In the present study, we have demonstrated that deoxygedunin directly binds the ECD of TrkB and promotes its dimerization and activation. Deoxygedunin provokes TrkB activation in primary neurons and in mouse brain. It strongly protects neurons from apoptosis in a TrkB- dependent manner. Further, it activates TrkB in BDNF conditional knockout mice, indicating that BDNF is not implicated in the stimulatory effect of deoxygedunin. When it is injected in animals, deoxygedunin mimics BDNF and exerts neuroprotective and antidepressant actions and enhances learning processes. Moreover, administration of 7,8-dihydroxyflavone and deoxygedunin into pregnant BDNF +/− mothers substantially rescues vestibular ganglia, which are significantly degenerated in BDNF −/− pups.
To determine which chemical group or moiety is critical for the neurotrophic activity of gedunin, we have conducted a structure-activity relationship study. The results suggest that the epoxy ring confirmation in the D ring is essential for gedunin's agonistic activity. It has to be in a down conformation for the agonist to be effective. Nevertheless, the alpha-beta unsaturated ketone in ring A is variable. Reduction of the unsaturated C
C bond did not impair its stimulatory activity, suggesting that this site can tolerate a potential chemical modification, which might be useful for improving its water solubility and biological activity. When the epoxy group is in down conformation, changing ketone into alcohol did not cripple its agonistic effect either regardless of alpha or beta conformation of the resulting hydroxy group. Nevertheless, deoxygedunin displays the strongest agonistic effect, suggesting that ketone might be more active than alcohol in ring A (Figure S1
). We have recently shown that 7,8-DHF and NAS (N-acetylserotonin) potently activate TrkB receptor. Structurally, these two compounds share the hydroxyphenol group. Our structure-activity study in 7,8-DHF shows that the 7-position hydroxyl group is critical for its agonist effect. This position parallels the 5-position hydroxyl group on the indole ring in NAS. Conceivably, these two compounds share the same binding pocket on TrkB ECD. However, they are quite different from deoxygedunin in structure. The latter belongs to terpenoid family members. Deoxygedunin might bind to different motif on TrkB ECD from that of 7,8-DHF and NAS.
Gedunin, a tetranortriterpenoid isolated from the Indian neem tree (Azadirachta indica), and was recently shown to manifest anticancer activity via inhibition of the 90 kDa heat shock protein (Hsp90) folding machinery and to induce the degradation of Hsp90-dependent client proteins similar to other Hsp90 inhibitors 
. Intraperitoneal injection of deoxygedunin (5 mg/kg) into mice triggers TrkB activation in mouse brain after 2 h and peaked at 4–8 h, indicating that this compound or its metabolites can pass brain blood barrier and has a fairly long effective duration. To monitor its tissue distribution and kinetics of degradation after i.p. injection, we collected different brain regions and body organs 4 h after 3
H-deoxygedunin administration. 3
H-labeled compounds were concentrated in olfactory bulb and hippocampus in the brain. Clearly, the brain as a whole organ was the major tissue in the body where 3
H-labeled compound was accumulated, followed by the lung (Figure S4
). Obviously, substantial amount of 3
H-deoxygedunin and its metabolites remained in the brain 4 h after drug administration, indicating this compound might possess a favorable half-life in the body. Further, it can activate TrkB in mouse brain via oral administration. Nevertheless, deoxygedunin and other derivatives have poor water solubility. In order to alleviate this issue, we are now synthesizing various novel derivatives directly from deoxygedunin that possess better water solubility while maintaining its biological efficacy. Our rationale for modifying the skeleton of deoxygedunin will occur in two phases. In the first phase, we will modify the furan ring in the hopes of increasing solubility. In the second phase, we will modify the ABCD rings of the solubilized deoxygedunin derivatives (from phase one) in order to improve their biological activity.
Deoxygedunin binds to the ECD domain of TrkB (). Interestingly, deoxygedunin and alpha-dihydrogedunol (epoxy ring down) provoked GST-TrkB to bind HA-TrkB in cotransfected HEK293 cells, leading to TrkB receptor association and autophosphorylation. However, competition assay shows that deoxygedunin is unable to compete off BDNF bound to TrkB (data not shown). Presumably, it is due to its much weaker binding affinity (1.4 µM) than BDNF (Kd
M). Although the detailed molecular mechanism of how a ligand provokes Trk receptor activation remains unclear, a conformational change hypothesis has been proposed. For instance, a mutation at P203A can cause spontaneous dimerization and activation of TrkA 
, suggesting that ligand-provoked receptor dimerization might be through triggering conformational change. Conceivably, deoxygedunin binding to TrkB might elicit its conformational change, leading to its dimerization. Remarkably, although dihydrodeoxygedunol, 3-alpha-acetoxy-dihydro-deoxygedunin and 3-deoxo-3-beta-acetoxy-deoxydihydrogedunin failed to provoke HA-TrkB to bind GST-TrkB, the precipitated GST-TrkB was strongly phosphorylated by these chemicals. This partial agonistic effect fits with their weak stimulatory activity on TrkB receptor ( and ). The maximal activation following ligand binding requires initial autophosphorylation of tyrosine residues 670, 674, and 675 on TrkB 
. Previous studies suggest the following model for Trk activation: in the ligand-unbound state, the activation loop blocks access of substrates to the active site of the kinase domain. Ligand binding permits autophosphorylation of the activation-loop tyrosines. Once phosphorylated, each of these forms specific charge-pair interactions with nearby standing positive charges. Such interactions in turn stabilize an “open” conformation in which the activation loop no longer blocks access of substrates to the kinase. This stabilized conformation now effectively phosphorylates inter- and intramolecular target 
. Conceivably, the robust GST-TrkB phosphorylation might result from the intramolecular reaction in the stabilized conformation by the 3 chemicals mentioned above.
We have shown that both 7,8-DHF and deoxygedunin bind to the ECD of TrkB and provoke the receptor dimerization autophosphorylation. They both display comparable agonistic activity on TrkB receptor. Moreover, they both minic BDNF by exhibiting robust neuroprotective effect in stroke, neuroexcitotoxicity, and vestibular ganglia survival assays. In addition, these two compounds also display strong activity in learning and memory and depression in animal models. Forced swim test revealed that deoxygedunin might possess more robust antidepressant effect than 7,8-DHF ().
Taken together, these two small molecules basically share the same biological and therapeutic actions. Nonetheless, 7,8-DHF is water soluble, but deoxygedunin is not. The latter has more complicated structure scaffold than the former. The flavonoids are easier for medicinal modification to improve their biological effect than gedunin family members.
BDNF and TrkB receptor are targets for therapeutic intervention in various neurological diseases including neuroexcitotoxicity, stroke, depression, anxiety, neurodegeneration etc. Nevertheless, BDNF is not useful as a therapeutic agent because of its poor pharmacokinetic properties. To search for small molecules that possess robust TrkB agonistic activity, here, we invented a cell-based survival functional assay. Our high-throughput screen was focused on the neuronal survival function provided by the TrkB receptor. Only the compounds that selectively protect TrkB expressing cells but not parental cells without TrkB from apoptosis were subjected to next round functional analysis. Hence, the positive hits from the first round screen either directly activate TrkB as an agonist or facilitate the downstream survival machinery mediated by TrkB receptor. The second round screening in primary neurons and follow-up TrkB association and autophosphorylation analyses eliminate compounds that did not directly target TrkB receptor. Through the in vitro
receptor/ligand binding assay, in vivo
TrkB activation and neuronal survival experiments, we finally obtained a few potent and selective TrkB agonists that virtually mimic BDNF's biochemical and physiological actions, and 7,8-DHF and deoxygedunin are the most promising lead compounds. Employing a variety of animal models, we have established that deoxygedunin exhibits potent neuroprotective actions in kainic acid neuroexcitotoxicity and stroke animal models (see also Supplemental Materials S1
and Figure S3
). Moreover, we found that vestibular ganglion loss in BDNF −/− pups was significantly blocked by the drug treatment (). This finding demonstrates that TrkB agonists can protect vestibular ganglia from degeneration in BDNF-lacking mice. Further, we show that both 7,8-DHF and deoxygedunin display prominent antidepressant action in forced swim test, which is also TrkB dependent (). BDNF/TrkB signaling is also associated with and required for the acquisition of classical conditioned fear in rodent models 
. Once again, we demonstrated that deoxygedunin enhances acquisition of conditioned fear, which is a BDNF-dependent learning process (). Therefore, all of these animal models strongly support the notion that deoxygedunin mimics BDNF in vitro
and in vivo
and reveal remarkably therapeutic activities in various neurological diseases.
In CNS, physiological activities regulate local BDNF synthesis and secretion, providing neurotrophic support necessary for the BDNF-responsive tissues in a timely, dynamic fashion. In some regions, neurotrophins act by either a paracrine or an autocrine mechanism, while other neurons could compete for the same factor in the more classical long-range, target-derived paradigm 
. However, it is impossible for the small molecular TrkB agonists to meet the needs of TrkB activation in tissues that require temporal and spatial regulation by like BDNF or NT-4/5. The finding that 7,8-DHF and deoxygedunin rescue the survival of vestibular ganglia in BDNF −/− pups supports that these TrkB agonists partially supplement the deficiency of BDNF in mice. Undoubtedly, further chemical modification on these lead compounds to improve their agonistic activities and binding affinities might provide powerful tools for us to dissect the physiological functions of BDNF/TrkB signalings.