Antidepressants act by binding to the transmembrane domain of TRKB receptor

It is unclear how binding of antidepressant drugs to their targets gives rise to the clinical antidepressant effect. We have found that both typical and fast-acting antidepressants bind to a cholesterol interaction motif in the brain-derived neurotrophic factor (BDNF) receptor (tropomyosin-related kinase B - TRKB), a known mediator of neuronal plasticity and antidepressant responses ​ 1–4 ​ . Antidepressants bind to a cross-shaped configuration of dimerized TRKB transmembrane domains and facilitate synaptic localization and activation of TRKB by BDNF. Mutation of the TRKB cholesterol interaction site or cholesterol depletion impaired BDNF-mediated plasticity as well as cellular and behavioral responses to antidepressants ​ in vitro ​ and ​ in vivo ​ . We suggest that binding to and facilitation of BDNF signaling through TRKB is the common mechanism for antidepressant action, which proposes a framework for how molecular effects of antidepressants can transiently signal in these domains ​ 13 ​ . Mutation in the TRKB cholesterol-interaction site limits BDNF signaling, inhibits theta-burst- induced LTP, and prevents antidepressant-induced learning and ocular dominance plasticity. Our findings indicate that TRKB may be the critical target through which astrocyte-derived cholesterol promotes synaptic function and plasticity ​ 16 ​ . BDNF signaling is crucial to the action of essentially all antidepressant drugs, but this effect has been assumed to be indirect, through other molecules such as 5HTT or NMDA receptors. We now show that antidepressants directly bind to the transmembrane region of TRKB dimers with a therapeutically relevant affinity ​ 24,25 ​ , favoring translocation of TRKB to the plasma membrane, where it is accessible to BDNF. This binding is shared by typical tricyclic and SSRI antidepressants, as well as by the rapid-acting ketamine and its metabolite R,R-HNK. Mutagenesis demonstrates the critical role of the CARC motif as the binding site for antidepressants to TRKB, and simulation data reveals that antidepressants was plug-in ​ 52 ​ . Electrophysiology (details in supplement): in vivo and ex vivo In vivo ​ BDNF-induced LTP was performed according to literature ​ 27 ​ . Briefly, pravastatin was administered in drinking water at a dose of 10mg/kg/day for 15-17 days calculated on the basis of daily body weight. Rats were anesthetized with an intraperitoneal injection of urethane 1.5 g/kg body weight, positioned in a stereotaxic frame, and Teflon-coated tungsten wire recording electrode was placed in the dentate hilus. The tip of the infusion cannula was located in the deep stratum lacunosum-moleculare of field CA1, 800μm above the hilar recording site and 300-400μm above the medial perforant synapse. After baseline recording for 20 min, infusion of 2µl of 1 μg/μl BDNF over 30 min at a rate of 0.067µl/min. Evoked responses were recorded for 120 min after infusion. Ex vivo ​ activity-induced LTP: TBS and tetanus stimulus was performed according to literature ​ 53 ​ . Briefly, mice were deeply anaesthetized with isoflurane, the brains were dissected and horizontal 350 micrometers brain slices of the hippocampus were cut on a vibratome. Field excitatory postsynaptic currents (EPSPs) were recorded at the stratum radiatum of the CA1 region. Electric stimulation was delivered by a bipolar concentric stimulation electrode placed at the Schaffer collateral. After obtaining a 15 min stable baseline theta burst stimulation (TBS: 10 bursts of four pulses at 100 Hz, with an interburst interval of 200 msec) or tetanic stimulation (200ms pulse interval;


Introduction
Several targets for antidepressant drug action have been identified but it is not clear how binding to these targets are translated into the clinical effects. Classical antidepressants increase the synaptic levels of serotonin and noradrenaline by inhibition of reuptake or metabolism, while the rapid antidepressant effect of ketamine is attributed to inhibition of NMDA receptors 5,6 . However, the low affinity of 2R,6R-hydroxynorketamine (R,R-HNK), a ketamine metabolite with antidepressant-like activity, to NMDA receptors has called this mechanism into question 7,8 . Therefore, it is not clear whether there is a common mechanism behind the antidepressant effect. Essentially all antidepressants, including ketamine and R,R-HNK, increase the expression of brain-derived neurotrophic factor (BDNF) and activate BDNF signaling through Neurotrophin Tyrosine Kinase Receptor 2 (TRKB) [2][3][4] , but this effect has been considered to be indirect through the inhibition of serotonin transporter (5HTT) and NMDA receptors. BDNF mimics the effects of antidepressants in rodents while inhibiting TRKB signaling prevents these effects [2][3][4]9 . BDNF signaling through TRKB is a critical mediator of activity-dependent synaptic plasticity 1 and the antidepressant-induced BDNF-TRKB signaling reactivates a juvenile-like state of plasticity in the adult brain, which has been suggested to underlie the effects of antidepressant treatments on mood 4,10,11 . BDNF signaling is bidirectionally linked to brain cholesterol metabolism. BDNF, acting through TRKB, promotes production of cholesterol in neurons 12,13 and cholesterol regulates TRKB signaling 14 . Cholesterol is essential for neuronal maturation and proper synaptic transmission 15 . Since cholesterol does not pass the blood-brain barrier, neurons are dependent on locally synthesized cholesterol, mostly by astrocytes 16 . Synaptic cholesterol levels are low during the embryonic and early postnatal life but strongly increase during the 3rd postnatal week in mice 17 , coinciding with the increase in BDNF expression and appearance of antidepressants effects on TRKB 4,18 . Many antidepressants interact with phospholipids and accumulate in cholesterol-rich membrane microdomains, such as lipid rafts 19 . These data prompted us to investigate the potential interactions between TRKB, cholesterol and antidepressants.

CARC motif regulates TRKB activation
Bioinformatic mining revealed a region in the TRKB transmembrane (TM) region that fulfils the criteria for inverted cholesterol recognition and alignment consensus motif (CARC) 20 . This motif is specific to TRKB and is not present in other TRK receptors (Fig. S1A). In primary cortical neurons, cholesterol at 20µM enhanced the effects of BDNF (10 ng/ml) on TRKB phosphorylation (pTRKB), but at higher concentrations (50-100 µM), cholesterol suppressed the effects of BDNF (Fig. 1A). Cholesterol promotes TRKB:PLC-γ1 interaction and TRKB surface exposure ( Fig. S1B-F). The effects of BDNF on TRKB-PLC-γ1 interaction (Fig. 1E), as well as neurite branching (Fig. S1H), were prevented by prior treatment with pravastatin (1 µM/3d), an inhibitor of cholesterol synthesis 14 . A higher concentration of pravastatin (2 µM/5d) reduced survival of cultured cortical neurons and this effect was attenuated by exogenous cholesterol (20 µM), but not by BDNF (Fig. S1M,N), indicating that the effects of pravastatin are mediated through inhibition of cholesterol synthesis. Mutation of the critical tyrosine 433 in the CARC domain 21 into phenylalanine (TRKB.Y433F) did not influence the binding affinity of BDNF (TRKB.wt= 0.081ng/ml; TRKB.Y433F= 0.076ng/ml; Fig. S4H). However, this mutation did reduce pTRKB at the PLC-γ1 interaction site Y816, but not at Y515 in fibroblasts (Fig. 1F, S1G). Protein complementation assay in neuroblastoma cells indicated that the Y433F mutation compromised BDNF-induced dimerization of TRKB, TRKB interaction with raft-restricted FYN (Fig. 1G,H), as well as TRKB translocation to lipid rafts (Fig. S3I). Atomistic simulation of dimerization of TRKB TM helices disclosed five possible structures, but only the conformation where TM domains cross each other at the A443-G439 motif was stable in phosphatidylcholine bilayer with 20-40 mol% cholesterol (Fig. 1B,C). The Y433 residue stabilizes the structure, and Y433F mutation destabilized it by rotating the TM monomers 40 degrees relative to each other (Fig. 1D). Modeling (Table S1) suggests that the increase in membrane thickness by cholesterol regulates the TM dimer structure (Fig. S2A), although cholesterol binding to the CARC motif is transient (residence times ~100ns or less). In the absence of cholesterol, the distance between the TRKB TM C-termini was about 2.2 GFP-TRKB in spines within two minutes, but no facilitation was seen in cells transfected with GFP-tagged TRKB.Y433F (Fig. 3E-J). Furthermore, both BDNF and fluoxetine increased the size of TRKB clusters at the plasma membrane in fibroblasts, as measured by dSTORM/TIRF superresolution microscopy, and these effects were attenuated in cells expressing TRKB.Y433F (Fig. 3O). Thus, BDNF 14 or antidepressant 26 induced TRKB translocation to the plasma membrane and clustering are dependent on the TRKB CARC domain (Fig. 3, S3I). However, the TRKB.Y433F mutation does not change the basal surface localization of TRKB (Fig. 3O). Infusion of BDNF into the dentate gyrus of anesthetized rats significantly increased synaptic strength, as previously reported 27 . However, this effect of BDNF was partially prevented when rats were co-treated with pravastatin (10 mg/kg/day for two weeks; Fig. 4A). To investigate whether this effect of BDNF on LTP involves the TRKB CARC motif, we generated a mouse carrying the TRKB.Y433F mutation. Theta-burst stimulation of hippocampal slices from wild-type mice induced a robust LTP in the CA3-CA1 synapse.
Remarkably, similar stimulation of slices from heterozygous TRKB.Y433F mice failed to induce any potentiation (Fig. 4B), indicating that TRKB CARC motif is critical for LTP. However, tetanic stimulation induced LTP in both WT and TRKB.Y433F slices (Fig. S5C,E), consistent with the central role of BDNF role particularly in theta-burst mediated LTP 28 . Fluoxetine (15mg/kg, 7 days) improved performance of mice in object location memory (OLM) test. The improved performance by fluoxetine was blocked in heterozygous TRKB.Y433F (Fig. 4C) and in BDNF haploinsufficient mice (Fig. S6E). Remarkably, the serotonin transporter knockout (5HTT.ko) mice lacking the primary site of action of SSRIs respond to fluoxetine normally in the OLM test (Fig. S6F). This is consistent with the recent finding that the behavioral and electrophysiological effects of SSRIs are preserved in 5HTT.ko mice 29 (however, see 30 ). Ketamine (10mg/kg single i.p. injection) and fluoxetine (15mg/kg in the drinking water, for 3 weeks) reduced immobility in the forced-swimming test, but they were ineffective in TRKB.Y433F mice (Fig.  4E,F). Furthermore, fluoxetine (15mg/kg in the drinking water, for 2 weeks) facilitated extinction of the freezing response in the contextual fear conditioning paradigm, but this effect was lost in mice carrying the TRKB.Y433F mutation (Fig. 4D). Finally, as observed before, chronic fluoxetine treatment (10mg/kg, in the drinking water for 4 weeks) reactivated critical period-like plasticity in the visual cortex of adult mice 31 (Fig. 4G). A similar induction of ocular dominance plasticity was seen in response to ketamine and R,R-HNK (10mg/kg, ip, three injections in alternate days; Fig. 4G). The effect of fluoxetine on the shift in ocular dominance was lost in TRKB.Y433F heterozygous mice (Fig. 4H), indicating that the plasticity-inducing effects of antidepressants are mediated by their direct binding to TRKB.

Discussion
Here we demonstrate that cholesterol-induced changes in the membrane thickness alter the relative orientation of TRKB dimers, which promotes TRKB surface expression and BDNF signaling. Cholesterol played a permissive role in BDNF-induced TRKB activation, but high cholesterol concentrations limited BDNF signaling. This agrees with previous studies showing that TRKB predominantly resides outside lipid rafts but can transiently signal in these domains 13 . Mutation in the TRKB cholesterol-interaction site limits BDNF signaling, inhibits theta-burst-induced LTP, and prevents antidepressant-induced learning and ocular dominance plasticity. Our findings indicate that TRKB may be the critical target through which astrocyte-derived cholesterol promotes synaptic function and plasticity 16 . BDNF signaling is crucial to the action of essentially all antidepressant drugs, but this effect has been assumed to be indirect, through other molecules such as 5HTT or NMDA receptors. We now show that antidepressants directly bind to the transmembrane region of TRKB dimers with a therapeutically relevant affinity 24,25 , favoring translocation of TRKB to the plasma membrane, where it is accessible to BDNF. This binding is shared by typical tricyclic and SSRI antidepressants, as well as by the rapid-acting ketamine and its metabolite R,R-HNK. Mutagenesis demonstrates the critical role of the CARC motif as the binding site for antidepressants to TRKB, and simulation data reveals that antidepressants stabilize TRKB dimers in a BDNF-activatable conformation that is stable in synaptic membranes with high cholesterol concentrations.
The affinity of ketamine to TRKB is comparable to its affinity to NMDA receptors 8 , but SSRIs bind to TRKB with a much lower affinity that they bind to the 5HTT. However, micromolar concentrations of SSRIs are reached in the brain during chronic treatment 24,25 . The gradual brain accumulation of typical antidepressants to brain at the concentration needed for TRKB binding might be at least one reason why typical antidepressants take so long to act, while the rapid brain penetration of ketamine and R,R-HNK enables fast action. The present findings that antidepressants bind to TRKB and thereby prolong BDNF signaling directly link the effects of antidepressants to neuronal plasticity 4 . Juvenile-like plasticity induced by these drugs in the adult brain facilitates beneficial adaptation of neuronal networks that have been abnormally wired during development or by stress 10,11 , which may explain the superiority of the antidepressant and psychotherapy combination 32 . Our data suggests a new framework for antidepressant action where the drugs bind to the cholesterol-interaction site of TRKB thereby promoting BDNF-induced plasticity, which is permissive for the activity-dependent network rewiring that underlies the clinical antidepressant response.
Animals: adult male rats were used for in vivo BDNF-induced LTP. Adult male and female mice (12-20 weeks): BDNF haploinsufficient, SERT.ko, TRKB.Y433F.het (detailed description in supplement), with respective wild-type littermates, were used in object-location memory, contextual fear conditioning or ocular dominance plasticity. All animals were kept group housed with free access to food and water, except during the experimental sessions. All protocols were approved by local ethical committees (Finland: ESAVI/10300/04.10.07/2016; Norway: 6159; Germany: G-18-88).
Cell culture and transfections: HEK293T (used for production of GFP-tagged TRKB), MG87.TRKB cells, Neuroblastoma-2A (N2A), and primary cultures of E18 rat embryos (DIV8 for biochemical analysis and DIV14 for FRAP) were cultivated according to previously described protocols 26,45,46 . The cells were transfected (DIV13 for hippocampal neurons) to express GFP-or Luciferase-tagged TRKB (wt or Y433F or Luciferase-tagged FYN fragment, using lipofectamine 2000 according to manufacturer's instructions (#11668019, Thermo Fisher), or calcium phosphate co-precipitation 47 , 24-48h prior to the experimental sessions or sample collection. After washing with PBS and blocking with PBS containing 5% nonfat dry milk and 5% BSA, the samples were incubated with primary anti-TRKB antibody (R&D Systems, #AF1494, 1:1000 in blocking buffer) ON at 4 o C. Following washing, the samples were incubated with HRP-conjugated anti goat IgG (1:5000 in blocking buffer) for 1h at RT. The cells were washed 4x with 200µl of PBS for 10 min each. Finally, the chemiluminescent signal generated by reaction with ECL was analyzed in a plate reader. The levels of total and phosphorylated TRKB at Y515 or Y816 in MG87.TRKB cells, challenged with BDNF, were measured by western-blotting 48 . For the analysis of TRKB migration to lipid rafts, the samples from transfected N2A cells to express GFP-tagged TRKB.wt or TRKB.Y433F, challenged with BDNF, were processed to isolate detergent-resistant membrane (DRM) fractions in sucrose gradient 49 . N2A cells were seeded at a density of 2.5 million per plate on 10 cm plates and transfected after 24 hours with either wild-type GFP-tagged full length TRKB or the GFP-tagged Y433F TRKB mutant. 48 hours after plating, cells were washed with ice cold 1x PBS and scraped in extraction buffer (25 mM Tris-HCl pH 8, 150 mM NaCl, 5 mM EDTA) with the addition of 0.5% v/v Lubrol (Serva) and a cocktail of protease and phosphatase inhibitors (Sigma). Cellular membranes were mechanically broken by passing the cell suspension through a 23G needle five times. Protein concentration was measured for each sample and equal amounts of proteins were transferred to Eppendorf tubes and mixed with sucrose in extraction buffer to a final concentration of 72% . The samples were then transferred to the bottom of Beckman 2.2 ml ultracentrifuge tubes and carefully covered with equal volumes of 35% sucrose and 5% sucrose in extraction buffer. The samples were centrifuged at 52000 x g for 18 hours at +4°C with a TLS-55 rotor in a Beckman Coulter XP Optima ultracentrifuge. Finally, 12 fractions per sample, collected from the top of the tube, were transferred to clean tubes, sonicated for 10 minutes in 0.25% SDS and prepared for western blotting, where the levels of GFP-tagged TRKB and flotillin-2 were analyzed. biotinylated drugs fluoxetine or R,R-HNK (0.1-100µM), tritiated imipramine (0.01-30μM) or biotinylated BDNF (0.1-100ng/ml) were added after blocking with 2% BSA in PBS for 1h and the signal developed by incubation with HRP-conjugated streptavidin followed by ECL. For the competitive assay, increased doses of non-biotinylated drugs (0.1-10µM) or BDNF (10-100ng/ml) were mixed with 1µM of biotinylated fluoxetine for 1h, and the signal from HRP-conjugated streptavidin developed as described above. Microscale Thermophoresis -MST -was performed according to literature 23 . Briefly, the changes in fluorescence following temperature gradient emitted by GFP-TRKB, incubated with fluoxetine (0-100μM) were conducted in premium coated capillaries using LED source with 470nm and 50% infrared-laser power.

Determination of TRKB activation and coupling
FRAP: hippocampal cells from E18 rat embryos (DIV14) were transfected to express GFP-TRKB.wt or GFP-TRKB.Y433F. Upon identification of GFP-positive cells, the cells were treated with fluoxetine, ketamine or BDNF for 15min; the whole spine head or the neurite shaft was bleached with high laser power 51 and the latency for fluorescence recovery was quantified. TIRF/dSTORM: MG87.TRKB cells were transfected to overexpress GFP-TRKB.wt or GFP-TRKB.Y433F, challenged with BDNFor fluoxetine for 15min and fixed to immunostaining for GFP. The area of GFP-tagged TRKB cell surface clusters was determined by total internal reflection fluorescence microscopy (TIRF) coupled with direct stochastic optical reconstruction microscopy (dSTORM).
Immunostaining and Sholl analysis: cortical cells (E18 rat embryo, DIV8) were treated with BDNF (10ng/ml/3d) and pravastatin (1μM/3d), fixed and labeled for actin. The number of intersections from the cell body was counted using FIJI plug-in 52 .

Electrophysiology (details in supplement): in vivo and ex vivo
In vivo BDNF-induced LTP was performed according to literature 27 . Briefly, pravastatin was administered in drinking water at a dose of 10mg/kg/day for 15-17 days calculated on the basis of daily body weight. Rats were anesthetized with an intraperitoneal injection of urethane 1.5 g/kg body weight, positioned in a stereotaxic frame, and Teflon-coated tungsten wire recording electrode was placed in the dentate hilus. The tip of the infusion cannula was located in the deep stratum lacunosum-moleculare of field CA1, 800μm above the hilar recording site and 300-400μm above the medial perforant synapse. After baseline recording for 20 min, infusion of 2µl of 1 μg/μl BDNF over 30 min at a rate of 0.067µl/min. Evoked responses were recorded for 120 min after infusion. Ex vivo activity-induced LTP: TBS and tetanus stimulus was performed according to literature 53 . Briefly, mice were deeply anaesthetized with isoflurane, the brains were dissected and horizontal 350 micrometers brain slices of the hippocampus were cut on a vibratome. Field excitatory postsynaptic currents (EPSPs) were recorded at the stratum radiatum of the CA1 region. Electric stimulation was delivered by a bipolar concentric stimulation electrode placed at the Schaffer collateral. After obtaining a 15 min stable baseline theta burst stimulation (TBS: 10 bursts of four pulses at 100 Hz, with an interburst interval of 200 msec) or tetanic stimulation (200ms pulse interval; 100 pulses; 0.1ms pulse duration) was delivered and field potentials were recorded for 45 min.

Ocular dominance plasticity assay (details in supplement):
The shift in ocular dominance induced by drug treatment was performed according to literature 31 . Briefly, animals were anesthetized via i.p. injection with a mixture containing: 0.05 mg/kg fentanyl; 5 mg/kg midazolam; 0.5 mg/kg medetomidine; diluted in saline and fixed in the stereotaxic frame. After cleaning and polishing, a thin layer of cyanoacrylate glue was applied to the surface of the skull, in order to make it transparent. The next day, the acryl layer was polished over the area of interest. A metal head holder was first glued on the skull, carefully keeping the area of interest at the center of the holder, and then fixed with a mixture of cyanoacrylate glue and dental cement. Finally, transparent nail polisher (#72180, Electron Microscopy Sciences) was applied inside the metal holder above the area of interest. Monocular deprivation was carried on the left eye, the eyelashes were cut and the eye sutured shut with 3 mattress sutures. The monocular deprivation lasted 8 days and the animals were checked daily and resutured if needed to prevent reopening of the eyes. Two sessions of imaging were performed: one before the beginning of the treatment with fluoxetine, ketamine or R,R-HNK (IOS I) and one on the 8th day after monocular deprivation (IOS II). For imaging, animals were kept on a heating pad located in front of and within 25 cm from the stimulus monitor. The visual stimulus was a 2°wide horizontal bar moving upwards with a temporal frequency of 0.125 Hz and a spatial frequency of 1/80 degree, displayed in the central part of a high refresh rate monitor (-15 to 5 degrees azimuth, relative to the animal visual field) in order to preferentially stimulate the binocular part of the visual field. The continuous-periodic stimulation was synchronized with a continuous frame acquisition, frames were collected independently for each eye. After obtaining cortical maps for both contralateral (C) and ipsilateral (I) eyes and computing Ocular Dominance score as (C−I)/(C+I), finally, the Optical Dominance Index (ODI) was calculated as the mean of the OD score for all responsive pixels. The ODI values are comprised in an interval going from -1 to +1: positive values indicate a contralateral bias, negative ones indicate ipsilateral bias and ODI values of 0 indicate that ipsilateral and contralateral eyes are equally strong.

Behavioral analysis:
Object-location memory (OLM): this test was performed in a square arena (28cm side) with opaque walls containing cues (black stripes or spots). The mice were placed for 3 consecutive days (15min per session) in the arena with two identical objects (table tennis balls glued to caps of 50ml Falcon tubes) in the same position throughout the sessions (pretest). Following the drug administration (starting immediately after the last pretest session). At the test session one of the objects is moved to a different position and the number of visits (counted as sniffing or interacting with the object) to the old (A) or newly located (A') object was determined by an observer blind to the conditions 54,55 . Contextual fear conditioning: this test was modified from previous studies 11 . Briefly, the mice were conditioned to 5 scrambled foot shocks (0.6mA/2s) during the 8 min session (arena: 23×23×35cm) under constant illumination (100 lux). The treatment with fluoxetine started immediately after the conditioning session throughout the last extinction trial. During the extinction trials, the animals were exposed to the same context where the shocks were delivered and the time spent in freezing during the 8 min session was automatically determined by the software (TSE Systems, Germany).
Forced swimming test -FST: animals were submitted to a 6min session of FST in 5-liter glass beaker cylinders (19cm diameter, with 20cm water column 25±1 o C) 56 . Fluoxetine was administered for 3weeks (15mg/kg, in the drinking water), and ketamine was injected 2h (ip) prior to the FST. The immobility was assessed in the last 4min of the session, and the water was changed between each test. After swimming, animals were kept in a warm cage until dried, then returned to their home cages. Test was videotaped and analyzed by a trained observer blind to treatment.
Statistics: Student's t test (two-tailed), one-or two-way ANOVA were used, followed by Fisher's LSD post hoc test. F and p values are indicated in Table S2.
author/funder. All rights reserved. No reuse allowed without permission.
The copyright holder for this preprint (which was not peer-reviewed) is the . https://doi.org/10.1101/757989 doi: bioRxiv preprint