Myostatin, Activin and Follistatin: The Muscle Brake Explained

Peptide science

Myostatin is the body’s brake on skeletal muscle. Identified in 1997 as GDF-8, a member of the TGF-beta superfamily, it does something unusual for a growth factor: it restrains growth rather than driving it. Delete the gene in a mouse and individual muscles weigh two to three times more than normal. That single result launched nearly three decades of drug development, an enormous amount of marketing, and a clinical record that is far thinner than the biology suggests it should be.

What myostatin actually is

Myostatin is encoded by MSTN on chromosome 2q32.2 and catalogued in UniProt as O14793. It is made as a 375-amino-acid precursor: a short signal sequence, a long propeptide, and a C-terminal region that pairs with a second copy of itself through a disulfide bond to form the active dimer.

The part that matters for anyone trying to interpret a blood test is what happens next. Myostatin does not circulate as a free, active molecule. It circulates as a latent complex, still bound to its own propeptide, and it requires a second cleavage by a tolloid-family metalloproteinase before it can signal at all. Most of the myostatin in your blood is switched off. That is why a “myostatin level” is a much weaker piece of information than it sounds — the regulated step is release from latency, not production.

The receptor is shared, and that changes everything

Active myostatin binds the type II receptor ACVR2B, and to a lesser degree ACVR2A. The type II receptor then recruits a type I partner — ALK4 or ALK5 — which phosphorylates SMAD2 and SMAD3, and the SMAD complex moves to the nucleus and suppresses the muscle growth programme.

Nothing about that chain belongs to myostatin alone. Activin A signals through the same type II receptors and the same SMADs, with comparable potency. ACVR2B also binds activins B, C and E, GDF11, BMP9 and BMP10. And GDF11 shares 89 per cent sequence identity with myostatin across the mature domain, with an identical set of type-II-binding residues.

Mice are a myostatin-dominant system. Primates are not.

The single most useful paper for understanding why this field disappointed is Latres and colleagues in Nature Communications in 2017, which ran the comparison head to head. In mice, blocking myostatin alone increased tibialis anterior weight by 20.1 per cent over 21 days; blocking activin A alone produced a non-significant 5.0 per cent; blocking both produced 43.9 per cent. Adding an anti-GDF11 antibody on top did not increase it further.

The reason is a straightforward difference in the ligand balance. Circulating myostatin runs four to eighteen times lower in monkeys, rats and humans than in mice, while activin A runs three to four times higher in monkeys and humans. Remove myostatin from a mouse and you have removed most of the brake. Remove it from a person and you have removed part of it.

The human numbers are correspondingly modest. A 2024 meta-analysis of seven randomised trials of bimagrumab — an antibody against the type II receptor, so broad blockade rather than myostatin-specific — found thigh muscle volume up 5.29 per cent and fat-free mass up 1.90 kg, with no significant improvement in muscle strength or physical performance. That last clause is the whole story of this drug class in one line.

Follistatin is a sponge, not a receptor blocker

Follistatin is the body’s own neutraliser for this pathway. It is a secreted glycoprotein encoded by FST on chromosome 5q11.2, made as a 344-amino-acid precursor. It does not compete at the receptor. Two follistatin molecules wrap around a single activin dimer, burying roughly a third of its surface and blocking both the type I and type II binding sites at once. The same trick works on myostatin.

Two isoforms exist, and confusing them is the most common error in consumer writing about follistatin. FS-288 carries a heparan-sulfate-binding region that tethers it to cell surfaces and extracellular matrix. FS-315 has a longer acidic C-terminal tail, does not stick to the matrix, and is the predominant follistatin isoform in serum. The names FST317 and FST344 refer to the corresponding precursor lengths, which is where the marketing name comes from.

There is a good pharmacological reason nobody injects it. A 2013 study in the Journal of Pharmacology and Experimental Therapeutics concluded that native FS315 is “poorly suited for acting as a parenterally administered biotherapeutic with broad systemic effects”; an engineered Fc-fusion with heparan-sulfate binding removed needed roughly a hundredfold longer terminal half-life and about sixteen-hundredfold greater exposure before it worked. Follow-up work attributed the problem to rapid heparin-mediated hepatic clearance. The closest thing to a real follistatin drug, ACE-083, sidestepped this entirely by being injected directly into a single muscle — and even then produced a 14.5 per cent increase in muscle volume with no significant change in mean strength before its trials were terminated.

What happens when the brake is missing

The natural experiments are more informative than most of the trials. Belgian Blue cattle carry an eleven-nucleotide deletion that frameshifts away the entire active region; Piedmontese cattle carry a missense change replacing an invariant cysteine. Both breeds show roughly a 20 per cent increase in muscle mass, driven by more fibres rather than bigger ones.

In whippets, a two-base deletion creates a stop codon at amino acid 313. Dogs carrying one copy averaged 17 per cent more mass per centimetre of height, and the racing data are striking: twelve of forty-one dogs in the fastest two grades carried the deletion, against one of forty-three in the slowest two.

The human case, reported in the New England Journal of Medicine in 2004, involved a child homozygous for a splice-donor variant in intron 1. His quadriceps cross-sectional area was 7.2 standard deviations above age- and sex-matched controls (6.72 versus 3.13 square centimetres) and his subcutaneous fat pad 2.88 standard deviations below. At four and a half years he could hold two three-kilogram dumbbells in horizontal suspension with his arms extended, with normal motor and mental development and a normal echocardiogram — though the authors noted explicitly that he was still too young for cardiac abnormalities to be definitively excluded. No long-term follow-up has been published.

Twenty years of trials: mass up, function flat

The clinical record is unusually consistent, and unusually disappointing. MYO-029, the first anti-myostatin antibody in humans, raised lean mass 2.4 per cent at its top dose and produced no strength improvement in any dystrophy subgroup at any dose. ACE-031, a soluble ACVR2B decoy, raised lean mass more but was halted after epistaxis and telangiectasia appeared. Bimagrumab missed its primary endpoint in inclusion-body myositis by a wide margin. Domagrozumab missed in Duchenne muscular dystrophy with a p-value of 0.94. Taldefgrobep alfa did not separate from placebo in spinal muscular atrophy.

The one clear exception is apitegromab, which targets the latent, pro-form of myostatin rather than the mature dimer. In non-ambulatory spinal muscular atrophy patients already on SMN-directed therapy, it improved the Hammersmith Functional Motor Scale Expanded by 1.8 points over placebo. It is worth noting that its earlier ambulatory cohort was negative, which is why the phase 3 enrolled non-ambulatory patients only. It received a complete response letter in September 2025 relating to an inspection of a third-party fill-finish site — not to efficacy or safety — and after resubmission carries an FDA decision date of 30 September 2026. As of this writing no myostatin-targeting drug is approved anywhere in the world.

The idea that actually changed

The interesting turn is that the field stopped trying to treat muscle disease and started trying to protect muscle during weight loss — a setting where “more lean mass” is the endpoint itself rather than a surrogate for strength.

In Regeneron’s COURAGE trial, roughly a third of the weight lost on semaglutide alone was lean mass (33.0 per cent of total weight lost at week 26). Adding trevogrumab, an anti-myostatin antibody, cut that lean-mass loss roughly in half. A triplet adding garetosmab, an anti-activin A antibody, pushed fat to 92.6 per cent of the weight lost. But discontinuation for adverse events climbed from 4.6 per cent on semaglutide alone to 28.3 per cent on the triplet, with two deaths in the triplet arm.

Lilly’s BELIEVE trial, published in Nature Medicine in March 2026, showed a similar shape: bimagrumab alone produced 9.3 kg of weight loss and a lean-mass gain of 2.3 to 2.7 per cent; semaglutide alone produced 14.2 kg with a lean-mass loss of 5.3 to 7.9 per cent; the combination produced 17.8 kg with only 1.1 to 2.6 per cent lean-mass loss, and 92.3 per cent of the weight lost was fat. Discontinuations for adverse events ran 14.0 to 21.4 per cent in the bimagrumab groups.

The safety questions that come with blocking this pathway

• Bleeding and telangiectasia. In the ACE-031 Duchenne trial, epistaxis occurred in five of nine and telangiectasia in five of nine patients at the 1.0 mg/kg dose, against none of six on placebo, with no serious adverse events. The proposed mechanism — and it is explicitly a proposal, not a demonstrated finding — involves interference with BMP9 signalling through the endoglin-ALK1 axis, the same axis mutated in hereditary haemorrhagic telangiectasia.
• The closest real-world readout. Sotatercept, an ActRIIA-Fc approved for pulmonary arterial hypertension, carries labelled warnings for erythrocytosis, severe thrombocytopenia, serious bleeding, embryo-fetal toxicity and impaired fertility. Serious bleeding was reported in 4 per cent versus 1 per cent and 7 per cent versus 5 per cent in its two pivotal trials, and telangiectasia in 16.6 versus 4.4 per cent and 25.6 versus 3.5 per cent.
• The reproductive axis. Activin was named for stimulating follicle-stimulating hormone, and follistatin for suppressing it. In humans, a single dose of an ActRIIA-Fc produced a dose-dependent fall in serum FSH. This pathway is not confined to muscle.
• Tendon. Myostatin-null mice have smaller tendons with lower fibroblast density and less type I collagen, and their muscle fibres are more susceptible to contraction-induced injury. A stronger muscle pulling on a stiffer, more brittle tendon is a plausible mismatch that has never been studied in humans.
• Cardiac effects. Genuinely unresolved. One well-known paper is titled “Myostatin does not regulate cardiac hypertrophy or fibrosis”; other groups report eccentric hypertrophy in null mice. Do not treat either position as settled.
• The cancer setting. In a pancreatic cancer cachexia trial, the high-dose landogrozumab arm was terminated in 2014 for an imbalance in death rates, with a hazard ratio of 1.70 (90 per cent CI 1.1 to 2.7) versus placebo. Preserving lean mass did not preserve survival, and directionally went the wrong way.

Where the anti-doping rules stand

The 2026 WADA Prohibited List covers this entire pathway under section S4.3, “Agents preventing activin receptor IIB activation”, prohibited at all times, in and out of competition. The list names activin A-neutralising antibodies, decoy activin receptors such as ACE-031, anti-ActRIIB antibodies such as bimagrumab, agents that reduce myostatin expression, myostatin-binding proteins — follistatin is named explicitly, alongside myostatin propeptide — and neutralising antibodies including apitegromab, domagrozumab, landogrozumab and stamulumab. Substances in S4.3 are non-Specified Substances, which carries the harsher sanctioning presumption. Gene-transfer approaches are covered separately under M3.1.

Frequently asked questions

Is there an approved myostatin drug?

No. As of August 2026 no myostatin- or activin-receptor-targeting drug is approved for a muscle indication anywhere in the world. Apitegromab is under FDA review with a decision date of 30 September 2026. Sotatercept is approved, but for pulmonary arterial hypertension, not for muscle.

Does “follistatin-344” do what its name implies?

The number 344 is the length of the follistatin precursor protein and the label used for the cDNA insert in gene-therapy vectors. It is not a peptide with a clinical dossier. No registered human trial has ever injected follistatin protein, and the preclinical pharmacokinetics explain why: native follistatin is cleared from the circulation far too quickly to work systemically.

Why did the spectacular mouse result not carry over to people?

Because mice and primates weight the two ligands differently. Myostatin dominates the brake in a mouse; in a monkey or a human, activin A shares the job, so removing myostatin alone leaves much of the brake in place.

Is myostatin inhibition banned in sport?

Yes — under WADA section S4.3, at all times, as a non-Specified Substance. Follistatin is listed by name.

Informational only — not medical advice · 21+