1. Introduction: The Wrong Mental Model

Ask most injectors how long botulinum toxin lasts and the answer arrives as a product attribute: three to four months, perhaps longer in the glabella, shorter in the crow's feet, shorter again in a gym-going man with a heavy corrugator. Ask why it stops working and the answer is usually some version of 'it wears off' or 'the body breaks it down'. This is a serviceable clinical heuristic and a poor description of what happens.

Botulinum neurotoxin is not a receptor antagonist that dissociates. The heavy chain binds SV2 and polysialogangliosides on the cholinergic terminal, the complex is internalised, and the light chain — a zinc-dependent endopeptidase — translocates into the cytosol, where it proteolytically cleaves a SNARE protein. For serotypes A and E that target is SNAP-25. Once cut, the protein is cut. There is no reversal reaction, no antidote that reaches the intracellular compartment, and no clinically meaningful way to accelerate recovery in a patient who dislikes their result.

The clinical effect therefore ends for one of two reasons: the intracellular protease is finally degraded and functional SNAP-25 is resynthesised faster than it is destroyed, or the nerve constructs alternative release machinery — new terminal sprouts forming functional contacts with the muscle fibre. In practice both occur, on different timescales, and the balance between them differs by serotype.

The July 2026 European approval of trenibotulinumtoxinE (Boey), the first serotype E product licensed in the EEA, makes this more than an academic distinction. Approved across all 30 EEA countries following Canadian approval in June 2026, it produces visible effect from around eight hours and resolves in approximately two to three weeks. That is not a weaker toxin or a smaller dose. It is a different serotype cutting the same substrate at a different residue, and the duration difference is a direct consequence of that.

2. Why Serotype A Lasts and Serotype E Does Not

BoNT/A cleaves SNAP-25 between residues 197 and 198, removing only nine C-terminal amino acids and leaving a SNAP-25(1–197) fragment. BoNT/E cleaves between 180 and 181, removing twenty-six residues to yield SNAP-25(1–180). This small difference in cut site is the single most important determinant of clinical duration, for two reasons.

First, the fragments behave differently. SNAP-25(1–197) retains sufficient structure to enter SNARE complexes with syntaxin-1A and VAMP but cannot support productive fusion. It therefore acts as a dominant-negative inhibitor, poisoning the assembly rather than simply being absent from it — which explains the long-observed paradox that cleavage of only a minority of the total SNAP-25 pool is enough to abolish release. Work published in 2025 in PLOS Pathogens demonstrated that this fragment is heavily phosphorylated at Ser187, that phosphorylation enhances its binding to syntaxin-1A, and that the fragment outlives the light-chain protease itself in the cell. The truncated SNAP-25(1–180) generated by BoNT/E lacks the domain needed for that stable interaction and does not inhibit in the same way; once the protease is gone, recovery proceeds rapidly.

Second, the light chains themselves have radically different intracellular half-lives. In primary rat spinal cord neurons, enzymatic activity of BoNT/A1, A2, A4 and A5 persisted for at least ten months, while BoNT/A3 resolved at around five months; BoNT/E activity was detectable for only two to three weeks. The LC/A protease is localised to the plasma membrane and appears to evade the ubiquitin–proteasome and autophagic pathways that clear LC/E efficiently. Nothing about dose explains this; it is a structural property of the molecule.

The consequence for practice is worth stating plainly. Serotype E's short duration is a designed-in feature of its molecular biology, not a limitation to be overcome by injecting more of it. A patient asking whether the rapid-onset product can be 'topped up' to last longer is asking for a serotype it is not.

It also follows that the two properties clinicians most want — rapid onset and long duration — are not intrinsically linked to one molecule. BoNT/E enters and acts faster; BoNT/A persists. Chimeric constructs combining the entry and translocation efficiency of E with the light-chain stability of A have been produced experimentally, and represent the most rational route to the product the market keeps asking for.

3. Axonal Sprouting: What It Is and What Triggers It

Intramuscular botulinum toxin induces profuse sprouting of motor nerve terminals, first characterised in detail by Brown, Holland and Hopkins in the early 1980s. Sprouts extend from the parent terminal, from the pre-terminal axon and from nodes of Ranvier, grow beneath the guidance of terminal (perisynaptic) Schwann cell processes, and form new contacts with the underlying muscle fibre.

The critical mechanistic point — and the one that answers the clinical question directly — is that sprouting is not a response to the toxin. It is a response to muscle inactivity. Blocking nerve conduction with tetrodotoxin, which never enters the terminal and never touches SNAP-25, produces sprouting just as reliably. So does surgical paralysis and so does simple disuse. The muscle fibre, deprived of activity, upregulates extrajunctional acetylcholine receptors and secretes sprouting-promoting signals; the classic demonstration by Pestronk and Drachman showed that the amount of sprouting correlates with the level of extrajunctional receptor expression and that blocking those receptors with α-bungarotoxin inhibits sprouting.

This means the muscle, not the nerve and not the drug, is running the repair programme. The nerve is responding to a chemical advertisement placed by a fibre that has stopped receiving instructions.

How much sprouts actually contribute to recovery has been contested and the answer has shifted. In vivo imaging by de Paiva, Meunier, Aoki and Dolly in 1999 described an elegant biphasic sequence in mouse sternomastoid: by day 28, when nerve stimulation again produced contraction, regulated vesicle recycling occurred exclusively in the sprouts and not at the parent terminals — the sprouts were, at that moment, the entire functioning synapse. A second phase then followed, with vesicle turnover returning to the original terminals and the now-superfluous sprouts being eliminated. Later focal recording work by Rogozhin and colleagues in 2008, however, found that the original synaptic sites carry the predominant share of restored transmission, prompting Ko's accompanying commentary in The Journal of Physiology to question how much of functional recovery the sprouts truly own.

The reconciliation most consistent with the data is that sprouts provide an early bridge while the parent terminal is rebuilt, and that the parent terminal ultimately reclaims the junction. Meunier's 2003 study made the further point that both the extent and the duration of this remodelling are dictated by the SNAP-25 truncation site — the same variable that sets duration. Sprouting, in other words, is downstream of the same molecular decision, not an independent clock.

4. Can Sprouting Be Slowed? What the Evidence Actually Shows

Four experimental levers have been demonstrated to alter this process. None is clinically available and it is important to be honest about why.

Muscle stimulation, in reverse. Because sprouting is driven by inactivity, restoring activity abolishes the signal. Brown and colleagues showed as early as 1977 that direct electrical stimulation of botulinum-poisoned mouse soleus prevents motor nerve sprouting. This is mechanistically clean and clinically inverted: the intervention that suppresses sprouting requires artificially contracting the very muscle the patient paid to have relaxed. It is a proof of principle for the mechanism, not a treatment. It does, however, offer a testable hypothesis for why habitually hyperkinetic patients — those who unconsciously recruit residual fibres hard against the block — may report shorter duration than dose alone predicts.

IGF-1 receptor signalling. A 2023 study demonstrated that administering an anti-IGF1R antibody after botulinum toxin inhibited recovery from neurogenic paralysis, with the postsynaptic components of the junction most affected and mTOR/S6 kinase translational signalling implicated. This is the clearest existing demonstration that duration can be pharmacologically extended by targeting the regeneration pathway rather than the toxin. It is also a systemic growth-factor pathway central to muscle maintenance, tissue repair and — a non-trivial concern — tumour biology. Extending a glabellar result from four months to six is not a proportionate reason to interfere with it.

Fragment stabilisation. If the dominant-negative SNAP-25(1–197) fragment is what sustains blockade, and Ser187 phosphorylation stabilises the fragment's interaction with syntaxin-1A, then modulating that phosphorylation is a duration lever that does not touch nerve growth at all. The authors of the 2025 PLOS Pathogens work explicitly raise this for future formulation development. This is the most promising direction precisely because it acts on the toxin's own footprint rather than on the host's repair machinery.

Light-chain persistence engineering. Subtype and serotype selection already produce durations ranging from weeks to many months in the same cell system. Selecting or engineering light chains for slower cytosolic degradation is the least biologically invasive route to longer duration, and is broadly what the long-acting pipeline — peptide-stabilised and novel-excipient formulations among them — is attempting by other means.

Against these sits the one intervention practitioners actually encounter in commerce: zinc and phytase supplementation. The rationale is not absurd, since the light chain is a zinc endopeptidase and requires zinc for catalysis. A small trial published in 2012 reported an increase in duration of roughly a third with four days of pre-treatment zinc plus phytase. It has not been convincingly replicated, the effect size is implausibly large for a nutrient not ordinarily limiting in a Western diet, and it should not be presented to patients as established. If duration is set by protease longevity and terminal remodelling rather than by substrate-limited catalysis, the mechanism is difficult to defend.

5. What Actually Changes Duration in Clinic Today

Stripped of speculation, the variables a practitioner genuinely controls are unglamorous and well established.

Dose is the dominant one. Duration is dose-dependent within the licensed range, and a substantial proportion of complaints about short duration are complaints about under-dosing a muscle whose bulk was underestimated — most commonly a heavy male corrugator or a hypertrophic masseter.

Accuracy of placement matters as much as quantity. Toxin deposited outside the target muscle contributes to the total dose administered, to the antigenic load and to the risk of unwanted effect, while contributing nothing to duration at the intended site. The corollary developed elsewhere in this journal applies here too: entry angle and vector determine where a bolus finishes, and a technically identical dose delivered 5 mm off-target behaves as a smaller dose.

Product selection is a real variable rather than a marketing one, given documented differences in light-chain persistence between subtypes and in formulation between products. It is legitimate to choose a different product for a patient who reliably reports short duration, and to document that reasoning.

Neutralising antibody formation is the most consequential avoidable cause of progressively shortening duration. Risk is associated with higher cumulative dose, shorter inter-treatment intervals and repeated top-up dosing. The patient who returns at eight weeks each time for a small correction is following a schedule that maximises antigenic exposure. Respecting a minimum three-month interval, avoiding routine touch-ups, and using the lowest effective dose remain the only immunogenicity strategy available.

Finally, patient factors — muscle mass, metabolic rate, habitual expressiveness, prior treatment history — modulate the same underlying process by modulating the inactivity signal and the number of junctions requiring blockade. They are worth recording, because they make the third short-duration complaint interpretable rather than mysterious.

6. Implications of Short-Duration Toxin for Practice

The arrival of a licensed short-duration serotype E product changes the consultation more than it changes the pharmacology. Three points deserve to be made explicitly to patients.

The first is that short duration is the product, not a failure of the product. A patient who returns at three weeks disappointed that a rapid-onset toxin has resolved has not had a poor result; they have had the licensed result. Consent and expectation-setting must state the two-to-three week window before treatment, not after.

The second is that a short-acting trial is a genuine clinical good. The most common reason a first-time patient declines treatment is fear of an irreversible-feeling change to their face for a season. A product that resolves within weeks converts an irreversible-feeling decision into a reversible one, and that is the correct use case: the anxious first-timer, the patient with an event and no treatment history, the patient exploring whether they tolerate the aesthetic of a relaxed glabella.

The third is that it should not be positioned as a safety intervention. Short duration does not mean low risk. The onset, spread and adverse-effect profile of a neurotoxin do not become benign because the effect resolves sooner, and the same anatomical discipline applies.

There is also a subtler point that will be missed in the marketing. Repeated short-interval exposure to any neurotoxin raises the immunogenicity question, and a product designed to be used 'occasionally', as the European product information puts it, invites exactly the frequent-redosing behaviour that antibody formation punishes. Practitioners should track cumulative exposure across serotypes in the same patient rather than treating them as independent accounts.

7. Conclusion

The question 'how long does Botox last?' has a more interesting answer than the number we usually give. Duration is set by how long the light-chain protease survives inside the terminal, how long the dominant-negative SNAP-25 fragment persists, and how quickly the muscle's inactivity signal recruits axonal sprouts and rebuilds the junction. The molecule does not wear off; the nerve works around it.

Sprouting can be slowed. It has been slowed experimentally by restoring muscle activity, and recovery has been delayed pharmacologically by blocking IGF-1 receptor signalling. Neither is a candidate for cosmetic use, and the reason is proportionality rather than feasibility: interfering with a systemic nerve and muscle repair programme to extend a glabellar result is not a defensible trade.

The rational path to longer duration runs through the toxin, not the patient — light-chain stability, cleavage-site selection, and stabilisation of the cleaved fragment. The rational path available today runs through correct dose, accurate placement, appropriate product selection and disciplined intervals to protect against neutralising antibodies.

And the short-duration serotype E products now entering European practice are best understood not as a weaker option but as evidence for the whole argument. Change the residue you cut, and you change the biology of recovery. That is what duration has always been.