Neurotrophic Factors: NGF, BDNF and the Neurogenesis Question

Four letters do an enormous amount of work in writing about cognitive research compounds: BDNF. Raise it, the argument goes, and you get new neurons, better memory, a more plastic brain. Each link in that chain is a real research subject. None of them is as settled as the shorthand implies — and one of them, whether adult humans make new hippocampal neurons at all, has been the subject of a public scientific argument for the past eight years that is not yet over.

This guide covers the neurotrophin family properly: where NGF and BDNF came from, which receptors they use, why the precursor forms do the opposite of the mature ones, what the neurogenesis dispute is actually about, which neurotrophin-based medicine is approved and why it worked where others failed — and then, compound by compound, exactly what the published work on Semax, Selank and P021 measured.

Discovery: sarcomas, snake venom and a misquoted prize

Nerve growth factor was the first growth factor found, and the route to it is more interesting than the usual one-line version. In 1951 Rita Levi-Montalcini and Viktor Hamburger reported that a transplanted mouse sarcoma selectively stimulated the sensory and sympathetic nervous system of the chick embryo. In 1954 Stanley Cohen, with Levi-Montalcini and Hamburger, isolated a nerve-growth-stimulating factor from the sarcoma tissue itself. In 1956 Cohen and Levi-Montalcini found the same kind of activity in snake venom, and in 1960 Cohen purified a nerve-growth-promoting protein from mouse salivary gland.

So NGF was not “discovered in snake venom”, a claim that circulates widely. The discovery was made in a tumour-and-chick-embryo system; venom and salivary gland were unusually rich sources that made purification practical.

The 1986 Nobel Prize in Physiology or Medicine went jointly to Cohen and Levi-Montalcini, and the official citation reads “for their discoveries of growth factors” — not “for the discovery of nerve growth factor”. That is the single most common misquote of this prize.

BDNF arrived three decades later. Barde, Edgar and Thoenen published “Purification of a new neurotrophic factor from mammalian brain” in The EMBO Journal in 1982. Two notes on citing it: the acronym BDNF does not appear in the title, and the record we could verify does not state the source species — so we say mammalian brain, not pig brain, however often the latter appears.

The family and its receptors

There are four human neurotrophins and, usefully, their receptor preferences are recorded in curated protein entries rather than resting on recollection.

That last row is the one most worth noticing. The receptor shared by the whole family belongs to the same superfamily as the TNF receptors discussed in our NF-κB guide — and like them, it can route either to survival or to death depending on what else is in the complex.

Why precursors matter

Neurotrophins are made as larger precursors. The intuitive assumption is that the precursor is an inactive intermediate, trimmed to release the real signal. That assumption is wrong, and the correction is the single most useful thing in this article for reading BDNF claims.

A 2005 study in The Journal of Neuroscience established it directly. Working in rat sympathetic neurons, neurons from p75NTR-null mice, cultured rat cortical neurons and TrkB-expressing cells, the authors reported that proBDNF is an apoptotic ligand that induces death at subnanomolar concentrations, and that mature BDNF, but not proBDNF, is effective in inducing TrkB phosphorylation. proBDNF killed sympathetic neurons roughly ten to twenty times more potently than mature BDNF did, the death required both p75NTR and sortilin, and cortical neurons were shown to secrete proBDNF — so the precursor is a real extracellular species, not merely an intracellular stage.

Put the two halves together. Mature BDNF acts through TrkB and supports survival and plasticity. proBDNF acts through p75NTR with sortilin and triggers apoptosis. They are the same gene product in different states, with opposite consequences.

Which means that “raise BDNF” is not a coherent objective unless the form is specified. And, as the measurement section below sets out, most commercial assays do not specify it.

Val66Met: solid cell biology, shaky associations

The BDNF variant rs6265, a valine-to-methionine substitution at position 66, is the most discussed polymorphism in this literature, and it is a good case study in how the strength of evidence can differ enormously between two halves of the same story.

The cell biology holds up. A 2004 study in The Journal of Neuroscience found that when the Met variant is expressed by itself in hippocampal neurons, less BDNF is secreted in an activity-dependent manner; and that when both variants are present in the same cell, around 70% of the variant BDNF forms Val·Met heterodimers, which are inefficiently sorted into secretory granules, producing a quantitative decrease in secretion. The defect was cell-type specific, absent in non-neuronal cells, and it was a problem of quantity rather than quality — the forms themselves looked normal. A single Met allele drags wild-type BDNF into poorly sorted heterodimers.

The human association literature is much weaker. A meta-analysis of 14 studies, with 2,812 cases of major depressive disorder and 10,843 non-depressed controls, found that in the total sample the Val66Met polymorphism was not significantly associated with depression; a signal appeared only in men, and stratifying by ethnicity showed no significant effects. For hippocampal volume, a joint meta-analysis found no association in a new sample of 643 healthy young subjects, a small effect across roughly 5,300 people that appeared only for manually traced and not automated volumes, evidence of publication bias, and effect sizes that shrank in larger and more recent samples. A 2015 review in the field has the word “controversy” in its title and names small effect sizes among its unresolved issues.

So: the trafficking mechanism is well supported. The downstream claims that Val66Met causes low BDNF, depression, poor memory or a smaller hippocampus are not.

The neurogenesis question

This is the part of the chain that most popular writing treats as settled, and it is the part that is least settled.

The case for. In 1998 Eriksson and colleagues reported in Nature Medicine that new neurons, as defined by their markers, are generated from dividing progenitor cells in the dentate gyrus of adult humans, concluding that the human hippocampus retains its ability to generate neurons throughout life — and note their own hedge, “as defined by these markers”. In 2013, Spalding and colleagues used nuclear-bomb-test-derived carbon-14 in genomic DNA to estimate that roughly one third of hippocampal neurons are subject to exchange, with around 700 new neurons added per day in each hippocampus and about 1.75% annual turnover in the renewing fraction. In 2018, Boldrini and colleagues published a paper titled “Human hippocampal neurogenesis persists throughout aging”. In 2019, Moreno-Jiménez and colleagues reported thousands of immature neurons in the dentate gyrus of neurologically healthy subjects into the ninth decade, declining as Alzheimer’s disease advanced.

The case against. Also in 2018, Sorrells and colleagues published in Nature under the title “Human hippocampal neurogenesis drops sharply in children to undetectable levels in adults”. Their conclusion was that neurogenesis in the dentate gyrus does not continue, or is extremely rare, in adult humans, and that a defined population of progenitor cells does not coalesce in the subgranular zone. Methodologically it was a serious effort: post-mortem tissue from fetal through adult, plus surgically resected hippocampi from epilepsy patients, which sidesteps post-mortem objections; immunohistochemistry for six markers; electron microscopy with immunogold; RNAscope; and rhesus macaque comparison.

In March 2021 the two groups published back-to-back perspectives in the same issue of The Journal of Neuroscience, which is the best single place to see what the disagreement is actually about. From those and the surrounding literature, four distinct and documented reasons the studies disagree:

One explanation worth not reaching for: post-mortem interval. The group best placed to test it reported no variation in marker-positive cell numbers, staining quality or distribution across their delays, while calling for study of longer intervals.

Where it stands now. In July 2025, a Science paper from the same laboratory as the 2013 carbon-14 work reported identification of proliferating neural progenitors in the adult human hippocampus, using single-nucleus RNA sequencing together with Ki67 antibodies and machine learning, placing progenitors within the dentate gyrus. That directly answers the challenge the sceptical group had set — identify dividing neuronal precursors — and it is a meaningful shift. It is also one study, and a 2025 review still describes the existence of newly generated immature dentate granule cells as under debate. The accurate summary is contested, with recent evidence favouring low-rate persistence — not settled in either direction.

Worth remembering throughout: every one of these studies is post-mortem or surgical tissue. There is no way to count new neurons in a living person, which is why the sceptical perspective says settling the question needs new methods to evaluate the age of a neuron at single-cell level. Our guide to reading a peptide study covers how to weigh this kind of evidence, and clinical trial phases explained covers what a human trial would have to look like.

What happened when neurotrophins were tried as medicines

A claim circulates that no neurotrophin has ever become a medicine. That is false, and the true version is more instructive.

Cenegermin is approved. Marketed as Oxervate, it is a recombinant form of human nerve growth factor produced in Escherichia coli, authorized by the European Medicines Agency on 6 July 2017 and by the US Food and Drug Administration on 22 August 2018, indicated for the treatment of neurotrophic keratitis. Note the delivery route: eye drops, applied topically to an exposed surface tissue. That is precisely why it works where systemic attempts did not.

The systemic trials failed. A phase III trial of recombinant methionyl human BDNF in amyotrophic lateral sclerosis, in 1,135 patients over nine months, failed to show benefit on the primary endpoints; survival across all groups was higher than the planners expected, which reduced statistical power, and the authors noted that trials using different delivery routes were already under way — the field’s own reading was that delivery was the problem. A phase III trial of recombinant human NGF in diabetic polyneuropathy, with 1,019 patients across 84 US centres, likewise did not demonstrate a beneficial effect, in contrast to earlier phase II results. Its most mechanistically telling finding was the safety profile: generally safe apart from injection-site pain or hyperalgesia.

And blocking NGF had the mirror-image problem. Tanezumab, a monoclonal antibody targeting NGF, was developed for osteoarthritis pain. The EMA refused its marketing authorisation, with a CHMP opinion dated 16 September 2021: although it showed better pain relief and improved physical functioning than placebo, there was no improvement compared with NSAIDs, patients had an increased risk of side effects such as rapidly progressive osteoarthritis and joint replacement, and the benefits did not outweigh the risks. The FDA convened a joint advisory committee on it in March 2021; tanezumab is not an approved medicine anywhere we could document. (One 2026 review states incorrectly that it was FDA-approved — a reminder to check secondary sources.)

The symmetry is neat and worth stating plainly: add NGF and you get pain; block NGF and joints degrade. A signalling molecule that does something important is rarely safe to push hard in either direction.

What blood BDNF actually measures

Almost every human study reporting that something “increased BDNF” measured it in serum or plasma. That number is not what most readers assume.

• Ninety per cent or more of blood BDNF is stored in platelets. The amount in serum is nearly identical to the amount found in washed platelet lysates — so serum BDNF is substantially a readout of platelet release, not of brain output. Only around 30% of platelet BDNF is secreted into blood, and plasma BDNF has very low retest stability.
• Drawing the blood can release it. Platelet BDNF is released by thrombin, collagen, calcium ionophore and shear stress — and a needle applies shear stress. Serum BDNF is also around a hundredfold higher than plasma.
• The assays disagree with each other. A comparison of six commercial assays on sera from 40 healthy adults found inter-assay variation ranging from 5% to 20%, higher than manufacturers declared in all but one case, and judged only two of the six to give reliable measurements of human serum BDNF.
• And most kits cannot tell proBDNF from mature BDNF — the two forms from the section above with opposite effects. In that six-assay comparison, only two kits selectively recognised the mature form; the others also reacted with the precursor.

This is why the familiar chain — exercise raises BDNF, therefore new neurons grow — does not hold end to end in humans. A meta-analysis of 29 studies and 910 participants found exercise training did raise resting peripheral BDNF, with a small-to-moderate effect that was significant for aerobic but not resistance training. And then its authors wrote the decisive sentence themselves: it cannot be inferred that BDNF concentration is increased in the brain. They add that evidence linking peripheral to central BDNF in humans is limited, that animal blood-brain-barrier data cannot be assumed to transfer, and that serum and plasma measure different pools. Combine that with the previous section and the popular chain is three unproven steps past the evidence: a small rise in blood BDNF, of unknown relationship to brain BDNF, feeding into a neurogenesis process whose very existence in adult humans is contested.

What the catalogue compounds actually showed

Four compounds in this catalogue carry neurotrophic claims. Here is what the published work measured, with the species, tissue and route stated, because that is where the claims and the evidence part company.

A note on P021’s description, since primary sources conflict: the 2017 paper calls it a pentapeptide and the 2024 paper calls it a tetrapeptide, while both give the same sequence. We describe it as a short peptide derived from CNTF residues 148–151 and leave the prefix off.

The general construction worth insisting on, for any compound in this class: in this species, by this route, researchers measured this marker and reported this direction — and there is no evidence the same happens in humans. Our guides to nootropic peptides and Semax versus Selank cover the compounds themselves, and neuropeptides explained covers the wider class.

What’s established and what isn’t

Sourcing neuro research reagents in Canada

The documentation questions for neuro work are the usual ones, with two that bite particularly hard for short modified peptides.

• Identity testing matters for modified variants. An acetylated or amidated analogue differs from its parent by a small mass, so a mass-spectrometry identity check is the step that distinguishes them. See reading a mass spectrum and reading an HPLC chromatogram.
• Match the batch number on the certificate to your vial — our COA guide covers what else a useful certificate contains.
• Account for net peptide content, since synthetic peptides arrive as salts carrying counter-ions and water: net peptide content explained.
• Store cold, dry and dark, and keep material lyophilized until needed: the storage guide, what lyophilized means and peptide stability and degradation.
• Be sceptical of mechanism claims on product pages, including ours. Our guide to spotting fake research peptides covers supplier red flags; the table above is the same scepticism applied to the literature.

SemaxThe compound with the best-characterised rodent BDNF and trkB dataView product N-Acetyl SemaxThe acetylated variant; the published neurotrophic work used standard SemaxView product SelankTwo rodent BDNF papers, one of which points the other wayView product P21 (P021)A CNTF-derived peptide; mouse neurogenesis data, contested BDNF dataView product

Frequently asked questions