BDNF & neuroplasticity

BDNF is a member of a small family of secreted proteins that also includes nerve growth factor, neurotrophin 3 and neurotrophin 4 ^[1]. It is synthesised as a precursor, proBDNF, and cleaved to the mature form. That detail is not bookkeeping, because the two forms do close to opposite things: mature BDNF acting at TrkB promotes survival, growth and synaptic strengthening, while the uncleaved precursor acting through the p75 receptor complex pushes toward pruning and cell death. A measurement or a claim that does not distinguish them is ambiguous by construction, and most of them do not.

The receptor is TrkB, the product of the NTRK2 gene, a receptor tyrosine kinase. Mature BDNF binds, the receptor dimerises, the two halves phosphorylate each other, and three downstream cascades run from there: phospholipase C gamma, the Ras and MAPK route, and PI3K to Akt to mTOR. That last one reappears below, because it is the arm ketamine engages.

The single most important property of the system is its activity dependence. BDNF transcription rises with neuronal firing, its secretion is triggered by depolarisation, and its actions are gated by the state of the target synapse ^[1]. This is why exercise and learning are the best-evidenced levers, and why a pill that raises a circulating concentration is not obviously doing the same thing as a circuit that is working hard.

The clearest human demonstration of that principle is a common genetic variant. Val66met substitutes a methionine for a valine in the pro-domain of the protein. Carriers of the met allele showed poorer episodic memory, abnormal hippocampal activation on functional MRI and lower hippocampal N-acetylaspartate. In transfected neurons, met-BDNF showed lower depolarisation-induced secretion while constitutive secretion was unchanged, and it failed to localise to secretory granules or to synapses ^[3]. The variant does not reduce how much BDNF a neuron makes. It reduces whether the protein is in the right place to be released at the right moment, and that alone is enough to show up in memory testing.

BDNF and the long-term potentiation that builds memories are tightly linked; BDNF is part of what makes the synaptic strengthening of learning persist rather than decay. See NMDA, glutamate & memory for the machinery it acts on.

The larger claim is the neurotrophic hypothesis of depression. Stress reduces BDNF expression in limbic structures that control mood, antidepressant treatment reverses or blocks that reduction, and the reduced neurotrophic support could contribute to the atrophy of hippocampus and prefrontal cortex observed in depressed people ^[2]. That is a coherent framework and the review that laid it out was explicitly a critical examination rather than a victory lap; the authors noted the limitations of the preclinical models and imaging studies it rests on.

The strongest causal evidence came from ketamine. Blocking NMDA receptors at rest deactivates eEF2 kinase, which reduces eEF2 phosphorylation and de-suppresses the translation of BDNF; the fast antidepressant-like behavioural effects in mice depend on that rapid BDNF synthesis, and inhibitors of eEF2 kinase alone reproduce the behavioural effect ^[8]. In parallel, ketamine rapidly activates mTOR, raising synaptic signalling proteins and the number and function of new spine synapses in rat prefrontal cortex, and blocking mTOR abolishes both the synaptogenesis and the behavioural response ^[7].

Those two papers are the reason BDNF is treated as a mechanism rather than a correlate. A molecule whose removal abolishes an effect is doing work; a molecule that merely goes up alongside an effect is not yet.

What is not settled is whether raising BDNF is sufficient. The obvious therapeutic move, giving BDNF itself, has never worked, and the reason is pharmacokinetic rather than conceptual: the protein does not cross the blood-brain barrier usefully, it distributes poorly, and delivery has been the central obstacle to every BDNF-based programme in neurological and psychiatric disease ^[11][17]. Hold on to that, because it is exactly why the next best idea became so attractive.

The most reliable levers are not pills, and the gap is not close. A meta-analysis of 29 studies covering 1,111 participants found a moderate effect of a single exercise session on BDNF, with a Hedges' g of 0.46. Regular training intensified the response to a subsequent session, g of 0.59, and produced a smaller rise in resting BDNF, g of 0.27. Sex moderated the effect: studies with more women showed less BDNF change ^[4].

Learning and sustained cognitive work belong on the same list, and they follow directly from the activity dependence described above rather than from a separate literature ^[1]. This is the least glamorous section of this page and the one with the best evidence behind it.

On the drug side, the honest ranking runs: strong preclinical causal evidence for ketamine ^[7][8], reasonable preclinical evidence for psychedelics ^[9], an established but slow effect for classic antidepressants ^[2], and thin or single-study evidence for everything sold as a BDNF supplement.

Semax has a real and specific animal result: intranasal dosing raised BDNF protein in rat basal forebrain within three hours and not in cerebellum, with specific binding sites in that region at 2.4 nanomolar ^[18]. One brain region, one species, one route, and no human BDNF data. See peptides & bioregulators.

Lion's mane is routinely miscited here. Its neurotrophic story concerns nerve growth factor rather than BDNF, and its one placebo-controlled human trial measured a cognitive scale and no neurotrophin at all ^[19].

7,8-DHF is the one everyone asks about, and it gets its own section below.

Almost every human BDNF number comes from blood, and two separate problems sit between that number and the brain.

The first is whether blood reflects brain. The most direct test compared blood, serum, plasma and brain tissue across three species. Whole-blood BDNF correlated with hippocampal BDNF in rats with an r-squared of 0.44, plasma BDNF correlated with hippocampal BDNF in pigs at 0.41, and in mice frontal cortex correlated with hippocampus at 0.81. The authors concluded that blood measures do reflect brain tissue levels ^[5]. Read the same numbers the other way and an r-squared of 0.44 means more than half the variance is unexplained, and the study also found BDNF undetectable in mouse blood entirely and unmeasurable in pig whole blood with a commercial kit. The relationship is real, modest and species-dependent.

The second is whether the assay works. Six commercial assays were run on sera from the same 40 healthy adults. All showed complete sample recovery and comparable ranges, but inter-assay variation ranged from 5% to 20%, and every kit except one exceeded the variation its manufacturer declared. Two of the six recognised mature BDNF selectively; the others reacted with proBDNF as well ^[6]. Only two were judged suitable for reliable measurement.

Put those together and the practical rule follows. A statement that a compound raised serum BDNF by some percentage is a statement about a measurement with an assay-dependent error bar, on a protein that exists in two functionally opposite forms which several kits cannot tell apart, most of which in blood comes from platelets rather than from neurons. That does not make the literature worthless. It does mean that small percentage differences reported across studies using different kits are not comparable, and that a single study reporting a modest serum rise is very weak evidence on its own.

Start with why this idea is so appealing. BDNF is a protein with poor pharmacokinetics that does not reach the brain usefully from the bloodstream, and that single fact has blocked every attempt to use it as a medicine ^[11][17]. A small molecule that switched the TrkB receptor on directly would step around the entire problem. That is an enormous prize, which is exactly why the claim deserves the hardest look on this page.

In 2010 a group at Emory reported 7,8-dihydroxyflavone as a bioactive high-affinity TrkB agonist that provoked receptor dimerisation and autophosphorylation and activated downstream signalling; it protected wild-type but not TrkB-deficient neurons from apoptosis, activated TrkB in mouse brain, reduced infarct volume in stroke in a TrkB-dependent way and was neuroprotective in a Parkinson's model ^[11]. That is a specific, strong, mechanistically complete claim, and it is the origin of essentially every subsequent description of 7,8-DHF as a TrkB agonist.

It has not held up at the receptor. At least three independent teams, two of them industrial groups running quantitative assays, failed to reproduce receptor-level activation. A group at BioFocus and Galapagos working for the CHDI Foundation evaluated TrkB antibodies alongside a panel of reported small-molecule TrkB agonists in both recombinant and native assay formats, and the candidate they selected to take into Huntington's models was an antibody ^[12]. A Columbia and Broad Institute collaboration built a set of complementary quantitative assays measuring TrkB receptor activation, TrkB-dependent downstream signalling and gene expression in several cellular contexts; BDNF and other neurotrophic factors produced robust dose-dependent responses, and the reported small-molecule agonists did not, with the authors calling explicitly for those compounds to be re-evaluated ^[13]. A Celon Pharma group in Poland then ran the fullest package: 7,8-DHF bound TrkB with a dissociation constant of 1.3 micromolar, but beyond that binding no activity against the receptor was detectable in either of two cell lines, its pharmacokinetics were poor, no activation of TrkB-dependent signalling was observed in brain after either intravenous or oral dosing, and a screen across 133 molecular targets showed a marked lack of selectivity. Their conclusion was that the published reference compounds including 7,8-DHF do not activate TrkB ^[14].

The fair reading is narrower than either camp's headline. 7,8-DHF plainly does things in animals, and many of those behavioural and neuroprotective effects are real. What is unresolved is the mechanism they are attributed to. A flavonoid can have genuine effects and not be a TrkB agonist, and the accurate sentence is that 7,8-DHF has neuroprotective activity in animal models whose mechanism is not established.

A pathway is not a receptor, and that distinction is the whole of this section. Saying that ketamine or a psychedelic raises BDNF and that BDNF then signals through TrkB and mTOR is ordinary biology with good support behind it ^[7][8][9]. Saying that a small molecule binds and turns on the TrkB receptor itself is a separate, stronger claim requiring separate, stronger evidence, and it is that second claim which has failed to replicate.

One further thread should not be confused with either. A Helsinki group reported that antidepressants including fluoxetine and ketamine bind the transmembrane domain of TrkB dimers, facilitating the receptor's synaptic localisation and its activation by BDNF ^[15], and later extended the finding to LSD and psilocin, reporting affinities around a thousandfold higher than those of the antidepressants and effects that depended on TrkB binding and on promotion of endogenous BDNF signalling ^[16]. Read the claim precisely: this is positive allosteric modulation, making BDNF work better, not agonism. It is a coherent and interesting hypothesis, it comes essentially from one laboratory, and it has not been independently replicated, so it belongs on this page as a hypothesis with that label attached.

Plasticity is double-edged. The same machinery that lets a person learn a language lets a bad habit or a traumatic association dig in, so more plasticity is not a goal by itself. What matters is what happens while the window is open.

That framing stopped being a metaphor recently. In mice, the ability to reopen the social reward learning critical period turned out to be a property shared across chemically diverse psychedelics, and the duration of the reopened state was proportional to the duration of the acute subjective effects those drugs produce in humans. The open state was accompanied by restored oxytocin-mediated long-term depression in the nucleus accumbens, and the genes differentially expressed between the open and closed states pointed at reorganisation of the extracellular matrix as the common downstream mechanism ^[10].

That is a strong result and it carries a specific implication: an open window has a duration, and it closes. It is the clearest available argument for pairing these agents with structured therapy or deliberate practice rather than dosing them alone, and it also explains why the acute subjective experience is not incidental to the effect. See psychedelics & 5-HT2A and dissociatives.

The structural side of the same story is that psychedelics increase neuritogenesis and spinogenesis in cultured neurons and in vivo, with a corresponding rise in synapse number and function, apparently through TrkB, mTOR and 5-HT2A signalling ^[9]. Note the register: signalling pathways, downstream of receptor engagement, which is exactly the kind of claim this evidence supports.

The summary a reader should leave with. Neuroplasticity is real, BDNF is genuinely central to it, and the causal evidence in animals is good. The interventions with the best human evidence are behavioural, led by aerobic exercise ^[4]. Human measurements of BDNF are noisy enough that small reported percentage changes should be discounted ^[5][6]. The drug evidence is mostly preclinical, and the specific shortcut everyone wants, a small molecule that switches TrkB on, is the exact claim that three independent teams could not reproduce ^[12][13][14]. None of this is medical advice.