In the brain, BDNF levels increase significantly after birth in mammals as a result of developing connectivity and increased neuronal activity, the major driver of BDNF expression in neurons (Zafra et al., 1990; Patterson et al., 1992; Hong et al., 2008). In the developing visual system, unilateral section of the optic stalk before innervation of the tectum, or injection of tetrodotoxin into one eye, markedly delays the increase of Bdnf expression in the tectum compared with the contralateral side (Herzog et al., 1994). Given the very low abundance of BDNF even in the adult brain, a reliable detection of BDNF by Western blot remained difficult for a long period of time. It is only recently that a specific monoclonal antibody suitable for Western blot detection became available and its specificity validated using brain extracts prepared from mice lacking Bdnf (Andreska et al., 2020; Wosnitzka et al., 2020). Likewise, the cellular and sub-cellular distribution of endogenous BDNF and of its mRNA remain challenging, both on brain sections and in cultured neurons (Dieni et al., 2012; Will et al., 2013). The straightforward notion that the Bdnf mRNA is primarily localized in neuronal cell bodies, as opposed to axons and dendrites (Will et al., 2013), and that the BDNF protein primarily accumulates in presynaptic terminals (Dieni et al., 2012) is not universally shared and has been challenged by a number of studies relying on overexpression paradigms. Some of these studies include GFP-tagged versions of BDNF, under the assumption that the tag would report the distribution of the endogenous protein (Egan et al., 2003; Matsuda et al., 2009; Leschik et al., 2019). However, large tags of this kind interfere with the processing and secretion of BDNF in transfected cells (Wosnitzka et al., 2020), making it unlikely that such tools reflect the distribution of the endogenous protein. Contrasting with the very low levels of BDNF in the brain, its receptor, the tyrosine kinase TrkB, can be readily detected. TrkB is not only expressed by most neurons in the CNS but also at levels far higher than those of BDNF. RNAseq data indicate that in the adult human brain, the ratio of NTRK2, the gene encoding TrkB, to BDNF is about 25–30 to 1 in sub-regions of the hippocampus, the visual and prefrontal cortex and more than 500 to 1 in the parietal cortex. The second endogenous TrkB ligand NT4 does not account for the ligand-receptor imbalance as NT4 is barely detectable in the adult CNS.¹ Observations indicating that TrkB can be activated by non-neurotrophin ligands, including antidepressants (see below), may contribute to explain the quantitative discrepancies between the levels of TrkB and those of BDNF. In addition, zinc, various G-protein coupled receptor agonists as well as EGF receptor ligands have all been shown to activate TrkB (Lee and Chao, 2001; Huang et al., 2008; Puehringer et al., 2013; Zagrebelsky et al., 2018; Casarotto et al., 2021).

The notion of neurotrophin limited availability, however technically inconvenient and challenging, is a key aspect of BDNF’s physiology, just as is the case with NGF in the PNS (see Introduction). This notion was supported early on by observations made with brain slices prepared from mouse mutants. The loss of just one Bdnf allele was found to cause an impairment of long-term potentiation (LTP) of a magnitude comparable to the loss of both Bdnf alleles (Korte et al., 1995). This important result was later extended to humans: a chromosomal inversion silencing one BDNF allele causes cognitive dysfunction and abnormal weight gain in the affected individual (Gray et al., 2006). Even if BDNF turned out not to be a major survival factor for developing CNS neurons (see above), it profoundly affects the extension of dendrites as well as the number of spines on dendrites (McAllister et al., 1995; Tyler and Pozzo-Miller, 2003; Rauskolb et al., 2010; Zagrebelsky et al., 2018). Given the notion that BDNF is implicated in synaptic plasticity in rodents (for a recent review, see Costa et al., 2022 and references therein), it was of particular interest to explore the relevance of these findings to the function of the human brain. In this regard, the discovery of a polymorphism in the human BDNF gene correlating with mild memory impairments was a key development (Egan et al., 2003). This missense mutation, designated rs6265, causes a replacement of valine (Val) by methionine (Met) in position 66 in the BDNF precursor pro-BDNF (Egan et al., 2003). The association of this polymorphism with modest impairments of episodic memory was later independently confirmed (Cathomas et al., 2010). While the biochemical impact of the Val → Met is still not entirely clear, it has been noted that the Met pro-peptide binds with higher affinity to the mature protein than the Val pro-peptide (Uegaki et al., 2017). Also, in vitro results obtained with a mouse model engineered to replicate the human BDNF polymorphism indicate a reduced secretion of BDNF from cultured neurons (Chen et al., 2006). In addition, the steady state levels of BDNF in adult mutant animals are decreased in the hippocampus (Bath et al., 2012), but not in whole brain lysates (Chen et al., 2006) whereby the reasons for this differential effect are not clear. While Val → Met mutant mice do not show clear memory deficits, they display an anxiety-type phenotype that cannot be ameliorated by the administration of the antidepressant fluoxetine (Chen et al., 2006; Bath et al., 2012). This finding is of special interest given recent observations that antidepressants potentiate TrkB activation by BDNF (see below). Regarding the incidence of the BDNF polymorphism in humans, it is of note that the frequency of the Met66 allele varies considerably across populations, from 0 to 72% (Petryshen et al., 2010). The polymorphism is virtually absent in all sub-Saharan African and some American indigenous populations, but dominant in various Asian ethnic groups (Petryshen et al., 2010). In addition, it has been noted that there are 3 additional protein-encoding genes, designated LIN7C, LGR4, and CCDC34, in the 500 kB region on Chromosome 11 that include the BDNF gene (Petryshen et al., 2010), itself spanning around 70 kb (Cattaneo et al., 2016). Both the close proximity of these genes and the geographical distribution of the rs6265 polymorphism caution a direct causality between the BDNF polymorphism and behavioral alterations in humans. These observations also emphasize the importance of animal models carrying a mutation limited to the Bdnf gene (Chen et al., 2006). Meanwhile, a second mouse Val → Met mutant has been generated (Warnault et al., 2016) as well as a rat mutant, the latter showing a selective impairment of fear memory (Jaehne et al., 2022). BDNF measurements in hippocampal lysates of these rats indicate no differences between Val/Val and Met/Met homozygote animals, neither in males nor in females (Jaehne et al., 2022).

In the context of phenotypes associated with the Val → Met polymorphism, it is also worth noting that Met/Met mice gain weight abnormally when placed on a high fat diet (Xie et al., 2023). These findings fit well with the notion that the release of BDNF may be decreased in these animals (see above) and extensive mouse work has mapped in some detail the hypothalamic circuitry involved in body weight regulation by BDNF/TrkB signalling (for review, see Xu and Xie, 2016). A slight reduction of TrkB activation by BDNF compromises the anti-orexic function of BDNF as noted in a recent mouse model in which a 22-amino acid carboxy terminal tag has been added to Val/Val alleles of BDNF (Wosnitzka et al., 2020). This genetic manipulation leads to a modest impairment of TrkB activation and results in a progressive weight gain that becomes measurable by about 6 months (Wosnitzka et al., 2020). In humans, the impact of BDNF and TrkB mutations on body weight regulation have been well documented (Yeo et al., 2004; Sonoyama et al., 2020), thus confirming that the availability of BDNF is a major component of body weight regulation.

In sum, this body of work indicates that BDNF levels in the CNS are critical and that even a modest decrease of BDNF/TrkB signalling in humans impacts memory performance and body weight regulation.

Extensive work with rodent models has long suggested that decreased BDNF/TrkB signaling may cause CNS dysfunction in humans and BDNF is often regarded as an endogenous neuroprotectant (recently reviewed in Azman and Zakaria, 2022). In addition to the BDNF haploinsufficiency mentioned in the above, exome-sequencing of 197 patients with developmental and epileptic encephalopathy of unknown origin revealed that 5 of these patients have missense mutations in NTRK2, including a Tyr434Cys mutation in 4 of them (Hamdan et al., 2014). The results of this study also indicate dominant inheritance. With regard to age-related forms of neurodegeneration and Alzheimer’s disease in particular, it has been noted that the progression of cognitive decline monitored over a period of 6 years negatively correlates with the levels of BDNF mRNA quantified in the dorsolateral prefrontal cortex, in a large postmortem analysis involving over 500 patients (Buchman et al., 2016). However, BDNF/TrkB variants have not been identified thus far as risk factors associated with Alzheimer’s disease in human genetic studies (de Rojas et al., 2021). Beyond Alzheimer’s disease, impairment of BDNF/TrkB signaling is also likely to play a role and other conditions. Largely based on work in rodent models, it would seem likely that a compromised anterograde transport of BDNF from the cerebral cortex to the basal ganglia may play a role in the pathogenesis of Huntington’s disease (for review, Saudou and Humbert, 2016). Likewise, the failure of TrkB intracellular transport caused by decreased levels of dopamine may play an aggravating role in Parkinson’s disease (Andreska et al., 2023). In line with this, post-mortem examination of the brain of Parkinson’s disease patients revealed abnormal intracellular clusters of TrkB (Andreska et al., 2023).

With growing evidence in rodent models for a role of BDNF in synaptic plasticity in the adult CNS, attention progressively focused on the possible role of BDNF/TrkB signalling in mood disorders and especially depression (Duman, 2002). Classical antidepressants take a long time to act, suggesting that processes other than the blockade of neurotransmitter re-uptake may be at work such as synaptic remodeling (Castren and Monteggia, 2021). Early findings also associated chronic stress with decreased levels of both BDNF and TrkB while by contrast, electroconvulsive therapies and long-term administration of antidepressants correlate with increased TrkB and BDNF levels (Nibuya et al., 1995). Additional key findings comprised the discovery of a link between BDNF and the serotoninergic system (Mamounas et al., 1995) and the observation that BDNF injections into rat brains improve depressive-like symptoms (Siuciak et al., 1997). Importantly, the activation of TrkB was shown to be required for antidepressants to work (Saarelainen et al., 2003). With regard to human genetics, the expected association of the Val/Met polymorphism with the frequency or intensity of depressive episodes has been intensely scrutinised, but failed to convincingly materialize (Castren and Monteggia, 2021). By contrast, several human genetic TrkB polymorphisms were found in patients presenting a lifetime history of depression and suicide attempts (Kohli et al., 2010). The role of TrkB in depression has received additional support following the astonishing observation that all anti-depressants bind to TrkB, in spite of a striking lack of structural relatedness among some of them (Casarotto et al., 2021 and see below). This work has been further extended with the observation that psychedelics bind to TrkB, with an affinity even higher than conventional antidepressants (Moliner et al., 2023).

Given the presence of readily measurable levels of BDNF in human serum, a very large number of studies have attempted to correlate the levels of BDNF with depression. The overall conclusion of metanalyses is that a weak inverse correlation seems to exist between the severity of depressive episodes and the levels of BDNF in serum (Tiwari et al., 2022). Yet it remains difficult to understand the logic of the assumption that BDNF levels in serum may reflect those in the brain, given that it has long been recognized that the main source of BDNF in human serum are circulating blood platelets (Yamamoto and Gurney, 1990). Furthermore, the levels of BDNF in human platelets are far higher than those in the brain, a consequence of the expression of the BDNF gene at comparatively high levels by mature human megakaryocytes (Chacon-Fernandez et al., 2016). The generation of serum involves blood coagulation, a process during which platelets release their content, including BDNF. Platelets are short-lived and their protein content decrease with their age (Allan et al., 2021). As physical exercise can change the composition of blood platelets and include a higher proportion of larger, younger platelets (Heber and Volf, 2015), it is conceivable that the weak correlation observed between depressive states and serum levels of BDNF may merely reflect different degree of physical activity between patient and control groups. The unquestioned benefits of physical exercise on mood and mental health may nonetheless involve BDNF, but rather as a result of the modulation of its levels in the brain by mechanisms involving metabolic intermediates such as beta-hydroxybutyrate (Sleiman et al., 2016) or the muscle-derived cytokine irisin (for review, see Jo and Song, 2021). Consistent with the lack of expression of the Bdnf gene in mouse megakaryocytes (Chacon-Fernandez et al., 2016), BDNF levels in mouse serum are very low, about 3 orders of magnitude lower than in human serum whereby validated BDNF measurements in mouse serum have only been made recently possible following the development of a highly sensitive BDNF ELISA (Want et al., 2023a)². Furthermore, in mouse blood, BDNF levels were found to be identical in plasma compared with serum, unlike is the case in primates (Want et al., 2023a). A main source of BDNF in mouse blood is the skeletal musculature (Fulgenzi et al., 2020) and this muscle-derived BDNF has been shown to trigger the release of insulin from pancreatic β cells expressing a truncated form of TrkB (Fulgenzi et al., 2020 for a recent review on trunacted forms of TrkB, see Tessarollo and Yanpallewar, 2022). Whether or not serum levels of BDNF correlate with mouse models of depression, or indeed with any mouse model of CNS dysfunction, can now be investigated, given that the confounding factor resulting from the presence of platelet-derived BDNF is absent in the mouse. The availability of a highly sensitive BDNF ELISA also opens the possibility to perform measurements of BDNF levels in more relevant body fluids than has been the case so far, including in particular the cerebrospinal fluid and the aqueous humor. With regard to possible functions of BDNF in human platelets, recent work suggests a neuroprotective role after nerve lesion. Using a mouse line engineered to express BDNF in megakaryocytes and consequently in platelets (Dingsdale et al., 2021), the retraction of dendrites of retinal ganglion cells caused by lesions of the optic nerve has been observed to be retarded in mice containing BDNF in platelets compared with wild-type animals (Want et al., 2023b). These results suggest that in humans, the impact of lesions of the nervous system may be attenuated by the release of BDNF from platelets accompanying blood coagulation, a contribution that would be missed in traditional mouse models of nerve lesion. Together with the notion that physical exercise changes the composition of blood platelets (see above), it is thus conceivable that in humans, the benefits of exercise in conditions such as glaucoma (Tribble et al., 2021) may be explained by an increased availability of BDNF originating from platelets.

So far, the administration of BDNF to restore compromised neuronal function has not worked well in humans. Repeated failures triggered a number of attempts to develop small compounds able to activate TrkB, such as a BDNF loop 2 mimic designated LM22A-A, 7,8-dihydroxyflavone and the antidepressants Deprenyl and Amitriptyline. However, none of these agents turned out to activate TrkB in assays specifically designed to report TrkB activation (Boltaev et al., 2017). The reasons for these failures are unclear whereby in the meantime it has been reported that all known antidepressants bind to and activate TrkB in a BDNF-and cholesterol-dependent manner (Casarotto et al., 2021). This unexpected finding is all the more surprising given that none of the antidepressants reported to activate TrkB have been developed based on their TrkB-activating ability and that some are structurally unrelated. Meanwhile, we tested the ability of some of these compounds, including fluoxetine and imipramine for their ability to activate human TrkB in cultured neurons, both in the presence or absence of added BDNF, given that antidepressants are thought to potentiate the action of endogenous BDNF (Casarotto et al., 2021). The preliminary results turned out to be negative, also in the presence of exogenous cholesterol supplemented to the cultures (Ateaque et al., unpublished results). As the successful incorporation of cholesterol into neurons was not monitored, these negative results should not be overinterpreted. Also, the virtual absence of glial cells in our cultures and the immature connectivity between the neurons are likely to be complicating factors whereby the neurons were electrically excitable at the time when they were tested (Telezhkin et al., 2016). With regard to the activation of TrkB by macromolecules other than neurotrophins, a number of agonist antibodies have been developed causing TrkB dimerization and activation, in addition to ZEB85 discussed in the above (Bai et al., 2010; Traub et al., 2017; Guo et al., 2019). Importantly, some of these reagents have been already used in relevant animal models including glaucoma. These were shown to demonstrate positive effects with regard to the prevention of the death of retinal ganglion cells (Bai et al., 2010). The selectivity of such reagents for TrkB and their lack of activation of the neurotrophin receptor p75 pathway suggests that they may be promising reagents towards ameliorating a number of conditions in humans.