BDNF-P21: P021, a four-amino-acid fragment that raises BDNF and works when eaten

The name on the vial and the name in the papers are different

Search the literature for "BDNF-P21" and you will find almost nothing. The compound in every study below is called P021, and its written-out sequence is Ac-DGGL(A)G-NH2 — four amino acids with a chemical cap on each end.

The vendor name is a description of what the molecule does downstream, not its identity. P021 is not a piece of BDNF and it is not a BDNF copy. It came out of a different protein, and it raises BDNF indirectly. Anyone comparing prices or reading research needs the number, not the marketing name, or they will find nothing and assume nothing exists.

Four amino acids cut out of a different growth factor

P021 was designed in Khalid Iqbal's laboratory at the New York State Institute for Basic Research in Developmental Disabilities on Staten Island, and the rights sit with a small company, Phanes Biotech. The starting material was ciliary neurotrophic factor, or CNTF — a growth factor that keeps nerve cells alive and pushes new ones to mature. The researchers identified the biologically active stretch of CNTF and built a four-amino-acid mimic of just that stretch.

So the chain is: a fragment modelled on CNTF's active region, which in a nerve cell drives more BDNF to be made, which then acts on nerve connections. Two steps, both documented, with the compound sitting at the top rather than in the middle.

Why a small piece instead of the whole protein

Three separate attempts to treat brain disease by giving people whole growth-factor proteins failed, and they failed for the same three reasons every time. The proteins could not cross out of the blood into the brain. They were cleared from the body too fast to matter. And the doses needed to overcome both problems caused serious side effects.

That is the entire reason small mimics exist. A four-amino-acid molecule with caps on both ends is small enough to reach brain tissue, stable enough to survive the trip, and — the part that matters most in practice — it works mixed into food. Every long study below was run by putting P021 in the animals' diet, not by injecting them.

For a compound bought online, oral dosing is a genuine practical advantage over almost everything else in this category. It is also the reason the compound is sold as tablets rather than as a powder and a syringe.

What it does inside a nerve cell, step by step

The measurements in the mouse studies trace a specific chain, and each link was recorded separately rather than assumed:

• BDNF levels rise in the brain tissue.
• CREB, a switch that turns on the genes needed to build a synapse, gets phosphorylated — which is the chemical marker of it being switched on.
• Proteins found on the sending side of the synapse stop disappearing, which in an untreated animal they do.
• GSK3β activity falls. That enzyme is what puts too many phosphate groups onto tau, the structural protein that tangles in Alzheimer's disease.

That last item is why the researchers describe P021 as possibly disease-modifying rather than merely propping up function. Raising a growth factor could produce a temporary benefit that disappears when you stop. Turning down the enzyme that damages tau is a change to the process itself.

Twelve months of it in the diet of a mouse bred for Alzheimer's

The longest experiment: triple-transgenic Alzheimer's mice — animals carrying three human disease mutations — plus matched normal mice, fed P021 or a plain diet for twelve months, starting at nine to ten months of age, which in that model is well into established disease.

What changed: abnormal tau phosphorylation and tau build-up fell significantly at the specific sites where Alzheimer's tangles form.

What barely changed: the amyloid side. The effect on amyloid beta was limited to a fall in the soluble form, with the plaques essentially unmoved.

Both halves of that result matter. A compound that reduced tau damage across a year of treatment in an animal that was already sick is a real finding. A compound that leaves plaques alone is not the cure the category gets sold as.

Started at birth instead of at symptoms

A second experiment ran the same Alzheimer's model the other way round: treatment in the diet from birth to day 120 of life, long before any symptoms appear. Mice treated that way did not develop the cognitive impairment the model produces, and the molecular markers of nerve-connection health were higher.

That is a prevention result in an animal with a known genetic fate, and it says something useful about the mechanism: the compound appears to act on how connections are built and maintained, rather than clearing away damage after the fact.

The Down syndrome experiment, which is the strongest result here

The most striking study used Ts65Dn mice, the standard model of Down syndrome, which show developmental delay in infancy and Alzheimer-type memory problems in adulthood. Treatment ran from before birth through early infancy — given to the mothers, then to the pups.

The reported outcome: the developmental delay was rescued, and the memory failures that appear in adult life did not appear. Alongside that, the loss of sending-side synaptic proteins was prevented, GSK3β activity fell, and BDNF and switched-on CREB were higher — measured twice, at three weeks old and again at about seven months.

The design is what gives that result its weight. The treatment stopped in infancy and the benefit was still measurable in middle age. That is not a drug propping up a symptom while it is in the body; it is an early intervention changing how a brain got built.

It is also, without qualification, mice.

The 2024 test in a mouse with a broken CDKL5 gene

The most recent work took P021 into another genetic disorder. Mutations in the CDKL5 gene cause a severe childhood epilepsy with motor and intellectual disability. Without that gene, nerve cells multiply, survive and mature poorly, and earlier work had shown that anything raising brain BDNF improves those animals.

The researchers tested P021 in cells and in the CDKL5 knockout mouse for exactly that reason: it was already documented as a BDNF-raiser in the Alzheimer's and Down syndrome models. That study is worth knowing about because of what it establishes about the compound's identity — three independent research groups now use it as a reliable way to raise brain BDNF in an animal. Its reputation inside the literature is as a research tool, and a well-behaved one.

What has never been done: no person, no injury, no trial

A search of the clinical trials registry for P021 and for BDNF-P21 returns no study of this compound. Not a Phase 1, not a safety study, not a dose-finding study. It is not approved anywhere and it has not been submitted for approval.

Beyond that, there is a second gap. Every study is a genetic disease of brain development or degeneration: Alzheimer's mutations, trisomy, a missing CDKL5 gene. There is no study of P021 in an injured nerve of any kind — not a crushed nerve, not a cut nerve, not a spinal cord injury, not a concussion, not nerve damage from chemotherapy or diabetes.

Where a compound has no safety record in people, unknown is not the same as safe. Nobody has measured what a year of raising BDNF does in a healthy adult brain, and BDNF signalling is not a system with only good outputs — the same pathway is involved in pain sensitisation.

The reasoning people use to get from these mice to a person, and where it breaks

The argument runs: damaged nerve tissue is short of BDNF, BDNF is what rebuilds nerve connections, P021 raises BDNF and can be swallowed, so P021 should help nerve tissue recover.

Every step in that chain is separately defensible. The chain has never been tested end to end, and there are two specific places it could fail.

First, all the measurements are in brain tissue — hippocampus and cortex. A peripheral nerve is a different cell type with a different repair programme, and nothing in this literature measured one.

Second, the timing is inverted. Both of the strongest results came from treating before damage appeared. The one study that treated an animal already sick got a partial result: tau improved, plaques did not. Damage that is already years old is the second situation, not the first.

The finding that cuts against the whole idea: the growth factor it raises has a second job

Raising BDNF is presented as unambiguously good. In nerve pain it is not, and the evidence for that is stronger and more specific than the evidence behind this compound.

A 2024 review in Neuroscience by Peter Smith lays out the pathway. When a peripheral nerve is injured, its terminals in the spinal cord release signals that change the local immune cells — microglia — so they start carrying a particular receptor, P2X4. When nerve-released ATP hits that receptor, those microglia release BDNF. That BDNF then makes the excitatory neurons in the spinal cord's dorsal horn fire more easily and makes the inhibitory ones fire less easily, and it shifts the chloride balance inside neurons in a way that turns a normally calming signal into an exciting one. The result is amplified pain from a normal input.

The review's title is blunt about the whole line of work: BDNF is "the culprit that cannot be apprehended". And it reports something that ought to be in every conversation about this class of compound: in males, BDNF plays an obligatory role in the onset and maintenance of nerve pain, and in females it does not.

So the same molecule builds nerve connections in the brain and amplifies pain signalling in the spinal cord, and which of those you get depends on where it goes up, which cells release it, and — for pain specifically — sex. A compound whose entire mechanism is "raise BDNF" is running toward both effects at once, with no way to steer between them. Nobody has measured which one wins in a person, because nobody has given this compound to a person.

Where it sits next to the other BDNF-raisers sold beside it

Three other compounds on this site work through the same growth factor, and the differences are worth having straight.

Semax has been a hospital medicine in Russia for decades and has human data — small, mostly Russian, including a 110-patient stroke study where blood BDNF rose. Its rat measurement is a 1.4-fold rise in brain BDNF from a single dose into the nose. It also has one 2025 mouse study of a crushed spinal cord, which is the only nerve-injury result in this group.

Selank keeps BDNF up in a different setting: in rats given alcohol long-term, seven days of it protected memory and kept BDNF up in the hippocampus and prefrontal cortex. Its human record is anxiety, not nerve repair.

P021 has no human data at all, and the most rigorous animal designs of the three — treat before damage, measure decades-equivalent later, watch the marker stay changed.

Read together, the group has an awkward shape. The compound with the best experimental design has never been near a person. The compound with human data has the weakest mechanism numbers. Nobody has run a head-to-head comparison of any two of them, in any species.

Oral, which is unusual and is the point

Almost everything sold in this category is a powder you mix with water and inject. P021 was designed to work in food, and every long-term study delivered it that way, which is why it is sold as tablets.

Practically, that removes the two most common failure points in self-administered peptide use: mixing arithmetic and injection technique. It does not remove the more basic problem, which is that no human dose exists to copy. The mouse studies dosed by mixing a set concentration into diet across months. There is no published translation of that into a milligram figure for a person, and any number on a label is the seller's, not a researcher's.

What it would take to know

A Phase 1 study in healthy adults would settle the safety question — what dose is tolerated, what happens to it in the body, whether measurable BDNF changes at all in a person. It has not been run, on a compound with mouse data going back to 2014.

For the nerve-injury question the study is different and harder: an animal model of actual nerve damage — a crushed sciatic nerve is the standard — with recovery measured against untreated controls. That study is straightforward, cheap by the standards of this field, and appears nowhere in the literature. Until somebody runs it, the case for using this compound on an injured nerve rests entirely on a mechanism, and the honest version of that sentence is that a mechanism is a hypothesis with a diagram attached.

There is a second study that would cost almost nothing and would settle the question that matters most on this page. Give the compound to male and female animals with an injured nerve, and measure pain behaviour rather than memory. If raising BDNF helps the nerve repair, pain should fall as function returns. If the spinal-cord pathway dominates, pain should rise, and rise in males more than females. Those two outcomes point in opposite directions and the experiment separates them in weeks. Twelve years after the first mouse paper, nobody has published it.

Until then, what can be said about BDNF-P21 is narrow and worth stating without decoration. It is a four-amino-acid piece modelled on CNTF that reliably raises BDNF in mouse brain tissue and can be eaten rather than injected. In three separate genetic diseases of the brain, given early, it changed how the animals' brains developed and the change outlasted the treatment. In an animal already sick it did about half of what it does when given early. Nobody has measured it in a person, nobody has measured it in an injured nerve, and the growth factor it raises has a documented second job of making nerve pain worse.

Seven published studies are attached to this page, along with the registry search that returns nothing. Every claim above traces to one of them.