This category has a nickname no label will ever print. It belongs to an injectable prescription drug with a trademark and a legal department, so the carton says “expression lines” and “relaxing” and “the look of”, and the shopper finishes the sentence. As copywriting that’s efficient. As a claim it skips the only question worth asking: what would a molecule sitting on the stratum corneum physically have to do before it changed a line across your forehead?
The answer runs through a protein zipper, and the zipper is far more interesting than the marketing on top of it.
A dynamic line is a record of folding
Formulators sort wrinkles early. A static line sits there whatever the face is doing. A dynamic one, the vertical pair between the brows or the fan at the outer eye corner, appears when a specific muscle shortens, running perpendicular to the pull. Corrugator supercilii draws the brows together. Frontalis creases the forehead. Orbicularis oculi purses the skin around the eye every time someone squints at a screen.
Fold the same sheet along the same crease a few thousand times a week for thirty years and the sheet keeps the crease.
That leaves two places to intervene: the skin that folds, or the pull that folds it. Nearly everything in a serum works on the first. This category claims the second, which is a nerve problem.
The pull comes down to a four-helix zipper
Skeletal muscle has no opinions. It shortens when a motor nerve terminal releases acetylcholine into the synaptic cleft, and that release isn’t a leak. Acetylcholine sits packaged in vesicles, and a vesicle has to fuse with the terminal membrane before any escapes.
Fusion is the hard part, because two lipid bilayers a few nanometres apart don’t want to merge. The proteins that force the issue are the SNAREs: syntaxin-1A and SNAP-25 on the nerve terminal, synaptobrevin on the vesicle. Sutton and colleagues crystallised the core complex at 2.4 angstrom resolution for a 1998 Nature paper: a twisted parallel four-helix bundle, leucine-zipper-like layers along its length, an ionic layer of one arginine and three glutamines at the centre. The helices zip from the far end inward, dragging the membranes together.
One detail makes the rest of this make sense: SNAP-25 contributes two of those four helices. Half the machine is a single protein.
One approach dismantles the machine, the other tries to crowd it
Botulinum toxin injections don’t block a receptor or calm a muscle. They perform surgery on that protein. Blasi and colleagues established the mechanism in a 1993 Nature paper, showing that the type A neurotoxin acts as a zinc-dependent protease that selectively cleaves SNAP-25 near its carboxy terminus. Later work pinned the bond: Q197-R198, which lops nine residues off a 206-residue protein.
Nine. Out of 206.
A 2017 paper in Toxins from Montecucco’s group in Padua notes in passing that a satisfactory rationale for why removing only nine residues is sufficient has still to be provided. Three decades in, the fine mechanism isn’t closed, in a field far better funded than cosmetic chemistry.
The synthetic peptides in this family were drawn from the opposite direction: not an enzyme that destroys a SNARE protein, but a short sequence copied from SNAP-25 itself, present in enough quantity to get in the way. A 2025 review in the International Journal of Molecular Sciences describes acetyl hexapeptide-8, the shorter and better studied member, in exactly those terms: modelled on the N-terminus of SNAP-25, competing with it for binding to synaptobrevin and destabilising assembly of the complex. Acetyl octapeptide-3 is the longer sibling of that idea, and the same sequence lives in two very different inventories. On a cosmetic ingredient list it carries an INCI name. In laboratory supply it’s catalogued as a research compound, sold by the vial with a certificate of analysis and a disclaimer that it isn’t for food, drug or cosmetic use.
Competition and proteolysis are not the same order of claim. An enzyme is catalytic: one molecule cleaves substrate after substrate, and the effect lasts until the neuron rebuilds the protein. A competitive mimic is not catalytic. It has to be present, in quantity, at the site, continuously, and the interference ends when it diffuses away. One is destruction. The other is crowding.
Worth saying plainly while research catalogues are on the table: substances with no human approval anywhere fall under WADA’s S0 non-approved substances category, which captures essentially everything sold as a research material. Anyone subject to testing should read that as a closed door.
Then it has to actually get there
The site of action is a motor nerve terminal at the neuromuscular junction: inside muscle, under the dermis, under the epidermis, under the stratum corneum, whose job is stopping water-soluble molecules crossing it.
The 2025 review is blunt here. Across the permeability studies it gathered, 0.22 percent of the applied peptide reached the stratum corneum while 99.7 percent was recovered from the surface, and in one experiment none was detected in the receptor fluid at all. Reaching the outermost dead layer isn’t the same as crossing it, and crossing it isn’t the same as reaching a synapse.
What the record supports, and what it doesn’t
The same review states the uncomfortable part directly: none of the in vivo studies it examined confirmed any inhibitory effect on muscle contraction. Wrinkle-depth measurements exist, but from small, short trials with the ingredient’s developers close to the work, and a change in a photographed crease is not evidence that a synapse was touched. The reviewers go as far as hypothesising that some different mechanism, related to permeation into the epidermis, explains it.
So the mechanism is real, elegant and well characterised at the bench. What hasn’t been shown is that any of it happens in a face.
That gap isn’t a scandal. It’s ordinary, and it’s where most ingredient stories sit if you read far enough down. The move is to notice which half of the sentence the evidence actually covers.
