In this study, we identified what we believe is the first low–molecular weight antagonist of TrkB using a structure-based virtual approach. The identified molecule binds the extracellular domain of TrkB, prevents BDNF-induced TrkB activation, and abolishes the biological effects of BDNF on TrkB-expressing cells but not those of NGF or NT-3 on TrkA- and TrkC-expressing cells. Moreover, the compound inhibits TrkB function in the brain after systemic administration and demonstrates anxiolytic and antidepressant potential.
The iterative process of rationale-based in silico screenings combined with functional assays illustrates the power of modeling specific structural interactions between 2 proteins for the rapid identification of functional modulators. Finding low–molecular weight compounds that selectively disturb protein-protein interactions is notably a difficult endeavor. As of today, only 1% of the whole human proteome has been successfully targeted with small-molecule compounds (36
). Moreover, most of the fruitful reports concern proteins that naturally have small organic molecules for ligands and not another protein. The 2-digit nanomolar affinity observed for ANA-12 is therefore particularly encouraging for other protein-protein interfaces considering the pace and small resources engaged here to discover the current lead.
In order to allow rapid and efficient selection of functionally active ligands, we used a discrete and specific area of the TrkB receptor for the virtual docking of a large database of 1.6 million structurally unrelated compounds. This region is located in the fifth subdomain of the extracellular domain of TrkB and has been described as the “specificity patch” of the receptor, as it interacts with the N-terminal region of BDNF (26
). In fact, the N-terminal end of neurotrophins has been shown (a) to be highly variable in terms of amino acid composition, (b) to affect Trk receptor binding and activation capacities, and (c) to fit into a binding pocket formed by a patch of charged amino acids in the fifth subdomain of Trk receptors. We therefore hypothesized that modeling the TrkB specificity patch to perform an in silico screening would lead to active TrkB-selective ligands. Conventional high-throughput screenings using functional assays, such as neurite outgrowth and cell survival, are demanding and generally yield less than 1% active compounds (37
). Moreover, the identified active compounds are likely to act through any mechanism and not necessarily through TrkB receptors. In contrast, the structure-based virtual strategies used in this study obtained a high rate of recovery, with 2 active molecules out of 12 tested in the first round of screening and 4 out of 14 in the second round of screening for more potent analogs. Very recently, Massa and colleagues have discovered active TrkB agonists using a pharmacophore-based approach by modeling the variable and specific loop II in BDNF (24
). This strategy yielded 5 active compounds out of 7 tested that can activate TrkB both in vitro and in vivo. Together, these findings demonstrate the efficiency of structure-based virtual screenings for the discovery of functional and selective Trk receptor ligands.
The data presented here show that low–molecular weight heterocyclic compounds affect the formation of a functional complex between BDNF and TrkB. Two of the compounds described here, ANA-12 and N-T19, structurally differ only by a benzene moiety, yet this modification showed important consequences on the pharmacological properties of ANA-12. For instance, while N-T19 demonstrated a second higher-affinity site when tested in neurons, this site was present in both recombinant and native systems for ANA-12. Here, we can hypothesize that N-T19 needs p75NTR to modulate the conformation of TrkB-d5 in order to open a binding pocket that will facilitate its entry and result in an apparent higher potency. The addition of a benzene cycle to ANA-12 may overcome the need of p75NTR to access this binding pocket. This would result in an apparent higher-affinity site even in the absence of p75NTR. Binding studies using TrkBECD-Fc showed that ANA-12 selectively binds to the extracellular domain of TrkB. Interestingly, 2 binding sites could be detected with constants similar to those observed in functional assays (high-affinity site, Kd ≈ 10 nM, fraction 20%–30%; low-affinity site, Kd ≈ 10 μM, fraction 70%–80%). The presence of 2 binding sites on purified TrkB confirms the idea that ANA-12 may access a binding pocket independently of p75NTR-induced transconformation of TrkB.
Modeling of ANA-12 docking in the specificity patch showed that the molecule nicely fits into the TrkB-d5 ADEB β-sheet and behaves like the N terminus region of BDNF in the binding pocket (26
). Similarly to NT-4/5 and BDNF, ANA-12 lies against the disulfide bridge formed by Cys302 and Cys347 and interacts with Asp298, His299, and His300. Cocrystallization of the compound with TrkB, in the presence or absence of BDNF, would be of invaluable help for a better understanding of the fine molecular events triggering TrkB activation. In fact, binding experiments showed that even high concentrations of ANA-12 cannot overcome its displacement from TrkB by BDNF, suggesting a noncompetitive mechanism. Given the size of BDNF relative to ANA-12 and its many binding sites on TrkB (26
), one can hypothesize that while the N-terminal arm of the neurotrophin competes with ANA-12 for the TrkB-d5–binding pocket, the small compound is not able to destabilize the other sites of interactions between BDNF and TrkB.
Systemic administration of low doses of ANA-12 to mice efficiently inhibited TrkB activity in the striatum, cortex, and hippocampus, and showed predictive anxiolytic and antidepressant properties in a panel of behavioral tests that included the elevated plus maze, the novelty-suppressed feeding, the tail suspension, and the Porsolt swimming tests. These tests are commonly used for their predictive validity to assess anxiolytic-like and antidepressant-like activities of compounds (38
). Because BDNF has been described as a modulator of memory formation, the behavioral alterations observed in these tests could therefore theoretically be a result of a memory deficit. However, it has been shown that memory can be addressed in the elevated plus maze only when repeated trials and preexposure sessions are used (38
), which is not the case here. Similarly, the novelty-suppressed feeding test measures the latency to first approach, which by definition cannot be affected by memory. We therefore exclude potential memory impairments as a confounding factor in our behavioral analysis.
The neurotrophin hypothesis of depression postulates that BDNF is strongly linked to depression and depression-like behaviors (9
). However, while positive modulation of BDNF signaling may be a key target for antidepressant actions, inhibition of this pathway does not seem to induce depression (16
). In contrast, inhibition of BDNF and TrkB signaling in the reward system has shown strong antidepressant effects in rodent models of stress and depression (7
). Given the dynamic activity of the BDNF/TrkB coupling in these disorders, it has been proposed that BDNF antagonists would provide a novel class of efficient therapeutic agents to treat anxiety and depression in humans (7
). Accordingly, we propose that ANA-12, antagonist of BDNF/TrkB signaling, may be effective for the treatment of anxiety disorders and depression in humans.
We previously observed similar behavioral effects on the elevated plus maze using cyclotraxin-B, a BDNF-derived peptide that inhibits both BDNF-dependent and -independent TrkB activity in the brain (20
). Interestingly, cyclotraxin-B had no effect in the forced-swim test. Further investigations will be required to better understand the different mechanisms underlying the effects of these 2 compounds and more generally to better understand the role of BDNF and TrkB in mood disorders. A potential issue with the use of TrkB inhibitors for clinical application as anxiolytic and/or antidepressive medication could be the induction of cell death. Chronic treatments with such compounds could result in the loss of neurons, the survival of which depends on BDNF signaling (2
). However, as demonstrated in this study with ANA-12, this side effect can potentially be prevented by using low doses of the antagonist that will efficiently alter behavior without affecting the survival of neurons.
Finally, other compounds sharing the 3-(acylamino)-ε-caprolactam scaffold of N-T19 and ANA-12 have been described as procognitive agents in mice (41
), although the precise mechanism of action is not described (42
). Given the central role of the BDNF/TrkB coupling in cognition and memory, it would be interesting to evaluate the effect of these compounds on TrkB receptors and to compare them with ANA-12 and N-T19.
To the best of our knowledge, ANA-12 is the first nonpeptide antagonist of TrkB receptor that elicits strong and specific effects in vivo. This proteolytically stable small molecule constitutes a valuable pharmacological tool to enable us to better investigate the role of BDNF/TrkB signaling in pathophysiological situations and will serve as a lead compound for the design of potent orally bioavailable TrkB modulators.