REF-1424 · Serrano et al., 2025 · The Journal of Physiology · Paper · PubMed · Free full text

Do muscles keep the nuclei they gain as a memory for the next block?

Not reliably: this critical evaluation finds them lost after longer detraining in mice and a few small human studies; it points to lasting DNA and protein changes.

Main caveat: Opinion format. A short response with no search, appraisal or weighting; the studies cited were chosen by the authors, largely their own laboratory's.
limited, mixed evidenceMeasured on non-athletes and recreational people

Serrano, Dupont-Versteegden and Murach respond to a 2025 Opinion piece on muscle memory and sports integrity (Sitko and colleagues), which raised whether former steroid users keep a long-term advantage through extra myonuclei.

They define muscle memory as a larger or faster response to a growth stimulus after an earlier one and state that "the question is not whether muscle memory exists as described" — citing REF-1426, REF-1425 and a mouse study — "but, instead, what is the mechanism controlling it?".

Their case against myonuclear permanence as that mechanism: (1) in the founding steroid study from the Gundersen laboratory, nuclei per fibre cross-section in a mouse muscle were about 0.7 in controls, about 1.0 after testosterone and about 0.8 three months after it stopped — about 20% lower, "a non-trivial loss"; (2) brief unloading (days to weeks in rodents, weeks to a few months in humans) usually leaves myonuclear number unchanged, but over longer periods of unloading, inactivity, ageing and especially paralysis there is "mounting evidence" of loss across species; (3) in adult mice, 3 months or more after 8 weeks of loaded wheel running, myonuclei in two leg muscles were about 20% lower as the fibres returned to their untrained size, by two counting methods; (4) in humans, REF-1360 added no nuclei and so did not test permanence, but a re-analysis of the participants who did add nuclei found them lost on detraining, and two studies in older adults found myonuclear density back at baseline after 12 weeks or a year off.

Counting myonuclei is hard (interstitial nuclei, nuclei that lengthen with training and age, unspecific markers, small samples, juvenile rodents). As alternatives they point to lasting changes in DNA methylation (REF-1426; mouse myonuclei), in proteins (a human study with some proteins still raised after 10 weeks of detraining) and in microRNAs.

They conclude that "myonuclear permanence therefore may fail to uphold the theory of muscle memory". Our reading: the current, critical statement of the myonuclear question, from the group whose data most challenge permanence; together with REF-1359 it frames the dispute.

Its acceptance of muscle memory as a phenomenon rests on citations — one uncontrolled human study, a review by that study's senior author and a mouse study from its own laboratory — while a human strength trial with an untrained control leg found no faster regain (REF-1360).

Protocol

Design: critical evaluation in response to Sitko et al. (2025, J Physiol); narrative, no search strategy, no appraisal, no pooled data. n = not applicable (an argument from cited studies; its table lists six rodent studies).
Topics: the origin of the myonuclear theory (the 2013 steroid study and earlier denervation work); myonuclear loss with unloading, atrophy, ageing, paralysis and detraining in rodents and humans; the mechanisms of nuclear removal (apoptosis, autophagy, degeneration); methods for counting myonuclei; alternative memory mechanisms (methylome, transcriptome, proteome, microRNA); doping policy.
Table 1 (rodent studies reporting myonuclear loss): rat spinal-cord transection, 10 days (down 25%) and 8 weeks (down 40%); rat hindlimb suspension, 1-2 weeks (down about 30%) and 2 weeks (down about 30%); mouse detraining after loaded wheel running, 12 weeks (plantaris down 18%) and 24 weeks (gastrocnemius down about 15%, soleus no loss but no atrophy).
Funding and interests: "supported by National Institutes of Health R01 AG080047 and K02 AG088465 to KAM"; "No competing interests declared". The mouse detraining studies (Dungan 2019, Murach 2020) and the human re-analysis (Murach 2019) the article relies on are co-authored by the senior author (by its reference list); the article itself calls only Wen 2021 "our laboratory".

Key Results

QuestionWhat the authors stateEvidence they cite (second-hand)
Does muscle memory exist?"the question is not whether muscle memory exists as described ... but, instead, what is the mechanism controlling it?"REF-1426 (7 men, no control), REF-1425, a mouse methylation study (Wen 2021)
The founding steroid studynuclei per fibre cross-section in mouse EDL about 0.7 (controls), about 1.0 after testosterone, about 0.8 after 3 months off — about 20% lower, "a non-trivial loss of myonuclei"rereading of the Gundersen laboratory's histology
Brief unloadingrodents (days to weeks) and humans (weeks to a few months): "typically not associated with a lower number of myonuclei", with exceptionsseveral rodent and human studies
Longer unloading, inactivity, ageing, paralysis"mounting evidence that myonuclei are lost" across species; "strong evidence" after spinal cord injury or paralysisreviews; Table 1
Detraining after training, miceabout 20% fewer myonuclei in plantaris and gastrocnemius 3 months or more after 8 weeks of loaded wheel running, as fibre size returned; soleus kept its nuclei but also did not detrainthe senior author's laboratory; two counting methods
HumansREF-1360 added no nuclei, so "did not address the permanency"; a re-analysis found that participants who did add nuclei lost them on detraining; older adults after 6 months of training and a year off: type 2 fibre size and myonuclear density back to baseline; similar after 12 weeks offMurach 2019 (re-analysis); Snijders 2019; Blocquiaux 2020
Methodssingle isolated fibres are the gold standard; histology can mistake interstitial nuclei; myonuclei lengthen or shorten with training and age; the PCM1 marker becomes unspecific under stress; detecting a 5-15% difference needs an adequately powered sample; several memory studies used juvenile rodents or did not state age—
Alternativeslasting changes in the methylome (UBR5 in humans; growth genes in mouse myonuclei), the proteome (CBR1 and calcium-binding proteins such as CAPN2 raised after 10 weeks of detraining in humans) and microRNA (miR-1, mice)REF-1426; Wen 2021; Hulmi 2025; mouse data
Dopingexpelling an athlete on the basis of a possible myonuclear advantage "need[s] to be approached with caution"—
Conclusion"Myonuclear permanence therefore may fail to uphold the theory of muscle memory"; look to the methylome, proteome and transcriptome—

Mechanisms

Measured (in the studies cited): myonuclear number unchanged after short unloading; lower after longer detraining or atrophy in rats and mice (15-40% in Table 1) and in a few small human studies; in mice, methylation of growth genes in muscle-fibre nuclei that persisted after detraining; in humans, some proteins still raised after 10 weeks of detraining.
Authors' interpretation: nuclei are added in growth and probably removed when a fibre shrinks for long enough (by apoptosis, autophagy or degeneration — debated and not well understood), so they cannot carry a long memory; if muscle memory is real, lasting molecular changes are the likelier carrier.
Inference: two leading groups read largely the same myonuclear data in opposite directions (REF-1359 against this paper); neither side has a human trial in which a previously trained muscle regrew faster than an untrained one under a control. In humans the cellular question is open, as the proponents of the epigenetic view also say (REF-1425).
Inference: the authors' alternatives are themselves measured as markers in small studies (REF-1426: one group, no control; REF-1427: marks kept, fitness not).

Practical Application

Do muscles keep the nuclei they gain as a memory for the next block? Not reliably: this critical evaluation finds them lost after longer detraining in mice and a few small human studies; it points to lasting DNA and protein changes.
Read "muscle memory" as a mechanism debate, not a training rule: the authors accept a faster regrowth as shown but dispute how it works; the base holds that the human evidence for a faster comeback is weak (REF-1360: no faster regain against an untrained control leg) and that returns follow SYNTH-36 whatever the mechanism (the base's own practice, not from this paper).
What it is good for as a source: the current case against myonuclear permanence, with the methodological traps in counting nuclei and the alternatives under study. What it does not show: that muscle memory exists in humans (it assumes so from citations), how long any memory lasts, or anything about endurance fitness or trained athletes.

Limitations

Opinion format. A short response with no search, appraisal or weighting; the studies cited were chosen by the authors, largely their own laboratory's.
Mostly rodent. All six studies in the table of myonuclear loss are in rats or mice; the human statements rest on two small studies in older adults and a post hoc re-analysis of a subgroup of REF-1360.
Muscle memory taken as given. The existence of faster regrowth is asserted from citations (REF-1426, a single group without control; REF-1425, a review by the same group; a mouse study), not argued.
Two descriptions that do not add up. It gives REF-1360's detraining as 16 weeks; the trial reports about 20 weeks. It describes "a persistent change to UBR5 promoter region hypomethylation following detraining" in REF-1426; in that study UBR5 methylation was above baseline after unloading (1.27 ± 0.02, P = 0.013) and low again only after reloading, and the epigenetic-memory review itself says UBR5 methylation returned to baseline during detraining (REF-1425).
Funding and interests: NIH grants R01 AG080047 and K02 AG088465 to the senior author, 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

the rubric's B+ for an authoritative narrative review from a leading group (a leading laboratory in myonuclear and satellite-cell research), held two steps lower: a short response article with no systematic search that argues its own laboratory's side of a live dispute (the mouse detraining data and the re-analysis of REF-1360 it leans on are the senior author's own); the six studies of myonuclear loss in its table are all rodent, and its human statements rest on three small studies (two in older adults, one a secondary re-analysis of a subgroup); and it misdescribes two studies (see Limitations). Graded above the opposing model's C+ (REF-1359) because it was read in full and cites human data for its human claim. Read from the full text (NIH author manuscript in PubMed Central)

Full text (checked 2026-10-29)
In our opinion, the question is not whether muscle memory exists as described, but what is the mechanism controlling it (full text)
Myonuclear number in the EDL muscle is ~0.7 per fibre cross-section in controls and ~1.0 immediately after testosterone; after a 3 month cessation it is ~0.8, an ~20% lower value, which implies a non-trivial loss of myonuclei (full text)
After 3 months or more of detraining following 8 weeks of PoWeR in adult female mice, myonuclear numbers in the plantaris and gastrocnemius were significantly lower (~20%) (full text)
One recent study aimed to test muscle memory after 16 weeks of detraining following resistance training in humans (Psilander et al., 2019) (full text)
Older adults resistance trained for 6 months, then a year of training cessation: type 2 fibre hypertrophy and myonuclear density both returned to baseline (full text)
To detect a relatively modest 5–15% difference in myonuclear density an appropriately powered sample size should be used (full text)
Myonuclear permanence therefore may fail to uphold the theory of muscle memory (full text)
Supported by National Institutes of Health R01 AG080047 and K02 AG088465 to KAM; no competing interests declared (full text)