lifting), lifted a significantly higher average load in reloading (P = 0.043), and were 14 weeks further on.Neural and skill learning raise strength and the training load a man can lift; a heavier load can build more muscle. Separating memory from order needs a group or limb that does only the second block — REF-1360 had an untrained control leg and found no faster regain.
Inference (the counts): if the per-site tests were unadjusted, about 42,500 of roughly 850,000 sites would pass P < 0.05 by chance alone (our arithmetic), more than any count reported here (17,365 to 27,155). The counts by themselves therefore do not show a training effect; the case rests on the direction of change (about two hypomethylated sites to each hypermethylated one after reloading), the temporal clusters and the gene-expression follow-up.
Inference (function): nothing here shows that the retained marks speed regrowth. The best-documented "memory" gene, UBR5, was not retained low — its methylation was above baseline after unloading — and the retained-cluster genes rose most in expression while nothing was being trained and did not track lean mass.
Practical Application
Does muscle remember an earlier strength block? In 8 untrained men some DNA marks of a 7-week block outlasted 7 weeks off; a second block built more lean mass (7 finished). One group, no control: order, skill and load are not separated.
Muscle and strength came back with retraining: after 7 weeks off, 7 more weeks of the same programme took lower-limb lean mass to 12.4% and knee-extension torque to 18% above the starting values (B-, 7 untrained men). What it shows is that a return to training rebuilds what a short break took; it does not show that the earlier block made this faster.
Do not promise a faster comeback from this study: the human strength trial with an untrained control limb found no faster regain (REF-1360), and a later interval-training study with the same senior author found the marks kept and no physiological memory (REF-1427). Not from this paper; the base's own practice for returns is in SYNTH-36.
What it is good for as a source: the first human evidence that some DNA-methylation changes from training persist after the muscle they came with is gone, and a measured time course of lean mass and torque through 7 weeks on, 7 off and 7 on in untrained men. What it does not show: that the marks cause faster or larger regrowth, that the larger second gain came from the first block rather than from doing the programme twice, or anything for trained athletes, endurance or longer breaks.
Limitations
No control. One group, one order: the reloading block always followed the loading block. Learning the lifts, the higher average load in reloading (P = 0.043), time of year and the men's expectations all go with "second block" and cannot be separated from a memory of the first; no untrained limb or second-block-only group was included.
Small. 8 enrolled, 7 finished; mean ± SEM throughout, so the spread between men is narrower on the page than it was (an SEM of 2.4 years for age is an SD of about 6.8, our arithmetic).
A marker, tested many times. Hundreds of thousands of CpG sites at P < 0.05; the Methods mention a Bonferroni correction without saying what it covered, and the clustering used unadjusted P values; two outlier samples were removed without saying which in the main text (Supplementary Figure 1, not read, may), and the stated n per time point (8, 8, 8, 8, 7) does not show where they fell.The "retained" values of the four named genes were not significantly different from baseline during unloading.
Unmeasured break. Activity during unloading was checked by conversation only.
Tissue and measures. Methylation from whole-muscle biopsies (fibres plus other cells; the authors judge the samples homogeneous by principal component analysis); lean mass by DEXA is not muscle alone; repeated biopsies from the same muscle; the torque method is in a supplement that was not read.
Interpretation by the model's proponents. The senior author conceived the study and leads the epigenetic-memory account (REF-1425).
Funding and interests: a PhD studentship from the Doctoral Training Alliance UK/LJMU/Keele University and a GlaxoSmithKline grant to the senior author for the methylation and gene-expression work, as printed; no competing interests declared. No device maker, federation or data platform is named.
How this entry was checked and changed
Why this level
an uncontrolled within-person study (n = 7 completers of 8 enrolled; the rubric's B), held a step lower for limitations beyond its design: with no control group or untrained limb, the larger second gain cannot be separated from simply doing the programme a second time (skill with the lifts, a significantly higher average training load in reloading, P = 0.043, time); the memory claim rests on a molecular marker counted at P < 0.05 per site across hundreds of thousands of sites, and the "retained" marks of the four named genes were not significantly below baseline during unloading; written by the proponents of the epigenetic-memory model (REF-1425). Read from the full text (open access, CC BY 4.0, via Europe PMC); the supplementary files were not read
Full text (checked 2026-10-29)
Eight healthy males gave written, informed consent; one participant withdrew at experimental week 17 of 21 (full text)
Eight previously untrained male participants (27.6 ± 2.4 yr, 82.5 ± 6.0 kg, 178.1 ± 2.8 cm, means ± SEM); methylome n = 8 baseline, acute, loading, unloading, n = 7 reloading; DEXA and isometric peak torque for n = 7 (full text)
Lean mass: an increase of 6.5% (± 1.0%; P = 0.013) after loading (20.74 ± 1.11 kg vs. 19.47 ± 1.01 kg baseline); reduced by 4.6% ± 0.6% (P = 0.02) with unloading (19.83 ± 1.06 kg); 12.4 ± 1.3% above baseline after reloading (21.85 ± 2.78 kg, P = 0.001), an increase of 5.9 ± 1.0% compared to loading (P = 0.005) (full text)
Reloading phase (unloading to reloading) compared to the loading phase (baseline to loading): P = 0.022 (full text)
Isometric peak torque increased by 9.3 ± 3.5% from 296.2 ± 22.1 Nm to 324.5 ± 27.3 Nm, not statistically significant; reduced by 8.3 ± 2.8% with unloading; an 18 ± 3.6% increase (349.6 ± 27.7 Nm) vs baseline after reloading (P = 0.015) (full text)
The reloading phase displaying a significant (P = 0.043) increase in average load (full text)
17,365 CpG sites changed after loading, 9,153 hypomethylated and 8,212 hypermethylated; 8,891 hypomethylated after unloading; 27,155 changed and 18,816 hypomethylated after reloading (full text)
AXIN1, GRIK2, CAMK4, TRAF1: hypomethylated after loading (0.78 ± 0.09; P = 0.036), maintained during unloading (0.84 ± 0.09) and reloading (0.83 ± 0.05), albeit non-significantly (full text)
UBR5 CpG methylation 1.27 ± 0.02 after unloading (P = 0.013) and 0.78 ± 0.02 after reloading (P = 0.039); expression 1.84 ± 0.5 vs 0.82 ± 0.27 (P = 0.035) (full text)
Funded by a PhD studentship (Doctoral Training Alliance UK/LJMU/Keele University) and a GlaxoSmithKline grant awarded to Adam P. Sharples; no competing interests declared (full text)