Pharmacologically diverse antidepressants facilitate TRKB receptor activation by disrupting its interaction with the endocytic adaptor complex AP-2

Several antidepressant drugs activate tropomyosin-related kinase B (TRKB) receptor, but it remains unclear whether these compounds employ a common mechanism for TRKB activation. Here, using MS, we found that a single intraperitoneal injection of fluoxetine disrupts the interaction of several proteins with TRKB in the hippocampus of mice. These proteins included members of adaptor protein complex-2 (AP-2) involved in vesicular endocytosis. The interaction of TRKB with the cargo-docking μ subunit of the AP-2 complex (AP2M) was confirmed to be disrupted by both acute and repeated fluoxetine treatments. Of note, fluoxetine disrupted the coupling between full-length TRKB and AP2M, but not the interaction between AP2M and the TRKB C-terminal region, indicating that the fluoxetine-binding site in TRKB lies outside the TRKB:AP2M interface. ELISA experiments revealed that in addition to fluoxetine, other chemically diverse antidepressants, such as imipramine, rolipram, phenelzine, ketamine, and its metabolite 2R,6R-hydroxynorketamine, also decreased the interaction between TRKB and AP2M in vitro. Silencing the expression of AP2M in a TRKB-expressing mouse fibroblast cell line (MG87.TRKB) increased cell-surface expression of TRKB and facilitated its activation by brain-derived neurotrophic factor (BDNF), observed as levels of phosphorylated TRKB. Moreover, animals haploinsufficient for the Ap2m1 gene displayed increased levels of active TRKB, along with enhanced cell-surface expression of the receptor in cultured hippocampal neurons. Taken together, our results suggest that disruption of the TRKB:AP2M interaction is a common mechanism underlying TRKB activation by several chemically diverse antidepressants.

Compromised plasticity has been suggested as one of the major causes of depression (1). Supporting this idea, morpho-logical and functional deficits have been observed in the brains of patients with mood disorders (2)(3)(4). Reduced neurotrophic support, such as by brain-derived neurotrophic factor (BDNF) 2 signaling via TRKB, has been linked to the atrophy observed in these patients in the neuronal networks regulating mood and cognition (5,6).
Antidepressant drugs (AD) have been suggested to act by improving neuronal connectivity, plasticity, and information processing mainly by targeting monoamine neurotransmission, such as serotonin (5-HT) and noradrenaline (4). In line with this, an intact BDNF-TRKB signaling system is crucial for the efficacy of antidepressant compounds, as mice overexpressing a dominant negative TRKB isoform (TRKB.T1) or haploinsufficient for BDNF (Bdnf ϩ/Ϫ ) are not responsive to AD (7). Furthermore, the activation of TRKB receptor signaling (and putatively the reinstatement of plasticity) is associated with the response to AD (7)(8)(9).
The binding of BDNF to TRKB initiates dimerization and phosphorylation of tyrosine residues in the intracellular portion of the receptor (10), which allows docking of adaptor proteins to these sites. Phosphorylated tyrosine 515, for example, serves as a docking site for SHC adaptor protein, whereas phospholipase C␥ (PLC-␥1) binds to the phosphorylated Tyr-816 residue (11). Interestingly, the phosphorylation of TRKB receptors at Tyr-706/707 and Tyr-816 residues are enhanced by several antidepressants in adult mouse hippocampus, whereas no change was detected at the Tyr-515 residue (7,9). Despite the evidence on the necessity of the BDNF/TRKB system for antidepressant responses, the mechanism by which these drugs might trigger TRKB activation remains unclear.
TRKB is mainly localized intracellularly in vesicles that are translocated to the neuronal surface through neuronal activity (12)(13)(14). In this scenario, proteins able to modulate TRKB transit to and from the cell surface, therefore modulating its exposure to BDNF, provide an interesting opportunity to understand the antidepressant-induced effects. In line with this idea, our MS-based screen for TRKB interactors that change upon fluoxetine treatment identified the AP-2 complex as a TRKB interactor. AP-2 is crucial for clathrin-dependent endocytosis and participates in the removal of several transmembrane receptors and their ligands from the cell surface (15). AP-2 is a heterotetrameric complex that consists of two large chains (␣and ␤-adaptin), one medium chain (, AP2M), and one small chain () (16). TRK receptors have been identified among the targets for rapid clathrin-mediated endocytosis upon binding to their ligands (17). Earlier studies demonstrated that BDNFinduced phosphorylation of TRKB receptors recruits AP-2 complex and clathrin to the plasma membrane in cultured hippocampal neurons (18). More recently, a novel relationship has been established between TRKB receptors and AP-2, suggesting that autophagosomes containing activated TRKB are transported to the cell soma in an AP-2-dependent manner (19).
In the present study, we identified a putative binding site on the TRKB receptor for AP2M, which provides a known interface for interaction between these proteins. Thus, we aimed at investigating whether the interface between TRKB and AP2M is a site of action for AD. Specifically, we tested the effects of several AD, such as fluoxetine, imipramine, rolipram, phenelzine, ketamine, and RR-HNK, as well as nonantidepressants with psychoactive effects (diazepam and chlorpromazine) on the interaction of TRKB receptors with AP2M.

AP2M is an interaction partner of TRKB receptors
Samples from mouse hippocampus (vehicle (n ϭ 3) and fluoxetine-treated (30 mg/kg, single i.p. injection, n ϭ 3)) were processed for co-IP with TRKB-specific antibody and analyzed by MS to investigate the TRKB interactome and the changes in the interactome induced by acute fluoxetine administration. The complete list of interactors can be found in FigShare (Table  S1). The identified interactors that are components of the synaptic and endocytic machineries are listed with their score from MS in Fig. 1a. In the present study, we decided to focus on the interaction with AP-2 as a putative modulator of TRKB surface localization and activation because three of four subunits of the complex were found associated with the TRKB receptor (Fig.  1a). Moreover, acute fluoxetine administration induced uncoupling of TRKB from AP-2, which is evidenced by the reduced scores of the AP-2 complex subunits ␣-2 (P17427), ␤ (Q9DBG3), and (P84091) in the samples from fluoxetinetreated mice compared with the vehicle-treated control animals (Fig. 1a).
Several cargo proteins docked to the AP-2 complex are bound via recognition of single or multiple tyrosine-based motifs by the AP2M subunit. These motifs in the cytosolic tails of transmembrane receptors consist of four amino acid residues (20). The initial tyrosine residue of the motif is followed by two random amino acids, and the signal ends with one of the hydrophobic residues leucine, isoleucine, phenylalanine, valine, or methionine: YXX (27). We searched for such tyrosine-based motifs in mature full-length TRKB. The Eukaryotic Linear Motif library (21) indicated the presence of multiple putative tyrosine-based sorting signals (classified as TRG ENDOCYTIC 2 motifs, TRG2) in the cytosolic domain of the rat and mouse canonical TRKB sequence: Y 515 FGI, Y 705 YRV, Y 726 RKF, Y 782 ELM, Y 816 LDI, as seen in Fig. 1f. Interestingly, one of the presumed AP2M-binding sites, Y 816 LDI, which is located within the C-terminal portion of TRKB, is phosphorylated following antidepressant administration (9). Thus, we focused on the TRKB interaction with the cargo-docking subunit of the AP-2 complex (AP2M). We further confirmed the interaction between TRKB and AP2M in lysates of mouse hippocampus and the TRKB-overexpressing MG87 cell line by co-IP and Western blotting (Fig. 1b).
To assess the interaction mechanism between AP2M and TRKB, we generated a model of full-length AP2M in the Rap-torX server (23,24). For the docking simulations, the AP2M model and the last 26 amino acid (aa) residues from the C-terminal portion of TRKB (Fig. 1g), including variants with either a phospho-dead (Y816A) or a phosphomimetic mutation (Y816E) of the critical tyrosine, were uploaded to the CABSdock server (25). We obtained the Gibbs' free energy for the AP2M:TRKB peptide interaction, and the lowest 10 values from each group were used in statistical analysis. The interaction energy of the AP2M:TRKB.Y816E complex was lower than that for the AP2M:TRKB.ctrl complex, suggesting a more stable interaction, whereas it was significantly higher for the AP2M: TRKB.Y816A complex compared with AP2M:TRKB.ctrl (Kruskal-Wallis ϭ 25.83; p Ͻ 0.0001, n ϭ 10/group) (Fig. 1i). Based on these in silico results, we proceeded to analyze the AP2M:TRKB C terminus interaction in vitro by ELISA. For this purpose, three variants of N-biotinylated synthetic peptides of the last 26 aa of rat TRKB were used (ctrl, pY, or Y816A) (Fig.  1h). The interaction between AP2M and the phosphorylated TRKB peptide (TRKB.pY) was stronger compared with control peptide. In line with our in silico data, the AP2M:TRKB.Y816A interaction was weaker compared with the control peptide (Kruskal-Wallis ϭ 20.48; p Ͻ 0.0001, n ϭ 8/group) (Fig. 1j).

Role of AP2M in TRKB function
Previous studies indicate that AP2M depletion in mammalian cells prevents the assembly of functional AP-2 complexes (26) and results in a loss of AP-2 function (17). Thus, we investigated the impact of AP2M down-regulation on activation of TRKB receptor in MG87.TRKB cells overexpressing TRKB receptor. To induce TRKB signaling, the cells were challenged with BDNF following the silencing of AP2M, as there was no detectable phosphorylation of the receptor without BDNF. The depletion of AP2M (t(14) ϭ 5.521; p Ͻ 0.0001, n ϭ 8/group) respectively. e, repeated administration of fluoxetine (7 days) also disrupts the TRKB:AP2M interaction, as measured by ELISA. f, the C-terminal region of TRKB exhibits AP2M-binding motifs identified by the Eukaryotic Linear Motif library (TRG_ENDOCYTIC_2): Y 515 FGI, Y 705 YRV, Y 726 RKF, Y 782 ELM, Y 816 LDI (blue circle). Yellow, signal peptide (SP); gray, extracellular portion (N-terminal); blue, transmembrane (TM); green, intracellular portion (C-terminal). Red circles, sites of tyrosine phosphorylation in TRKB. g, the docking simulations were performed using sequences of the peptides spanning the last 26 amino acids of TRKB receptor. h, synthetic peptides (0.1 g/ml ϭ 34 M) were used in cell-free ELISA to measure their binding to AP2M protein in MG.87.TRKB cell lysate. Red residues represent the TRG_ENDOCYTIC_2-positive motif and mutated sites. i, the phosphomimetic peptide (TRKB.Y816E, red circles) shows lower free energy for the receptor:peptide docking simulations than WT peptide (TRKB.ctrl, gray circles), whereas TRKB.Y816A mutant (blue circles) displayed the highest free energy values (n ϭ 10/group). j, phosphorylated peptide (TRKB.pY, red circles) showed a higher interaction with AP2M compared with control peptide (TRKB.ctrl, gray circles), whereas TRKB.Y816A mutant (blue circles) displayed the lowest interaction levels, as measured by cell-free ELISA (n ϭ 8/group). *, p Ͻ 0.05 from the WT peptide TRKB.ctrl. Black crosses, mean Ϯ S.E.

Discussion
Sub-chronic (29) and chronic fluoxetine treatment (30,31) can induce changes in the proteomics profile in rodent brains. This compound altered the expression of proteins involved in neuroprotection, serotonin biosynthesis, and axonal transport in hippocampus and frontal cortex (30), as well as in endocytosis and transport in visual cortex (31).
The present study identified the endocytic complex AP-2 as a target whose interaction with TRKB is reduced following acute fluoxetine treatment. In line with a robust interaction between these proteins, we found three of four subunits of the AP-2 complex interacting with TRKB, whereas in the case of related complexes (i.e. AP-3 and AP-1), only one subunit was co-precipitated for each: AP3D1 and AP1B1. Furthermore, we identi-

Antidepressant-induced disruption of the TRKB:AP2M interaction
fied tyrosine-based motifs in the TRKB receptor as putative binding sites for the AP2M subunit of the AP-2 adaptor complex. Especially the region around the Tyr-816 residue of TRKB conformed to the classical consensus for an AP2M-interacting motif. In line with its classical role as an endocytic adaptor complex, AP-2 regulates the surface localization of TRKB. Decreased expression of AP2M resulted in increased surface levels of TRKB, facilitating its activation by its ligand BDNF. We also found increased levels of activated TRKB in the hippocampus of AP2M haploinsufficient mice, as well as an increased exposure of TRKB at the surface of cultured hippocampal neurons from these animals. Finally, we observed that classical AD, such as fluoxetine, imipramine, rolipram, and phenelzine, as well as the fast acting antidepressants ketamine and RR-HNK are able to disrupt the coupling between AP2M and TRKB, resulting in increased exposure of TRKB at the cell surface, at concentrations found in the central nervous system after systemic administration (32)(33)(34).
It has been established that activation, endocytosis, and trafficking of TRKB receptors regulate the propagation of the downstream signaling response (35). There are three main cascades downstream of TRKB receptor activation. MAPK/ERK signaling regulating neuronal differentiation and growth is activated by recruitment of adaptors to the Tyr-515 site of the receptor. Phosphorylation of the Tyr-515 residue also induces a PI3K/Akt cascade promoting cell survival and growth. Finally, phosphorylation of the Tyr-816 residue is the key step for docking PLC-␥1, which regulates protein kinase C isoforms and induces the release of Ca 2ϩ from internal sources to activate Ca 2ϩ /calmodulin (11,36). Eventually, the post-endocytic sorting machinery plays a major role during the fate-determination process of the receptor signaling (37). In this scenario, some of the vesicles carrying the receptors are destined for retrograde transport toward the cell body (38,39), whereas some are recycled back to the cell surface (40). For instance, Pincher, a membrane-trafficking protein, regulates the formation of TRK-carrying endosomes from plasma membrane ruffles in soma, axons, and dendrites (41). Pincher-mediated retrograde transport of signaling TRK receptors is rescued from lysosomal degradation (42). Moreover, BDNF-induced survival signaling in hippocampal neurons has been shown to be regulated by endophilin-A, which is involved in endosomal sorting of active TRKB receptors (27). Evidence suggests that the maintenance of long-term potentiation in hippocampal slices depends on the recycling and re-secretion of BDNF, thus emphasizing the necessity of TRKB receptor recycling back to the cell surface (43). Another study indicates that Slitrk5, by coupling to TRKB receptors, promotes its recycling to the cell surface in striatal neurons (44).
AP-2, as a member of the endocytic machinery (45), contributes to the formation of clathrin-coated vesicles around cargo proteins for endocytosis (46). In line with this idea, our results indicate a BDNF-induced enhancement of the TRKB:AP2M interaction, suggesting that the receptor is endocytosed upon activation. Although acute treatment with AD (e.g. 30 min) can induce a rapid phosphorylation of TRKB receptor in the mouse brain (7), these drugs disrupted the interaction of TRKB:AP2M. Thus, we suggest that the differential regulation of TRKB: AP2M by BDNF and AD may underlie the differential surface expression of the receptor. Zheng et al. (47), as well as Sommerfeld et al. (48), demonstrated that BDNF quickly reduces the TRKB receptor surface level, whereas our data indicate that antidepressants increase such levels. Interestingly, AMPA receptors, also committed to AP-2 complex-dependent endocytosis in the brain (49, 50), are internalized upon activation by glutamate. However, whereas the activation of NMDA receptors results in rapid reinsertion of AMPAR subunits to the cell surface, AMPARs activated in the absence of NMDAR activation are destined for degradation in lysosomes (51). Thus, the source of the activation signal recruits alternative machineries to differentially regulate the fate of the target receptor.
Although the AP-2 complex is best known for its major role in endocytosis (52), it remains doubtful whether the main purpose of the identified AP2M:TRKB interaction is facilitated endocytosis of TRKB. A recent study performing endocytosis assays with WT and AP2M null neurons overexpressing EGFP-TRKB and treated with BDNF could not detect any significant defect in the uptake of TRKB upon loss of AP-2. Instead, the authors showed that the retrograde axonal transport of autophagosomes containing TRKB receptor depends on AP-2 linking the autophagy adaptor LC3 to p150.Glued, an activator of the dynein motor required for autophagosome transport (19). This led us to conclude that the action of the TRKB-AP-2 complex is not necessarily restricted to the neuronal surface, but can take place in different neuronal compartments. Therefore, antidepressants could disrupt the interaction of TRKB receptors with AP2M at the cell surface, as well as in the autophagosomes.
Disruption of the AP2M:TRKB interaction in internal vesicles could prevent the interaction of the complex with the members of the autophagy-lysosome pathway that can lead to degradation of TRKB receptors (19). The delayed delivery of the signaling endosomes to the soma may locally enhance the specific signaling pathways that can recruit more plasticity-related proteins to the synapses. The two above-mentioned scenarios are not exclusive, although our TRKB:RAB colocalization data point toward changes that take place mainly at the cell-surface level. RAB11-positive endosomes have been shown to accumulate in dendrites of hippocampal neurons after BDNF challenge to support dendritic TRKB recycling (28). As we failed to observe an enhancement of TRKB:RAB11 colocalization in spines and neurites under fluoxetine, we suggest that the druginduced regulation of TRKB surface exposure takes place in the cell membrane, at least shortly after the drug administration. On the other hand, the increase of TRKB:RAB7 colocalization only in spines but not in neurites suggests that TRKB receptors directed toward degradation could be stalled, which could contribute to the maintenance of previously activated and internalized receptors.
In the present study, we combined in vivo, in vitro, and in silico methods to gain insight into putative interaction domains between TRKB and AP-2. The in silico techniques allowed us to obtain a priori knowledge about possible interactions, especially involving the Tyr-816 region and the AP2M subunit. In fact, in the present study, the in silico results on the AP2M interaction with the TRKB C terminus matched the in vitro Antidepressant-induced disruption of the TRKB:AP2M interaction observations. As predicted by the CABS-dock server, AP2M displayed an increased interaction with phosphorylated TRKB (simulated using the phosphomimetic mutation Y816E). Interestingly, the Y816A mutation led to a decrease below the levels of control peptide (WT, nonphosphorylated) observed by both in silico (interpreted as a less stable complex) and in vitro assays. This suggests that there is some level of interaction between AP2M and nonactivated TRKB that is further potentiated upon TRKB phosphorylation. However, the role of such interaction, if any, or its occurrence in vivo remains to be investigated.
At this point, we speculated that the interaction of AP2M and TRKB could be a putative target for AD. Supporting this idea, cell-free assays indicated that the disruption of the TRKB: AP2M complex by fluoxetine happens in a dose-dependent manner. However, the same effect was not replicated in the AP2M:TRKB.peptide interaction assay. Therefore, although our data support the idea of a binding site for AD in the TRKB complex, it is unlikely that this binding site is between the receptor C-terminal portion and the AP2M subunit. Moreover, this site displays a low affinity for fluoxetine, given that effective doses are found above 0.3 M. In fact, we have recently identified a low-affinity binding site for antidepressants in the transmembrane domain of TRKB (53). However, it is presently unclear how the conformational changes triggered by antidepressant binding to TRKB could alter the interaction of TRKB with AP-2.
Altogether, we suggest a novel mechanism where the antidepressant-induced disruption of the TRKB:AP2M interaction promotes TRKB cell-surface exposure and consequently BDNF signaling. However, further investigation will be necessary to address the precise mechanism and the long-term consequences of such a disruption.

Animals
Adult (7-week-old) male C57BL/6RccHsd mice from Harlan (Horst aan de Maas, Netherlands), kept with free access to food and water, were used for TRKB immunoprecipitation and MS analysis. Briefly, the animals received a single i.p. injection of fluoxetine (30 mg/kg) and were euthanized by CO 2 . An independent cohort of female and another one of male mice from the same supplier (12-18 weeks old) received fluoxetine for 7 days or single i.p. injection, respectively. The hippocampi were collected and processed for ELISA and co-IP as described below. All procedures were in accordance with international guidelines for animal experimentation and the County Administrative Board of Southern Finland (ESLH-2007-09085/Ym-23, ESAVI/7551/04.10.07/2013).
The generation of floxed AP2M mice carrying loxP sites in front of exon 2 and behind exon 3 of the AP2M encoding gene Ap2m1 has been described (54). These conditional knockout (KO) mice were crossed with CMV-Cre deleter mice to achieve ubiquitous excision of exons 2 and 3. The resulting AP2M.het animals were crossed with AP2M.wt mice, thus excluding the generation of AP2M homozygous KO, which are embryonically lethal (55). All animals were genotyped prior to experiments using PCR on genomic DNA extracted from tissue biopsies and confirmed by Western blot analysis of hippocampus. The KO allele was detected using the forward primer GCTCTAAAGG-TTATGCCTGGTGG and the reverse primer CCAAGGGAC-CTACAGGACTTC, which generate a fragment of 404 bp (PCR at 58°C, 25 cycles). All experiments involving AP2M.het mice were reviewed and approved by the ethics committee of the "Landesamt für Gesundheit und Soziales" (LAGeSo) Berlin and were conducted according to the committee's guidelines. For brain dissection, animals (15 weeks old) were anesthetized with isoflurane prior to cervical dislocation. The hippocampus was dissected on ice and processed as described below.

Immunoprecipitation and mass spectrometry
For TRKB proteome quantitative MS analysis, adult mouse hippocampal samples (n ϭ 3) were mechanically homogenized in NP lysis buffer (20 mM Tris-HCl, 137 mM NaCl, 10% glycerol, 50 mM NaF, 1% Nonidet P-40, 0.05 mM Na 3 VO 4 , containing a mixture of protease and phosphatase inhibitors (#P2714 and #P0044, respectively, Sigma-Aldrich)). The samples were centrifuged (16,000 ϫ g) at 4°C for 15 min, and the supernatant was subjected to preclearing (control) and TRKB immunoprecipitation using Protein G-Sepharose-containing spin columns (#69725, Pierce Spin Columns Snap Cap). Briefly, 30 l of Protein G-Sepharose were added to the columns and washed twice with 200 l of NP buffer. Samples (400 g of total protein) were cleared through the columns without antibody (1 h, under rotation at 4°C). Unbound samples were collected (1000 ϫ g, 1 min at 4°C), preincubated with 5 l of TRKB antibody (goat antimouse TRKB antibody, #AF1494, R&D Systems), and cleared through new Protein G-Sepharose tubes. After 1 h of rotation in the cold, the columns were washed twice with NP buffer and twice with TBS buffer (1000 ϫ g, 1 min at 4°C). The proteins were then eluted from the column with 0.1 M glycine (3 ϫ 1000 ϫ g, 1 min, 4°C) or with Laemmli buffer for samples submitted to Western blotting and stored at Ϫ80°C until further analysis. Cysteine bonds of the proteins were reduced with tris(2-carboxyethyl)-phosphine hydrochloride salt (#C4706, Sigma-Aldrich) and alkylated with iodoacetamide (#57670, Fluka, Sigma-Aldrich). Samples were digested overnight with trypsin (sequencing grade modified trypsin, #V5111, Promega) that specifically cleaves the carboxylic side of lysine and arginine amino acids. The digested peptides were then purified with C18 microspin columns (Harvard Apparatus) (58).

Antidepressant-induced disruption of the TRKB:AP2M interaction
Mass spectrometry analysis was performed on an Orbitrap Elite hybrid mass spectrometer (Thermo Fisher Scientific) coupled to EASY-nLC II system (Xcalibur version 2.7.0 SP1, Thermo Fisher Scientific). The peptides were separated with a 60-min linear gradient from 5 to 35% of buffer B (98% acetonitrile and 0.1% formic acid in MS grade water). The analysis was performed in data-dependent acquisition: a high-resolution (60,000) FTMS full scan (m/z 300 -1700) was followed by top20 CID-MS2 scans in an ion trap (Energy 35). The maximum fill time was 200 ms for both Fourier-transform ion cyclotron resonance MS (full AGC target 1,000,000) and the ion trap (MSn AGC target of 50,000). The MS/MS spectra were searched against the mouse component of the UniProtKB database (release 2012_08, 16,943 entries) using the SEQUEST search engine in Proteome Discoverer TM software (version 1.4, Thermo Fisher Scientific). Carbamidomethylation (ϩ57.021464 Da) of cysteine residues was used as static modification, and oxidation (ϩ15.994491 Da) of methionine was used as dynamic modification. Precursor mass tolerance and fragment mass tolerance were set to Ͻ15 ppm and Ͻ0.8 Da, respectively. A maximum of two missed cleavages was allowed, and the results were filtered to a maximum false discovery rate of 0.05 that was calculated by Target Decoy PSM validator, a tool of the Proteome Discoverer software. Raw MS data files can be found in FigShare.
For each protein-protein interaction, besides spectral counting values, a D R -score was calculated, and the specific interactions were filtered as in PMID (59). The full list of identified interactions is presented in Table S1. Proteins differing between saline-and fluoxetine-treated samples are highlighted in Table S2.
For silencing AP2M expression, MG87.TRKB cells were transfected with a mix of four AP2M-specific shRNA sequences (#TG712191 (Origene) or scrambled sequence as control) using Lipofectamine 2000 (#11668019, Thermo Fisher Scientific) as transfecting agent according to the manufacturer's instructions. The expression of GFP reporter was confirmed by microscopy 24 h after transfection, and AP2M expression levels were determined by Western blotting. The cells were challenged with BDNF (0 or 25 ng/ml/15 min) and lysed for ELISA or Western blotting as described below.
Primary cultures of cortical and hippocampal cells from rat embryos (embryonic day 18) or postnatal day 0 mice (for AP2M.het and wt) were prepared as described in the literature (60) and maintained in NeuroBasal medium (Thermo Fisher Scientific) supplemented with 1% B-27 supplement (Thermo Fisher Scientific), 1% penicillin/streptomycin, and 1% L-glutamine at 37°C (60).

Sample processing and Western blotting
Samples from hippocampi of male and female mice were sonicated in NP lysis buffer and centrifuged (16,000 ϫ g at 4°C for 15 min). The supernatants were either submitted to TRKB: AP2M ELISA, or processed for immunoprecipitation of TRKB or AP2M, and submitted to SDS-PAGE. Supernatant of samples from AP2M.wt and AP2M.het mice, as well as samples from MG87.TRKB cells, were also submitted to SDS-PAGE. For cell-free ELISAs described below, the samples from MG87.TRKB cells were acidified with 1 M HCl (pH ϳ2) at room temperature (RT) for 15 min, and the pH was restored to 7.4 with 1 M NaOH. This procedure aims to disrupt noncovalent interactions between proteins in a given complex.
The surface levels of TRKB were also determined by ELISA (47) in MG87.TRKB cells treated with fluoxetine or imipramine (10 M; 15 min or 2 h) and also after AP2M down-regulation. Hippocampal neurons (10 DIV) of haploinsufficient Ap2m mice were also tested for their TRKB surface expression. Briefly, cells cultivated in clear-bottom 96-well plates (View-Plate 96, PerkinElmer Life Sciences) were washed with ice-cold PBS and fixed with 100 l of 4% paraformaldehyde per well. 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 (1:1000 in blocking buffer, #AF1494, R&D Systems) ON at 4°C. Following a brief washing with PBS, the samples were incubated with HRP-conjugated anti-Gt antibody (1:5000 in blocking buffer, #61-1620, Invitrogen) for 1 h at RT. The cells were washed four times with 200 l of PBS for 10 min each. Finally, the chemiluminescent signal generated by reaction with ECL was acquired in a plate reader.
The cell-free assays were performed in white 96-well plates. The plates were precoated with anti TRKB antibody (1:1000, #AF1494, R&D Systems) in carbonate buffer (pH 9.8) ON at 4°C. Following blocking with 3% BSA in PBST buffer (2 h at RT), 120 g of total protein from each sample (treated as for cell-free ELISA) were added and incubated ON at 4°C under agitation. The plates were then washed three times with PBST buffer, and fluoxetine (0.01-100 M) was added for 15 min. Following the washing protocol, the secondary antibody against AP2M was added (1:2000, #sc-515920, Santa Cruz Biotechnology) ON at 4°C under agitation. The plates were then incubated with HRP-conjugated mouse IgGbinding protein (1:5000, #516102, Santa Cruz Biotechnology) for 2 h at RT. The HRP activity was detected by luminescence following incubation with ECL on a plate reader.
For the cell-free AP2M:TRKB C-terminal interaction, the plates were precoated with antibody against AP2M (1:500, #516102, Santa Cruz Biotechnology) in carbonate buffer, and then samples (treated as for cell-free ELISA) from MG87.TRKB cells were added ON at 4°C. N-Biotinylated synthetic peptides (ctrl, pY, or Y816A; Genscript; sequences described in Fig. 1h) of the last 26 aa of rat TRKB (0.1 g/ml ϭ 34 nM, in PBST) were added for 1 h at RT, followed by HRP-conjugated streptavidin (1:10,000, #21126, Thermo Fisher Scientific) for 1 h at RT. The luminescence was determined upon incubation with ECL as described. In an independent experiment, immobilized AP2M from MG87.TRKB lysates was incubated with pY TRKB C-terminal peptide and fluoxetine (0 or 10 M) for 1 h at RT, and the luminescence was determined via HRP-conjugated streptavidin activity reaction with ECL by a plate reader. The background-subtracted signal from each sample in each of the described assays was expressed as a percentage of the control group (vehicle-treated or WT peptide).
Imaging was performed with a Leica TCS SP8 X inverted confocal microscope, ϫ63 oil objective (1.40 numeric aperture) at 1024 ϫ 1024-pixel resolution. At least 6 -10 Z-stack steps were acquired with 0.30-m intervals from each image. The ImageJ FIJI Coloc2 plug-in was used for the colocalization analysis. The experimenters selected and analyzed the regions of interest (spines or neurites) blindly to the treatments. The regions of interest that passed the requirement of Costes' p value Ͼ 0.95 were selected for further statistical analysis of Pearson's correlation coefficient values from ctrl and fluoxetine groups.
The antibodies used in Western blotting, ELISA, or immunostaining were validated in previous studies (63-65) (e.g. against pY residues of TRKB) (7,9,57). The specificity of AP2M antibody was assessed in this study in transgenic mice and AP2M-silenced cells. For the antibody used to detect total TRKB, we performed Western blotting with samples from MG87 cells expressing TRKB or TRKA, and the results can be found on the FigShare repository.

Structural bioinformatics data mining and docking assay
To address a putative interaction between AP2M and TRKB, we submitted the full-length, canonical sequence of rat and mouse TRKB (Uniprot: Q63604 or P15209, respectively) to the Eukaryotic Linear Motif library (21). Several TRG_ENDO-CYTIC_2 motifs (TRG2; a tyrosine-based sorting signal responsible for the interaction with AP2M subunit) were found for both mouse and rat TRKB. Of particular interest, the Tyr-816 residue is also activated by BDNF and antidepressants. Thus, the sequence of the last 26 aa residues of rat TRKB (WT, Y/E, or Y/A) and the full-length AP2M model generated through the RaptorX server (23) were submitted to docking simulations using the CABS-dock server (25).

Statistical analysis
Data were analyzed by Student's t test and one-or two-way analysis of variance (or their nonparametric equivalents) followed by Fisher's LSD post hoc test, when appropriate. Values