key: cord-0021042-3ko2095k
authors: Plenge, Per; Yang, Dongxue; Salomon, Kristine; Laursen, Louise; Kalenderoglou, Iris E.; Newman, Amy H.; Gouaux, Eric; Coleman, Jonathan A.; Loland, Claus J.
title: The antidepressant drug vilazodone is an allosteric inhibitor of the serotonin transporter
date: 2021-08-20
journal: Nat Commun
DOI: 10.1038/s41467-021-25363-3
sha: baf0d4b1d06f1606c148433f5fb40145eff68d36
doc_id: 21042
cord_uid: 3ko2095k

Depression is a common mental disorder. The standard medical treatment is the selective serotonin reuptake inhibitors (SSRIs). All characterized SSRIs are competitive inhibitors of the serotonin transporter (SERT). A non-competitive inhibitor may produce a more favorable therapeutic profile. Vilazodone is an antidepressant with limited information on its molecular interactions with SERT. Here we use molecular pharmacology and cryo-EM structural elucidation to characterize vilazodone binding to SERT. We find that it exhibits non-competitive inhibition of serotonin uptake and impedes dissociation of [(3)H]imipramine at low nanomolar concentrations. Our SERT structure with bound imipramine and vilazodone reveals a unique binding pocket for vilazodone, expanding the boundaries of the extracellular vestibule. Characterization of the binding site is substantiated with molecular dynamics simulations and systematic mutagenesis of interacting residues resulting in decreased vilazodone binding to the allosteric site. Our findings underline the versatility of SERT allosteric ligands and describe the unique binding characteristics of vilazodone.

S ynaptic transmission is a fundamental process which underlies neuronal communication 1 . Neurotransmitters within synaptic vesicles are released from the pre-synaptic neuron and diffuse across the synapse, binding to post-synaptic receptors and thereby mediate downstream signaling 2, 3 . Serotonin is a neurotransmitter which is involved in modulating many key brain functions including sleep, cognition, mood, appetite, and sex drive as well as other important physiological processes such as gastrointestinal mobility and blood clotting 4 . The serotonin transporter (SERT) removes serotonin from synapses, recycling it into the pre-synaptic neuron and terminating signaling [5] [6] [7] . SERT belongs to the family of neurotransmitter transporters known as neurotransmitter sodium symporters (NSSs) which also includes the dopamine, norepinephrine, GABA, and glycine transporters as well as numerous bacterial homologs 8 . NSSs use the energy stored in the sodium gradient to drive the thermodynamically "uphill" transport of the substrate. NSSs harbor a conserved 3D fold which consists of a 10 transmembrane (TM) core region arranged into two invertedtopological repeats, known as the LeuT fold 7, [9] [10] [11] [12] [13] [14] [15] [16] , as well as one or two additional C-terminal TMs. NSSs are proposed to utilize an alternating access mechanism to transport substrate 17 where a rocking bundle of TM1, 2, 6, and 7 has been suggested to control inward versus outward accessibility to the central substratebinding site relative to a stable scaffold domain of TM3, 4, 8, and 9 14, 18, 19 . It is hypothesized that in the transporter's apo form, the conformational equilibrium will be biased towards an outwardopen conformation, primed to bind ligands, such as neurotransmitter and ions (Na + , Cl -). Upon binding, the protein will rearrange to an occluded conformation, followed by opening to the cytosol where the substrate and ions will be released. The transporter then transitions back to the outward-open conformation 6, 20, 21 . For SERT and LeuT, the return step can involve a counter-transport of K + 22,23 .

SERT is the target of therapeutics used to treat major depressive, anxiety, obsessive-compulsive, eating, and post-traumatic stress disorders 6, 24, 25 . Imipramine (IMI, Supplementary Fig. 1 ) was the first tricyclic antidepressant (TCA) to be medically used and was found to block both SERT and the norepinephrine transporter (NET) 26 . However, IMI was also found to exhibit a large number of side effects including neurological, pulmonary, and gastrointestinal complications, as well as toxicity 27 . Selective serotonin reuptake inhibitors (SSRIs) were developed to bind SERT with high affinity and specificity, resulting in better tolerance and fewer side effects 28, 29 . However, many individuals who are prescribed SSRIs also still report a myriad of side effects including sexual dysfunction, weight loss or gain, anxiety, nausea, headaches, dizziness, insomnia, and dry mouth, many of which pose barriers to adherence 30 . All investigated drugs have been shown to exert their inhibitory effect as competitive inhibitors by binding to the central binding site for serotonin, located about halfway across the membrane, also known as the S1 site [31] [32] [33] .

NSSs possess an extracellular vestibule, serving as an entryway to the S1 site, which also harbors a potential allosteric binding site for ligands, deemed the S2 site 31, 32, [34] [35] [36] . Drugs that target allosteric sites can be superior to competitive drugs that target orthosteric sites because the allosteric sites are typically less conserved and can therefore demonstrate higher drug target selectivity 37, 38 , resulting in fewer side effects. Drugs that act at allosteric sites can also provide a different therapeutic indication because they can act as more than an on/off switch and can function as activity-modulators (e.g., benzodiazepines) or possess activity only in specific body regions due to environmental selectivity (e.g., acetaminophen/paracetamol). Accordingly, the allosteric site in SERT, as well as in other NSS proteins, is an attractive target that has recently gained attention [39] [40] [41] [42] [43] .

Several drugs bind to the S2 site in SERT, including S-citalopram (S-CIT, Supplementary Fig. 1) 34, 44 and clomipramine 34 . However, so far all therapeutic drugs that possess an allosteric component bind to the S2 site with low affinity, thus rendering their allosteric binding therapeutically irrelevant. For example, S-CIT binds to the S1 site of SERT with an affinity of 5 nM and to the S2 site with 5 µM 34, 45 . We have reported on the first small-molecule inhibitor possessing high affinity to the S2 site, Lu AF60097 ( Supplementary Fig. 1) , that inhibits 5-HT transport by SERT via a mixed competitive and non-competitive mechanism, suggesting it binds both the allosteric and central sites 40 . Lu AF60097 has an allosteric potency of~30 nM when [ 3 H]IMI was bound to S1. The allosteric interaction has a synergistic effect, which could alleviate side effects from IMI treatment while preserving its therapeutic effects.

Vilazodone (viibryd ® ; VLZ, Supplementary Fig. 1 ) is a clinically approved SSRI which also has activity as a 5-HT 1A receptor partial agonist 46 . Its action on the 5-HT 1A receptor is thought to provide for its faster onset of action relative to the classical SSRIs 47 , as has been observed for vortioxetine 48 . The common side effects of VLZ include nausea, diarrhea, and headaches, which are also common for other SSRIs. In contrast, VLZ is not as frequently associated with a reduction in sexual dysfunction [49] [50] [51] , the most common cause of discontinuation of SSRI treatment 52 . Also, weight gain is reduced with VLZ treatment relative to other SSRIs [49] [50] [51] . However, despite these intriguing properties, the molecular basis for VLZ binding to SERT remains an open question. The only report addressing the VLZ binding site is a computational approach using ligand docking and molecular dynamics (MD) simulations 53 . In this study, VLZ binds to the S1 site with the main interacting residues overlapping with those that bind other TCAs and SSRIs. However, due to the long and potentially bitopic nature of the molecule, its 5-cyanoindole group protrudes into the S2 site and disrupts an extracellular salt bridge between Arg104 in TM1 and Glu493 in TM10.

Here we use molecular pharmacology and cryo-EM structural analysis to elucidate the SERT:VLZ interaction. We find that VLZ exhibits non-competitive inhibition of serotonin transport. We show that VLZ impedes the dissociation of S1-bound [ 3 H]IMI with low nanomolar affinity suggesting a strong association to the S2 site without involving residues in the S1 site. This is supported by a cryo-EM structure of the SERT:IMI:VLZ complex, which reveals that VLZ binds to the allosteric (S2) site that includes an aromatic pocket formed by TM10, 11, and 12. Accordingly, our studies expand the three-dimension volume of the allosteric site of SERT and demonstrate a mode of action of the clinically approved antidepressant drug VLZ, which could account for its distinct therapeutic profile relative to other SSRIs.

The VLZ binding site is distinct from other antidepressants. VLZ is one of the newer SSRIs on the market. It was approved for the treatment of major depressive disorders in the US in 2011 and reached more than two million prescriptions per year by 2015 54 . Despite the therapeutic success, understanding its mechanistic underpinnings at SERT is limited. To examine the SERT:VLZ interaction in detail, we first measured its ability to inhibit [ 3 H]5-HT transport. COS-7 cells were transiently transfected with SERT WT and preincubated with increasing concentrations of VLZ to obtain binding equilibrium before adding [ 3 H]5-HT (Fig. 1a) . We found that VLZ inhibits [ 3 H]5-HT transport with an apparent affinity (K i ) of 1.1 nM (Supplementary Table 1 ). This is in accordance with previously reported data 55 and approximately 5 times as potent as S-CIT, which has an apparent affinity (K i ) of Table 1 ). In equilibrium binding experiments using [ 3 H]S-CIT (Fig. 1b) , VLZ showed a similar affinity with a K i of 0.7 nM, which is also about 5-times the affinity of S-CIT. All SSRIs and TCAs inhibiting SERT bind with high affinity to the S1 site 56-62 and a recent MD simulation also suggested that VLZ binds at this site 53 . S1-site drug binding is dependent on the side-chain interactions with at least two out of three key residues: Tyr95, Ile172, and Ser438 and the MD simulation of VLZ binding also suggested that all three of these residues are involved in its binding pose. We therefore wished to test this experimentally by measuring the impact of mutagenesis of these residues on VLZ affinity. Using the same experimental setup as in Fig. 1a , we found that all three mutants are capable of transporting 5-HT with a K M similar or slightly higher than observed for SERT WT ( Supplementary Fig. 2 59 . Surprisingly, we did not find any effect of the mutations on the apparent affinity for VLZ (Fig. 1d, Supplementary Table 1 ), suggesting that the drug binds in a different site than the other tested antidepressants.

SERT possesses a high-affinity allosteric binding site for VLZ. Our results from the S1 site mutants raised the possibility that VLZ may bind to a different site in SERT. A classical approach to determine whether VLZ binds competitively or noncompetitively to 5-HT is to perform a [ 3 H]5-HT uptake experiment in the presence of increasing VLZ concentrations (Fig. 2a) . The experiment showed that VLZ causes a decrease in the V MAX of [ 3 H]5-HT without having any significant effect on K M (Supplementary Table 2 ). This is in accordance with non-competitive binding between 5-HT and VLZ suggesting that VLZ may not bind to the central S1 site. We therefore investigated VLZ binding to the S2 site. To investigate S2 binding, we and others have previously shown that binding of SERT inhibitors to the S2 site can be characterized by their ability to impede the dissociation of an S1-bound radioligand 34, 40, 63, 64 . Thus, we prepared membranes of COS-7 cells transiently expressing SERT WT and preincubated them with [ 3 H]IMI or [ 3 H]S-CIT, followed by the addition of VLZ. We then monitored the effect of VLZ on the dissociation rate of the radiolabeled drug. Indeed, over a range of concentrations, VLZ dose-dependently slowed the rate of [ 3 H]IMI unbinding (Fig. 2b) . The results show that VLZ inhibits [ 3 H]IMI dissociation with an allosteric potency of 14 nM (Fig. 2c , Table 1 ). This is several orders of magnitude more potent than previously reported for any other drug 34, 63 . VLZ also inhibits [ 3 H]S-CIT dissociation, albeit with a lower potency of 250 nM (Fig. 2c) . The high allosteric potency of VLZ could suggest that it binds in the extracellular vestibule and thereby slows dissociation by occlusion of the exit pathway from the S1 site.

Cryo-EM structures of SERT-15B8-Fab:IMI:VLZ complex. To investigate the VLZ binding location and conformation, we determined the structure of the SERT:IMI:VLZ complex (PDB: 7LWD). We purified human SERT using an N-and C-terminally truncated construct (ΔN72/C13) in the presence of VLZ and IMI. The 15B8-Fab was used to guide particle alignment for singleparticle cryo-EM analysis. We have previously shown that the binding of this Fab to SERT does not perturb SERT transport activity 20 . Here we further investigated whether Fab binding to SERT affects the binding of VLZ. The results show that the VLZ dissociation rate is unaffected by the presence of Fab, irrespective of the S1-bound radioligand, suggesting that the Fab fragment does not influence VLZ binding to SERT ( Supplementary Fig. 3) . A density map of the SERT:15B8-Fab:IMI:VLZ complex was obtained at overall resolution of 3.65 Å with the highest local resolution extending to 2.9 Å in the core TMs near the drug binding sites (Fig. 3a, Supplementary Fig. 4 ). The SERT and drug densities were of good quality for most key functional regions with the side chains of most large residues well-fitted into their corresponding density features. The overall structure is similar to the X-ray structure of SERT complexed with S-CIT (PDB: 5I73) and paroxetine (PDB:5I6X) 44 and the transporter has been stabilized in an outward-open conformation with an intact Arg104-Glu493 salt bridge (Fig. 3b ). IMI exhibits a binding pose similar to the one found previously by MD simualtions 40 and also aligns with S-CIT and paroxetine binding in the S1 site 44 (Fig. 3e) . The chemically equivalent amine group of IMI occupies subsite A as it does for S-CIT, while the dibenzazepine group localizes in subsites B and C (Fig. 3e) .

In the allosteric site, about 12 Å above the S1-bound IMI, the bulky backbone of VLZ adopts a nearly linear binding pose across the extracellular vestibule. The heteroaryl piperazine moiety of VLZ overlaps with S2-bound S-CIT and Lu AF60097 (Supplementary Fig. 1 ), hindering the entry pathway to the S1 site and ensuring the blockade of IMI dissociation from SERT. VLZ binding is unlike the other allosteric site binding inhibitors, S-CIT and Lu AF60097, as these adopt a triangular pose with the cyanogroup extending toward the non-helical region of TM6. The quinolinone moiety of Lu AF60097 protrudes into a subsite near the tip of the extracellular loop (EL) 4 ( Fig. 3f ). In contrast, the indole ring of VLZ is nestled in a hydrophobic cavity constituted by residues from TM10, TM11, TM12, and EL6 (Fig. 3f) , allowing for an almost linear pose of the drug. The extension of the indole ring of VLZ, close to TM12, capped by EL6 and flanked by Tyr579, indicates the promiscuity of the allosteric site in NSSs. The diversity of chemical structures of allosteric inhibitors results in different binding modes in the S2 site, which may result in distinct pharmacological profiles 65 .

We examined detailed views of the SERT residues forming the ligand-binding sites (Fig. 4) , in the S1 site. The tertiary amine group of IMI interacts with the carboxylate of Asp98 and forms cation-π interaction with Tyr95 from TM1 (Fig. 4b) . The dibenzazepine group is encompassed by Tyr176, Phe341, and Thr497 and interacting with Ile172 and Tyr175 through hydrophobic and aromatic interactions (Fig. 4b ). In the allosteric site, Tyr495, Pro499, Ser559, Pro560, Pro561, and Tyr579 form extensive hydrophobic interactions with the indole on VLZ (Fig. 4c) . Glu494 likely forms a salt bridge with the protonatable piperazine nitrogen, whereas the piperazine ring system forms a hydrophobic interaction with Phe556 (Fig. 4c) . The benzofuran ring extends close to the central site allowing the formamide group to interact with IMI (3.8 Å to the tertiary amine group) and with Gln332 in TM6a. Glu493 and Arg104 form a salt bridge "above" the benzofuran moiety while Phe335, from "underneath", participates in aromatic interactions (Fig. 4c) .

MD simulation of VLZ binding reveals two possible poses. We calculated the local resolution of IMI and VLZ in our reconstructions to 2.9 Å and 3.4 Å, respectively (FSC, 0.143). Their modeling into their respective sites of the cryo-EM structure resulted in correlation coefficients (CC) of 0.81 and 0.79. However, both molecules are pseudosymmetric. Thus, the possibility of the ligands binding in a pose rotated by 180°rather than the hypothesized one, was not ruled out. When the ligands-in this "flipped" orientation-were fit into the density features, both had a CC of 0.69 after refinement ( Supplementary Fig. 5 (Fig. 5a, b) . Both poses were also fit back into the electron density without major clashes (Fig. 5c-f ). The molecular mechanics Poisson-Boltzmann surface area (MMPBSA) method was used to calculate the protein-ligand binding free energy of each VLZ pose. The free energy of binding (ΔG bind ) was not significantly different between the hypothesized and flipped poses with ΔG bind = −10 ± 3 kcal/ mol and −14 ± 3 kcal/mol, respectively. This opens the possibility that VLZ may bind with equally high affinity in both poses. However, when we viewed both fits of VLZ from the plane of the membrane (Fig. 5e, f) , the indole ring in the hypothesized pose complies better with the electron density. Taken together this suggests that VLZ can bind in both poses, but the hypothesized pose is predominant in our cryo-EM structure.

Mutation of VLZ binding residues decreases allosteric potency. Based on our cryo-EM structure and MD simulations, we define the predominant allosteric binding site for VLZ as the hypothesized one. It is composed of residues, Arg104 (TM1), Gln332 and Phe335 (TM6), Glu493, Glu494, and Tyr495 (TM10), Phe556 (TM11), Ser559 and Pro561 (EL6) and Tyr579 (TM12) (Fig. 6a) . All residues have hydrophobic interactions with VLZ. To further investigate the involvement of each side chain, we mutated each residue, one-by-one, and monitored the impact on the allosteric potency of VLZ (Fig. 6b, Table 1 ). All mutants possessed detectable 5-HT transport activity and most mutants had a K M for 5-HT transport with less than a two-fold deviation from SERT WT. The two exceptions were Q332A and Y495A, which showed a 3.5-and 4.8-fold increase in K M , respectively. The decreased apparent affinity coincided with a marked drop in transport velocity, suggesting that the two mutations may change the conformational equilibrium of SERT. All of the mutants showed a decreased allosteric potency relative to SERT WT, consistent with an impairment of VLZ binding. The decrease in allosteric potency ranged from~4-fold for Y579A to 430-fold for the F335A mutant ( Table 1 ). The mutants fall into three categories with respect to their effect: F579A and P561Q exhibit less than 7-fold decrease in allosteric potency. An intermediary group of R104K, Q332A, E493N, and E494Q show a 13-to 30-fold decrease in potency, whereas F335A, T495A, and F556A have 77-, 240-and 430-fold decrease, respectively. The fold change in allosteric potency is not only dependent on the interaction of the residue and VLZ, but also on the ability of the substituted residue to compensate for the WT environment. The isolated contribution of each residue might deviate somewhat from the observed affinity changes. Taken together, we find that all mutants investigated produced a significant decrease in the allosteric potency for VLZ, supporting the binding location of the cryo-EM structure.

Depression is a major worldwide problem. In the US population, above the age of 18 the prevalence was 7.8% in 2019 66 and with the entry of the COVID-19 pandemic, this is expected to rise further. More than a third of patients suffering from the major depressive disorder are non-responsive to classic antidepressant treatment 67 , supporting the value of a larger selection of drugs with a variety of pharmacological actions in identifying the optimal treatment regimen for each individual. So far, all antidepressant drugs targeting SERT have been reported to bind with high affinity to the orthosteric, central binding site. Here we report that VLZ binds with high affinity to the S2 site located to the SERT extracellular vestibule. The finding is substantiated by pharmacological and structural data showing (i) that VLZ binds non-competitively to 5-HT (Fig. 2a) , (ii) its affinity is unaffected by central S1 mutations (Fig. 1d) , (iii) it has a profound effect on [ 3 H]IMI dissociation (Fig. 2c) , which is decreased by mutation of its binding site (Fig. 6b) , and (iv) high-resolution structural evidence supported by MD simulation for the association of VLZ to the S2 site (Figs. [3] [4] [5] . With an allosteric potency of~14 nM, VLZ binds with higher S2 affinity than any other molecule reported. This is supported by its linear binding pose spanning approximately 19 Å, wedging between TM1 and 6 and capped by Tyr579 in TM12. Moreover, the VLZ binding site is different from the allosteric binding site for S-CIT 44 . Though there are overlapping regions in the binding sites, S-CIT is a smaller molecule and kinks "downwards" with its fluoro-group between TM10 and 11. The binding site is also markedly different from the experimental compound Lu AF60097, which has similarities to the VLZ structure as well as to S-CIT. Lu AF60097 binds similarly to S-CIT but has been suggested to create a~90°kink "upwards" with its N-substituted ring thus occluding the presumed exit pathway by interacting with TM10 40 . In contrast, VLZ twists around TM10 and protrudes all the way to TM12. This novel binding mode expands the size of the S2 site and highlights the versatility of ligands that could bind at this site (Fig. 3f) . IMI was first discovered in the 1950s and marked a new era of molecular pharmacology which revolutionized the treatment of psychiatric disorders. This not only gave clinicians better tools for treating mood disorders but also directly enabled advancements in our understanding of neurotransmission and neurotransmitter reuptake through the seminal work of Arvid Carlsson and Julius Axelrod 68,69 . Our cryo-EM structure provides a high-resolution view of IMI binding to SERT now more than 60 years later. We find that compared to the binding pose of other TCAs 56 and SSRIs 70 , the binding of IMI shows a similar association to the S1 site with an equal distribution of drug moieties in subsites A, B, and C (Fig. 3d) .

A recent MD simulation 53 has suggested a binding pose for VLZ to the S1 site with a protrusion into the S2 site. Based on our cryo-EM structure, we cannot rule out that VLZ would associate with the S1 site in the absence of IMI. However, if the suggested pose was predominant, we would expect our mutagenesis study of the S1 site to show a decrease in VLZ affinity (Fig. 1d) . We would also expect our [ 3 H]5-HT saturation uptake experiments to show, at least in part, a competitive inhibition by VLZ (Fig. 2a) . Another possible explanation is that VLZ has two binding sites of similar affinity. We note that the Hill slope of~2 for VLZ inhibition of [ 3 H]5-HT uptake (Fig. 1a) opens this possibility.

The solved SERT structure is similar to previously solved structures with antidepressants bound in the S1 site 44, 70, 71 though the Arg104-Asp493 salt bridge is intact the S1-bound ligands are solvent accessible to the extracellular environment. Accordingly, the conformation is outward-open. This is in contrast to the solved LeuT structures where an intact salt bridge excludes solvent to the S1 site and is thus an outward-occluded conformation. It is indeed possible that the SERT transport cycle also includes an outward-open conformation with a broken salt bridge. Could there be a connection between the VLZ binding pose and its reduction in adverse effects relative to other SSRIs? A linkage between binding pose and therapeutic effect is a wellknown phenomenon for biased agonism in G-protein coupled receptors, especially for bitopic ligands 72, 73 , similar to VLZ, but correlations have also been reported in transporters. The stimulant effect of cocaine is due to its inhibition of the DAT 74 , but the atypical DAT inhibitors possess a different binding pose, and are not correlated with any stimulant or rewarding effect in rodents [75] [76] [77] and humans 78 . Whether the distinct pharmacological profile for VLZ is due to its allosteric binding in SERT must await further investigation. Conversely, it is plausible that blocking transport alone may be responsible for beneficial as well as adverse outcomes of the SSRIs and that allosteric inhibition may provide no clinical or adherence benefit relative to orthosteric inhibitors.

Taken together, we show that the therapeutic drug VLZ binds with high affinity to the allosteric site of SERT and that it may be able to bind to the site in two distinct poses with equal affinities. We also show that the allosteric site is larger than previously anticipated, by defining a funnel that is wedged between TM10 and 11 and capped by Tyr579 in TM12 which extends the existing site. The allosteric potency for VLZ lies in the low nanomolar range, and hence, within a therapeutic dose regimen. VLZ interacts with polar, non-polar, hydrophobic, acidic, and basic residues at the allosteric site. The elucidated molecular interactions and high-affinity binding opens for the possibility of developing improved drugs with distinct pharmacodynamic profiles, which may translate into more beneficial therapeutic actions. The allosteric site of SERT and other NSSs has likely been underexplored in structure-based drug design and we anticipate 

Site-directed mutagenesis. Human SERT was cloned into the pUbi1z vector using the NotI and XbaI. Mutations herein were generated using the two-step PCR method or, for R104K, F335L, E493N, E494Q, Y495A, F556A, P561G, and Y579A, ordered at GeneArt, Thermo Fisher (Waltham, MA). All mutations were confirmed by DNA sequencing. For mutants generated by PCR, the sense primers were (antisense primers were complementary): Y95F: cagtgattggctttgcagtggacctgggc; I172M: gcatcattgccttttacatggcttcctactacaac; Q332N: gatgcagccgctaacatcttcttctctc; S438T: caagcctgcaaacgttgtgtccaagcc. SERT expression and purification. The human SERT construct used for the cryo-EM studies was the N-and C-terminally truncated WT transporter (ΔN72, ΔC13) 44, 79 . Cells were solubilized in 20 mM Tris-HCl, pH 8, 100 mM NaCl containing 20 mM DDM, 2.5 mM CHS in the presence of 10 μM VLZ and 10 μM IMI and were then purified into buffer A containing 20 mM Tris-HCl, pH 8.0, 100 mM NaCl, 1 mM DDM, 0.2 mM CHS, 10 μM VLZ, and 10 μM IMI by Strep-Tactin affinity chromatography. The N-and C-termini containing GFP and purification tags were removed by thrombin digestion. SERT was mixed with 15B8 Fab at a 1:1.2 molar ratio. The resulting complexes were further purified by size-exclusion chromatography into buffer A. The peak fraction containing the SERT-15B8 Fab was concentrated to 4 mg/ml and then 100 μM VLZ and 100 μM IMI were added before cryo-EM grid preparation. 83 were subjected to reference-free 2D classification followed by heterogenous refinement in cryoSPARC v3.2 84 . Homogeneous refinement local contrast transfer function (CTF) refinement and then non-uniform refinement was performed in cryoSPARC after recentering particles 85 (Supplementary Fig. 4 ). The maximum fit resolution for local CTF refinement was 3.6 Å. The 185,019 selected particles yielded a reconstruction at 3.65 Å. The resolution was estimated with the goldstandard Fourier shell correlation (FSC) 0.143 criterion 85 in cryoSPARC. The local resolution was also calculated in cryoSPARC.

Model building and refinement. A previous cryo-EM structure of the ts2-active SERT in complex with 15B8 Fab and 8B6 ScFv bound to ibogaine (6DZY) 20 was used as initial model; the 8B6 ScFv was removed before docking the PDB into the sharpened map in ChimeraX v0.9 86 . Manual adjustment was then performed in Coot v0.8.9.1 87 and VLZ and IMI were placed into the electron densities in Coot to generate a model. Model refinement of the coordinates was carried out in PHENIX v1.15.2-3472 88 using the real space refinement package. This iterative refinement process was repeated until the model reached optimal stereochemistry and geometric statistics as evaluated by MolProbity 89 . For cross-validation, the FSC curve between the refined model and half maps was calculated and compared to avoid overfitting.

Atomistic MD simulations. The SERT:IMI:VLZ complexes were prepared using the Maestro software tool (Schrödinger Release 2021-2: Schrödinger, LLC, New York, NY, 2021); antibody fragments were removed, and missing hydrogens atoms and side chains were added. Amino acid pK a was calculated using Epik 90 . Two Na + ions and one Cl − ion were modeled on the basis of PDB ID: 5I71 and 5I6X 44 by protein backbone superposition of the cryo-EM structures to the X-ray crystallography protein structure and deleting the latter ones. Final models were processed through the Orientations of Proteins in Membrane (OPM) tool 91 . The force field parameters of protonated VLZ and IMI were parameterized according to the CHARMM General Force Field (CGenFF) 92 . The insertion of the OPM-protein complexes into a bilayer was performed with CHARMM-GUI 93 . Each hSERT:I-MI:VLZ complex was inserted in a lipid bilayer consisting of 249 1-palmitoyl-2oleoyl-sn-glycero-3-phosphocholine (POPC) and 83 cholesterol molecules, followed by hydration and NaCl (150 mM). Approximate box dimensions: 110 × 110 x 115 Å 3 (~128000 atoms). All simulations were performed with GROMACS version 2020.3 using CHARMM36m force fields 94 for SERT, CHARMM36 force fields 95 for lipids, and the TIP3P model 96 for water. To maintain the temperature, a Nosé-Hoover temperature coupling method 97 with a tau-t of 1 ps was used, and for pressure coupling, a semi-isotropic Parrinello − Rahman method 98 with a tau-p of 5 ps and a compressibility of 4.5 × 10 −5 bar −1 was used. Fig. 6 Effect of SERT S2 mutants on VLZ allosteric potency. a Zoomed interactions between VLZ and SERT. Schematics are generated by LIGPLOT + 1.4. Each eyelash motif indicates a hydrophobic contact. b Effect of mutating residues in the S2 site on allosteric potency for VLZ. Point mutation of residues, shown in the cryo-EM structure to interact with VLZ, all significantly (P < 0.001, one-way ANOVA with Tukey's multiple comparisons test) decreases its allosteric potency to various degrees relative to SERT WT (dotted line, data shown in Fig. 2b ). See allosteric potencies for SERT WT and mutants in Table 1 . Experiments are performed on membrane preparations from COS-7 cells transiently expressing the indicated SERT mutants. Data are shown as means ± S.E. (error bars), n = 3-10. Wild-type is shown with a black dotted line. Allosteric site mutants are shown using colored symbols and solid lines defined in the legend. See Table 1 for quantitative data including allosteric potency and n-value for the individual experiments. Source data are provided as a Source Data file. maintained at 310 K and pressure at 1 bar. Non-bonded interactions were calculated in a pairwise manner within the 12 Å cutoff, with a switching function applied between 10-12 Å. Long-range non-bonded interactions were calculated with the particle mesh Ewald (PME) method 99 . The LINCS 100 method was applied to hydrogen bonds. Periodic boundary conditions were used. 5000 steps of steepest descent minimization were performed. Next, systems were equilibrated with two sets of NPT and NVT of simulations to smoothly relax the system with overall duration of 500 ps and 6 ns respectively, during which lipids, Cα atoms, protein side chains, ligand heavy atoms and the bound ions were restrained individually by harmonic potentials with decreasing force constants (from 4 kcal/mol/Å 2 to 0 kcal/ mol/Å 2 ) to allow for relaxation of protein side chains and hydration of the protein.

The equilibrated structures were subjected to 200 ns NPT MD simulations, integrated into 2-fs time steps, and trajectories recorded every 10 ps. VMD 101 was used for visualization. The RMSD was calculated using GROMACS version 2020.3 software tools.

MMPBSA calculation. The MMPBSA (Molecular Mechanics Poisson-Boltzmann Surface Area) method was used to calculate the protein-ligand binding free energy of each VLZ pose. For each system, in total, the final 750 frames were extracted from the trajectories and the free energy of binding was determined using the g_mmpbsa 102 tool developed for GROMACS.

Electrical synapses and their functional interactions with chemical synapses

Vesicular and plasma membrane transporters for neurotransmitters

Neurotransmitter transporters: structure meets function

The expanded biology of serotonin

Neurotransmitter transporters: molecular function of important drug targets

SLC6 neurotransmitter transporters: structure, function, and regulation

Biophysical approaches to the study of LeuT, a prokaryotic homolog of neurotransmitter sodium symporters

TCDB: the Transporter Classification Database for membrane transport protein analyses and information

Crystal structure of a bacterial homologue of Na+/Cl-dependent neurotransmitter transporters

X-ray structures of LeuT in substrate-free outward-open and apo inward-open states

Conformational dynamics of ligand-dependent alternating access in LeuT

A mechanism for intracellular release of Na+ by neurotransmitter/sodium symporters

Properties of an inwardfacing state of LeuT: conformational stability and substrate release

Substrate-modulated unwinding of transmembrane helices in the NSS transporter

A partially-open inward-facing intermediate conformation of LeuT is associated with Na(+) release and substrate transport

The use of LeuT as a model in elucidating binding sites for substrates and inhibitors in neurotransmitter transporters

Simple allosteric model for membrane pumps

Substrate-induced conformational dynamics of the dopamine transporter

Shared molecular mechanisms of membrane transporters

Serotonin transporter-ibogaine complexes illuminate mechanisms of inhibition and transport

Neurotransmitter transporters: recent progress

Unifying concept of serotonin transporter-associated currents

Transition metal ion FRET uncovers K(+) regulation of a neurotransmitter/sodium symporter

Cloning and expression of a functional serotonin transporter from rat brain

Cloning of a serotonin transporter affected by antidepressants

The serotonergic system in the neurobiology of depression: Relevance for novel antidepressants

National Institute of Diabetes and Digestive and Kidney Diseases. LiverTox: Clinical and Research Information on Drug-Induced Liver Injury (National Institute of Diabetes and Digestive and Kidney Diseases

Recent advances in the understanding of the interaction of antidepressant drugs with serotonin and norepinephrine transporters

Proportion of antidepressants prescribed without a psychiatric diagnosis is growing

Addressing the side effects of contemporary antidepressant drugs: a comprehensive review

The mechanism of a neurotransmitter:sodium symporter-inward release of Na+ and substrate is triggered by substrate in a second binding site

Direct assessment of substrate binding to the Neurotransmitter:Sodium Symporter LeuT by solid state NMR

X-ray structure of dopamine transporter elucidates antidepressant mechanism

Steric hindrance mutagenesis in the conserved extracellular vestibule impedes allosteric binding of antidepressants to the serotonin transporter

Antidepressant binding site in a bacterial homologue of neurotransmitter transporters

Nanopharmacological force sensing to reveal allosteric coupling in transporter binding sites

The concept of allosteric modulation: an overview

Novel allosteric modulators of G protein-coupled receptors

Identification of a novel allosteric modulator of the human dopamine transporter

The mechanism of a high-affinity allosteric inhibitor of the serotonin transporter

Thermostabilization and purification of the human dopamine transporter (hDAT) in an inhibitor and allosteric ligand bound conformation

Identification of an allosteric binding site on the human glycine transporter, GlyT2, for bioactive lipid analgesics

Allosteric modulation of neurotransmitter transporters as a therapeutic strategy

X-ray structures and mechanism of the human serotonin transporter

increases inhibitor binding to the human serotonin transporter by an allosteric mechanism. Comparison with other serotonin transporter inhibitors

Neurochemical evaluation of the novel 5-HT1A receptor partial agonist/serotonin reuptake inhibitor, vilazodone

Vilazodone for the treatment of depression: an update

Modes and nodes explain the mechanism of action of vortioxetine, a multimodal agent (MMA): enhancing serotonin release by combining serotonin (5HT) transporter inhibition with actions at 5HT receptors (5HT1A, 5HT1B, 5HT1D, 5HT7 receptors)

A 1-year, open-label study assessing the safety and tolerability of vilazodone in patients with major depressive disorder

A 12-week comparative prospective open-label randomized controlled study in depression patients treated with vilazodone and escitalopram in a Tertiary Care Hospital in North India

Efficacy and safety of vilazodone 20 and 40 mg in major depressive disorder: a randomized, double-blind, placebo-controlled trial

Antidepressant-induced sexual dysfunction during treatment with moclobemide, paroxetine, sertraline, and venlafaxine

The binding mode of vilazodone in the human serotonin transporter elucidated by ligand docking and molecular dynamics simulations

ClinCalc DrugStats Database, Version 21.2

Synthesis and structure-activity relationship in a class of indolebutylpiperazines as dual 5-HT(1A) receptor agonists and serotonin reuptake inhibitors

Binding and orientation of tricyclic antidepressants within the central substrate site of the human serotonin transporter

Genetically encoded photocrosslinkers locate the highaffinity binding site of antidepressant drugs in the human serotonin transporter

Binding of the multimodal antidepressant drug vortioxetine to the human serotonin transporter

Interaction of antidepressants with the serotonin and norepinephrine transporters: mutational studies of the S1 substrate binding pocket

High affinity recognition of serotonin transporter antagonists defined by species-scanning mutagenesis. An aromatic residue in transmembrane domain I dictates species-selective recognition of citalopram and mazindol

Tyr-95 and Ile-172 in transmembrane segments 1 and 3 of human serotonin transporters interact to establish high affinity recognition of antidepressants

The chicken serotonin transporter discriminates between serotonin-selective reuptake inhibitors. A speciesscanning mutagenesis study

Characterization of an allosteric citalopram-binding site at the serotonin transporter

Structure-activity relationship studies of citalopram derivatives: examining substituents conferring selectivity for the allosteric site in the 5-HT transporter

Antidepressants: past, present and future

Pharmacological treatments for patients with treatment-resistant depression

Recent studies on the mode of action of antidepressive drugs

Inhibition of uptake of tritiated-noradrenaline in the intact rat brain by imipramine and structurally related compounds

Structural basis for recognition of diverse antidepressants by the human serotonin transporter

Chemical and structural investigation of the paroxetinehuman serotonin transporter complex

Philip S. Portoghese Medicinal Chemistry Lectureship: designing bivalent or bitopic molecules for G-protein coupled receptors. the whole is greater than the sum of its parts

Novel and potent dopamine D2 receptor Go-protein biased agonists

Abolished cocaine reward in mice with a cocaine-insensitive dopamine transporter

Relationship between conformational changes in the dopamine transporter and cocaine-like subjective effects of uptake inhibitors

R-modafinil (armodafinil): a unique dopamine uptake inhibitor and potential medication for psychostimulant abuse

Translating the atypical dopamine uptake inhibitor hypothesis toward therapeutics for treatment of psychostimulant use disorders

Modafinil: preclinical, clinical, and post-marketing surveillance-a review of abuse liability issues

Screening and large-scale expression of membrane proteins in mammalian cells for structural studies

Automated electron microscope tomography using robust prediction of specimen movements

Refined Cryo-EM structure of the T4 tail tube: exploring the lowest dose limit

Real-time CTF determination and correction

DoG Picker and TiltPicker: software tools to facilitate particle selection in single particle electron microscopy

cryoSPARC: algorithms for rapid unsupervised cryo-EM structure determination

Optimal determination of particle orientation, absolute hand, and contrast loss in single-particle electron cryomicroscopy

UCSF Chimera-a visualization system for exploratory research and analysis

Coot: model-building tools for molecular graphics

PHENIX: a comprehensive Python-based system for macromolecular structure solution

MolProbity: all-atom structure validation for macromolecular crystallography

Epik: a software program for pKa prediction and protonation state generation for drug-like molecules

OPM database and PPM web server: resources for positioning of proteins in membranes

CHARMM general force field: a force field for drug-like molecules compatible with the CHARMM all-atom additive biological force fields

Simulations Using the CHARMM36 Additive Force Field

CHARMM36m: an improved force field for folded and intrinsically disordered proteins

Update of the CHARMM all-atom additive force field for lipids: validation on six lipid types

Comparison of simple potential functions for simulating liquid water

Canonical dynamics: equilibrium phase-space distributions

Constant pressure molecular dynamics for molecular systems

Particle mesh Ewald: An N⋅log(N) method for Ewald sums in large systems

LINCS: a linear constraint solver for molecular simulations

VMD: visual molecular dynamics

G-mmpbsa -A GROMACS tool for highthroughput MM-PBSA calculations

The authors thank Lone Rosenquist for excellent technical assistance, Vivek Kumar for discussion on vilazodone and to Prof. Antonios Kolocouris (U. Athens) for providing computational time. A portion of this research was supported by NIH grant U24GM129547 and performed at the Pacific Northwest Cryo-EM Center at OHSU and accessed through EMSL (grid. 

The authors declare no competing interests.

Supplementary information The online version contains supplementary material available at https://doi.org/10.1038/s41467-021-25363-3.Correspondence and requests for materials should be addressed to J.A.C. or C.J.L.Peer review information Nature Communications thanks Keith Henry, Baruch Kanner and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.Reprints and permission information is available at http://www.nature.com/reprintsPublisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/.