REF-1427 · Pilotto et al., 2025 · American Journal of Physiology-Cell Physiology · Paper · PubMed

Does my body remember interval training after a long break?

Not in fitness: in 20 untrained adults a second 8-week HIIT block after 3 months off raised VO2max no more than the first (+14% both times, preprint). DNA marks stayed.

Main caveat: No control group. Changes over 7 months in one group cannot be separated from time, season or repeated biopsies.
limited, mixed evidenceMeasured on non-athletes and recreational people

Pilotto, Porcelli, Sharples and colleagues asked whether high-intensity interval training, like resistance training in their group's earlier study (REF-1426), leaves an epigenetic memory in human muscle.

Twenty healthy, untrained adults (11 men, 9 women) did 8 weeks of HIIT on a cycle ergometer, three sessions a week and mostly at home, then 3 months of habitual life, then the identical 8 weeks again; incremental cycle tests and, in 15 of them, thigh-muscle biopsies were taken at baseline, after training, after detraining and after retraining.

Peak power rose 18% and VO2max 14% after the first block; after detraining every maximal variable was back at baseline; the second block raised peak power 13% and VO2max 14% — no more than the first (the published abstract: no differences between interventions, P > 0.58), with the same power actually ridden in both blocks.

In the DNA of 5 participants, 21,605 sites differed from baseline after training (two thirds hypomethylated), 3,854 still differed after detraining (87% hypomethylated) and 10,396 after retraining; 1,190 sites hypomethylated by the first block stayed hypomethylated through detraining and retraining.

Five genes with retained hypomethylation — ADAM19, INPP5a, MTHFD1L, CAPN2 and SLC16A3 — kept a raised expression through the break in 15 participants; three of them work in lactate transport and calcium signalling. The authors conclude that human muscle "possesses an epigenetic memory of HIIT", while noting that "significant memory was not observed in physiological parameters".

Our reading: the marks and the transcripts were retained; fitness was not, and it rose no more the second time. Retained epigenetic marks did not mean retained or enhanced physiology. With the marks counted in 5 people at an unadjusted threshold and no control group, the study shows a molecular trace, not a functional memory.

Protocol

Design: single group, within-subject, about 7 months (8 weeks training, 3 months detraining, 8 weeks retraining); University of Pavia; ethics approval Besta 64-19/07/2019. No control group.
Participants: n = 20 (11 men, 9 women; 25 ± 5 yr), recruited from the local community, never in a structured training programme, VO2max at screening below 40 (women) or 45 (men) ml·min−1·kg−1, free of major disease and medication. Biopsies at all time points: 15 of 20 (9 men, 6 women); gene expression in all 15; DNA methylome in 5 per time point (the authors' power analysis, from REF-1426, put 4 as enough to detect a 5% change in methylation).
Training (both blocks identical): 8 weeks, three sessions a week (24 sessions); the first three supervised, the rest at home on an electronically braked cycle ergometer with time, heart rate and power uploaded to an app (MyZone MZ3) and a weekly phone call; 10-min warm-up, then short (1-2 min) and long (3-4 min) high-intensity bouts, progressively incremented in power and repetitions; prescribed at an average of 120% of Wmax from the test before each block.
Detraining: 3 months; instructed to return to habitual life; physical activity monitored by accelerometer (the results are not reported in the preprint text).
Tests at each time point: incremental cycle test to exhaustion (10-15 W per minute, 70-80 rpm) with a verification bout at 90% of Wmax 30 min later; VO2max, gas exchange threshold, respiratory compensation point, heart rate, capillary lactate, RPE; 48 h without strenuous exercise before; biopsy on a separate visit.
Methylation: EPIC array, 791,084 probes after filtering (sex chromosomes, failed, SNP and cross-reactive probes removed); differentially methylated positions (DMPs) by one-way ANOVA at an unadjusted p ≤ 0.01; differentially methylated regions by DMRcate; temporal profiles by self-organising maps; no outlier samples found.
Gene expression: RT-qPCR of 11 genes with differentially methylated regions in the retained profiles, against each participant's own baseline.
Analysis: repeated-measures ANOVA with Tukey post hoc tests; mean ± SD.
Funding and interests: "Simone Porcelli was supported by a grant from Sports Medicine Italian Federation (FMSI01092021). Adam P. Sharples is currently supported by Research Council Norway (project no: 314157).Roberto Bottinelli is currently supported by Italian Ministry for Research (MUR, PRIN Prot. 2020EM9A8X)"; the acknowledgements thank Matrix Fitness Italia for access to FitActive facilities and for supplying MyZone MZ3 heart-rate chest bands; preprint statement: "The authors have declared no competing interest" (as printed in the preprint).

Key Results

Physiology (n = 20) and DNA methylation (n = 5) at each time point, against baseline:

MeasureAfter training (8 wk)After detraining (3 mo)After retraining (8 wk)
Wmax+18% (p < 0.001)back to baseline (all p > 0.99)+13% (p < 0.001)
VO2max+14% (p < 0.001)back to baseline (all p > 0.99)+14% (p < 0.001); no difference between the two blocks (p > 0.58, published abstract)
Power actually ridden in training245 ± 52 W; 202% of the power at the gas exchange threshold—247 ± 53 W (p = 0.653); 208% (p = 0.504)
DMPs vs baseline (hypo / hyper)21,605 (14,516 / 7,089)3,854 (3,349 / 505)10,396 (9,644 / 752)
Expression of the 6 signature-1 genes (n = 15, fold of baseline)2.25 ± 0.50 (p = 0.07)3.05 ± 1.17 (p < 0.01)2.50 ± 1.43 (p < 0.05)
Expression of the 5 signature-2 genes (n = 15)1.65 ± 0.18 (p < 0.05)1.94 ± 0.49 (p < 0.001)1.62 ± 0.29 (p < 0.05)
What was retained: of the 5,334 DMPs that overlapped across the three comparisons, 1,190 were hypomethylated after training and stayed so through detraining and retraining (signature 1), 2,000 stayed hypomethylated but slightly less during detraining (signature 2), 1,675 returned toward baseline in detraining and were hypomethylated again after retraining (signature 3) and 469 were hypermethylated (signature 4). Wnt signalling was the top pathway in signature 1.
Five "memory genes": ADAM19, INPP5a, MTHFD1L, CAPN2 and SLC16A3 showed an inverse pattern of methylation and expression, with hypomethylation and raised expression kept into detraining and retraining; SLC16A3, INPP5a and CAPN2 "are involved in lactate transport and calcium signaling" (published abstract).
What was not retained: after detraining, VO2max, Wmax and all other maximal variables were back at baseline (all p > 0.99); the second block raised them no more than the first; the authors report "no signs of memory" in VO2max, ventilation, RER, heart rate, lactate or RPE.
Overlap with resistance training: re-analysing REF-1426, the authors found 5,033 sites with the same retained profile after resistance training; only 5 CpG sites were shared with the 1,190 here (in RTTN, SRC, DDR1 and two unannotated).

Mechanisms

Measured: the first block shifted muscle DNA toward hypomethylation; part of that shift was still present after 3 months without training, alongside raised expression of genes in calcium signalling, lactate transport and folate metabolism; fitness itself returned to baseline and rose by the same amount on retraining.
Authors' interpretation: human muscle "possesses an epigenetic memory" of HIIT. For the missing physiological memory they offer a ceiling effect (untrained people near their trainability limit after the first block), blocks too short for the molecular memory to reach the phenotype, and a possibly faster adaptation they could not see because there was no mid-block test.
Inference (marks vs physiology): VO2max was at baseline after detraining while the marks were still there, and the second block raised it no more than the first; whatever the marks do, they did not appear as kept or extra fitness in this study.
Inference (the counts): at an unadjusted p ≤ 0.01 across 791,084 probes, about 7,900 would pass by chance alone if nothing had changed (our arithmetic). The detraining count (3,854) is below that, so "retained" rests on the overlap with sites changed by training, the direction of change (87% hypomethylated) and the expression follow-up in 15 — not on the count.
Inference (no control group): expression of the signature-1 genes was highest after 3 months without training (3.05 against 2.25 after training); without a control group, an effect of time, season or repeated biopsies of the same muscle cannot be excluded.

Practical Application

Does my body remember interval training after a long break? Not in fitness: in 20 untrained adults a second 8-week HIIT block after 3 months off raised VO2max no more than the first (+14% both times, preprint). DNA marks stayed.
Fitness came back with retraining, no more than the first time: after 3 months off, the same 8 weeks of HIIT rebuilt VO2max and peak power as the first block had built them (B-, 20 untrained adults; no mid-block test, so whether it came back faster is unknown). This fits the base's line that the aerobic qualities go first in a break while strength largely stays (REF-374, SYNTH-36).
Do not read "epigenetic memory" as a head start: the marks were retained, but the fitness they might carry was not. The only earlier human test of endurance memory the authors cite (Lindholm 2016, one-legged training with 9 months off; not in the base) found no retained gene expression.
What it is good for as a source: that interval training changes muscle DNA methylation and that part of the change outlasts 3 months off in untrained adults, with VO2max measured alongside. What it does not show: that the marks change performance, how fast fitness returns, or anything for trained athletes.

Limitations

No control group. Changes over 7 months in one group cannot be separated from time, season or repeated biopsies; the expression of the "memory genes" peaked during detraining.
Five methylomes. The methylation results come from 5 of the 20 participants at an unadjusted p ≤ 0.01 across 791,084 probes; the power analysis rested on the group's earlier study (REF-1426).
No mid-block test. A faster return of fitness could not be detected, the authors note; the end-of-block values were equal.
Home training and an unreported break. Adherence was checked through uploaded heart rate and power; accelerometer data from the detraining period are not reported in the preprint text.
Mixed nuclei. Methylation from biopsy homogenates (fibre and interstitial nuclei together).
Source. Read from the bioRxiv preprint v1, whose tables are images (their values are not in the saved text); the published version may differ in detail. The authors' own group proposed the epigenetic-memory model (REF-1425, REF-1426).
Funding and interests: Sports Medicine Italian Federation (FMSI01092021), Research Council of Norway (314157) and the Italian Ministry for Research (PRIN) as printed; Matrix Fitness Italia provided facility access and heart-rate chest bands (acknowledgements); no competing interest declared in the preprint.
How this entry was checked and changed
Why this level

an uncontrolled within-person study in 20 adults (the rubric's B), held a step lower because its headline claim, an epigenetic memory, is a molecular marker measured in 5 participants at an unadjusted p ≤ 0.01 with no control group for time, season or repeated biopsies, while the outcomes that matter to an athlete (VO2max, peak power) showed no memory. Read from the full text of the bioRxiv preprint v1 (posted 3 June 2024; an OpenAlex-listed copy); the published version (CC BY) sits behind a Cloudflare check at the publisher and was not read. The published abstract agrees with the preprint on n = 20, age, the VO2max statistics and the five genes; numbers not in that abstract are the preprint's

Full text (bioRxiv preprint v1, checked 2026-10-29)
Twenty male (n = 11) and female (n = 9) adults; twenty healthy subjects (25±5yrs) completed two HIIT interventions lasting 2 months, separated by 3 months of detraining (full text)
Fifteen participants out of 20 agreed to muscle biopsy; 15 analysed for gene expression and 5 subjects per time point for DNA methylome analysis (full text)
Wmax and VO2max increased after TRAINING (+18% and +14%, respectively; both p < 0.001) and RETRAINING (+13% and +14%, respectively; both p < 0.001); DETRAINING returned all maximal values to BASELINE levels (all p > 0.99) (full text)
Workload carried out 245 ± 52 W in training and 247 ± 53 W in retraining (p = 0.653); 202% and 208% of WGET (p = 0.504) (full text)
21,605 DMPs after TRAINING (14,516 hypomethylated, 7,089 hypermethylated); 3,854 after DETRAINING (3,349 hypomethylated, 505 hypermethylated); 10,396 after RETRAINING (9,644 hypomethylated, 752 hypermethylated) (full text)
Signature 1: 1,190 DMPs hypomethylated after TRAINING and retained even after 3 months of DETRAINING and RETRAINING (full text)
Signature 1 gene expression 2.25 ± 0.50 (p = 0.07) after training, 3.05 ± 1.17 (p < 0.01) after detraining, 2.50 ± 1.43 (p < 0.05) after retraining (full text)
DMPs called at an unadjusted p-value of ≤ 0.01 from a final list of 791,084 probes (full text)
Grants: Sports Medicine Italian Federation (FMSI01092021), Research Council Norway (project no: 314157), Italian Ministry for Research; the authors have declared no competing interest (full text)